4 ** The author disclaims copyright to this source code. In place of
5 ** a legal notice, here is a blessing:
7 ** May you do good and not evil.
8 ** May you find forgiveness for yourself and forgive others.
9 ** May you share freely, never taking more than you give.
11 *************************************************************************
12 ** This is the implementation of the page cache subsystem or "pager".
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"
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:
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
44 ** (b) The page was a freelist leaf page at the start of the transaction.
46 ** (c) The page number is greater than the largest page that existed in
47 ** the database file at the start of the transaction.
49 ** (1) A page of the database file is never overwritten unless one of the
50 ** following are true:
52 ** (a) The page and all other pages on the same sector are overwriteable.
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.
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
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.
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.
70 ** (5) All writes to the database file are synced prior to the rollback journal
71 ** being deleted, truncated, or zeroed.
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.
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
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.
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
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
116 int sqlite3PagerTrace
=1; /* True to enable tracing */
117 #define sqlite3DebugPrintf printf
118 #define PAGERTRACE(X) if( sqlite3PagerTrace ){ sqlite3DebugPrintf X; }
120 #define PAGERTRACE(X)
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
139 ** OPEN <------+------+
142 ** +---------> READER-------+ |
145 ** |<-------WRITER_LOCKED------> ERROR
148 ** |<------WRITER_CACHEMOD-------->|
151 ** |<-------WRITER_DBMOD---------->|
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]
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.
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.
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
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
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.
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.
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
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.
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
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 performed)
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
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
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
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.
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().
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
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; }
420 # define CODEC1(P,D,N,X,E) /* NO-OP */
421 # define CODEC2(P,D,N,X,E,O) O=(char*)D
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 */
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:
473 ** The current 'state' of the pager object. See the comment and state
474 ** diagram above for a description of the pager state.
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
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
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).
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.
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
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
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
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.
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
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
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.
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 */
709 int nRead
; /* Database pages read */
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 */
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 */
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.
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++
750 # define PAGER_INCR(v)
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
803 # define MEMDB pPager->memDb
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)
813 # define USEFETCH(x) 0
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) ){ ...
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;
847 #ifndef SQLITE_OMIT_WAL
851 rc
= sqlite3WalFindFrame(pPager
->pWal
, pgno
, &iRead
);
852 return (rc
==SQLITE_OK
&& iRead
==0);
859 #ifndef SQLITE_OMIT_WAL
860 # define pagerUseWal(x) ((x)->pWal!=0)
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
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
){
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
913 assert( !isOpen(p
->fd
) );
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
925 assert( pPager
->changeCountDone
==0 || pPager
->eLock
>=RESERVED_LOCK
);
926 assert( p
->eLock
!=PENDING_LOCK
);
931 assert( pPager
->errCode
==SQLITE_OK
);
932 assert( sqlite3PcacheRefCount(pPager
->pPCache
)==0 || pPager
->tempFile
);
936 assert( pPager
->errCode
==SQLITE_OK
);
937 assert( p
->eLock
!=UNKNOWN_LOCK
);
938 assert( p
->eLock
>=SHARED_LOCK
);
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 );
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
);
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
);
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
)
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
);
1008 #endif /* ifndef NDEBUG */
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
,
1028 "State: %s errCode=%d\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"
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?"
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
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);
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
1087 pPager
->xGet
= getPageMMap
;
1088 #endif /* SQLITE_MAX_MMAP_SIZE>0 */
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
;
1106 Pgno pgno
= pPg
->pgno
;
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;
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
);
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
);
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
){
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
){
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 */
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
1198 static int pagerLockDb(Pager
*pPager
, int eLock
){
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
))
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
){
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
);
1241 UNUSED_PARAMETER(pPager
);
1244 #ifdef SQLITE_ENABLE_BATCH_ATOMIC_WRITE
1245 if( pPager
->dbSize
>0 && (dc
&SQLITE_IOCAP_BATCH_ATOMIC
) ){
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
) ){
1262 return JOURNAL_HDR_SZ(pPager
) + JOURNAL_PG_SZ(pPager
);
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
){
1280 for(i
=0; i
<nByte
; i
++){
1281 hash
= (hash
*1039) + pData
[i
];
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
) );
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 */
1345 if( SQLITE_OK
!=(rc
= sqlite3OsFileSize(pJrnl
, &szJ
))
1347 || SQLITE_OK
!=(rc
= read32bits(pJrnl
, szJ
-16, &len
))
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
))
1359 /* See if the checksum matches the super-journal name */
1360 for(u
=0; u
<len
; u
++){
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.
1372 zSuper
[len
+1] = '\0';
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 ** ---------------------------------------
1392 static i64
journalHdrOffset(Pager
*pPager
){
1394 i64 c
= pPager
->journalOff
;
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
) );
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);
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 ){
1451 rc
= sqlite3OsFileSize(pPager
->jfd
, &sz
);
1452 if( rc
==SQLITE_OK
&& sz
>iLimit
){
1453 rc
= sqlite3OsTruncate(pPager
->jfd
, iLimit
);
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);
1527 memset(zHeader
, 0, sizeof(aJournalMagic
)+4);
1532 /* The random check-hash initializer */
1533 if( pPager
->journalMode
!=PAGER_JOURNALMODE_MEMORY
){
1534 sqlite3_randomness(sizeof(pPager
->cksumInit
), &pPager
->cksumInit
);
1538 /* The Pager.cksumInit variable is usually randomized above to protect
1539 ** against there being existing records in the journal file. This is
1540 ** dangerous, as following a crash they may be mistaken for records
1541 ** written by the current transaction and rolled back into the database
1542 ** file, causing corruption. The following assert statements verify
1543 ** that this is not required in "journal_mode=memory" mode, as in that
1544 ** case the journal file is always 0 bytes in size at this point.
1545 ** It is advantageous to avoid the sqlite3_randomness() call if possible
1546 ** as it takes the global PRNG mutex. */
1548 sqlite3OsFileSize(pPager
->jfd
, &sz
);
1550 assert( pPager
->journalOff
==journalHdrOffset(pPager
) );
1551 assert( sqlite3JournalIsInMemory(pPager
->jfd
) );
1554 put32bits(&zHeader
[sizeof(aJournalMagic
)+4], pPager
->cksumInit
);
1556 /* The initial database size */
1557 put32bits(&zHeader
[sizeof(aJournalMagic
)+8], pPager
->dbOrigSize
);
1558 /* The assumed sector size for this process */
1559 put32bits(&zHeader
[sizeof(aJournalMagic
)+12], pPager
->sectorSize
);
1562 put32bits(&zHeader
[sizeof(aJournalMagic
)+16], pPager
->pageSize
);
1564 /* Initializing the tail of the buffer is not necessary. Everything
1565 ** works find if the following memset() is omitted. But initializing
1566 ** the memory prevents valgrind from complaining, so we are willing to
1567 ** take the performance hit.
1569 memset(&zHeader
[sizeof(aJournalMagic
)+20], 0,
1570 nHeader
-(sizeof(aJournalMagic
)+20));
1572 /* In theory, it is only necessary to write the 28 bytes that the
1573 ** journal header consumes to the journal file here. Then increment the
1574 ** Pager.journalOff variable by JOURNAL_HDR_SZ so that the next
1575 ** record is written to the following sector (leaving a gap in the file
1576 ** that will be implicitly filled in by the OS).
1578 ** However it has been discovered that on some systems this pattern can
1579 ** be significantly slower than contiguously writing data to the file,
1580 ** even if that means explicitly writing data to the block of
1581 ** (JOURNAL_HDR_SZ - 28) bytes that will not be used. So that is what
1584 ** The loop is required here in case the sector-size is larger than the
1585 ** database page size. Since the zHeader buffer is only Pager.pageSize
1586 ** bytes in size, more than one call to sqlite3OsWrite() may be required
1587 ** to populate the entire journal header sector.
1589 for(nWrite
=0; rc
==SQLITE_OK
&&nWrite
<JOURNAL_HDR_SZ(pPager
); nWrite
+=nHeader
){
1590 IOTRACE(("JHDR %p %lld %d\n", pPager
, pPager
->journalHdr
, nHeader
))
1591 rc
= sqlite3OsWrite(pPager
->jfd
, zHeader
, nHeader
, pPager
->journalOff
);
1592 assert( pPager
->journalHdr
<= pPager
->journalOff
);
1593 pPager
->journalOff
+= nHeader
;
1600 ** The journal file must be open when this is called. A journal header file
1601 ** (JOURNAL_HDR_SZ bytes) is read from the current location in the journal
1602 ** file. The current location in the journal file is given by
1603 ** pPager->journalOff. See comments above function writeJournalHdr() for
1604 ** a description of the journal header format.
1606 ** If the header is read successfully, *pNRec is set to the number of
1607 ** page records following this header and *pDbSize is set to the size of the
1608 ** database before the transaction began, in pages. Also, pPager->cksumInit
1609 ** is set to the value read from the journal header. SQLITE_OK is returned
1612 ** If the journal header file appears to be corrupted, SQLITE_DONE is
1613 ** returned and *pNRec and *PDbSize are undefined. If JOURNAL_HDR_SZ bytes
1614 ** cannot be read from the journal file an error code is returned.
1616 static int readJournalHdr(
1617 Pager
*pPager
, /* Pager object */
1619 i64 journalSize
, /* Size of the open journal file in bytes */
1620 u32
*pNRec
, /* OUT: Value read from the nRec field */
1621 u32
*pDbSize
/* OUT: Value of original database size field */
1623 int rc
; /* Return code */
1624 unsigned char aMagic
[8]; /* A buffer to hold the magic header */
1625 i64 iHdrOff
; /* Offset of journal header being read */
1627 assert( isOpen(pPager
->jfd
) ); /* Journal file must be open. */
1629 /* Advance Pager.journalOff to the start of the next sector. If the
1630 ** journal file is too small for there to be a header stored at this
1631 ** point, return SQLITE_DONE.
1633 pPager
->journalOff
= journalHdrOffset(pPager
);
1634 if( pPager
->journalOff
+JOURNAL_HDR_SZ(pPager
) > journalSize
){
1637 iHdrOff
= pPager
->journalOff
;
1639 /* Read in the first 8 bytes of the journal header. If they do not match
1640 ** the magic string found at the start of each journal header, return
1641 ** SQLITE_DONE. If an IO error occurs, return an error code. Otherwise,
1644 if( isHot
|| iHdrOff
!=pPager
->journalHdr
){
1645 rc
= sqlite3OsRead(pPager
->jfd
, aMagic
, sizeof(aMagic
), iHdrOff
);
1649 if( memcmp(aMagic
, aJournalMagic
, sizeof(aMagic
))!=0 ){
1654 /* Read the first three 32-bit fields of the journal header: The nRec
1655 ** field, the checksum-initializer and the database size at the start
1656 ** of the transaction. Return an error code if anything goes wrong.
1658 if( SQLITE_OK
!=(rc
= read32bits(pPager
->jfd
, iHdrOff
+8, pNRec
))
1659 || SQLITE_OK
!=(rc
= read32bits(pPager
->jfd
, iHdrOff
+12, &pPager
->cksumInit
))
1660 || SQLITE_OK
!=(rc
= read32bits(pPager
->jfd
, iHdrOff
+16, pDbSize
))
1665 if( pPager
->journalOff
==0 ){
1666 u32 iPageSize
; /* Page-size field of journal header */
1667 u32 iSectorSize
; /* Sector-size field of journal header */
1669 /* Read the page-size and sector-size journal header fields. */
1670 if( SQLITE_OK
!=(rc
= read32bits(pPager
->jfd
, iHdrOff
+20, &iSectorSize
))
1671 || SQLITE_OK
!=(rc
= read32bits(pPager
->jfd
, iHdrOff
+24, &iPageSize
))
1676 /* Versions of SQLite prior to 3.5.8 set the page-size field of the
1677 ** journal header to zero. In this case, assume that the Pager.pageSize
1678 ** variable is already set to the correct page size.
1681 iPageSize
= pPager
->pageSize
;
1684 /* Check that the values read from the page-size and sector-size fields
1685 ** are within range. To be 'in range', both values need to be a power
1686 ** of two greater than or equal to 512 or 32, and not greater than their
1687 ** respective compile time maximum limits.
1689 if( iPageSize
<512 || iSectorSize
<32
1690 || iPageSize
>SQLITE_MAX_PAGE_SIZE
|| iSectorSize
>MAX_SECTOR_SIZE
1691 || ((iPageSize
-1)&iPageSize
)!=0 || ((iSectorSize
-1)&iSectorSize
)!=0
1693 /* If the either the page-size or sector-size in the journal-header is
1694 ** invalid, then the process that wrote the journal-header must have
1695 ** crashed before the header was synced. In this case stop reading
1696 ** the journal file here.
1701 /* Update the page-size to match the value read from the journal.
1702 ** Use a testcase() macro to make sure that malloc failure within
1703 ** PagerSetPagesize() is tested.
1705 rc
= sqlite3PagerSetPagesize(pPager
, &iPageSize
, -1);
1706 testcase( rc
!=SQLITE_OK
);
1708 /* Update the assumed sector-size to match the value used by
1709 ** the process that created this journal. If this journal was
1710 ** created by a process other than this one, then this routine
1711 ** is being called from within pager_playback(). The local value
1712 ** of Pager.sectorSize is restored at the end of that routine.
1714 pPager
->sectorSize
= iSectorSize
;
1717 pPager
->journalOff
+= JOURNAL_HDR_SZ(pPager
);
1723 ** Write the supplied super-journal name into the journal file for pager
1724 ** pPager at the current location. The super-journal name must be the last
1725 ** thing written to a journal file. If the pager is in full-sync mode, the
1726 ** journal file descriptor is advanced to the next sector boundary before
1727 ** anything is written. The format is:
1729 ** + 4 bytes: PAGER_SJ_PGNO.
1730 ** + N bytes: super-journal filename in utf-8.
1731 ** + 4 bytes: N (length of super-journal name in bytes, no nul-terminator).
1732 ** + 4 bytes: super-journal name checksum.
1733 ** + 8 bytes: aJournalMagic[].
1735 ** The super-journal page checksum is the sum of the bytes in the super-journal
1736 ** name, where each byte is interpreted as a signed 8-bit integer.
1738 ** If zSuper is a NULL pointer (occurs for a single database transaction),
1739 ** this call is a no-op.
1741 static int writeSuperJournal(Pager
*pPager
, const char *zSuper
){
1742 int rc
; /* Return code */
1743 int nSuper
; /* Length of string zSuper */
1744 i64 iHdrOff
; /* Offset of header in journal file */
1745 i64 jrnlSize
; /* Size of journal file on disk */
1746 u32 cksum
= 0; /* Checksum of string zSuper */
1748 assert( pPager
->setSuper
==0 );
1749 assert( !pagerUseWal(pPager
) );
1752 || pPager
->journalMode
==PAGER_JOURNALMODE_MEMORY
1753 || !isOpen(pPager
->jfd
)
1757 pPager
->setSuper
= 1;
1758 assert( pPager
->journalHdr
<= pPager
->journalOff
);
1760 /* Calculate the length in bytes and the checksum of zSuper */
1761 for(nSuper
=0; zSuper
[nSuper
]; nSuper
++){
1762 cksum
+= zSuper
[nSuper
];
1765 /* If in full-sync mode, advance to the next disk sector before writing
1766 ** the super-journal name. This is in case the previous page written to
1767 ** the journal has already been synced.
1769 if( pPager
->fullSync
){
1770 pPager
->journalOff
= journalHdrOffset(pPager
);
1772 iHdrOff
= pPager
->journalOff
;
1774 /* Write the super-journal data to the end of the journal file. If
1775 ** an error occurs, return the error code to the caller.
1777 if( (0 != (rc
= write32bits(pPager
->jfd
, iHdrOff
, PAGER_SJ_PGNO(pPager
))))
1778 || (0 != (rc
= sqlite3OsWrite(pPager
->jfd
, zSuper
, nSuper
, iHdrOff
+4)))
1779 || (0 != (rc
= write32bits(pPager
->jfd
, iHdrOff
+4+nSuper
, nSuper
)))
1780 || (0 != (rc
= write32bits(pPager
->jfd
, iHdrOff
+4+nSuper
+4, cksum
)))
1781 || (0 != (rc
= sqlite3OsWrite(pPager
->jfd
, aJournalMagic
, 8,
1782 iHdrOff
+4+nSuper
+8)))
1786 pPager
->journalOff
+= (nSuper
+20);
1788 /* If the pager is in persistent-journal mode, then the physical
1789 ** journal-file may extend past the end of the super-journal name
1790 ** and 8 bytes of magic data just written to the file. This is
1791 ** dangerous because the code to rollback a hot-journal file
1792 ** will not be able to find the super-journal name to determine
1793 ** whether or not the journal is hot.
1795 ** Easiest thing to do in this scenario is to truncate the journal
1796 ** file to the required size.
1798 if( SQLITE_OK
==(rc
= sqlite3OsFileSize(pPager
->jfd
, &jrnlSize
))
1799 && jrnlSize
>pPager
->journalOff
1801 rc
= sqlite3OsTruncate(pPager
->jfd
, pPager
->journalOff
);
1807 ** Discard the entire contents of the in-memory page-cache.
1809 static void pager_reset(Pager
*pPager
){
1810 pPager
->iDataVersion
++;
1811 sqlite3BackupRestart(pPager
->pBackup
);
1812 sqlite3PcacheClear(pPager
->pPCache
);
1816 ** Return the pPager->iDataVersion value
1818 u32
sqlite3PagerDataVersion(Pager
*pPager
){
1819 return pPager
->iDataVersion
;
1823 ** Free all structures in the Pager.aSavepoint[] array and set both
1824 ** Pager.aSavepoint and Pager.nSavepoint to zero. Close the sub-journal
1825 ** if it is open and the pager is not in exclusive mode.
1827 static void releaseAllSavepoints(Pager
*pPager
){
1828 int ii
; /* Iterator for looping through Pager.aSavepoint */
1829 for(ii
=0; ii
<pPager
->nSavepoint
; ii
++){
1830 sqlite3BitvecDestroy(pPager
->aSavepoint
[ii
].pInSavepoint
);
1832 if( !pPager
->exclusiveMode
|| sqlite3JournalIsInMemory(pPager
->sjfd
) ){
1833 sqlite3OsClose(pPager
->sjfd
);
1835 sqlite3_free(pPager
->aSavepoint
);
1836 pPager
->aSavepoint
= 0;
1837 pPager
->nSavepoint
= 0;
1838 pPager
->nSubRec
= 0;
1842 ** Set the bit number pgno in the PagerSavepoint.pInSavepoint
1843 ** bitvecs of all open savepoints. Return SQLITE_OK if successful
1844 ** or SQLITE_NOMEM if a malloc failure occurs.
1846 static int addToSavepointBitvecs(Pager
*pPager
, Pgno pgno
){
1847 int ii
; /* Loop counter */
1848 int rc
= SQLITE_OK
; /* Result code */
1850 for(ii
=0; ii
<pPager
->nSavepoint
; ii
++){
1851 PagerSavepoint
*p
= &pPager
->aSavepoint
[ii
];
1852 if( pgno
<=p
->nOrig
){
1853 rc
|= sqlite3BitvecSet(p
->pInSavepoint
, pgno
);
1854 testcase( rc
==SQLITE_NOMEM
);
1855 assert( rc
==SQLITE_OK
|| rc
==SQLITE_NOMEM
);
1862 ** This function is a no-op if the pager is in exclusive mode and not
1863 ** in the ERROR state. Otherwise, it switches the pager to PAGER_OPEN
1866 ** If the pager is not in exclusive-access mode, the database file is
1867 ** completely unlocked. If the file is unlocked and the file-system does
1868 ** not exhibit the UNDELETABLE_WHEN_OPEN property, the journal file is
1869 ** closed (if it is open).
1871 ** If the pager is in ERROR state when this function is called, the
1872 ** contents of the pager cache are discarded before switching back to
1873 ** the OPEN state. Regardless of whether the pager is in exclusive-mode
1874 ** or not, any journal file left in the file-system will be treated
1875 ** as a hot-journal and rolled back the next time a read-transaction
1876 ** is opened (by this or by any other connection).
1878 static void pager_unlock(Pager
*pPager
){
1880 assert( pPager
->eState
==PAGER_READER
1881 || pPager
->eState
==PAGER_OPEN
1882 || pPager
->eState
==PAGER_ERROR
1885 sqlite3BitvecDestroy(pPager
->pInJournal
);
1886 pPager
->pInJournal
= 0;
1887 releaseAllSavepoints(pPager
);
1889 if( pagerUseWal(pPager
) ){
1890 assert( !isOpen(pPager
->jfd
) );
1891 sqlite3WalEndReadTransaction(pPager
->pWal
);
1892 pPager
->eState
= PAGER_OPEN
;
1893 }else if( !pPager
->exclusiveMode
){
1894 int rc
; /* Error code returned by pagerUnlockDb() */
1895 int iDc
= isOpen(pPager
->fd
)?sqlite3OsDeviceCharacteristics(pPager
->fd
):0;
1897 /* If the operating system support deletion of open files, then
1898 ** close the journal file when dropping the database lock. Otherwise
1899 ** another connection with journal_mode=delete might delete the file
1900 ** out from under us.
1902 assert( (PAGER_JOURNALMODE_MEMORY
& 5)!=1 );
1903 assert( (PAGER_JOURNALMODE_OFF
& 5)!=1 );
1904 assert( (PAGER_JOURNALMODE_WAL
& 5)!=1 );
1905 assert( (PAGER_JOURNALMODE_DELETE
& 5)!=1 );
1906 assert( (PAGER_JOURNALMODE_TRUNCATE
& 5)==1 );
1907 assert( (PAGER_JOURNALMODE_PERSIST
& 5)==1 );
1908 if( 0==(iDc
& SQLITE_IOCAP_UNDELETABLE_WHEN_OPEN
)
1909 || 1!=(pPager
->journalMode
& 5)
1911 sqlite3OsClose(pPager
->jfd
);
1914 /* If the pager is in the ERROR state and the call to unlock the database
1915 ** file fails, set the current lock to UNKNOWN_LOCK. See the comment
1916 ** above the #define for UNKNOWN_LOCK for an explanation of why this
1919 rc
= pagerUnlockDb(pPager
, NO_LOCK
);
1920 if( rc
!=SQLITE_OK
&& pPager
->eState
==PAGER_ERROR
){
1921 pPager
->eLock
= UNKNOWN_LOCK
;
1924 /* The pager state may be changed from PAGER_ERROR to PAGER_OPEN here
1925 ** without clearing the error code. This is intentional - the error
1926 ** code is cleared and the cache reset in the block below.
1928 assert( pPager
->errCode
|| pPager
->eState
!=PAGER_ERROR
);
1929 pPager
->eState
= PAGER_OPEN
;
1932 /* If Pager.errCode is set, the contents of the pager cache cannot be
1933 ** trusted. Now that there are no outstanding references to the pager,
1934 ** it can safely move back to PAGER_OPEN state. This happens in both
1935 ** normal and exclusive-locking mode.
1937 assert( pPager
->errCode
==SQLITE_OK
|| !MEMDB
);
1938 if( pPager
->errCode
){
1939 if( pPager
->tempFile
==0 ){
1940 pager_reset(pPager
);
1941 pPager
->changeCountDone
= 0;
1942 pPager
->eState
= PAGER_OPEN
;
1944 pPager
->eState
= (isOpen(pPager
->jfd
) ? PAGER_OPEN
: PAGER_READER
);
1946 if( USEFETCH(pPager
) ) sqlite3OsUnfetch(pPager
->fd
, 0, 0);
1947 pPager
->errCode
= SQLITE_OK
;
1948 setGetterMethod(pPager
);
1951 pPager
->journalOff
= 0;
1952 pPager
->journalHdr
= 0;
1953 pPager
->setSuper
= 0;
1957 ** This function is called whenever an IOERR or FULL error that requires
1958 ** the pager to transition into the ERROR state may have occurred.
1959 ** The first argument is a pointer to the pager structure, the second
1960 ** the error-code about to be returned by a pager API function. The
1961 ** value returned is a copy of the second argument to this function.
1963 ** If the second argument is SQLITE_FULL, SQLITE_IOERR or one of the
1964 ** IOERR sub-codes, the pager enters the ERROR state and the error code
1965 ** is stored in Pager.errCode. While the pager remains in the ERROR state,
1966 ** all major API calls on the Pager will immediately return Pager.errCode.
1968 ** The ERROR state indicates that the contents of the pager-cache
1969 ** cannot be trusted. This state can be cleared by completely discarding
1970 ** the contents of the pager-cache. If a transaction was active when
1971 ** the persistent error occurred, then the rollback journal may need
1972 ** to be replayed to restore the contents of the database file (as if
1973 ** it were a hot-journal).
1975 static int pager_error(Pager
*pPager
, int rc
){
1976 int rc2
= rc
& 0xff;
1977 assert( rc
==SQLITE_OK
|| !MEMDB
);
1979 pPager
->errCode
==SQLITE_FULL
||
1980 pPager
->errCode
==SQLITE_OK
||
1981 (pPager
->errCode
& 0xff)==SQLITE_IOERR
1983 if( rc2
==SQLITE_FULL
|| rc2
==SQLITE_IOERR
){
1984 pPager
->errCode
= rc
;
1985 pPager
->eState
= PAGER_ERROR
;
1986 setGetterMethod(pPager
);
1991 static int pager_truncate(Pager
*pPager
, Pgno nPage
);
1994 ** The write transaction open on pPager is being committed (bCommit==1)
1995 ** or rolled back (bCommit==0).
1997 ** Return TRUE if and only if all dirty pages should be flushed to disk.
2001 ** * For non-TEMP databases, always sync to disk. This is necessary
2002 ** for transactions to be durable.
2004 ** * Sync TEMP database only on a COMMIT (not a ROLLBACK) when the backing
2005 ** file has been created already (via a spill on pagerStress()) and
2006 ** when the number of dirty pages in memory exceeds 25% of the total
2009 static int pagerFlushOnCommit(Pager
*pPager
, int bCommit
){
2010 if( pPager
->tempFile
==0 ) return 1;
2011 if( !bCommit
) return 0;
2012 if( !isOpen(pPager
->fd
) ) return 0;
2013 return (sqlite3PCachePercentDirty(pPager
->pPCache
)>=25);
2017 ** This routine ends a transaction. A transaction is usually ended by
2018 ** either a COMMIT or a ROLLBACK operation. This routine may be called
2019 ** after rollback of a hot-journal, or if an error occurs while opening
2020 ** the journal file or writing the very first journal-header of a
2021 ** database transaction.
2023 ** This routine is never called in PAGER_ERROR state. If it is called
2024 ** in PAGER_NONE or PAGER_SHARED state and the lock held is less
2025 ** exclusive than a RESERVED lock, it is a no-op.
2027 ** Otherwise, any active savepoints are released.
2029 ** If the journal file is open, then it is "finalized". Once a journal
2030 ** file has been finalized it is not possible to use it to roll back a
2031 ** transaction. Nor will it be considered to be a hot-journal by this
2032 ** or any other database connection. Exactly how a journal is finalized
2033 ** depends on whether or not the pager is running in exclusive mode and
2034 ** the current journal-mode (Pager.journalMode value), as follows:
2036 ** journalMode==MEMORY
2037 ** Journal file descriptor is simply closed. This destroys an
2038 ** in-memory journal.
2040 ** journalMode==TRUNCATE
2041 ** Journal file is truncated to zero bytes in size.
2043 ** journalMode==PERSIST
2044 ** The first 28 bytes of the journal file are zeroed. This invalidates
2045 ** the first journal header in the file, and hence the entire journal
2046 ** file. An invalid journal file cannot be rolled back.
2048 ** journalMode==DELETE
2049 ** The journal file is closed and deleted using sqlite3OsDelete().
2051 ** If the pager is running in exclusive mode, this method of finalizing
2052 ** the journal file is never used. Instead, if the journalMode is
2053 ** DELETE and the pager is in exclusive mode, the method described under
2054 ** journalMode==PERSIST is used instead.
2056 ** After the journal is finalized, the pager moves to PAGER_READER state.
2057 ** If running in non-exclusive rollback mode, the lock on the file is
2058 ** downgraded to a SHARED_LOCK.
2060 ** SQLITE_OK is returned if no error occurs. If an error occurs during
2061 ** any of the IO operations to finalize the journal file or unlock the
2062 ** database then the IO error code is returned to the user. If the
2063 ** operation to finalize the journal file fails, then the code still
2064 ** tries to unlock the database file if not in exclusive mode. If the
2065 ** unlock operation fails as well, then the first error code related
2066 ** to the first error encountered (the journal finalization one) is
2069 static int pager_end_transaction(Pager
*pPager
, int hasSuper
, int bCommit
){
2070 int rc
= SQLITE_OK
; /* Error code from journal finalization operation */
2071 int rc2
= SQLITE_OK
; /* Error code from db file unlock operation */
2073 /* Do nothing if the pager does not have an open write transaction
2074 ** or at least a RESERVED lock. This function may be called when there
2075 ** is no write-transaction active but a RESERVED or greater lock is
2076 ** held under two circumstances:
2078 ** 1. After a successful hot-journal rollback, it is called with
2079 ** eState==PAGER_NONE and eLock==EXCLUSIVE_LOCK.
2081 ** 2. If a connection with locking_mode=exclusive holding an EXCLUSIVE
2082 ** lock switches back to locking_mode=normal and then executes a
2083 ** read-transaction, this function is called with eState==PAGER_READER
2084 ** and eLock==EXCLUSIVE_LOCK when the read-transaction is closed.
2086 assert( assert_pager_state(pPager
) );
2087 assert( pPager
->eState
!=PAGER_ERROR
);
2088 if( pPager
->eState
<PAGER_WRITER_LOCKED
&& pPager
->eLock
<RESERVED_LOCK
){
2092 releaseAllSavepoints(pPager
);
2093 assert( isOpen(pPager
->jfd
) || pPager
->pInJournal
==0
2094 || (sqlite3OsDeviceCharacteristics(pPager
->fd
)&SQLITE_IOCAP_BATCH_ATOMIC
)
2096 if( isOpen(pPager
->jfd
) ){
2097 assert( !pagerUseWal(pPager
) );
2099 /* Finalize the journal file. */
2100 if( sqlite3JournalIsInMemory(pPager
->jfd
) ){
2101 /* assert( pPager->journalMode==PAGER_JOURNALMODE_MEMORY ); */
2102 sqlite3OsClose(pPager
->jfd
);
2103 }else if( pPager
->journalMode
==PAGER_JOURNALMODE_TRUNCATE
){
2104 if( pPager
->journalOff
==0 ){
2107 rc
= sqlite3OsTruncate(pPager
->jfd
, 0);
2108 if( rc
==SQLITE_OK
&& pPager
->fullSync
){
2109 /* Make sure the new file size is written into the inode right away.
2110 ** Otherwise the journal might resurrect following a power loss and
2111 ** cause the last transaction to roll back. See
2112 ** https://bugzilla.mozilla.org/show_bug.cgi?id=1072773
2114 rc
= sqlite3OsSync(pPager
->jfd
, pPager
->syncFlags
);
2117 pPager
->journalOff
= 0;
2118 }else if( pPager
->journalMode
==PAGER_JOURNALMODE_PERSIST
2119 || (pPager
->exclusiveMode
&& pPager
->journalMode
!=PAGER_JOURNALMODE_WAL
)
2121 rc
= zeroJournalHdr(pPager
, hasSuper
||pPager
->tempFile
);
2122 pPager
->journalOff
= 0;
2124 /* This branch may be executed with Pager.journalMode==MEMORY if
2125 ** a hot-journal was just rolled back. In this case the journal
2126 ** file should be closed and deleted. If this connection writes to
2127 ** the database file, it will do so using an in-memory journal.
2129 int bDelete
= !pPager
->tempFile
;
2130 assert( sqlite3JournalIsInMemory(pPager
->jfd
)==0 );
2131 assert( pPager
->journalMode
==PAGER_JOURNALMODE_DELETE
2132 || pPager
->journalMode
==PAGER_JOURNALMODE_MEMORY
2133 || pPager
->journalMode
==PAGER_JOURNALMODE_WAL
2135 sqlite3OsClose(pPager
->jfd
);
2137 rc
= sqlite3OsDelete(pPager
->pVfs
, pPager
->zJournal
, pPager
->extraSync
);
2142 #ifdef SQLITE_CHECK_PAGES
2143 sqlite3PcacheIterateDirty(pPager
->pPCache
, pager_set_pagehash
);
2144 if( pPager
->dbSize
==0 && sqlite3PcacheRefCount(pPager
->pPCache
)>0 ){
2145 PgHdr
*p
= sqlite3PagerLookup(pPager
, 1);
2148 sqlite3PagerUnrefNotNull(p
);
2153 sqlite3BitvecDestroy(pPager
->pInJournal
);
2154 pPager
->pInJournal
= 0;
2156 if( rc
==SQLITE_OK
){
2157 if( MEMDB
|| pagerFlushOnCommit(pPager
, bCommit
) ){
2158 sqlite3PcacheCleanAll(pPager
->pPCache
);
2160 sqlite3PcacheClearWritable(pPager
->pPCache
);
2162 sqlite3PcacheTruncate(pPager
->pPCache
, pPager
->dbSize
);
2165 if( pagerUseWal(pPager
) ){
2166 /* Drop the WAL write-lock, if any. Also, if the connection was in
2167 ** locking_mode=exclusive mode but is no longer, drop the EXCLUSIVE
2168 ** lock held on the database file.
2170 rc2
= sqlite3WalEndWriteTransaction(pPager
->pWal
);
2171 assert( rc2
==SQLITE_OK
);
2172 }else if( rc
==SQLITE_OK
&& bCommit
&& pPager
->dbFileSize
>pPager
->dbSize
){
2173 /* This branch is taken when committing a transaction in rollback-journal
2174 ** mode if the database file on disk is larger than the database image.
2175 ** At this point the journal has been finalized and the transaction
2176 ** successfully committed, but the EXCLUSIVE lock is still held on the
2177 ** file. So it is safe to truncate the database file to its minimum
2178 ** required size. */
2179 assert( pPager
->eLock
==EXCLUSIVE_LOCK
);
2180 rc
= pager_truncate(pPager
, pPager
->dbSize
);
2183 if( rc
==SQLITE_OK
&& bCommit
){
2184 rc
= sqlite3OsFileControl(pPager
->fd
, SQLITE_FCNTL_COMMIT_PHASETWO
, 0);
2185 if( rc
==SQLITE_NOTFOUND
) rc
= SQLITE_OK
;
2188 if( !pPager
->exclusiveMode
2189 && (!pagerUseWal(pPager
) || sqlite3WalExclusiveMode(pPager
->pWal
, 0))
2191 rc2
= pagerUnlockDb(pPager
, SHARED_LOCK
);
2193 pPager
->eState
= PAGER_READER
;
2194 pPager
->setSuper
= 0;
2196 return (rc
==SQLITE_OK
?rc2
:rc
);
2199 /* Forward reference */
2200 static int pager_playback(Pager
*pPager
, int isHot
);
2203 ** Execute a rollback if a transaction is active and unlock the
2206 ** If the pager has already entered the ERROR state, do not attempt
2207 ** the rollback at this time. Instead, pager_unlock() is called. The
2208 ** call to pager_unlock() will discard all in-memory pages, unlock
2209 ** the database file and move the pager back to OPEN state. If this
2210 ** means that there is a hot-journal left in the file-system, the next
2211 ** connection to obtain a shared lock on the pager (which may be this one)
2212 ** will roll it back.
2214 ** If the pager has not already entered the ERROR state, but an IO or
2215 ** malloc error occurs during a rollback, then this will itself cause
2216 ** the pager to enter the ERROR state. Which will be cleared by the
2217 ** call to pager_unlock(), as described above.
2219 static void pagerUnlockAndRollback(Pager
*pPager
){
2220 if( pPager
->eState
!=PAGER_ERROR
&& pPager
->eState
!=PAGER_OPEN
){
2221 assert( assert_pager_state(pPager
) );
2222 if( pPager
->eState
>=PAGER_WRITER_LOCKED
){
2223 sqlite3BeginBenignMalloc();
2224 sqlite3PagerRollback(pPager
);
2225 sqlite3EndBenignMalloc();
2226 }else if( !pPager
->exclusiveMode
){
2227 assert( pPager
->eState
==PAGER_READER
);
2228 pager_end_transaction(pPager
, 0, 0);
2230 }else if( pPager
->eState
==PAGER_ERROR
2231 && pPager
->journalMode
==PAGER_JOURNALMODE_MEMORY
2232 && isOpen(pPager
->jfd
)
2234 /* Special case for a ROLLBACK due to I/O error with an in-memory
2235 ** journal: We have to rollback immediately, before the journal is
2236 ** closed, because once it is closed, all content is forgotten. */
2237 int errCode
= pPager
->errCode
;
2238 u8 eLock
= pPager
->eLock
;
2239 pPager
->eState
= PAGER_OPEN
;
2240 pPager
->errCode
= SQLITE_OK
;
2241 pPager
->eLock
= EXCLUSIVE_LOCK
;
2242 pager_playback(pPager
, 1);
2243 pPager
->errCode
= errCode
;
2244 pPager
->eLock
= eLock
;
2246 pager_unlock(pPager
);
2250 ** Parameter aData must point to a buffer of pPager->pageSize bytes
2251 ** of data. Compute and return a checksum based on the contents of the
2252 ** page of data and the current value of pPager->cksumInit.
2254 ** This is not a real checksum. It is really just the sum of the
2255 ** random initial value (pPager->cksumInit) and every 200th byte
2256 ** of the page data, starting with byte offset (pPager->pageSize%200).
2257 ** Each byte is interpreted as an 8-bit unsigned integer.
2259 ** Changing the formula used to compute this checksum results in an
2260 ** incompatible journal file format.
2262 ** If journal corruption occurs due to a power failure, the most likely
2263 ** scenario is that one end or the other of the record will be changed.
2264 ** It is much less likely that the two ends of the journal record will be
2265 ** correct and the middle be corrupt. Thus, this "checksum" scheme,
2266 ** though fast and simple, catches the mostly likely kind of corruption.
2268 static u32
pager_cksum(Pager
*pPager
, const u8
*aData
){
2269 u32 cksum
= pPager
->cksumInit
; /* Checksum value to return */
2270 int i
= pPager
->pageSize
-200; /* Loop counter */
2279 ** Report the current page size and number of reserved bytes back
2282 /* BEGIN SQLCIPHER */
2283 #ifdef SQLITE_HAS_CODEC
2284 static void pagerReportSize(Pager
*pPager
){
2285 if( pPager
->xCodecSizeChng
){
2286 pPager
->xCodecSizeChng(pPager
->pCodec
, pPager
->pageSize
,
2287 (int)pPager
->nReserve
);
2291 # define pagerReportSize(X) /* No-op if we do not support a codec */
2295 /* BEGIN SQLCIPHER */
2296 #ifdef SQLITE_HAS_CODEC
2298 ** Make sure the number of reserved bits is the same in the destination
2299 ** pager as it is in the source. This comes up when a VACUUM changes the
2300 ** number of reserved bits to the "optimal" amount.
2302 void sqlite3PagerAlignReserve(Pager
*pDest
, Pager
*pSrc
){
2303 if( pDest
->nReserve
!=pSrc
->nReserve
){
2304 pDest
->nReserve
= pSrc
->nReserve
;
2305 pagerReportSize(pDest
);
2312 ** Read a single page from either the journal file (if isMainJrnl==1) or
2313 ** from the sub-journal (if isMainJrnl==0) and playback that page.
2314 ** The page begins at offset *pOffset into the file. The *pOffset
2315 ** value is increased to the start of the next page in the journal.
2317 ** The main rollback journal uses checksums - the statement journal does
2320 ** If the page number of the page record read from the (sub-)journal file
2321 ** is greater than the current value of Pager.dbSize, then playback is
2322 ** skipped and SQLITE_OK is returned.
2324 ** If pDone is not NULL, then it is a record of pages that have already
2325 ** been played back. If the page at *pOffset has already been played back
2326 ** (if the corresponding pDone bit is set) then skip the playback.
2327 ** Make sure the pDone bit corresponding to the *pOffset page is set
2328 ** prior to returning.
2330 ** If the page record is successfully read from the (sub-)journal file
2331 ** and played back, then SQLITE_OK is returned. If an IO error occurs
2332 ** while reading the record from the (sub-)journal file or while writing
2333 ** to the database file, then the IO error code is returned. If data
2334 ** is successfully read from the (sub-)journal file but appears to be
2335 ** corrupted, SQLITE_DONE is returned. Data is considered corrupted in
2336 ** two circumstances:
2338 ** * If the record page-number is illegal (0 or PAGER_SJ_PGNO), or
2339 ** * If the record is being rolled back from the main journal file
2340 ** and the checksum field does not match the record content.
2342 ** Neither of these two scenarios are possible during a savepoint rollback.
2344 ** If this is a savepoint rollback, then memory may have to be dynamically
2345 ** allocated by this function. If this is the case and an allocation fails,
2346 ** SQLITE_NOMEM is returned.
2348 static int pager_playback_one_page(
2349 Pager
*pPager
, /* The pager being played back */
2350 i64
*pOffset
, /* Offset of record to playback */
2351 Bitvec
*pDone
, /* Bitvec of pages already played back */
2352 int isMainJrnl
, /* 1 -> main journal. 0 -> sub-journal. */
2353 int isSavepnt
/* True for a savepoint rollback */
2356 PgHdr
*pPg
; /* An existing page in the cache */
2357 Pgno pgno
; /* The page number of a page in journal */
2358 u32 cksum
; /* Checksum used for sanity checking */
2359 char *aData
; /* Temporary storage for the page */
2360 sqlite3_file
*jfd
; /* The file descriptor for the journal file */
2361 int isSynced
; /* True if journal page is synced */
2362 /* BEGIN SQLCIPHER */
2363 #ifdef SQLITE_HAS_CODEC
2364 /* The jrnlEnc flag is true if Journal pages should be passed through
2365 ** the codec. It is false for pure in-memory journals. */
2366 const int jrnlEnc
= (isMainJrnl
|| pPager
->subjInMemory
==0);
2370 assert( (isMainJrnl
&~1)==0 ); /* isMainJrnl is 0 or 1 */
2371 assert( (isSavepnt
&~1)==0 ); /* isSavepnt is 0 or 1 */
2372 assert( isMainJrnl
|| pDone
); /* pDone always used on sub-journals */
2373 assert( isSavepnt
|| pDone
==0 ); /* pDone never used on non-savepoint */
2375 aData
= pPager
->pTmpSpace
;
2376 assert( aData
); /* Temp storage must have already been allocated */
2377 assert( pagerUseWal(pPager
)==0 || (!isMainJrnl
&& isSavepnt
) );
2379 /* Either the state is greater than PAGER_WRITER_CACHEMOD (a transaction
2380 ** or savepoint rollback done at the request of the caller) or this is
2381 ** a hot-journal rollback. If it is a hot-journal rollback, the pager
2382 ** is in state OPEN and holds an EXCLUSIVE lock. Hot-journal rollback
2383 ** only reads from the main journal, not the sub-journal.
2385 assert( pPager
->eState
>=PAGER_WRITER_CACHEMOD
2386 || (pPager
->eState
==PAGER_OPEN
&& pPager
->eLock
==EXCLUSIVE_LOCK
)
2388 assert( pPager
->eState
>=PAGER_WRITER_CACHEMOD
|| isMainJrnl
);
2390 /* Read the page number and page data from the journal or sub-journal
2391 ** file. Return an error code to the caller if an IO error occurs.
2393 jfd
= isMainJrnl
? pPager
->jfd
: pPager
->sjfd
;
2394 rc
= read32bits(jfd
, *pOffset
, &pgno
);
2395 if( rc
!=SQLITE_OK
) return rc
;
2396 rc
= sqlite3OsRead(jfd
, (u8
*)aData
, pPager
->pageSize
, (*pOffset
)+4);
2397 if( rc
!=SQLITE_OK
) return rc
;
2398 *pOffset
+= pPager
->pageSize
+ 4 + isMainJrnl
*4;
2400 /* Sanity checking on the page. This is more important that I originally
2401 ** thought. If a power failure occurs while the journal is being written,
2402 ** it could cause invalid data to be written into the journal. We need to
2403 ** detect this invalid data (with high probability) and ignore it.
2405 if( pgno
==0 || pgno
==PAGER_SJ_PGNO(pPager
) ){
2406 assert( !isSavepnt
);
2409 if( pgno
>(Pgno
)pPager
->dbSize
|| sqlite3BitvecTest(pDone
, pgno
) ){
2413 rc
= read32bits(jfd
, (*pOffset
)-4, &cksum
);
2415 if( !isSavepnt
&& pager_cksum(pPager
, (u8
*)aData
)!=cksum
){
2420 /* If this page has already been played back before during the current
2421 ** rollback, then don't bother to play it back again.
2423 if( pDone
&& (rc
= sqlite3BitvecSet(pDone
, pgno
))!=SQLITE_OK
){
2427 /* When playing back page 1, restore the nReserve setting
2429 if( pgno
==1 && pPager
->nReserve
!=((u8
*)aData
)[20] ){
2430 pPager
->nReserve
= ((u8
*)aData
)[20];
2431 pagerReportSize(pPager
);
2434 /* If the pager is in CACHEMOD state, then there must be a copy of this
2435 ** page in the pager cache. In this case just update the pager cache,
2436 ** not the database file. The page is left marked dirty in this case.
2438 ** An exception to the above rule: If the database is in no-sync mode
2439 ** and a page is moved during an incremental vacuum then the page may
2440 ** not be in the pager cache. Later: if a malloc() or IO error occurs
2441 ** during a Movepage() call, then the page may not be in the cache
2442 ** either. So the condition described in the above paragraph is not
2445 ** If in WRITER_DBMOD, WRITER_FINISHED or OPEN state, then we update the
2446 ** pager cache if it exists and the main file. The page is then marked
2447 ** not dirty. Since this code is only executed in PAGER_OPEN state for
2448 ** a hot-journal rollback, it is guaranteed that the page-cache is empty
2449 ** if the pager is in OPEN state.
2451 ** Ticket #1171: The statement journal might contain page content that is
2452 ** different from the page content at the start of the transaction.
2453 ** This occurs when a page is changed prior to the start of a statement
2454 ** then changed again within the statement. When rolling back such a
2455 ** statement we must not write to the original database unless we know
2456 ** for certain that original page contents are synced into the main rollback
2457 ** journal. Otherwise, a power loss might leave modified data in the
2458 ** database file without an entry in the rollback journal that can
2459 ** restore the database to its original form. Two conditions must be
2460 ** met before writing to the database files. (1) the database must be
2461 ** locked. (2) we know that the original page content is fully synced
2462 ** in the main journal either because the page is not in cache or else
2463 ** the page is marked as needSync==0.
2465 ** 2008-04-14: When attempting to vacuum a corrupt database file, it
2466 ** is possible to fail a statement on a database that does not yet exist.
2467 ** Do not attempt to write if database file has never been opened.
2469 if( pagerUseWal(pPager
) ){
2472 pPg
= sqlite3PagerLookup(pPager
, pgno
);
2474 assert( pPg
|| !MEMDB
);
2475 assert( pPager
->eState
!=PAGER_OPEN
|| pPg
==0 || pPager
->tempFile
);
2476 PAGERTRACE(("PLAYBACK %d page %d hash(%08x) %s\n",
2477 PAGERID(pPager
), pgno
, pager_datahash(pPager
->pageSize
, (u8
*)aData
),
2478 (isMainJrnl
?"main-journal":"sub-journal")
2481 isSynced
= pPager
->noSync
|| (*pOffset
<= pPager
->journalHdr
);
2483 isSynced
= (pPg
==0 || 0==(pPg
->flags
& PGHDR_NEED_SYNC
));
2485 if( isOpen(pPager
->fd
)
2486 && (pPager
->eState
>=PAGER_WRITER_DBMOD
|| pPager
->eState
==PAGER_OPEN
)
2489 i64 ofst
= (pgno
-1)*(i64
)pPager
->pageSize
;
2490 testcase( !isSavepnt
&& pPg
!=0 && (pPg
->flags
&PGHDR_NEED_SYNC
)!=0 );
2491 assert( !pagerUseWal(pPager
) );
2493 /* Write the data read from the journal back into the database file.
2494 ** This is usually safe even for an encrypted database - as the data
2495 ** was encrypted before it was written to the journal file. The exception
2496 ** is if the data was just read from an in-memory sub-journal. In that
2497 ** case it must be encrypted here before it is copied into the database
2499 /* BEGIN SQLCIPHER */
2500 #ifdef SQLITE_HAS_CODEC
2502 CODEC2(pPager
, aData
, pgno
, 7, rc
=SQLITE_NOMEM_BKPT
, aData
);
2503 rc
= sqlite3OsWrite(pPager
->fd
, (u8
*)aData
, pPager
->pageSize
, ofst
);
2504 CODEC1(pPager
, aData
, pgno
, 3, rc
=SQLITE_NOMEM_BKPT
);
2508 rc
= sqlite3OsWrite(pPager
->fd
, (u8
*)aData
, pPager
->pageSize
, ofst
);
2510 if( pgno
>pPager
->dbFileSize
){
2511 pPager
->dbFileSize
= pgno
;
2513 if( pPager
->pBackup
){
2514 /* BEGIN SQLCIPHER */
2515 #ifdef SQLITE_HAS_CODEC
2517 CODEC1(pPager
, aData
, pgno
, 3, rc
=SQLITE_NOMEM_BKPT
);
2518 sqlite3BackupUpdate(pPager
->pBackup
, pgno
, (u8
*)aData
);
2519 CODEC2(pPager
, aData
, pgno
, 7, rc
=SQLITE_NOMEM_BKPT
,aData
);
2523 sqlite3BackupUpdate(pPager
->pBackup
, pgno
, (u8
*)aData
);
2525 }else if( !isMainJrnl
&& pPg
==0 ){
2526 /* If this is a rollback of a savepoint and data was not written to
2527 ** the database and the page is not in-memory, there is a potential
2528 ** problem. When the page is next fetched by the b-tree layer, it
2529 ** will be read from the database file, which may or may not be
2532 ** There are a couple of different ways this can happen. All are quite
2533 ** obscure. When running in synchronous mode, this can only happen
2534 ** if the page is on the free-list at the start of the transaction, then
2535 ** populated, then moved using sqlite3PagerMovepage().
2537 ** The solution is to add an in-memory page to the cache containing
2538 ** the data just read from the sub-journal. Mark the page as dirty
2539 ** and if the pager requires a journal-sync, then mark the page as
2540 ** requiring a journal-sync before it is written.
2542 assert( isSavepnt
);
2543 assert( (pPager
->doNotSpill
& SPILLFLAG_ROLLBACK
)==0 );
2544 pPager
->doNotSpill
|= SPILLFLAG_ROLLBACK
;
2545 rc
= sqlite3PagerGet(pPager
, pgno
, &pPg
, 1);
2546 assert( (pPager
->doNotSpill
& SPILLFLAG_ROLLBACK
)!=0 );
2547 pPager
->doNotSpill
&= ~SPILLFLAG_ROLLBACK
;
2548 if( rc
!=SQLITE_OK
) return rc
;
2549 sqlite3PcacheMakeDirty(pPg
);
2552 /* No page should ever be explicitly rolled back that is in use, except
2553 ** for page 1 which is held in use in order to keep the lock on the
2554 ** database active. However such a page may be rolled back as a result
2555 ** of an internal error resulting in an automatic call to
2556 ** sqlite3PagerRollback().
2560 memcpy(pData
, (u8
*)aData
, pPager
->pageSize
);
2561 pPager
->xReiniter(pPg
);
2562 /* It used to be that sqlite3PcacheMakeClean(pPg) was called here. But
2563 ** that call was dangerous and had no detectable benefit since the cache
2564 ** is normally cleaned by sqlite3PcacheCleanAll() after rollback and so
2565 ** has been removed. */
2566 pager_set_pagehash(pPg
);
2568 /* If this was page 1, then restore the value of Pager.dbFileVers.
2569 ** Do this before any decoding. */
2571 memcpy(&pPager
->dbFileVers
, &((u8
*)pData
)[24],sizeof(pPager
->dbFileVers
));
2574 /* Decode the page just read from disk */
2575 /* BEGIN SQLCIPHER */
2576 #if SQLITE_HAS_CODEC
2577 if( jrnlEnc
){ CODEC1(pPager
, pData
, pPg
->pgno
, 3, rc
=SQLITE_NOMEM_BKPT
); }
2580 sqlite3PcacheRelease(pPg
);
2586 ** Parameter zSuper is the name of a super-journal file. A single journal
2587 ** file that referred to the super-journal file has just been rolled back.
2588 ** This routine checks if it is possible to delete the super-journal file,
2589 ** and does so if it is.
2591 ** Argument zSuper may point to Pager.pTmpSpace. So that buffer is not
2592 ** available for use within this function.
2594 ** When a super-journal file is created, it is populated with the names
2595 ** of all of its child journals, one after another, formatted as utf-8
2596 ** encoded text. The end of each child journal file is marked with a
2597 ** nul-terminator byte (0x00). i.e. the entire contents of a super-journal
2598 ** file for a transaction involving two databases might be:
2600 ** "/home/bill/a.db-journal\x00/home/bill/b.db-journal\x00"
2602 ** A super-journal file may only be deleted once all of its child
2603 ** journals have been rolled back.
2605 ** This function reads the contents of the super-journal file into
2606 ** memory and loops through each of the child journal names. For
2607 ** each child journal, it checks if:
2609 ** * if the child journal exists, and if so
2610 ** * if the child journal contains a reference to super-journal
2613 ** If a child journal can be found that matches both of the criteria
2614 ** above, this function returns without doing anything. Otherwise, if
2615 ** no such child journal can be found, file zSuper is deleted from
2616 ** the file-system using sqlite3OsDelete().
2618 ** If an IO error within this function, an error code is returned. This
2619 ** function allocates memory by calling sqlite3Malloc(). If an allocation
2620 ** fails, SQLITE_NOMEM is returned. Otherwise, if no IO or malloc errors
2621 ** occur, SQLITE_OK is returned.
2623 ** TODO: This function allocates a single block of memory to load
2624 ** the entire contents of the super-journal file. This could be
2625 ** a couple of kilobytes or so - potentially larger than the page
2628 static int pager_delsuper(Pager
*pPager
, const char *zSuper
){
2629 sqlite3_vfs
*pVfs
= pPager
->pVfs
;
2630 int rc
; /* Return code */
2631 sqlite3_file
*pSuper
; /* Malloc'd super-journal file descriptor */
2632 sqlite3_file
*pJournal
; /* Malloc'd child-journal file descriptor */
2633 char *zSuperJournal
= 0; /* Contents of super-journal file */
2634 i64 nSuperJournal
; /* Size of super-journal file */
2635 char *zJournal
; /* Pointer to one journal within MJ file */
2636 char *zSuperPtr
; /* Space to hold super-journal filename */
2637 char *zFree
= 0; /* Free this buffer */
2638 int nSuperPtr
; /* Amount of space allocated to zSuperPtr[] */
2640 /* Allocate space for both the pJournal and pSuper file descriptors.
2641 ** If successful, open the super-journal file for reading.
2643 pSuper
= (sqlite3_file
*)sqlite3MallocZero(pVfs
->szOsFile
* 2);
2645 rc
= SQLITE_NOMEM_BKPT
;
2648 const int flags
= (SQLITE_OPEN_READONLY
|SQLITE_OPEN_SUPER_JOURNAL
);
2649 rc
= sqlite3OsOpen(pVfs
, zSuper
, pSuper
, flags
, 0);
2650 pJournal
= (sqlite3_file
*)(((u8
*)pSuper
) + pVfs
->szOsFile
);
2652 if( rc
!=SQLITE_OK
) goto delsuper_out
;
2654 /* Load the entire super-journal file into space obtained from
2655 ** sqlite3_malloc() and pointed to by zSuperJournal. Also obtain
2656 ** sufficient space (in zSuperPtr) to hold the names of super-journal
2657 ** files extracted from regular rollback-journals.
2659 rc
= sqlite3OsFileSize(pSuper
, &nSuperJournal
);
2660 if( rc
!=SQLITE_OK
) goto delsuper_out
;
2661 nSuperPtr
= pVfs
->mxPathname
+1;
2662 zFree
= sqlite3Malloc(4 + nSuperJournal
+ nSuperPtr
+ 2);
2664 rc
= SQLITE_NOMEM_BKPT
;
2667 zFree
[0] = zFree
[1] = zFree
[2] = zFree
[3] = 0;
2668 zSuperJournal
= &zFree
[4];
2669 zSuperPtr
= &zSuperJournal
[nSuperJournal
+2];
2670 rc
= sqlite3OsRead(pSuper
, zSuperJournal
, (int)nSuperJournal
, 0);
2671 if( rc
!=SQLITE_OK
) goto delsuper_out
;
2672 zSuperJournal
[nSuperJournal
] = 0;
2673 zSuperJournal
[nSuperJournal
+1] = 0;
2675 zJournal
= zSuperJournal
;
2676 while( (zJournal
-zSuperJournal
)<nSuperJournal
){
2678 rc
= sqlite3OsAccess(pVfs
, zJournal
, SQLITE_ACCESS_EXISTS
, &exists
);
2679 if( rc
!=SQLITE_OK
){
2683 /* One of the journals pointed to by the super-journal exists.
2684 ** Open it and check if it points at the super-journal. If
2685 ** so, return without deleting the super-journal file.
2686 ** NB: zJournal is really a MAIN_JOURNAL. But call it a
2687 ** SUPER_JOURNAL here so that the VFS will not send the zJournal
2688 ** name into sqlite3_database_file_object().
2691 int flags
= (SQLITE_OPEN_READONLY
|SQLITE_OPEN_SUPER_JOURNAL
);
2692 rc
= sqlite3OsOpen(pVfs
, zJournal
, pJournal
, flags
, 0);
2693 if( rc
!=SQLITE_OK
){
2697 rc
= readSuperJournal(pJournal
, zSuperPtr
, nSuperPtr
);
2698 sqlite3OsClose(pJournal
);
2699 if( rc
!=SQLITE_OK
){
2703 c
= zSuperPtr
[0]!=0 && strcmp(zSuperPtr
, zSuper
)==0;
2705 /* We have a match. Do not delete the super-journal file. */
2709 zJournal
+= (sqlite3Strlen30(zJournal
)+1);
2712 sqlite3OsClose(pSuper
);
2713 rc
= sqlite3OsDelete(pVfs
, zSuper
, 0);
2716 sqlite3_free(zFree
);
2718 sqlite3OsClose(pSuper
);
2719 assert( !isOpen(pJournal
) );
2720 sqlite3_free(pSuper
);
2727 ** This function is used to change the actual size of the database
2728 ** file in the file-system. This only happens when committing a transaction,
2729 ** or rolling back a transaction (including rolling back a hot-journal).
2731 ** If the main database file is not open, or the pager is not in either
2732 ** DBMOD or OPEN state, this function is a no-op. Otherwise, the size
2733 ** of the file is changed to nPage pages (nPage*pPager->pageSize bytes).
2734 ** If the file on disk is currently larger than nPage pages, then use the VFS
2735 ** xTruncate() method to truncate it.
2737 ** Or, it might be the case that the file on disk is smaller than
2738 ** nPage pages. Some operating system implementations can get confused if
2739 ** you try to truncate a file to some size that is larger than it
2740 ** currently is, so detect this case and write a single zero byte to
2741 ** the end of the new file instead.
2743 ** If successful, return SQLITE_OK. If an IO error occurs while modifying
2744 ** the database file, return the error code to the caller.
2746 static int pager_truncate(Pager
*pPager
, Pgno nPage
){
2748 assert( pPager
->eState
!=PAGER_ERROR
);
2749 assert( pPager
->eState
!=PAGER_READER
);
2750 PAGERTRACE(("Truncate %d npage %u\n", PAGERID(pPager
), nPage
));
2753 if( isOpen(pPager
->fd
)
2754 && (pPager
->eState
>=PAGER_WRITER_DBMOD
|| pPager
->eState
==PAGER_OPEN
)
2756 i64 currentSize
, newSize
;
2757 int szPage
= pPager
->pageSize
;
2758 assert( pPager
->eLock
==EXCLUSIVE_LOCK
);
2759 /* TODO: Is it safe to use Pager.dbFileSize here? */
2760 rc
= sqlite3OsFileSize(pPager
->fd
, ¤tSize
);
2761 newSize
= szPage
*(i64
)nPage
;
2762 if( rc
==SQLITE_OK
&& currentSize
!=newSize
){
2763 if( currentSize
>newSize
){
2764 rc
= sqlite3OsTruncate(pPager
->fd
, newSize
);
2765 }else if( (currentSize
+szPage
)<=newSize
){
2766 char *pTmp
= pPager
->pTmpSpace
;
2767 memset(pTmp
, 0, szPage
);
2768 testcase( (newSize
-szPage
) == currentSize
);
2769 testcase( (newSize
-szPage
) > currentSize
);
2770 sqlite3OsFileControlHint(pPager
->fd
, SQLITE_FCNTL_SIZE_HINT
, &newSize
);
2771 rc
= sqlite3OsWrite(pPager
->fd
, pTmp
, szPage
, newSize
-szPage
);
2773 if( rc
==SQLITE_OK
){
2774 pPager
->dbFileSize
= nPage
;
2782 ** Return a sanitized version of the sector-size of OS file pFile. The
2783 ** return value is guaranteed to lie between 32 and MAX_SECTOR_SIZE.
2785 int sqlite3SectorSize(sqlite3_file
*pFile
){
2786 int iRet
= sqlite3OsSectorSize(pFile
);
2789 }else if( iRet
>MAX_SECTOR_SIZE
){
2790 assert( MAX_SECTOR_SIZE
>=512 );
2791 iRet
= MAX_SECTOR_SIZE
;
2797 ** Set the value of the Pager.sectorSize variable for the given
2798 ** pager based on the value returned by the xSectorSize method
2799 ** of the open database file. The sector size will be used
2800 ** to determine the size and alignment of journal header and
2801 ** super-journal pointers within created journal files.
2803 ** For temporary files the effective sector size is always 512 bytes.
2805 ** Otherwise, for non-temporary files, the effective sector size is
2806 ** the value returned by the xSectorSize() method rounded up to 32 if
2807 ** it is less than 32, or rounded down to MAX_SECTOR_SIZE if it
2808 ** is greater than MAX_SECTOR_SIZE.
2810 ** If the file has the SQLITE_IOCAP_POWERSAFE_OVERWRITE property, then set
2811 ** the effective sector size to its minimum value (512). The purpose of
2812 ** pPager->sectorSize is to define the "blast radius" of bytes that
2813 ** might change if a crash occurs while writing to a single byte in
2814 ** that range. But with POWERSAFE_OVERWRITE, the blast radius is zero
2815 ** (that is what POWERSAFE_OVERWRITE means), so we minimize the sector
2816 ** size. For backwards compatibility of the rollback journal file format,
2817 ** we cannot reduce the effective sector size below 512.
2819 static void setSectorSize(Pager
*pPager
){
2820 assert( isOpen(pPager
->fd
) || pPager
->tempFile
);
2822 if( pPager
->tempFile
2823 || (sqlite3OsDeviceCharacteristics(pPager
->fd
) &
2824 SQLITE_IOCAP_POWERSAFE_OVERWRITE
)!=0
2826 /* Sector size doesn't matter for temporary files. Also, the file
2827 ** may not have been opened yet, in which case the OsSectorSize()
2828 ** call will segfault. */
2829 pPager
->sectorSize
= 512;
2831 pPager
->sectorSize
= sqlite3SectorSize(pPager
->fd
);
2836 ** Playback the journal and thus restore the database file to
2837 ** the state it was in before we started making changes.
2839 ** The journal file format is as follows:
2841 ** (1) 8 byte prefix. A copy of aJournalMagic[].
2842 ** (2) 4 byte big-endian integer which is the number of valid page records
2843 ** in the journal. If this value is 0xffffffff, then compute the
2844 ** number of page records from the journal size.
2845 ** (3) 4 byte big-endian integer which is the initial value for the
2847 ** (4) 4 byte integer which is the number of pages to truncate the
2848 ** database to during a rollback.
2849 ** (5) 4 byte big-endian integer which is the sector size. The header
2850 ** is this many bytes in size.
2851 ** (6) 4 byte big-endian integer which is the page size.
2852 ** (7) zero padding out to the next sector size.
2853 ** (8) Zero or more pages instances, each as follows:
2854 ** + 4 byte page number.
2855 ** + pPager->pageSize bytes of data.
2856 ** + 4 byte checksum
2858 ** When we speak of the journal header, we mean the first 7 items above.
2859 ** Each entry in the journal is an instance of the 8th item.
2861 ** Call the value from the second bullet "nRec". nRec is the number of
2862 ** valid page entries in the journal. In most cases, you can compute the
2863 ** value of nRec from the size of the journal file. But if a power
2864 ** failure occurred while the journal was being written, it could be the
2865 ** case that the size of the journal file had already been increased but
2866 ** the extra entries had not yet made it safely to disk. In such a case,
2867 ** the value of nRec computed from the file size would be too large. For
2868 ** that reason, we always use the nRec value in the header.
2870 ** If the nRec value is 0xffffffff it means that nRec should be computed
2871 ** from the file size. This value is used when the user selects the
2872 ** no-sync option for the journal. A power failure could lead to corruption
2873 ** in this case. But for things like temporary table (which will be
2874 ** deleted when the power is restored) we don't care.
2876 ** If the file opened as the journal file is not a well-formed
2877 ** journal file then all pages up to the first corrupted page are rolled
2878 ** back (or no pages if the journal header is corrupted). The journal file
2879 ** is then deleted and SQLITE_OK returned, just as if no corruption had
2880 ** been encountered.
2882 ** If an I/O or malloc() error occurs, the journal-file is not deleted
2883 ** and an error code is returned.
2885 ** The isHot parameter indicates that we are trying to rollback a journal
2886 ** that might be a hot journal. Or, it could be that the journal is
2887 ** preserved because of JOURNALMODE_PERSIST or JOURNALMODE_TRUNCATE.
2888 ** If the journal really is hot, reset the pager cache prior rolling
2889 ** back any content. If the journal is merely persistent, no reset is
2892 static int pager_playback(Pager
*pPager
, int isHot
){
2893 sqlite3_vfs
*pVfs
= pPager
->pVfs
;
2894 i64 szJ
; /* Size of the journal file in bytes */
2895 u32 nRec
; /* Number of Records in the journal */
2896 u32 u
; /* Unsigned loop counter */
2897 Pgno mxPg
= 0; /* Size of the original file in pages */
2898 int rc
; /* Result code of a subroutine */
2899 int res
= 1; /* Value returned by sqlite3OsAccess() */
2900 char *zSuper
= 0; /* Name of super-journal file if any */
2901 int needPagerReset
; /* True to reset page prior to first page rollback */
2902 int nPlayback
= 0; /* Total number of pages restored from journal */
2903 u32 savedPageSize
= pPager
->pageSize
;
2905 /* Figure out how many records are in the journal. Abort early if
2906 ** the journal is empty.
2908 assert( isOpen(pPager
->jfd
) );
2909 rc
= sqlite3OsFileSize(pPager
->jfd
, &szJ
);
2910 if( rc
!=SQLITE_OK
){
2914 /* Read the super-journal name from the journal, if it is present.
2915 ** If a super-journal file name is specified, but the file is not
2916 ** present on disk, then the journal is not hot and does not need to be
2919 ** TODO: Technically the following is an error because it assumes that
2920 ** buffer Pager.pTmpSpace is (mxPathname+1) bytes or larger. i.e. that
2921 ** (pPager->pageSize >= pPager->pVfs->mxPathname+1). Using os_unix.c,
2922 ** mxPathname is 512, which is the same as the minimum allowable value
2925 zSuper
= pPager
->pTmpSpace
;
2926 rc
= readSuperJournal(pPager
->jfd
, zSuper
, pPager
->pVfs
->mxPathname
+1);
2927 if( rc
==SQLITE_OK
&& zSuper
[0] ){
2928 rc
= sqlite3OsAccess(pVfs
, zSuper
, SQLITE_ACCESS_EXISTS
, &res
);
2931 if( rc
!=SQLITE_OK
|| !res
){
2934 pPager
->journalOff
= 0;
2935 needPagerReset
= isHot
;
2937 /* This loop terminates either when a readJournalHdr() or
2938 ** pager_playback_one_page() call returns SQLITE_DONE or an IO error
2942 /* Read the next journal header from the journal file. If there are
2943 ** not enough bytes left in the journal file for a complete header, or
2944 ** it is corrupted, then a process must have failed while writing it.
2945 ** This indicates nothing more needs to be rolled back.
2947 rc
= readJournalHdr(pPager
, isHot
, szJ
, &nRec
, &mxPg
);
2948 if( rc
!=SQLITE_OK
){
2949 if( rc
==SQLITE_DONE
){
2955 /* If nRec is 0xffffffff, then this journal was created by a process
2956 ** working in no-sync mode. This means that the rest of the journal
2957 ** file consists of pages, there are no more journal headers. Compute
2958 ** the value of nRec based on this assumption.
2960 if( nRec
==0xffffffff ){
2961 assert( pPager
->journalOff
==JOURNAL_HDR_SZ(pPager
) );
2962 nRec
= (int)((szJ
- JOURNAL_HDR_SZ(pPager
))/JOURNAL_PG_SZ(pPager
));
2965 /* If nRec is 0 and this rollback is of a transaction created by this
2966 ** process and if this is the final header in the journal, then it means
2967 ** that this part of the journal was being filled but has not yet been
2968 ** synced to disk. Compute the number of pages based on the remaining
2969 ** size of the file.
2971 ** The third term of the test was added to fix ticket #2565.
2972 ** When rolling back a hot journal, nRec==0 always means that the next
2973 ** chunk of the journal contains zero pages to be rolled back. But
2974 ** when doing a ROLLBACK and the nRec==0 chunk is the last chunk in
2975 ** the journal, it means that the journal might contain additional
2976 ** pages that need to be rolled back and that the number of pages
2977 ** should be computed based on the journal file size.
2979 if( nRec
==0 && !isHot
&&
2980 pPager
->journalHdr
+JOURNAL_HDR_SZ(pPager
)==pPager
->journalOff
){
2981 nRec
= (int)((szJ
- pPager
->journalOff
) / JOURNAL_PG_SZ(pPager
));
2984 /* If this is the first header read from the journal, truncate the
2985 ** database file back to its original size.
2987 if( pPager
->journalOff
==JOURNAL_HDR_SZ(pPager
) ){
2988 rc
= pager_truncate(pPager
, mxPg
);
2989 if( rc
!=SQLITE_OK
){
2992 pPager
->dbSize
= mxPg
;
2993 if( pPager
->mxPgno
<mxPg
){
2994 pPager
->mxPgno
= mxPg
;
2998 /* Copy original pages out of the journal and back into the
2999 ** database file and/or page cache.
3001 for(u
=0; u
<nRec
; u
++){
3002 if( needPagerReset
){
3003 pager_reset(pPager
);
3006 rc
= pager_playback_one_page(pPager
,&pPager
->journalOff
,0,1,0);
3007 if( rc
==SQLITE_OK
){
3010 if( rc
==SQLITE_DONE
){
3011 pPager
->journalOff
= szJ
;
3013 }else if( rc
==SQLITE_IOERR_SHORT_READ
){
3014 /* If the journal has been truncated, simply stop reading and
3015 ** processing the journal. This might happen if the journal was
3016 ** not completely written and synced prior to a crash. In that
3017 ** case, the database should have never been written in the
3018 ** first place so it is OK to simply abandon the rollback. */
3022 /* If we are unable to rollback, quit and return the error
3023 ** code. This will cause the pager to enter the error state
3024 ** so that no further harm will be done. Perhaps the next
3025 ** process to come along will be able to rollback the database.
3036 if( rc
==SQLITE_OK
){
3037 rc
= sqlite3PagerSetPagesize(pPager
, &savedPageSize
, -1);
3039 /* Following a rollback, the database file should be back in its original
3040 ** state prior to the start of the transaction, so invoke the
3041 ** SQLITE_FCNTL_DB_UNCHANGED file-control method to disable the
3042 ** assertion that the transaction counter was modified.
3045 sqlite3OsFileControlHint(pPager
->fd
,SQLITE_FCNTL_DB_UNCHANGED
,0);
3048 /* If this playback is happening automatically as a result of an IO or
3049 ** malloc error that occurred after the change-counter was updated but
3050 ** before the transaction was committed, then the change-counter
3051 ** modification may just have been reverted. If this happens in exclusive
3052 ** mode, then subsequent transactions performed by the connection will not
3053 ** update the change-counter at all. This may lead to cache inconsistency
3054 ** problems for other processes at some point in the future. So, just
3055 ** in case this has happened, clear the changeCountDone flag now.
3057 pPager
->changeCountDone
= pPager
->tempFile
;
3059 if( rc
==SQLITE_OK
){
3060 /* Leave 4 bytes of space before the super-journal filename in memory.
3061 ** This is because it may end up being passed to sqlite3OsOpen(), in
3062 ** which case it requires 4 0x00 bytes in memory immediately before
3064 zSuper
= &pPager
->pTmpSpace
[4];
3065 rc
= readSuperJournal(pPager
->jfd
, zSuper
, pPager
->pVfs
->mxPathname
+1);
3066 testcase( rc
!=SQLITE_OK
);
3069 && (pPager
->eState
>=PAGER_WRITER_DBMOD
|| pPager
->eState
==PAGER_OPEN
)
3071 rc
= sqlite3PagerSync(pPager
, 0);
3073 if( rc
==SQLITE_OK
){
3074 rc
= pager_end_transaction(pPager
, zSuper
[0]!='\0', 0);
3075 testcase( rc
!=SQLITE_OK
);
3077 if( rc
==SQLITE_OK
&& zSuper
[0] && res
){
3078 /* If there was a super-journal and this routine will return success,
3079 ** see if it is possible to delete the super-journal.
3081 assert( zSuper
==&pPager
->pTmpSpace
[4] );
3082 memset(pPager
->pTmpSpace
, 0, 4);
3083 rc
= pager_delsuper(pPager
, zSuper
);
3084 testcase( rc
!=SQLITE_OK
);
3086 if( isHot
&& nPlayback
){
3087 sqlite3_log(SQLITE_NOTICE_RECOVER_ROLLBACK
, "recovered %d pages from %s",
3088 nPlayback
, pPager
->zJournal
);
3091 /* The Pager.sectorSize variable may have been updated while rolling
3092 ** back a journal created by a process with a different sector size
3093 ** value. Reset it to the correct value for this process.
3095 setSectorSize(pPager
);
3101 ** Read the content for page pPg out of the database file (or out of
3102 ** the WAL if that is where the most recent copy if found) into
3103 ** pPg->pData. A shared lock or greater must be held on the database
3104 ** file before this function is called.
3106 ** If page 1 is read, then the value of Pager.dbFileVers[] is set to
3107 ** the value read from the database file.
3109 ** If an IO error occurs, then the IO error is returned to the caller.
3110 ** Otherwise, SQLITE_OK is returned.
3112 static int readDbPage(PgHdr
*pPg
){
3113 Pager
*pPager
= pPg
->pPager
; /* Pager object associated with page pPg */
3114 int rc
= SQLITE_OK
; /* Return code */
3116 #ifndef SQLITE_OMIT_WAL
3117 u32 iFrame
= 0; /* Frame of WAL containing pgno */
3119 assert( pPager
->eState
>=PAGER_READER
&& !MEMDB
);
3120 assert( isOpen(pPager
->fd
) );
3122 if( pagerUseWal(pPager
) ){
3123 rc
= sqlite3WalFindFrame(pPager
->pWal
, pPg
->pgno
, &iFrame
);
3127 rc
= sqlite3WalReadFrame(pPager
->pWal
, iFrame
,pPager
->pageSize
,pPg
->pData
);
3131 i64 iOffset
= (pPg
->pgno
-1)*(i64
)pPager
->pageSize
;
3132 rc
= sqlite3OsRead(pPager
->fd
, pPg
->pData
, pPager
->pageSize
, iOffset
);
3133 if( rc
==SQLITE_IOERR_SHORT_READ
){
3140 /* If the read is unsuccessful, set the dbFileVers[] to something
3141 ** that will never be a valid file version. dbFileVers[] is a copy
3142 ** of bytes 24..39 of the database. Bytes 28..31 should always be
3143 ** zero or the size of the database in page. Bytes 32..35 and 35..39
3144 ** should be page numbers which are never 0xffffffff. So filling
3145 ** pPager->dbFileVers[] with all 0xff bytes should suffice.
3147 ** For an encrypted database, the situation is more complex: bytes
3148 ** 24..39 of the database are white noise. But the probability of
3149 ** white noise equaling 16 bytes of 0xff is vanishingly small so
3150 ** we should still be ok.
3152 memset(pPager
->dbFileVers
, 0xff, sizeof(pPager
->dbFileVers
));
3154 u8
*dbFileVers
= &((u8
*)pPg
->pData
)[24];
3155 memcpy(&pPager
->dbFileVers
, dbFileVers
, sizeof(pPager
->dbFileVers
));
3158 CODEC1(pPager
, pPg
->pData
, pPg
->pgno
, 3, rc
= SQLITE_NOMEM_BKPT
);
3160 PAGER_INCR(sqlite3_pager_readdb_count
);
3161 PAGER_INCR(pPager
->nRead
);
3162 IOTRACE(("PGIN %p %d\n", pPager
, pPg
->pgno
));
3163 PAGERTRACE(("FETCH %d page %d hash(%08x)\n",
3164 PAGERID(pPager
), pPg
->pgno
, pager_pagehash(pPg
)));
3170 ** Update the value of the change-counter at offsets 24 and 92 in
3171 ** the header and the sqlite version number at offset 96.
3173 ** This is an unconditional update. See also the pager_incr_changecounter()
3174 ** routine which only updates the change-counter if the update is actually
3175 ** needed, as determined by the pPager->changeCountDone state variable.
3177 static void pager_write_changecounter(PgHdr
*pPg
){
3179 if( NEVER(pPg
==0) ) return;
3181 /* Increment the value just read and write it back to byte 24. */
3182 change_counter
= sqlite3Get4byte((u8
*)pPg
->pPager
->dbFileVers
)+1;
3183 put32bits(((char*)pPg
->pData
)+24, change_counter
);
3185 /* Also store the SQLite version number in bytes 96..99 and in
3186 ** bytes 92..95 store the change counter for which the version number
3188 put32bits(((char*)pPg
->pData
)+92, change_counter
);
3189 put32bits(((char*)pPg
->pData
)+96, SQLITE_VERSION_NUMBER
);
3192 #ifndef SQLITE_OMIT_WAL
3194 ** This function is invoked once for each page that has already been
3195 ** written into the log file when a WAL transaction is rolled back.
3196 ** Parameter iPg is the page number of said page. The pCtx argument
3197 ** is actually a pointer to the Pager structure.
3199 ** If page iPg is present in the cache, and has no outstanding references,
3200 ** it is discarded. Otherwise, if there are one or more outstanding
3201 ** references, the page content is reloaded from the database. If the
3202 ** attempt to reload content from the database is required and fails,
3203 ** return an SQLite error code. Otherwise, SQLITE_OK.
3205 static int pagerUndoCallback(void *pCtx
, Pgno iPg
){
3207 Pager
*pPager
= (Pager
*)pCtx
;
3210 assert( pagerUseWal(pPager
) );
3211 pPg
= sqlite3PagerLookup(pPager
, iPg
);
3213 if( sqlite3PcachePageRefcount(pPg
)==1 ){
3214 sqlite3PcacheDrop(pPg
);
3216 rc
= readDbPage(pPg
);
3217 if( rc
==SQLITE_OK
){
3218 pPager
->xReiniter(pPg
);
3220 sqlite3PagerUnrefNotNull(pPg
);
3224 /* Normally, if a transaction is rolled back, any backup processes are
3225 ** updated as data is copied out of the rollback journal and into the
3226 ** database. This is not generally possible with a WAL database, as
3227 ** rollback involves simply truncating the log file. Therefore, if one
3228 ** or more frames have already been written to the log (and therefore
3229 ** also copied into the backup databases) as part of this transaction,
3230 ** the backups must be restarted.
3232 sqlite3BackupRestart(pPager
->pBackup
);
3238 ** This function is called to rollback a transaction on a WAL database.
3240 static int pagerRollbackWal(Pager
*pPager
){
3241 int rc
; /* Return Code */
3242 PgHdr
*pList
; /* List of dirty pages to revert */
3244 /* For all pages in the cache that are currently dirty or have already
3245 ** been written (but not committed) to the log file, do one of the
3248 ** + Discard the cached page (if refcount==0), or
3249 ** + Reload page content from the database (if refcount>0).
3251 pPager
->dbSize
= pPager
->dbOrigSize
;
3252 rc
= sqlite3WalUndo(pPager
->pWal
, pagerUndoCallback
, (void *)pPager
);
3253 pList
= sqlite3PcacheDirtyList(pPager
->pPCache
);
3254 while( pList
&& rc
==SQLITE_OK
){
3255 PgHdr
*pNext
= pList
->pDirty
;
3256 rc
= pagerUndoCallback((void *)pPager
, pList
->pgno
);
3264 ** This function is a wrapper around sqlite3WalFrames(). As well as logging
3265 ** the contents of the list of pages headed by pList (connected by pDirty),
3266 ** this function notifies any active backup processes that the pages have
3269 ** The list of pages passed into this routine is always sorted by page number.
3270 ** Hence, if page 1 appears anywhere on the list, it will be the first page.
3272 static int pagerWalFrames(
3273 Pager
*pPager
, /* Pager object */
3274 PgHdr
*pList
, /* List of frames to log */
3275 Pgno nTruncate
, /* Database size after this commit */
3276 int isCommit
/* True if this is a commit */
3278 int rc
; /* Return code */
3279 int nList
; /* Number of pages in pList */
3280 PgHdr
*p
; /* For looping over pages */
3282 assert( pPager
->pWal
);
3285 /* Verify that the page list is in ascending order */
3286 for(p
=pList
; p
&& p
->pDirty
; p
=p
->pDirty
){
3287 assert( p
->pgno
< p
->pDirty
->pgno
);
3291 assert( pList
->pDirty
==0 || isCommit
);
3293 /* If a WAL transaction is being committed, there is no point in writing
3294 ** any pages with page numbers greater than nTruncate into the WAL file.
3295 ** They will never be read by any client. So remove them from the pDirty
3297 PgHdr
**ppNext
= &pList
;
3299 for(p
=pList
; (*ppNext
= p
)!=0; p
=p
->pDirty
){
3300 if( p
->pgno
<=nTruncate
){
3301 ppNext
= &p
->pDirty
;
3309 pPager
->aStat
[PAGER_STAT_WRITE
] += nList
;
3311 if( pList
->pgno
==1 ) pager_write_changecounter(pList
);
3312 rc
= sqlite3WalFrames(pPager
->pWal
,
3313 pPager
->pageSize
, pList
, nTruncate
, isCommit
, pPager
->walSyncFlags
3315 if( rc
==SQLITE_OK
&& pPager
->pBackup
){
3316 for(p
=pList
; p
; p
=p
->pDirty
){
3317 sqlite3BackupUpdate(pPager
->pBackup
, p
->pgno
, (u8
*)p
->pData
);
3321 #ifdef SQLITE_CHECK_PAGES
3322 pList
= sqlite3PcacheDirtyList(pPager
->pPCache
);
3323 for(p
=pList
; p
; p
=p
->pDirty
){
3324 pager_set_pagehash(p
);
3332 ** Begin a read transaction on the WAL.
3334 ** This routine used to be called "pagerOpenSnapshot()" because it essentially
3335 ** makes a snapshot of the database at the current point in time and preserves
3336 ** that snapshot for use by the reader in spite of concurrently changes by
3337 ** other writers or checkpointers.
3339 static int pagerBeginReadTransaction(Pager
*pPager
){
3340 int rc
; /* Return code */
3341 int changed
= 0; /* True if cache must be reset */
3343 assert( pagerUseWal(pPager
) );
3344 assert( pPager
->eState
==PAGER_OPEN
|| pPager
->eState
==PAGER_READER
);
3346 /* sqlite3WalEndReadTransaction() was not called for the previous
3347 ** transaction in locking_mode=EXCLUSIVE. So call it now. If we
3348 ** are in locking_mode=NORMAL and EndRead() was previously called,
3349 ** the duplicate call is harmless.
3351 sqlite3WalEndReadTransaction(pPager
->pWal
);
3353 rc
= sqlite3WalBeginReadTransaction(pPager
->pWal
, &changed
);
3354 if( rc
!=SQLITE_OK
|| changed
){
3355 pager_reset(pPager
);
3356 if( USEFETCH(pPager
) ) sqlite3OsUnfetch(pPager
->fd
, 0, 0);
3364 ** This function is called as part of the transition from PAGER_OPEN
3365 ** to PAGER_READER state to determine the size of the database file
3366 ** in pages (assuming the page size currently stored in Pager.pageSize).
3368 ** If no error occurs, SQLITE_OK is returned and the size of the database
3369 ** in pages is stored in *pnPage. Otherwise, an error code (perhaps
3370 ** SQLITE_IOERR_FSTAT) is returned and *pnPage is left unmodified.
3372 static int pagerPagecount(Pager
*pPager
, Pgno
*pnPage
){
3373 Pgno nPage
; /* Value to return via *pnPage */
3375 /* Query the WAL sub-system for the database size. The WalDbsize()
3376 ** function returns zero if the WAL is not open (i.e. Pager.pWal==0), or
3377 ** if the database size is not available. The database size is not
3378 ** available from the WAL sub-system if the log file is empty or
3379 ** contains no valid committed transactions.
3381 assert( pPager
->eState
==PAGER_OPEN
);
3382 assert( pPager
->eLock
>=SHARED_LOCK
);
3383 assert( isOpen(pPager
->fd
) );
3384 assert( pPager
->tempFile
==0 );
3385 nPage
= sqlite3WalDbsize(pPager
->pWal
);
3387 /* If the number of pages in the database is not available from the
3388 ** WAL sub-system, determine the page count based on the size of
3389 ** the database file. If the size of the database file is not an
3390 ** integer multiple of the page-size, round up the result.
3392 if( nPage
==0 && ALWAYS(isOpen(pPager
->fd
)) ){
3393 i64 n
= 0; /* Size of db file in bytes */
3394 int rc
= sqlite3OsFileSize(pPager
->fd
, &n
);
3395 if( rc
!=SQLITE_OK
){
3398 nPage
= (Pgno
)((n
+pPager
->pageSize
-1) / pPager
->pageSize
);
3401 /* If the current number of pages in the file is greater than the
3402 ** configured maximum pager number, increase the allowed limit so
3403 ** that the file can be read.
3405 if( nPage
>pPager
->mxPgno
){
3406 pPager
->mxPgno
= (Pgno
)nPage
;
3413 #ifndef SQLITE_OMIT_WAL
3415 ** Check if the *-wal file that corresponds to the database opened by pPager
3416 ** exists if the database is not empty, or verify that the *-wal file does
3417 ** not exist (by deleting it) if the database file is empty.
3419 ** If the database is not empty and the *-wal file exists, open the pager
3420 ** in WAL mode. If the database is empty or if no *-wal file exists and
3421 ** if no error occurs, make sure Pager.journalMode is not set to
3422 ** PAGER_JOURNALMODE_WAL.
3424 ** Return SQLITE_OK or an error code.
3426 ** The caller must hold a SHARED lock on the database file to call this
3427 ** function. Because an EXCLUSIVE lock on the db file is required to delete
3428 ** a WAL on a none-empty database, this ensures there is no race condition
3429 ** between the xAccess() below and an xDelete() being executed by some
3430 ** other connection.
3432 static int pagerOpenWalIfPresent(Pager
*pPager
){
3434 assert( pPager
->eState
==PAGER_OPEN
);
3435 assert( pPager
->eLock
>=SHARED_LOCK
);
3437 if( !pPager
->tempFile
){
3438 int isWal
; /* True if WAL file exists */
3439 rc
= sqlite3OsAccess(
3440 pPager
->pVfs
, pPager
->zWal
, SQLITE_ACCESS_EXISTS
, &isWal
3442 if( rc
==SQLITE_OK
){
3444 Pgno nPage
; /* Size of the database file */
3446 rc
= pagerPagecount(pPager
, &nPage
);
3449 rc
= sqlite3OsDelete(pPager
->pVfs
, pPager
->zWal
, 0);
3451 testcase( sqlite3PcachePagecount(pPager
->pPCache
)==0 );
3452 rc
= sqlite3PagerOpenWal(pPager
, 0);
3454 }else if( pPager
->journalMode
==PAGER_JOURNALMODE_WAL
){
3455 pPager
->journalMode
= PAGER_JOURNALMODE_DELETE
;
3464 ** Playback savepoint pSavepoint. Or, if pSavepoint==NULL, then playback
3465 ** the entire super-journal file. The case pSavepoint==NULL occurs when
3466 ** a ROLLBACK TO command is invoked on a SAVEPOINT that is a transaction
3469 ** When pSavepoint is not NULL (meaning a non-transaction savepoint is
3470 ** being rolled back), then the rollback consists of up to three stages,
3471 ** performed in the order specified:
3473 ** * Pages are played back from the main journal starting at byte
3474 ** offset PagerSavepoint.iOffset and continuing to
3475 ** PagerSavepoint.iHdrOffset, or to the end of the main journal
3476 ** file if PagerSavepoint.iHdrOffset is zero.
3478 ** * If PagerSavepoint.iHdrOffset is not zero, then pages are played
3479 ** back starting from the journal header immediately following
3480 ** PagerSavepoint.iHdrOffset to the end of the main journal file.
3482 ** * Pages are then played back from the sub-journal file, starting
3483 ** with the PagerSavepoint.iSubRec and continuing to the end of
3484 ** the journal file.
3486 ** Throughout the rollback process, each time a page is rolled back, the
3487 ** corresponding bit is set in a bitvec structure (variable pDone in the
3488 ** implementation below). This is used to ensure that a page is only
3489 ** rolled back the first time it is encountered in either journal.
3491 ** If pSavepoint is NULL, then pages are only played back from the main
3492 ** journal file. There is no need for a bitvec in this case.
3494 ** In either case, before playback commences the Pager.dbSize variable
3495 ** is reset to the value that it held at the start of the savepoint
3496 ** (or transaction). No page with a page-number greater than this value
3497 ** is played back. If one is encountered it is simply skipped.
3499 static int pagerPlaybackSavepoint(Pager
*pPager
, PagerSavepoint
*pSavepoint
){
3500 i64 szJ
; /* Effective size of the main journal */
3501 i64 iHdrOff
; /* End of first segment of main-journal records */
3502 int rc
= SQLITE_OK
; /* Return code */
3503 Bitvec
*pDone
= 0; /* Bitvec to ensure pages played back only once */
3505 assert( pPager
->eState
!=PAGER_ERROR
);
3506 assert( pPager
->eState
>=PAGER_WRITER_LOCKED
);
3508 /* Allocate a bitvec to use to store the set of pages rolled back */
3510 pDone
= sqlite3BitvecCreate(pSavepoint
->nOrig
);
3512 return SQLITE_NOMEM_BKPT
;
3516 /* Set the database size back to the value it was before the savepoint
3517 ** being reverted was opened.
3519 pPager
->dbSize
= pSavepoint
? pSavepoint
->nOrig
: pPager
->dbOrigSize
;
3520 pPager
->changeCountDone
= pPager
->tempFile
;
3522 if( !pSavepoint
&& pagerUseWal(pPager
) ){
3523 return pagerRollbackWal(pPager
);
3526 /* Use pPager->journalOff as the effective size of the main rollback
3527 ** journal. The actual file might be larger than this in
3528 ** PAGER_JOURNALMODE_TRUNCATE or PAGER_JOURNALMODE_PERSIST. But anything
3529 ** past pPager->journalOff is off-limits to us.
3531 szJ
= pPager
->journalOff
;
3532 assert( pagerUseWal(pPager
)==0 || szJ
==0 );
3534 /* Begin by rolling back records from the main journal starting at
3535 ** PagerSavepoint.iOffset and continuing to the next journal header.
3536 ** There might be records in the main journal that have a page number
3537 ** greater than the current database size (pPager->dbSize) but those
3538 ** will be skipped automatically. Pages are added to pDone as they
3541 if( pSavepoint
&& !pagerUseWal(pPager
) ){
3542 iHdrOff
= pSavepoint
->iHdrOffset
? pSavepoint
->iHdrOffset
: szJ
;
3543 pPager
->journalOff
= pSavepoint
->iOffset
;
3544 while( rc
==SQLITE_OK
&& pPager
->journalOff
<iHdrOff
){
3545 rc
= pager_playback_one_page(pPager
, &pPager
->journalOff
, pDone
, 1, 1);
3547 assert( rc
!=SQLITE_DONE
);
3549 pPager
->journalOff
= 0;
3552 /* Continue rolling back records out of the main journal starting at
3553 ** the first journal header seen and continuing until the effective end
3554 ** of the main journal file. Continue to skip out-of-range pages and
3555 ** continue adding pages rolled back to pDone.
3557 while( rc
==SQLITE_OK
&& pPager
->journalOff
<szJ
){
3558 u32 ii
; /* Loop counter */
3559 u32 nJRec
= 0; /* Number of Journal Records */
3561 rc
= readJournalHdr(pPager
, 0, szJ
, &nJRec
, &dummy
);
3562 assert( rc
!=SQLITE_DONE
);
3565 ** The "pPager->journalHdr+JOURNAL_HDR_SZ(pPager)==pPager->journalOff"
3566 ** test is related to ticket #2565. See the discussion in the
3567 ** pager_playback() function for additional information.
3570 && pPager
->journalHdr
+JOURNAL_HDR_SZ(pPager
)==pPager
->journalOff
3572 nJRec
= (u32
)((szJ
- pPager
->journalOff
)/JOURNAL_PG_SZ(pPager
));
3574 for(ii
=0; rc
==SQLITE_OK
&& ii
<nJRec
&& pPager
->journalOff
<szJ
; ii
++){
3575 rc
= pager_playback_one_page(pPager
, &pPager
->journalOff
, pDone
, 1, 1);
3577 assert( rc
!=SQLITE_DONE
);
3579 assert( rc
!=SQLITE_OK
|| pPager
->journalOff
>=szJ
);
3581 /* Finally, rollback pages from the sub-journal. Page that were
3582 ** previously rolled back out of the main journal (and are hence in pDone)
3583 ** will be skipped. Out-of-range pages are also skipped.
3586 u32 ii
; /* Loop counter */
3587 i64 offset
= (i64
)pSavepoint
->iSubRec
*(4+pPager
->pageSize
);
3589 if( pagerUseWal(pPager
) ){
3590 rc
= sqlite3WalSavepointUndo(pPager
->pWal
, pSavepoint
->aWalData
);
3592 for(ii
=pSavepoint
->iSubRec
; rc
==SQLITE_OK
&& ii
<pPager
->nSubRec
; ii
++){
3593 assert( offset
==(i64
)ii
*(4+pPager
->pageSize
) );
3594 rc
= pager_playback_one_page(pPager
, &offset
, pDone
, 0, 1);
3596 assert( rc
!=SQLITE_DONE
);
3599 sqlite3BitvecDestroy(pDone
);
3600 if( rc
==SQLITE_OK
){
3601 pPager
->journalOff
= szJ
;
3608 ** Change the maximum number of in-memory pages that are allowed
3609 ** before attempting to recycle clean and unused pages.
3611 void sqlite3PagerSetCachesize(Pager
*pPager
, int mxPage
){
3612 sqlite3PcacheSetCachesize(pPager
->pPCache
, mxPage
);
3616 ** Change the maximum number of in-memory pages that are allowed
3617 ** before attempting to spill pages to journal.
3619 int sqlite3PagerSetSpillsize(Pager
*pPager
, int mxPage
){
3620 return sqlite3PcacheSetSpillsize(pPager
->pPCache
, mxPage
);
3624 ** Invoke SQLITE_FCNTL_MMAP_SIZE based on the current value of szMmap.
3626 static void pagerFixMaplimit(Pager
*pPager
){
3627 #if SQLITE_MAX_MMAP_SIZE>0
3628 sqlite3_file
*fd
= pPager
->fd
;
3629 if( isOpen(fd
) && fd
->pMethods
->iVersion
>=3 ){
3631 sz
= pPager
->szMmap
;
3632 pPager
->bUseFetch
= (sz
>0);
3633 setGetterMethod(pPager
);
3634 sqlite3OsFileControlHint(pPager
->fd
, SQLITE_FCNTL_MMAP_SIZE
, &sz
);
3640 ** Change the maximum size of any memory mapping made of the database file.
3642 void sqlite3PagerSetMmapLimit(Pager
*pPager
, sqlite3_int64 szMmap
){
3643 pPager
->szMmap
= szMmap
;
3644 pagerFixMaplimit(pPager
);
3648 ** Free as much memory as possible from the pager.
3650 void sqlite3PagerShrink(Pager
*pPager
){
3651 sqlite3PcacheShrink(pPager
->pPCache
);
3655 ** Adjust settings of the pager to those specified in the pgFlags parameter.
3657 ** The "level" in pgFlags & PAGER_SYNCHRONOUS_MASK sets the robustness
3658 ** of the database to damage due to OS crashes or power failures by
3659 ** changing the number of syncs()s when writing the journals.
3660 ** There are four levels:
3662 ** OFF sqlite3OsSync() is never called. This is the default
3663 ** for temporary and transient files.
3665 ** NORMAL The journal is synced once before writes begin on the
3666 ** database. This is normally adequate protection, but
3667 ** it is theoretically possible, though very unlikely,
3668 ** that an inopertune power failure could leave the journal
3669 ** in a state which would cause damage to the database
3670 ** when it is rolled back.
3672 ** FULL The journal is synced twice before writes begin on the
3673 ** database (with some additional information - the nRec field
3674 ** of the journal header - being written in between the two
3675 ** syncs). If we assume that writing a
3676 ** single disk sector is atomic, then this mode provides
3677 ** assurance that the journal will not be corrupted to the
3678 ** point of causing damage to the database during rollback.
3680 ** EXTRA This is like FULL except that is also syncs the directory
3681 ** that contains the rollback journal after the rollback
3682 ** journal is unlinked.
3684 ** The above is for a rollback-journal mode. For WAL mode, OFF continues
3685 ** to mean that no syncs ever occur. NORMAL means that the WAL is synced
3686 ** prior to the start of checkpoint and that the database file is synced
3687 ** at the conclusion of the checkpoint if the entire content of the WAL
3688 ** was written back into the database. But no sync operations occur for
3689 ** an ordinary commit in NORMAL mode with WAL. FULL means that the WAL
3690 ** file is synced following each commit operation, in addition to the
3691 ** syncs associated with NORMAL. There is no difference between FULL
3692 ** and EXTRA for WAL mode.
3694 ** Do not confuse synchronous=FULL with SQLITE_SYNC_FULL. The
3695 ** SQLITE_SYNC_FULL macro means to use the MacOSX-style full-fsync
3696 ** using fcntl(F_FULLFSYNC). SQLITE_SYNC_NORMAL means to do an
3697 ** ordinary fsync() call. There is no difference between SQLITE_SYNC_FULL
3698 ** and SQLITE_SYNC_NORMAL on platforms other than MacOSX. But the
3699 ** synchronous=FULL versus synchronous=NORMAL setting determines when
3700 ** the xSync primitive is called and is relevant to all platforms.
3702 ** Numeric values associated with these states are OFF==1, NORMAL=2,
3705 void sqlite3PagerSetFlags(
3706 Pager
*pPager
, /* The pager to set safety level for */
3707 unsigned pgFlags
/* Various flags */
3709 unsigned level
= pgFlags
& PAGER_SYNCHRONOUS_MASK
;
3710 if( pPager
->tempFile
){
3712 pPager
->fullSync
= 0;
3713 pPager
->extraSync
= 0;
3715 pPager
->noSync
= level
==PAGER_SYNCHRONOUS_OFF
?1:0;
3716 pPager
->fullSync
= level
>=PAGER_SYNCHRONOUS_FULL
?1:0;
3717 pPager
->extraSync
= level
==PAGER_SYNCHRONOUS_EXTRA
?1:0;
3719 if( pPager
->noSync
){
3720 pPager
->syncFlags
= 0;
3721 }else if( pgFlags
& PAGER_FULLFSYNC
){
3722 pPager
->syncFlags
= SQLITE_SYNC_FULL
;
3724 pPager
->syncFlags
= SQLITE_SYNC_NORMAL
;
3726 pPager
->walSyncFlags
= (pPager
->syncFlags
<<2);
3727 if( pPager
->fullSync
){
3728 pPager
->walSyncFlags
|= pPager
->syncFlags
;
3730 if( (pgFlags
& PAGER_CKPT_FULLFSYNC
) && !pPager
->noSync
){
3731 pPager
->walSyncFlags
|= (SQLITE_SYNC_FULL
<<2);
3733 if( pgFlags
& PAGER_CACHESPILL
){
3734 pPager
->doNotSpill
&= ~SPILLFLAG_OFF
;
3736 pPager
->doNotSpill
|= SPILLFLAG_OFF
;
3741 ** The following global variable is incremented whenever the library
3742 ** attempts to open a temporary file. This information is used for
3743 ** testing and analysis only.
3746 int sqlite3_opentemp_count
= 0;
3750 ** Open a temporary file.
3752 ** Write the file descriptor into *pFile. Return SQLITE_OK on success
3753 ** or some other error code if we fail. The OS will automatically
3754 ** delete the temporary file when it is closed.
3756 ** The flags passed to the VFS layer xOpen() call are those specified
3757 ** by parameter vfsFlags ORed with the following:
3759 ** SQLITE_OPEN_READWRITE
3760 ** SQLITE_OPEN_CREATE
3761 ** SQLITE_OPEN_EXCLUSIVE
3762 ** SQLITE_OPEN_DELETEONCLOSE
3764 static int pagerOpentemp(
3765 Pager
*pPager
, /* The pager object */
3766 sqlite3_file
*pFile
, /* Write the file descriptor here */
3767 int vfsFlags
/* Flags passed through to the VFS */
3769 int rc
; /* Return code */
3772 sqlite3_opentemp_count
++; /* Used for testing and analysis only */
3775 vfsFlags
|= SQLITE_OPEN_READWRITE
| SQLITE_OPEN_CREATE
|
3776 SQLITE_OPEN_EXCLUSIVE
| SQLITE_OPEN_DELETEONCLOSE
;
3777 rc
= sqlite3OsOpen(pPager
->pVfs
, 0, pFile
, vfsFlags
, 0);
3778 assert( rc
!=SQLITE_OK
|| isOpen(pFile
) );
3783 ** Set the busy handler function.
3785 ** The pager invokes the busy-handler if sqlite3OsLock() returns
3786 ** SQLITE_BUSY when trying to upgrade from no-lock to a SHARED lock,
3787 ** or when trying to upgrade from a RESERVED lock to an EXCLUSIVE
3788 ** lock. It does *not* invoke the busy handler when upgrading from
3789 ** SHARED to RESERVED, or when upgrading from SHARED to EXCLUSIVE
3790 ** (which occurs during hot-journal rollback). Summary:
3792 ** Transition | Invokes xBusyHandler
3793 ** --------------------------------------------------------
3794 ** NO_LOCK -> SHARED_LOCK | Yes
3795 ** SHARED_LOCK -> RESERVED_LOCK | No
3796 ** SHARED_LOCK -> EXCLUSIVE_LOCK | No
3797 ** RESERVED_LOCK -> EXCLUSIVE_LOCK | Yes
3799 ** If the busy-handler callback returns non-zero, the lock is
3800 ** retried. If it returns zero, then the SQLITE_BUSY error is
3801 ** returned to the caller of the pager API function.
3803 void sqlite3PagerSetBusyHandler(
3804 Pager
*pPager
, /* Pager object */
3805 int (*xBusyHandler
)(void *), /* Pointer to busy-handler function */
3806 void *pBusyHandlerArg
/* Argument to pass to xBusyHandler */
3809 pPager
->xBusyHandler
= xBusyHandler
;
3810 pPager
->pBusyHandlerArg
= pBusyHandlerArg
;
3811 ap
= (void **)&pPager
->xBusyHandler
;
3812 assert( ((int(*)(void *))(ap
[0]))==xBusyHandler
);
3813 assert( ap
[1]==pBusyHandlerArg
);
3814 sqlite3OsFileControlHint(pPager
->fd
, SQLITE_FCNTL_BUSYHANDLER
, (void *)ap
);
3818 ** Change the page size used by the Pager object. The new page size
3819 ** is passed in *pPageSize.
3821 ** If the pager is in the error state when this function is called, it
3822 ** is a no-op. The value returned is the error state error code (i.e.
3823 ** one of SQLITE_IOERR, an SQLITE_IOERR_xxx sub-code or SQLITE_FULL).
3825 ** Otherwise, if all of the following are true:
3827 ** * the new page size (value of *pPageSize) is valid (a power
3828 ** of two between 512 and SQLITE_MAX_PAGE_SIZE, inclusive), and
3830 ** * there are no outstanding page references, and
3832 ** * the database is either not an in-memory database or it is
3833 ** an in-memory database that currently consists of zero pages.
3835 ** then the pager object page size is set to *pPageSize.
3837 ** If the page size is changed, then this function uses sqlite3PagerMalloc()
3838 ** to obtain a new Pager.pTmpSpace buffer. If this allocation attempt
3839 ** fails, SQLITE_NOMEM is returned and the page size remains unchanged.
3840 ** In all other cases, SQLITE_OK is returned.
3842 ** If the page size is not changed, either because one of the enumerated
3843 ** conditions above is not true, the pager was in error state when this
3844 ** function was called, or because the memory allocation attempt failed,
3845 ** then *pPageSize is set to the old, retained page size before returning.
3847 int sqlite3PagerSetPagesize(Pager
*pPager
, u32
*pPageSize
, int nReserve
){
3850 /* It is not possible to do a full assert_pager_state() here, as this
3851 ** function may be called from within PagerOpen(), before the state
3852 ** of the Pager object is internally consistent.
3854 ** At one point this function returned an error if the pager was in
3855 ** PAGER_ERROR state. But since PAGER_ERROR state guarantees that
3856 ** there is at least one outstanding page reference, this function
3857 ** is a no-op for that case anyhow.
3860 u32 pageSize
= *pPageSize
;
3861 assert( pageSize
==0 || (pageSize
>=512 && pageSize
<=SQLITE_MAX_PAGE_SIZE
) );
3862 if( (pPager
->memDb
==0 || pPager
->dbSize
==0)
3863 && sqlite3PcacheRefCount(pPager
->pPCache
)==0
3864 && pageSize
&& pageSize
!=(u32
)pPager
->pageSize
3866 char *pNew
= NULL
; /* New temp space */
3869 if( pPager
->eState
>PAGER_OPEN
&& isOpen(pPager
->fd
) ){
3870 rc
= sqlite3OsFileSize(pPager
->fd
, &nByte
);
3872 if( rc
==SQLITE_OK
){
3873 /* 8 bytes of zeroed overrun space is sufficient so that the b-tree
3874 * cell header parser will never run off the end of the allocation */
3875 pNew
= (char *)sqlite3PageMalloc(pageSize
+8);
3877 rc
= SQLITE_NOMEM_BKPT
;
3879 memset(pNew
+pageSize
, 0, 8);
3883 if( rc
==SQLITE_OK
){
3884 pager_reset(pPager
);
3885 rc
= sqlite3PcacheSetPageSize(pPager
->pPCache
, pageSize
);
3887 if( rc
==SQLITE_OK
){
3888 sqlite3PageFree(pPager
->pTmpSpace
);
3889 pPager
->pTmpSpace
= pNew
;
3890 pPager
->dbSize
= (Pgno
)((nByte
+pageSize
-1)/pageSize
);
3891 pPager
->pageSize
= pageSize
;
3892 pPager
->lckPgno
= (Pgno
)(PENDING_BYTE
/pageSize
) + 1;
3894 sqlite3PageFree(pNew
);
3898 *pPageSize
= pPager
->pageSize
;
3899 if( rc
==SQLITE_OK
){
3900 if( nReserve
<0 ) nReserve
= pPager
->nReserve
;
3901 assert( nReserve
>=0 && nReserve
<1000 );
3902 pPager
->nReserve
= (i16
)nReserve
;
3903 pagerReportSize(pPager
);
3904 pagerFixMaplimit(pPager
);
3910 ** Return a pointer to the "temporary page" buffer held internally
3911 ** by the pager. This is a buffer that is big enough to hold the
3912 ** entire content of a database page. This buffer is used internally
3913 ** during rollback and will be overwritten whenever a rollback
3914 ** occurs. But other modules are free to use it too, as long as
3915 ** no rollbacks are happening.
3917 void *sqlite3PagerTempSpace(Pager
*pPager
){
3918 return pPager
->pTmpSpace
;
3922 ** Attempt to set the maximum database page count if mxPage is positive.
3923 ** Make no changes if mxPage is zero or negative. And never reduce the
3924 ** maximum page count below the current size of the database.
3926 ** Regardless of mxPage, return the current maximum page count.
3928 Pgno
sqlite3PagerMaxPageCount(Pager
*pPager
, Pgno mxPage
){
3930 pPager
->mxPgno
= mxPage
;
3932 assert( pPager
->eState
!=PAGER_OPEN
); /* Called only by OP_MaxPgcnt */
3933 /* assert( pPager->mxPgno>=pPager->dbSize ); */
3934 /* OP_MaxPgcnt ensures that the parameter passed to this function is not
3935 ** less than the total number of valid pages in the database. But this
3936 ** may be less than Pager.dbSize, and so the assert() above is not valid */
3937 return pPager
->mxPgno
;
3941 ** The following set of routines are used to disable the simulated
3942 ** I/O error mechanism. These routines are used to avoid simulated
3943 ** errors in places where we do not care about errors.
3945 ** Unless -DSQLITE_TEST=1 is used, these routines are all no-ops
3946 ** and generate no code.
3949 extern int sqlite3_io_error_pending
;
3950 extern int sqlite3_io_error_hit
;
3951 static int saved_cnt
;
3952 void disable_simulated_io_errors(void){
3953 saved_cnt
= sqlite3_io_error_pending
;
3954 sqlite3_io_error_pending
= -1;
3956 void enable_simulated_io_errors(void){
3957 sqlite3_io_error_pending
= saved_cnt
;
3960 # define disable_simulated_io_errors()
3961 # define enable_simulated_io_errors()
3965 ** Read the first N bytes from the beginning of the file into memory
3966 ** that pDest points to.
3968 ** If the pager was opened on a transient file (zFilename==""), or
3969 ** opened on a file less than N bytes in size, the output buffer is
3970 ** zeroed and SQLITE_OK returned. The rationale for this is that this
3971 ** function is used to read database headers, and a new transient or
3972 ** zero sized database has a header than consists entirely of zeroes.
3974 ** If any IO error apart from SQLITE_IOERR_SHORT_READ is encountered,
3975 ** the error code is returned to the caller and the contents of the
3976 ** output buffer undefined.
3978 int sqlite3PagerReadFileheader(Pager
*pPager
, int N
, unsigned char *pDest
){
3980 memset(pDest
, 0, N
);
3981 assert( isOpen(pPager
->fd
) || pPager
->tempFile
);
3983 /* This routine is only called by btree immediately after creating
3984 ** the Pager object. There has not been an opportunity to transition
3987 assert( !pagerUseWal(pPager
) );
3989 if( isOpen(pPager
->fd
) ){
3990 IOTRACE(("DBHDR %p 0 %d\n", pPager
, N
))
3991 rc
= sqlite3OsRead(pPager
->fd
, pDest
, N
, 0);
3992 if( rc
==SQLITE_IOERR_SHORT_READ
){
4000 ** This function may only be called when a read-transaction is open on
4001 ** the pager. It returns the total number of pages in the database.
4003 ** However, if the file is between 1 and <page-size> bytes in size, then
4004 ** this is considered a 1 page file.
4006 void sqlite3PagerPagecount(Pager
*pPager
, int *pnPage
){
4007 assert( pPager
->eState
>=PAGER_READER
);
4008 assert( pPager
->eState
!=PAGER_WRITER_FINISHED
);
4009 *pnPage
= (int)pPager
->dbSize
;
4014 ** Try to obtain a lock of type locktype on the database file. If
4015 ** a similar or greater lock is already held, this function is a no-op
4016 ** (returning SQLITE_OK immediately).
4018 ** Otherwise, attempt to obtain the lock using sqlite3OsLock(). Invoke
4019 ** the busy callback if the lock is currently not available. Repeat
4020 ** until the busy callback returns false or until the attempt to
4021 ** obtain the lock succeeds.
4023 ** Return SQLITE_OK on success and an error code if we cannot obtain
4024 ** the lock. If the lock is obtained successfully, set the Pager.state
4025 ** variable to locktype before returning.
4027 static int pager_wait_on_lock(Pager
*pPager
, int locktype
){
4028 int rc
; /* Return code */
4030 /* Check that this is either a no-op (because the requested lock is
4031 ** already held), or one of the transitions that the busy-handler
4032 ** may be invoked during, according to the comment above
4033 ** sqlite3PagerSetBusyhandler().
4035 assert( (pPager
->eLock
>=locktype
)
4036 || (pPager
->eLock
==NO_LOCK
&& locktype
==SHARED_LOCK
)
4037 || (pPager
->eLock
==RESERVED_LOCK
&& locktype
==EXCLUSIVE_LOCK
)
4041 rc
= pagerLockDb(pPager
, locktype
);
4042 }while( rc
==SQLITE_BUSY
&& pPager
->xBusyHandler(pPager
->pBusyHandlerArg
) );
4047 ** Function assertTruncateConstraint(pPager) checks that one of the
4048 ** following is true for all dirty pages currently in the page-cache:
4050 ** a) The page number is less than or equal to the size of the
4051 ** current database image, in pages, OR
4053 ** b) if the page content were written at this time, it would not
4054 ** be necessary to write the current content out to the sub-journal.
4056 ** If the condition asserted by this function were not true, and the
4057 ** dirty page were to be discarded from the cache via the pagerStress()
4058 ** routine, pagerStress() would not write the current page content to
4059 ** the database file. If a savepoint transaction were rolled back after
4060 ** this happened, the correct behavior would be to restore the current
4061 ** content of the page. However, since this content is not present in either
4062 ** the database file or the portion of the rollback journal and
4063 ** sub-journal rolled back the content could not be restored and the
4064 ** database image would become corrupt. It is therefore fortunate that
4065 ** this circumstance cannot arise.
4067 #if defined(SQLITE_DEBUG)
4068 static void assertTruncateConstraintCb(PgHdr
*pPg
){
4069 Pager
*pPager
= pPg
->pPager
;
4070 assert( pPg
->flags
&PGHDR_DIRTY
);
4071 if( pPg
->pgno
>pPager
->dbSize
){ /* if (a) is false */
4072 Pgno pgno
= pPg
->pgno
;
4074 for(i
=0; i
<pPg
->pPager
->nSavepoint
; i
++){
4075 PagerSavepoint
*p
= &pPager
->aSavepoint
[i
];
4076 assert( p
->nOrig
<pgno
|| sqlite3BitvecTestNotNull(p
->pInSavepoint
,pgno
) );
4080 static void assertTruncateConstraint(Pager
*pPager
){
4081 sqlite3PcacheIterateDirty(pPager
->pPCache
, assertTruncateConstraintCb
);
4084 # define assertTruncateConstraint(pPager)
4088 ** Truncate the in-memory database file image to nPage pages. This
4089 ** function does not actually modify the database file on disk. It
4090 ** just sets the internal state of the pager object so that the
4091 ** truncation will be done when the current transaction is committed.
4093 ** This function is only called right before committing a transaction.
4094 ** Once this function has been called, the transaction must either be
4095 ** rolled back or committed. It is not safe to call this function and
4096 ** then continue writing to the database.
4098 void sqlite3PagerTruncateImage(Pager
*pPager
, Pgno nPage
){
4099 assert( pPager
->dbSize
>=nPage
|| CORRUPT_DB
);
4100 assert( pPager
->eState
>=PAGER_WRITER_CACHEMOD
);
4101 pPager
->dbSize
= nPage
;
4103 /* At one point the code here called assertTruncateConstraint() to
4104 ** ensure that all pages being truncated away by this operation are,
4105 ** if one or more savepoints are open, present in the savepoint
4106 ** journal so that they can be restored if the savepoint is rolled
4107 ** back. This is no longer necessary as this function is now only
4108 ** called right before committing a transaction. So although the
4109 ** Pager object may still have open savepoints (Pager.nSavepoint!=0),
4110 ** they cannot be rolled back. So the assertTruncateConstraint() call
4111 ** is no longer correct. */
4116 ** This function is called before attempting a hot-journal rollback. It
4117 ** syncs the journal file to disk, then sets pPager->journalHdr to the
4118 ** size of the journal file so that the pager_playback() routine knows
4119 ** that the entire journal file has been synced.
4121 ** Syncing a hot-journal to disk before attempting to roll it back ensures
4122 ** that if a power-failure occurs during the rollback, the process that
4123 ** attempts rollback following system recovery sees the same journal
4124 ** content as this process.
4126 ** If everything goes as planned, SQLITE_OK is returned. Otherwise,
4127 ** an SQLite error code.
4129 static int pagerSyncHotJournal(Pager
*pPager
){
4131 if( !pPager
->noSync
){
4132 rc
= sqlite3OsSync(pPager
->jfd
, SQLITE_SYNC_NORMAL
);
4134 if( rc
==SQLITE_OK
){
4135 rc
= sqlite3OsFileSize(pPager
->jfd
, &pPager
->journalHdr
);
4140 #if SQLITE_MAX_MMAP_SIZE>0
4142 ** Obtain a reference to a memory mapped page object for page number pgno.
4143 ** The new object will use the pointer pData, obtained from xFetch().
4144 ** If successful, set *ppPage to point to the new page reference
4145 ** and return SQLITE_OK. Otherwise, return an SQLite error code and set
4148 ** Page references obtained by calling this function should be released
4149 ** by calling pagerReleaseMapPage().
4151 static int pagerAcquireMapPage(
4152 Pager
*pPager
, /* Pager object */
4153 Pgno pgno
, /* Page number */
4154 void *pData
, /* xFetch()'d data for this page */
4155 PgHdr
**ppPage
/* OUT: Acquired page object */
4157 PgHdr
*p
; /* Memory mapped page to return */
4159 if( pPager
->pMmapFreelist
){
4160 *ppPage
= p
= pPager
->pMmapFreelist
;
4161 pPager
->pMmapFreelist
= p
->pDirty
;
4163 assert( pPager
->nExtra
>=8 );
4164 memset(p
->pExtra
, 0, 8);
4166 *ppPage
= p
= (PgHdr
*)sqlite3MallocZero(sizeof(PgHdr
) + pPager
->nExtra
);
4168 sqlite3OsUnfetch(pPager
->fd
, (i64
)(pgno
-1) * pPager
->pageSize
, pData
);
4169 return SQLITE_NOMEM_BKPT
;
4171 p
->pExtra
= (void *)&p
[1];
4172 p
->flags
= PGHDR_MMAP
;
4177 assert( p
->pExtra
==(void *)&p
[1] );
4178 assert( p
->pPage
==0 );
4179 assert( p
->flags
==PGHDR_MMAP
);
4180 assert( p
->pPager
==pPager
);
4181 assert( p
->nRef
==1 );
4192 ** Release a reference to page pPg. pPg must have been returned by an
4193 ** earlier call to pagerAcquireMapPage().
4195 static void pagerReleaseMapPage(PgHdr
*pPg
){
4196 Pager
*pPager
= pPg
->pPager
;
4198 pPg
->pDirty
= pPager
->pMmapFreelist
;
4199 pPager
->pMmapFreelist
= pPg
;
4201 assert( pPager
->fd
->pMethods
->iVersion
>=3 );
4202 sqlite3OsUnfetch(pPager
->fd
, (i64
)(pPg
->pgno
-1)*pPager
->pageSize
, pPg
->pData
);
4206 ** Free all PgHdr objects stored in the Pager.pMmapFreelist list.
4208 static void pagerFreeMapHdrs(Pager
*pPager
){
4211 for(p
=pPager
->pMmapFreelist
; p
; p
=pNext
){
4217 /* Verify that the database file has not be deleted or renamed out from
4218 ** under the pager. Return SQLITE_OK if the database is still where it ought
4219 ** to be on disk. Return non-zero (SQLITE_READONLY_DBMOVED or some other error
4220 ** code from sqlite3OsAccess()) if the database has gone missing.
4222 static int databaseIsUnmoved(Pager
*pPager
){
4226 if( pPager
->tempFile
) return SQLITE_OK
;
4227 if( pPager
->dbSize
==0 ) return SQLITE_OK
;
4228 assert( pPager
->zFilename
&& pPager
->zFilename
[0] );
4229 rc
= sqlite3OsFileControl(pPager
->fd
, SQLITE_FCNTL_HAS_MOVED
, &bHasMoved
);
4230 if( rc
==SQLITE_NOTFOUND
){
4231 /* If the HAS_MOVED file-control is unimplemented, assume that the file
4232 ** has not been moved. That is the historical behavior of SQLite: prior to
4233 ** version 3.8.3, it never checked */
4235 }else if( rc
==SQLITE_OK
&& bHasMoved
){
4236 rc
= SQLITE_READONLY_DBMOVED
;
4243 ** Shutdown the page cache. Free all memory and close all files.
4245 ** If a transaction was in progress when this routine is called, that
4246 ** transaction is rolled back. All outstanding pages are invalidated
4247 ** and their memory is freed. Any attempt to use a page associated
4248 ** with this page cache after this function returns will likely
4249 ** result in a coredump.
4251 ** This function always succeeds. If a transaction is active an attempt
4252 ** is made to roll it back. If an error occurs during the rollback
4253 ** a hot journal may be left in the filesystem but no error is returned
4256 int sqlite3PagerClose(Pager
*pPager
, sqlite3
*db
){
4257 u8
*pTmp
= (u8
*)pPager
->pTmpSpace
;
4258 assert( db
|| pagerUseWal(pPager
)==0 );
4259 assert( assert_pager_state(pPager
) );
4260 disable_simulated_io_errors();
4261 sqlite3BeginBenignMalloc();
4262 pagerFreeMapHdrs(pPager
);
4263 /* pPager->errCode = 0; */
4264 pPager
->exclusiveMode
= 0;
4265 #ifndef SQLITE_OMIT_WAL
4268 assert( db
|| pPager
->pWal
==0 );
4269 if( db
&& 0==(db
->flags
& SQLITE_NoCkptOnClose
)
4270 && SQLITE_OK
==databaseIsUnmoved(pPager
)
4274 sqlite3WalClose(pPager
->pWal
, db
, pPager
->walSyncFlags
, pPager
->pageSize
,a
);
4278 pager_reset(pPager
);
4280 pager_unlock(pPager
);
4282 /* If it is open, sync the journal file before calling UnlockAndRollback.
4283 ** If this is not done, then an unsynced portion of the open journal
4284 ** file may be played back into the database. If a power failure occurs
4285 ** while this is happening, the database could become corrupt.
4287 ** If an error occurs while trying to sync the journal, shift the pager
4288 ** into the ERROR state. This causes UnlockAndRollback to unlock the
4289 ** database and close the journal file without attempting to roll it
4290 ** back or finalize it. The next database user will have to do hot-journal
4291 ** rollback before accessing the database file.
4293 if( isOpen(pPager
->jfd
) ){
4294 pager_error(pPager
, pagerSyncHotJournal(pPager
));
4296 pagerUnlockAndRollback(pPager
);
4298 sqlite3EndBenignMalloc();
4299 enable_simulated_io_errors();
4300 PAGERTRACE(("CLOSE %d\n", PAGERID(pPager
)));
4301 IOTRACE(("CLOSE %p\n", pPager
))
4302 sqlite3OsClose(pPager
->jfd
);
4303 sqlite3OsClose(pPager
->fd
);
4304 sqlite3PageFree(pTmp
);
4305 sqlite3PcacheClose(pPager
->pPCache
);
4307 /* BEGIN SQLCIPHER */
4308 #ifdef SQLITE_HAS_CODEC
4309 if( pPager
->xCodecFree
) pPager
->xCodecFree(pPager
->pCodec
);
4313 assert( !pPager
->aSavepoint
&& !pPager
->pInJournal
);
4314 assert( !isOpen(pPager
->jfd
) && !isOpen(pPager
->sjfd
) );
4316 sqlite3_free(pPager
);
4320 #if !defined(NDEBUG) || defined(SQLITE_TEST)
4322 ** Return the page number for page pPg.
4324 Pgno
sqlite3PagerPagenumber(DbPage
*pPg
){
4330 ** Increment the reference count for page pPg.
4332 void sqlite3PagerRef(DbPage
*pPg
){
4333 sqlite3PcacheRef(pPg
);
4337 ** Sync the journal. In other words, make sure all the pages that have
4338 ** been written to the journal have actually reached the surface of the
4339 ** disk and can be restored in the event of a hot-journal rollback.
4341 ** If the Pager.noSync flag is set, then this function is a no-op.
4342 ** Otherwise, the actions required depend on the journal-mode and the
4343 ** device characteristics of the file-system, as follows:
4345 ** * If the journal file is an in-memory journal file, no action need
4348 ** * Otherwise, if the device does not support the SAFE_APPEND property,
4349 ** then the nRec field of the most recently written journal header
4350 ** is updated to contain the number of journal records that have
4351 ** been written following it. If the pager is operating in full-sync
4352 ** mode, then the journal file is synced before this field is updated.
4354 ** * If the device does not support the SEQUENTIAL property, then
4355 ** journal file is synced.
4357 ** Or, in pseudo-code:
4359 ** if( NOT <in-memory journal> ){
4360 ** if( NOT SAFE_APPEND ){
4361 ** if( <full-sync mode> ) xSync(<journal file>);
4362 ** <update nRec field>
4364 ** if( NOT SEQUENTIAL ) xSync(<journal file>);
4367 ** If successful, this routine clears the PGHDR_NEED_SYNC flag of every
4368 ** page currently held in memory before returning SQLITE_OK. If an IO
4369 ** error is encountered, then the IO error code is returned to the caller.
4371 static int syncJournal(Pager
*pPager
, int newHdr
){
4372 int rc
; /* Return code */
4374 assert( pPager
->eState
==PAGER_WRITER_CACHEMOD
4375 || pPager
->eState
==PAGER_WRITER_DBMOD
4377 assert( assert_pager_state(pPager
) );
4378 assert( !pagerUseWal(pPager
) );
4380 rc
= sqlite3PagerExclusiveLock(pPager
);
4381 if( rc
!=SQLITE_OK
) return rc
;
4383 if( !pPager
->noSync
){
4384 assert( !pPager
->tempFile
);
4385 if( isOpen(pPager
->jfd
) && pPager
->journalMode
!=PAGER_JOURNALMODE_MEMORY
){
4386 const int iDc
= sqlite3OsDeviceCharacteristics(pPager
->fd
);
4387 assert( isOpen(pPager
->jfd
) );
4389 if( 0==(iDc
&SQLITE_IOCAP_SAFE_APPEND
) ){
4390 /* This block deals with an obscure problem. If the last connection
4391 ** that wrote to this database was operating in persistent-journal
4392 ** mode, then the journal file may at this point actually be larger
4393 ** than Pager.journalOff bytes. If the next thing in the journal
4394 ** file happens to be a journal-header (written as part of the
4395 ** previous connection's transaction), and a crash or power-failure
4396 ** occurs after nRec is updated but before this connection writes
4397 ** anything else to the journal file (or commits/rolls back its
4398 ** transaction), then SQLite may become confused when doing the
4399 ** hot-journal rollback following recovery. It may roll back all
4400 ** of this connections data, then proceed to rolling back the old,
4401 ** out-of-date data that follows it. Database corruption.
4403 ** To work around this, if the journal file does appear to contain
4404 ** a valid header following Pager.journalOff, then write a 0x00
4405 ** byte to the start of it to prevent it from being recognized.
4407 ** Variable iNextHdrOffset is set to the offset at which this
4408 ** problematic header will occur, if it exists. aMagic is used
4409 ** as a temporary buffer to inspect the first couple of bytes of
4410 ** the potential journal header.
4414 u8 zHeader
[sizeof(aJournalMagic
)+4];
4416 memcpy(zHeader
, aJournalMagic
, sizeof(aJournalMagic
));
4417 put32bits(&zHeader
[sizeof(aJournalMagic
)], pPager
->nRec
);
4419 iNextHdrOffset
= journalHdrOffset(pPager
);
4420 rc
= sqlite3OsRead(pPager
->jfd
, aMagic
, 8, iNextHdrOffset
);
4421 if( rc
==SQLITE_OK
&& 0==memcmp(aMagic
, aJournalMagic
, 8) ){
4422 static const u8 zerobyte
= 0;
4423 rc
= sqlite3OsWrite(pPager
->jfd
, &zerobyte
, 1, iNextHdrOffset
);
4425 if( rc
!=SQLITE_OK
&& rc
!=SQLITE_IOERR_SHORT_READ
){
4429 /* Write the nRec value into the journal file header. If in
4430 ** full-synchronous mode, sync the journal first. This ensures that
4431 ** all data has really hit the disk before nRec is updated to mark
4432 ** it as a candidate for rollback.
4434 ** This is not required if the persistent media supports the
4435 ** SAFE_APPEND property. Because in this case it is not possible
4436 ** for garbage data to be appended to the file, the nRec field
4437 ** is populated with 0xFFFFFFFF when the journal header is written
4438 ** and never needs to be updated.
4440 if( pPager
->fullSync
&& 0==(iDc
&SQLITE_IOCAP_SEQUENTIAL
) ){
4441 PAGERTRACE(("SYNC journal of %d\n", PAGERID(pPager
)));
4442 IOTRACE(("JSYNC %p\n", pPager
))
4443 rc
= sqlite3OsSync(pPager
->jfd
, pPager
->syncFlags
);
4444 if( rc
!=SQLITE_OK
) return rc
;
4446 IOTRACE(("JHDR %p %lld\n", pPager
, pPager
->journalHdr
));
4447 rc
= sqlite3OsWrite(
4448 pPager
->jfd
, zHeader
, sizeof(zHeader
), pPager
->journalHdr
4450 if( rc
!=SQLITE_OK
) return rc
;
4452 if( 0==(iDc
&SQLITE_IOCAP_SEQUENTIAL
) ){
4453 PAGERTRACE(("SYNC journal of %d\n", PAGERID(pPager
)));
4454 IOTRACE(("JSYNC %p\n", pPager
))
4455 rc
= sqlite3OsSync(pPager
->jfd
, pPager
->syncFlags
|
4456 (pPager
->syncFlags
==SQLITE_SYNC_FULL
?SQLITE_SYNC_DATAONLY
:0)
4458 if( rc
!=SQLITE_OK
) return rc
;
4461 pPager
->journalHdr
= pPager
->journalOff
;
4462 if( newHdr
&& 0==(iDc
&SQLITE_IOCAP_SAFE_APPEND
) ){
4464 rc
= writeJournalHdr(pPager
);
4465 if( rc
!=SQLITE_OK
) return rc
;
4468 pPager
->journalHdr
= pPager
->journalOff
;
4472 /* Unless the pager is in noSync mode, the journal file was just
4473 ** successfully synced. Either way, clear the PGHDR_NEED_SYNC flag on
4476 sqlite3PcacheClearSyncFlags(pPager
->pPCache
);
4477 pPager
->eState
= PAGER_WRITER_DBMOD
;
4478 assert( assert_pager_state(pPager
) );
4483 ** The argument is the first in a linked list of dirty pages connected
4484 ** by the PgHdr.pDirty pointer. This function writes each one of the
4485 ** in-memory pages in the list to the database file. The argument may
4486 ** be NULL, representing an empty list. In this case this function is
4489 ** The pager must hold at least a RESERVED lock when this function
4490 ** is called. Before writing anything to the database file, this lock
4491 ** is upgraded to an EXCLUSIVE lock. If the lock cannot be obtained,
4492 ** SQLITE_BUSY is returned and no data is written to the database file.
4494 ** If the pager is a temp-file pager and the actual file-system file
4495 ** is not yet open, it is created and opened before any data is
4498 ** Once the lock has been upgraded and, if necessary, the file opened,
4499 ** the pages are written out to the database file in list order. Writing
4500 ** a page is skipped if it meets either of the following criteria:
4502 ** * The page number is greater than Pager.dbSize, or
4503 ** * The PGHDR_DONT_WRITE flag is set on the page.
4505 ** If writing out a page causes the database file to grow, Pager.dbFileSize
4506 ** is updated accordingly. If page 1 is written out, then the value cached
4507 ** in Pager.dbFileVers[] is updated to match the new value stored in
4508 ** the database file.
4510 ** If everything is successful, SQLITE_OK is returned. If an IO error
4511 ** occurs, an IO error code is returned. Or, if the EXCLUSIVE lock cannot
4512 ** be obtained, SQLITE_BUSY is returned.
4514 static int pager_write_pagelist(Pager
*pPager
, PgHdr
*pList
){
4515 int rc
= SQLITE_OK
; /* Return code */
4517 /* This function is only called for rollback pagers in WRITER_DBMOD state. */
4518 assert( !pagerUseWal(pPager
) );
4519 assert( pPager
->tempFile
|| pPager
->eState
==PAGER_WRITER_DBMOD
);
4520 assert( pPager
->eLock
==EXCLUSIVE_LOCK
);
4521 assert( isOpen(pPager
->fd
) || pList
->pDirty
==0 );
4523 /* If the file is a temp-file has not yet been opened, open it now. It
4524 ** is not possible for rc to be other than SQLITE_OK if this branch
4525 ** is taken, as pager_wait_on_lock() is a no-op for temp-files.
4527 if( !isOpen(pPager
->fd
) ){
4528 assert( pPager
->tempFile
&& rc
==SQLITE_OK
);
4529 rc
= pagerOpentemp(pPager
, pPager
->fd
, pPager
->vfsFlags
);
4532 /* Before the first write, give the VFS a hint of what the final
4533 ** file size will be.
4535 assert( rc
!=SQLITE_OK
|| isOpen(pPager
->fd
) );
4537 && pPager
->dbHintSize
<pPager
->dbSize
4538 && (pList
->pDirty
|| pList
->pgno
>pPager
->dbHintSize
)
4540 sqlite3_int64 szFile
= pPager
->pageSize
* (sqlite3_int64
)pPager
->dbSize
;
4541 sqlite3OsFileControlHint(pPager
->fd
, SQLITE_FCNTL_SIZE_HINT
, &szFile
);
4542 pPager
->dbHintSize
= pPager
->dbSize
;
4545 while( rc
==SQLITE_OK
&& pList
){
4546 Pgno pgno
= pList
->pgno
;
4548 /* If there are dirty pages in the page cache with page numbers greater
4549 ** than Pager.dbSize, this means sqlite3PagerTruncateImage() was called to
4550 ** make the file smaller (presumably by auto-vacuum code). Do not write
4551 ** any such pages to the file.
4553 ** Also, do not write out any page that has the PGHDR_DONT_WRITE flag
4554 ** set (set by sqlite3PagerDontWrite()).
4556 if( pgno
<=pPager
->dbSize
&& 0==(pList
->flags
&PGHDR_DONT_WRITE
) ){
4557 i64 offset
= (pgno
-1)*(i64
)pPager
->pageSize
; /* Offset to write */
4558 char *pData
; /* Data to write */
4560 assert( (pList
->flags
&PGHDR_NEED_SYNC
)==0 );
4561 if( pList
->pgno
==1 ) pager_write_changecounter(pList
);
4563 /* Encode the database */
4564 CODEC2(pPager
, pList
->pData
, pgno
, 6, return SQLITE_NOMEM_BKPT
, pData
);
4566 /* Write out the page data. */
4567 rc
= sqlite3OsWrite(pPager
->fd
, pData
, pPager
->pageSize
, offset
);
4569 /* If page 1 was just written, update Pager.dbFileVers to match
4570 ** the value now stored in the database file. If writing this
4571 ** page caused the database file to grow, update dbFileSize.
4574 memcpy(&pPager
->dbFileVers
, &pData
[24], sizeof(pPager
->dbFileVers
));
4576 if( pgno
>pPager
->dbFileSize
){
4577 pPager
->dbFileSize
= pgno
;
4579 pPager
->aStat
[PAGER_STAT_WRITE
]++;
4581 /* Update any backup objects copying the contents of this pager. */
4582 sqlite3BackupUpdate(pPager
->pBackup
, pgno
, (u8
*)pList
->pData
);
4584 PAGERTRACE(("STORE %d page %d hash(%08x)\n",
4585 PAGERID(pPager
), pgno
, pager_pagehash(pList
)));
4586 IOTRACE(("PGOUT %p %d\n", pPager
, pgno
));
4587 PAGER_INCR(sqlite3_pager_writedb_count
);
4589 PAGERTRACE(("NOSTORE %d page %d\n", PAGERID(pPager
), pgno
));
4591 pager_set_pagehash(pList
);
4592 pList
= pList
->pDirty
;
4599 ** Ensure that the sub-journal file is open. If it is already open, this
4600 ** function is a no-op.
4602 ** SQLITE_OK is returned if everything goes according to plan. An
4603 ** SQLITE_IOERR_XXX error code is returned if a call to sqlite3OsOpen()
4606 static int openSubJournal(Pager
*pPager
){
4608 if( !isOpen(pPager
->sjfd
) ){
4609 const int flags
= SQLITE_OPEN_SUBJOURNAL
| SQLITE_OPEN_READWRITE
4610 | SQLITE_OPEN_CREATE
| SQLITE_OPEN_EXCLUSIVE
4611 | SQLITE_OPEN_DELETEONCLOSE
;
4612 int nStmtSpill
= sqlite3Config
.nStmtSpill
;
4613 if( pPager
->journalMode
==PAGER_JOURNALMODE_MEMORY
|| pPager
->subjInMemory
){
4616 rc
= sqlite3JournalOpen(pPager
->pVfs
, 0, pPager
->sjfd
, flags
, nStmtSpill
);
4622 ** Append a record of the current state of page pPg to the sub-journal.
4624 ** If successful, set the bit corresponding to pPg->pgno in the bitvecs
4625 ** for all open savepoints before returning.
4627 ** This function returns SQLITE_OK if everything is successful, an IO
4628 ** error code if the attempt to write to the sub-journal fails, or
4629 ** SQLITE_NOMEM if a malloc fails while setting a bit in a savepoint
4632 static int subjournalPage(PgHdr
*pPg
){
4634 Pager
*pPager
= pPg
->pPager
;
4635 if( pPager
->journalMode
!=PAGER_JOURNALMODE_OFF
){
4637 /* Open the sub-journal, if it has not already been opened */
4638 assert( pPager
->useJournal
);
4639 assert( isOpen(pPager
->jfd
) || pagerUseWal(pPager
) );
4640 assert( isOpen(pPager
->sjfd
) || pPager
->nSubRec
==0 );
4641 assert( pagerUseWal(pPager
)
4642 || pageInJournal(pPager
, pPg
)
4643 || pPg
->pgno
>pPager
->dbOrigSize
4645 rc
= openSubJournal(pPager
);
4647 /* If the sub-journal was opened successfully (or was already open),
4648 ** write the journal record into the file. */
4649 if( rc
==SQLITE_OK
){
4650 void *pData
= pPg
->pData
;
4651 i64 offset
= (i64
)pPager
->nSubRec
*(4+pPager
->pageSize
);
4654 /* BEGIN SQLCIPHER */
4655 #if SQLITE_HAS_CODEC
4656 if( !pPager
->subjInMemory
){
4657 CODEC2(pPager
, pData
, pPg
->pgno
, 7, return SQLITE_NOMEM_BKPT
, pData2
);
4662 PAGERTRACE(("STMT-JOURNAL %d page %d\n", PAGERID(pPager
), pPg
->pgno
));
4663 rc
= write32bits(pPager
->sjfd
, offset
, pPg
->pgno
);
4664 if( rc
==SQLITE_OK
){
4665 rc
= sqlite3OsWrite(pPager
->sjfd
, pData2
, pPager
->pageSize
, offset
+4);
4669 if( rc
==SQLITE_OK
){
4671 assert( pPager
->nSavepoint
>0 );
4672 rc
= addToSavepointBitvecs(pPager
, pPg
->pgno
);
4676 static int subjournalPageIfRequired(PgHdr
*pPg
){
4677 if( subjRequiresPage(pPg
) ){
4678 return subjournalPage(pPg
);
4685 ** This function is called by the pcache layer when it has reached some
4686 ** soft memory limit. The first argument is a pointer to a Pager object
4687 ** (cast as a void*). The pager is always 'purgeable' (not an in-memory
4688 ** database). The second argument is a reference to a page that is
4689 ** currently dirty but has no outstanding references. The page
4690 ** is always associated with the Pager object passed as the first
4693 ** The job of this function is to make pPg clean by writing its contents
4694 ** out to the database file, if possible. This may involve syncing the
4697 ** If successful, sqlite3PcacheMakeClean() is called on the page and
4698 ** SQLITE_OK returned. If an IO error occurs while trying to make the
4699 ** page clean, the IO error code is returned. If the page cannot be
4700 ** made clean for some other reason, but no error occurs, then SQLITE_OK
4701 ** is returned by sqlite3PcacheMakeClean() is not called.
4703 static int pagerStress(void *p
, PgHdr
*pPg
){
4704 Pager
*pPager
= (Pager
*)p
;
4707 assert( pPg
->pPager
==pPager
);
4708 assert( pPg
->flags
&PGHDR_DIRTY
);
4710 /* The doNotSpill NOSYNC bit is set during times when doing a sync of
4711 ** journal (and adding a new header) is not allowed. This occurs
4712 ** during calls to sqlite3PagerWrite() while trying to journal multiple
4713 ** pages belonging to the same sector.
4715 ** The doNotSpill ROLLBACK and OFF bits inhibits all cache spilling
4716 ** regardless of whether or not a sync is required. This is set during
4717 ** a rollback or by user request, respectively.
4719 ** Spilling is also prohibited when in an error state since that could
4720 ** lead to database corruption. In the current implementation it
4721 ** is impossible for sqlite3PcacheFetch() to be called with createFlag==3
4722 ** while in the error state, hence it is impossible for this routine to
4723 ** be called in the error state. Nevertheless, we include a NEVER()
4724 ** test for the error state as a safeguard against future changes.
4726 if( NEVER(pPager
->errCode
) ) return SQLITE_OK
;
4727 testcase( pPager
->doNotSpill
& SPILLFLAG_ROLLBACK
);
4728 testcase( pPager
->doNotSpill
& SPILLFLAG_OFF
);
4729 testcase( pPager
->doNotSpill
& SPILLFLAG_NOSYNC
);
4730 if( pPager
->doNotSpill
4731 && ((pPager
->doNotSpill
& (SPILLFLAG_ROLLBACK
|SPILLFLAG_OFF
))!=0
4732 || (pPg
->flags
& PGHDR_NEED_SYNC
)!=0)
4737 pPager
->aStat
[PAGER_STAT_SPILL
]++;
4739 if( pagerUseWal(pPager
) ){
4740 /* Write a single frame for this page to the log. */
4741 rc
= subjournalPageIfRequired(pPg
);
4742 if( rc
==SQLITE_OK
){
4743 rc
= pagerWalFrames(pPager
, pPg
, 0, 0);
4747 #ifdef SQLITE_ENABLE_BATCH_ATOMIC_WRITE
4748 if( pPager
->tempFile
==0 ){
4749 rc
= sqlite3JournalCreate(pPager
->jfd
);
4750 if( rc
!=SQLITE_OK
) return pager_error(pPager
, rc
);
4754 /* Sync the journal file if required. */
4755 if( pPg
->flags
&PGHDR_NEED_SYNC
4756 || pPager
->eState
==PAGER_WRITER_CACHEMOD
4758 rc
= syncJournal(pPager
, 1);
4761 /* Write the contents of the page out to the database file. */
4762 if( rc
==SQLITE_OK
){
4763 assert( (pPg
->flags
&PGHDR_NEED_SYNC
)==0 );
4764 rc
= pager_write_pagelist(pPager
, pPg
);
4768 /* Mark the page as clean. */
4769 if( rc
==SQLITE_OK
){
4770 PAGERTRACE(("STRESS %d page %d\n", PAGERID(pPager
), pPg
->pgno
));
4771 sqlite3PcacheMakeClean(pPg
);
4774 return pager_error(pPager
, rc
);
4778 ** Flush all unreferenced dirty pages to disk.
4780 int sqlite3PagerFlush(Pager
*pPager
){
4781 int rc
= pPager
->errCode
;
4783 PgHdr
*pList
= sqlite3PcacheDirtyList(pPager
->pPCache
);
4784 assert( assert_pager_state(pPager
) );
4785 while( rc
==SQLITE_OK
&& pList
){
4786 PgHdr
*pNext
= pList
->pDirty
;
4787 if( pList
->nRef
==0 ){
4788 rc
= pagerStress((void*)pPager
, pList
);
4798 ** Allocate and initialize a new Pager object and put a pointer to it
4799 ** in *ppPager. The pager should eventually be freed by passing it
4800 ** to sqlite3PagerClose().
4802 ** The zFilename argument is the path to the database file to open.
4803 ** If zFilename is NULL then a randomly-named temporary file is created
4804 ** and used as the file to be cached. Temporary files are be deleted
4805 ** automatically when they are closed. If zFilename is ":memory:" then
4806 ** all information is held in cache. It is never written to disk.
4807 ** This can be used to implement an in-memory database.
4809 ** The nExtra parameter specifies the number of bytes of space allocated
4810 ** along with each page reference. This space is available to the user
4811 ** via the sqlite3PagerGetExtra() API. When a new page is allocated, the
4812 ** first 8 bytes of this space are zeroed but the remainder is uninitialized.
4813 ** (The extra space is used by btree as the MemPage object.)
4815 ** The flags argument is used to specify properties that affect the
4816 ** operation of the pager. It should be passed some bitwise combination
4817 ** of the PAGER_* flags.
4819 ** The vfsFlags parameter is a bitmask to pass to the flags parameter
4820 ** of the xOpen() method of the supplied VFS when opening files.
4822 ** If the pager object is allocated and the specified file opened
4823 ** successfully, SQLITE_OK is returned and *ppPager set to point to
4824 ** the new pager object. If an error occurs, *ppPager is set to NULL
4825 ** and error code returned. This function may return SQLITE_NOMEM
4826 ** (sqlite3Malloc() is used to allocate memory), SQLITE_CANTOPEN or
4827 ** various SQLITE_IO_XXX errors.
4829 int sqlite3PagerOpen(
4830 sqlite3_vfs
*pVfs
, /* The virtual file system to use */
4831 Pager
**ppPager
, /* OUT: Return the Pager structure here */
4832 const char *zFilename
, /* Name of the database file to open */
4833 int nExtra
, /* Extra bytes append to each in-memory page */
4834 int flags
, /* flags controlling this file */
4835 int vfsFlags
, /* flags passed through to sqlite3_vfs.xOpen() */
4836 void (*xReinit
)(DbPage
*) /* Function to reinitialize pages */
4839 Pager
*pPager
= 0; /* Pager object to allocate and return */
4840 int rc
= SQLITE_OK
; /* Return code */
4841 int tempFile
= 0; /* True for temp files (incl. in-memory files) */
4842 int memDb
= 0; /* True if this is an in-memory file */
4843 int memJM
= 0; /* Memory journal mode */
4844 int readOnly
= 0; /* True if this is a read-only file */
4845 int journalFileSize
; /* Bytes to allocate for each journal fd */
4846 char *zPathname
= 0; /* Full path to database file */
4847 int nPathname
= 0; /* Number of bytes in zPathname */
4848 int useJournal
= (flags
& PAGER_OMIT_JOURNAL
)==0; /* False to omit journal */
4849 int pcacheSize
= sqlite3PcacheSize(); /* Bytes to allocate for PCache */
4850 u32 szPageDflt
= SQLITE_DEFAULT_PAGE_SIZE
; /* Default page size */
4851 const char *zUri
= 0; /* URI args to copy */
4852 int nUriByte
= 1; /* Number of bytes of URI args at *zUri */
4854 /* Figure out how much space is required for each journal file-handle
4855 ** (there are two of them, the main journal and the sub-journal). */
4856 journalFileSize
= ROUND8(sqlite3JournalSize(pVfs
));
4858 /* Set the output variable to NULL in case an error occurs. */
4861 #ifndef SQLITE_OMIT_MEMORYDB
4862 if( flags
& PAGER_MEMORY
){
4864 if( zFilename
&& zFilename
[0] ){
4865 zPathname
= sqlite3DbStrDup(0, zFilename
);
4866 if( zPathname
==0 ) return SQLITE_NOMEM_BKPT
;
4867 nPathname
= sqlite3Strlen30(zPathname
);
4873 /* Compute and store the full pathname in an allocated buffer pointed
4874 ** to by zPathname, length nPathname. Or, if this is a temporary file,
4875 ** leave both nPathname and zPathname set to 0.
4877 if( zFilename
&& zFilename
[0] ){
4879 nPathname
= pVfs
->mxPathname
+1;
4880 zPathname
= sqlite3DbMallocRaw(0, nPathname
*2);
4882 return SQLITE_NOMEM_BKPT
;
4884 zPathname
[0] = 0; /* Make sure initialized even if FullPathname() fails */
4885 rc
= sqlite3OsFullPathname(pVfs
, zFilename
, nPathname
, zPathname
);
4886 if( rc
!=SQLITE_OK
){
4887 if( rc
==SQLITE_OK_SYMLINK
){
4888 if( vfsFlags
& SQLITE_OPEN_NOFOLLOW
){
4889 rc
= SQLITE_CANTOPEN_SYMLINK
;
4895 nPathname
= sqlite3Strlen30(zPathname
);
4896 z
= zUri
= &zFilename
[sqlite3Strlen30(zFilename
)+1];
4901 nUriByte
= (int)(&z
[1] - zUri
);
4902 assert( nUriByte
>=1 );
4903 if( rc
==SQLITE_OK
&& nPathname
+8>pVfs
->mxPathname
){
4904 /* This branch is taken when the journal path required by
4905 ** the database being opened will be more than pVfs->mxPathname
4906 ** bytes in length. This means the database cannot be opened,
4907 ** as it will not be possible to open the journal file or even
4908 ** check for a hot-journal before reading.
4910 rc
= SQLITE_CANTOPEN_BKPT
;
4912 if( rc
!=SQLITE_OK
){
4913 sqlite3DbFree(0, zPathname
);
4918 /* Allocate memory for the Pager structure, PCache object, the
4919 ** three file descriptors, the database file name and the journal
4920 ** file name. The layout in memory is as follows:
4922 ** Pager object (sizeof(Pager) bytes)
4923 ** PCache object (sqlite3PcacheSize() bytes)
4924 ** Database file handle (pVfs->szOsFile bytes)
4925 ** Sub-journal file handle (journalFileSize bytes)
4926 ** Main journal file handle (journalFileSize bytes)
4927 ** Ptr back to the Pager (sizeof(Pager*) bytes)
4928 ** \0\0\0\0 database prefix (4 bytes)
4929 ** Database file name (nPathname+1 bytes)
4930 ** URI query parameters (nUriByte bytes)
4931 ** Journal filename (nPathname+8+1 bytes)
4932 ** WAL filename (nPathname+4+1 bytes)
4933 ** \0\0\0 terminator (3 bytes)
4935 ** Some 3rd-party software, over which we have no control, depends on
4936 ** the specific order of the filenames and the \0 separators between them
4937 ** so that it can (for example) find the database filename given the WAL
4938 ** filename without using the sqlite3_filename_database() API. This is a
4939 ** misuse of SQLite and a bug in the 3rd-party software, but the 3rd-party
4940 ** software is in widespread use, so we try to avoid changing the filename
4941 ** order and formatting if possible. In particular, the details of the
4942 ** filename format expected by 3rd-party software should be as follows:
4944 ** - Main Database Path
4946 ** - Multiple URI components consisting of:
4954 ** - WAL Path (zWALName)
4957 ** The sqlite3_create_filename() interface and the databaseFilename() utility
4958 ** that is used by sqlite3_filename_database() and kin also depend on the
4959 ** specific formatting and order of the various filenames, so if the format
4960 ** changes here, be sure to change it there as well.
4962 assert( SQLITE_PTRSIZE
==sizeof(Pager
*) );
4963 pPtr
= (u8
*)sqlite3MallocZero(
4964 ROUND8(sizeof(*pPager
)) + /* Pager structure */
4965 ROUND8(pcacheSize
) + /* PCache object */
4966 ROUND8(pVfs
->szOsFile
) + /* The main db file */
4967 journalFileSize
* 2 + /* The two journal files */
4968 SQLITE_PTRSIZE
+ /* Space to hold a pointer */
4969 4 + /* Database prefix */
4970 nPathname
+ 1 + /* database filename */
4971 nUriByte
+ /* query parameters */
4972 nPathname
+ 8 + 1 + /* Journal filename */
4973 #ifndef SQLITE_OMIT_WAL
4974 nPathname
+ 4 + 1 + /* WAL filename */
4978 assert( EIGHT_BYTE_ALIGNMENT(SQLITE_INT_TO_PTR(journalFileSize
)) );
4980 sqlite3DbFree(0, zPathname
);
4981 return SQLITE_NOMEM_BKPT
;
4983 pPager
= (Pager
*)pPtr
; pPtr
+= ROUND8(sizeof(*pPager
));
4984 pPager
->pPCache
= (PCache
*)pPtr
; pPtr
+= ROUND8(pcacheSize
);
4985 pPager
->fd
= (sqlite3_file
*)pPtr
; pPtr
+= ROUND8(pVfs
->szOsFile
);
4986 pPager
->sjfd
= (sqlite3_file
*)pPtr
; pPtr
+= journalFileSize
;
4987 pPager
->jfd
= (sqlite3_file
*)pPtr
; pPtr
+= journalFileSize
;
4988 assert( EIGHT_BYTE_ALIGNMENT(pPager
->jfd
) );
4989 memcpy(pPtr
, &pPager
, SQLITE_PTRSIZE
); pPtr
+= SQLITE_PTRSIZE
;
4991 /* Fill in the Pager.zFilename and pPager.zQueryParam fields */
4992 pPtr
+= 4; /* Skip zero prefix */
4993 pPager
->zFilename
= (char*)pPtr
;
4995 memcpy(pPtr
, zPathname
, nPathname
); pPtr
+= nPathname
+ 1;
4997 memcpy(pPtr
, zUri
, nUriByte
); pPtr
+= nUriByte
;
5004 /* Fill in Pager.zJournal */
5006 pPager
->zJournal
= (char*)pPtr
;
5007 memcpy(pPtr
, zPathname
, nPathname
); pPtr
+= nPathname
;
5008 memcpy(pPtr
, "-journal",8); pPtr
+= 8 + 1;
5009 #ifdef SQLITE_ENABLE_8_3_NAMES
5010 sqlite3FileSuffix3(zFilename
,pPager
->zJournal
);
5011 pPtr
= (u8
*)(pPager
->zJournal
+ sqlite3Strlen30(pPager
->zJournal
)+1);
5014 pPager
->zJournal
= 0;
5017 #ifndef SQLITE_OMIT_WAL
5018 /* Fill in Pager.zWal */
5020 pPager
->zWal
= (char*)pPtr
;
5021 memcpy(pPtr
, zPathname
, nPathname
); pPtr
+= nPathname
;
5022 memcpy(pPtr
, "-wal", 4); pPtr
+= 4 + 1;
5023 #ifdef SQLITE_ENABLE_8_3_NAMES
5024 sqlite3FileSuffix3(zFilename
, pPager
->zWal
);
5025 pPtr
= (u8
*)(pPager
->zWal
+ sqlite3Strlen30(pPager
->zWal
)+1);
5031 (void)pPtr
; /* Suppress warning about unused pPtr value */
5033 if( nPathname
) sqlite3DbFree(0, zPathname
);
5034 pPager
->pVfs
= pVfs
;
5035 pPager
->vfsFlags
= vfsFlags
;
5037 /* Open the pager file.
5039 if( zFilename
&& zFilename
[0] ){
5040 int fout
= 0; /* VFS flags returned by xOpen() */
5041 rc
= sqlite3OsOpen(pVfs
, pPager
->zFilename
, pPager
->fd
, vfsFlags
, &fout
);
5043 pPager
->memVfs
= memJM
= (fout
&SQLITE_OPEN_MEMORY
)!=0;
5044 readOnly
= (fout
&SQLITE_OPEN_READONLY
)!=0;
5046 /* If the file was successfully opened for read/write access,
5047 ** choose a default page size in case we have to create the
5048 ** database file. The default page size is the maximum of:
5050 ** + SQLITE_DEFAULT_PAGE_SIZE,
5051 ** + The value returned by sqlite3OsSectorSize()
5052 ** + The largest page size that can be written atomically.
5054 if( rc
==SQLITE_OK
){
5055 int iDc
= sqlite3OsDeviceCharacteristics(pPager
->fd
);
5057 setSectorSize(pPager
);
5058 assert(SQLITE_DEFAULT_PAGE_SIZE
<=SQLITE_MAX_DEFAULT_PAGE_SIZE
);
5059 if( szPageDflt
<pPager
->sectorSize
){
5060 if( pPager
->sectorSize
>SQLITE_MAX_DEFAULT_PAGE_SIZE
){
5061 szPageDflt
= SQLITE_MAX_DEFAULT_PAGE_SIZE
;
5063 szPageDflt
= (u32
)pPager
->sectorSize
;
5066 #ifdef SQLITE_ENABLE_ATOMIC_WRITE
5069 assert(SQLITE_IOCAP_ATOMIC512
==(512>>8));
5070 assert(SQLITE_IOCAP_ATOMIC64K
==(65536>>8));
5071 assert(SQLITE_MAX_DEFAULT_PAGE_SIZE
<=65536);
5072 for(ii
=szPageDflt
; ii
<=SQLITE_MAX_DEFAULT_PAGE_SIZE
; ii
=ii
*2){
5073 if( iDc
&(SQLITE_IOCAP_ATOMIC
|(ii
>>8)) ){
5080 pPager
->noLock
= sqlite3_uri_boolean(pPager
->zFilename
, "nolock", 0);
5081 if( (iDc
& SQLITE_IOCAP_IMMUTABLE
)!=0
5082 || sqlite3_uri_boolean(pPager
->zFilename
, "immutable", 0) ){
5083 vfsFlags
|= SQLITE_OPEN_READONLY
;
5084 goto act_like_temp_file
;
5088 /* If a temporary file is requested, it is not opened immediately.
5089 ** In this case we accept the default page size and delay actually
5090 ** opening the file until the first call to OsWrite().
5092 ** This branch is also run for an in-memory database. An in-memory
5093 ** database is the same as a temp-file that is never written out to
5094 ** disk and uses an in-memory rollback journal.
5096 ** This branch also runs for files marked as immutable.
5100 pPager
->eState
= PAGER_READER
; /* Pretend we already have a lock */
5101 pPager
->eLock
= EXCLUSIVE_LOCK
; /* Pretend we are in EXCLUSIVE mode */
5102 pPager
->noLock
= 1; /* Do no locking */
5103 readOnly
= (vfsFlags
&SQLITE_OPEN_READONLY
);
5106 /* The following call to PagerSetPagesize() serves to set the value of
5107 ** Pager.pageSize and to allocate the Pager.pTmpSpace buffer.
5109 if( rc
==SQLITE_OK
){
5110 assert( pPager
->memDb
==0 );
5111 rc
= sqlite3PagerSetPagesize(pPager
, &szPageDflt
, -1);
5112 testcase( rc
!=SQLITE_OK
);
5115 /* Initialize the PCache object. */
5116 if( rc
==SQLITE_OK
){
5117 nExtra
= ROUND8(nExtra
);
5118 assert( nExtra
>=8 && nExtra
<1000 );
5119 rc
= sqlite3PcacheOpen(szPageDflt
, nExtra
, !memDb
,
5120 !memDb
?pagerStress
:0, (void *)pPager
, pPager
->pPCache
);
5123 /* If an error occurred above, free the Pager structure and close the file.
5125 if( rc
!=SQLITE_OK
){
5126 sqlite3OsClose(pPager
->fd
);
5127 sqlite3PageFree(pPager
->pTmpSpace
);
5128 sqlite3_free(pPager
);
5132 PAGERTRACE(("OPEN %d %s\n", FILEHANDLEID(pPager
->fd
), pPager
->zFilename
));
5133 IOTRACE(("OPEN %p %s\n", pPager
, pPager
->zFilename
))
5135 pPager
->useJournal
= (u8
)useJournal
;
5136 /* pPager->stmtOpen = 0; */
5137 /* pPager->stmtInUse = 0; */
5138 /* pPager->nRef = 0; */
5139 /* pPager->stmtSize = 0; */
5140 /* pPager->stmtJSize = 0; */
5141 /* pPager->nPage = 0; */
5142 pPager
->mxPgno
= SQLITE_MAX_PAGE_COUNT
;
5143 /* pPager->state = PAGER_UNLOCK; */
5144 /* pPager->errMask = 0; */
5145 pPager
->tempFile
= (u8
)tempFile
;
5146 assert( tempFile
==PAGER_LOCKINGMODE_NORMAL
5147 || tempFile
==PAGER_LOCKINGMODE_EXCLUSIVE
);
5148 assert( PAGER_LOCKINGMODE_EXCLUSIVE
==1 );
5149 pPager
->exclusiveMode
= (u8
)tempFile
;
5150 pPager
->changeCountDone
= pPager
->tempFile
;
5151 pPager
->memDb
= (u8
)memDb
;
5152 pPager
->readOnly
= (u8
)readOnly
;
5153 assert( useJournal
|| pPager
->tempFile
);
5154 sqlite3PagerSetFlags(pPager
, (SQLITE_DEFAULT_SYNCHRONOUS
+1)|PAGER_CACHESPILL
);
5155 /* pPager->pFirst = 0; */
5156 /* pPager->pFirstSynced = 0; */
5157 /* pPager->pLast = 0; */
5158 pPager
->nExtra
= (u16
)nExtra
;
5159 pPager
->journalSizeLimit
= SQLITE_DEFAULT_JOURNAL_SIZE_LIMIT
;
5160 assert( isOpen(pPager
->fd
) || tempFile
);
5161 setSectorSize(pPager
);
5163 pPager
->journalMode
= PAGER_JOURNALMODE_OFF
;
5164 }else if( memDb
|| memJM
){
5165 pPager
->journalMode
= PAGER_JOURNALMODE_MEMORY
;
5167 /* pPager->xBusyHandler = 0; */
5168 /* pPager->pBusyHandlerArg = 0; */
5169 pPager
->xReiniter
= xReinit
;
5170 setGetterMethod(pPager
);
5171 /* memset(pPager->aHash, 0, sizeof(pPager->aHash)); */
5172 /* pPager->szMmap = SQLITE_DEFAULT_MMAP_SIZE // will be set by btree.c */
5179 ** Return the sqlite3_file for the main database given the name
5180 ** of the corresponding WAL or Journal name as passed into
5183 sqlite3_file
*sqlite3_database_file_object(const char *zName
){
5186 while( zName
[-1]!=0 || zName
[-2]!=0 || zName
[-3]!=0 || zName
[-4]!=0 ){
5189 p
= zName
- 4 - sizeof(Pager
*);
5190 assert( EIGHT_BYTE_ALIGNMENT(p
) );
5191 pPager
= *(Pager
**)p
;
5197 ** This function is called after transitioning from PAGER_UNLOCK to
5198 ** PAGER_SHARED state. It tests if there is a hot journal present in
5199 ** the file-system for the given pager. A hot journal is one that
5200 ** needs to be played back. According to this function, a hot-journal
5201 ** file exists if the following criteria are met:
5203 ** * The journal file exists in the file system, and
5204 ** * No process holds a RESERVED or greater lock on the database file, and
5205 ** * The database file itself is greater than 0 bytes in size, and
5206 ** * The first byte of the journal file exists and is not 0x00.
5208 ** If the current size of the database file is 0 but a journal file
5209 ** exists, that is probably an old journal left over from a prior
5210 ** database with the same name. In this case the journal file is
5211 ** just deleted using OsDelete, *pExists is set to 0 and SQLITE_OK
5214 ** This routine does not check if there is a super-journal filename
5215 ** at the end of the file. If there is, and that super-journal file
5216 ** does not exist, then the journal file is not really hot. In this
5217 ** case this routine will return a false-positive. The pager_playback()
5218 ** routine will discover that the journal file is not really hot and
5219 ** will not roll it back.
5221 ** If a hot-journal file is found to exist, *pExists is set to 1 and
5222 ** SQLITE_OK returned. If no hot-journal file is present, *pExists is
5223 ** set to 0 and SQLITE_OK returned. If an IO error occurs while trying
5224 ** to determine whether or not a hot-journal file exists, the IO error
5225 ** code is returned and the value of *pExists is undefined.
5227 static int hasHotJournal(Pager
*pPager
, int *pExists
){
5228 sqlite3_vfs
* const pVfs
= pPager
->pVfs
;
5229 int rc
= SQLITE_OK
; /* Return code */
5230 int exists
= 1; /* True if a journal file is present */
5231 int jrnlOpen
= !!isOpen(pPager
->jfd
);
5233 assert( pPager
->useJournal
);
5234 assert( isOpen(pPager
->fd
) );
5235 assert( pPager
->eState
==PAGER_OPEN
);
5237 assert( jrnlOpen
==0 || ( sqlite3OsDeviceCharacteristics(pPager
->jfd
) &
5238 SQLITE_IOCAP_UNDELETABLE_WHEN_OPEN
5243 rc
= sqlite3OsAccess(pVfs
, pPager
->zJournal
, SQLITE_ACCESS_EXISTS
, &exists
);
5245 if( rc
==SQLITE_OK
&& exists
){
5246 int locked
= 0; /* True if some process holds a RESERVED lock */
5248 /* Race condition here: Another process might have been holding the
5249 ** the RESERVED lock and have a journal open at the sqlite3OsAccess()
5250 ** call above, but then delete the journal and drop the lock before
5251 ** we get to the following sqlite3OsCheckReservedLock() call. If that
5252 ** is the case, this routine might think there is a hot journal when
5253 ** in fact there is none. This results in a false-positive which will
5254 ** be dealt with by the playback routine. Ticket #3883.
5256 rc
= sqlite3OsCheckReservedLock(pPager
->fd
, &locked
);
5257 if( rc
==SQLITE_OK
&& !locked
){
5258 Pgno nPage
; /* Number of pages in database file */
5260 assert( pPager
->tempFile
==0 );
5261 rc
= pagerPagecount(pPager
, &nPage
);
5262 if( rc
==SQLITE_OK
){
5263 /* If the database is zero pages in size, that means that either (1) the
5264 ** journal is a remnant from a prior database with the same name where
5265 ** the database file but not the journal was deleted, or (2) the initial
5266 ** transaction that populates a new database is being rolled back.
5267 ** In either case, the journal file can be deleted. However, take care
5268 ** not to delete the journal file if it is already open due to
5269 ** journal_mode=PERSIST.
5271 if( nPage
==0 && !jrnlOpen
){
5272 sqlite3BeginBenignMalloc();
5273 if( pagerLockDb(pPager
, RESERVED_LOCK
)==SQLITE_OK
){
5274 sqlite3OsDelete(pVfs
, pPager
->zJournal
, 0);
5275 if( !pPager
->exclusiveMode
) pagerUnlockDb(pPager
, SHARED_LOCK
);
5277 sqlite3EndBenignMalloc();
5279 /* The journal file exists and no other connection has a reserved
5280 ** or greater lock on the database file. Now check that there is
5281 ** at least one non-zero bytes at the start of the journal file.
5282 ** If there is, then we consider this journal to be hot. If not,
5283 ** it can be ignored.
5286 int f
= SQLITE_OPEN_READONLY
|SQLITE_OPEN_MAIN_JOURNAL
;
5287 rc
= sqlite3OsOpen(pVfs
, pPager
->zJournal
, pPager
->jfd
, f
, &f
);
5289 if( rc
==SQLITE_OK
){
5291 rc
= sqlite3OsRead(pPager
->jfd
, (void *)&first
, 1, 0);
5292 if( rc
==SQLITE_IOERR_SHORT_READ
){
5296 sqlite3OsClose(pPager
->jfd
);
5298 *pExists
= (first
!=0);
5299 }else if( rc
==SQLITE_CANTOPEN
){
5300 /* If we cannot open the rollback journal file in order to see if
5301 ** it has a zero header, that might be due to an I/O error, or
5302 ** it might be due to the race condition described above and in
5303 ** ticket #3883. Either way, assume that the journal is hot.
5304 ** This might be a false positive. But if it is, then the
5305 ** automatic journal playback and recovery mechanism will deal
5306 ** with it under an EXCLUSIVE lock where we do not need to
5307 ** worry so much with race conditions.
5321 ** This function is called to obtain a shared lock on the database file.
5322 ** It is illegal to call sqlite3PagerGet() until after this function
5323 ** has been successfully called. If a shared-lock is already held when
5324 ** this function is called, it is a no-op.
5326 ** The following operations are also performed by this function.
5328 ** 1) If the pager is currently in PAGER_OPEN state (no lock held
5329 ** on the database file), then an attempt is made to obtain a
5330 ** SHARED lock on the database file. Immediately after obtaining
5331 ** the SHARED lock, the file-system is checked for a hot-journal,
5332 ** which is played back if present. Following any hot-journal
5333 ** rollback, the contents of the cache are validated by checking
5334 ** the 'change-counter' field of the database file header and
5335 ** discarded if they are found to be invalid.
5337 ** 2) If the pager is running in exclusive-mode, and there are currently
5338 ** no outstanding references to any pages, and is in the error state,
5339 ** then an attempt is made to clear the error state by discarding
5340 ** the contents of the page cache and rolling back any open journal
5343 ** If everything is successful, SQLITE_OK is returned. If an IO error
5344 ** occurs while locking the database, checking for a hot-journal file or
5345 ** rolling back a journal file, the IO error code is returned.
5347 int sqlite3PagerSharedLock(Pager
*pPager
){
5348 int rc
= SQLITE_OK
; /* Return code */
5350 /* This routine is only called from b-tree and only when there are no
5351 ** outstanding pages. This implies that the pager state should either
5352 ** be OPEN or READER. READER is only possible if the pager is or was in
5353 ** exclusive access mode. */
5354 assert( sqlite3PcacheRefCount(pPager
->pPCache
)==0 );
5355 assert( assert_pager_state(pPager
) );
5356 assert( pPager
->eState
==PAGER_OPEN
|| pPager
->eState
==PAGER_READER
);
5357 assert( pPager
->errCode
==SQLITE_OK
);
5359 if( !pagerUseWal(pPager
) && pPager
->eState
==PAGER_OPEN
){
5360 int bHotJournal
= 1; /* True if there exists a hot journal-file */
5363 assert( pPager
->tempFile
==0 || pPager
->eLock
==EXCLUSIVE_LOCK
);
5365 rc
= pager_wait_on_lock(pPager
, SHARED_LOCK
);
5366 if( rc
!=SQLITE_OK
){
5367 assert( pPager
->eLock
==NO_LOCK
|| pPager
->eLock
==UNKNOWN_LOCK
);
5371 /* If a journal file exists, and there is no RESERVED lock on the
5372 ** database file, then it either needs to be played back or deleted.
5374 if( pPager
->eLock
<=SHARED_LOCK
){
5375 rc
= hasHotJournal(pPager
, &bHotJournal
);
5377 if( rc
!=SQLITE_OK
){
5381 if( pPager
->readOnly
){
5382 rc
= SQLITE_READONLY_ROLLBACK
;
5386 /* Get an EXCLUSIVE lock on the database file. At this point it is
5387 ** important that a RESERVED lock is not obtained on the way to the
5388 ** EXCLUSIVE lock. If it were, another process might open the
5389 ** database file, detect the RESERVED lock, and conclude that the
5390 ** database is safe to read while this process is still rolling the
5391 ** hot-journal back.
5393 ** Because the intermediate RESERVED lock is not requested, any
5394 ** other process attempting to access the database file will get to
5395 ** this point in the code and fail to obtain its own EXCLUSIVE lock
5396 ** on the database file.
5398 ** Unless the pager is in locking_mode=exclusive mode, the lock is
5399 ** downgraded to SHARED_LOCK before this function returns.
5401 rc
= pagerLockDb(pPager
, EXCLUSIVE_LOCK
);
5402 if( rc
!=SQLITE_OK
){
5406 /* If it is not already open and the file exists on disk, open the
5407 ** journal for read/write access. Write access is required because
5408 ** in exclusive-access mode the file descriptor will be kept open
5409 ** and possibly used for a transaction later on. Also, write-access
5410 ** is usually required to finalize the journal in journal_mode=persist
5411 ** mode (and also for journal_mode=truncate on some systems).
5413 ** If the journal does not exist, it usually means that some
5414 ** other connection managed to get in and roll it back before
5415 ** this connection obtained the exclusive lock above. Or, it
5416 ** may mean that the pager was in the error-state when this
5417 ** function was called and the journal file does not exist.
5419 if( !isOpen(pPager
->jfd
) && pPager
->journalMode
!=PAGER_JOURNALMODE_OFF
){
5420 sqlite3_vfs
* const pVfs
= pPager
->pVfs
;
5421 int bExists
; /* True if journal file exists */
5422 rc
= sqlite3OsAccess(
5423 pVfs
, pPager
->zJournal
, SQLITE_ACCESS_EXISTS
, &bExists
);
5424 if( rc
==SQLITE_OK
&& bExists
){
5426 int f
= SQLITE_OPEN_READWRITE
|SQLITE_OPEN_MAIN_JOURNAL
;
5427 assert( !pPager
->tempFile
);
5428 rc
= sqlite3OsOpen(pVfs
, pPager
->zJournal
, pPager
->jfd
, f
, &fout
);
5429 assert( rc
!=SQLITE_OK
|| isOpen(pPager
->jfd
) );
5430 if( rc
==SQLITE_OK
&& fout
&SQLITE_OPEN_READONLY
){
5431 rc
= SQLITE_CANTOPEN_BKPT
;
5432 sqlite3OsClose(pPager
->jfd
);
5437 /* Playback and delete the journal. Drop the database write
5438 ** lock and reacquire the read lock. Purge the cache before
5439 ** playing back the hot-journal so that we don't end up with
5440 ** an inconsistent cache. Sync the hot journal before playing
5441 ** it back since the process that crashed and left the hot journal
5442 ** probably did not sync it and we are required to always sync
5443 ** the journal before playing it back.
5445 if( isOpen(pPager
->jfd
) ){
5446 assert( rc
==SQLITE_OK
);
5447 rc
= pagerSyncHotJournal(pPager
);
5448 if( rc
==SQLITE_OK
){
5449 rc
= pager_playback(pPager
, !pPager
->tempFile
);
5450 pPager
->eState
= PAGER_OPEN
;
5452 }else if( !pPager
->exclusiveMode
){
5453 pagerUnlockDb(pPager
, SHARED_LOCK
);
5456 if( rc
!=SQLITE_OK
){
5457 /* This branch is taken if an error occurs while trying to open
5458 ** or roll back a hot-journal while holding an EXCLUSIVE lock. The
5459 ** pager_unlock() routine will be called before returning to unlock
5460 ** the file. If the unlock attempt fails, then Pager.eLock must be
5461 ** set to UNKNOWN_LOCK (see the comment above the #define for
5462 ** UNKNOWN_LOCK above for an explanation).
5464 ** In order to get pager_unlock() to do this, set Pager.eState to
5465 ** PAGER_ERROR now. This is not actually counted as a transition
5466 ** to ERROR state in the state diagram at the top of this file,
5467 ** since we know that the same call to pager_unlock() will very
5468 ** shortly transition the pager object to the OPEN state. Calling
5469 ** assert_pager_state() would fail now, as it should not be possible
5470 ** to be in ERROR state when there are zero outstanding page
5473 pager_error(pPager
, rc
);
5477 assert( pPager
->eState
==PAGER_OPEN
);
5478 assert( (pPager
->eLock
==SHARED_LOCK
)
5479 || (pPager
->exclusiveMode
&& pPager
->eLock
>SHARED_LOCK
)
5483 if( !pPager
->tempFile
&& pPager
->hasHeldSharedLock
){
5484 /* The shared-lock has just been acquired then check to
5485 ** see if the database has been modified. If the database has changed,
5486 ** flush the cache. The hasHeldSharedLock flag prevents this from
5487 ** occurring on the very first access to a file, in order to save a
5488 ** single unnecessary sqlite3OsRead() call at the start-up.
5490 ** Database changes are detected by looking at 15 bytes beginning
5491 ** at offset 24 into the file. The first 4 of these 16 bytes are
5492 ** a 32-bit counter that is incremented with each change. The
5493 ** other bytes change randomly with each file change when
5494 ** a codec is in use.
5496 ** There is a vanishingly small chance that a change will not be
5497 ** detected. The chance of an undetected change is so small that
5498 ** it can be neglected.
5500 char dbFileVers
[sizeof(pPager
->dbFileVers
)];
5502 IOTRACE(("CKVERS %p %d\n", pPager
, sizeof(dbFileVers
)));
5503 rc
= sqlite3OsRead(pPager
->fd
, &dbFileVers
, sizeof(dbFileVers
), 24);
5504 if( rc
!=SQLITE_OK
){
5505 if( rc
!=SQLITE_IOERR_SHORT_READ
){
5508 memset(dbFileVers
, 0, sizeof(dbFileVers
));
5511 if( memcmp(pPager
->dbFileVers
, dbFileVers
, sizeof(dbFileVers
))!=0 ){
5512 pager_reset(pPager
);
5514 /* Unmap the database file. It is possible that external processes
5515 ** may have truncated the database file and then extended it back
5516 ** to its original size while this process was not holding a lock.
5517 ** In this case there may exist a Pager.pMap mapping that appears
5518 ** to be the right size but is not actually valid. Avoid this
5519 ** possibility by unmapping the db here. */
5520 if( USEFETCH(pPager
) ){
5521 sqlite3OsUnfetch(pPager
->fd
, 0, 0);
5526 /* If there is a WAL file in the file-system, open this database in WAL
5527 ** mode. Otherwise, the following function call is a no-op.
5529 rc
= pagerOpenWalIfPresent(pPager
);
5530 #ifndef SQLITE_OMIT_WAL
5531 assert( pPager
->pWal
==0 || rc
==SQLITE_OK
);
5535 if( pagerUseWal(pPager
) ){
5536 assert( rc
==SQLITE_OK
);
5537 rc
= pagerBeginReadTransaction(pPager
);
5540 if( pPager
->tempFile
==0 && pPager
->eState
==PAGER_OPEN
&& rc
==SQLITE_OK
){
5541 rc
= pagerPagecount(pPager
, &pPager
->dbSize
);
5545 if( rc
!=SQLITE_OK
){
5547 pager_unlock(pPager
);
5548 assert( pPager
->eState
==PAGER_OPEN
);
5550 pPager
->eState
= PAGER_READER
;
5551 pPager
->hasHeldSharedLock
= 1;
5557 ** If the reference count has reached zero, rollback any active
5558 ** transaction and unlock the pager.
5560 ** Except, in locking_mode=EXCLUSIVE when there is nothing to in
5561 ** the rollback journal, the unlock is not performed and there is
5562 ** nothing to rollback, so this routine is a no-op.
5564 static void pagerUnlockIfUnused(Pager
*pPager
){
5565 if( sqlite3PcacheRefCount(pPager
->pPCache
)==0 ){
5566 assert( pPager
->nMmapOut
==0 ); /* because page1 is never memory mapped */
5567 pagerUnlockAndRollback(pPager
);
5572 ** The page getter methods each try to acquire a reference to a
5573 ** page with page number pgno. If the requested reference is
5574 ** successfully obtained, it is copied to *ppPage and SQLITE_OK returned.
5576 ** There are different implementations of the getter method depending
5577 ** on the current state of the pager.
5579 ** getPageNormal() -- The normal getter
5580 ** getPageError() -- Used if the pager is in an error state
5581 ** getPageMmap() -- Used if memory-mapped I/O is enabled
5583 ** If the requested page is already in the cache, it is returned.
5584 ** Otherwise, a new page object is allocated and populated with data
5585 ** read from the database file. In some cases, the pcache module may
5586 ** choose not to allocate a new page object and may reuse an existing
5587 ** object with no outstanding references.
5589 ** The extra data appended to a page is always initialized to zeros the
5590 ** first time a page is loaded into memory. If the page requested is
5591 ** already in the cache when this function is called, then the extra
5592 ** data is left as it was when the page object was last used.
5594 ** If the database image is smaller than the requested page or if
5595 ** the flags parameter contains the PAGER_GET_NOCONTENT bit and the
5596 ** requested page is not already stored in the cache, then no
5597 ** actual disk read occurs. In this case the memory image of the
5598 ** page is initialized to all zeros.
5600 ** If PAGER_GET_NOCONTENT is true, it means that we do not care about
5601 ** the contents of the page. This occurs in two scenarios:
5603 ** a) When reading a free-list leaf page from the database, and
5605 ** b) When a savepoint is being rolled back and we need to load
5606 ** a new page into the cache to be filled with the data read
5607 ** from the savepoint journal.
5609 ** If PAGER_GET_NOCONTENT is true, then the data returned is zeroed instead
5610 ** of being read from the database. Additionally, the bits corresponding
5611 ** to pgno in Pager.pInJournal (bitvec of pages already written to the
5612 ** journal file) and the PagerSavepoint.pInSavepoint bitvecs of any open
5613 ** savepoints are set. This means if the page is made writable at any
5614 ** point in the future, using a call to sqlite3PagerWrite(), its contents
5615 ** will not be journaled. This saves IO.
5617 ** The acquisition might fail for several reasons. In all cases,
5618 ** an appropriate error code is returned and *ppPage is set to NULL.
5620 ** See also sqlite3PagerLookup(). Both this routine and Lookup() attempt
5621 ** to find a page in the in-memory cache first. If the page is not already
5622 ** in memory, this routine goes to disk to read it in whereas Lookup()
5623 ** just returns 0. This routine acquires a read-lock the first time it
5624 ** has to go to disk, and could also playback an old journal if necessary.
5625 ** Since Lookup() never goes to disk, it never has to deal with locks
5626 ** or journal files.
5628 static int getPageNormal(
5629 Pager
*pPager
, /* The pager open on the database file */
5630 Pgno pgno
, /* Page number to fetch */
5631 DbPage
**ppPage
, /* Write a pointer to the page here */
5632 int flags
/* PAGER_GET_XXX flags */
5636 u8 noContent
; /* True if PAGER_GET_NOCONTENT is set */
5637 sqlite3_pcache_page
*pBase
;
5639 assert( pPager
->errCode
==SQLITE_OK
);
5640 assert( pPager
->eState
>=PAGER_READER
);
5641 assert( assert_pager_state(pPager
) );
5642 assert( pPager
->hasHeldSharedLock
==1 );
5644 if( pgno
==0 ) return SQLITE_CORRUPT_BKPT
;
5645 pBase
= sqlite3PcacheFetch(pPager
->pPCache
, pgno
, 3);
5648 rc
= sqlite3PcacheFetchStress(pPager
->pPCache
, pgno
, &pBase
);
5649 if( rc
!=SQLITE_OK
) goto pager_acquire_err
;
5651 rc
= SQLITE_NOMEM_BKPT
;
5652 goto pager_acquire_err
;
5655 pPg
= *ppPage
= sqlite3PcacheFetchFinish(pPager
->pPCache
, pgno
, pBase
);
5656 assert( pPg
==(*ppPage
) );
5657 assert( pPg
->pgno
==pgno
);
5658 assert( pPg
->pPager
==pPager
|| pPg
->pPager
==0 );
5660 noContent
= (flags
& PAGER_GET_NOCONTENT
)!=0;
5661 if( pPg
->pPager
&& !noContent
){
5662 /* In this case the pcache already contains an initialized copy of
5663 ** the page. Return without further ado. */
5664 assert( pgno
!=PAGER_SJ_PGNO(pPager
) );
5665 pPager
->aStat
[PAGER_STAT_HIT
]++;
5669 /* The pager cache has created a new page. Its content needs to
5670 ** be initialized. But first some error checks:
5672 ** (*) obsolete. Was: maximum page number is 2^31
5673 ** (2) Never try to fetch the locking page
5675 if( pgno
==PAGER_SJ_PGNO(pPager
) ){
5676 rc
= SQLITE_CORRUPT_BKPT
;
5677 goto pager_acquire_err
;
5680 pPg
->pPager
= pPager
;
5682 assert( !isOpen(pPager
->fd
) || !MEMDB
);
5683 if( !isOpen(pPager
->fd
) || pPager
->dbSize
<pgno
|| noContent
){
5684 if( pgno
>pPager
->mxPgno
){
5686 if( pgno
<=pPager
->dbSize
){
5687 sqlite3PcacheRelease(pPg
);
5690 goto pager_acquire_err
;
5693 /* Failure to set the bits in the InJournal bit-vectors is benign.
5694 ** It merely means that we might do some extra work to journal a
5695 ** page that does not need to be journaled. Nevertheless, be sure
5696 ** to test the case where a malloc error occurs while trying to set
5697 ** a bit in a bit vector.
5699 sqlite3BeginBenignMalloc();
5700 if( pgno
<=pPager
->dbOrigSize
){
5701 TESTONLY( rc
= ) sqlite3BitvecSet(pPager
->pInJournal
, pgno
);
5702 testcase( rc
==SQLITE_NOMEM
);
5704 TESTONLY( rc
= ) addToSavepointBitvecs(pPager
, pgno
);
5705 testcase( rc
==SQLITE_NOMEM
);
5706 sqlite3EndBenignMalloc();
5708 memset(pPg
->pData
, 0, pPager
->pageSize
);
5709 IOTRACE(("ZERO %p %d\n", pPager
, pgno
));
5711 assert( pPg
->pPager
==pPager
);
5712 pPager
->aStat
[PAGER_STAT_MISS
]++;
5713 rc
= readDbPage(pPg
);
5714 if( rc
!=SQLITE_OK
){
5715 goto pager_acquire_err
;
5718 pager_set_pagehash(pPg
);
5723 assert( rc
!=SQLITE_OK
);
5725 sqlite3PcacheDrop(pPg
);
5727 pagerUnlockIfUnused(pPager
);
5732 #if SQLITE_MAX_MMAP_SIZE>0
5733 /* The page getter for when memory-mapped I/O is enabled */
5734 static int getPageMMap(
5735 Pager
*pPager
, /* The pager open on the database file */
5736 Pgno pgno
, /* Page number to fetch */
5737 DbPage
**ppPage
, /* Write a pointer to the page here */
5738 int flags
/* PAGER_GET_XXX flags */
5742 u32 iFrame
= 0; /* Frame to read from WAL file */
5744 /* It is acceptable to use a read-only (mmap) page for any page except
5745 ** page 1 if there is no write-transaction open or the ACQUIRE_READONLY
5746 ** flag was specified by the caller. And so long as the db is not a
5747 ** temporary or in-memory database. */
5748 const int bMmapOk
= (pgno
>1
5749 && (pPager
->eState
==PAGER_READER
|| (flags
& PAGER_GET_READONLY
))
5752 assert( USEFETCH(pPager
) );
5753 /* BEGIN SQLCIPHER */
5754 #ifdef SQLITE_HAS_CODEC
5755 assert( pPager
->xCodec
==0 );
5759 /* Optimization note: Adding the "pgno<=1" term before "pgno==0" here
5760 ** allows the compiler optimizer to reuse the results of the "pgno>1"
5761 ** test in the previous statement, and avoid testing pgno==0 in the
5762 ** common case where pgno is large. */
5763 if( pgno
<=1 && pgno
==0 ){
5764 return SQLITE_CORRUPT_BKPT
;
5766 assert( pPager
->eState
>=PAGER_READER
);
5767 assert( assert_pager_state(pPager
) );
5768 assert( pPager
->hasHeldSharedLock
==1 );
5769 assert( pPager
->errCode
==SQLITE_OK
);
5771 if( bMmapOk
&& pagerUseWal(pPager
) ){
5772 rc
= sqlite3WalFindFrame(pPager
->pWal
, pgno
, &iFrame
);
5773 if( rc
!=SQLITE_OK
){
5778 if( bMmapOk
&& iFrame
==0 ){
5780 rc
= sqlite3OsFetch(pPager
->fd
,
5781 (i64
)(pgno
-1) * pPager
->pageSize
, pPager
->pageSize
, &pData
5783 if( rc
==SQLITE_OK
&& pData
){
5784 if( pPager
->eState
>PAGER_READER
|| pPager
->tempFile
){
5785 pPg
= sqlite3PagerLookup(pPager
, pgno
);
5788 rc
= pagerAcquireMapPage(pPager
, pgno
, pData
, &pPg
);
5790 sqlite3OsUnfetch(pPager
->fd
, (i64
)(pgno
-1)*pPager
->pageSize
, pData
);
5793 assert( rc
==SQLITE_OK
);
5798 if( rc
!=SQLITE_OK
){
5803 return getPageNormal(pPager
, pgno
, ppPage
, flags
);
5805 #endif /* SQLITE_MAX_MMAP_SIZE>0 */
5807 /* The page getter method for when the pager is an error state */
5808 static int getPageError(
5809 Pager
*pPager
, /* The pager open on the database file */
5810 Pgno pgno
, /* Page number to fetch */
5811 DbPage
**ppPage
, /* Write a pointer to the page here */
5812 int flags
/* PAGER_GET_XXX flags */
5814 UNUSED_PARAMETER(pgno
);
5815 UNUSED_PARAMETER(flags
);
5816 assert( pPager
->errCode
!=SQLITE_OK
);
5818 return pPager
->errCode
;
5822 /* Dispatch all page fetch requests to the appropriate getter method.
5824 int sqlite3PagerGet(
5825 Pager
*pPager
, /* The pager open on the database file */
5826 Pgno pgno
, /* Page number to fetch */
5827 DbPage
**ppPage
, /* Write a pointer to the page here */
5828 int flags
/* PAGER_GET_XXX flags */
5830 #if 0 /* Trace page fetch by setting to 1 */
5832 printf("PAGE %u\n", pgno
);
5834 rc
= pPager
->xGet(pPager
, pgno
, ppPage
, flags
);
5836 printf("PAGE %u failed with 0x%02x\n", pgno
, rc
);
5841 /* Normal, high-speed version of sqlite3PagerGet() */
5842 return pPager
->xGet(pPager
, pgno
, ppPage
, flags
);
5847 ** Acquire a page if it is already in the in-memory cache. Do
5848 ** not read the page from disk. Return a pointer to the page,
5849 ** or 0 if the page is not in cache.
5851 ** See also sqlite3PagerGet(). The difference between this routine
5852 ** and sqlite3PagerGet() is that _get() will go to the disk and read
5853 ** in the page if the page is not already in cache. This routine
5854 ** returns NULL if the page is not in cache or if a disk I/O error
5855 ** has ever happened.
5857 DbPage
*sqlite3PagerLookup(Pager
*pPager
, Pgno pgno
){
5858 sqlite3_pcache_page
*pPage
;
5859 assert( pPager
!=0 );
5861 assert( pPager
->pPCache
!=0 );
5862 pPage
= sqlite3PcacheFetch(pPager
->pPCache
, pgno
, 0);
5863 assert( pPage
==0 || pPager
->hasHeldSharedLock
);
5864 if( pPage
==0 ) return 0;
5865 return sqlite3PcacheFetchFinish(pPager
->pPCache
, pgno
, pPage
);
5869 ** Release a page reference.
5871 ** The sqlite3PagerUnref() and sqlite3PagerUnrefNotNull() may only be used
5872 ** if we know that the page being released is not the last reference to page1.
5873 ** The btree layer always holds page1 open until the end, so these first
5874 ** two routines can be used to release any page other than BtShared.pPage1.
5875 ** The assert() at tag-20230419-2 proves that this constraint is always
5878 ** Use sqlite3PagerUnrefPageOne() to release page1. This latter routine
5879 ** checks the total number of outstanding pages and if the number of
5880 ** pages reaches zero it drops the database lock.
5882 void sqlite3PagerUnrefNotNull(DbPage
*pPg
){
5883 TESTONLY( Pager
*pPager
= pPg
->pPager
; )
5885 if( pPg
->flags
& PGHDR_MMAP
){
5886 assert( pPg
->pgno
!=1 ); /* Page1 is never memory mapped */
5887 pagerReleaseMapPage(pPg
);
5889 sqlite3PcacheRelease(pPg
);
5891 /* Do not use this routine to release the last reference to page1 */
5892 assert( sqlite3PcacheRefCount(pPager
->pPCache
)>0 ); /* tag-20230419-2 */
5894 void sqlite3PagerUnref(DbPage
*pPg
){
5895 if( pPg
) sqlite3PagerUnrefNotNull(pPg
);
5897 void sqlite3PagerUnrefPageOne(DbPage
*pPg
){
5900 assert( pPg
->pgno
==1 );
5901 assert( (pPg
->flags
& PGHDR_MMAP
)==0 ); /* Page1 is never memory mapped */
5902 pPager
= pPg
->pPager
;
5903 sqlite3PcacheRelease(pPg
);
5904 pagerUnlockIfUnused(pPager
);
5908 ** This function is called at the start of every write transaction.
5909 ** There must already be a RESERVED or EXCLUSIVE lock on the database
5910 ** file when this routine is called.
5912 ** Open the journal file for pager pPager and write a journal header
5913 ** to the start of it. If there are active savepoints, open the sub-journal
5914 ** as well. This function is only used when the journal file is being
5915 ** opened to write a rollback log for a transaction. It is not used
5916 ** when opening a hot journal file to roll it back.
5918 ** If the journal file is already open (as it may be in exclusive mode),
5919 ** then this function just writes a journal header to the start of the
5920 ** already open file.
5922 ** Whether or not the journal file is opened by this function, the
5923 ** Pager.pInJournal bitvec structure is allocated.
5925 ** Return SQLITE_OK if everything is successful. Otherwise, return
5926 ** SQLITE_NOMEM if the attempt to allocate Pager.pInJournal fails, or
5927 ** an IO error code if opening or writing the journal file fails.
5929 static int pager_open_journal(Pager
*pPager
){
5930 int rc
= SQLITE_OK
; /* Return code */
5931 sqlite3_vfs
* const pVfs
= pPager
->pVfs
; /* Local cache of vfs pointer */
5933 assert( pPager
->eState
==PAGER_WRITER_LOCKED
);
5934 assert( assert_pager_state(pPager
) );
5935 assert( pPager
->pInJournal
==0 );
5937 /* If already in the error state, this function is a no-op. But on
5938 ** the other hand, this routine is never called if we are already in
5939 ** an error state. */
5940 if( NEVER(pPager
->errCode
) ) return pPager
->errCode
;
5942 if( !pagerUseWal(pPager
) && pPager
->journalMode
!=PAGER_JOURNALMODE_OFF
){
5943 pPager
->pInJournal
= sqlite3BitvecCreate(pPager
->dbSize
);
5944 if( pPager
->pInJournal
==0 ){
5945 return SQLITE_NOMEM_BKPT
;
5948 /* Open the journal file if it is not already open. */
5949 if( !isOpen(pPager
->jfd
) ){
5950 if( pPager
->journalMode
==PAGER_JOURNALMODE_MEMORY
){
5951 sqlite3MemJournalOpen(pPager
->jfd
);
5953 int flags
= SQLITE_OPEN_READWRITE
|SQLITE_OPEN_CREATE
;
5956 if( pPager
->tempFile
){
5957 flags
|= (SQLITE_OPEN_DELETEONCLOSE
|SQLITE_OPEN_TEMP_JOURNAL
);
5958 flags
|= SQLITE_OPEN_EXCLUSIVE
;
5959 nSpill
= sqlite3Config
.nStmtSpill
;
5961 flags
|= SQLITE_OPEN_MAIN_JOURNAL
;
5962 nSpill
= jrnlBufferSize(pPager
);
5965 /* Verify that the database still has the same name as it did when
5966 ** it was originally opened. */
5967 rc
= databaseIsUnmoved(pPager
);
5968 if( rc
==SQLITE_OK
){
5969 rc
= sqlite3JournalOpen (
5970 pVfs
, pPager
->zJournal
, pPager
->jfd
, flags
, nSpill
5974 assert( rc
!=SQLITE_OK
|| isOpen(pPager
->jfd
) );
5978 /* Write the first journal header to the journal file and open
5979 ** the sub-journal if necessary.
5981 if( rc
==SQLITE_OK
){
5982 /* TODO: Check if all of these are really required. */
5984 pPager
->journalOff
= 0;
5985 pPager
->setSuper
= 0;
5986 pPager
->journalHdr
= 0;
5987 rc
= writeJournalHdr(pPager
);
5991 if( rc
!=SQLITE_OK
){
5992 sqlite3BitvecDestroy(pPager
->pInJournal
);
5993 pPager
->pInJournal
= 0;
5994 pPager
->journalOff
= 0;
5996 assert( pPager
->eState
==PAGER_WRITER_LOCKED
);
5997 pPager
->eState
= PAGER_WRITER_CACHEMOD
;
6004 ** Begin a write-transaction on the specified pager object. If a
6005 ** write-transaction has already been opened, this function is a no-op.
6007 ** If the exFlag argument is false, then acquire at least a RESERVED
6008 ** lock on the database file. If exFlag is true, then acquire at least
6009 ** an EXCLUSIVE lock. If such a lock is already held, no locking
6010 ** functions need be called.
6012 ** If the subjInMemory argument is non-zero, then any sub-journal opened
6013 ** within this transaction will be opened as an in-memory file. This
6014 ** has no effect if the sub-journal is already opened (as it may be when
6015 ** running in exclusive mode) or if the transaction does not require a
6016 ** sub-journal. If the subjInMemory argument is zero, then any required
6017 ** sub-journal is implemented in-memory if pPager is an in-memory database,
6018 ** or using a temporary file otherwise.
6020 int sqlite3PagerBegin(Pager
*pPager
, int exFlag
, int subjInMemory
){
6023 if( pPager
->errCode
) return pPager
->errCode
;
6024 assert( pPager
->eState
>=PAGER_READER
&& pPager
->eState
<PAGER_ERROR
);
6025 pPager
->subjInMemory
= (u8
)subjInMemory
;
6027 if( pPager
->eState
==PAGER_READER
){
6028 assert( pPager
->pInJournal
==0 );
6030 if( pagerUseWal(pPager
) ){
6031 /* If the pager is configured to use locking_mode=exclusive, and an
6032 ** exclusive lock on the database is not already held, obtain it now.
6034 if( pPager
->exclusiveMode
&& sqlite3WalExclusiveMode(pPager
->pWal
, -1) ){
6035 rc
= pagerLockDb(pPager
, EXCLUSIVE_LOCK
);
6036 if( rc
!=SQLITE_OK
){
6039 (void)sqlite3WalExclusiveMode(pPager
->pWal
, 1);
6042 /* Grab the write lock on the log file. If successful, upgrade to
6043 ** PAGER_RESERVED state. Otherwise, return an error code to the caller.
6044 ** The busy-handler is not invoked if another connection already
6045 ** holds the write-lock. If possible, the upper layer will call it.
6047 rc
= sqlite3WalBeginWriteTransaction(pPager
->pWal
);
6049 /* Obtain a RESERVED lock on the database file. If the exFlag parameter
6050 ** is true, then immediately upgrade this to an EXCLUSIVE lock. The
6051 ** busy-handler callback can be used when upgrading to the EXCLUSIVE
6052 ** lock, but not when obtaining the RESERVED lock.
6054 rc
= pagerLockDb(pPager
, RESERVED_LOCK
);
6055 if( rc
==SQLITE_OK
&& exFlag
){
6056 rc
= pager_wait_on_lock(pPager
, EXCLUSIVE_LOCK
);
6060 if( rc
==SQLITE_OK
){
6061 /* Change to WRITER_LOCKED state.
6063 ** WAL mode sets Pager.eState to PAGER_WRITER_LOCKED or CACHEMOD
6064 ** when it has an open transaction, but never to DBMOD or FINISHED.
6065 ** This is because in those states the code to roll back savepoint
6066 ** transactions may copy data from the sub-journal into the database
6067 ** file as well as into the page cache. Which would be incorrect in
6070 pPager
->eState
= PAGER_WRITER_LOCKED
;
6071 pPager
->dbHintSize
= pPager
->dbSize
;
6072 pPager
->dbFileSize
= pPager
->dbSize
;
6073 pPager
->dbOrigSize
= pPager
->dbSize
;
6074 pPager
->journalOff
= 0;
6077 assert( rc
==SQLITE_OK
|| pPager
->eState
==PAGER_READER
);
6078 assert( rc
!=SQLITE_OK
|| pPager
->eState
==PAGER_WRITER_LOCKED
);
6079 assert( assert_pager_state(pPager
) );
6082 PAGERTRACE(("TRANSACTION %d\n", PAGERID(pPager
)));
6087 ** Write page pPg onto the end of the rollback journal.
6089 static SQLITE_NOINLINE
int pagerAddPageToRollbackJournal(PgHdr
*pPg
){
6090 Pager
*pPager
= pPg
->pPager
;
6094 i64 iOff
= pPager
->journalOff
;
6096 /* We should never write to the journal file the page that
6097 ** contains the database locks. The following assert verifies
6098 ** that we do not. */
6099 assert( pPg
->pgno
!=PAGER_SJ_PGNO(pPager
) );
6101 assert( pPager
->journalHdr
<=pPager
->journalOff
);
6102 CODEC2(pPager
, pPg
->pData
, pPg
->pgno
, 7, return SQLITE_NOMEM_BKPT
, pData2
);
6103 cksum
= pager_cksum(pPager
, (u8
*)pData2
);
6105 /* Even if an IO or diskfull error occurs while journalling the
6106 ** page in the block above, set the need-sync flag for the page.
6107 ** Otherwise, when the transaction is rolled back, the logic in
6108 ** playback_one_page() will think that the page needs to be restored
6109 ** in the database file. And if an IO error occurs while doing so,
6110 ** then corruption may follow.
6112 pPg
->flags
|= PGHDR_NEED_SYNC
;
6114 rc
= write32bits(pPager
->jfd
, iOff
, pPg
->pgno
);
6115 if( rc
!=SQLITE_OK
) return rc
;
6116 rc
= sqlite3OsWrite(pPager
->jfd
, pData2
, pPager
->pageSize
, iOff
+4);
6117 if( rc
!=SQLITE_OK
) return rc
;
6118 rc
= write32bits(pPager
->jfd
, iOff
+pPager
->pageSize
+4, cksum
);
6119 if( rc
!=SQLITE_OK
) return rc
;
6121 IOTRACE(("JOUT %p %d %lld %d\n", pPager
, pPg
->pgno
,
6122 pPager
->journalOff
, pPager
->pageSize
));
6123 PAGER_INCR(sqlite3_pager_writej_count
);
6124 PAGERTRACE(("JOURNAL %d page %d needSync=%d hash(%08x)\n",
6125 PAGERID(pPager
), pPg
->pgno
,
6126 ((pPg
->flags
&PGHDR_NEED_SYNC
)?1:0), pager_pagehash(pPg
)));
6128 pPager
->journalOff
+= 8 + pPager
->pageSize
;
6130 assert( pPager
->pInJournal
!=0 );
6131 rc
= sqlite3BitvecSet(pPager
->pInJournal
, pPg
->pgno
);
6132 testcase( rc
==SQLITE_NOMEM
);
6133 assert( rc
==SQLITE_OK
|| rc
==SQLITE_NOMEM
);
6134 rc
|= addToSavepointBitvecs(pPager
, pPg
->pgno
);
6135 assert( rc
==SQLITE_OK
|| rc
==SQLITE_NOMEM
);
6140 ** Mark a single data page as writeable. The page is written into the
6141 ** main journal or sub-journal as required. If the page is written into
6142 ** one of the journals, the corresponding bit is set in the
6143 ** Pager.pInJournal bitvec and the PagerSavepoint.pInSavepoint bitvecs
6144 ** of any open savepoints as appropriate.
6146 static int pager_write(PgHdr
*pPg
){
6147 Pager
*pPager
= pPg
->pPager
;
6150 /* This routine is not called unless a write-transaction has already
6151 ** been started. The journal file may or may not be open at this point.
6152 ** It is never called in the ERROR state.
6154 assert( pPager
->eState
==PAGER_WRITER_LOCKED
6155 || pPager
->eState
==PAGER_WRITER_CACHEMOD
6156 || pPager
->eState
==PAGER_WRITER_DBMOD
6158 assert( assert_pager_state(pPager
) );
6159 assert( pPager
->errCode
==0 );
6160 assert( pPager
->readOnly
==0 );
6163 /* The journal file needs to be opened. Higher level routines have already
6164 ** obtained the necessary locks to begin the write-transaction, but the
6165 ** rollback journal might not yet be open. Open it now if this is the case.
6167 ** This is done before calling sqlite3PcacheMakeDirty() on the page.
6168 ** Otherwise, if it were done after calling sqlite3PcacheMakeDirty(), then
6169 ** an error might occur and the pager would end up in WRITER_LOCKED state
6170 ** with pages marked as dirty in the cache.
6172 if( pPager
->eState
==PAGER_WRITER_LOCKED
){
6173 rc
= pager_open_journal(pPager
);
6174 if( rc
!=SQLITE_OK
) return rc
;
6176 assert( pPager
->eState
>=PAGER_WRITER_CACHEMOD
);
6177 assert( assert_pager_state(pPager
) );
6179 /* Mark the page that is about to be modified as dirty. */
6180 sqlite3PcacheMakeDirty(pPg
);
6182 /* If a rollback journal is in use, them make sure the page that is about
6183 ** to change is in the rollback journal, or if the page is a new page off
6184 ** then end of the file, make sure it is marked as PGHDR_NEED_SYNC.
6186 assert( (pPager
->pInJournal
!=0) == isOpen(pPager
->jfd
) );
6187 if( pPager
->pInJournal
!=0
6188 && sqlite3BitvecTestNotNull(pPager
->pInJournal
, pPg
->pgno
)==0
6190 assert( pagerUseWal(pPager
)==0 );
6191 if( pPg
->pgno
<=pPager
->dbOrigSize
){
6192 rc
= pagerAddPageToRollbackJournal(pPg
);
6193 if( rc
!=SQLITE_OK
){
6197 if( pPager
->eState
!=PAGER_WRITER_DBMOD
){
6198 pPg
->flags
|= PGHDR_NEED_SYNC
;
6200 PAGERTRACE(("APPEND %d page %d needSync=%d\n",
6201 PAGERID(pPager
), pPg
->pgno
,
6202 ((pPg
->flags
&PGHDR_NEED_SYNC
)?1:0)));
6206 /* The PGHDR_DIRTY bit is set above when the page was added to the dirty-list
6207 ** and before writing the page into the rollback journal. Wait until now,
6208 ** after the page has been successfully journalled, before setting the
6209 ** PGHDR_WRITEABLE bit that indicates that the page can be safely modified.
6211 pPg
->flags
|= PGHDR_WRITEABLE
;
6213 /* If the statement journal is open and the page is not in it,
6214 ** then write the page into the statement journal.
6216 if( pPager
->nSavepoint
>0 ){
6217 rc
= subjournalPageIfRequired(pPg
);
6220 /* Update the database size and return. */
6221 if( pPager
->dbSize
<pPg
->pgno
){
6222 pPager
->dbSize
= pPg
->pgno
;
6228 ** This is a variant of sqlite3PagerWrite() that runs when the sector size
6229 ** is larger than the page size. SQLite makes the (reasonable) assumption that
6230 ** all bytes of a sector are written together by hardware. Hence, all bytes of
6231 ** a sector need to be journalled in case of a power loss in the middle of
6234 ** Usually, the sector size is less than or equal to the page size, in which
6235 ** case pages can be individually written. This routine only runs in the
6236 ** exceptional case where the page size is smaller than the sector size.
6238 static SQLITE_NOINLINE
int pagerWriteLargeSector(PgHdr
*pPg
){
6239 int rc
= SQLITE_OK
; /* Return code */
6240 Pgno nPageCount
; /* Total number of pages in database file */
6241 Pgno pg1
; /* First page of the sector pPg is located on. */
6242 int nPage
= 0; /* Number of pages starting at pg1 to journal */
6243 int ii
; /* Loop counter */
6244 int needSync
= 0; /* True if any page has PGHDR_NEED_SYNC */
6245 Pager
*pPager
= pPg
->pPager
; /* The pager that owns pPg */
6246 Pgno nPagePerSector
= (pPager
->sectorSize
/pPager
->pageSize
);
6248 /* Set the doNotSpill NOSYNC bit to 1. This is because we cannot allow
6249 ** a journal header to be written between the pages journaled by
6253 assert( (pPager
->doNotSpill
& SPILLFLAG_NOSYNC
)==0 );
6254 pPager
->doNotSpill
|= SPILLFLAG_NOSYNC
;
6256 /* This trick assumes that both the page-size and sector-size are
6257 ** an integer power of 2. It sets variable pg1 to the identifier
6258 ** of the first page of the sector pPg is located on.
6260 pg1
= ((pPg
->pgno
-1) & ~(nPagePerSector
-1)) + 1;
6262 nPageCount
= pPager
->dbSize
;
6263 if( pPg
->pgno
>nPageCount
){
6264 nPage
= (pPg
->pgno
- pg1
)+1;
6265 }else if( (pg1
+nPagePerSector
-1)>nPageCount
){
6266 nPage
= nPageCount
+1-pg1
;
6268 nPage
= nPagePerSector
;
6271 assert(pg1
<=pPg
->pgno
);
6272 assert((pg1
+nPage
)>pPg
->pgno
);
6274 for(ii
=0; ii
<nPage
&& rc
==SQLITE_OK
; ii
++){
6277 if( pg
==pPg
->pgno
|| !sqlite3BitvecTest(pPager
->pInJournal
, pg
) ){
6278 if( pg
!=PAGER_SJ_PGNO(pPager
) ){
6279 rc
= sqlite3PagerGet(pPager
, pg
, &pPage
, 0);
6280 if( rc
==SQLITE_OK
){
6281 rc
= pager_write(pPage
);
6282 if( pPage
->flags
&PGHDR_NEED_SYNC
){
6285 sqlite3PagerUnrefNotNull(pPage
);
6288 }else if( (pPage
= sqlite3PagerLookup(pPager
, pg
))!=0 ){
6289 if( pPage
->flags
&PGHDR_NEED_SYNC
){
6292 sqlite3PagerUnrefNotNull(pPage
);
6296 /* If the PGHDR_NEED_SYNC flag is set for any of the nPage pages
6297 ** starting at pg1, then it needs to be set for all of them. Because
6298 ** writing to any of these nPage pages may damage the others, the
6299 ** journal file must contain sync()ed copies of all of them
6300 ** before any of them can be written out to the database file.
6302 if( rc
==SQLITE_OK
&& needSync
){
6304 for(ii
=0; ii
<nPage
; ii
++){
6305 PgHdr
*pPage
= sqlite3PagerLookup(pPager
, pg1
+ii
);
6307 pPage
->flags
|= PGHDR_NEED_SYNC
;
6308 sqlite3PagerUnrefNotNull(pPage
);
6313 assert( (pPager
->doNotSpill
& SPILLFLAG_NOSYNC
)!=0 );
6314 pPager
->doNotSpill
&= ~SPILLFLAG_NOSYNC
;
6319 ** Mark a data page as writeable. This routine must be called before
6320 ** making changes to a page. The caller must check the return value
6321 ** of this function and be careful not to change any page data unless
6322 ** this routine returns SQLITE_OK.
6324 ** The difference between this function and pager_write() is that this
6325 ** function also deals with the special case where 2 or more pages
6326 ** fit on a single disk sector. In this case all co-resident pages
6327 ** must have been written to the journal file before returning.
6329 ** If an error occurs, SQLITE_NOMEM or an IO error code is returned
6330 ** as appropriate. Otherwise, SQLITE_OK.
6332 int sqlite3PagerWrite(PgHdr
*pPg
){
6333 Pager
*pPager
= pPg
->pPager
;
6334 assert( (pPg
->flags
& PGHDR_MMAP
)==0 );
6335 assert( pPager
->eState
>=PAGER_WRITER_LOCKED
);
6336 assert( assert_pager_state(pPager
) );
6337 if( (pPg
->flags
& PGHDR_WRITEABLE
)!=0 && pPager
->dbSize
>=pPg
->pgno
){
6338 if( pPager
->nSavepoint
) return subjournalPageIfRequired(pPg
);
6340 }else if( pPager
->errCode
){
6341 return pPager
->errCode
;
6342 }else if( pPager
->sectorSize
> (u32
)pPager
->pageSize
){
6343 assert( pPager
->tempFile
==0 );
6344 return pagerWriteLargeSector(pPg
);
6346 return pager_write(pPg
);
6351 ** Return TRUE if the page given in the argument was previously passed
6352 ** to sqlite3PagerWrite(). In other words, return TRUE if it is ok
6353 ** to change the content of the page.
6356 int sqlite3PagerIswriteable(DbPage
*pPg
){
6357 return pPg
->flags
& PGHDR_WRITEABLE
;
6362 ** A call to this routine tells the pager that it is not necessary to
6363 ** write the information on page pPg back to the disk, even though
6364 ** that page might be marked as dirty. This happens, for example, when
6365 ** the page has been added as a leaf of the freelist and so its
6366 ** content no longer matters.
6368 ** The overlying software layer calls this routine when all of the data
6369 ** on the given page is unused. The pager marks the page as clean so
6370 ** that it does not get written to disk.
6372 ** Tests show that this optimization can quadruple the speed of large
6373 ** DELETE operations.
6375 ** This optimization cannot be used with a temp-file, as the page may
6376 ** have been dirty at the start of the transaction. In that case, if
6377 ** memory pressure forces page pPg out of the cache, the data does need
6378 ** to be written out to disk so that it may be read back in if the
6379 ** current transaction is rolled back.
6381 void sqlite3PagerDontWrite(PgHdr
*pPg
){
6382 Pager
*pPager
= pPg
->pPager
;
6383 if( !pPager
->tempFile
&& (pPg
->flags
&PGHDR_DIRTY
) && pPager
->nSavepoint
==0 ){
6384 PAGERTRACE(("DONT_WRITE page %d of %d\n", pPg
->pgno
, PAGERID(pPager
)));
6385 IOTRACE(("CLEAN %p %d\n", pPager
, pPg
->pgno
))
6386 pPg
->flags
|= PGHDR_DONT_WRITE
;
6387 pPg
->flags
&= ~PGHDR_WRITEABLE
;
6388 testcase( pPg
->flags
& PGHDR_NEED_SYNC
);
6389 pager_set_pagehash(pPg
);
6394 ** This routine is called to increment the value of the database file
6395 ** change-counter, stored as a 4-byte big-endian integer starting at
6396 ** byte offset 24 of the pager file. The secondary change counter at
6397 ** 92 is also updated, as is the SQLite version number at offset 96.
6399 ** But this only happens if the pPager->changeCountDone flag is false.
6400 ** To avoid excess churning of page 1, the update only happens once.
6401 ** See also the pager_write_changecounter() routine that does an
6402 ** unconditional update of the change counters.
6404 ** If the isDirectMode flag is zero, then this is done by calling
6405 ** sqlite3PagerWrite() on page 1, then modifying the contents of the
6406 ** page data. In this case the file will be updated when the current
6407 ** transaction is committed.
6409 ** The isDirectMode flag may only be non-zero if the library was compiled
6410 ** with the SQLITE_ENABLE_ATOMIC_WRITE macro defined. In this case,
6411 ** if isDirect is non-zero, then the database file is updated directly
6412 ** by writing an updated version of page 1 using a call to the
6413 ** sqlite3OsWrite() function.
6415 static int pager_incr_changecounter(Pager
*pPager
, int isDirectMode
){
6418 assert( pPager
->eState
==PAGER_WRITER_CACHEMOD
6419 || pPager
->eState
==PAGER_WRITER_DBMOD
6421 assert( assert_pager_state(pPager
) );
6423 /* Declare and initialize constant integer 'isDirect'. If the
6424 ** atomic-write optimization is enabled in this build, then isDirect
6425 ** is initialized to the value passed as the isDirectMode parameter
6426 ** to this function. Otherwise, it is always set to zero.
6428 ** The idea is that if the atomic-write optimization is not
6429 ** enabled at compile time, the compiler can omit the tests of
6430 ** 'isDirect' below, as well as the block enclosed in the
6431 ** "if( isDirect )" condition.
6433 #ifndef SQLITE_ENABLE_ATOMIC_WRITE
6434 # define DIRECT_MODE 0
6435 assert( isDirectMode
==0 );
6436 UNUSED_PARAMETER(isDirectMode
);
6438 # define DIRECT_MODE isDirectMode
6441 if( !pPager
->changeCountDone
&& pPager
->dbSize
>0 ){
6442 PgHdr
*pPgHdr
; /* Reference to page 1 */
6444 assert( !pPager
->tempFile
&& isOpen(pPager
->fd
) );
6446 /* Open page 1 of the file for writing. */
6447 rc
= sqlite3PagerGet(pPager
, 1, &pPgHdr
, 0);
6448 assert( pPgHdr
==0 || rc
==SQLITE_OK
);
6450 /* If page one was fetched successfully, and this function is not
6451 ** operating in direct-mode, make page 1 writable. When not in
6452 ** direct mode, page 1 is always held in cache and hence the PagerGet()
6453 ** above is always successful - hence the ALWAYS on rc==SQLITE_OK.
6455 if( !DIRECT_MODE
&& ALWAYS(rc
==SQLITE_OK
) ){
6456 rc
= sqlite3PagerWrite(pPgHdr
);
6459 if( rc
==SQLITE_OK
){
6460 /* Actually do the update of the change counter */
6461 pager_write_changecounter(pPgHdr
);
6463 /* If running in direct mode, write the contents of page 1 to the file. */
6466 assert( pPager
->dbFileSize
>0 );
6467 CODEC2(pPager
, pPgHdr
->pData
, 1, 6, rc
=SQLITE_NOMEM_BKPT
, zBuf
);
6468 if( rc
==SQLITE_OK
){
6469 rc
= sqlite3OsWrite(pPager
->fd
, zBuf
, pPager
->pageSize
, 0);
6470 pPager
->aStat
[PAGER_STAT_WRITE
]++;
6472 if( rc
==SQLITE_OK
){
6473 /* Update the pager's copy of the change-counter. Otherwise, the
6474 ** next time a read transaction is opened the cache will be
6475 ** flushed (as the change-counter values will not match). */
6476 const void *pCopy
= (const void *)&((const char *)zBuf
)[24];
6477 memcpy(&pPager
->dbFileVers
, pCopy
, sizeof(pPager
->dbFileVers
));
6478 pPager
->changeCountDone
= 1;
6481 pPager
->changeCountDone
= 1;
6485 /* Release the page reference. */
6486 sqlite3PagerUnref(pPgHdr
);
6492 ** Sync the database file to disk. This is a no-op for in-memory databases
6493 ** or pages with the Pager.noSync flag set.
6495 ** If successful, or if called on a pager for which it is a no-op, this
6496 ** function returns SQLITE_OK. Otherwise, an IO error code is returned.
6498 int sqlite3PagerSync(Pager
*pPager
, const char *zSuper
){
6500 void *pArg
= (void*)zSuper
;
6501 rc
= sqlite3OsFileControl(pPager
->fd
, SQLITE_FCNTL_SYNC
, pArg
);
6502 if( rc
==SQLITE_NOTFOUND
) rc
= SQLITE_OK
;
6503 if( rc
==SQLITE_OK
&& !pPager
->noSync
){
6505 rc
= sqlite3OsSync(pPager
->fd
, pPager
->syncFlags
);
6511 ** This function may only be called while a write-transaction is active in
6512 ** rollback. If the connection is in WAL mode, this call is a no-op.
6513 ** Otherwise, if the connection does not already have an EXCLUSIVE lock on
6514 ** the database file, an attempt is made to obtain one.
6516 ** If the EXCLUSIVE lock is already held or the attempt to obtain it is
6517 ** successful, or the connection is in WAL mode, SQLITE_OK is returned.
6518 ** Otherwise, either SQLITE_BUSY or an SQLITE_IOERR_XXX error code is
6521 int sqlite3PagerExclusiveLock(Pager
*pPager
){
6522 int rc
= pPager
->errCode
;
6523 assert( assert_pager_state(pPager
) );
6524 if( rc
==SQLITE_OK
){
6525 assert( pPager
->eState
==PAGER_WRITER_CACHEMOD
6526 || pPager
->eState
==PAGER_WRITER_DBMOD
6527 || pPager
->eState
==PAGER_WRITER_LOCKED
6529 assert( assert_pager_state(pPager
) );
6530 if( 0==pagerUseWal(pPager
) ){
6531 rc
= pager_wait_on_lock(pPager
, EXCLUSIVE_LOCK
);
6538 ** Sync the database file for the pager pPager. zSuper points to the name
6539 ** of a super-journal file that should be written into the individual
6540 ** journal file. zSuper may be NULL, which is interpreted as no
6541 ** super-journal (a single database transaction).
6543 ** This routine ensures that:
6545 ** * The database file change-counter is updated,
6546 ** * the journal is synced (unless the atomic-write optimization is used),
6547 ** * all dirty pages are written to the database file,
6548 ** * the database file is truncated (if required), and
6549 ** * the database file synced.
6551 ** The only thing that remains to commit the transaction is to finalize
6552 ** (delete, truncate or zero the first part of) the journal file (or
6553 ** delete the super-journal file if specified).
6555 ** Note that if zSuper==NULL, this does not overwrite a previous value
6556 ** passed to an sqlite3PagerCommitPhaseOne() call.
6558 ** If the final parameter - noSync - is true, then the database file itself
6559 ** is not synced. The caller must call sqlite3PagerSync() directly to
6560 ** sync the database file before calling CommitPhaseTwo() to delete the
6561 ** journal file in this case.
6563 int sqlite3PagerCommitPhaseOne(
6564 Pager
*pPager
, /* Pager object */
6565 const char *zSuper
, /* If not NULL, the super-journal name */
6566 int noSync
/* True to omit the xSync on the db file */
6568 int rc
= SQLITE_OK
; /* Return code */
6570 assert( pPager
->eState
==PAGER_WRITER_LOCKED
6571 || pPager
->eState
==PAGER_WRITER_CACHEMOD
6572 || pPager
->eState
==PAGER_WRITER_DBMOD
6573 || pPager
->eState
==PAGER_ERROR
6575 assert( assert_pager_state(pPager
) );
6577 /* If a prior error occurred, report that error again. */
6578 if( NEVER(pPager
->errCode
) ) return pPager
->errCode
;
6580 /* Provide the ability to easily simulate an I/O error during testing */
6581 if( sqlite3FaultSim(400) ) return SQLITE_IOERR
;
6583 PAGERTRACE(("DATABASE SYNC: File=%s zSuper=%s nSize=%d\n",
6584 pPager
->zFilename
, zSuper
, pPager
->dbSize
));
6586 /* If no database changes have been made, return early. */
6587 if( pPager
->eState
<PAGER_WRITER_CACHEMOD
) return SQLITE_OK
;
6589 assert( MEMDB
==0 || pPager
->tempFile
);
6590 assert( isOpen(pPager
->fd
) || pPager
->tempFile
);
6591 if( 0==pagerFlushOnCommit(pPager
, 1) ){
6592 /* If this is an in-memory db, or no pages have been written to, or this
6593 ** function has already been called, it is mostly a no-op. However, any
6594 ** backup in progress needs to be restarted. */
6595 sqlite3BackupRestart(pPager
->pBackup
);
6598 if( pagerUseWal(pPager
) ){
6599 PgHdr
*pPageOne
= 0;
6600 pList
= sqlite3PcacheDirtyList(pPager
->pPCache
);
6602 /* Must have at least one page for the WAL commit flag.
6603 ** Ticket [2d1a5c67dfc2363e44f29d9bbd57f] 2011-05-18 */
6604 rc
= sqlite3PagerGet(pPager
, 1, &pPageOne
, 0);
6608 assert( rc
==SQLITE_OK
);
6609 if( ALWAYS(pList
) ){
6610 rc
= pagerWalFrames(pPager
, pList
, pPager
->dbSize
, 1);
6612 sqlite3PagerUnref(pPageOne
);
6613 if( rc
==SQLITE_OK
){
6614 sqlite3PcacheCleanAll(pPager
->pPCache
);
6617 /* The bBatch boolean is true if the batch-atomic-write commit method
6618 ** should be used. No rollback journal is created if batch-atomic-write
6621 #ifdef SQLITE_ENABLE_BATCH_ATOMIC_WRITE
6622 sqlite3_file
*fd
= pPager
->fd
;
6623 int bBatch
= zSuper
==0 /* An SQLITE_IOCAP_BATCH_ATOMIC commit */
6624 && (sqlite3OsDeviceCharacteristics(fd
) & SQLITE_IOCAP_BATCH_ATOMIC
)
6626 && sqlite3JournalIsInMemory(pPager
->jfd
);
6631 #ifdef SQLITE_ENABLE_ATOMIC_WRITE
6632 /* The following block updates the change-counter. Exactly how it
6633 ** does this depends on whether or not the atomic-update optimization
6634 ** was enabled at compile time, and if this transaction meets the
6635 ** runtime criteria to use the operation:
6637 ** * The file-system supports the atomic-write property for
6638 ** blocks of size page-size, and
6639 ** * This commit is not part of a multi-file transaction, and
6640 ** * Exactly one page has been modified and store in the journal file.
6642 ** If the optimization was not enabled at compile time, then the
6643 ** pager_incr_changecounter() function is called to update the change
6644 ** counter in 'indirect-mode'. If the optimization is compiled in but
6645 ** is not applicable to this transaction, call sqlite3JournalCreate()
6646 ** to make sure the journal file has actually been created, then call
6647 ** pager_incr_changecounter() to update the change-counter in indirect
6650 ** Otherwise, if the optimization is both enabled and applicable,
6651 ** then call pager_incr_changecounter() to update the change-counter
6652 ** in 'direct' mode. In this case the journal file will never be
6653 ** created for this transaction.
6657 assert( isOpen(pPager
->jfd
)
6658 || pPager
->journalMode
==PAGER_JOURNALMODE_OFF
6659 || pPager
->journalMode
==PAGER_JOURNALMODE_WAL
6661 if( !zSuper
&& isOpen(pPager
->jfd
)
6662 && pPager
->journalOff
==jrnlBufferSize(pPager
)
6663 && pPager
->dbSize
>=pPager
->dbOrigSize
6664 && (!(pPg
= sqlite3PcacheDirtyList(pPager
->pPCache
)) || 0==pPg
->pDirty
)
6666 /* Update the db file change counter via the direct-write method. The
6667 ** following call will modify the in-memory representation of page 1
6668 ** to include the updated change counter and then write page 1
6669 ** directly to the database file. Because of the atomic-write
6670 ** property of the host file-system, this is safe.
6672 rc
= pager_incr_changecounter(pPager
, 1);
6674 rc
= sqlite3JournalCreate(pPager
->jfd
);
6675 if( rc
==SQLITE_OK
){
6676 rc
= pager_incr_changecounter(pPager
, 0);
6680 #else /* SQLITE_ENABLE_ATOMIC_WRITE */
6681 #ifdef SQLITE_ENABLE_BATCH_ATOMIC_WRITE
6683 rc
= sqlite3JournalCreate(pPager
->jfd
);
6684 if( rc
!=SQLITE_OK
) goto commit_phase_one_exit
;
6685 assert( bBatch
==0 );
6688 rc
= pager_incr_changecounter(pPager
, 0);
6689 #endif /* !SQLITE_ENABLE_ATOMIC_WRITE */
6690 if( rc
!=SQLITE_OK
) goto commit_phase_one_exit
;
6692 /* Write the super-journal name into the journal file. If a
6693 ** super-journal file name has already been written to the journal file,
6694 ** or if zSuper is NULL (no super-journal), then this call is a no-op.
6696 rc
= writeSuperJournal(pPager
, zSuper
);
6697 if( rc
!=SQLITE_OK
) goto commit_phase_one_exit
;
6699 /* Sync the journal file and write all dirty pages to the database.
6700 ** If the atomic-update optimization is being used, this sync will not
6701 ** create the journal file or perform any real IO.
6703 ** Because the change-counter page was just modified, unless the
6704 ** atomic-update optimization is used it is almost certain that the
6705 ** journal requires a sync here. However, in locking_mode=exclusive
6706 ** on a system under memory pressure it is just possible that this is
6707 ** not the case. In this case it is likely enough that the redundant
6708 ** xSync() call will be changed to a no-op by the OS anyhow.
6710 rc
= syncJournal(pPager
, 0);
6711 if( rc
!=SQLITE_OK
) goto commit_phase_one_exit
;
6713 pList
= sqlite3PcacheDirtyList(pPager
->pPCache
);
6714 #ifdef SQLITE_ENABLE_BATCH_ATOMIC_WRITE
6716 rc
= sqlite3OsFileControl(fd
, SQLITE_FCNTL_BEGIN_ATOMIC_WRITE
, 0);
6717 if( rc
==SQLITE_OK
){
6718 rc
= pager_write_pagelist(pPager
, pList
);
6719 if( rc
==SQLITE_OK
&& pPager
->dbSize
>pPager
->dbFileSize
){
6720 char *pTmp
= pPager
->pTmpSpace
;
6721 int szPage
= (int)pPager
->pageSize
;
6722 memset(pTmp
, 0, szPage
);
6723 rc
= sqlite3OsWrite(pPager
->fd
, pTmp
, szPage
,
6724 ((i64
)pPager
->dbSize
*pPager
->pageSize
)-szPage
);
6726 if( rc
==SQLITE_OK
){
6727 rc
= sqlite3OsFileControl(fd
, SQLITE_FCNTL_COMMIT_ATOMIC_WRITE
, 0);
6729 if( rc
!=SQLITE_OK
){
6730 sqlite3OsFileControlHint(fd
, SQLITE_FCNTL_ROLLBACK_ATOMIC_WRITE
, 0);
6734 if( (rc
&0xFF)==SQLITE_IOERR
&& rc
!=SQLITE_IOERR_NOMEM
){
6735 rc
= sqlite3JournalCreate(pPager
->jfd
);
6736 if( rc
!=SQLITE_OK
){
6737 sqlite3OsClose(pPager
->jfd
);
6738 goto commit_phase_one_exit
;
6742 sqlite3OsClose(pPager
->jfd
);
6745 #endif /* SQLITE_ENABLE_BATCH_ATOMIC_WRITE */
6748 rc
= pager_write_pagelist(pPager
, pList
);
6750 if( rc
!=SQLITE_OK
){
6751 assert( rc
!=SQLITE_IOERR_BLOCKED
);
6752 goto commit_phase_one_exit
;
6754 sqlite3PcacheCleanAll(pPager
->pPCache
);
6756 /* If the file on disk is smaller than the database image, use
6757 ** pager_truncate to grow the file here. This can happen if the database
6758 ** image was extended as part of the current transaction and then the
6759 ** last page in the db image moved to the free-list. In this case the
6760 ** last page is never written out to disk, leaving the database file
6761 ** undersized. Fix this now if it is the case. */
6762 if( pPager
->dbSize
>pPager
->dbFileSize
){
6763 Pgno nNew
= pPager
->dbSize
- (pPager
->dbSize
==PAGER_SJ_PGNO(pPager
));
6764 assert( pPager
->eState
==PAGER_WRITER_DBMOD
);
6765 rc
= pager_truncate(pPager
, nNew
);
6766 if( rc
!=SQLITE_OK
) goto commit_phase_one_exit
;
6769 /* Finally, sync the database file. */
6771 rc
= sqlite3PagerSync(pPager
, zSuper
);
6773 IOTRACE(("DBSYNC %p\n", pPager
))
6777 commit_phase_one_exit
:
6778 if( rc
==SQLITE_OK
&& !pagerUseWal(pPager
) ){
6779 pPager
->eState
= PAGER_WRITER_FINISHED
;
6786 ** When this function is called, the database file has been completely
6787 ** updated to reflect the changes made by the current transaction and
6788 ** synced to disk. The journal file still exists in the file-system
6789 ** though, and if a failure occurs at this point it will eventually
6790 ** be used as a hot-journal and the current transaction rolled back.
6792 ** This function finalizes the journal file, either by deleting,
6793 ** truncating or partially zeroing it, so that it cannot be used
6794 ** for hot-journal rollback. Once this is done the transaction is
6795 ** irrevocably committed.
6797 ** If an error occurs, an IO error code is returned and the pager
6798 ** moves into the error state. Otherwise, SQLITE_OK is returned.
6800 int sqlite3PagerCommitPhaseTwo(Pager
*pPager
){
6801 int rc
= SQLITE_OK
; /* Return code */
6803 /* This routine should not be called if a prior error has occurred.
6804 ** But if (due to a coding error elsewhere in the system) it does get
6805 ** called, just return the same error code without doing anything. */
6806 if( NEVER(pPager
->errCode
) ) return pPager
->errCode
;
6807 pPager
->iDataVersion
++;
6809 assert( pPager
->eState
==PAGER_WRITER_LOCKED
6810 || pPager
->eState
==PAGER_WRITER_FINISHED
6811 || (pagerUseWal(pPager
) && pPager
->eState
==PAGER_WRITER_CACHEMOD
)
6813 assert( assert_pager_state(pPager
) );
6815 /* An optimization. If the database was not actually modified during
6816 ** this transaction, the pager is running in exclusive-mode and is
6817 ** using persistent journals, then this function is a no-op.
6819 ** The start of the journal file currently contains a single journal
6820 ** header with the nRec field set to 0. If such a journal is used as
6821 ** a hot-journal during hot-journal rollback, 0 changes will be made
6822 ** to the database file. So there is no need to zero the journal
6823 ** header. Since the pager is in exclusive mode, there is no need
6824 ** to drop any locks either.
6826 if( pPager
->eState
==PAGER_WRITER_LOCKED
6827 && pPager
->exclusiveMode
6828 && pPager
->journalMode
==PAGER_JOURNALMODE_PERSIST
6830 assert( pPager
->journalOff
==JOURNAL_HDR_SZ(pPager
) || !pPager
->journalOff
);
6831 pPager
->eState
= PAGER_READER
;
6835 PAGERTRACE(("COMMIT %d\n", PAGERID(pPager
)));
6836 rc
= pager_end_transaction(pPager
, pPager
->setSuper
, 1);
6837 return pager_error(pPager
, rc
);
6841 ** If a write transaction is open, then all changes made within the
6842 ** transaction are reverted and the current write-transaction is closed.
6843 ** The pager falls back to PAGER_READER state if successful, or PAGER_ERROR
6844 ** state if an error occurs.
6846 ** If the pager is already in PAGER_ERROR state when this function is called,
6847 ** it returns Pager.errCode immediately. No work is performed in this case.
6849 ** Otherwise, in rollback mode, this function performs two functions:
6851 ** 1) It rolls back the journal file, restoring all database file and
6852 ** in-memory cache pages to the state they were in when the transaction
6855 ** 2) It finalizes the journal file, so that it is not used for hot
6856 ** rollback at any point in the future.
6858 ** Finalization of the journal file (task 2) is only performed if the
6859 ** rollback is successful.
6861 ** In WAL mode, all cache-entries containing data modified within the
6862 ** current transaction are either expelled from the cache or reverted to
6863 ** their pre-transaction state by re-reading data from the database or
6864 ** WAL files. The WAL transaction is then closed.
6866 int sqlite3PagerRollback(Pager
*pPager
){
6867 int rc
= SQLITE_OK
; /* Return code */
6868 PAGERTRACE(("ROLLBACK %d\n", PAGERID(pPager
)));
6870 /* PagerRollback() is a no-op if called in READER or OPEN state. If
6871 ** the pager is already in the ERROR state, the rollback is not
6872 ** attempted here. Instead, the error code is returned to the caller.
6874 assert( assert_pager_state(pPager
) );
6875 if( pPager
->eState
==PAGER_ERROR
) return pPager
->errCode
;
6876 if( pPager
->eState
<=PAGER_READER
) return SQLITE_OK
;
6878 if( pagerUseWal(pPager
) ){
6880 rc
= sqlite3PagerSavepoint(pPager
, SAVEPOINT_ROLLBACK
, -1);
6881 rc2
= pager_end_transaction(pPager
, pPager
->setSuper
, 0);
6882 if( rc
==SQLITE_OK
) rc
= rc2
;
6883 }else if( !isOpen(pPager
->jfd
) || pPager
->eState
==PAGER_WRITER_LOCKED
){
6884 int eState
= pPager
->eState
;
6885 rc
= pager_end_transaction(pPager
, 0, 0);
6886 if( !MEMDB
&& eState
>PAGER_WRITER_LOCKED
){
6887 /* This can happen using journal_mode=off. Move the pager to the error
6888 ** state to indicate that the contents of the cache may not be trusted.
6889 ** Any active readers will get SQLITE_ABORT.
6891 pPager
->errCode
= SQLITE_ABORT
;
6892 pPager
->eState
= PAGER_ERROR
;
6893 setGetterMethod(pPager
);
6897 rc
= pager_playback(pPager
, 0);
6900 assert( pPager
->eState
==PAGER_READER
|| rc
!=SQLITE_OK
);
6901 assert( rc
==SQLITE_OK
|| rc
==SQLITE_FULL
|| rc
==SQLITE_CORRUPT
6902 || rc
==SQLITE_NOMEM
|| (rc
&0xFF)==SQLITE_IOERR
6903 || rc
==SQLITE_CANTOPEN
6906 /* If an error occurs during a ROLLBACK, we can no longer trust the pager
6907 ** cache. So call pager_error() on the way out to make any error persistent.
6909 return pager_error(pPager
, rc
);
6913 ** Return TRUE if the database file is opened read-only. Return FALSE
6914 ** if the database is (in theory) writable.
6916 u8
sqlite3PagerIsreadonly(Pager
*pPager
){
6917 return pPager
->readOnly
;
6922 ** Return the sum of the reference counts for all pages held by pPager.
6924 int sqlite3PagerRefcount(Pager
*pPager
){
6925 return sqlite3PcacheRefCount(pPager
->pPCache
);
6930 ** Return the approximate number of bytes of memory currently
6931 ** used by the pager and its associated cache.
6933 int sqlite3PagerMemUsed(Pager
*pPager
){
6934 int perPageSize
= pPager
->pageSize
+ pPager
->nExtra
6935 + (int)(sizeof(PgHdr
) + 5*sizeof(void*));
6936 return perPageSize
*sqlite3PcachePagecount(pPager
->pPCache
)
6937 + sqlite3MallocSize(pPager
)
6942 ** Return the number of references to the specified page.
6944 int sqlite3PagerPageRefcount(DbPage
*pPage
){
6945 return sqlite3PcachePageRefcount(pPage
);
6950 ** This routine is used for testing and analysis only.
6952 int *sqlite3PagerStats(Pager
*pPager
){
6954 a
[0] = sqlite3PcacheRefCount(pPager
->pPCache
);
6955 a
[1] = sqlite3PcachePagecount(pPager
->pPCache
);
6956 a
[2] = sqlite3PcacheGetCachesize(pPager
->pPCache
);
6957 a
[3] = pPager
->eState
==PAGER_OPEN
? -1 : (int) pPager
->dbSize
;
6958 a
[4] = pPager
->eState
;
6959 a
[5] = pPager
->errCode
;
6960 a
[6] = pPager
->aStat
[PAGER_STAT_HIT
];
6961 a
[7] = pPager
->aStat
[PAGER_STAT_MISS
];
6962 a
[8] = 0; /* Used to be pPager->nOvfl */
6963 a
[9] = pPager
->nRead
;
6964 a
[10] = pPager
->aStat
[PAGER_STAT_WRITE
];
6970 ** Parameter eStat must be one of SQLITE_DBSTATUS_CACHE_HIT, _MISS, _WRITE,
6971 ** or _WRITE+1. The SQLITE_DBSTATUS_CACHE_WRITE+1 case is a translation
6972 ** of SQLITE_DBSTATUS_CACHE_SPILL. The _SPILL case is not contiguous because
6973 ** it was added later.
6975 ** Before returning, *pnVal is incremented by the
6976 ** current cache hit or miss count, according to the value of eStat. If the
6977 ** reset parameter is non-zero, the cache hit or miss count is zeroed before
6980 void sqlite3PagerCacheStat(Pager
*pPager
, int eStat
, int reset
, int *pnVal
){
6982 assert( eStat
==SQLITE_DBSTATUS_CACHE_HIT
6983 || eStat
==SQLITE_DBSTATUS_CACHE_MISS
6984 || eStat
==SQLITE_DBSTATUS_CACHE_WRITE
6985 || eStat
==SQLITE_DBSTATUS_CACHE_WRITE
+1
6988 assert( SQLITE_DBSTATUS_CACHE_HIT
+1==SQLITE_DBSTATUS_CACHE_MISS
);
6989 assert( SQLITE_DBSTATUS_CACHE_HIT
+2==SQLITE_DBSTATUS_CACHE_WRITE
);
6990 assert( PAGER_STAT_HIT
==0 && PAGER_STAT_MISS
==1
6991 && PAGER_STAT_WRITE
==2 && PAGER_STAT_SPILL
==3 );
6993 eStat
-= SQLITE_DBSTATUS_CACHE_HIT
;
6994 *pnVal
+= pPager
->aStat
[eStat
];
6996 pPager
->aStat
[eStat
] = 0;
7001 ** Return true if this is an in-memory or temp-file backed pager.
7003 int sqlite3PagerIsMemdb(Pager
*pPager
){
7004 return pPager
->tempFile
|| pPager
->memVfs
;
7008 ** Check that there are at least nSavepoint savepoints open. If there are
7009 ** currently less than nSavepoints open, then open one or more savepoints
7010 ** to make up the difference. If the number of savepoints is already
7011 ** equal to nSavepoint, then this function is a no-op.
7013 ** If a memory allocation fails, SQLITE_NOMEM is returned. If an error
7014 ** occurs while opening the sub-journal file, then an IO error code is
7015 ** returned. Otherwise, SQLITE_OK.
7017 static SQLITE_NOINLINE
int pagerOpenSavepoint(Pager
*pPager
, int nSavepoint
){
7018 int rc
= SQLITE_OK
; /* Return code */
7019 int nCurrent
= pPager
->nSavepoint
; /* Current number of savepoints */
7020 int ii
; /* Iterator variable */
7021 PagerSavepoint
*aNew
; /* New Pager.aSavepoint array */
7023 assert( pPager
->eState
>=PAGER_WRITER_LOCKED
);
7024 assert( assert_pager_state(pPager
) );
7025 assert( nSavepoint
>nCurrent
&& pPager
->useJournal
);
7027 /* Grow the Pager.aSavepoint array using realloc(). Return SQLITE_NOMEM
7028 ** if the allocation fails. Otherwise, zero the new portion in case a
7029 ** malloc failure occurs while populating it in the for(...) loop below.
7031 aNew
= (PagerSavepoint
*)sqlite3Realloc(
7032 pPager
->aSavepoint
, sizeof(PagerSavepoint
)*nSavepoint
7035 return SQLITE_NOMEM_BKPT
;
7037 memset(&aNew
[nCurrent
], 0, (nSavepoint
-nCurrent
) * sizeof(PagerSavepoint
));
7038 pPager
->aSavepoint
= aNew
;
7040 /* Populate the PagerSavepoint structures just allocated. */
7041 for(ii
=nCurrent
; ii
<nSavepoint
; ii
++){
7042 aNew
[ii
].nOrig
= pPager
->dbSize
;
7043 if( isOpen(pPager
->jfd
) && pPager
->journalOff
>0 ){
7044 aNew
[ii
].iOffset
= pPager
->journalOff
;
7046 aNew
[ii
].iOffset
= JOURNAL_HDR_SZ(pPager
);
7048 aNew
[ii
].iSubRec
= pPager
->nSubRec
;
7049 aNew
[ii
].pInSavepoint
= sqlite3BitvecCreate(pPager
->dbSize
);
7050 aNew
[ii
].bTruncateOnRelease
= 1;
7051 if( !aNew
[ii
].pInSavepoint
){
7052 return SQLITE_NOMEM_BKPT
;
7054 if( pagerUseWal(pPager
) ){
7055 sqlite3WalSavepoint(pPager
->pWal
, aNew
[ii
].aWalData
);
7057 pPager
->nSavepoint
= ii
+1;
7059 assert( pPager
->nSavepoint
==nSavepoint
);
7060 assertTruncateConstraint(pPager
);
7063 int sqlite3PagerOpenSavepoint(Pager
*pPager
, int nSavepoint
){
7064 assert( pPager
->eState
>=PAGER_WRITER_LOCKED
);
7065 assert( assert_pager_state(pPager
) );
7067 if( nSavepoint
>pPager
->nSavepoint
&& pPager
->useJournal
){
7068 return pagerOpenSavepoint(pPager
, nSavepoint
);
7076 ** This function is called to rollback or release (commit) a savepoint.
7077 ** The savepoint to release or rollback need not be the most recently
7078 ** created savepoint.
7080 ** Parameter op is always either SAVEPOINT_ROLLBACK or SAVEPOINT_RELEASE.
7081 ** If it is SAVEPOINT_RELEASE, then release and destroy the savepoint with
7082 ** index iSavepoint. If it is SAVEPOINT_ROLLBACK, then rollback all changes
7083 ** that have occurred since the specified savepoint was created.
7085 ** The savepoint to rollback or release is identified by parameter
7086 ** iSavepoint. A value of 0 means to operate on the outermost savepoint
7087 ** (the first created). A value of (Pager.nSavepoint-1) means operate
7088 ** on the most recently created savepoint. If iSavepoint is greater than
7089 ** (Pager.nSavepoint-1), then this function is a no-op.
7091 ** If a negative value is passed to this function, then the current
7092 ** transaction is rolled back. This is different to calling
7093 ** sqlite3PagerRollback() because this function does not terminate
7094 ** the transaction or unlock the database, it just restores the
7095 ** contents of the database to its original state.
7097 ** In any case, all savepoints with an index greater than iSavepoint
7098 ** are destroyed. If this is a release operation (op==SAVEPOINT_RELEASE),
7099 ** then savepoint iSavepoint is also destroyed.
7101 ** This function may return SQLITE_NOMEM if a memory allocation fails,
7102 ** or an IO error code if an IO error occurs while rolling back a
7103 ** savepoint. If no errors occur, SQLITE_OK is returned.
7105 int sqlite3PagerSavepoint(Pager
*pPager
, int op
, int iSavepoint
){
7106 int rc
= pPager
->errCode
;
7108 #ifdef SQLITE_ENABLE_ZIPVFS
7109 if( op
==SAVEPOINT_RELEASE
) rc
= SQLITE_OK
;
7112 assert( op
==SAVEPOINT_RELEASE
|| op
==SAVEPOINT_ROLLBACK
);
7113 assert( iSavepoint
>=0 || op
==SAVEPOINT_ROLLBACK
);
7115 if( rc
==SQLITE_OK
&& iSavepoint
<pPager
->nSavepoint
){
7116 int ii
; /* Iterator variable */
7117 int nNew
; /* Number of remaining savepoints after this op. */
7119 /* Figure out how many savepoints will still be active after this
7120 ** operation. Store this value in nNew. Then free resources associated
7121 ** with any savepoints that are destroyed by this operation.
7123 nNew
= iSavepoint
+ (( op
==SAVEPOINT_RELEASE
) ? 0 : 1);
7124 for(ii
=nNew
; ii
<pPager
->nSavepoint
; ii
++){
7125 sqlite3BitvecDestroy(pPager
->aSavepoint
[ii
].pInSavepoint
);
7127 pPager
->nSavepoint
= nNew
;
7129 /* Truncate the sub-journal so that it only includes the parts
7130 ** that are still in use. */
7131 if( op
==SAVEPOINT_RELEASE
){
7132 PagerSavepoint
*pRel
= &pPager
->aSavepoint
[nNew
];
7133 if( pRel
->bTruncateOnRelease
&& isOpen(pPager
->sjfd
) ){
7134 /* Only truncate if it is an in-memory sub-journal. */
7135 if( sqlite3JournalIsInMemory(pPager
->sjfd
) ){
7136 i64 sz
= (pPager
->pageSize
+4)*(i64
)pRel
->iSubRec
;
7137 rc
= sqlite3OsTruncate(pPager
->sjfd
, sz
);
7138 assert( rc
==SQLITE_OK
);
7140 pPager
->nSubRec
= pRel
->iSubRec
;
7143 /* Else this is a rollback operation, playback the specified savepoint.
7144 ** If this is a temp-file, it is possible that the journal file has
7145 ** not yet been opened. In this case there have been no changes to
7146 ** the database file, so the playback operation can be skipped.
7148 else if( pagerUseWal(pPager
) || isOpen(pPager
->jfd
) ){
7149 PagerSavepoint
*pSavepoint
= (nNew
==0)?0:&pPager
->aSavepoint
[nNew
-1];
7150 rc
= pagerPlaybackSavepoint(pPager
, pSavepoint
);
7151 assert(rc
!=SQLITE_DONE
);
7154 #ifdef SQLITE_ENABLE_ZIPVFS
7155 /* If the cache has been modified but the savepoint cannot be rolled
7156 ** back journal_mode=off, put the pager in the error state. This way,
7157 ** if the VFS used by this pager includes ZipVFS, the entire transaction
7158 ** can be rolled back at the ZipVFS level. */
7160 pPager
->journalMode
==PAGER_JOURNALMODE_OFF
7161 && pPager
->eState
>=PAGER_WRITER_CACHEMOD
7163 pPager
->errCode
= SQLITE_ABORT
;
7164 pPager
->eState
= PAGER_ERROR
;
7165 setGetterMethod(pPager
);
7174 ** Return the full pathname of the database file.
7176 ** Except, if the pager is in-memory only, then return an empty string if
7177 ** nullIfMemDb is true. This routine is called with nullIfMemDb==1 when
7178 ** used to report the filename to the user, for compatibility with legacy
7179 ** behavior. But when the Btree needs to know the filename for matching to
7180 ** shared cache, it uses nullIfMemDb==0 so that in-memory databases can
7181 ** participate in shared-cache.
7183 ** The return value to this routine is always safe to use with
7184 ** sqlite3_uri_parameter() and sqlite3_filename_database() and friends.
7186 const char *sqlite3PagerFilename(const Pager
*pPager
, int nullIfMemDb
){
7187 static const char zFake
[8] = { 0, 0, 0, 0, 0, 0, 0, 0 };
7188 if( nullIfMemDb
&& (pPager
->memDb
|| sqlite3IsMemdb(pPager
->pVfs
)) ){
7191 return pPager
->zFilename
;
7196 ** Return the VFS structure for the pager.
7198 sqlite3_vfs
*sqlite3PagerVfs(Pager
*pPager
){
7199 return pPager
->pVfs
;
7203 ** Return the file handle for the database file associated
7204 ** with the pager. This might return NULL if the file has
7205 ** not yet been opened.
7207 sqlite3_file
*sqlite3PagerFile(Pager
*pPager
){
7212 ** Return the file handle for the journal file (if it exists).
7213 ** This will be either the rollback journal or the WAL file.
7215 sqlite3_file
*sqlite3PagerJrnlFile(Pager
*pPager
){
7219 return pPager
->pWal
? sqlite3WalFile(pPager
->pWal
) : pPager
->jfd
;
7224 ** Return the full pathname of the journal file.
7226 const char *sqlite3PagerJournalname(Pager
*pPager
){
7227 return pPager
->zJournal
;
7230 /* BEGIN SQLCIPHER */
7231 #ifdef SQLITE_HAS_CODEC
7233 ** Set or retrieve the codec for this pager
7235 void sqlcipherPagerSetCodec(
7237 void *(*xCodec
)(void*,void*,Pgno
,int),
7238 void (*xCodecSizeChng
)(void*,int,int),
7239 void (*xCodecFree
)(void*),
7242 if( pPager
->xCodecFree
){
7243 pPager
->xCodecFree(pPager
->pCodec
);
7245 pager_reset(pPager
);
7247 pPager
->xCodec
= pPager
->memDb
? 0 : xCodec
;
7248 pPager
->xCodecSizeChng
= xCodecSizeChng
;
7249 pPager
->xCodecFree
= xCodecFree
;
7250 pPager
->pCodec
= pCodec
;
7251 setGetterMethod(pPager
);
7252 pagerReportSize(pPager
);
7254 void *sqlcipherPagerGetCodec(Pager
*pPager
){
7255 return pPager
->pCodec
;
7259 ** This function is called by the wal module when writing page content
7260 ** into the log file.
7262 ** This function returns a pointer to a buffer containing the encrypted
7263 ** page content. If a malloc fails, this function may return NULL.
7265 void *sqlcipherPagerCodec(PgHdr
*pPg
){
7267 CODEC2(pPg
->pPager
, pPg
->pData
, pPg
->pgno
, 6, return 0, aData
);
7270 #endif /* SQLITE_HAS_CODEC */
7273 #ifndef SQLITE_OMIT_AUTOVACUUM
7275 ** Move the page pPg to location pgno in the file.
7277 ** There must be no references to the page previously located at
7278 ** pgno (which we call pPgOld) though that page is allowed to be
7279 ** in cache. If the page previously located at pgno is not already
7280 ** in the rollback journal, it is not put there by by this routine.
7282 ** References to the page pPg remain valid. Updating any
7283 ** meta-data associated with pPg (i.e. data stored in the nExtra bytes
7284 ** allocated along with the page) is the responsibility of the caller.
7286 ** A transaction must be active when this routine is called. It used to be
7287 ** required that a statement transaction was not active, but this restriction
7288 ** has been removed (CREATE INDEX needs to move a page when a statement
7289 ** transaction is active).
7291 ** If the fourth argument, isCommit, is non-zero, then this page is being
7292 ** moved as part of a database reorganization just before the transaction
7293 ** is being committed. In this case, it is guaranteed that the database page
7294 ** pPg refers to will not be written to again within this transaction.
7296 ** This function may return SQLITE_NOMEM or an IO error code if an error
7297 ** occurs. Otherwise, it returns SQLITE_OK.
7299 int sqlite3PagerMovepage(Pager
*pPager
, DbPage
*pPg
, Pgno pgno
, int isCommit
){
7300 PgHdr
*pPgOld
; /* The page being overwritten. */
7301 Pgno needSyncPgno
= 0; /* Old value of pPg->pgno, if sync is required */
7302 int rc
; /* Return code */
7303 Pgno origPgno
; /* The original page number */
7305 assert( pPg
->nRef
>0 );
7306 assert( pPager
->eState
==PAGER_WRITER_CACHEMOD
7307 || pPager
->eState
==PAGER_WRITER_DBMOD
7309 assert( assert_pager_state(pPager
) );
7311 /* In order to be able to rollback, an in-memory database must journal
7312 ** the page we are moving from.
7314 assert( pPager
->tempFile
|| !MEMDB
);
7315 if( pPager
->tempFile
){
7316 rc
= sqlite3PagerWrite(pPg
);
7320 /* If the page being moved is dirty and has not been saved by the latest
7321 ** savepoint, then save the current contents of the page into the
7322 ** sub-journal now. This is required to handle the following scenario:
7325 ** <journal page X, then modify it in memory>
7327 ** <Move page X to location Y>
7330 ** If page X were not written to the sub-journal here, it would not
7331 ** be possible to restore its contents when the "ROLLBACK TO one"
7332 ** statement were is processed.
7334 ** subjournalPage() may need to allocate space to store pPg->pgno into
7335 ** one or more savepoint bitvecs. This is the reason this function
7336 ** may return SQLITE_NOMEM.
7338 if( (pPg
->flags
& PGHDR_DIRTY
)!=0
7339 && SQLITE_OK
!=(rc
= subjournalPageIfRequired(pPg
))
7344 PAGERTRACE(("MOVE %d page %d (needSync=%d) moves to %d\n",
7345 PAGERID(pPager
), pPg
->pgno
, (pPg
->flags
&PGHDR_NEED_SYNC
)?1:0, pgno
));
7346 IOTRACE(("MOVE %p %d %d\n", pPager
, pPg
->pgno
, pgno
))
7348 /* If the journal needs to be sync()ed before page pPg->pgno can
7349 ** be written to, store pPg->pgno in local variable needSyncPgno.
7351 ** If the isCommit flag is set, there is no need to remember that
7352 ** the journal needs to be sync()ed before database page pPg->pgno
7353 ** can be written to. The caller has already promised not to write to it.
7355 if( (pPg
->flags
&PGHDR_NEED_SYNC
) && !isCommit
){
7356 needSyncPgno
= pPg
->pgno
;
7357 assert( pPager
->journalMode
==PAGER_JOURNALMODE_OFF
||
7358 pageInJournal(pPager
, pPg
) || pPg
->pgno
>pPager
->dbOrigSize
);
7359 assert( pPg
->flags
&PGHDR_DIRTY
);
7362 /* If the cache contains a page with page-number pgno, remove it
7363 ** from its hash chain. Also, if the PGHDR_NEED_SYNC flag was set for
7364 ** page pgno before the 'move' operation, it needs to be retained
7365 ** for the page moved there.
7367 pPg
->flags
&= ~PGHDR_NEED_SYNC
;
7368 pPgOld
= sqlite3PagerLookup(pPager
, pgno
);
7369 assert( !pPgOld
|| pPgOld
->nRef
==1 || CORRUPT_DB
);
7371 if( NEVER(pPgOld
->nRef
>1) ){
7372 sqlite3PagerUnrefNotNull(pPgOld
);
7373 return SQLITE_CORRUPT_BKPT
;
7375 pPg
->flags
|= (pPgOld
->flags
&PGHDR_NEED_SYNC
);
7376 if( pPager
->tempFile
){
7377 /* Do not discard pages from an in-memory database since we might
7378 ** need to rollback later. Just move the page out of the way. */
7379 sqlite3PcacheMove(pPgOld
, pPager
->dbSize
+1);
7381 sqlite3PcacheDrop(pPgOld
);
7385 origPgno
= pPg
->pgno
;
7386 sqlite3PcacheMove(pPg
, pgno
);
7387 sqlite3PcacheMakeDirty(pPg
);
7389 /* For an in-memory database, make sure the original page continues
7390 ** to exist, in case the transaction needs to roll back. Use pPgOld
7391 ** as the original page since it has already been allocated.
7393 if( pPager
->tempFile
&& pPgOld
){
7394 sqlite3PcacheMove(pPgOld
, origPgno
);
7395 sqlite3PagerUnrefNotNull(pPgOld
);
7399 /* If needSyncPgno is non-zero, then the journal file needs to be
7400 ** sync()ed before any data is written to database file page needSyncPgno.
7401 ** Currently, no such page exists in the page-cache and the
7402 ** "is journaled" bitvec flag has been set. This needs to be remedied by
7403 ** loading the page into the pager-cache and setting the PGHDR_NEED_SYNC
7406 ** If the attempt to load the page into the page-cache fails, (due
7407 ** to a malloc() or IO failure), clear the bit in the pInJournal[]
7408 ** array. Otherwise, if the page is loaded and written again in
7409 ** this transaction, it may be written to the database file before
7410 ** it is synced into the journal file. This way, it may end up in
7411 ** the journal file twice, but that is not a problem.
7414 rc
= sqlite3PagerGet(pPager
, needSyncPgno
, &pPgHdr
, 0);
7415 if( rc
!=SQLITE_OK
){
7416 if( needSyncPgno
<=pPager
->dbOrigSize
){
7417 assert( pPager
->pTmpSpace
!=0 );
7418 sqlite3BitvecClear(pPager
->pInJournal
, needSyncPgno
, pPager
->pTmpSpace
);
7422 pPgHdr
->flags
|= PGHDR_NEED_SYNC
;
7423 sqlite3PcacheMakeDirty(pPgHdr
);
7424 sqlite3PagerUnrefNotNull(pPgHdr
);
7432 ** The page handle passed as the first argument refers to a dirty page
7433 ** with a page number other than iNew. This function changes the page's
7434 ** page number to iNew and sets the value of the PgHdr.flags field to
7435 ** the value passed as the third parameter.
7437 void sqlite3PagerRekey(DbPage
*pPg
, Pgno iNew
, u16 flags
){
7438 assert( pPg
->pgno
!=iNew
);
7440 sqlite3PcacheMove(pPg
, iNew
);
7444 ** Return a pointer to the data for the specified page.
7446 void *sqlite3PagerGetData(DbPage
*pPg
){
7447 assert( pPg
->nRef
>0 || pPg
->pPager
->memDb
);
7452 ** Return a pointer to the Pager.nExtra bytes of "extra" space
7453 ** allocated along with the specified page.
7455 void *sqlite3PagerGetExtra(DbPage
*pPg
){
7460 ** Get/set the locking-mode for this pager. Parameter eMode must be one
7461 ** of PAGER_LOCKINGMODE_QUERY, PAGER_LOCKINGMODE_NORMAL or
7462 ** PAGER_LOCKINGMODE_EXCLUSIVE. If the parameter is not _QUERY, then
7463 ** the locking-mode is set to the value specified.
7465 ** The returned value is either PAGER_LOCKINGMODE_NORMAL or
7466 ** PAGER_LOCKINGMODE_EXCLUSIVE, indicating the current (possibly updated)
7469 int sqlite3PagerLockingMode(Pager
*pPager
, int eMode
){
7470 assert( eMode
==PAGER_LOCKINGMODE_QUERY
7471 || eMode
==PAGER_LOCKINGMODE_NORMAL
7472 || eMode
==PAGER_LOCKINGMODE_EXCLUSIVE
);
7473 assert( PAGER_LOCKINGMODE_QUERY
<0 );
7474 assert( PAGER_LOCKINGMODE_NORMAL
>=0 && PAGER_LOCKINGMODE_EXCLUSIVE
>=0 );
7475 assert( pPager
->exclusiveMode
|| 0==sqlite3WalHeapMemory(pPager
->pWal
) );
7476 if( eMode
>=0 && !pPager
->tempFile
&& !sqlite3WalHeapMemory(pPager
->pWal
) ){
7477 pPager
->exclusiveMode
= (u8
)eMode
;
7479 return (int)pPager
->exclusiveMode
;
7483 ** Set the journal-mode for this pager. Parameter eMode must be one of:
7485 ** PAGER_JOURNALMODE_DELETE
7486 ** PAGER_JOURNALMODE_TRUNCATE
7487 ** PAGER_JOURNALMODE_PERSIST
7488 ** PAGER_JOURNALMODE_OFF
7489 ** PAGER_JOURNALMODE_MEMORY
7490 ** PAGER_JOURNALMODE_WAL
7492 ** The journalmode is set to the value specified if the change is allowed.
7493 ** The change may be disallowed for the following reasons:
7495 ** * An in-memory database can only have its journal_mode set to _OFF
7498 ** * Temporary databases cannot have _WAL journalmode.
7500 ** The returned indicate the current (possibly updated) journal-mode.
7502 int sqlite3PagerSetJournalMode(Pager
*pPager
, int eMode
){
7503 u8 eOld
= pPager
->journalMode
; /* Prior journalmode */
7505 /* The eMode parameter is always valid */
7506 assert( eMode
==PAGER_JOURNALMODE_DELETE
/* 0 */
7507 || eMode
==PAGER_JOURNALMODE_PERSIST
/* 1 */
7508 || eMode
==PAGER_JOURNALMODE_OFF
/* 2 */
7509 || eMode
==PAGER_JOURNALMODE_TRUNCATE
/* 3 */
7510 || eMode
==PAGER_JOURNALMODE_MEMORY
/* 4 */
7511 || eMode
==PAGER_JOURNALMODE_WAL
/* 5 */ );
7513 /* This routine is only called from the OP_JournalMode opcode, and
7514 ** the logic there will never allow a temporary file to be changed
7517 assert( pPager
->tempFile
==0 || eMode
!=PAGER_JOURNALMODE_WAL
);
7519 /* Do allow the journalmode of an in-memory database to be set to
7520 ** anything other than MEMORY or OFF
7523 assert( eOld
==PAGER_JOURNALMODE_MEMORY
|| eOld
==PAGER_JOURNALMODE_OFF
);
7524 if( eMode
!=PAGER_JOURNALMODE_MEMORY
&& eMode
!=PAGER_JOURNALMODE_OFF
){
7531 /* Change the journal mode. */
7532 assert( pPager
->eState
!=PAGER_ERROR
);
7533 pPager
->journalMode
= (u8
)eMode
;
7535 /* When transitioning from TRUNCATE or PERSIST to any other journal
7536 ** mode except WAL, unless the pager is in locking_mode=exclusive mode,
7537 ** delete the journal file.
7539 assert( (PAGER_JOURNALMODE_TRUNCATE
& 5)==1 );
7540 assert( (PAGER_JOURNALMODE_PERSIST
& 5)==1 );
7541 assert( (PAGER_JOURNALMODE_DELETE
& 5)==0 );
7542 assert( (PAGER_JOURNALMODE_MEMORY
& 5)==4 );
7543 assert( (PAGER_JOURNALMODE_OFF
& 5)==0 );
7544 assert( (PAGER_JOURNALMODE_WAL
& 5)==5 );
7546 assert( isOpen(pPager
->fd
) || pPager
->exclusiveMode
);
7547 if( !pPager
->exclusiveMode
&& (eOld
& 5)==1 && (eMode
& 1)==0 ){
7548 /* In this case we would like to delete the journal file. If it is
7549 ** not possible, then that is not a problem. Deleting the journal file
7550 ** here is an optimization only.
7552 ** Before deleting the journal file, obtain a RESERVED lock on the
7553 ** database file. This ensures that the journal file is not deleted
7554 ** while it is in use by some other client.
7556 sqlite3OsClose(pPager
->jfd
);
7557 if( pPager
->eLock
>=RESERVED_LOCK
){
7558 sqlite3OsDelete(pPager
->pVfs
, pPager
->zJournal
, 0);
7561 int state
= pPager
->eState
;
7562 assert( state
==PAGER_OPEN
|| state
==PAGER_READER
);
7563 if( state
==PAGER_OPEN
){
7564 rc
= sqlite3PagerSharedLock(pPager
);
7566 if( pPager
->eState
==PAGER_READER
){
7567 assert( rc
==SQLITE_OK
);
7568 rc
= pagerLockDb(pPager
, RESERVED_LOCK
);
7570 if( rc
==SQLITE_OK
){
7571 sqlite3OsDelete(pPager
->pVfs
, pPager
->zJournal
, 0);
7573 if( rc
==SQLITE_OK
&& state
==PAGER_READER
){
7574 pagerUnlockDb(pPager
, SHARED_LOCK
);
7575 }else if( state
==PAGER_OPEN
){
7576 pager_unlock(pPager
);
7578 assert( state
==pPager
->eState
);
7580 }else if( eMode
==PAGER_JOURNALMODE_OFF
|| eMode
==PAGER_JOURNALMODE_MEMORY
){
7581 sqlite3OsClose(pPager
->jfd
);
7585 /* Return the new journal mode */
7586 return (int)pPager
->journalMode
;
7590 ** Return the current journal mode.
7592 int sqlite3PagerGetJournalMode(Pager
*pPager
){
7593 return (int)pPager
->journalMode
;
7597 ** Return TRUE if the pager is in a state where it is OK to change the
7598 ** journalmode. Journalmode changes can only happen when the database
7601 int sqlite3PagerOkToChangeJournalMode(Pager
*pPager
){
7602 assert( assert_pager_state(pPager
) );
7603 if( pPager
->eState
>=PAGER_WRITER_CACHEMOD
) return 0;
7604 if( NEVER(isOpen(pPager
->jfd
) && pPager
->journalOff
>0) ) return 0;
7609 ** Get/set the size-limit used for persistent journal files.
7611 ** Setting the size limit to -1 means no limit is enforced.
7612 ** An attempt to set a limit smaller than -1 is a no-op.
7614 i64
sqlite3PagerJournalSizeLimit(Pager
*pPager
, i64 iLimit
){
7616 pPager
->journalSizeLimit
= iLimit
;
7617 sqlite3WalLimit(pPager
->pWal
, iLimit
);
7619 return pPager
->journalSizeLimit
;
7623 ** Return a pointer to the pPager->pBackup variable. The backup module
7624 ** in backup.c maintains the content of this variable. This module
7625 ** uses it opaquely as an argument to sqlite3BackupRestart() and
7626 ** sqlite3BackupUpdate() only.
7628 sqlite3_backup
**sqlite3PagerBackupPtr(Pager
*pPager
){
7629 return &pPager
->pBackup
;
7632 #ifndef SQLITE_OMIT_VACUUM
7634 ** Unless this is an in-memory or temporary database, clear the pager cache.
7636 void sqlite3PagerClearCache(Pager
*pPager
){
7637 assert( MEMDB
==0 || pPager
->tempFile
);
7638 if( pPager
->tempFile
==0 ) pager_reset(pPager
);
7643 #ifndef SQLITE_OMIT_WAL
7645 ** This function is called when the user invokes "PRAGMA wal_checkpoint",
7646 ** "PRAGMA wal_blocking_checkpoint" or calls the sqlite3_wal_checkpoint()
7647 ** or wal_blocking_checkpoint() API functions.
7649 ** Parameter eMode is one of SQLITE_CHECKPOINT_PASSIVE, FULL or RESTART.
7651 int sqlite3PagerCheckpoint(
7652 Pager
*pPager
, /* Checkpoint on this pager */
7653 sqlite3
*db
, /* Db handle used to check for interrupts */
7654 int eMode
, /* Type of checkpoint */
7655 int *pnLog
, /* OUT: Final number of frames in log */
7656 int *pnCkpt
/* OUT: Final number of checkpointed frames */
7659 if( pPager
->pWal
==0 && pPager
->journalMode
==PAGER_JOURNALMODE_WAL
){
7660 /* This only happens when a database file is zero bytes in size opened and
7661 ** then "PRAGMA journal_mode=WAL" is run and then sqlite3_wal_checkpoint()
7662 ** is invoked without any intervening transactions. We need to start
7663 ** a transaction to initialize pWal. The PRAGMA table_list statement is
7664 ** used for this since it starts transactions on every database file,
7665 ** including all ATTACHed databases. This seems expensive for a single
7666 ** sqlite3_wal_checkpoint() call, but it happens very rarely.
7667 ** https://sqlite.org/forum/forumpost/fd0f19d229156939
7669 sqlite3_exec(db
, "PRAGMA table_list",0,0,0);
7672 rc
= sqlite3WalCheckpoint(pPager
->pWal
, db
, eMode
,
7673 (eMode
==SQLITE_CHECKPOINT_PASSIVE
? 0 : pPager
->xBusyHandler
),
7674 pPager
->pBusyHandlerArg
,
7675 pPager
->walSyncFlags
, pPager
->pageSize
, (u8
*)pPager
->pTmpSpace
,
7682 int sqlite3PagerWalCallback(Pager
*pPager
){
7683 return sqlite3WalCallback(pPager
->pWal
);
7687 ** Return true if the underlying VFS for the given pager supports the
7688 ** primitives necessary for write-ahead logging.
7690 int sqlite3PagerWalSupported(Pager
*pPager
){
7691 const sqlite3_io_methods
*pMethods
= pPager
->fd
->pMethods
;
7692 if( pPager
->noLock
) return 0;
7693 return pPager
->exclusiveMode
|| (pMethods
->iVersion
>=2 && pMethods
->xShmMap
);
7697 ** Attempt to take an exclusive lock on the database file. If a PENDING lock
7698 ** is obtained instead, immediately release it.
7700 static int pagerExclusiveLock(Pager
*pPager
){
7701 int rc
; /* Return code */
7702 u8 eOrigLock
; /* Original lock */
7704 assert( pPager
->eLock
>=SHARED_LOCK
);
7705 eOrigLock
= pPager
->eLock
;
7706 rc
= pagerLockDb(pPager
, EXCLUSIVE_LOCK
);
7707 if( rc
!=SQLITE_OK
){
7708 /* If the attempt to grab the exclusive lock failed, release the
7709 ** pending lock that may have been obtained instead. */
7710 pagerUnlockDb(pPager
, eOrigLock
);
7717 ** Call sqlite3WalOpen() to open the WAL handle. If the pager is in
7718 ** exclusive-locking mode when this function is called, take an EXCLUSIVE
7719 ** lock on the database file and use heap-memory to store the wal-index
7720 ** in. Otherwise, use the normal shared-memory.
7722 static int pagerOpenWal(Pager
*pPager
){
7725 assert( pPager
->pWal
==0 && pPager
->tempFile
==0 );
7726 assert( pPager
->eLock
==SHARED_LOCK
|| pPager
->eLock
==EXCLUSIVE_LOCK
);
7728 /* If the pager is already in exclusive-mode, the WAL module will use
7729 ** heap-memory for the wal-index instead of the VFS shared-memory
7730 ** implementation. Take the exclusive lock now, before opening the WAL
7731 ** file, to make sure this is safe.
7733 if( pPager
->exclusiveMode
){
7734 rc
= pagerExclusiveLock(pPager
);
7737 /* Open the connection to the log file. If this operation fails,
7738 ** (e.g. due to malloc() failure), return an error code.
7740 if( rc
==SQLITE_OK
){
7741 rc
= sqlite3WalOpen(pPager
->pVfs
,
7742 pPager
->fd
, pPager
->zWal
, pPager
->exclusiveMode
,
7743 pPager
->journalSizeLimit
, &pPager
->pWal
7746 pagerFixMaplimit(pPager
);
7753 ** The caller must be holding a SHARED lock on the database file to call
7756 ** If the pager passed as the first argument is open on a real database
7757 ** file (not a temp file or an in-memory database), and the WAL file
7758 ** is not already open, make an attempt to open it now. If successful,
7759 ** return SQLITE_OK. If an error occurs or the VFS used by the pager does
7760 ** not support the xShmXXX() methods, return an error code. *pbOpen is
7761 ** not modified in either case.
7763 ** If the pager is open on a temp-file (or in-memory database), or if
7764 ** the WAL file is already open, set *pbOpen to 1 and return SQLITE_OK
7765 ** without doing anything.
7767 int sqlite3PagerOpenWal(
7768 Pager
*pPager
, /* Pager object */
7769 int *pbOpen
/* OUT: Set to true if call is a no-op */
7771 int rc
= SQLITE_OK
; /* Return code */
7773 assert( assert_pager_state(pPager
) );
7774 assert( pPager
->eState
==PAGER_OPEN
|| pbOpen
);
7775 assert( pPager
->eState
==PAGER_READER
|| !pbOpen
);
7776 assert( pbOpen
==0 || *pbOpen
==0 );
7777 assert( pbOpen
!=0 || (!pPager
->tempFile
&& !pPager
->pWal
) );
7779 if( !pPager
->tempFile
&& !pPager
->pWal
){
7780 if( !sqlite3PagerWalSupported(pPager
) ) return SQLITE_CANTOPEN
;
7782 /* Close any rollback journal previously open */
7783 sqlite3OsClose(pPager
->jfd
);
7785 rc
= pagerOpenWal(pPager
);
7786 if( rc
==SQLITE_OK
){
7787 pPager
->journalMode
= PAGER_JOURNALMODE_WAL
;
7788 pPager
->eState
= PAGER_OPEN
;
7798 ** This function is called to close the connection to the log file prior
7799 ** to switching from WAL to rollback mode.
7801 ** Before closing the log file, this function attempts to take an
7802 ** EXCLUSIVE lock on the database file. If this cannot be obtained, an
7803 ** error (SQLITE_BUSY) is returned and the log connection is not closed.
7804 ** If successful, the EXCLUSIVE lock is not released before returning.
7806 int sqlite3PagerCloseWal(Pager
*pPager
, sqlite3
*db
){
7809 assert( pPager
->journalMode
==PAGER_JOURNALMODE_WAL
);
7811 /* If the log file is not already open, but does exist in the file-system,
7812 ** it may need to be checkpointed before the connection can switch to
7813 ** rollback mode. Open it now so this can happen.
7815 if( !pPager
->pWal
){
7817 rc
= pagerLockDb(pPager
, SHARED_LOCK
);
7818 if( rc
==SQLITE_OK
){
7819 rc
= sqlite3OsAccess(
7820 pPager
->pVfs
, pPager
->zWal
, SQLITE_ACCESS_EXISTS
, &logexists
7823 if( rc
==SQLITE_OK
&& logexists
){
7824 rc
= pagerOpenWal(pPager
);
7828 /* Checkpoint and close the log. Because an EXCLUSIVE lock is held on
7829 ** the database file, the log and log-summary files will be deleted.
7831 if( rc
==SQLITE_OK
&& pPager
->pWal
){
7832 rc
= pagerExclusiveLock(pPager
);
7833 if( rc
==SQLITE_OK
){
7834 rc
= sqlite3WalClose(pPager
->pWal
, db
, pPager
->walSyncFlags
,
7835 pPager
->pageSize
, (u8
*)pPager
->pTmpSpace
);
7837 pagerFixMaplimit(pPager
);
7838 if( rc
&& !pPager
->exclusiveMode
) pagerUnlockDb(pPager
, SHARED_LOCK
);
7844 #ifdef SQLITE_ENABLE_SETLK_TIMEOUT
7846 ** If pager pPager is a wal-mode database not in exclusive locking mode,
7847 ** invoke the sqlite3WalWriteLock() function on the associated Wal object
7848 ** with the same db and bLock parameters as were passed to this function.
7849 ** Return an SQLite error code if an error occurs, or SQLITE_OK otherwise.
7851 int sqlite3PagerWalWriteLock(Pager
*pPager
, int bLock
){
7853 if( pagerUseWal(pPager
) && pPager
->exclusiveMode
==0 ){
7854 rc
= sqlite3WalWriteLock(pPager
->pWal
, bLock
);
7860 ** Set the database handle used by the wal layer to determine if
7861 ** blocking locks are required.
7863 void sqlite3PagerWalDb(Pager
*pPager
, sqlite3
*db
){
7864 if( pagerUseWal(pPager
) ){
7865 sqlite3WalDb(pPager
->pWal
, db
);
7870 #ifdef SQLITE_ENABLE_SNAPSHOT
7872 ** If this is a WAL database, obtain a snapshot handle for the snapshot
7873 ** currently open. Otherwise, return an error.
7875 int sqlite3PagerSnapshotGet(Pager
*pPager
, sqlite3_snapshot
**ppSnapshot
){
7876 int rc
= SQLITE_ERROR
;
7878 rc
= sqlite3WalSnapshotGet(pPager
->pWal
, ppSnapshot
);
7884 ** If this is a WAL database, store a pointer to pSnapshot. Next time a
7885 ** read transaction is opened, attempt to read from the snapshot it
7886 ** identifies. If this is not a WAL database, return an error.
7888 int sqlite3PagerSnapshotOpen(
7890 sqlite3_snapshot
*pSnapshot
7894 sqlite3WalSnapshotOpen(pPager
->pWal
, pSnapshot
);
7902 ** If this is a WAL database, call sqlite3WalSnapshotRecover(). If this
7903 ** is not a WAL database, return an error.
7905 int sqlite3PagerSnapshotRecover(Pager
*pPager
){
7908 rc
= sqlite3WalSnapshotRecover(pPager
->pWal
);
7916 ** The caller currently has a read transaction open on the database.
7917 ** If this is not a WAL database, SQLITE_ERROR is returned. Otherwise,
7918 ** this function takes a SHARED lock on the CHECKPOINTER slot and then
7919 ** checks if the snapshot passed as the second argument is still
7920 ** available. If so, SQLITE_OK is returned.
7922 ** If the snapshot is not available, SQLITE_ERROR is returned. Or, if
7923 ** the CHECKPOINTER lock cannot be obtained, SQLITE_BUSY. If any error
7924 ** occurs (any value other than SQLITE_OK is returned), the CHECKPOINTER
7925 ** lock is released before returning.
7927 int sqlite3PagerSnapshotCheck(Pager
*pPager
, sqlite3_snapshot
*pSnapshot
){
7930 rc
= sqlite3WalSnapshotCheck(pPager
->pWal
, pSnapshot
);
7938 ** Release a lock obtained by an earlier successful call to
7939 ** sqlite3PagerSnapshotCheck().
7941 void sqlite3PagerSnapshotUnlock(Pager
*pPager
){
7942 assert( pPager
->pWal
);
7943 sqlite3WalSnapshotUnlock(pPager
->pWal
);
7946 #endif /* SQLITE_ENABLE_SNAPSHOT */
7947 #endif /* !SQLITE_OMIT_WAL */
7949 #ifdef SQLITE_ENABLE_ZIPVFS
7951 ** A read-lock must be held on the pager when this function is called. If
7952 ** the pager is in WAL mode and the WAL file currently contains one or more
7953 ** frames, return the size in bytes of the page images stored within the
7954 ** WAL frames. Otherwise, if this is not a WAL database or the WAL file
7955 ** is empty, return 0.
7957 int sqlite3PagerWalFramesize(Pager
*pPager
){
7958 assert( pPager
->eState
>=PAGER_READER
);
7959 return sqlite3WalFramesize(pPager
->pWal
);
7963 #ifdef SQLITE_USE_SEH
7964 int sqlite3PagerWalSystemErrno(Pager
*pPager
){
7965 return sqlite3WalSystemErrno(pPager
->pWal
);
7969 #endif /* SQLITE_OMIT_DISKIO */
7971 /* BEGIN SQLCIPHER */
7972 #ifdef SQLITE_HAS_CODEC
7974 int sqlite3pager_is_sj_pgno(Pager
*pPager
, Pgno pgno
) {
7975 return (PAGER_SJ_PGNO(pPager
) == pgno
) ? 1 : 0;
7978 void sqlite3pager_error(Pager
*pPager
, int error
) {
7979 pPager
->errCode
= error
;
7980 pPager
->eState
= PAGER_ERROR
;
7981 setGetterMethod(pPager
);
7984 void sqlite3pager_reset(Pager
*pPager
){
7985 pager_reset(pPager
);