1/*
2** 2001 September 15
3**
4** The author disclaims copyright to this source code. In place of
5** a legal notice, here is a blessing:
6**
7** May you do good and not evil.
8** May you find forgiveness for yourself and forgive others.
9** May you share freely, never taking more than you give.
10**
11*************************************************************************
12** This is the implementation of the page cache subsystem or "pager".
13**
14** The pager is used to access a database disk file. It implements
15** atomic commit and rollback through the use of a journal file that
16** is separate from the database file. The pager also implements file
17** locking to prevent two processes from writing the same database
18** file simultaneously, or one process from reading the database while
19** another is writing.
20*/
21#ifndef SQLITE_OMIT_DISKIO
22#include "sqliteInt.h"
23#include "wal.h"
24
25
26/******************* NOTES ON THE DESIGN OF THE PAGER ************************
27**
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.
31**
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.
36**
37** Definition: A page of the database file is said to be "overwriteable" if
38** one or more of the following are true about the page:
39**
40** (a) The original content of the page as it was at the beginning of
41** the transaction has been written into the rollback journal and
42** synced.
43**
44** (b) The page was a freelist leaf page at the start of the transaction.
45**
46** (c) The page number is greater than the largest page that existed in
47** the database file at the start of the transaction.
48**
49** (1) A page of the database file is never overwritten unless one of the
50** following are true:
51**
52** (a) The page and all other pages on the same sector are overwriteable.
53**
54** (b) The atomic page write optimization is enabled, and the entire
55** transaction other than the update of the transaction sequence
56** number consists of a single page change.
57**
58** (2) The content of a page written into the rollback journal exactly matches
59** both the content in the database when the rollback journal was written
60** and the content in the database at the beginning of the current
61** transaction.
62**
63** (3) Writes to the database file are an integer multiple of the page size
64** in length and are aligned on a page boundary.
65**
66** (4) Reads from the database file are either aligned on a page boundary and
67** an integer multiple of the page size in length or are taken from the
68** first 100 bytes of the database file.
69**
70** (5) All writes to the database file are synced prior to the rollback journal
71** being deleted, truncated, or zeroed.
72**
73** (6) If a super-journal file is used, then all writes to the database file
74** are synced prior to the super-journal being deleted.
75**
76** Definition: Two databases (or the same database at two points it time)
77** are said to be "logically equivalent" if they give the same answer to
78** all queries. Note in particular the content of freelist leaf
79** pages can be changed arbitrarily without affecting the logical equivalence
80** of the database.
81**
82** (7) At any time, if any subset, including the empty set and the total set,
83** of the unsynced changes to a rollback journal are removed and the
84** journal is rolled back, the resulting database file will be logically
85** equivalent to the database file at the beginning of the transaction.
86**
87** (8) When a transaction is rolled back, the xTruncate method of the VFS
88** is called to restore the database file to the same size it was at
89** the beginning of the transaction. (In some VFSes, the xTruncate
90** method is a no-op, but that does not change the fact the SQLite will
91** invoke it.)
92**
93** (9) Whenever the database file is modified, at least one bit in the range
94** of bytes from 24 through 39 inclusive will be changed prior to releasing
95** the EXCLUSIVE lock, thus signaling other connections on the same
96** database to flush their caches.
97**
98** (10) The pattern of bits in bytes 24 through 39 shall not repeat in less
99** than one billion transactions.
100**
101** (11) A database file is well-formed at the beginning and at the conclusion
102** of every transaction.
103**
104** (12) An EXCLUSIVE lock is held on the database file when writing to
105** the database file.
106**
107** (13) A SHARED lock is held on the database file while reading any
108** content out of the database file.
109**
110******************************************************************************/
111
112/*
113** Macros for troubleshooting. Normally turned off
114*/
115#if 0
116int sqlite3PagerTrace=1; /* True to enable tracing */
117#define sqlite3DebugPrintf printf
118#define PAGERTRACE(X) if( sqlite3PagerTrace ){ sqlite3DebugPrintf X; }
119#else
120#define PAGERTRACE(X)
121#endif
122
123/*
124** The following two macros are used within the PAGERTRACE() macros above
125** to print out file-descriptors.
126**
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.
130*/
131#define PAGERID(p) (SQLITE_PTR_TO_INT(p->fd))
132#define FILEHANDLEID(fd) (SQLITE_PTR_TO_INT(fd))
133
134/*
135** The Pager.eState variable stores the current 'state' of a pager. A
136** pager may be in any one of the seven states shown in the following
137** state diagram.
138**
139** OPEN <------+------+
140** | | |
141** V | |
142** +---------> READER-------+ |
143** | | |
144** | V |
145** |<-------WRITER_LOCKED------> ERROR
146** | | ^
147** | V |
148** |<------WRITER_CACHEMOD-------->|
149** | | |
150** | V |
151** |<-------WRITER_DBMOD---------->|
152** | | |
153** | V |
154** +<------WRITER_FINISHED-------->+
155**
156**
157** List of state transitions and the C [function] that performs each:
158**
159** OPEN -> READER [sqlite3PagerSharedLock]
160** READER -> OPEN [pager_unlock]
161**
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]
167**
168** WRITER_*** -> ERROR [pager_error]
169** ERROR -> OPEN [pager_unlock]
170**
171**
172** OPEN:
173**
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.
177**
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.
181**
182** READER:
183**
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.
188**
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).
196**
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.
205**
206** WRITER_LOCKED:
207**
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.
212**
213** In rollback mode, a RESERVED or (if the transaction was opened with
214** BEGIN EXCLUSIVE) EXCLUSIVE lock is obtained on the database file when
215** moving to this state, but the journal file is not written to or opened
216** to in this state. If the transaction is committed or rolled back while
217** in WRITER_LOCKED state, all that is required is to unlock the database
218** file.
219**
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.
223**
224** * A write transaction is active.
225** * If the connection is open in rollback-mode, a RESERVED or greater
226** lock is held on the database file.
227** * If the connection is open in WAL-mode, a WAL write transaction
228** is open (i.e. sqlite3WalBeginWriteTransaction() has been successfully
229** called).
230** * The dbSize, dbOrigSize and dbFileSize variables are all valid.
231** * The contents of the pager cache have not been modified.
232** * The journal file may or may not be open.
233** * Nothing (not even the first header) has been written to the journal.
234**
235** WRITER_CACHEMOD:
236**
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.
241**
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.
247**
248** WRITER_DBMOD:
249**
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).
254**
255** * A write transaction is active.
256** * An EXCLUSIVE or greater lock is held on the database file.
257** * The journal file is open and the first header has been written
258** and synced to disk.
259** * The contents of the page cache have been modified (and possibly
260** written to disk).
261**
262** WRITER_FINISHED:
263**
264** It is not possible for a WAL connection to enter this state.
265**
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.
272**
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.
279**
280** ERROR:
281**
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.
286**
287** Temporary pager files may enter the ERROR state, but in-memory pagers
288** cannot.
289**
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.
298**
299** Once it has entered the ERROR state, any attempt to use the pager
300** to read or write data returns an error. Eventually, once all
301** outstanding transactions have been abandoned, the pager is able to
302** transition back to OPEN state, discarding the contents of the
303** page-cache and any other in-memory state at the same time. Everything
304** is reloaded from disk (and, if necessary, hot-journal rollback peformed)
305** when a read-transaction is next opened on the pager (transitioning
306** the pager into READER state). At that point the system has recovered
307** from the error.
308**
309** Specifically, the pager jumps into the ERROR state if:
310**
311** 1. An error occurs while attempting a rollback. This happens in
312** function sqlite3PagerRollback().
313**
314** 2. An error occurs while attempting to finalize a journal file
315** following a commit in function sqlite3PagerCommitPhaseTwo().
316**
317** 3. An error occurs while attempting to write to the journal or
318** database file in function pagerStress() in order to free up
319** memory.
320**
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.
324**
325** Condition (3) is necessary because it can be triggered by a read-only
326** statement executed within a transaction. In this case, if the error
327** code were simply returned to the user, the b-tree layer would not
328** automatically attempt a rollback, as it assumes that an error in a
329** read-only statement cannot leave the pager in an internally inconsistent
330** state.
331**
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.
336**
337**
338** Notes:
339**
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.
343**
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".
348**
349** * See also: assert_pager_state().
350*/
351#define PAGER_OPEN 0
352#define PAGER_READER 1
353#define PAGER_WRITER_LOCKED 2
354#define PAGER_WRITER_CACHEMOD 3
355#define PAGER_WRITER_DBMOD 4
356#define PAGER_WRITER_FINISHED 5
357#define PAGER_ERROR 6
358
359/*
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.
365**
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.
374**
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.
378**
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.
386**
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.
394**
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.
403**
404** Pager.eLock may only be set to UNKNOWN_LOCK when the pager is in
405** PAGER_OPEN state.
406*/
407#define UNKNOWN_LOCK (EXCLUSIVE_LOCK+1)
408
409/*
410** The maximum allowed sector size. 64KiB. If the xSectorsize() method
411** returns a value larger than this, then MAX_SECTOR_SIZE is used instead.
412** This could conceivably cause corruption following a power failure on
413** such a system. This is currently an undocumented limit.
414*/
415#define MAX_SECTOR_SIZE 0x10000
416
417
418/*
419** An instance of the following structure is allocated for each active
420** savepoint and statement transaction in the system. All such structures
421** are stored in the Pager.aSavepoint[] array, which is allocated and
422** resized using sqlite3Realloc().
423**
424** When a savepoint is created, the PagerSavepoint.iHdrOffset field is
425** set to 0. If a journal-header is written into the main journal while
426** the savepoint is active, then iHdrOffset is set to the byte offset
427** immediately following the last journal record written into the main
428** journal before the journal-header. This is required during savepoint
429** rollback (see pagerPlaybackSavepoint()).
430*/
431typedef struct PagerSavepoint PagerSavepoint;
432struct PagerSavepoint {
433 i64 iOffset; /* Starting offset in main journal */
434 i64 iHdrOffset; /* See above */
435 Bitvec *pInSavepoint; /* Set of pages in this savepoint */
436 Pgno nOrig; /* Original number of pages in file */
437 Pgno iSubRec; /* Index of first record in sub-journal */
438 int bTruncateOnRelease; /* If stmt journal may be truncated on RELEASE */
439#ifndef SQLITE_OMIT_WAL
440 u32 aWalData[WAL_SAVEPOINT_NDATA]; /* WAL savepoint context */
441#endif
442};
443
444/*
445** Bits of the Pager.doNotSpill flag. See further description below.
446*/
447#define SPILLFLAG_OFF 0x01 /* Never spill cache. Set via pragma */
448#define SPILLFLAG_ROLLBACK 0x02 /* Current rolling back, so do not spill */
449#define SPILLFLAG_NOSYNC 0x04 /* Spill is ok, but do not sync */
450
451/*
452** An open page cache is an instance of struct Pager. A description of
453** some of the more important member variables follows:
454**
455** eState
456**
457** The current 'state' of the pager object. See the comment and state
458** diagram above for a description of the pager state.
459**
460** eLock
461**
462** For a real on-disk database, the current lock held on the database file -
463** NO_LOCK, SHARED_LOCK, RESERVED_LOCK or EXCLUSIVE_LOCK.
464**
465** For a temporary or in-memory database (neither of which require any
466** locks), this variable is always set to EXCLUSIVE_LOCK. Since such
467** databases always have Pager.exclusiveMode==1, this tricks the pager
468** logic into thinking that it already has all the locks it will ever
469** need (and no reason to release them).
470**
471** In some (obscure) circumstances, this variable may also be set to
472** UNKNOWN_LOCK. See the comment above the #define of UNKNOWN_LOCK for
473** details.
474**
475** changeCountDone
476**
477** This boolean variable is used to make sure that the change-counter
478** (the 4-byte header field at byte offset 24 of the database file) is
479** not updated more often than necessary.
480**
481** It is set to true when the change-counter field is updated, which
482** can only happen if an exclusive lock is held on the database file.
483** It is cleared (set to false) whenever an exclusive lock is
484** relinquished on the database file. Each time a transaction is committed,
485** The changeCountDone flag is inspected. If it is true, the work of
486** updating the change-counter is omitted for the current transaction.
487**
488** This mechanism means that when running in exclusive mode, a connection
489** need only update the change-counter once, for the first transaction
490** committed.
491**
492** setSuper
493**
494** When PagerCommitPhaseOne() is called to commit a transaction, it may
495** (or may not) specify a super-journal name to be written into the
496** journal file before it is synced to disk.
497**
498** Whether or not a journal file contains a super-journal pointer affects
499** the way in which the journal file is finalized after the transaction is
500** committed or rolled back when running in "journal_mode=PERSIST" mode.
501** If a journal file does not contain a super-journal pointer, it is
502** finalized by overwriting the first journal header with zeroes. If
503** it does contain a super-journal pointer the journal file is finalized
504** by truncating it to zero bytes, just as if the connection were
505** running in "journal_mode=truncate" mode.
506**
507** Journal files that contain super-journal pointers cannot be finalized
508** simply by overwriting the first journal-header with zeroes, as the
509** super-journal pointer could interfere with hot-journal rollback of any
510** subsequently interrupted transaction that reuses the journal file.
511**
512** The flag is cleared as soon as the journal file is finalized (either
513** by PagerCommitPhaseTwo or PagerRollback). If an IO error prevents the
514** journal file from being successfully finalized, the setSuper flag
515** is cleared anyway (and the pager will move to ERROR state).
516**
517** doNotSpill
518**
519** This variables control the behavior of cache-spills (calls made by
520** the pcache module to the pagerStress() routine to write cached data
521** to the file-system in order to free up memory).
522**
523** When bits SPILLFLAG_OFF or SPILLFLAG_ROLLBACK of doNotSpill are set,
524** writing to the database from pagerStress() is disabled altogether.
525** The SPILLFLAG_ROLLBACK case is done in a very obscure case that
526** comes up during savepoint rollback that requires the pcache module
527** to allocate a new page to prevent the journal file from being written
528** while it is being traversed by code in pager_playback(). The SPILLFLAG_OFF
529** case is a user preference.
530**
531** If the SPILLFLAG_NOSYNC bit is set, writing to the database from
532** pagerStress() is permitted, but syncing the journal file is not.
533** This flag is set by sqlite3PagerWrite() when the file-system sector-size
534** is larger than the database page-size in order to prevent a journal sync
535** from happening in between the journalling of two pages on the same sector.
536**
537** subjInMemory
538**
539** This is a boolean variable. If true, then any required sub-journal
540** is opened as an in-memory journal file. If false, then in-memory
541** sub-journals are only used for in-memory pager files.
542**
543** This variable is updated by the upper layer each time a new
544** write-transaction is opened.
545**
546** dbSize, dbOrigSize, dbFileSize
547**
548** Variable dbSize is set to the number of pages in the database file.
549** It is valid in PAGER_READER and higher states (all states except for
550** OPEN and ERROR).
551**
552** dbSize is set based on the size of the database file, which may be
553** larger than the size of the database (the value stored at offset
554** 28 of the database header by the btree). If the size of the file
555** is not an integer multiple of the page-size, the value stored in
556** dbSize is rounded down (i.e. a 5KB file with 2K page-size has dbSize==2).
557** Except, any file that is greater than 0 bytes in size is considered
558** to have at least one page. (i.e. a 1KB file with 2K page-size leads
559** to dbSize==1).
560**
561** During a write-transaction, if pages with page-numbers greater than
562** dbSize are modified in the cache, dbSize is updated accordingly.
563** Similarly, if the database is truncated using PagerTruncateImage(),
564** dbSize is updated.
565**
566** Variables dbOrigSize and dbFileSize are valid in states
567** PAGER_WRITER_LOCKED and higher. dbOrigSize is a copy of the dbSize
568** variable at the start of the transaction. It is used during rollback,
569** and to determine whether or not pages need to be journalled before
570** being modified.
571**
572** Throughout a write-transaction, dbFileSize contains the size of
573** the file on disk in pages. It is set to a copy of dbSize when the
574** write-transaction is first opened, and updated when VFS calls are made
575** to write or truncate the database file on disk.
576**
577** The only reason the dbFileSize variable is required is to suppress
578** unnecessary calls to xTruncate() after committing a transaction. If,
579** when a transaction is committed, the dbFileSize variable indicates
580** that the database file is larger than the database image (Pager.dbSize),
581** pager_truncate() is called. The pager_truncate() call uses xFilesize()
582** to measure the database file on disk, and then truncates it if required.
583** dbFileSize is not used when rolling back a transaction. In this case
584** pager_truncate() is called unconditionally (which means there may be
585** a call to xFilesize() that is not strictly required). In either case,
586** pager_truncate() may cause the file to become smaller or larger.
587**
588** dbHintSize
589**
590** The dbHintSize variable is used to limit the number of calls made to
591** the VFS xFileControl(FCNTL_SIZE_HINT) method.
592**
593** dbHintSize is set to a copy of the dbSize variable when a
594** write-transaction is opened (at the same time as dbFileSize and
595** dbOrigSize). If the xFileControl(FCNTL_SIZE_HINT) method is called,
596** dbHintSize is increased to the number of pages that correspond to the
597** size-hint passed to the method call. See pager_write_pagelist() for
598** details.
599**
600** errCode
601**
602** The Pager.errCode variable is only ever used in PAGER_ERROR state. It
603** is set to zero in all other states. In PAGER_ERROR state, Pager.errCode
604** is always set to SQLITE_FULL, SQLITE_IOERR or one of the SQLITE_IOERR_XXX
605** sub-codes.
606**
607** syncFlags, walSyncFlags
608**
609** syncFlags is either SQLITE_SYNC_NORMAL (0x02) or SQLITE_SYNC_FULL (0x03).
610** syncFlags is used for rollback mode. walSyncFlags is used for WAL mode
611** and contains the flags used to sync the checkpoint operations in the
612** lower two bits, and sync flags used for transaction commits in the WAL
613** file in bits 0x04 and 0x08. In other words, to get the correct sync flags
614** for checkpoint operations, use (walSyncFlags&0x03) and to get the correct
615** sync flags for transaction commit, use ((walSyncFlags>>2)&0x03). Note
616** that with synchronous=NORMAL in WAL mode, transaction commit is not synced
617** meaning that the 0x04 and 0x08 bits are both zero.
618*/
619struct Pager {
620 sqlite3_vfs *pVfs; /* OS functions to use for IO */
621 u8 exclusiveMode; /* Boolean. True if locking_mode==EXCLUSIVE */
622 u8 journalMode; /* One of the PAGER_JOURNALMODE_* values */
623 u8 useJournal; /* Use a rollback journal on this file */
624 u8 noSync; /* Do not sync the journal if true */
625 u8 fullSync; /* Do extra syncs of the journal for robustness */
626 u8 extraSync; /* sync directory after journal delete */
627 u8 syncFlags; /* SYNC_NORMAL or SYNC_FULL otherwise */
628 u8 walSyncFlags; /* See description above */
629 u8 tempFile; /* zFilename is a temporary or immutable file */
630 u8 noLock; /* Do not lock (except in WAL mode) */
631 u8 readOnly; /* True for a read-only database */
632 u8 memDb; /* True to inhibit all file I/O */
633 u8 memVfs; /* VFS-implemented memory database */
634
635 /**************************************************************************
636 ** The following block contains those class members that change during
637 ** routine operation. Class members not in this block are either fixed
638 ** when the pager is first created or else only change when there is a
639 ** significant mode change (such as changing the page_size, locking_mode,
640 ** or the journal_mode). From another view, these class members describe
641 ** the "state" of the pager, while other class members describe the
642 ** "configuration" of the pager.
643 */
644 u8 eState; /* Pager state (OPEN, READER, WRITER_LOCKED..) */
645 u8 eLock; /* Current lock held on database file */
646 u8 changeCountDone; /* Set after incrementing the change-counter */
647 u8 setSuper; /* Super-jrnl name is written into jrnl */
648 u8 doNotSpill; /* Do not spill the cache when non-zero */
649 u8 subjInMemory; /* True to use in-memory sub-journals */
650 u8 bUseFetch; /* True to use xFetch() */
651 u8 hasHeldSharedLock; /* True if a shared lock has ever been held */
652 Pgno dbSize; /* Number of pages in the database */
653 Pgno dbOrigSize; /* dbSize before the current transaction */
654 Pgno dbFileSize; /* Number of pages in the database file */
655 Pgno dbHintSize; /* Value passed to FCNTL_SIZE_HINT call */
656 int errCode; /* One of several kinds of errors */
657 int nRec; /* Pages journalled since last j-header written */
658 u32 cksumInit; /* Quasi-random value added to every checksum */
659 u32 nSubRec; /* Number of records written to sub-journal */
660 Bitvec *pInJournal; /* One bit for each page in the database file */
661 sqlite3_file *fd; /* File descriptor for database */
662 sqlite3_file *jfd; /* File descriptor for main journal */
663 sqlite3_file *sjfd; /* File descriptor for sub-journal */
664 i64 journalOff; /* Current write offset in the journal file */
665 i64 journalHdr; /* Byte offset to previous journal header */
666 sqlite3_backup *pBackup; /* Pointer to list of ongoing backup processes */
667 PagerSavepoint *aSavepoint; /* Array of active savepoints */
668 int nSavepoint; /* Number of elements in aSavepoint[] */
669 u32 iDataVersion; /* Changes whenever database content changes */
670 char dbFileVers[16]; /* Changes whenever database file changes */
671
672 int nMmapOut; /* Number of mmap pages currently outstanding */
673 sqlite3_int64 szMmap; /* Desired maximum mmap size */
674 PgHdr *pMmapFreelist; /* List of free mmap page headers (pDirty) */
675 /*
676 ** End of the routinely-changing class members
677 ***************************************************************************/
678
679 u16 nExtra; /* Add this many bytes to each in-memory page */
680 i16 nReserve; /* Number of unused bytes at end of each page */
681 u32 vfsFlags; /* Flags for sqlite3_vfs.xOpen() */
682 u32 sectorSize; /* Assumed sector size during rollback */
683 Pgno mxPgno; /* Maximum allowed size of the database */
684 Pgno lckPgno; /* Page number for the locking page */
685 i64 pageSize; /* Number of bytes in a page */
686 i64 journalSizeLimit; /* Size limit for persistent journal files */
687 char *zFilename; /* Name of the database file */
688 char *zJournal; /* Name of the journal file */
689 int (*xBusyHandler)(void*); /* Function to call when busy */
690 void *pBusyHandlerArg; /* Context argument for xBusyHandler */
691 int aStat[4]; /* Total cache hits, misses, writes, spills */
692#ifdef SQLITE_TEST
693 int nRead; /* Database pages read */
694#endif
695 void (*xReiniter)(DbPage*); /* Call this routine when reloading pages */
696 int (*xGet)(Pager*,Pgno,DbPage**,int); /* Routine to fetch a patch */
697 char *pTmpSpace; /* Pager.pageSize bytes of space for tmp use */
698 PCache *pPCache; /* Pointer to page cache object */
699#ifndef SQLITE_OMIT_WAL
700 Wal *pWal; /* Write-ahead log used by "journal_mode=wal" */
701 char *zWal; /* File name for write-ahead log */
702#endif
703};
704
705/*
706** Indexes for use with Pager.aStat[]. The Pager.aStat[] array contains
707** the values accessed by passing SQLITE_DBSTATUS_CACHE_HIT, CACHE_MISS
708** or CACHE_WRITE to sqlite3_db_status().
709*/
710#define PAGER_STAT_HIT 0
711#define PAGER_STAT_MISS 1
712#define PAGER_STAT_WRITE 2
713#define PAGER_STAT_SPILL 3
714
715/*
716** The following global variables hold counters used for
717** testing purposes only. These variables do not exist in
718** a non-testing build. These variables are not thread-safe.
719*/
720#ifdef SQLITE_TEST
721int sqlite3_pager_readdb_count = 0; /* Number of full pages read from DB */
722int sqlite3_pager_writedb_count = 0; /* Number of full pages written to DB */
723int sqlite3_pager_writej_count = 0; /* Number of pages written to journal */
724# define PAGER_INCR(v) v++
725#else
726# define PAGER_INCR(v)
727#endif
728
729
730
731/*
732** Journal files begin with the following magic string. The data
733** was obtained from /dev/random. It is used only as a sanity check.
734**
735** Since version 2.8.0, the journal format contains additional sanity
736** checking information. If the power fails while the journal is being
737** written, semi-random garbage data might appear in the journal
738** file after power is restored. If an attempt is then made
739** to roll the journal back, the database could be corrupted. The additional
740** sanity checking data is an attempt to discover the garbage in the
741** journal and ignore it.
742**
743** The sanity checking information for the new journal format consists
744** of a 32-bit checksum on each page of data. The checksum covers both
745** the page number and the pPager->pageSize bytes of data for the page.
746** This cksum is initialized to a 32-bit random value that appears in the
747** journal file right after the header. The random initializer is important,
748** because garbage data that appears at the end of a journal is likely
749** data that was once in other files that have now been deleted. If the
750** garbage data came from an obsolete journal file, the checksums might
751** be correct. But by initializing the checksum to random value which
752** is different for every journal, we minimize that risk.
753*/
754static const unsigned char aJournalMagic[] = {
755 0xd9, 0xd5, 0x05, 0xf9, 0x20, 0xa1, 0x63, 0xd7,
756};
757
758/*
759** The size of the of each page record in the journal is given by
760** the following macro.
761*/
762#define JOURNAL_PG_SZ(pPager) ((pPager->pageSize) + 8)
763
764/*
765** The journal header size for this pager. This is usually the same
766** size as a single disk sector. See also setSectorSize().
767*/
768#define JOURNAL_HDR_SZ(pPager) (pPager->sectorSize)
769
770/*
771** The macro MEMDB is true if we are dealing with an in-memory database.
772** We do this as a macro so that if the SQLITE_OMIT_MEMORYDB macro is set,
773** the value of MEMDB will be a constant and the compiler will optimize
774** out code that would never execute.
775*/
776#ifdef SQLITE_OMIT_MEMORYDB
777# define MEMDB 0
778#else
779# define MEMDB pPager->memDb
780#endif
781
782/*
783** The macro USEFETCH is true if we are allowed to use the xFetch and xUnfetch
784** interfaces to access the database using memory-mapped I/O.
785*/
786#if SQLITE_MAX_MMAP_SIZE>0
787# define USEFETCH(x) ((x)->bUseFetch)
788#else
789# define USEFETCH(x) 0
790#endif
791
792/*
793** The argument to this macro is a file descriptor (type sqlite3_file*).
794** Return 0 if it is not open, or non-zero (but not 1) if it is.
795**
796** This is so that expressions can be written as:
797**
798** if( isOpen(pPager->jfd) ){ ...
799**
800** instead of
801**
802** if( pPager->jfd->pMethods ){ ...
803*/
804#define isOpen(pFd) ((pFd)->pMethods!=0)
805
806#ifdef SQLITE_DIRECT_OVERFLOW_READ
807/*
808** Return true if page pgno can be read directly from the database file
809** by the b-tree layer. This is the case if:
810**
811** * the database file is open,
812** * there are no dirty pages in the cache, and
813** * the desired page is not currently in the wal file.
814*/
815int sqlite3PagerDirectReadOk(Pager *pPager, Pgno pgno){
816 if( pPager->fd->pMethods==0 ) return 0;
817 if( sqlite3PCacheIsDirty(pPager->pPCache) ) return 0;
818#ifndef SQLITE_OMIT_WAL
819 if( pPager->pWal ){
820 u32 iRead = 0;
821 int rc;
822 rc = sqlite3WalFindFrame(pPager->pWal, pgno, &iRead);
823 return (rc==SQLITE_OK && iRead==0);
824 }
825#endif
826 return 1;
827}
828#endif
829
830#ifndef SQLITE_OMIT_WAL
831# define pagerUseWal(x) ((x)->pWal!=0)
832#else
833# define pagerUseWal(x) 0
834# define pagerRollbackWal(x) 0
835# define pagerWalFrames(v,w,x,y) 0
836# define pagerOpenWalIfPresent(z) SQLITE_OK
837# define pagerBeginReadTransaction(z) SQLITE_OK
838#endif
839
840#ifndef NDEBUG
841/*
842** Usage:
843**
844** assert( assert_pager_state(pPager) );
845**
846** This function runs many asserts to try to find inconsistencies in
847** the internal state of the Pager object.
848*/
849static int assert_pager_state(Pager *p){
850 Pager *pPager = p;
851
852 /* State must be valid. */
853 assert( p->eState==PAGER_OPEN
854 || p->eState==PAGER_READER
855 || p->eState==PAGER_WRITER_LOCKED
856 || p->eState==PAGER_WRITER_CACHEMOD
857 || p->eState==PAGER_WRITER_DBMOD
858 || p->eState==PAGER_WRITER_FINISHED
859 || p->eState==PAGER_ERROR
860 );
861
862 /* Regardless of the current state, a temp-file connection always behaves
863 ** as if it has an exclusive lock on the database file. It never updates
864 ** the change-counter field, so the changeCountDone flag is always set.
865 */
866 assert( p->tempFile==0 || p->eLock==EXCLUSIVE_LOCK );
867 assert( p->tempFile==0 || pPager->changeCountDone );
868
869 /* If the useJournal flag is clear, the journal-mode must be "OFF".
870 ** And if the journal-mode is "OFF", the journal file must not be open.
871 */
872 assert( p->journalMode==PAGER_JOURNALMODE_OFF || p->useJournal );
873 assert( p->journalMode!=PAGER_JOURNALMODE_OFF || !isOpen(p->jfd) );
874
875 /* Check that MEMDB implies noSync. And an in-memory journal. Since
876 ** this means an in-memory pager performs no IO at all, it cannot encounter
877 ** either SQLITE_IOERR or SQLITE_FULL during rollback or while finalizing
878 ** a journal file. (although the in-memory journal implementation may
879 ** return SQLITE_IOERR_NOMEM while the journal file is being written). It
880 ** is therefore not possible for an in-memory pager to enter the ERROR
881 ** state.
882 */
883 if( MEMDB ){
884 assert( !isOpen(p->fd) );
885 assert( p->noSync );
886 assert( p->journalMode==PAGER_JOURNALMODE_OFF
887 || p->journalMode==PAGER_JOURNALMODE_MEMORY
888 );
889 assert( p->eState!=PAGER_ERROR && p->eState!=PAGER_OPEN );
890 assert( pagerUseWal(p)==0 );
891 }
892
893 /* If changeCountDone is set, a RESERVED lock or greater must be held
894 ** on the file.
895 */
896 assert( pPager->changeCountDone==0 || pPager->eLock>=RESERVED_LOCK );
897 assert( p->eLock!=PENDING_LOCK );
898
899 switch( p->eState ){
900 case PAGER_OPEN:
901 assert( !MEMDB );
902 assert( pPager->errCode==SQLITE_OK );
903 assert( sqlite3PcacheRefCount(pPager->pPCache)==0 || pPager->tempFile );
904 break;
905
906 case PAGER_READER:
907 assert( pPager->errCode==SQLITE_OK );
908 assert( p->eLock!=UNKNOWN_LOCK );
909 assert( p->eLock>=SHARED_LOCK );
910 break;
911
912 case PAGER_WRITER_LOCKED:
913 assert( p->eLock!=UNKNOWN_LOCK );
914 assert( pPager->errCode==SQLITE_OK );
915 if( !pagerUseWal(pPager) ){
916 assert( p->eLock>=RESERVED_LOCK );
917 }
918 assert( pPager->dbSize==pPager->dbOrigSize );
919 assert( pPager->dbOrigSize==pPager->dbFileSize );
920 assert( pPager->dbOrigSize==pPager->dbHintSize );
921 assert( pPager->setSuper==0 );
922 break;
923
924 case PAGER_WRITER_CACHEMOD:
925 assert( p->eLock!=UNKNOWN_LOCK );
926 assert( pPager->errCode==SQLITE_OK );
927 if( !pagerUseWal(pPager) ){
928 /* It is possible that if journal_mode=wal here that neither the
929 ** journal file nor the WAL file are open. This happens during
930 ** a rollback transaction that switches from journal_mode=off
931 ** to journal_mode=wal.
932 */
933 assert( p->eLock>=RESERVED_LOCK );
934 assert( isOpen(p->jfd)
935 || p->journalMode==PAGER_JOURNALMODE_OFF
936 || p->journalMode==PAGER_JOURNALMODE_WAL
937 );
938 }
939 assert( pPager->dbOrigSize==pPager->dbFileSize );
940 assert( pPager->dbOrigSize==pPager->dbHintSize );
941 break;
942
943 case PAGER_WRITER_DBMOD:
944 assert( p->eLock==EXCLUSIVE_LOCK );
945 assert( pPager->errCode==SQLITE_OK );
946 assert( !pagerUseWal(pPager) );
947 assert( p->eLock>=EXCLUSIVE_LOCK );
948 assert( isOpen(p->jfd)
949 || p->journalMode==PAGER_JOURNALMODE_OFF
950 || p->journalMode==PAGER_JOURNALMODE_WAL
951 || (sqlite3OsDeviceCharacteristics(p->fd)&SQLITE_IOCAP_BATCH_ATOMIC)
952 );
953 assert( pPager->dbOrigSize<=pPager->dbHintSize );
954 break;
955
956 case PAGER_WRITER_FINISHED:
957 assert( p->eLock==EXCLUSIVE_LOCK );
958 assert( pPager->errCode==SQLITE_OK );
959 assert( !pagerUseWal(pPager) );
960 assert( isOpen(p->jfd)
961 || p->journalMode==PAGER_JOURNALMODE_OFF
962 || p->journalMode==PAGER_JOURNALMODE_WAL
963 || (sqlite3OsDeviceCharacteristics(p->fd)&SQLITE_IOCAP_BATCH_ATOMIC)
964 );
965 break;
966
967 case PAGER_ERROR:
968 /* There must be at least one outstanding reference to the pager if
969 ** in ERROR state. Otherwise the pager should have already dropped
970 ** back to OPEN state.
971 */
972 assert( pPager->errCode!=SQLITE_OK );
973 assert( sqlite3PcacheRefCount(pPager->pPCache)>0 || pPager->tempFile );
974 break;
975 }
976
977 return 1;
978}
979#endif /* ifndef NDEBUG */
980
981#ifdef SQLITE_DEBUG
982/*
983** Return a pointer to a human readable string in a static buffer
984** containing the state of the Pager object passed as an argument. This
985** is intended to be used within debuggers. For example, as an alternative
986** to "print *pPager" in gdb:
987**
988** (gdb) printf "%s", print_pager_state(pPager)
989**
990** This routine has external linkage in order to suppress compiler warnings
991** about an unused function. It is enclosed within SQLITE_DEBUG and so does
992** not appear in normal builds.
993*/
994char *print_pager_state(Pager *p){
995 static char zRet[1024];
996
997 sqlite3_snprintf(1024, zRet,
998 "Filename: %s\n"
999 "State: %s errCode=%d\n"
1000 "Lock: %s\n"
1001 "Locking mode: locking_mode=%s\n"
1002 "Journal mode: journal_mode=%s\n"
1003 "Backing store: tempFile=%d memDb=%d useJournal=%d\n"
1004 "Journal: journalOff=%lld journalHdr=%lld\n"
1005 "Size: dbsize=%d dbOrigSize=%d dbFileSize=%d\n"
1006 , p->zFilename
1007 , p->eState==PAGER_OPEN ? "OPEN" :
1008 p->eState==PAGER_READER ? "READER" :
1009 p->eState==PAGER_WRITER_LOCKED ? "WRITER_LOCKED" :
1010 p->eState==PAGER_WRITER_CACHEMOD ? "WRITER_CACHEMOD" :
1011 p->eState==PAGER_WRITER_DBMOD ? "WRITER_DBMOD" :
1012 p->eState==PAGER_WRITER_FINISHED ? "WRITER_FINISHED" :
1013 p->eState==PAGER_ERROR ? "ERROR" : "?error?"
1014 , (int)p->errCode
1015 , p->eLock==NO_LOCK ? "NO_LOCK" :
1016 p->eLock==RESERVED_LOCK ? "RESERVED" :
1017 p->eLock==EXCLUSIVE_LOCK ? "EXCLUSIVE" :
1018 p->eLock==SHARED_LOCK ? "SHARED" :
1019 p->eLock==UNKNOWN_LOCK ? "UNKNOWN" : "?error?"
1020 , p->exclusiveMode ? "exclusive" : "normal"
1021 , p->journalMode==PAGER_JOURNALMODE_MEMORY ? "memory" :
1022 p->journalMode==PAGER_JOURNALMODE_OFF ? "off" :
1023 p->journalMode==PAGER_JOURNALMODE_DELETE ? "delete" :
1024 p->journalMode==PAGER_JOURNALMODE_PERSIST ? "persist" :
1025 p->journalMode==PAGER_JOURNALMODE_TRUNCATE ? "truncate" :
1026 p->journalMode==PAGER_JOURNALMODE_WAL ? "wal" : "?error?"
1027 , (int)p->tempFile, (int)p->memDb, (int)p->useJournal
1028 , p->journalOff, p->journalHdr
1029 , (int)p->dbSize, (int)p->dbOrigSize, (int)p->dbFileSize
1030 );
1031
1032 return zRet;
1033}
1034#endif
1035
1036/* Forward references to the various page getters */
1037static int getPageNormal(Pager*,Pgno,DbPage**,int);
1038static int getPageError(Pager*,Pgno,DbPage**,int);
1039#if SQLITE_MAX_MMAP_SIZE>0
1040static int getPageMMap(Pager*,Pgno,DbPage**,int);
1041#endif
1042
1043/*
1044** Set the Pager.xGet method for the appropriate routine used to fetch
1045** content from the pager.
1046*/
1047static void setGetterMethod(Pager *pPager){
1048 if( pPager->errCode ){
1049 pPager->xGet = getPageError;
1050#if SQLITE_MAX_MMAP_SIZE>0
1051 }else if( USEFETCH(pPager) ){
1052 pPager->xGet = getPageMMap;
1053#endif /* SQLITE_MAX_MMAP_SIZE>0 */
1054 }else{
1055 pPager->xGet = getPageNormal;
1056 }
1057}
1058
1059/*
1060** Return true if it is necessary to write page *pPg into the sub-journal.
1061** A page needs to be written into the sub-journal if there exists one
1062** or more open savepoints for which:
1063**
1064** * The page-number is less than or equal to PagerSavepoint.nOrig, and
1065** * The bit corresponding to the page-number is not set in
1066** PagerSavepoint.pInSavepoint.
1067*/
1068static int subjRequiresPage(PgHdr *pPg){
1069 Pager *pPager = pPg->pPager;
1070 PagerSavepoint *p;
1071 Pgno pgno = pPg->pgno;
1072 int i;
1073 for(i=0; i<pPager->nSavepoint; i++){
1074 p = &pPager->aSavepoint[i];
1075 if( p->nOrig>=pgno && 0==sqlite3BitvecTestNotNull(p->pInSavepoint, pgno) ){
1076 for(i=i+1; i<pPager->nSavepoint; i++){
1077 pPager->aSavepoint[i].bTruncateOnRelease = 0;
1078 }
1079 return 1;
1080 }
1081 }
1082 return 0;
1083}
1084
1085#ifdef SQLITE_DEBUG
1086/*
1087** Return true if the page is already in the journal file.
1088*/
1089static int pageInJournal(Pager *pPager, PgHdr *pPg){
1090 return sqlite3BitvecTest(pPager->pInJournal, pPg->pgno);
1091}
1092#endif
1093
1094/*
1095** Read a 32-bit integer from the given file descriptor. Store the integer
1096** that is read in *pRes. Return SQLITE_OK if everything worked, or an
1097** error code is something goes wrong.
1098**
1099** All values are stored on disk as big-endian.
1100*/
1101static int read32bits(sqlite3_file *fd, i64 offset, u32 *pRes){
1102 unsigned char ac[4];
1103 int rc = sqlite3OsRead(fd, ac, sizeof(ac), offset);
1104 if( rc==SQLITE_OK ){
1105 *pRes = sqlite3Get4byte(ac);
1106 }
1107 return rc;
1108}
1109
1110/*
1111** Write a 32-bit integer into a string buffer in big-endian byte order.
1112*/
1113#define put32bits(A,B) sqlite3Put4byte((u8*)A,B)
1114
1115
1116/*
1117** Write a 32-bit integer into the given file descriptor. Return SQLITE_OK
1118** on success or an error code is something goes wrong.
1119*/
1120static int write32bits(sqlite3_file *fd, i64 offset, u32 val){
1121 char ac[4];
1122 put32bits(ac, val);
1123 return sqlite3OsWrite(fd, ac, 4, offset);
1124}
1125
1126/*
1127** Unlock the database file to level eLock, which must be either NO_LOCK
1128** or SHARED_LOCK. Regardless of whether or not the call to xUnlock()
1129** succeeds, set the Pager.eLock variable to match the (attempted) new lock.
1130**
1131** Except, if Pager.eLock is set to UNKNOWN_LOCK when this function is
1132** called, do not modify it. See the comment above the #define of
1133** UNKNOWN_LOCK for an explanation of this.
1134*/
1135static int pagerUnlockDb(Pager *pPager, int eLock){
1136 int rc = SQLITE_OK;
1137
1138 assert( !pPager->exclusiveMode || pPager->eLock==eLock );
1139 assert( eLock==NO_LOCK || eLock==SHARED_LOCK );
1140 assert( eLock!=NO_LOCK || pagerUseWal(pPager)==0 );
1141 if( isOpen(pPager->fd) ){
1142 assert( pPager->eLock>=eLock );
1143 rc = pPager->noLock ? SQLITE_OK : sqlite3OsUnlock(pPager->fd, eLock);
1144 if( pPager->eLock!=UNKNOWN_LOCK ){
1145 pPager->eLock = (u8)eLock;
1146 }
1147 IOTRACE(("UNLOCK %p %d\n", pPager, eLock))
1148 }
1149 pPager->changeCountDone = pPager->tempFile; /* ticket fb3b3024ea238d5c */
1150 return rc;
1151}
1152
1153/*
1154** Lock the database file to level eLock, which must be either SHARED_LOCK,
1155** RESERVED_LOCK or EXCLUSIVE_LOCK. If the caller is successful, set the
1156** Pager.eLock variable to the new locking state.
1157**
1158** Except, if Pager.eLock is set to UNKNOWN_LOCK when this function is
1159** called, do not modify it unless the new locking state is EXCLUSIVE_LOCK.
1160** See the comment above the #define of UNKNOWN_LOCK for an explanation
1161** of this.
1162*/
1163static int pagerLockDb(Pager *pPager, int eLock){
1164 int rc = SQLITE_OK;
1165
1166 assert( eLock==SHARED_LOCK || eLock==RESERVED_LOCK || eLock==EXCLUSIVE_LOCK );
1167 if( pPager->eLock<eLock || pPager->eLock==UNKNOWN_LOCK ){
1168 rc = pPager->noLock ? SQLITE_OK : sqlite3OsLock(pPager->fd, eLock);
1169 if( rc==SQLITE_OK && (pPager->eLock!=UNKNOWN_LOCK||eLock==EXCLUSIVE_LOCK) ){
1170 pPager->eLock = (u8)eLock;
1171 IOTRACE(("LOCK %p %d\n", pPager, eLock))
1172 }
1173 }
1174 return rc;
1175}
1176
1177/*
1178** This function determines whether or not the atomic-write or
1179** atomic-batch-write optimizations can be used with this pager. The
1180** atomic-write optimization can be used if:
1181**
1182** (a) the value returned by OsDeviceCharacteristics() indicates that
1183** a database page may be written atomically, and
1184** (b) the value returned by OsSectorSize() is less than or equal
1185** to the page size.
1186**
1187** If it can be used, then the value returned is the size of the journal
1188** file when it contains rollback data for exactly one page.
1189**
1190** The atomic-batch-write optimization can be used if OsDeviceCharacteristics()
1191** returns a value with the SQLITE_IOCAP_BATCH_ATOMIC bit set. -1 is
1192** returned in this case.
1193**
1194** If neither optimization can be used, 0 is returned.
1195*/
1196static int jrnlBufferSize(Pager *pPager){
1197 assert( !MEMDB );
1198
1199#if defined(SQLITE_ENABLE_ATOMIC_WRITE) \
1200 || defined(SQLITE_ENABLE_BATCH_ATOMIC_WRITE)
1201 int dc; /* Device characteristics */
1202
1203 assert( isOpen(pPager->fd) );
1204 dc = sqlite3OsDeviceCharacteristics(pPager->fd);
1205#else
1206 UNUSED_PARAMETER(pPager);
1207#endif
1208
1209#ifdef SQLITE_ENABLE_BATCH_ATOMIC_WRITE
1210 if( pPager->dbSize>0 && (dc&SQLITE_IOCAP_BATCH_ATOMIC) ){
1211 return -1;
1212 }
1213#endif
1214
1215#ifdef SQLITE_ENABLE_ATOMIC_WRITE
1216 {
1217 int nSector = pPager->sectorSize;
1218 int szPage = pPager->pageSize;
1219
1220 assert(SQLITE_IOCAP_ATOMIC512==(512>>8));
1221 assert(SQLITE_IOCAP_ATOMIC64K==(65536>>8));
1222 if( 0==(dc&(SQLITE_IOCAP_ATOMIC|(szPage>>8)) || nSector>szPage) ){
1223 return 0;
1224 }
1225 }
1226
1227 return JOURNAL_HDR_SZ(pPager) + JOURNAL_PG_SZ(pPager);
1228#endif
1229
1230 return 0;
1231}
1232
1233/*
1234** If SQLITE_CHECK_PAGES is defined then we do some sanity checking
1235** on the cache using a hash function. This is used for testing
1236** and debugging only.
1237*/
1238#ifdef SQLITE_CHECK_PAGES
1239/*
1240** Return a 32-bit hash of the page data for pPage.
1241*/
1242static u32 pager_datahash(int nByte, unsigned char *pData){
1243 u32 hash = 0;
1244 int i;
1245 for(i=0; i<nByte; i++){
1246 hash = (hash*1039) + pData[i];
1247 }
1248 return hash;
1249}
1250static u32 pager_pagehash(PgHdr *pPage){
1251 return pager_datahash(pPage->pPager->pageSize, (unsigned char *)pPage->pData);
1252}
1253static void pager_set_pagehash(PgHdr *pPage){
1254 pPage->pageHash = pager_pagehash(pPage);
1255}
1256
1257/*
1258** The CHECK_PAGE macro takes a PgHdr* as an argument. If SQLITE_CHECK_PAGES
1259** is defined, and NDEBUG is not defined, an assert() statement checks
1260** that the page is either dirty or still matches the calculated page-hash.
1261*/
1262#define CHECK_PAGE(x) checkPage(x)
1263static void checkPage(PgHdr *pPg){
1264 Pager *pPager = pPg->pPager;
1265 assert( pPager->eState!=PAGER_ERROR );
1266 assert( (pPg->flags&PGHDR_DIRTY) || pPg->pageHash==pager_pagehash(pPg) );
1267}
1268
1269#else
1270#define pager_datahash(X,Y) 0
1271#define pager_pagehash(X) 0
1272#define pager_set_pagehash(X)
1273#define CHECK_PAGE(x)
1274#endif /* SQLITE_CHECK_PAGES */
1275
1276/*
1277** When this is called the journal file for pager pPager must be open.
1278** This function attempts to read a super-journal file name from the
1279** end of the file and, if successful, copies it into memory supplied
1280** by the caller. See comments above writeSuperJournal() for the format
1281** used to store a super-journal file name at the end of a journal file.
1282**
1283** zSuper must point to a buffer of at least nSuper bytes allocated by
1284** the caller. This should be sqlite3_vfs.mxPathname+1 (to ensure there is
1285** enough space to write the super-journal name). If the super-journal
1286** name in the journal is longer than nSuper bytes (including a
1287** nul-terminator), then this is handled as if no super-journal name
1288** were present in the journal.
1289**
1290** If a super-journal file name is present at the end of the journal
1291** file, then it is copied into the buffer pointed to by zSuper. A
1292** nul-terminator byte is appended to the buffer following the
1293** super-journal file name.
1294**
1295** If it is determined that no super-journal file name is present
1296** zSuper[0] is set to 0 and SQLITE_OK returned.
1297**
1298** If an error occurs while reading from the journal file, an SQLite
1299** error code is returned.
1300*/
1301static int readSuperJournal(sqlite3_file *pJrnl, char *zSuper, u32 nSuper){
1302 int rc; /* Return code */
1303 u32 len; /* Length in bytes of super-journal name */
1304 i64 szJ; /* Total size in bytes of journal file pJrnl */
1305 u32 cksum; /* MJ checksum value read from journal */
1306 u32 u; /* Unsigned loop counter */
1307 unsigned char aMagic[8]; /* A buffer to hold the magic header */
1308 zSuper[0] = '\0';
1309
1310 if( SQLITE_OK!=(rc = sqlite3OsFileSize(pJrnl, &szJ))
1311 || szJ<16
1312 || SQLITE_OK!=(rc = read32bits(pJrnl, szJ-16, &len))
1313 || len>=nSuper
1314 || len>szJ-16
1315 || len==0
1316 || SQLITE_OK!=(rc = read32bits(pJrnl, szJ-12, &cksum))
1317 || SQLITE_OK!=(rc = sqlite3OsRead(pJrnl, aMagic, 8, szJ-8))
1318 || memcmp(aMagic, aJournalMagic, 8)
1319 || SQLITE_OK!=(rc = sqlite3OsRead(pJrnl, zSuper, len, szJ-16-len))
1320 ){
1321 return rc;
1322 }
1323
1324 /* See if the checksum matches the super-journal name */
1325 for(u=0; u<len; u++){
1326 cksum -= zSuper[u];
1327 }
1328 if( cksum ){
1329 /* If the checksum doesn't add up, then one or more of the disk sectors
1330 ** containing the super-journal filename is corrupted. This means
1331 ** definitely roll back, so just return SQLITE_OK and report a (nul)
1332 ** super-journal filename.
1333 */
1334 len = 0;
1335 }
1336 zSuper[len] = '\0';
1337 zSuper[len+1] = '\0';
1338
1339 return SQLITE_OK;
1340}
1341
1342/*
1343** Return the offset of the sector boundary at or immediately
1344** following the value in pPager->journalOff, assuming a sector
1345** size of pPager->sectorSize bytes.
1346**
1347** i.e for a sector size of 512:
1348**
1349** Pager.journalOff Return value
1350** ---------------------------------------
1351** 0 0
1352** 512 512
1353** 100 512
1354** 2000 2048
1355**
1356*/
1357static i64 journalHdrOffset(Pager *pPager){
1358 i64 offset = 0;
1359 i64 c = pPager->journalOff;
1360 if( c ){
1361 offset = ((c-1)/JOURNAL_HDR_SZ(pPager) + 1) * JOURNAL_HDR_SZ(pPager);
1362 }
1363 assert( offset%JOURNAL_HDR_SZ(pPager)==0 );
1364 assert( offset>=c );
1365 assert( (offset-c)<JOURNAL_HDR_SZ(pPager) );
1366 return offset;
1367}
1368
1369/*
1370** The journal file must be open when this function is called.
1371**
1372** This function is a no-op if the journal file has not been written to
1373** within the current transaction (i.e. if Pager.journalOff==0).
1374**
1375** If doTruncate is non-zero or the Pager.journalSizeLimit variable is
1376** set to 0, then truncate the journal file to zero bytes in size. Otherwise,
1377** zero the 28-byte header at the start of the journal file. In either case,
1378** if the pager is not in no-sync mode, sync the journal file immediately
1379** after writing or truncating it.
1380**
1381** If Pager.journalSizeLimit is set to a positive, non-zero value, and
1382** following the truncation or zeroing described above the size of the
1383** journal file in bytes is larger than this value, then truncate the
1384** journal file to Pager.journalSizeLimit bytes. The journal file does
1385** not need to be synced following this operation.
1386**
1387** If an IO error occurs, abandon processing and return the IO error code.
1388** Otherwise, return SQLITE_OK.
1389*/
1390static int zeroJournalHdr(Pager *pPager, int doTruncate){
1391 int rc = SQLITE_OK; /* Return code */
1392 assert( isOpen(pPager->jfd) );
1393 assert( !sqlite3JournalIsInMemory(pPager->jfd) );
1394 if( pPager->journalOff ){
1395 const i64 iLimit = pPager->journalSizeLimit; /* Local cache of jsl */
1396
1397 IOTRACE(("JZEROHDR %p\n", pPager))
1398 if( doTruncate || iLimit==0 ){
1399 rc = sqlite3OsTruncate(pPager->jfd, 0);
1400 }else{
1401 static const char zeroHdr[28] = {0};
1402 rc = sqlite3OsWrite(pPager->jfd, zeroHdr, sizeof(zeroHdr), 0);
1403 }
1404 if( rc==SQLITE_OK && !pPager->noSync ){
1405 rc = sqlite3OsSync(pPager->jfd, SQLITE_SYNC_DATAONLY|pPager->syncFlags);
1406 }
1407
1408 /* At this point the transaction is committed but the write lock
1409 ** is still held on the file. If there is a size limit configured for
1410 ** the persistent journal and the journal file currently consumes more
1411 ** space than that limit allows for, truncate it now. There is no need
1412 ** to sync the file following this operation.
1413 */
1414 if( rc==SQLITE_OK && iLimit>0 ){
1415 i64 sz;
1416 rc = sqlite3OsFileSize(pPager->jfd, &sz);
1417 if( rc==SQLITE_OK && sz>iLimit ){
1418 rc = sqlite3OsTruncate(pPager->jfd, iLimit);
1419 }
1420 }
1421 }
1422 return rc;
1423}
1424
1425/*
1426** The journal file must be open when this routine is called. A journal
1427** header (JOURNAL_HDR_SZ bytes) is written into the journal file at the
1428** current location.
1429**
1430** The format for the journal header is as follows:
1431** - 8 bytes: Magic identifying journal format.
1432** - 4 bytes: Number of records in journal, or -1 no-sync mode is on.
1433** - 4 bytes: Random number used for page hash.
1434** - 4 bytes: Initial database page count.
1435** - 4 bytes: Sector size used by the process that wrote this journal.
1436** - 4 bytes: Database page size.
1437**
1438** Followed by (JOURNAL_HDR_SZ - 28) bytes of unused space.
1439*/
1440static int writeJournalHdr(Pager *pPager){
1441 int rc = SQLITE_OK; /* Return code */
1442 char *zHeader = pPager->pTmpSpace; /* Temporary space used to build header */
1443 u32 nHeader = (u32)pPager->pageSize;/* Size of buffer pointed to by zHeader */
1444 u32 nWrite; /* Bytes of header sector written */
1445 int ii; /* Loop counter */
1446
1447 assert( isOpen(pPager->jfd) ); /* Journal file must be open. */
1448
1449 if( nHeader>JOURNAL_HDR_SZ(pPager) ){
1450 nHeader = JOURNAL_HDR_SZ(pPager);
1451 }
1452
1453 /* If there are active savepoints and any of them were created
1454 ** since the most recent journal header was written, update the
1455 ** PagerSavepoint.iHdrOffset fields now.
1456 */
1457 for(ii=0; ii<pPager->nSavepoint; ii++){
1458 if( pPager->aSavepoint[ii].iHdrOffset==0 ){
1459 pPager->aSavepoint[ii].iHdrOffset = pPager->journalOff;
1460 }
1461 }
1462
1463 pPager->journalHdr = pPager->journalOff = journalHdrOffset(pPager);
1464
1465 /*
1466 ** Write the nRec Field - the number of page records that follow this
1467 ** journal header. Normally, zero is written to this value at this time.
1468 ** After the records are added to the journal (and the journal synced,
1469 ** if in full-sync mode), the zero is overwritten with the true number
1470 ** of records (see syncJournal()).
1471 **
1472 ** A faster alternative is to write 0xFFFFFFFF to the nRec field. When
1473 ** reading the journal this value tells SQLite to assume that the
1474 ** rest of the journal file contains valid page records. This assumption
1475 ** is dangerous, as if a failure occurred whilst writing to the journal
1476 ** file it may contain some garbage data. There are two scenarios
1477 ** where this risk can be ignored:
1478 **
1479 ** * When the pager is in no-sync mode. Corruption can follow a
1480 ** power failure in this case anyway.
1481 **
1482 ** * When the SQLITE_IOCAP_SAFE_APPEND flag is set. This guarantees
1483 ** that garbage data is never appended to the journal file.
1484 */
1485 assert( isOpen(pPager->fd) || pPager->noSync );
1486 if( pPager->noSync || (pPager->journalMode==PAGER_JOURNALMODE_MEMORY)
1487 || (sqlite3OsDeviceCharacteristics(pPager->fd)&SQLITE_IOCAP_SAFE_APPEND)
1488 ){
1489 memcpy(zHeader, aJournalMagic, sizeof(aJournalMagic));
1490 put32bits(&zHeader[sizeof(aJournalMagic)], 0xffffffff);
1491 }else{
1492 memset(zHeader, 0, sizeof(aJournalMagic)+4);
1493 }
1494
1495 /* The random check-hash initializer */
1496 sqlite3_randomness(sizeof(pPager->cksumInit), &pPager->cksumInit);
1497 put32bits(&zHeader[sizeof(aJournalMagic)+4], pPager->cksumInit);
1498 /* The initial database size */
1499 put32bits(&zHeader[sizeof(aJournalMagic)+8], pPager->dbOrigSize);
1500 /* The assumed sector size for this process */
1501 put32bits(&zHeader[sizeof(aJournalMagic)+12], pPager->sectorSize);
1502
1503 /* The page size */
1504 put32bits(&zHeader[sizeof(aJournalMagic)+16], pPager->pageSize);
1505
1506 /* Initializing the tail of the buffer is not necessary. Everything
1507 ** works find if the following memset() is omitted. But initializing
1508 ** the memory prevents valgrind from complaining, so we are willing to
1509 ** take the performance hit.
1510 */
1511 memset(&zHeader[sizeof(aJournalMagic)+20], 0,
1512 nHeader-(sizeof(aJournalMagic)+20));
1513
1514 /* In theory, it is only necessary to write the 28 bytes that the
1515 ** journal header consumes to the journal file here. Then increment the
1516 ** Pager.journalOff variable by JOURNAL_HDR_SZ so that the next
1517 ** record is written to the following sector (leaving a gap in the file
1518 ** that will be implicitly filled in by the OS).
1519 **
1520 ** However it has been discovered that on some systems this pattern can
1521 ** be significantly slower than contiguously writing data to the file,
1522 ** even if that means explicitly writing data to the block of
1523 ** (JOURNAL_HDR_SZ - 28) bytes that will not be used. So that is what
1524 ** is done.
1525 **
1526 ** The loop is required here in case the sector-size is larger than the
1527 ** database page size. Since the zHeader buffer is only Pager.pageSize
1528 ** bytes in size, more than one call to sqlite3OsWrite() may be required
1529 ** to populate the entire journal header sector.
1530 */
1531 for(nWrite=0; rc==SQLITE_OK&&nWrite<JOURNAL_HDR_SZ(pPager); nWrite+=nHeader){
1532 IOTRACE(("JHDR %p %lld %d\n", pPager, pPager->journalHdr, nHeader))
1533 rc = sqlite3OsWrite(pPager->jfd, zHeader, nHeader, pPager->journalOff);
1534 assert( pPager->journalHdr <= pPager->journalOff );
1535 pPager->journalOff += nHeader;
1536 }
1537
1538 return rc;
1539}
1540
1541/*
1542** The journal file must be open when this is called. A journal header file
1543** (JOURNAL_HDR_SZ bytes) is read from the current location in the journal
1544** file. The current location in the journal file is given by
1545** pPager->journalOff. See comments above function writeJournalHdr() for
1546** a description of the journal header format.
1547**
1548** If the header is read successfully, *pNRec is set to the number of
1549** page records following this header and *pDbSize is set to the size of the
1550** database before the transaction began, in pages. Also, pPager->cksumInit
1551** is set to the value read from the journal header. SQLITE_OK is returned
1552** in this case.
1553**
1554** If the journal header file appears to be corrupted, SQLITE_DONE is
1555** returned and *pNRec and *PDbSize are undefined. If JOURNAL_HDR_SZ bytes
1556** cannot be read from the journal file an error code is returned.
1557*/
1558static int readJournalHdr(
1559 Pager *pPager, /* Pager object */
1560 int isHot,
1561 i64 journalSize, /* Size of the open journal file in bytes */
1562 u32 *pNRec, /* OUT: Value read from the nRec field */
1563 u32 *pDbSize /* OUT: Value of original database size field */
1564){
1565 int rc; /* Return code */
1566 unsigned char aMagic[8]; /* A buffer to hold the magic header */
1567 i64 iHdrOff; /* Offset of journal header being read */
1568
1569 assert( isOpen(pPager->jfd) ); /* Journal file must be open. */
1570
1571 /* Advance Pager.journalOff to the start of the next sector. If the
1572 ** journal file is too small for there to be a header stored at this
1573 ** point, return SQLITE_DONE.
1574 */
1575 pPager->journalOff = journalHdrOffset(pPager);
1576 if( pPager->journalOff+JOURNAL_HDR_SZ(pPager) > journalSize ){
1577 return SQLITE_DONE;
1578 }
1579 iHdrOff = pPager->journalOff;
1580
1581 /* Read in the first 8 bytes of the journal header. If they do not match
1582 ** the magic string found at the start of each journal header, return
1583 ** SQLITE_DONE. If an IO error occurs, return an error code. Otherwise,
1584 ** proceed.
1585 */
1586 if( isHot || iHdrOff!=pPager->journalHdr ){
1587 rc = sqlite3OsRead(pPager->jfd, aMagic, sizeof(aMagic), iHdrOff);
1588 if( rc ){
1589 return rc;
1590 }
1591 if( memcmp(aMagic, aJournalMagic, sizeof(aMagic))!=0 ){
1592 return SQLITE_DONE;
1593 }
1594 }
1595
1596 /* Read the first three 32-bit fields of the journal header: The nRec
1597 ** field, the checksum-initializer and the database size at the start
1598 ** of the transaction. Return an error code if anything goes wrong.
1599 */
1600 if( SQLITE_OK!=(rc = read32bits(pPager->jfd, iHdrOff+8, pNRec))
1601 || SQLITE_OK!=(rc = read32bits(pPager->jfd, iHdrOff+12, &pPager->cksumInit))
1602 || SQLITE_OK!=(rc = read32bits(pPager->jfd, iHdrOff+16, pDbSize))
1603 ){
1604 return rc;
1605 }
1606
1607 if( pPager->journalOff==0 ){
1608 u32 iPageSize; /* Page-size field of journal header */
1609 u32 iSectorSize; /* Sector-size field of journal header */
1610
1611 /* Read the page-size and sector-size journal header fields. */
1612 if( SQLITE_OK!=(rc = read32bits(pPager->jfd, iHdrOff+20, &iSectorSize))
1613 || SQLITE_OK!=(rc = read32bits(pPager->jfd, iHdrOff+24, &iPageSize))
1614 ){
1615 return rc;
1616 }
1617
1618 /* Versions of SQLite prior to 3.5.8 set the page-size field of the
1619 ** journal header to zero. In this case, assume that the Pager.pageSize
1620 ** variable is already set to the correct page size.
1621 */
1622 if( iPageSize==0 ){
1623 iPageSize = pPager->pageSize;
1624 }
1625
1626 /* Check that the values read from the page-size and sector-size fields
1627 ** are within range. To be 'in range', both values need to be a power
1628 ** of two greater than or equal to 512 or 32, and not greater than their
1629 ** respective compile time maximum limits.
1630 */
1631 if( iPageSize<512 || iSectorSize<32
1632 || iPageSize>SQLITE_MAX_PAGE_SIZE || iSectorSize>MAX_SECTOR_SIZE
1633 || ((iPageSize-1)&iPageSize)!=0 || ((iSectorSize-1)&iSectorSize)!=0
1634 ){
1635 /* If the either the page-size or sector-size in the journal-header is
1636 ** invalid, then the process that wrote the journal-header must have
1637 ** crashed before the header was synced. In this case stop reading
1638 ** the journal file here.
1639 */
1640 return SQLITE_DONE;
1641 }
1642
1643 /* Update the page-size to match the value read from the journal.
1644 ** Use a testcase() macro to make sure that malloc failure within
1645 ** PagerSetPagesize() is tested.
1646 */
1647 rc = sqlite3PagerSetPagesize(pPager, &iPageSize, -1);
1648 testcase( rc!=SQLITE_OK );
1649
1650 /* Update the assumed sector-size to match the value used by
1651 ** the process that created this journal. If this journal was
1652 ** created by a process other than this one, then this routine
1653 ** is being called from within pager_playback(). The local value
1654 ** of Pager.sectorSize is restored at the end of that routine.
1655 */
1656 pPager->sectorSize = iSectorSize;
1657 }
1658
1659 pPager->journalOff += JOURNAL_HDR_SZ(pPager);
1660 return rc;
1661}
1662
1663
1664/*
1665** Write the supplied super-journal name into the journal file for pager
1666** pPager at the current location. The super-journal name must be the last
1667** thing written to a journal file. If the pager is in full-sync mode, the
1668** journal file descriptor is advanced to the next sector boundary before
1669** anything is written. The format is:
1670**
1671** + 4 bytes: PAGER_SJ_PGNO.
1672** + N bytes: super-journal filename in utf-8.
1673** + 4 bytes: N (length of super-journal name in bytes, no nul-terminator).
1674** + 4 bytes: super-journal name checksum.
1675** + 8 bytes: aJournalMagic[].
1676**
1677** The super-journal page checksum is the sum of the bytes in thesuper-journal
1678** name, where each byte is interpreted as a signed 8-bit integer.
1679**
1680** If zSuper is a NULL pointer (occurs for a single database transaction),
1681** this call is a no-op.
1682*/
1683static int writeSuperJournal(Pager *pPager, const char *zSuper){
1684 int rc; /* Return code */
1685 int nSuper; /* Length of string zSuper */
1686 i64 iHdrOff; /* Offset of header in journal file */
1687 i64 jrnlSize; /* Size of journal file on disk */
1688 u32 cksum = 0; /* Checksum of string zSuper */
1689
1690 assert( pPager->setSuper==0 );
1691 assert( !pagerUseWal(pPager) );
1692
1693 if( !zSuper
1694 || pPager->journalMode==PAGER_JOURNALMODE_MEMORY
1695 || !isOpen(pPager->jfd)
1696 ){
1697 return SQLITE_OK;
1698 }
1699 pPager->setSuper = 1;
1700 assert( pPager->journalHdr <= pPager->journalOff );
1701
1702 /* Calculate the length in bytes and the checksum of zSuper */
1703 for(nSuper=0; zSuper[nSuper]; nSuper++){
1704 cksum += zSuper[nSuper];
1705 }
1706
1707 /* If in full-sync mode, advance to the next disk sector before writing
1708 ** the super-journal name. This is in case the previous page written to
1709 ** the journal has already been synced.
1710 */
1711 if( pPager->fullSync ){
1712 pPager->journalOff = journalHdrOffset(pPager);
1713 }
1714 iHdrOff = pPager->journalOff;
1715
1716 /* Write the super-journal data to the end of the journal file. If
1717 ** an error occurs, return the error code to the caller.
1718 */
1719 if( (0 != (rc = write32bits(pPager->jfd, iHdrOff, PAGER_SJ_PGNO(pPager))))
1720 || (0 != (rc = sqlite3OsWrite(pPager->jfd, zSuper, nSuper, iHdrOff+4)))
1721 || (0 != (rc = write32bits(pPager->jfd, iHdrOff+4+nSuper, nSuper)))
1722 || (0 != (rc = write32bits(pPager->jfd, iHdrOff+4+nSuper+4, cksum)))
1723 || (0 != (rc = sqlite3OsWrite(pPager->jfd, aJournalMagic, 8,
1724 iHdrOff+4+nSuper+8)))
1725 ){
1726 return rc;
1727 }
1728 pPager->journalOff += (nSuper+20);
1729
1730 /* If the pager is in peristent-journal mode, then the physical
1731 ** journal-file may extend past the end of the super-journal name
1732 ** and 8 bytes of magic data just written to the file. This is
1733 ** dangerous because the code to rollback a hot-journal file
1734 ** will not be able to find the super-journal name to determine
1735 ** whether or not the journal is hot.
1736 **
1737 ** Easiest thing to do in this scenario is to truncate the journal
1738 ** file to the required size.
1739 */
1740 if( SQLITE_OK==(rc = sqlite3OsFileSize(pPager->jfd, &jrnlSize))
1741 && jrnlSize>pPager->journalOff
1742 ){
1743 rc = sqlite3OsTruncate(pPager->jfd, pPager->journalOff);
1744 }
1745 return rc;
1746}
1747
1748/*
1749** Discard the entire contents of the in-memory page-cache.
1750*/
1751static void pager_reset(Pager *pPager){
1752 pPager->iDataVersion++;
1753 sqlite3BackupRestart(pPager->pBackup);
1754 sqlite3PcacheClear(pPager->pPCache);
1755}
1756
1757/*
1758** Return the pPager->iDataVersion value
1759*/
1760u32 sqlite3PagerDataVersion(Pager *pPager){
1761 return pPager->iDataVersion;
1762}
1763
1764/*
1765** Free all structures in the Pager.aSavepoint[] array and set both
1766** Pager.aSavepoint and Pager.nSavepoint to zero. Close the sub-journal
1767** if it is open and the pager is not in exclusive mode.
1768*/
1769static void releaseAllSavepoints(Pager *pPager){
1770 int ii; /* Iterator for looping through Pager.aSavepoint */
1771 for(ii=0; ii<pPager->nSavepoint; ii++){
1772 sqlite3BitvecDestroy(pPager->aSavepoint[ii].pInSavepoint);
1773 }
1774 if( !pPager->exclusiveMode || sqlite3JournalIsInMemory(pPager->sjfd) ){
1775 sqlite3OsClose(pPager->sjfd);
1776 }
1777 sqlite3_free(pPager->aSavepoint);
1778 pPager->aSavepoint = 0;
1779 pPager->nSavepoint = 0;
1780 pPager->nSubRec = 0;
1781}
1782
1783/*
1784** Set the bit number pgno in the PagerSavepoint.pInSavepoint
1785** bitvecs of all open savepoints. Return SQLITE_OK if successful
1786** or SQLITE_NOMEM if a malloc failure occurs.
1787*/
1788static int addToSavepointBitvecs(Pager *pPager, Pgno pgno){
1789 int ii; /* Loop counter */
1790 int rc = SQLITE_OK; /* Result code */
1791
1792 for(ii=0; ii<pPager->nSavepoint; ii++){
1793 PagerSavepoint *p = &pPager->aSavepoint[ii];
1794 if( pgno<=p->nOrig ){
1795 rc |= sqlite3BitvecSet(p->pInSavepoint, pgno);
1796 testcase( rc==SQLITE_NOMEM );
1797 assert( rc==SQLITE_OK || rc==SQLITE_NOMEM );
1798 }
1799 }
1800 return rc;
1801}
1802
1803/*
1804** This function is a no-op if the pager is in exclusive mode and not
1805** in the ERROR state. Otherwise, it switches the pager to PAGER_OPEN
1806** state.
1807**
1808** If the pager is not in exclusive-access mode, the database file is
1809** completely unlocked. If the file is unlocked and the file-system does
1810** not exhibit the UNDELETABLE_WHEN_OPEN property, the journal file is
1811** closed (if it is open).
1812**
1813** If the pager is in ERROR state when this function is called, the
1814** contents of the pager cache are discarded before switching back to
1815** the OPEN state. Regardless of whether the pager is in exclusive-mode
1816** or not, any journal file left in the file-system will be treated
1817** as a hot-journal and rolled back the next time a read-transaction
1818** is opened (by this or by any other connection).
1819*/
1820static void pager_unlock(Pager *pPager){
1821
1822 assert( pPager->eState==PAGER_READER
1823 || pPager->eState==PAGER_OPEN
1824 || pPager->eState==PAGER_ERROR
1825 );
1826
1827 sqlite3BitvecDestroy(pPager->pInJournal);
1828 pPager->pInJournal = 0;
1829 releaseAllSavepoints(pPager);
1830
1831 if( pagerUseWal(pPager) ){
1832 assert( !isOpen(pPager->jfd) );
1833 sqlite3WalEndReadTransaction(pPager->pWal);
1834 pPager->eState = PAGER_OPEN;
1835 }else if( !pPager->exclusiveMode ){
1836 int rc; /* Error code returned by pagerUnlockDb() */
1837 int iDc = isOpen(pPager->fd)?sqlite3OsDeviceCharacteristics(pPager->fd):0;
1838
1839 /* If the operating system support deletion of open files, then
1840 ** close the journal file when dropping the database lock. Otherwise
1841 ** another connection with journal_mode=delete might delete the file
1842 ** out from under us.
1843 */
1844 assert( (PAGER_JOURNALMODE_MEMORY & 5)!=1 );
1845 assert( (PAGER_JOURNALMODE_OFF & 5)!=1 );
1846 assert( (PAGER_JOURNALMODE_WAL & 5)!=1 );
1847 assert( (PAGER_JOURNALMODE_DELETE & 5)!=1 );
1848 assert( (PAGER_JOURNALMODE_TRUNCATE & 5)==1 );
1849 assert( (PAGER_JOURNALMODE_PERSIST & 5)==1 );
1850 if( 0==(iDc & SQLITE_IOCAP_UNDELETABLE_WHEN_OPEN)
1851 || 1!=(pPager->journalMode & 5)
1852 ){
1853 sqlite3OsClose(pPager->jfd);
1854 }
1855
1856 /* If the pager is in the ERROR state and the call to unlock the database
1857 ** file fails, set the current lock to UNKNOWN_LOCK. See the comment
1858 ** above the #define for UNKNOWN_LOCK for an explanation of why this
1859 ** is necessary.
1860 */
1861 rc = pagerUnlockDb(pPager, NO_LOCK);
1862 if( rc!=SQLITE_OK && pPager->eState==PAGER_ERROR ){
1863 pPager->eLock = UNKNOWN_LOCK;
1864 }
1865
1866 /* The pager state may be changed from PAGER_ERROR to PAGER_OPEN here
1867 ** without clearing the error code. This is intentional - the error
1868 ** code is cleared and the cache reset in the block below.
1869 */
1870 assert( pPager->errCode || pPager->eState!=PAGER_ERROR );
1871 pPager->eState = PAGER_OPEN;
1872 }
1873
1874 /* If Pager.errCode is set, the contents of the pager cache cannot be
1875 ** trusted. Now that there are no outstanding references to the pager,
1876 ** it can safely move back to PAGER_OPEN state. This happens in both
1877 ** normal and exclusive-locking mode.
1878 */
1879 assert( pPager->errCode==SQLITE_OK || !MEMDB );
1880 if( pPager->errCode ){
1881 if( pPager->tempFile==0 ){
1882 pager_reset(pPager);
1883 pPager->changeCountDone = 0;
1884 pPager->eState = PAGER_OPEN;
1885 }else{
1886 pPager->eState = (isOpen(pPager->jfd) ? PAGER_OPEN : PAGER_READER);
1887 }
1888 if( USEFETCH(pPager) ) sqlite3OsUnfetch(pPager->fd, 0, 0);
1889 pPager->errCode = SQLITE_OK;
1890 setGetterMethod(pPager);
1891 }
1892
1893 pPager->journalOff = 0;
1894 pPager->journalHdr = 0;
1895 pPager->setSuper = 0;
1896}
1897
1898/*
1899** This function is called whenever an IOERR or FULL error that requires
1900** the pager to transition into the ERROR state may ahve occurred.
1901** The first argument is a pointer to the pager structure, the second
1902** the error-code about to be returned by a pager API function. The
1903** value returned is a copy of the second argument to this function.
1904**
1905** If the second argument is SQLITE_FULL, SQLITE_IOERR or one of the
1906** IOERR sub-codes, the pager enters the ERROR state and the error code
1907** is stored in Pager.errCode. While the pager remains in the ERROR state,
1908** all major API calls on the Pager will immediately return Pager.errCode.
1909**
1910** The ERROR state indicates that the contents of the pager-cache
1911** cannot be trusted. This state can be cleared by completely discarding
1912** the contents of the pager-cache. If a transaction was active when
1913** the persistent error occurred, then the rollback journal may need
1914** to be replayed to restore the contents of the database file (as if
1915** it were a hot-journal).
1916*/
1917static int pager_error(Pager *pPager, int rc){
1918 int rc2 = rc & 0xff;
1919 assert( rc==SQLITE_OK || !MEMDB );
1920 assert(
1921 pPager->errCode==SQLITE_FULL ||
1922 pPager->errCode==SQLITE_OK ||
1923 (pPager->errCode & 0xff)==SQLITE_IOERR
1924 );
1925 if( rc2==SQLITE_FULL || rc2==SQLITE_IOERR ){
1926 pPager->errCode = rc;
1927 pPager->eState = PAGER_ERROR;
1928 setGetterMethod(pPager);
1929 }
1930 return rc;
1931}
1932
1933static int pager_truncate(Pager *pPager, Pgno nPage);
1934
1935/*
1936** The write transaction open on pPager is being committed (bCommit==1)
1937** or rolled back (bCommit==0).
1938**
1939** Return TRUE if and only if all dirty pages should be flushed to disk.
1940**
1941** Rules:
1942**
1943** * For non-TEMP databases, always sync to disk. This is necessary
1944** for transactions to be durable.
1945**
1946** * Sync TEMP database only on a COMMIT (not a ROLLBACK) when the backing
1947** file has been created already (via a spill on pagerStress()) and
1948** when the number of dirty pages in memory exceeds 25% of the total
1949** cache size.
1950*/
1951static int pagerFlushOnCommit(Pager *pPager, int bCommit){
1952 if( pPager->tempFile==0 ) return 1;
1953 if( !bCommit ) return 0;
1954 if( !isOpen(pPager->fd) ) return 0;
1955 return (sqlite3PCachePercentDirty(pPager->pPCache)>=25);
1956}
1957
1958/*
1959** This routine ends a transaction. A transaction is usually ended by
1960** either a COMMIT or a ROLLBACK operation. This routine may be called
1961** after rollback of a hot-journal, or if an error occurs while opening
1962** the journal file or writing the very first journal-header of a
1963** database transaction.
1964**
1965** This routine is never called in PAGER_ERROR state. If it is called
1966** in PAGER_NONE or PAGER_SHARED state and the lock held is less
1967** exclusive than a RESERVED lock, it is a no-op.
1968**
1969** Otherwise, any active savepoints are released.
1970**
1971** If the journal file is open, then it is "finalized". Once a journal
1972** file has been finalized it is not possible to use it to roll back a
1973** transaction. Nor will it be considered to be a hot-journal by this
1974** or any other database connection. Exactly how a journal is finalized
1975** depends on whether or not the pager is running in exclusive mode and
1976** the current journal-mode (Pager.journalMode value), as follows:
1977**
1978** journalMode==MEMORY
1979** Journal file descriptor is simply closed. This destroys an
1980** in-memory journal.
1981**
1982** journalMode==TRUNCATE
1983** Journal file is truncated to zero bytes in size.
1984**
1985** journalMode==PERSIST
1986** The first 28 bytes of the journal file are zeroed. This invalidates
1987** the first journal header in the file, and hence the entire journal
1988** file. An invalid journal file cannot be rolled back.
1989**
1990** journalMode==DELETE
1991** The journal file is closed and deleted using sqlite3OsDelete().
1992**
1993** If the pager is running in exclusive mode, this method of finalizing
1994** the journal file is never used. Instead, if the journalMode is
1995** DELETE and the pager is in exclusive mode, the method described under
1996** journalMode==PERSIST is used instead.
1997**
1998** After the journal is finalized, the pager moves to PAGER_READER state.
1999** If running in non-exclusive rollback mode, the lock on the file is
2000** downgraded to a SHARED_LOCK.
2001**
2002** SQLITE_OK is returned if no error occurs. If an error occurs during
2003** any of the IO operations to finalize the journal file or unlock the
2004** database then the IO error code is returned to the user. If the
2005** operation to finalize the journal file fails, then the code still
2006** tries to unlock the database file if not in exclusive mode. If the
2007** unlock operation fails as well, then the first error code related
2008** to the first error encountered (the journal finalization one) is
2009** returned.
2010*/
2011static int pager_end_transaction(Pager *pPager, int hasSuper, int bCommit){
2012 int rc = SQLITE_OK; /* Error code from journal finalization operation */
2013 int rc2 = SQLITE_OK; /* Error code from db file unlock operation */
2014
2015 /* Do nothing if the pager does not have an open write transaction
2016 ** or at least a RESERVED lock. This function may be called when there
2017 ** is no write-transaction active but a RESERVED or greater lock is
2018 ** held under two circumstances:
2019 **
2020 ** 1. After a successful hot-journal rollback, it is called with
2021 ** eState==PAGER_NONE and eLock==EXCLUSIVE_LOCK.
2022 **
2023 ** 2. If a connection with locking_mode=exclusive holding an EXCLUSIVE
2024 ** lock switches back to locking_mode=normal and then executes a
2025 ** read-transaction, this function is called with eState==PAGER_READER
2026 ** and eLock==EXCLUSIVE_LOCK when the read-transaction is closed.
2027 */
2028 assert( assert_pager_state(pPager) );
2029 assert( pPager->eState!=PAGER_ERROR );
2030 if( pPager->eState<PAGER_WRITER_LOCKED && pPager->eLock<RESERVED_LOCK ){
2031 return SQLITE_OK;
2032 }
2033
2034 releaseAllSavepoints(pPager);
2035 assert( isOpen(pPager->jfd) || pPager->pInJournal==0
2036 || (sqlite3OsDeviceCharacteristics(pPager->fd)&SQLITE_IOCAP_BATCH_ATOMIC)
2037 );
2038 if( isOpen(pPager->jfd) ){
2039 assert( !pagerUseWal(pPager) );
2040
2041 /* Finalize the journal file. */
2042 if( sqlite3JournalIsInMemory(pPager->jfd) ){
2043 /* assert( pPager->journalMode==PAGER_JOURNALMODE_MEMORY ); */
2044 sqlite3OsClose(pPager->jfd);
2045 }else if( pPager->journalMode==PAGER_JOURNALMODE_TRUNCATE ){
2046 if( pPager->journalOff==0 ){
2047 rc = SQLITE_OK;
2048 }else{
2049 rc = sqlite3OsTruncate(pPager->jfd, 0);
2050 if( rc==SQLITE_OK && pPager->fullSync ){
2051 /* Make sure the new file size is written into the inode right away.
2052 ** Otherwise the journal might resurrect following a power loss and
2053 ** cause the last transaction to roll back. See
2054 ** https://bugzilla.mozilla.org/show_bug.cgi?id=1072773
2055 */
2056 rc = sqlite3OsSync(pPager->jfd, pPager->syncFlags);
2057 }
2058 }
2059 pPager->journalOff = 0;
2060 }else if( pPager->journalMode==PAGER_JOURNALMODE_PERSIST
2061 || (pPager->exclusiveMode && pPager->journalMode!=PAGER_JOURNALMODE_WAL)
2062 ){
2063 rc = zeroJournalHdr(pPager, hasSuper||pPager->tempFile);
2064 pPager->journalOff = 0;
2065 }else{
2066 /* This branch may be executed with Pager.journalMode==MEMORY if
2067 ** a hot-journal was just rolled back. In this case the journal
2068 ** file should be closed and deleted. If this connection writes to
2069 ** the database file, it will do so using an in-memory journal.
2070 */
2071 int bDelete = !pPager->tempFile;
2072 assert( sqlite3JournalIsInMemory(pPager->jfd)==0 );
2073 assert( pPager->journalMode==PAGER_JOURNALMODE_DELETE
2074 || pPager->journalMode==PAGER_JOURNALMODE_MEMORY
2075 || pPager->journalMode==PAGER_JOURNALMODE_WAL
2076 );
2077 sqlite3OsClose(pPager->jfd);
2078 if( bDelete ){
2079 rc = sqlite3OsDelete(pPager->pVfs, pPager->zJournal, pPager->extraSync);
2080 }
2081 }
2082 }
2083
2084#ifdef SQLITE_CHECK_PAGES
2085 sqlite3PcacheIterateDirty(pPager->pPCache, pager_set_pagehash);
2086 if( pPager->dbSize==0 && sqlite3PcacheRefCount(pPager->pPCache)>0 ){
2087 PgHdr *p = sqlite3PagerLookup(pPager, 1);
2088 if( p ){
2089 p->pageHash = 0;
2090 sqlite3PagerUnrefNotNull(p);
2091 }
2092 }
2093#endif
2094
2095 sqlite3BitvecDestroy(pPager->pInJournal);
2096 pPager->pInJournal = 0;
2097 pPager->nRec = 0;
2098 if( rc==SQLITE_OK ){
2099 if( MEMDB || pagerFlushOnCommit(pPager, bCommit) ){
2100 sqlite3PcacheCleanAll(pPager->pPCache);
2101 }else{
2102 sqlite3PcacheClearWritable(pPager->pPCache);
2103 }
2104 sqlite3PcacheTruncate(pPager->pPCache, pPager->dbSize);
2105 }
2106
2107 if( pagerUseWal(pPager) ){
2108 /* Drop the WAL write-lock, if any. Also, if the connection was in
2109 ** locking_mode=exclusive mode but is no longer, drop the EXCLUSIVE
2110 ** lock held on the database file.
2111 */
2112 rc2 = sqlite3WalEndWriteTransaction(pPager->pWal);
2113 assert( rc2==SQLITE_OK );
2114 }else if( rc==SQLITE_OK && bCommit && pPager->dbFileSize>pPager->dbSize ){
2115 /* This branch is taken when committing a transaction in rollback-journal
2116 ** mode if the database file on disk is larger than the database image.
2117 ** At this point the journal has been finalized and the transaction
2118 ** successfully committed, but the EXCLUSIVE lock is still held on the
2119 ** file. So it is safe to truncate the database file to its minimum
2120 ** required size. */
2121 assert( pPager->eLock==EXCLUSIVE_LOCK );
2122 rc = pager_truncate(pPager, pPager->dbSize);
2123 }
2124
2125 if( rc==SQLITE_OK && bCommit ){
2126 rc = sqlite3OsFileControl(pPager->fd, SQLITE_FCNTL_COMMIT_PHASETWO, 0);
2127 if( rc==SQLITE_NOTFOUND ) rc = SQLITE_OK;
2128 }
2129
2130 if( !pPager->exclusiveMode
2131 && (!pagerUseWal(pPager) || sqlite3WalExclusiveMode(pPager->pWal, 0))
2132 ){
2133 rc2 = pagerUnlockDb(pPager, SHARED_LOCK);
2134 }
2135 pPager->eState = PAGER_READER;
2136 pPager->setSuper = 0;
2137
2138 return (rc==SQLITE_OK?rc2:rc);
2139}
2140
2141/*
2142** Execute a rollback if a transaction is active and unlock the
2143** database file.
2144**
2145** If the pager has already entered the ERROR state, do not attempt
2146** the rollback at this time. Instead, pager_unlock() is called. The
2147** call to pager_unlock() will discard all in-memory pages, unlock
2148** the database file and move the pager back to OPEN state. If this
2149** means that there is a hot-journal left in the file-system, the next
2150** connection to obtain a shared lock on the pager (which may be this one)
2151** will roll it back.
2152**
2153** If the pager has not already entered the ERROR state, but an IO or
2154** malloc error occurs during a rollback, then this will itself cause
2155** the pager to enter the ERROR state. Which will be cleared by the
2156** call to pager_unlock(), as described above.
2157*/
2158static void pagerUnlockAndRollback(Pager *pPager){
2159 if( pPager->eState!=PAGER_ERROR && pPager->eState!=PAGER_OPEN ){
2160 assert( assert_pager_state(pPager) );
2161 if( pPager->eState>=PAGER_WRITER_LOCKED ){
2162 sqlite3BeginBenignMalloc();
2163 sqlite3PagerRollback(pPager);
2164 sqlite3EndBenignMalloc();
2165 }else if( !pPager->exclusiveMode ){
2166 assert( pPager->eState==PAGER_READER );
2167 pager_end_transaction(pPager, 0, 0);
2168 }
2169 }
2170 pager_unlock(pPager);
2171}
2172
2173/*
2174** Parameter aData must point to a buffer of pPager->pageSize bytes
2175** of data. Compute and return a checksum based ont the contents of the
2176** page of data and the current value of pPager->cksumInit.
2177**
2178** This is not a real checksum. It is really just the sum of the
2179** random initial value (pPager->cksumInit) and every 200th byte
2180** of the page data, starting with byte offset (pPager->pageSize%200).
2181** Each byte is interpreted as an 8-bit unsigned integer.
2182**
2183** Changing the formula used to compute this checksum results in an
2184** incompatible journal file format.
2185**
2186** If journal corruption occurs due to a power failure, the most likely
2187** scenario is that one end or the other of the record will be changed.
2188** It is much less likely that the two ends of the journal record will be
2189** correct and the middle be corrupt. Thus, this "checksum" scheme,
2190** though fast and simple, catches the mostly likely kind of corruption.
2191*/
2192static u32 pager_cksum(Pager *pPager, const u8 *aData){
2193 u32 cksum = pPager->cksumInit; /* Checksum value to return */
2194 int i = pPager->pageSize-200; /* Loop counter */
2195 while( i>0 ){
2196 cksum += aData[i];
2197 i -= 200;
2198 }
2199 return cksum;
2200}
2201
2202/*
2203** Read a single page from either the journal file (if isMainJrnl==1) or
2204** from the sub-journal (if isMainJrnl==0) and playback that page.
2205** The page begins at offset *pOffset into the file. The *pOffset
2206** value is increased to the start of the next page in the journal.
2207**
2208** The main rollback journal uses checksums - the statement journal does
2209** not.
2210**
2211** If the page number of the page record read from the (sub-)journal file
2212** is greater than the current value of Pager.dbSize, then playback is
2213** skipped and SQLITE_OK is returned.
2214**
2215** If pDone is not NULL, then it is a record of pages that have already
2216** been played back. If the page at *pOffset has already been played back
2217** (if the corresponding pDone bit is set) then skip the playback.
2218** Make sure the pDone bit corresponding to the *pOffset page is set
2219** prior to returning.
2220**
2221** If the page record is successfully read from the (sub-)journal file
2222** and played back, then SQLITE_OK is returned. If an IO error occurs
2223** while reading the record from the (sub-)journal file or while writing
2224** to the database file, then the IO error code is returned. If data
2225** is successfully read from the (sub-)journal file but appears to be
2226** corrupted, SQLITE_DONE is returned. Data is considered corrupted in
2227** two circumstances:
2228**
2229** * If the record page-number is illegal (0 or PAGER_SJ_PGNO), or
2230** * If the record is being rolled back from the main journal file
2231** and the checksum field does not match the record content.
2232**
2233** Neither of these two scenarios are possible during a savepoint rollback.
2234**
2235** If this is a savepoint rollback, then memory may have to be dynamically
2236** allocated by this function. If this is the case and an allocation fails,
2237** SQLITE_NOMEM is returned.
2238*/
2239static int pager_playback_one_page(
2240 Pager *pPager, /* The pager being played back */
2241 i64 *pOffset, /* Offset of record to playback */
2242 Bitvec *pDone, /* Bitvec of pages already played back */
2243 int isMainJrnl, /* 1 -> main journal. 0 -> sub-journal. */
2244 int isSavepnt /* True for a savepoint rollback */
2245){
2246 int rc;
2247 PgHdr *pPg; /* An existing page in the cache */
2248 Pgno pgno; /* The page number of a page in journal */
2249 u32 cksum; /* Checksum used for sanity checking */
2250 char *aData; /* Temporary storage for the page */
2251 sqlite3_file *jfd; /* The file descriptor for the journal file */
2252 int isSynced; /* True if journal page is synced */
2253
2254 assert( (isMainJrnl&~1)==0 ); /* isMainJrnl is 0 or 1 */
2255 assert( (isSavepnt&~1)==0 ); /* isSavepnt is 0 or 1 */
2256 assert( isMainJrnl || pDone ); /* pDone always used on sub-journals */
2257 assert( isSavepnt || pDone==0 ); /* pDone never used on non-savepoint */
2258
2259 aData = pPager->pTmpSpace;
2260 assert( aData ); /* Temp storage must have already been allocated */
2261 assert( pagerUseWal(pPager)==0 || (!isMainJrnl && isSavepnt) );
2262
2263 /* Either the state is greater than PAGER_WRITER_CACHEMOD (a transaction
2264 ** or savepoint rollback done at the request of the caller) or this is
2265 ** a hot-journal rollback. If it is a hot-journal rollback, the pager
2266 ** is in state OPEN and holds an EXCLUSIVE lock. Hot-journal rollback
2267 ** only reads from the main journal, not the sub-journal.
2268 */
2269 assert( pPager->eState>=PAGER_WRITER_CACHEMOD
2270 || (pPager->eState==PAGER_OPEN && pPager->eLock==EXCLUSIVE_LOCK)
2271 );
2272 assert( pPager->eState>=PAGER_WRITER_CACHEMOD || isMainJrnl );
2273
2274 /* Read the page number and page data from the journal or sub-journal
2275 ** file. Return an error code to the caller if an IO error occurs.
2276 */
2277 jfd = isMainJrnl ? pPager->jfd : pPager->sjfd;
2278 rc = read32bits(jfd, *pOffset, &pgno);
2279 if( rc!=SQLITE_OK ) return rc;
2280 rc = sqlite3OsRead(jfd, (u8*)aData, pPager->pageSize, (*pOffset)+4);
2281 if( rc!=SQLITE_OK ) return rc;
2282 *pOffset += pPager->pageSize + 4 + isMainJrnl*4;
2283
2284 /* Sanity checking on the page. This is more important that I originally
2285 ** thought. If a power failure occurs while the journal is being written,
2286 ** it could cause invalid data to be written into the journal. We need to
2287 ** detect this invalid data (with high probability) and ignore it.
2288 */
2289 if( pgno==0 || pgno==PAGER_SJ_PGNO(pPager) ){
2290 assert( !isSavepnt );
2291 return SQLITE_DONE;
2292 }
2293 if( pgno>(Pgno)pPager->dbSize || sqlite3BitvecTest(pDone, pgno) ){
2294 return SQLITE_OK;
2295 }
2296 if( isMainJrnl ){
2297 rc = read32bits(jfd, (*pOffset)-4, &cksum);
2298 if( rc ) return rc;
2299 if( !isSavepnt && pager_cksum(pPager, (u8*)aData)!=cksum ){
2300 return SQLITE_DONE;
2301 }
2302 }
2303
2304 /* If this page has already been played back before during the current
2305 ** rollback, then don't bother to play it back again.
2306 */
2307 if( pDone && (rc = sqlite3BitvecSet(pDone, pgno))!=SQLITE_OK ){
2308 return rc;
2309 }
2310
2311 /* When playing back page 1, restore the nReserve setting
2312 */
2313 if( pgno==1 && pPager->nReserve!=((u8*)aData)[20] ){
2314 pPager->nReserve = ((u8*)aData)[20];
2315 }
2316
2317 /* If the pager is in CACHEMOD state, then there must be a copy of this
2318 ** page in the pager cache. In this case just update the pager cache,
2319 ** not the database file. The page is left marked dirty in this case.
2320 **
2321 ** An exception to the above rule: If the database is in no-sync mode
2322 ** and a page is moved during an incremental vacuum then the page may
2323 ** not be in the pager cache. Later: if a malloc() or IO error occurs
2324 ** during a Movepage() call, then the page may not be in the cache
2325 ** either. So the condition described in the above paragraph is not
2326 ** assert()able.
2327 **
2328 ** If in WRITER_DBMOD, WRITER_FINISHED or OPEN state, then we update the
2329 ** pager cache if it exists and the main file. The page is then marked
2330 ** not dirty. Since this code is only executed in PAGER_OPEN state for
2331 ** a hot-journal rollback, it is guaranteed that the page-cache is empty
2332 ** if the pager is in OPEN state.
2333 **
2334 ** Ticket #1171: The statement journal might contain page content that is
2335 ** different from the page content at the start of the transaction.
2336 ** This occurs when a page is changed prior to the start of a statement
2337 ** then changed again within the statement. When rolling back such a
2338 ** statement we must not write to the original database unless we know
2339 ** for certain that original page contents are synced into the main rollback
2340 ** journal. Otherwise, a power loss might leave modified data in the
2341 ** database file without an entry in the rollback journal that can
2342 ** restore the database to its original form. Two conditions must be
2343 ** met before writing to the database files. (1) the database must be
2344 ** locked. (2) we know that the original page content is fully synced
2345 ** in the main journal either because the page is not in cache or else
2346 ** the page is marked as needSync==0.
2347 **
2348 ** 2008-04-14: When attempting to vacuum a corrupt database file, it
2349 ** is possible to fail a statement on a database that does not yet exist.
2350 ** Do not attempt to write if database file has never been opened.
2351 */
2352 if( pagerUseWal(pPager) ){
2353 pPg = 0;
2354 }else{
2355 pPg = sqlite3PagerLookup(pPager, pgno);
2356 }
2357 assert( pPg || !MEMDB );
2358 assert( pPager->eState!=PAGER_OPEN || pPg==0 || pPager->tempFile );
2359 PAGERTRACE(("PLAYBACK %d page %d hash(%08x) %s\n",
2360 PAGERID(pPager), pgno, pager_datahash(pPager->pageSize, (u8*)aData),
2361 (isMainJrnl?"main-journal":"sub-journal")
2362 ));
2363 if( isMainJrnl ){
2364 isSynced = pPager->noSync || (*pOffset <= pPager->journalHdr);
2365 }else{
2366 isSynced = (pPg==0 || 0==(pPg->flags & PGHDR_NEED_SYNC));
2367 }
2368 if( isOpen(pPager->fd)
2369 && (pPager->eState>=PAGER_WRITER_DBMOD || pPager->eState==PAGER_OPEN)
2370 && isSynced
2371 ){
2372 i64 ofst = (pgno-1)*(i64)pPager->pageSize;
2373 testcase( !isSavepnt && pPg!=0 && (pPg->flags&PGHDR_NEED_SYNC)!=0 );
2374 assert( !pagerUseWal(pPager) );
2375
2376 /* Write the data read from the journal back into the database file.
2377 ** This is usually safe even for an encrypted database - as the data
2378 ** was encrypted before it was written to the journal file. The exception
2379 ** is if the data was just read from an in-memory sub-journal. In that
2380 ** case it must be encrypted here before it is copied into the database
2381 ** file. */
2382 rc = sqlite3OsWrite(pPager->fd, (u8 *)aData, pPager->pageSize, ofst);
2383
2384 if( pgno>pPager->dbFileSize ){
2385 pPager->dbFileSize = pgno;
2386 }
2387 if( pPager->pBackup ){
2388 sqlite3BackupUpdate(pPager->pBackup, pgno, (u8*)aData);
2389 }
2390 }else if( !isMainJrnl && pPg==0 ){
2391 /* If this is a rollback of a savepoint and data was not written to
2392 ** the database and the page is not in-memory, there is a potential
2393 ** problem. When the page is next fetched by the b-tree layer, it
2394 ** will be read from the database file, which may or may not be
2395 ** current.
2396 **
2397 ** There are a couple of different ways this can happen. All are quite
2398 ** obscure. When running in synchronous mode, this can only happen
2399 ** if the page is on the free-list at the start of the transaction, then
2400 ** populated, then moved using sqlite3PagerMovepage().
2401 **
2402 ** The solution is to add an in-memory page to the cache containing
2403 ** the data just read from the sub-journal. Mark the page as dirty
2404 ** and if the pager requires a journal-sync, then mark the page as
2405 ** requiring a journal-sync before it is written.
2406 */
2407 assert( isSavepnt );
2408 assert( (pPager->doNotSpill & SPILLFLAG_ROLLBACK)==0 );
2409 pPager->doNotSpill |= SPILLFLAG_ROLLBACK;
2410 rc = sqlite3PagerGet(pPager, pgno, &pPg, 1);
2411 assert( (pPager->doNotSpill & SPILLFLAG_ROLLBACK)!=0 );
2412 pPager->doNotSpill &= ~SPILLFLAG_ROLLBACK;
2413 if( rc!=SQLITE_OK ) return rc;
2414 sqlite3PcacheMakeDirty(pPg);
2415 }
2416 if( pPg ){
2417 /* No page should ever be explicitly rolled back that is in use, except
2418 ** for page 1 which is held in use in order to keep the lock on the
2419 ** database active. However such a page may be rolled back as a result
2420 ** of an internal error resulting in an automatic call to
2421 ** sqlite3PagerRollback().
2422 */
2423 void *pData;
2424 pData = pPg->pData;
2425 memcpy(pData, (u8*)aData, pPager->pageSize);
2426 pPager->xReiniter(pPg);
2427 /* It used to be that sqlite3PcacheMakeClean(pPg) was called here. But
2428 ** that call was dangerous and had no detectable benefit since the cache
2429 ** is normally cleaned by sqlite3PcacheCleanAll() after rollback and so
2430 ** has been removed. */
2431 pager_set_pagehash(pPg);
2432
2433 /* If this was page 1, then restore the value of Pager.dbFileVers.
2434 ** Do this before any decoding. */
2435 if( pgno==1 ){
2436 memcpy(&pPager->dbFileVers, &((u8*)pData)[24],sizeof(pPager->dbFileVers));
2437 }
2438 sqlite3PcacheRelease(pPg);
2439 }
2440 return rc;
2441}
2442
2443/*
2444** Parameter zSuper is the name of a super-journal file. A single journal
2445** file that referred to the super-journal file has just been rolled back.
2446** This routine checks if it is possible to delete the super-journal file,
2447** and does so if it is.
2448**
2449** Argument zSuper may point to Pager.pTmpSpace. So that buffer is not
2450** available for use within this function.
2451**
2452** When a super-journal file is created, it is populated with the names
2453** of all of its child journals, one after another, formatted as utf-8
2454** encoded text. The end of each child journal file is marked with a
2455** nul-terminator byte (0x00). i.e. the entire contents of a super-journal
2456** file for a transaction involving two databases might be:
2457**
2458** "/home/bill/a.db-journal\x00/home/bill/b.db-journal\x00"
2459**
2460** A super-journal file may only be deleted once all of its child
2461** journals have been rolled back.
2462**
2463** This function reads the contents of the super-journal file into
2464** memory and loops through each of the child journal names. For
2465** each child journal, it checks if:
2466**
2467** * if the child journal exists, and if so
2468** * if the child journal contains a reference to super-journal
2469** file zSuper
2470**
2471** If a child journal can be found that matches both of the criteria
2472** above, this function returns without doing anything. Otherwise, if
2473** no such child journal can be found, file zSuper is deleted from
2474** the file-system using sqlite3OsDelete().
2475**
2476** If an IO error within this function, an error code is returned. This
2477** function allocates memory by calling sqlite3Malloc(). If an allocation
2478** fails, SQLITE_NOMEM is returned. Otherwise, if no IO or malloc errors
2479** occur, SQLITE_OK is returned.
2480**
2481** TODO: This function allocates a single block of memory to load
2482** the entire contents of the super-journal file. This could be
2483** a couple of kilobytes or so - potentially larger than the page
2484** size.
2485*/
2486static int pager_delsuper(Pager *pPager, const char *zSuper){
2487 sqlite3_vfs *pVfs = pPager->pVfs;
2488 int rc; /* Return code */
2489 sqlite3_file *pSuper; /* Malloc'd super-journal file descriptor */
2490 sqlite3_file *pJournal; /* Malloc'd child-journal file descriptor */
2491 char *zSuperJournal = 0; /* Contents of super-journal file */
2492 i64 nSuperJournal; /* Size of super-journal file */
2493 char *zJournal; /* Pointer to one journal within MJ file */
2494 char *zSuperPtr; /* Space to hold super-journal filename */
2495 char *zFree = 0; /* Free this buffer */
2496 int nSuperPtr; /* Amount of space allocated to zSuperPtr[] */
2497
2498 /* Allocate space for both the pJournal and pSuper file descriptors.
2499 ** If successful, open the super-journal file for reading.
2500 */
2501 pSuper = (sqlite3_file *)sqlite3MallocZero(pVfs->szOsFile * 2);
2502 if( !pSuper ){
2503 rc = SQLITE_NOMEM_BKPT;
2504 pJournal = 0;
2505 }else{
2506 const int flags = (SQLITE_OPEN_READONLY|SQLITE_OPEN_SUPER_JOURNAL);
2507 rc = sqlite3OsOpen(pVfs, zSuper, pSuper, flags, 0);
2508 pJournal = (sqlite3_file *)(((u8 *)pSuper) + pVfs->szOsFile);
2509 }
2510 if( rc!=SQLITE_OK ) goto delsuper_out;
2511
2512 /* Load the entire super-journal file into space obtained from
2513 ** sqlite3_malloc() and pointed to by zSuperJournal. Also obtain
2514 ** sufficient space (in zSuperPtr) to hold the names of super-journal
2515 ** files extracted from regular rollback-journals.
2516 */
2517 rc = sqlite3OsFileSize(pSuper, &nSuperJournal);
2518 if( rc!=SQLITE_OK ) goto delsuper_out;
2519 nSuperPtr = pVfs->mxPathname+1;
2520 zFree = sqlite3Malloc(4 + nSuperJournal + nSuperPtr + 2);
2521 if( !zFree ){
2522 rc = SQLITE_NOMEM_BKPT;
2523 goto delsuper_out;
2524 }
2525 zFree[0] = zFree[1] = zFree[2] = zFree[3] = 0;
2526 zSuperJournal = &zFree[4];
2527 zSuperPtr = &zSuperJournal[nSuperJournal+2];
2528 rc = sqlite3OsRead(pSuper, zSuperJournal, (int)nSuperJournal, 0);
2529 if( rc!=SQLITE_OK ) goto delsuper_out;
2530 zSuperJournal[nSuperJournal] = 0;
2531 zSuperJournal[nSuperJournal+1] = 0;
2532
2533 zJournal = zSuperJournal;
2534 while( (zJournal-zSuperJournal)<nSuperJournal ){
2535 int exists;
2536 rc = sqlite3OsAccess(pVfs, zJournal, SQLITE_ACCESS_EXISTS, &exists);
2537 if( rc!=SQLITE_OK ){
2538 goto delsuper_out;
2539 }
2540 if( exists ){
2541 /* One of the journals pointed to by the super-journal exists.
2542 ** Open it and check if it points at the super-journal. If
2543 ** so, return without deleting the super-journal file.
2544 ** NB: zJournal is really a MAIN_JOURNAL. But call it a
2545 ** SUPER_JOURNAL here so that the VFS will not send the zJournal
2546 ** name into sqlite3_database_file_object().
2547 */
2548 int c;
2549 int flags = (SQLITE_OPEN_READONLY|SQLITE_OPEN_SUPER_JOURNAL);
2550 rc = sqlite3OsOpen(pVfs, zJournal, pJournal, flags, 0);
2551 if( rc!=SQLITE_OK ){
2552 goto delsuper_out;
2553 }
2554
2555 rc = readSuperJournal(pJournal, zSuperPtr, nSuperPtr);
2556 sqlite3OsClose(pJournal);
2557 if( rc!=SQLITE_OK ){
2558 goto delsuper_out;
2559 }
2560
2561 c = zSuperPtr[0]!=0 && strcmp(zSuperPtr, zSuper)==0;
2562 if( c ){
2563 /* We have a match. Do not delete the super-journal file. */
2564 goto delsuper_out;
2565 }
2566 }
2567 zJournal += (sqlite3Strlen30(zJournal)+1);
2568 }
2569
2570 sqlite3OsClose(pSuper);
2571 rc = sqlite3OsDelete(pVfs, zSuper, 0);
2572
2573delsuper_out:
2574 sqlite3_free(zFree);
2575 if( pSuper ){
2576 sqlite3OsClose(pSuper);
2577 assert( !isOpen(pJournal) );
2578 sqlite3_free(pSuper);
2579 }
2580 return rc;
2581}
2582
2583
2584/*
2585** This function is used to change the actual size of the database
2586** file in the file-system. This only happens when committing a transaction,
2587** or rolling back a transaction (including rolling back a hot-journal).
2588**
2589** If the main database file is not open, or the pager is not in either
2590** DBMOD or OPEN state, this function is a no-op. Otherwise, the size
2591** of the file is changed to nPage pages (nPage*pPager->pageSize bytes).
2592** If the file on disk is currently larger than nPage pages, then use the VFS
2593** xTruncate() method to truncate it.
2594**
2595** Or, it might be the case that the file on disk is smaller than
2596** nPage pages. Some operating system implementations can get confused if
2597** you try to truncate a file to some size that is larger than it
2598** currently is, so detect this case and write a single zero byte to
2599** the end of the new file instead.
2600**
2601** If successful, return SQLITE_OK. If an IO error occurs while modifying
2602** the database file, return the error code to the caller.
2603*/
2604static int pager_truncate(Pager *pPager, Pgno nPage){
2605 int rc = SQLITE_OK;
2606 assert( pPager->eState!=PAGER_ERROR );
2607 assert( pPager->eState!=PAGER_READER );
2608
2609 if( isOpen(pPager->fd)
2610 && (pPager->eState>=PAGER_WRITER_DBMOD || pPager->eState==PAGER_OPEN)
2611 ){
2612 i64 currentSize, newSize;
2613 int szPage = pPager->pageSize;
2614 assert( pPager->eLock==EXCLUSIVE_LOCK );
2615 /* TODO: Is it safe to use Pager.dbFileSize here? */
2616 rc = sqlite3OsFileSize(pPager->fd, &currentSize);
2617 newSize = szPage*(i64)nPage;
2618 if( rc==SQLITE_OK && currentSize!=newSize ){
2619 if( currentSize>newSize ){
2620 rc = sqlite3OsTruncate(pPager->fd, newSize);
2621 }else if( (currentSize+szPage)<=newSize ){
2622 char *pTmp = pPager->pTmpSpace;
2623 memset(pTmp, 0, szPage);
2624 testcase( (newSize-szPage) == currentSize );
2625 testcase( (newSize-szPage) > currentSize );
2626 sqlite3OsFileControlHint(pPager->fd, SQLITE_FCNTL_SIZE_HINT, &newSize);
2627 rc = sqlite3OsWrite(pPager->fd, pTmp, szPage, newSize-szPage);
2628 }
2629 if( rc==SQLITE_OK ){
2630 pPager->dbFileSize = nPage;
2631 }
2632 }
2633 }
2634 return rc;
2635}
2636
2637/*
2638** Return a sanitized version of the sector-size of OS file pFile. The
2639** return value is guaranteed to lie between 32 and MAX_SECTOR_SIZE.
2640*/
2641int sqlite3SectorSize(sqlite3_file *pFile){
2642 int iRet = sqlite3OsSectorSize(pFile);
2643 if( iRet<32 ){
2644 iRet = 512;
2645 }else if( iRet>MAX_SECTOR_SIZE ){
2646 assert( MAX_SECTOR_SIZE>=512 );
2647 iRet = MAX_SECTOR_SIZE;
2648 }
2649 return iRet;
2650}
2651
2652/*
2653** Set the value of the Pager.sectorSize variable for the given
2654** pager based on the value returned by the xSectorSize method
2655** of the open database file. The sector size will be used
2656** to determine the size and alignment of journal header and
2657** super-journal pointers within created journal files.
2658**
2659** For temporary files the effective sector size is always 512 bytes.
2660**
2661** Otherwise, for non-temporary files, the effective sector size is
2662** the value returned by the xSectorSize() method rounded up to 32 if
2663** it is less than 32, or rounded down to MAX_SECTOR_SIZE if it
2664** is greater than MAX_SECTOR_SIZE.
2665**
2666** If the file has the SQLITE_IOCAP_POWERSAFE_OVERWRITE property, then set
2667** the effective sector size to its minimum value (512). The purpose of
2668** pPager->sectorSize is to define the "blast radius" of bytes that
2669** might change if a crash occurs while writing to a single byte in
2670** that range. But with POWERSAFE_OVERWRITE, the blast radius is zero
2671** (that is what POWERSAFE_OVERWRITE means), so we minimize the sector
2672** size. For backwards compatibility of the rollback journal file format,
2673** we cannot reduce the effective sector size below 512.
2674*/
2675static void setSectorSize(Pager *pPager){
2676 assert( isOpen(pPager->fd) || pPager->tempFile );
2677
2678 if( pPager->tempFile
2679 || (sqlite3OsDeviceCharacteristics(pPager->fd) &
2680 SQLITE_IOCAP_POWERSAFE_OVERWRITE)!=0
2681 ){
2682 /* Sector size doesn't matter for temporary files. Also, the file
2683 ** may not have been opened yet, in which case the OsSectorSize()
2684 ** call will segfault. */
2685 pPager->sectorSize = 512;
2686 }else{
2687 pPager->sectorSize = sqlite3SectorSize(pPager->fd);
2688 }
2689}
2690
2691/*
2692** Playback the journal and thus restore the database file to
2693** the state it was in before we started making changes.
2694**
2695** The journal file format is as follows:
2696**
2697** (1) 8 byte prefix. A copy of aJournalMagic[].
2698** (2) 4 byte big-endian integer which is the number of valid page records
2699** in the journal. If this value is 0xffffffff, then compute the
2700** number of page records from the journal size.
2701** (3) 4 byte big-endian integer which is the initial value for the
2702** sanity checksum.
2703** (4) 4 byte integer which is the number of pages to truncate the
2704** database to during a rollback.
2705** (5) 4 byte big-endian integer which is the sector size. The header
2706** is this many bytes in size.
2707** (6) 4 byte big-endian integer which is the page size.
2708** (7) zero padding out to the next sector size.
2709** (8) Zero or more pages instances, each as follows:
2710** + 4 byte page number.
2711** + pPager->pageSize bytes of data.
2712** + 4 byte checksum
2713**
2714** When we speak of the journal header, we mean the first 7 items above.
2715** Each entry in the journal is an instance of the 8th item.
2716**
2717** Call the value from the second bullet "nRec". nRec is the number of
2718** valid page entries in the journal. In most cases, you can compute the
2719** value of nRec from the size of the journal file. But if a power
2720** failure occurred while the journal was being written, it could be the
2721** case that the size of the journal file had already been increased but
2722** the extra entries had not yet made it safely to disk. In such a case,
2723** the value of nRec computed from the file size would be too large. For
2724** that reason, we always use the nRec value in the header.
2725**
2726** If the nRec value is 0xffffffff it means that nRec should be computed
2727** from the file size. This value is used when the user selects the
2728** no-sync option for the journal. A power failure could lead to corruption
2729** in this case. But for things like temporary table (which will be
2730** deleted when the power is restored) we don't care.
2731**
2732** If the file opened as the journal file is not a well-formed
2733** journal file then all pages up to the first corrupted page are rolled
2734** back (or no pages if the journal header is corrupted). The journal file
2735** is then deleted and SQLITE_OK returned, just as if no corruption had
2736** been encountered.
2737**
2738** If an I/O or malloc() error occurs, the journal-file is not deleted
2739** and an error code is returned.
2740**
2741** The isHot parameter indicates that we are trying to rollback a journal
2742** that might be a hot journal. Or, it could be that the journal is
2743** preserved because of JOURNALMODE_PERSIST or JOURNALMODE_TRUNCATE.
2744** If the journal really is hot, reset the pager cache prior rolling
2745** back any content. If the journal is merely persistent, no reset is
2746** needed.
2747*/
2748static int pager_playback(Pager *pPager, int isHot){
2749 sqlite3_vfs *pVfs = pPager->pVfs;
2750 i64 szJ; /* Size of the journal file in bytes */
2751 u32 nRec; /* Number of Records in the journal */
2752 u32 u; /* Unsigned loop counter */
2753 Pgno mxPg = 0; /* Size of the original file in pages */
2754 int rc; /* Result code of a subroutine */
2755 int res = 1; /* Value returned by sqlite3OsAccess() */
2756 char *zSuper = 0; /* Name of super-journal file if any */
2757 int needPagerReset; /* True to reset page prior to first page rollback */
2758 int nPlayback = 0; /* Total number of pages restored from journal */
2759 u32 savedPageSize = pPager->pageSize;
2760
2761 /* Figure out how many records are in the journal. Abort early if
2762 ** the journal is empty.
2763 */
2764 assert( isOpen(pPager->jfd) );
2765 rc = sqlite3OsFileSize(pPager->jfd, &szJ);
2766 if( rc!=SQLITE_OK ){
2767 goto end_playback;
2768 }
2769
2770 /* Read the super-journal name from the journal, if it is present.
2771 ** If a super-journal file name is specified, but the file is not
2772 ** present on disk, then the journal is not hot and does not need to be
2773 ** played back.
2774 **
2775 ** TODO: Technically the following is an error because it assumes that
2776 ** buffer Pager.pTmpSpace is (mxPathname+1) bytes or larger. i.e. that
2777 ** (pPager->pageSize >= pPager->pVfs->mxPathname+1). Using os_unix.c,
2778 ** mxPathname is 512, which is the same as the minimum allowable value
2779 ** for pageSize.
2780 */
2781 zSuper = pPager->pTmpSpace;
2782 rc = readSuperJournal(pPager->jfd, zSuper, pPager->pVfs->mxPathname+1);
2783 if( rc==SQLITE_OK && zSuper[0] ){
2784 rc = sqlite3OsAccess(pVfs, zSuper, SQLITE_ACCESS_EXISTS, &res);
2785 }
2786 zSuper = 0;
2787 if( rc!=SQLITE_OK || !res ){
2788 goto end_playback;
2789 }
2790 pPager->journalOff = 0;
2791 needPagerReset = isHot;
2792
2793 /* This loop terminates either when a readJournalHdr() or
2794 ** pager_playback_one_page() call returns SQLITE_DONE or an IO error
2795 ** occurs.
2796 */
2797 while( 1 ){
2798 /* Read the next journal header from the journal file. If there are
2799 ** not enough bytes left in the journal file for a complete header, or
2800 ** it is corrupted, then a process must have failed while writing it.
2801 ** This indicates nothing more needs to be rolled back.
2802 */
2803 rc = readJournalHdr(pPager, isHot, szJ, &nRec, &mxPg);
2804 if( rc!=SQLITE_OK ){
2805 if( rc==SQLITE_DONE ){
2806 rc = SQLITE_OK;
2807 }
2808 goto end_playback;
2809 }
2810
2811 /* If nRec is 0xffffffff, then this journal was created by a process
2812 ** working in no-sync mode. This means that the rest of the journal
2813 ** file consists of pages, there are no more journal headers. Compute
2814 ** the value of nRec based on this assumption.
2815 */
2816 if( nRec==0xffffffff ){
2817 assert( pPager->journalOff==JOURNAL_HDR_SZ(pPager) );
2818 nRec = (int)((szJ - JOURNAL_HDR_SZ(pPager))/JOURNAL_PG_SZ(pPager));
2819 }
2820
2821 /* If nRec is 0 and this rollback is of a transaction created by this
2822 ** process and if this is the final header in the journal, then it means
2823 ** that this part of the journal was being filled but has not yet been
2824 ** synced to disk. Compute the number of pages based on the remaining
2825 ** size of the file.
2826 **
2827 ** The third term of the test was added to fix ticket #2565.
2828 ** When rolling back a hot journal, nRec==0 always means that the next
2829 ** chunk of the journal contains zero pages to be rolled back. But
2830 ** when doing a ROLLBACK and the nRec==0 chunk is the last chunk in
2831 ** the journal, it means that the journal might contain additional
2832 ** pages that need to be rolled back and that the number of pages
2833 ** should be computed based on the journal file size.
2834 */
2835 if( nRec==0 && !isHot &&
2836 pPager->journalHdr+JOURNAL_HDR_SZ(pPager)==pPager->journalOff ){
2837 nRec = (int)((szJ - pPager->journalOff) / JOURNAL_PG_SZ(pPager));
2838 }
2839
2840 /* If this is the first header read from the journal, truncate the
2841 ** database file back to its original size.
2842 */
2843 if( pPager->journalOff==JOURNAL_HDR_SZ(pPager) ){
2844 rc = pager_truncate(pPager, mxPg);
2845 if( rc!=SQLITE_OK ){
2846 goto end_playback;
2847 }
2848 pPager->dbSize = mxPg;
2849 if( pPager->mxPgno<mxPg ){
2850 pPager->mxPgno = mxPg;
2851 }
2852 }
2853
2854 /* Copy original pages out of the journal and back into the
2855 ** database file and/or page cache.
2856 */
2857 for(u=0; u<nRec; u++){
2858 if( needPagerReset ){
2859 pager_reset(pPager);
2860 needPagerReset = 0;
2861 }
2862 rc = pager_playback_one_page(pPager,&pPager->journalOff,0,1,0);
2863 if( rc==SQLITE_OK ){
2864 nPlayback++;
2865 }else{
2866 if( rc==SQLITE_DONE ){
2867 pPager->journalOff = szJ;
2868 break;
2869 }else if( rc==SQLITE_IOERR_SHORT_READ ){
2870 /* If the journal has been truncated, simply stop reading and
2871 ** processing the journal. This might happen if the journal was
2872 ** not completely written and synced prior to a crash. In that
2873 ** case, the database should have never been written in the
2874 ** first place so it is OK to simply abandon the rollback. */
2875 rc = SQLITE_OK;
2876 goto end_playback;
2877 }else{
2878 /* If we are unable to rollback, quit and return the error
2879 ** code. This will cause the pager to enter the error state
2880 ** so that no further harm will be done. Perhaps the next
2881 ** process to come along will be able to rollback the database.
2882 */
2883 goto end_playback;
2884 }
2885 }
2886 }
2887 }
2888 /*NOTREACHED*/
2889 assert( 0 );
2890
2891end_playback:
2892 if( rc==SQLITE_OK ){
2893 rc = sqlite3PagerSetPagesize(pPager, &savedPageSize, -1);
2894 }
2895 /* Following a rollback, the database file should be back in its original
2896 ** state prior to the start of the transaction, so invoke the
2897 ** SQLITE_FCNTL_DB_UNCHANGED file-control method to disable the
2898 ** assertion that the transaction counter was modified.
2899 */
2900#ifdef SQLITE_DEBUG
2901 sqlite3OsFileControlHint(pPager->fd,SQLITE_FCNTL_DB_UNCHANGED,0);
2902#endif
2903
2904 /* If this playback is happening automatically as a result of an IO or
2905 ** malloc error that occurred after the change-counter was updated but
2906 ** before the transaction was committed, then the change-counter
2907 ** modification may just have been reverted. If this happens in exclusive
2908 ** mode, then subsequent transactions performed by the connection will not
2909 ** update the change-counter at all. This may lead to cache inconsistency
2910 ** problems for other processes at some point in the future. So, just
2911 ** in case this has happened, clear the changeCountDone flag now.
2912 */
2913 pPager->changeCountDone = pPager->tempFile;
2914
2915 if( rc==SQLITE_OK ){
2916 /* Leave 4 bytes of space before the super-journal filename in memory.
2917 ** This is because it may end up being passed to sqlite3OsOpen(), in
2918 ** which case it requires 4 0x00 bytes in memory immediately before
2919 ** the filename. */
2920 zSuper = &pPager->pTmpSpace[4];
2921 rc = readSuperJournal(pPager->jfd, zSuper, pPager->pVfs->mxPathname+1);
2922 testcase( rc!=SQLITE_OK );
2923 }
2924 if( rc==SQLITE_OK
2925 && (pPager->eState>=PAGER_WRITER_DBMOD || pPager->eState==PAGER_OPEN)
2926 ){
2927 rc = sqlite3PagerSync(pPager, 0);
2928 }
2929 if( rc==SQLITE_OK ){
2930 rc = pager_end_transaction(pPager, zSuper[0]!='\0', 0);
2931 testcase( rc!=SQLITE_OK );
2932 }
2933 if( rc==SQLITE_OK && zSuper[0] && res ){
2934 /* If there was a super-journal and this routine will return success,
2935 ** see if it is possible to delete the super-journal.
2936 */
2937 assert( zSuper==&pPager->pTmpSpace[4] );
2938 memset(&zSuper[-4], 0, 4);
2939 rc = pager_delsuper(pPager, zSuper);
2940 testcase( rc!=SQLITE_OK );
2941 }
2942 if( isHot && nPlayback ){
2943 sqlite3_log(SQLITE_NOTICE_RECOVER_ROLLBACK, "recovered %d pages from %s",
2944 nPlayback, pPager->zJournal);
2945 }
2946
2947 /* The Pager.sectorSize variable may have been updated while rolling
2948 ** back a journal created by a process with a different sector size
2949 ** value. Reset it to the correct value for this process.
2950 */
2951 setSectorSize(pPager);
2952 return rc;
2953}
2954
2955
2956/*
2957** Read the content for page pPg out of the database file (or out of
2958** the WAL if that is where the most recent copy if found) into
2959** pPg->pData. A shared lock or greater must be held on the database
2960** file before this function is called.
2961**
2962** If page 1 is read, then the value of Pager.dbFileVers[] is set to
2963** the value read from the database file.
2964**
2965** If an IO error occurs, then the IO error is returned to the caller.
2966** Otherwise, SQLITE_OK is returned.
2967*/
2968static int readDbPage(PgHdr *pPg){
2969 Pager *pPager = pPg->pPager; /* Pager object associated with page pPg */
2970 int rc = SQLITE_OK; /* Return code */
2971
2972#ifndef SQLITE_OMIT_WAL
2973 u32 iFrame = 0; /* Frame of WAL containing pgno */
2974
2975 assert( pPager->eState>=PAGER_READER && !MEMDB );
2976 assert( isOpen(pPager->fd) );
2977
2978 if( pagerUseWal(pPager) ){
2979 rc = sqlite3WalFindFrame(pPager->pWal, pPg->pgno, &iFrame);
2980 if( rc ) return rc;
2981 }
2982 if( iFrame ){
2983 rc = sqlite3WalReadFrame(pPager->pWal, iFrame,pPager->pageSize,pPg->pData);
2984 }else
2985#endif
2986 {
2987 i64 iOffset = (pPg->pgno-1)*(i64)pPager->pageSize;
2988 rc = sqlite3OsRead(pPager->fd, pPg->pData, pPager->pageSize, iOffset);
2989 if( rc==SQLITE_IOERR_SHORT_READ ){
2990 rc = SQLITE_OK;
2991 }
2992 }
2993
2994 if( pPg->pgno==1 ){
2995 if( rc ){
2996 /* If the read is unsuccessful, set the dbFileVers[] to something
2997 ** that will never be a valid file version. dbFileVers[] is a copy
2998 ** of bytes 24..39 of the database. Bytes 28..31 should always be
2999 ** zero or the size of the database in page. Bytes 32..35 and 35..39
3000 ** should be page numbers which are never 0xffffffff. So filling
3001 ** pPager->dbFileVers[] with all 0xff bytes should suffice.
3002 **
3003 ** For an encrypted database, the situation is more complex: bytes
3004 ** 24..39 of the database are white noise. But the probability of
3005 ** white noise equaling 16 bytes of 0xff is vanishingly small so
3006 ** we should still be ok.
3007 */
3008 memset(pPager->dbFileVers, 0xff, sizeof(pPager->dbFileVers));
3009 }else{
3010 u8 *dbFileVers = &((u8*)pPg->pData)[24];
3011 memcpy(&pPager->dbFileVers, dbFileVers, sizeof(pPager->dbFileVers));
3012 }
3013 }
3014 PAGER_INCR(sqlite3_pager_readdb_count);
3015 PAGER_INCR(pPager->nRead);
3016 IOTRACE(("PGIN %p %d\n", pPager, pPg->pgno));
3017 PAGERTRACE(("FETCH %d page %d hash(%08x)\n",
3018 PAGERID(pPager), pPg->pgno, pager_pagehash(pPg)));
3019
3020 return rc;
3021}
3022
3023/*
3024** Update the value of the change-counter at offsets 24 and 92 in
3025** the header and the sqlite version number at offset 96.
3026**
3027** This is an unconditional update. See also the pager_incr_changecounter()
3028** routine which only updates the change-counter if the update is actually
3029** needed, as determined by the pPager->changeCountDone state variable.
3030*/
3031static void pager_write_changecounter(PgHdr *pPg){
3032 u32 change_counter;
3033 if( NEVER(pPg==0) ) return;
3034
3035 /* Increment the value just read and write it back to byte 24. */
3036 change_counter = sqlite3Get4byte((u8*)pPg->pPager->dbFileVers)+1;
3037 put32bits(((char*)pPg->pData)+24, change_counter);
3038
3039 /* Also store the SQLite version number in bytes 96..99 and in
3040 ** bytes 92..95 store the change counter for which the version number
3041 ** is valid. */
3042 put32bits(((char*)pPg->pData)+92, change_counter);
3043 put32bits(((char*)pPg->pData)+96, SQLITE_VERSION_NUMBER);
3044}
3045
3046#ifndef SQLITE_OMIT_WAL
3047/*
3048** This function is invoked once for each page that has already been
3049** written into the log file when a WAL transaction is rolled back.
3050** Parameter iPg is the page number of said page. The pCtx argument
3051** is actually a pointer to the Pager structure.
3052**
3053** If page iPg is present in the cache, and has no outstanding references,
3054** it is discarded. Otherwise, if there are one or more outstanding
3055** references, the page content is reloaded from the database. If the
3056** attempt to reload content from the database is required and fails,
3057** return an SQLite error code. Otherwise, SQLITE_OK.
3058*/
3059static int pagerUndoCallback(void *pCtx, Pgno iPg){
3060 int rc = SQLITE_OK;
3061 Pager *pPager = (Pager *)pCtx;
3062 PgHdr *pPg;
3063
3064 assert( pagerUseWal(pPager) );
3065 pPg = sqlite3PagerLookup(pPager, iPg);
3066 if( pPg ){
3067 if( sqlite3PcachePageRefcount(pPg)==1 ){
3068 sqlite3PcacheDrop(pPg);
3069 }else{
3070 rc = readDbPage(pPg);
3071 if( rc==SQLITE_OK ){
3072 pPager->xReiniter(pPg);
3073 }
3074 sqlite3PagerUnrefNotNull(pPg);
3075 }
3076 }
3077
3078 /* Normally, if a transaction is rolled back, any backup processes are
3079 ** updated as data is copied out of the rollback journal and into the
3080 ** database. This is not generally possible with a WAL database, as
3081 ** rollback involves simply truncating the log file. Therefore, if one
3082 ** or more frames have already been written to the log (and therefore
3083 ** also copied into the backup databases) as part of this transaction,
3084 ** the backups must be restarted.
3085 */
3086 sqlite3BackupRestart(pPager->pBackup);
3087
3088 return rc;
3089}
3090
3091/*
3092** This function is called to rollback a transaction on a WAL database.
3093*/
3094static int pagerRollbackWal(Pager *pPager){
3095 int rc; /* Return Code */
3096 PgHdr *pList; /* List of dirty pages to revert */
3097
3098 /* For all pages in the cache that are currently dirty or have already
3099 ** been written (but not committed) to the log file, do one of the
3100 ** following:
3101 **
3102 ** + Discard the cached page (if refcount==0), or
3103 ** + Reload page content from the database (if refcount>0).
3104 */
3105 pPager->dbSize = pPager->dbOrigSize;
3106 rc = sqlite3WalUndo(pPager->pWal, pagerUndoCallback, (void *)pPager);
3107 pList = sqlite3PcacheDirtyList(pPager->pPCache);
3108 while( pList && rc==SQLITE_OK ){
3109 PgHdr *pNext = pList->pDirty;
3110 rc = pagerUndoCallback((void *)pPager, pList->pgno);
3111 pList = pNext;
3112 }
3113
3114 return rc;
3115}
3116
3117/*
3118** This function is a wrapper around sqlite3WalFrames(). As well as logging
3119** the contents of the list of pages headed by pList (connected by pDirty),
3120** this function notifies any active backup processes that the pages have
3121** changed.
3122**
3123** The list of pages passed into this routine is always sorted by page number.
3124** Hence, if page 1 appears anywhere on the list, it will be the first page.
3125*/
3126static int pagerWalFrames(
3127 Pager *pPager, /* Pager object */
3128 PgHdr *pList, /* List of frames to log */
3129 Pgno nTruncate, /* Database size after this commit */
3130 int isCommit /* True if this is a commit */
3131){
3132 int rc; /* Return code */
3133 int nList; /* Number of pages in pList */
3134 PgHdr *p; /* For looping over pages */
3135
3136 assert( pPager->pWal );
3137 assert( pList );
3138#ifdef SQLITE_DEBUG
3139 /* Verify that the page list is in accending order */
3140 for(p=pList; p && p->pDirty; p=p->pDirty){
3141 assert( p->pgno < p->pDirty->pgno );
3142 }
3143#endif
3144
3145 assert( pList->pDirty==0 || isCommit );
3146 if( isCommit ){
3147 /* If a WAL transaction is being committed, there is no point in writing
3148 ** any pages with page numbers greater than nTruncate into the WAL file.
3149 ** They will never be read by any client. So remove them from the pDirty
3150 ** list here. */
3151 PgHdr **ppNext = &pList;
3152 nList = 0;
3153 for(p=pList; (*ppNext = p)!=0; p=p->pDirty){
3154 if( p->pgno<=nTruncate ){
3155 ppNext = &p->pDirty;
3156 nList++;
3157 }
3158 }
3159 assert( pList );
3160 }else{
3161 nList = 1;
3162 }
3163 pPager->aStat[PAGER_STAT_WRITE] += nList;
3164
3165 if( pList->pgno==1 ) pager_write_changecounter(pList);
3166 rc = sqlite3WalFrames(pPager->pWal,
3167 pPager->pageSize, pList, nTruncate, isCommit, pPager->walSyncFlags
3168 );
3169 if( rc==SQLITE_OK && pPager->pBackup ){
3170 for(p=pList; p; p=p->pDirty){
3171 sqlite3BackupUpdate(pPager->pBackup, p->pgno, (u8 *)p->pData);
3172 }
3173 }
3174
3175#ifdef SQLITE_CHECK_PAGES
3176 pList = sqlite3PcacheDirtyList(pPager->pPCache);
3177 for(p=pList; p; p=p->pDirty){
3178 pager_set_pagehash(p);
3179 }
3180#endif
3181
3182 return rc;
3183}
3184
3185/*
3186** Begin a read transaction on the WAL.
3187**
3188** This routine used to be called "pagerOpenSnapshot()" because it essentially
3189** makes a snapshot of the database at the current point in time and preserves
3190** that snapshot for use by the reader in spite of concurrently changes by
3191** other writers or checkpointers.
3192*/
3193static int pagerBeginReadTransaction(Pager *pPager){
3194 int rc; /* Return code */
3195 int changed = 0; /* True if cache must be reset */
3196
3197 assert( pagerUseWal(pPager) );
3198 assert( pPager->eState==PAGER_OPEN || pPager->eState==PAGER_READER );
3199
3200 /* sqlite3WalEndReadTransaction() was not called for the previous
3201 ** transaction in locking_mode=EXCLUSIVE. So call it now. If we
3202 ** are in locking_mode=NORMAL and EndRead() was previously called,
3203 ** the duplicate call is harmless.
3204 */
3205 sqlite3WalEndReadTransaction(pPager->pWal);
3206
3207 rc = sqlite3WalBeginReadTransaction(pPager->pWal, &changed);
3208 if( rc!=SQLITE_OK || changed ){
3209 pager_reset(pPager);
3210 if( USEFETCH(pPager) ) sqlite3OsUnfetch(pPager->fd, 0, 0);
3211 }
3212
3213 return rc;
3214}
3215#endif
3216
3217/*
3218** This function is called as part of the transition from PAGER_OPEN
3219** to PAGER_READER state to determine the size of the database file
3220** in pages (assuming the page size currently stored in Pager.pageSize).
3221**
3222** If no error occurs, SQLITE_OK is returned and the size of the database
3223** in pages is stored in *pnPage. Otherwise, an error code (perhaps
3224** SQLITE_IOERR_FSTAT) is returned and *pnPage is left unmodified.
3225*/
3226static int pagerPagecount(Pager *pPager, Pgno *pnPage){
3227 Pgno nPage; /* Value to return via *pnPage */
3228
3229 /* Query the WAL sub-system for the database size. The WalDbsize()
3230 ** function returns zero if the WAL is not open (i.e. Pager.pWal==0), or
3231 ** if the database size is not available. The database size is not
3232 ** available from the WAL sub-system if the log file is empty or
3233 ** contains no valid committed transactions.
3234 */
3235 assert( pPager->eState==PAGER_OPEN );
3236 assert( pPager->eLock>=SHARED_LOCK );
3237 assert( isOpen(pPager->fd) );
3238 assert( pPager->tempFile==0 );
3239 nPage = sqlite3WalDbsize(pPager->pWal);
3240
3241 /* If the number of pages in the database is not available from the
3242 ** WAL sub-system, determine the page count based on the size of
3243 ** the database file. If the size of the database file is not an
3244 ** integer multiple of the page-size, round up the result.
3245 */
3246 if( nPage==0 && ALWAYS(isOpen(pPager->fd)) ){
3247 i64 n = 0; /* Size of db file in bytes */
3248 int rc = sqlite3OsFileSize(pPager->fd, &n);
3249 if( rc!=SQLITE_OK ){
3250 return rc;
3251 }
3252 nPage = (Pgno)((n+pPager->pageSize-1) / pPager->pageSize);
3253 }
3254
3255 /* If the current number of pages in the file is greater than the
3256 ** configured maximum pager number, increase the allowed limit so
3257 ** that the file can be read.
3258 */
3259 if( nPage>pPager->mxPgno ){
3260 pPager->mxPgno = (Pgno)nPage;
3261 }
3262
3263 *pnPage = nPage;
3264 return SQLITE_OK;
3265}
3266
3267#ifndef SQLITE_OMIT_WAL
3268/*
3269** Check if the *-wal file that corresponds to the database opened by pPager
3270** exists if the database is not empy, or verify that the *-wal file does
3271** not exist (by deleting it) if the database file is empty.
3272**
3273** If the database is not empty and the *-wal file exists, open the pager
3274** in WAL mode. If the database is empty or if no *-wal file exists and
3275** if no error occurs, make sure Pager.journalMode is not set to
3276** PAGER_JOURNALMODE_WAL.
3277**
3278** Return SQLITE_OK or an error code.
3279**
3280** The caller must hold a SHARED lock on the database file to call this
3281** function. Because an EXCLUSIVE lock on the db file is required to delete
3282** a WAL on a none-empty database, this ensures there is no race condition
3283** between the xAccess() below and an xDelete() being executed by some
3284** other connection.
3285*/
3286static int pagerOpenWalIfPresent(Pager *pPager){
3287 int rc = SQLITE_OK;
3288 assert( pPager->eState==PAGER_OPEN );
3289 assert( pPager->eLock>=SHARED_LOCK );
3290
3291 if( !pPager->tempFile ){
3292 int isWal; /* True if WAL file exists */
3293 rc = sqlite3OsAccess(
3294 pPager->pVfs, pPager->zWal, SQLITE_ACCESS_EXISTS, &isWal
3295 );
3296 if( rc==SQLITE_OK ){
3297 if( isWal ){
3298 Pgno nPage; /* Size of the database file */
3299
3300 rc = pagerPagecount(pPager, &nPage);
3301 if( rc ) return rc;
3302 if( nPage==0 ){
3303 rc = sqlite3OsDelete(pPager->pVfs, pPager->zWal, 0);
3304 }else{
3305 testcase( sqlite3PcachePagecount(pPager->pPCache)==0 );
3306 rc = sqlite3PagerOpenWal(pPager, 0);
3307 }
3308 }else if( pPager->journalMode==PAGER_JOURNALMODE_WAL ){
3309 pPager->journalMode = PAGER_JOURNALMODE_DELETE;
3310 }
3311 }
3312 }
3313 return rc;
3314}
3315#endif
3316
3317/*
3318** Playback savepoint pSavepoint. Or, if pSavepoint==NULL, then playback
3319** the entire super-journal file. The case pSavepoint==NULL occurs when
3320** a ROLLBACK TO command is invoked on a SAVEPOINT that is a transaction
3321** savepoint.
3322**
3323** When pSavepoint is not NULL (meaning a non-transaction savepoint is
3324** being rolled back), then the rollback consists of up to three stages,
3325** performed in the order specified:
3326**
3327** * Pages are played back from the main journal starting at byte
3328** offset PagerSavepoint.iOffset and continuing to
3329** PagerSavepoint.iHdrOffset, or to the end of the main journal
3330** file if PagerSavepoint.iHdrOffset is zero.
3331**
3332** * If PagerSavepoint.iHdrOffset is not zero, then pages are played
3333** back starting from the journal header immediately following
3334** PagerSavepoint.iHdrOffset to the end of the main journal file.
3335**
3336** * Pages are then played back from the sub-journal file, starting
3337** with the PagerSavepoint.iSubRec and continuing to the end of
3338** the journal file.
3339**
3340** Throughout the rollback process, each time a page is rolled back, the
3341** corresponding bit is set in a bitvec structure (variable pDone in the
3342** implementation below). This is used to ensure that a page is only
3343** rolled back the first time it is encountered in either journal.
3344**
3345** If pSavepoint is NULL, then pages are only played back from the main
3346** journal file. There is no need for a bitvec in this case.
3347**
3348** In either case, before playback commences the Pager.dbSize variable
3349** is reset to the value that it held at the start of the savepoint
3350** (or transaction). No page with a page-number greater than this value
3351** is played back. If one is encountered it is simply skipped.
3352*/
3353static int pagerPlaybackSavepoint(Pager *pPager, PagerSavepoint *pSavepoint){
3354 i64 szJ; /* Effective size of the main journal */
3355 i64 iHdrOff; /* End of first segment of main-journal records */
3356 int rc = SQLITE_OK; /* Return code */
3357 Bitvec *pDone = 0; /* Bitvec to ensure pages played back only once */
3358
3359 assert( pPager->eState!=PAGER_ERROR );
3360 assert( pPager->eState>=PAGER_WRITER_LOCKED );
3361
3362 /* Allocate a bitvec to use to store the set of pages rolled back */
3363 if( pSavepoint ){
3364 pDone = sqlite3BitvecCreate(pSavepoint->nOrig);
3365 if( !pDone ){
3366 return SQLITE_NOMEM_BKPT;
3367 }
3368 }
3369
3370 /* Set the database size back to the value it was before the savepoint
3371 ** being reverted was opened.
3372 */
3373 pPager->dbSize = pSavepoint ? pSavepoint->nOrig : pPager->dbOrigSize;
3374 pPager->changeCountDone = pPager->tempFile;
3375
3376 if( !pSavepoint && pagerUseWal(pPager) ){
3377 return pagerRollbackWal(pPager);
3378 }
3379
3380 /* Use pPager->journalOff as the effective size of the main rollback
3381 ** journal. The actual file might be larger than this in
3382 ** PAGER_JOURNALMODE_TRUNCATE or PAGER_JOURNALMODE_PERSIST. But anything
3383 ** past pPager->journalOff is off-limits to us.
3384 */
3385 szJ = pPager->journalOff;
3386 assert( pagerUseWal(pPager)==0 || szJ==0 );
3387
3388 /* Begin by rolling back records from the main journal starting at
3389 ** PagerSavepoint.iOffset and continuing to the next journal header.
3390 ** There might be records in the main journal that have a page number
3391 ** greater than the current database size (pPager->dbSize) but those
3392 ** will be skipped automatically. Pages are added to pDone as they
3393 ** are played back.
3394 */
3395 if( pSavepoint && !pagerUseWal(pPager) ){
3396 iHdrOff = pSavepoint->iHdrOffset ? pSavepoint->iHdrOffset : szJ;
3397 pPager->journalOff = pSavepoint->iOffset;
3398 while( rc==SQLITE_OK && pPager->journalOff<iHdrOff ){
3399 rc = pager_playback_one_page(pPager, &pPager->journalOff, pDone, 1, 1);
3400 }
3401 assert( rc!=SQLITE_DONE );
3402 }else{
3403 pPager->journalOff = 0;
3404 }
3405
3406 /* Continue rolling back records out of the main journal starting at
3407 ** the first journal header seen and continuing until the effective end
3408 ** of the main journal file. Continue to skip out-of-range pages and
3409 ** continue adding pages rolled back to pDone.
3410 */
3411 while( rc==SQLITE_OK && pPager->journalOff<szJ ){
3412 u32 ii; /* Loop counter */
3413 u32 nJRec = 0; /* Number of Journal Records */
3414 u32 dummy;
3415 rc = readJournalHdr(pPager, 0, szJ, &nJRec, &dummy);
3416 assert( rc!=SQLITE_DONE );
3417
3418 /*
3419 ** The "pPager->journalHdr+JOURNAL_HDR_SZ(pPager)==pPager->journalOff"
3420 ** test is related to ticket #2565. See the discussion in the
3421 ** pager_playback() function for additional information.
3422 */
3423 if( nJRec==0
3424 && pPager->journalHdr+JOURNAL_HDR_SZ(pPager)==pPager->journalOff
3425 ){
3426 nJRec = (u32)((szJ - pPager->journalOff)/JOURNAL_PG_SZ(pPager));
3427 }
3428 for(ii=0; rc==SQLITE_OK && ii<nJRec && pPager->journalOff<szJ; ii++){
3429 rc = pager_playback_one_page(pPager, &pPager->journalOff, pDone, 1, 1);
3430 }
3431 assert( rc!=SQLITE_DONE );
3432 }
3433 assert( rc!=SQLITE_OK || pPager->journalOff>=szJ );
3434
3435 /* Finally, rollback pages from the sub-journal. Page that were
3436 ** previously rolled back out of the main journal (and are hence in pDone)
3437 ** will be skipped. Out-of-range pages are also skipped.
3438 */
3439 if( pSavepoint ){
3440 u32 ii; /* Loop counter */
3441 i64 offset = (i64)pSavepoint->iSubRec*(4+pPager->pageSize);
3442
3443 if( pagerUseWal(pPager) ){
3444 rc = sqlite3WalSavepointUndo(pPager->pWal, pSavepoint->aWalData);
3445 }
3446 for(ii=pSavepoint->iSubRec; rc==SQLITE_OK && ii<pPager->nSubRec; ii++){
3447 assert( offset==(i64)ii*(4+pPager->pageSize) );
3448 rc = pager_playback_one_page(pPager, &offset, pDone, 0, 1);
3449 }
3450 assert( rc!=SQLITE_DONE );
3451 }
3452
3453 sqlite3BitvecDestroy(pDone);
3454 if( rc==SQLITE_OK ){
3455 pPager->journalOff = szJ;
3456 }
3457
3458 return rc;
3459}
3460
3461/*
3462** Change the maximum number of in-memory pages that are allowed
3463** before attempting to recycle clean and unused pages.
3464*/
3465void sqlite3PagerSetCachesize(Pager *pPager, int mxPage){
3466 sqlite3PcacheSetCachesize(pPager->pPCache, mxPage);
3467}
3468
3469/*
3470** Change the maximum number of in-memory pages that are allowed
3471** before attempting to spill pages to journal.
3472*/
3473int sqlite3PagerSetSpillsize(Pager *pPager, int mxPage){
3474 return sqlite3PcacheSetSpillsize(pPager->pPCache, mxPage);
3475}
3476
3477/*
3478** Invoke SQLITE_FCNTL_MMAP_SIZE based on the current value of szMmap.
3479*/
3480static void pagerFixMaplimit(Pager *pPager){
3481#if SQLITE_MAX_MMAP_SIZE>0
3482 sqlite3_file *fd = pPager->fd;
3483 if( isOpen(fd) && fd->pMethods->iVersion>=3 ){
3484 sqlite3_int64 sz;
3485 sz = pPager->szMmap;
3486 pPager->bUseFetch = (sz>0);
3487 setGetterMethod(pPager);
3488 sqlite3OsFileControlHint(pPager->fd, SQLITE_FCNTL_MMAP_SIZE, &sz);
3489 }
3490#endif
3491}
3492
3493/*
3494** Change the maximum size of any memory mapping made of the database file.
3495*/
3496void sqlite3PagerSetMmapLimit(Pager *pPager, sqlite3_int64 szMmap){
3497 pPager->szMmap = szMmap;
3498 pagerFixMaplimit(pPager);
3499}
3500
3501/*
3502** Free as much memory as possible from the pager.
3503*/
3504void sqlite3PagerShrink(Pager *pPager){
3505 sqlite3PcacheShrink(pPager->pPCache);
3506}
3507
3508/*
3509** Adjust settings of the pager to those specified in the pgFlags parameter.
3510**
3511** The "level" in pgFlags & PAGER_SYNCHRONOUS_MASK sets the robustness
3512** of the database to damage due to OS crashes or power failures by
3513** changing the number of syncs()s when writing the journals.
3514** There are four levels:
3515**
3516** OFF sqlite3OsSync() is never called. This is the default
3517** for temporary and transient files.
3518**
3519** NORMAL The journal is synced once before writes begin on the
3520** database. This is normally adequate protection, but
3521** it is theoretically possible, though very unlikely,
3522** that an inopertune power failure could leave the journal
3523** in a state which would cause damage to the database
3524** when it is rolled back.
3525**
3526** FULL The journal is synced twice before writes begin on the
3527** database (with some additional information - the nRec field
3528** of the journal header - being written in between the two
3529** syncs). If we assume that writing a
3530** single disk sector is atomic, then this mode provides
3531** assurance that the journal will not be corrupted to the
3532** point of causing damage to the database during rollback.
3533**
3534** EXTRA This is like FULL except that is also syncs the directory
3535** that contains the rollback journal after the rollback
3536** journal is unlinked.
3537**
3538** The above is for a rollback-journal mode. For WAL mode, OFF continues
3539** to mean that no syncs ever occur. NORMAL means that the WAL is synced
3540** prior to the start of checkpoint and that the database file is synced
3541** at the conclusion of the checkpoint if the entire content of the WAL
3542** was written back into the database. But no sync operations occur for
3543** an ordinary commit in NORMAL mode with WAL. FULL means that the WAL
3544** file is synced following each commit operation, in addition to the
3545** syncs associated with NORMAL. There is no difference between FULL
3546** and EXTRA for WAL mode.
3547**
3548** Do not confuse synchronous=FULL with SQLITE_SYNC_FULL. The
3549** SQLITE_SYNC_FULL macro means to use the MacOSX-style full-fsync
3550** using fcntl(F_FULLFSYNC). SQLITE_SYNC_NORMAL means to do an
3551** ordinary fsync() call. There is no difference between SQLITE_SYNC_FULL
3552** and SQLITE_SYNC_NORMAL on platforms other than MacOSX. But the
3553** synchronous=FULL versus synchronous=NORMAL setting determines when
3554** the xSync primitive is called and is relevant to all platforms.
3555**
3556** Numeric values associated with these states are OFF==1, NORMAL=2,
3557** and FULL=3.
3558*/
3559#ifndef SQLITE_OMIT_PAGER_PRAGMAS
3560void sqlite3PagerSetFlags(
3561 Pager *pPager, /* The pager to set safety level for */
3562 unsigned pgFlags /* Various flags */
3563){
3564 unsigned level = pgFlags & PAGER_SYNCHRONOUS_MASK;
3565 if( pPager->tempFile ){
3566 pPager->noSync = 1;
3567 pPager->fullSync = 0;
3568 pPager->extraSync = 0;
3569 }else{
3570 pPager->noSync = level==PAGER_SYNCHRONOUS_OFF ?1:0;
3571 pPager->fullSync = level>=PAGER_SYNCHRONOUS_FULL ?1:0;
3572 pPager->extraSync = level==PAGER_SYNCHRONOUS_EXTRA ?1:0;
3573 }
3574 if( pPager->noSync ){
3575 pPager->syncFlags = 0;
3576 }else if( pgFlags & PAGER_FULLFSYNC ){
3577 pPager->syncFlags = SQLITE_SYNC_FULL;
3578 }else{
3579 pPager->syncFlags = SQLITE_SYNC_NORMAL;
3580 }
3581 pPager->walSyncFlags = (pPager->syncFlags<<2);
3582 if( pPager->fullSync ){
3583 pPager->walSyncFlags |= pPager->syncFlags;
3584 }
3585 if( (pgFlags & PAGER_CKPT_FULLFSYNC) && !pPager->noSync ){
3586 pPager->walSyncFlags |= (SQLITE_SYNC_FULL<<2);
3587 }
3588 if( pgFlags & PAGER_CACHESPILL ){
3589 pPager->doNotSpill &= ~SPILLFLAG_OFF;
3590 }else{
3591 pPager->doNotSpill |= SPILLFLAG_OFF;
3592 }
3593}
3594#endif
3595
3596/*
3597** The following global variable is incremented whenever the library
3598** attempts to open a temporary file. This information is used for
3599** testing and analysis only.
3600*/
3601#ifdef SQLITE_TEST
3602int sqlite3_opentemp_count = 0;
3603#endif
3604
3605/*
3606** Open a temporary file.
3607**
3608** Write the file descriptor into *pFile. Return SQLITE_OK on success
3609** or some other error code if we fail. The OS will automatically
3610** delete the temporary file when it is closed.
3611**
3612** The flags passed to the VFS layer xOpen() call are those specified
3613** by parameter vfsFlags ORed with the following:
3614**
3615** SQLITE_OPEN_READWRITE
3616** SQLITE_OPEN_CREATE
3617** SQLITE_OPEN_EXCLUSIVE
3618** SQLITE_OPEN_DELETEONCLOSE
3619*/
3620static int pagerOpentemp(
3621 Pager *pPager, /* The pager object */
3622 sqlite3_file *pFile, /* Write the file descriptor here */
3623 int vfsFlags /* Flags passed through to the VFS */
3624){
3625 int rc; /* Return code */
3626
3627#ifdef SQLITE_TEST
3628 sqlite3_opentemp_count++; /* Used for testing and analysis only */
3629#endif
3630
3631 vfsFlags |= SQLITE_OPEN_READWRITE | SQLITE_OPEN_CREATE |
3632 SQLITE_OPEN_EXCLUSIVE | SQLITE_OPEN_DELETEONCLOSE;
3633 rc = sqlite3OsOpen(pPager->pVfs, 0, pFile, vfsFlags, 0);
3634 assert( rc!=SQLITE_OK || isOpen(pFile) );
3635 return rc;
3636}
3637
3638/*
3639** Set the busy handler function.
3640**
3641** The pager invokes the busy-handler if sqlite3OsLock() returns
3642** SQLITE_BUSY when trying to upgrade from no-lock to a SHARED lock,
3643** or when trying to upgrade from a RESERVED lock to an EXCLUSIVE
3644** lock. It does *not* invoke the busy handler when upgrading from
3645** SHARED to RESERVED, or when upgrading from SHARED to EXCLUSIVE
3646** (which occurs during hot-journal rollback). Summary:
3647**
3648** Transition | Invokes xBusyHandler
3649** --------------------------------------------------------
3650** NO_LOCK -> SHARED_LOCK | Yes
3651** SHARED_LOCK -> RESERVED_LOCK | No
3652** SHARED_LOCK -> EXCLUSIVE_LOCK | No
3653** RESERVED_LOCK -> EXCLUSIVE_LOCK | Yes
3654**
3655** If the busy-handler callback returns non-zero, the lock is
3656** retried. If it returns zero, then the SQLITE_BUSY error is
3657** returned to the caller of the pager API function.
3658*/
3659void sqlite3PagerSetBusyHandler(
3660 Pager *pPager, /* Pager object */
3661 int (*xBusyHandler)(void *), /* Pointer to busy-handler function */
3662 void *pBusyHandlerArg /* Argument to pass to xBusyHandler */
3663){
3664 void **ap;
3665 pPager->xBusyHandler = xBusyHandler;
3666 pPager->pBusyHandlerArg = pBusyHandlerArg;
3667 ap = (void **)&pPager->xBusyHandler;
3668 assert( ((int(*)(void *))(ap[0]))==xBusyHandler );
3669 assert( ap[1]==pBusyHandlerArg );
3670 sqlite3OsFileControlHint(pPager->fd, SQLITE_FCNTL_BUSYHANDLER, (void *)ap);
3671}
3672
3673/*
3674** Change the page size used by the Pager object. The new page size
3675** is passed in *pPageSize.
3676**
3677** If the pager is in the error state when this function is called, it
3678** is a no-op. The value returned is the error state error code (i.e.
3679** one of SQLITE_IOERR, an SQLITE_IOERR_xxx sub-code or SQLITE_FULL).
3680**
3681** Otherwise, if all of the following are true:
3682**
3683** * the new page size (value of *pPageSize) is valid (a power
3684** of two between 512 and SQLITE_MAX_PAGE_SIZE, inclusive), and
3685**
3686** * there are no outstanding page references, and
3687**
3688** * the database is either not an in-memory database or it is
3689** an in-memory database that currently consists of zero pages.
3690**
3691** then the pager object page size is set to *pPageSize.
3692**
3693** If the page size is changed, then this function uses sqlite3PagerMalloc()
3694** to obtain a new Pager.pTmpSpace buffer. If this allocation attempt
3695** fails, SQLITE_NOMEM is returned and the page size remains unchanged.
3696** In all other cases, SQLITE_OK is returned.
3697**
3698** If the page size is not changed, either because one of the enumerated
3699** conditions above is not true, the pager was in error state when this
3700** function was called, or because the memory allocation attempt failed,
3701** then *pPageSize is set to the old, retained page size before returning.
3702*/
3703int sqlite3PagerSetPagesize(Pager *pPager, u32 *pPageSize, int nReserve){
3704 int rc = SQLITE_OK;
3705
3706 /* It is not possible to do a full assert_pager_state() here, as this
3707 ** function may be called from within PagerOpen(), before the state
3708 ** of the Pager object is internally consistent.
3709 **
3710 ** At one point this function returned an error if the pager was in
3711 ** PAGER_ERROR state. But since PAGER_ERROR state guarantees that
3712 ** there is at least one outstanding page reference, this function
3713 ** is a no-op for that case anyhow.
3714 */
3715
3716 u32 pageSize = *pPageSize;
3717 assert( pageSize==0 || (pageSize>=512 && pageSize<=SQLITE_MAX_PAGE_SIZE) );
3718 if( (pPager->memDb==0 || pPager->dbSize==0)
3719 && sqlite3PcacheRefCount(pPager->pPCache)==0
3720 && pageSize && pageSize!=(u32)pPager->pageSize
3721 ){
3722 char *pNew = NULL; /* New temp space */
3723 i64 nByte = 0;
3724
3725 if( pPager->eState>PAGER_OPEN && isOpen(pPager->fd) ){
3726 rc = sqlite3OsFileSize(pPager->fd, &nByte);
3727 }
3728 if( rc==SQLITE_OK ){
3729 /* 8 bytes of zeroed overrun space is sufficient so that the b-tree
3730 * cell header parser will never run off the end of the allocation */
3731 pNew = (char *)sqlite3PageMalloc(pageSize+8);
3732 if( !pNew ){
3733 rc = SQLITE_NOMEM_BKPT;
3734 }else{
3735 memset(pNew+pageSize, 0, 8);
3736 }
3737 }
3738
3739 if( rc==SQLITE_OK ){
3740 pager_reset(pPager);
3741 rc = sqlite3PcacheSetPageSize(pPager->pPCache, pageSize);
3742 }
3743 if( rc==SQLITE_OK ){
3744 sqlite3PageFree(pPager->pTmpSpace);
3745 pPager->pTmpSpace = pNew;
3746 pPager->dbSize = (Pgno)((nByte+pageSize-1)/pageSize);
3747 pPager->pageSize = pageSize;
3748 pPager->lckPgno = (Pgno)(PENDING_BYTE/pageSize) + 1;
3749 }else{
3750 sqlite3PageFree(pNew);
3751 }
3752 }
3753
3754 *pPageSize = pPager->pageSize;
3755 if( rc==SQLITE_OK ){
3756 if( nReserve<0 ) nReserve = pPager->nReserve;
3757 assert( nReserve>=0 && nReserve<1000 );
3758 pPager->nReserve = (i16)nReserve;
3759 pagerFixMaplimit(pPager);
3760 }
3761 return rc;
3762}
3763
3764/*
3765** Return a pointer to the "temporary page" buffer held internally
3766** by the pager. This is a buffer that is big enough to hold the
3767** entire content of a database page. This buffer is used internally
3768** during rollback and will be overwritten whenever a rollback
3769** occurs. But other modules are free to use it too, as long as
3770** no rollbacks are happening.
3771*/
3772void *sqlite3PagerTempSpace(Pager *pPager){
3773 return pPager->pTmpSpace;
3774}
3775
3776/*
3777** Attempt to set the maximum database page count if mxPage is positive.
3778** Make no changes if mxPage is zero or negative. And never reduce the
3779** maximum page count below the current size of the database.
3780**
3781** Regardless of mxPage, return the current maximum page count.
3782*/
3783Pgno sqlite3PagerMaxPageCount(Pager *pPager, Pgno mxPage){
3784 if( mxPage>0 ){
3785 pPager->mxPgno = mxPage;
3786 }
3787 assert( pPager->eState!=PAGER_OPEN ); /* Called only by OP_MaxPgcnt */
3788 /* assert( pPager->mxPgno>=pPager->dbSize ); */
3789 /* OP_MaxPgcnt ensures that the parameter passed to this function is not
3790 ** less than the total number of valid pages in the database. But this
3791 ** may be less than Pager.dbSize, and so the assert() above is not valid */
3792 return pPager->mxPgno;
3793}
3794
3795/*
3796** The following set of routines are used to disable the simulated
3797** I/O error mechanism. These routines are used to avoid simulated
3798** errors in places where we do not care about errors.
3799**
3800** Unless -DSQLITE_TEST=1 is used, these routines are all no-ops
3801** and generate no code.
3802*/
3803#ifdef SQLITE_TEST
3804extern int sqlite3_io_error_pending;
3805extern int sqlite3_io_error_hit;
3806static int saved_cnt;
3807void disable_simulated_io_errors(void){
3808 saved_cnt = sqlite3_io_error_pending;
3809 sqlite3_io_error_pending = -1;
3810}
3811void enable_simulated_io_errors(void){
3812 sqlite3_io_error_pending = saved_cnt;
3813}
3814#else
3815# define disable_simulated_io_errors()
3816# define enable_simulated_io_errors()
3817#endif
3818
3819/*
3820** Read the first N bytes from the beginning of the file into memory
3821** that pDest points to.
3822**
3823** If the pager was opened on a transient file (zFilename==""), or
3824** opened on a file less than N bytes in size, the output buffer is
3825** zeroed and SQLITE_OK returned. The rationale for this is that this
3826** function is used to read database headers, and a new transient or
3827** zero sized database has a header than consists entirely of zeroes.
3828**
3829** If any IO error apart from SQLITE_IOERR_SHORT_READ is encountered,
3830** the error code is returned to the caller and the contents of the
3831** output buffer undefined.
3832*/
3833int sqlite3PagerReadFileheader(Pager *pPager, int N, unsigned char *pDest){
3834 int rc = SQLITE_OK;
3835 memset(pDest, 0, N);
3836 assert( isOpen(pPager->fd) || pPager->tempFile );
3837
3838 /* This routine is only called by btree immediately after creating
3839 ** the Pager object. There has not been an opportunity to transition
3840 ** to WAL mode yet.
3841 */
3842 assert( !pagerUseWal(pPager) );
3843
3844 if( isOpen(pPager->fd) ){
3845 IOTRACE(("DBHDR %p 0 %d\n", pPager, N))
3846 rc = sqlite3OsRead(pPager->fd, pDest, N, 0);
3847 if( rc==SQLITE_IOERR_SHORT_READ ){
3848 rc = SQLITE_OK;
3849 }
3850 }
3851 return rc;
3852}
3853
3854/*
3855** This function may only be called when a read-transaction is open on
3856** the pager. It returns the total number of pages in the database.
3857**
3858** However, if the file is between 1 and <page-size> bytes in size, then
3859** this is considered a 1 page file.
3860*/
3861void sqlite3PagerPagecount(Pager *pPager, int *pnPage){
3862 assert( pPager->eState>=PAGER_READER );
3863 assert( pPager->eState!=PAGER_WRITER_FINISHED );
3864 *pnPage = (int)pPager->dbSize;
3865}
3866
3867
3868/*
3869** Try to obtain a lock of type locktype on the database file. If
3870** a similar or greater lock is already held, this function is a no-op
3871** (returning SQLITE_OK immediately).
3872**
3873** Otherwise, attempt to obtain the lock using sqlite3OsLock(). Invoke
3874** the busy callback if the lock is currently not available. Repeat
3875** until the busy callback returns false or until the attempt to
3876** obtain the lock succeeds.
3877**
3878** Return SQLITE_OK on success and an error code if we cannot obtain
3879** the lock. If the lock is obtained successfully, set the Pager.state
3880** variable to locktype before returning.
3881*/
3882static int pager_wait_on_lock(Pager *pPager, int locktype){
3883 int rc; /* Return code */
3884
3885 /* Check that this is either a no-op (because the requested lock is
3886 ** already held), or one of the transitions that the busy-handler
3887 ** may be invoked during, according to the comment above
3888 ** sqlite3PagerSetBusyhandler().
3889 */
3890 assert( (pPager->eLock>=locktype)
3891 || (pPager->eLock==NO_LOCK && locktype==SHARED_LOCK)
3892 || (pPager->eLock==RESERVED_LOCK && locktype==EXCLUSIVE_LOCK)
3893 );
3894
3895 do {
3896 rc = pagerLockDb(pPager, locktype);
3897 }while( rc==SQLITE_BUSY && pPager->xBusyHandler(pPager->pBusyHandlerArg) );
3898 return rc;
3899}
3900
3901/*
3902** Function assertTruncateConstraint(pPager) checks that one of the
3903** following is true for all dirty pages currently in the page-cache:
3904**
3905** a) The page number is less than or equal to the size of the
3906** current database image, in pages, OR
3907**
3908** b) if the page content were written at this time, it would not
3909** be necessary to write the current content out to the sub-journal.
3910**
3911** If the condition asserted by this function were not true, and the
3912** dirty page were to be discarded from the cache via the pagerStress()
3913** routine, pagerStress() would not write the current page content to
3914** the database file. If a savepoint transaction were rolled back after
3915** this happened, the correct behavior would be to restore the current
3916** content of the page. However, since this content is not present in either
3917** the database file or the portion of the rollback journal and
3918** sub-journal rolled back the content could not be restored and the
3919** database image would become corrupt. It is therefore fortunate that
3920** this circumstance cannot arise.
3921*/
3922#if defined(SQLITE_DEBUG)
3923static void assertTruncateConstraintCb(PgHdr *pPg){
3924 Pager *pPager = pPg->pPager;
3925 assert( pPg->flags&PGHDR_DIRTY );
3926 if( pPg->pgno>pPager->dbSize ){ /* if (a) is false */
3927 Pgno pgno = pPg->pgno;
3928 int i;
3929 for(i=0; i<pPg->pPager->nSavepoint; i++){
3930 PagerSavepoint *p = &pPager->aSavepoint[i];
3931 assert( p->nOrig<pgno || sqlite3BitvecTestNotNull(p->pInSavepoint,pgno) );
3932 }
3933 }
3934}
3935static void assertTruncateConstraint(Pager *pPager){
3936 sqlite3PcacheIterateDirty(pPager->pPCache, assertTruncateConstraintCb);
3937}
3938#else
3939# define assertTruncateConstraint(pPager)
3940#endif
3941
3942/*
3943** Truncate the in-memory database file image to nPage pages. This
3944** function does not actually modify the database file on disk. It
3945** just sets the internal state of the pager object so that the
3946** truncation will be done when the current transaction is committed.
3947**
3948** This function is only called right before committing a transaction.
3949** Once this function has been called, the transaction must either be
3950** rolled back or committed. It is not safe to call this function and
3951** then continue writing to the database.
3952*/
3953void sqlite3PagerTruncateImage(Pager *pPager, Pgno nPage){
3954 assert( pPager->dbSize>=nPage || CORRUPT_DB );
3955 assert( pPager->eState>=PAGER_WRITER_CACHEMOD );
3956 pPager->dbSize = nPage;
3957
3958 /* At one point the code here called assertTruncateConstraint() to
3959 ** ensure that all pages being truncated away by this operation are,
3960 ** if one or more savepoints are open, present in the savepoint
3961 ** journal so that they can be restored if the savepoint is rolled
3962 ** back. This is no longer necessary as this function is now only
3963 ** called right before committing a transaction. So although the
3964 ** Pager object may still have open savepoints (Pager.nSavepoint!=0),
3965 ** they cannot be rolled back. So the assertTruncateConstraint() call
3966 ** is no longer correct. */
3967}
3968
3969
3970/*
3971** This function is called before attempting a hot-journal rollback. It
3972** syncs the journal file to disk, then sets pPager->journalHdr to the
3973** size of the journal file so that the pager_playback() routine knows
3974** that the entire journal file has been synced.
3975**
3976** Syncing a hot-journal to disk before attempting to roll it back ensures
3977** that if a power-failure occurs during the rollback, the process that
3978** attempts rollback following system recovery sees the same journal
3979** content as this process.
3980**
3981** If everything goes as planned, SQLITE_OK is returned. Otherwise,
3982** an SQLite error code.
3983*/
3984static int pagerSyncHotJournal(Pager *pPager){
3985 int rc = SQLITE_OK;
3986 if( !pPager->noSync ){
3987 rc = sqlite3OsSync(pPager->jfd, SQLITE_SYNC_NORMAL);
3988 }
3989 if( rc==SQLITE_OK ){
3990 rc = sqlite3OsFileSize(pPager->jfd, &pPager->journalHdr);
3991 }
3992 return rc;
3993}
3994
3995#if SQLITE_MAX_MMAP_SIZE>0
3996/*
3997** Obtain a reference to a memory mapped page object for page number pgno.
3998** The new object will use the pointer pData, obtained from xFetch().
3999** If successful, set *ppPage to point to the new page reference
4000** and return SQLITE_OK. Otherwise, return an SQLite error code and set
4001** *ppPage to zero.
4002**
4003** Page references obtained by calling this function should be released
4004** by calling pagerReleaseMapPage().
4005*/
4006static int pagerAcquireMapPage(
4007 Pager *pPager, /* Pager object */
4008 Pgno pgno, /* Page number */
4009 void *pData, /* xFetch()'d data for this page */
4010 PgHdr **ppPage /* OUT: Acquired page object */
4011){
4012 PgHdr *p; /* Memory mapped page to return */
4013
4014 if( pPager->pMmapFreelist ){
4015 *ppPage = p = pPager->pMmapFreelist;
4016 pPager->pMmapFreelist = p->pDirty;
4017 p->pDirty = 0;
4018 assert( pPager->nExtra>=8 );
4019 memset(p->pExtra, 0, 8);
4020 }else{
4021 *ppPage = p = (PgHdr *)sqlite3MallocZero(sizeof(PgHdr) + pPager->nExtra);
4022 if( p==0 ){
4023 sqlite3OsUnfetch(pPager->fd, (i64)(pgno-1) * pPager->pageSize, pData);
4024 return SQLITE_NOMEM_BKPT;
4025 }
4026 p->pExtra = (void *)&p[1];
4027 p->flags = PGHDR_MMAP;
4028 p->nRef = 1;
4029 p->pPager = pPager;
4030 }
4031
4032 assert( p->pExtra==(void *)&p[1] );
4033 assert( p->pPage==0 );
4034 assert( p->flags==PGHDR_MMAP );
4035 assert( p->pPager==pPager );
4036 assert( p->nRef==1 );
4037
4038 p->pgno = pgno;
4039 p->pData = pData;
4040 pPager->nMmapOut++;
4041
4042 return SQLITE_OK;
4043}
4044#endif
4045
4046/*
4047** Release a reference to page pPg. pPg must have been returned by an
4048** earlier call to pagerAcquireMapPage().
4049*/
4050static void pagerReleaseMapPage(PgHdr *pPg){
4051 Pager *pPager = pPg->pPager;
4052 pPager->nMmapOut--;
4053 pPg->pDirty = pPager->pMmapFreelist;
4054 pPager->pMmapFreelist = pPg;
4055
4056 assert( pPager->fd->pMethods->iVersion>=3 );
4057 sqlite3OsUnfetch(pPager->fd, (i64)(pPg->pgno-1)*pPager->pageSize, pPg->pData);
4058}
4059
4060/*
4061** Free all PgHdr objects stored in the Pager.pMmapFreelist list.
4062*/
4063static void pagerFreeMapHdrs(Pager *pPager){
4064 PgHdr *p;
4065 PgHdr *pNext;
4066 for(p=pPager->pMmapFreelist; p; p=pNext){
4067 pNext = p->pDirty;
4068 sqlite3_free(p);
4069 }
4070}
4071
4072/* Verify that the database file has not be deleted or renamed out from
4073** under the pager. Return SQLITE_OK if the database is still where it ought
4074** to be on disk. Return non-zero (SQLITE_READONLY_DBMOVED or some other error
4075** code from sqlite3OsAccess()) if the database has gone missing.
4076*/
4077static int databaseIsUnmoved(Pager *pPager){
4078 int bHasMoved = 0;
4079 int rc;
4080
4081 if( pPager->tempFile ) return SQLITE_OK;
4082 if( pPager->dbSize==0 ) return SQLITE_OK;
4083 assert( pPager->zFilename && pPager->zFilename[0] );
4084 rc = sqlite3OsFileControl(pPager->fd, SQLITE_FCNTL_HAS_MOVED, &bHasMoved);
4085 if( rc==SQLITE_NOTFOUND ){
4086 /* If the HAS_MOVED file-control is unimplemented, assume that the file
4087 ** has not been moved. That is the historical behavior of SQLite: prior to
4088 ** version 3.8.3, it never checked */
4089 rc = SQLITE_OK;
4090 }else if( rc==SQLITE_OK && bHasMoved ){
4091 rc = SQLITE_READONLY_DBMOVED;
4092 }
4093 return rc;
4094}
4095
4096
4097/*
4098** Shutdown the page cache. Free all memory and close all files.
4099**
4100** If a transaction was in progress when this routine is called, that
4101** transaction is rolled back. All outstanding pages are invalidated
4102** and their memory is freed. Any attempt to use a page associated
4103** with this page cache after this function returns will likely
4104** result in a coredump.
4105**
4106** This function always succeeds. If a transaction is active an attempt
4107** is made to roll it back. If an error occurs during the rollback
4108** a hot journal may be left in the filesystem but no error is returned
4109** to the caller.
4110*/
4111int sqlite3PagerClose(Pager *pPager, sqlite3 *db){
4112 u8 *pTmp = (u8*)pPager->pTmpSpace;
4113 assert( db || pagerUseWal(pPager)==0 );
4114 assert( assert_pager_state(pPager) );
4115 disable_simulated_io_errors();
4116 sqlite3BeginBenignMalloc();
4117 pagerFreeMapHdrs(pPager);
4118 /* pPager->errCode = 0; */
4119 pPager->exclusiveMode = 0;
4120#ifndef SQLITE_OMIT_WAL
4121 {
4122 u8 *a = 0;
4123 assert( db || pPager->pWal==0 );
4124 if( db && 0==(db->flags & SQLITE_NoCkptOnClose)
4125 && SQLITE_OK==databaseIsUnmoved(pPager)
4126 ){
4127 a = pTmp;
4128 }
4129 sqlite3WalClose(pPager->pWal, db, pPager->walSyncFlags, pPager->pageSize,a);
4130 pPager->pWal = 0;
4131 }
4132#endif
4133 pager_reset(pPager);
4134 if( MEMDB ){
4135 pager_unlock(pPager);
4136 }else{
4137 /* If it is open, sync the journal file before calling UnlockAndRollback.
4138 ** If this is not done, then an unsynced portion of the open journal
4139 ** file may be played back into the database. If a power failure occurs
4140 ** while this is happening, the database could become corrupt.
4141 **
4142 ** If an error occurs while trying to sync the journal, shift the pager
4143 ** into the ERROR state. This causes UnlockAndRollback to unlock the
4144 ** database and close the journal file without attempting to roll it
4145 ** back or finalize it. The next database user will have to do hot-journal
4146 ** rollback before accessing the database file.
4147 */
4148 if( isOpen(pPager->jfd) ){
4149 pager_error(pPager, pagerSyncHotJournal(pPager));
4150 }
4151 pagerUnlockAndRollback(pPager);
4152 }
4153 sqlite3EndBenignMalloc();
4154 enable_simulated_io_errors();
4155 PAGERTRACE(("CLOSE %d\n", PAGERID(pPager)));
4156 IOTRACE(("CLOSE %p\n", pPager))
4157 sqlite3OsClose(pPager->jfd);
4158 sqlite3OsClose(pPager->fd);
4159 sqlite3PageFree(pTmp);
4160 sqlite3PcacheClose(pPager->pPCache);
4161 assert( !pPager->aSavepoint && !pPager->pInJournal );
4162 assert( !isOpen(pPager->jfd) && !isOpen(pPager->sjfd) );
4163
4164 sqlite3_free(pPager);
4165 return SQLITE_OK;
4166}
4167
4168#if !defined(NDEBUG) || defined(SQLITE_TEST)
4169/*
4170** Return the page number for page pPg.
4171*/
4172Pgno sqlite3PagerPagenumber(DbPage *pPg){
4173 return pPg->pgno;
4174}
4175#endif
4176
4177/*
4178** Increment the reference count for page pPg.
4179*/
4180void sqlite3PagerRef(DbPage *pPg){
4181 sqlite3PcacheRef(pPg);
4182}
4183
4184/*
4185** Sync the journal. In other words, make sure all the pages that have
4186** been written to the journal have actually reached the surface of the
4187** disk and can be restored in the event of a hot-journal rollback.
4188**
4189** If the Pager.noSync flag is set, then this function is a no-op.
4190** Otherwise, the actions required depend on the journal-mode and the
4191** device characteristics of the file-system, as follows:
4192**
4193** * If the journal file is an in-memory journal file, no action need
4194** be taken.
4195**
4196** * Otherwise, if the device does not support the SAFE_APPEND property,
4197** then the nRec field of the most recently written journal header
4198** is updated to contain the number of journal records that have
4199** been written following it. If the pager is operating in full-sync
4200** mode, then the journal file is synced before this field is updated.
4201**
4202** * If the device does not support the SEQUENTIAL property, then
4203** journal file is synced.
4204**
4205** Or, in pseudo-code:
4206**
4207** if( NOT <in-memory journal> ){
4208** if( NOT SAFE_APPEND ){
4209** if( <full-sync mode> ) xSync(<journal file>);
4210** <update nRec field>
4211** }
4212** if( NOT SEQUENTIAL ) xSync(<journal file>);
4213** }
4214**
4215** If successful, this routine clears the PGHDR_NEED_SYNC flag of every
4216** page currently held in memory before returning SQLITE_OK. If an IO
4217** error is encountered, then the IO error code is returned to the caller.
4218*/
4219static int syncJournal(Pager *pPager, int newHdr){
4220 int rc; /* Return code */
4221
4222 assert( pPager->eState==PAGER_WRITER_CACHEMOD
4223 || pPager->eState==PAGER_WRITER_DBMOD
4224 );
4225 assert( assert_pager_state(pPager) );
4226 assert( !pagerUseWal(pPager) );
4227
4228 rc = sqlite3PagerExclusiveLock(pPager);
4229 if( rc!=SQLITE_OK ) return rc;
4230
4231 if( !pPager->noSync ){
4232 assert( !pPager->tempFile );
4233 if( isOpen(pPager->jfd) && pPager->journalMode!=PAGER_JOURNALMODE_MEMORY ){
4234 const int iDc = sqlite3OsDeviceCharacteristics(pPager->fd);
4235 assert( isOpen(pPager->jfd) );
4236
4237 if( 0==(iDc&SQLITE_IOCAP_SAFE_APPEND) ){
4238 /* This block deals with an obscure problem. If the last connection
4239 ** that wrote to this database was operating in persistent-journal
4240 ** mode, then the journal file may at this point actually be larger
4241 ** than Pager.journalOff bytes. If the next thing in the journal
4242 ** file happens to be a journal-header (written as part of the
4243 ** previous connection's transaction), and a crash or power-failure
4244 ** occurs after nRec is updated but before this connection writes
4245 ** anything else to the journal file (or commits/rolls back its
4246 ** transaction), then SQLite may become confused when doing the
4247 ** hot-journal rollback following recovery. It may roll back all
4248 ** of this connections data, then proceed to rolling back the old,
4249 ** out-of-date data that follows it. Database corruption.
4250 **
4251 ** To work around this, if the journal file does appear to contain
4252 ** a valid header following Pager.journalOff, then write a 0x00
4253 ** byte to the start of it to prevent it from being recognized.
4254 **
4255 ** Variable iNextHdrOffset is set to the offset at which this
4256 ** problematic header will occur, if it exists. aMagic is used
4257 ** as a temporary buffer to inspect the first couple of bytes of
4258 ** the potential journal header.
4259 */
4260 i64 iNextHdrOffset;
4261 u8 aMagic[8];
4262 u8 zHeader[sizeof(aJournalMagic)+4];
4263
4264 memcpy(zHeader, aJournalMagic, sizeof(aJournalMagic));
4265 put32bits(&zHeader[sizeof(aJournalMagic)], pPager->nRec);
4266
4267 iNextHdrOffset = journalHdrOffset(pPager);
4268 rc = sqlite3OsRead(pPager->jfd, aMagic, 8, iNextHdrOffset);
4269 if( rc==SQLITE_OK && 0==memcmp(aMagic, aJournalMagic, 8) ){
4270 static const u8 zerobyte = 0;
4271 rc = sqlite3OsWrite(pPager->jfd, &zerobyte, 1, iNextHdrOffset);
4272 }
4273 if( rc!=SQLITE_OK && rc!=SQLITE_IOERR_SHORT_READ ){
4274 return rc;
4275 }
4276
4277 /* Write the nRec value into the journal file header. If in
4278 ** full-synchronous mode, sync the journal first. This ensures that
4279 ** all data has really hit the disk before nRec is updated to mark
4280 ** it as a candidate for rollback.
4281 **
4282 ** This is not required if the persistent media supports the
4283 ** SAFE_APPEND property. Because in this case it is not possible
4284 ** for garbage data to be appended to the file, the nRec field
4285 ** is populated with 0xFFFFFFFF when the journal header is written
4286 ** and never needs to be updated.
4287 */
4288 if( pPager->fullSync && 0==(iDc&SQLITE_IOCAP_SEQUENTIAL) ){
4289 PAGERTRACE(("SYNC journal of %d\n", PAGERID(pPager)));
4290 IOTRACE(("JSYNC %p\n", pPager))
4291 rc = sqlite3OsSync(pPager->jfd, pPager->syncFlags);
4292 if( rc!=SQLITE_OK ) return rc;
4293 }
4294 IOTRACE(("JHDR %p %lld\n", pPager, pPager->journalHdr));
4295 rc = sqlite3OsWrite(
4296 pPager->jfd, zHeader, sizeof(zHeader), pPager->journalHdr
4297 );
4298 if( rc!=SQLITE_OK ) return rc;
4299 }
4300 if( 0==(iDc&SQLITE_IOCAP_SEQUENTIAL) ){
4301 PAGERTRACE(("SYNC journal of %d\n", PAGERID(pPager)));
4302 IOTRACE(("JSYNC %p\n", pPager))
4303 rc = sqlite3OsSync(pPager->jfd, pPager->syncFlags|
4304 (pPager->syncFlags==SQLITE_SYNC_FULL?SQLITE_SYNC_DATAONLY:0)
4305 );
4306 if( rc!=SQLITE_OK ) return rc;
4307 }
4308
4309 pPager->journalHdr = pPager->journalOff;
4310 if( newHdr && 0==(iDc&SQLITE_IOCAP_SAFE_APPEND) ){
4311 pPager->nRec = 0;
4312 rc = writeJournalHdr(pPager);
4313 if( rc!=SQLITE_OK ) return rc;
4314 }
4315 }else{
4316 pPager->journalHdr = pPager->journalOff;
4317 }
4318 }
4319
4320 /* Unless the pager is in noSync mode, the journal file was just
4321 ** successfully synced. Either way, clear the PGHDR_NEED_SYNC flag on
4322 ** all pages.
4323 */
4324 sqlite3PcacheClearSyncFlags(pPager->pPCache);
4325 pPager->eState = PAGER_WRITER_DBMOD;
4326 assert( assert_pager_state(pPager) );
4327 return SQLITE_OK;
4328}
4329
4330/*
4331** The argument is the first in a linked list of dirty pages connected
4332** by the PgHdr.pDirty pointer. This function writes each one of the
4333** in-memory pages in the list to the database file. The argument may
4334** be NULL, representing an empty list. In this case this function is
4335** a no-op.
4336**
4337** The pager must hold at least a RESERVED lock when this function
4338** is called. Before writing anything to the database file, this lock
4339** is upgraded to an EXCLUSIVE lock. If the lock cannot be obtained,
4340** SQLITE_BUSY is returned and no data is written to the database file.
4341**
4342** If the pager is a temp-file pager and the actual file-system file
4343** is not yet open, it is created and opened before any data is
4344** written out.
4345**
4346** Once the lock has been upgraded and, if necessary, the file opened,
4347** the pages are written out to the database file in list order. Writing
4348** a page is skipped if it meets either of the following criteria:
4349**
4350** * The page number is greater than Pager.dbSize, or
4351** * The PGHDR_DONT_WRITE flag is set on the page.
4352**
4353** If writing out a page causes the database file to grow, Pager.dbFileSize
4354** is updated accordingly. If page 1 is written out, then the value cached
4355** in Pager.dbFileVers[] is updated to match the new value stored in
4356** the database file.
4357**
4358** If everything is successful, SQLITE_OK is returned. If an IO error
4359** occurs, an IO error code is returned. Or, if the EXCLUSIVE lock cannot
4360** be obtained, SQLITE_BUSY is returned.
4361*/
4362static int pager_write_pagelist(Pager *pPager, PgHdr *pList){
4363 int rc = SQLITE_OK; /* Return code */
4364
4365 /* This function is only called for rollback pagers in WRITER_DBMOD state. */
4366 assert( !pagerUseWal(pPager) );
4367 assert( pPager->tempFile || pPager->eState==PAGER_WRITER_DBMOD );
4368 assert( pPager->eLock==EXCLUSIVE_LOCK );
4369 assert( isOpen(pPager->fd) || pList->pDirty==0 );
4370
4371 /* If the file is a temp-file has not yet been opened, open it now. It
4372 ** is not possible for rc to be other than SQLITE_OK if this branch
4373 ** is taken, as pager_wait_on_lock() is a no-op for temp-files.
4374 */
4375 if( !isOpen(pPager->fd) ){
4376 assert( pPager->tempFile && rc==SQLITE_OK );
4377 rc = pagerOpentemp(pPager, pPager->fd, pPager->vfsFlags);
4378 }
4379
4380 /* Before the first write, give the VFS a hint of what the final
4381 ** file size will be.
4382 */
4383 assert( rc!=SQLITE_OK || isOpen(pPager->fd) );
4384 if( rc==SQLITE_OK
4385 && pPager->dbHintSize<pPager->dbSize
4386 && (pList->pDirty || pList->pgno>pPager->dbHintSize)
4387 ){
4388 sqlite3_int64 szFile = pPager->pageSize * (sqlite3_int64)pPager->dbSize;
4389 sqlite3OsFileControlHint(pPager->fd, SQLITE_FCNTL_SIZE_HINT, &szFile);
4390 pPager->dbHintSize = pPager->dbSize;
4391 }
4392
4393 while( rc==SQLITE_OK && pList ){
4394 Pgno pgno = pList->pgno;
4395
4396 /* If there are dirty pages in the page cache with page numbers greater
4397 ** than Pager.dbSize, this means sqlite3PagerTruncateImage() was called to
4398 ** make the file smaller (presumably by auto-vacuum code). Do not write
4399 ** any such pages to the file.
4400 **
4401 ** Also, do not write out any page that has the PGHDR_DONT_WRITE flag
4402 ** set (set by sqlite3PagerDontWrite()).
4403 */
4404 if( pgno<=pPager->dbSize && 0==(pList->flags&PGHDR_DONT_WRITE) ){
4405 i64 offset = (pgno-1)*(i64)pPager->pageSize; /* Offset to write */
4406 char *pData; /* Data to write */
4407
4408 assert( (pList->flags&PGHDR_NEED_SYNC)==0 );
4409 if( pList->pgno==1 ) pager_write_changecounter(pList);
4410
4411 pData = pList->pData;
4412
4413 /* Write out the page data. */
4414 rc = sqlite3OsWrite(pPager->fd, pData, pPager->pageSize, offset);
4415
4416 /* If page 1 was just written, update Pager.dbFileVers to match
4417 ** the value now stored in the database file. If writing this
4418 ** page caused the database file to grow, update dbFileSize.
4419 */
4420 if( pgno==1 ){
4421 memcpy(&pPager->dbFileVers, &pData[24], sizeof(pPager->dbFileVers));
4422 }
4423 if( pgno>pPager->dbFileSize ){
4424 pPager->dbFileSize = pgno;
4425 }
4426 pPager->aStat[PAGER_STAT_WRITE]++;
4427
4428 /* Update any backup objects copying the contents of this pager. */
4429 sqlite3BackupUpdate(pPager->pBackup, pgno, (u8*)pList->pData);
4430
4431 PAGERTRACE(("STORE %d page %d hash(%08x)\n",
4432 PAGERID(pPager), pgno, pager_pagehash(pList)));
4433 IOTRACE(("PGOUT %p %d\n", pPager, pgno));
4434 PAGER_INCR(sqlite3_pager_writedb_count);
4435 }else{
4436 PAGERTRACE(("NOSTORE %d page %d\n", PAGERID(pPager), pgno));
4437 }
4438 pager_set_pagehash(pList);
4439 pList = pList->pDirty;
4440 }
4441
4442 return rc;
4443}
4444
4445/*
4446** Ensure that the sub-journal file is open. If it is already open, this
4447** function is a no-op.
4448**
4449** SQLITE_OK is returned if everything goes according to plan. An
4450** SQLITE_IOERR_XXX error code is returned if a call to sqlite3OsOpen()
4451** fails.
4452*/
4453static int openSubJournal(Pager *pPager){
4454 int rc = SQLITE_OK;
4455 if( !isOpen(pPager->sjfd) ){
4456 const int flags = SQLITE_OPEN_SUBJOURNAL | SQLITE_OPEN_READWRITE
4457 | SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE
4458 | SQLITE_OPEN_DELETEONCLOSE;
4459 int nStmtSpill = sqlite3Config.nStmtSpill;
4460 if( pPager->journalMode==PAGER_JOURNALMODE_MEMORY || pPager->subjInMemory ){
4461 nStmtSpill = -1;
4462 }
4463 rc = sqlite3JournalOpen(pPager->pVfs, 0, pPager->sjfd, flags, nStmtSpill);
4464 }
4465 return rc;
4466}
4467
4468/*
4469** Append a record of the current state of page pPg to the sub-journal.
4470**
4471** If successful, set the bit corresponding to pPg->pgno in the bitvecs
4472** for all open savepoints before returning.
4473**
4474** This function returns SQLITE_OK if everything is successful, an IO
4475** error code if the attempt to write to the sub-journal fails, or
4476** SQLITE_NOMEM if a malloc fails while setting a bit in a savepoint
4477** bitvec.
4478*/
4479static int subjournalPage(PgHdr *pPg){
4480 int rc = SQLITE_OK;
4481 Pager *pPager = pPg->pPager;
4482 if( pPager->journalMode!=PAGER_JOURNALMODE_OFF ){
4483
4484 /* Open the sub-journal, if it has not already been opened */
4485 assert( pPager->useJournal );
4486 assert( isOpen(pPager->jfd) || pagerUseWal(pPager) );
4487 assert( isOpen(pPager->sjfd) || pPager->nSubRec==0 );
4488 assert( pagerUseWal(pPager)
4489 || pageInJournal(pPager, pPg)
4490 || pPg->pgno>pPager->dbOrigSize
4491 );
4492 rc = openSubJournal(pPager);
4493
4494 /* If the sub-journal was opened successfully (or was already open),
4495 ** write the journal record into the file. */
4496 if( rc==SQLITE_OK ){
4497 void *pData = pPg->pData;
4498 i64 offset = (i64)pPager->nSubRec*(4+pPager->pageSize);
4499 char *pData2;
4500 pData2 = pData;
4501 PAGERTRACE(("STMT-JOURNAL %d page %d\n", PAGERID(pPager), pPg->pgno));
4502 rc = write32bits(pPager->sjfd, offset, pPg->pgno);
4503 if( rc==SQLITE_OK ){
4504 rc = sqlite3OsWrite(pPager->sjfd, pData2, pPager->pageSize, offset+4);
4505 }
4506 }
4507 }
4508 if( rc==SQLITE_OK ){
4509 pPager->nSubRec++;
4510 assert( pPager->nSavepoint>0 );
4511 rc = addToSavepointBitvecs(pPager, pPg->pgno);
4512 }
4513 return rc;
4514}
4515static int subjournalPageIfRequired(PgHdr *pPg){
4516 if( subjRequiresPage(pPg) ){
4517 return subjournalPage(pPg);
4518 }else{
4519 return SQLITE_OK;
4520 }
4521}
4522
4523/*
4524** This function is called by the pcache layer when it has reached some
4525** soft memory limit. The first argument is a pointer to a Pager object
4526** (cast as a void*). The pager is always 'purgeable' (not an in-memory
4527** database). The second argument is a reference to a page that is
4528** currently dirty but has no outstanding references. The page
4529** is always associated with the Pager object passed as the first
4530** argument.
4531**
4532** The job of this function is to make pPg clean by writing its contents
4533** out to the database file, if possible. This may involve syncing the
4534** journal file.
4535**
4536** If successful, sqlite3PcacheMakeClean() is called on the page and
4537** SQLITE_OK returned. If an IO error occurs while trying to make the
4538** page clean, the IO error code is returned. If the page cannot be
4539** made clean for some other reason, but no error occurs, then SQLITE_OK
4540** is returned by sqlite3PcacheMakeClean() is not called.
4541*/
4542static int pagerStress(void *p, PgHdr *pPg){
4543 Pager *pPager = (Pager *)p;
4544 int rc = SQLITE_OK;
4545
4546 assert( pPg->pPager==pPager );
4547 assert( pPg->flags&PGHDR_DIRTY );
4548
4549 /* The doNotSpill NOSYNC bit is set during times when doing a sync of
4550 ** journal (and adding a new header) is not allowed. This occurs
4551 ** during calls to sqlite3PagerWrite() while trying to journal multiple
4552 ** pages belonging to the same sector.
4553 **
4554 ** The doNotSpill ROLLBACK and OFF bits inhibits all cache spilling
4555 ** regardless of whether or not a sync is required. This is set during
4556 ** a rollback or by user request, respectively.
4557 **
4558 ** Spilling is also prohibited when in an error state since that could
4559 ** lead to database corruption. In the current implementation it
4560 ** is impossible for sqlite3PcacheFetch() to be called with createFlag==3
4561 ** while in the error state, hence it is impossible for this routine to
4562 ** be called in the error state. Nevertheless, we include a NEVER()
4563 ** test for the error state as a safeguard against future changes.
4564 */
4565 if( NEVER(pPager->errCode) ) return SQLITE_OK;
4566 testcase( pPager->doNotSpill & SPILLFLAG_ROLLBACK );
4567 testcase( pPager->doNotSpill & SPILLFLAG_OFF );
4568 testcase( pPager->doNotSpill & SPILLFLAG_NOSYNC );
4569 if( pPager->doNotSpill
4570 && ((pPager->doNotSpill & (SPILLFLAG_ROLLBACK|SPILLFLAG_OFF))!=0
4571 || (pPg->flags & PGHDR_NEED_SYNC)!=0)
4572 ){
4573 return SQLITE_OK;
4574 }
4575
4576 pPager->aStat[PAGER_STAT_SPILL]++;
4577 pPg->pDirty = 0;
4578 if( pagerUseWal(pPager) ){
4579 /* Write a single frame for this page to the log. */
4580 rc = subjournalPageIfRequired(pPg);
4581 if( rc==SQLITE_OK ){
4582 rc = pagerWalFrames(pPager, pPg, 0, 0);
4583 }
4584 }else{
4585
4586#ifdef SQLITE_ENABLE_BATCH_ATOMIC_WRITE
4587 if( pPager->tempFile==0 ){
4588 rc = sqlite3JournalCreate(pPager->jfd);
4589 if( rc!=SQLITE_OK ) return pager_error(pPager, rc);
4590 }
4591#endif
4592
4593 /* Sync the journal file if required. */
4594 if( pPg->flags&PGHDR_NEED_SYNC
4595 || pPager->eState==PAGER_WRITER_CACHEMOD
4596 ){
4597 rc = syncJournal(pPager, 1);
4598 }
4599
4600 /* Write the contents of the page out to the database file. */
4601 if( rc==SQLITE_OK ){
4602 assert( (pPg->flags&PGHDR_NEED_SYNC)==0 );
4603 rc = pager_write_pagelist(pPager, pPg);
4604 }
4605 }
4606
4607 /* Mark the page as clean. */
4608 if( rc==SQLITE_OK ){
4609 PAGERTRACE(("STRESS %d page %d\n", PAGERID(pPager), pPg->pgno));
4610 sqlite3PcacheMakeClean(pPg);
4611 }
4612
4613 return pager_error(pPager, rc);
4614}
4615
4616/*
4617** Flush all unreferenced dirty pages to disk.
4618*/
4619int sqlite3PagerFlush(Pager *pPager){
4620 int rc = pPager->errCode;
4621 if( !MEMDB ){
4622 PgHdr *pList = sqlite3PcacheDirtyList(pPager->pPCache);
4623 assert( assert_pager_state(pPager) );
4624 while( rc==SQLITE_OK && pList ){
4625 PgHdr *pNext = pList->pDirty;
4626 if( pList->nRef==0 ){
4627 rc = pagerStress((void*)pPager, pList);
4628 }
4629 pList = pNext;
4630 }
4631 }
4632
4633 return rc;
4634}
4635
4636/*
4637** Allocate and initialize a new Pager object and put a pointer to it
4638** in *ppPager. The pager should eventually be freed by passing it
4639** to sqlite3PagerClose().
4640**
4641** The zFilename argument is the path to the database file to open.
4642** If zFilename is NULL then a randomly-named temporary file is created
4643** and used as the file to be cached. Temporary files are be deleted
4644** automatically when they are closed. If zFilename is ":memory:" then
4645** all information is held in cache. It is never written to disk.
4646** This can be used to implement an in-memory database.
4647**
4648** The nExtra parameter specifies the number of bytes of space allocated
4649** along with each page reference. This space is available to the user
4650** via the sqlite3PagerGetExtra() API. When a new page is allocated, the
4651** first 8 bytes of this space are zeroed but the remainder is uninitialized.
4652** (The extra space is used by btree as the MemPage object.)
4653**
4654** The flags argument is used to specify properties that affect the
4655** operation of the pager. It should be passed some bitwise combination
4656** of the PAGER_* flags.
4657**
4658** The vfsFlags parameter is a bitmask to pass to the flags parameter
4659** of the xOpen() method of the supplied VFS when opening files.
4660**
4661** If the pager object is allocated and the specified file opened
4662** successfully, SQLITE_OK is returned and *ppPager set to point to
4663** the new pager object. If an error occurs, *ppPager is set to NULL
4664** and error code returned. This function may return SQLITE_NOMEM
4665** (sqlite3Malloc() is used to allocate memory), SQLITE_CANTOPEN or
4666** various SQLITE_IO_XXX errors.
4667*/
4668int sqlite3PagerOpen(
4669 sqlite3_vfs *pVfs, /* The virtual file system to use */
4670 Pager **ppPager, /* OUT: Return the Pager structure here */
4671 const char *zFilename, /* Name of the database file to open */
4672 int nExtra, /* Extra bytes append to each in-memory page */
4673 int flags, /* flags controlling this file */
4674 int vfsFlags, /* flags passed through to sqlite3_vfs.xOpen() */
4675 void (*xReinit)(DbPage*) /* Function to reinitialize pages */
4676){
4677 u8 *pPtr;
4678 Pager *pPager = 0; /* Pager object to allocate and return */
4679 int rc = SQLITE_OK; /* Return code */
4680 int tempFile = 0; /* True for temp files (incl. in-memory files) */
4681 int memDb = 0; /* True if this is an in-memory file */
4682#ifndef SQLITE_OMIT_DESERIALIZE
4683 int memJM = 0; /* Memory journal mode */
4684#else
4685# define memJM 0
4686#endif
4687 int readOnly = 0; /* True if this is a read-only file */
4688 int journalFileSize; /* Bytes to allocate for each journal fd */
4689 char *zPathname = 0; /* Full path to database file */
4690 int nPathname = 0; /* Number of bytes in zPathname */
4691 int useJournal = (flags & PAGER_OMIT_JOURNAL)==0; /* False to omit journal */
4692 int pcacheSize = sqlite3PcacheSize(); /* Bytes to allocate for PCache */
4693 u32 szPageDflt = SQLITE_DEFAULT_PAGE_SIZE; /* Default page size */
4694 const char *zUri = 0; /* URI args to copy */
4695 int nUriByte = 1; /* Number of bytes of URI args at *zUri */
4696 int nUri = 0; /* Number of URI parameters */
4697
4698 /* Figure out how much space is required for each journal file-handle
4699 ** (there are two of them, the main journal and the sub-journal). */
4700 journalFileSize = ROUND8(sqlite3JournalSize(pVfs));
4701
4702 /* Set the output variable to NULL in case an error occurs. */
4703 *ppPager = 0;
4704
4705#ifndef SQLITE_OMIT_MEMORYDB
4706 if( flags & PAGER_MEMORY ){
4707 memDb = 1;
4708 if( zFilename && zFilename[0] ){
4709 zPathname = sqlite3DbStrDup(0, zFilename);
4710 if( zPathname==0 ) return SQLITE_NOMEM_BKPT;
4711 nPathname = sqlite3Strlen30(zPathname);
4712 zFilename = 0;
4713 }
4714 }
4715#endif
4716
4717 /* Compute and store the full pathname in an allocated buffer pointed
4718 ** to by zPathname, length nPathname. Or, if this is a temporary file,
4719 ** leave both nPathname and zPathname set to 0.
4720 */
4721 if( zFilename && zFilename[0] ){
4722 const char *z;
4723 nPathname = pVfs->mxPathname+1;
4724 zPathname = sqlite3DbMallocRaw(0, nPathname*2);
4725 if( zPathname==0 ){
4726 return SQLITE_NOMEM_BKPT;
4727 }
4728 zPathname[0] = 0; /* Make sure initialized even if FullPathname() fails */
4729 rc = sqlite3OsFullPathname(pVfs, zFilename, nPathname, zPathname);
4730 if( rc!=SQLITE_OK ){
4731 if( rc==SQLITE_OK_SYMLINK ){
4732 if( vfsFlags & SQLITE_OPEN_NOFOLLOW ){
4733 rc = SQLITE_CANTOPEN_SYMLINK;
4734 }else{
4735 rc = SQLITE_OK;
4736 }
4737 }
4738 }
4739 nPathname = sqlite3Strlen30(zPathname);
4740 z = zUri = &zFilename[sqlite3Strlen30(zFilename)+1];
4741 while( *z ){
4742 z += strlen(z)+1;
4743 z += strlen(z)+1;
4744 nUri++;
4745 }
4746 nUriByte = (int)(&z[1] - zUri);
4747 assert( nUriByte>=1 );
4748 if( rc==SQLITE_OK && nPathname+8>pVfs->mxPathname ){
4749 /* This branch is taken when the journal path required by
4750 ** the database being opened will be more than pVfs->mxPathname
4751 ** bytes in length. This means the database cannot be opened,
4752 ** as it will not be possible to open the journal file or even
4753 ** check for a hot-journal before reading.
4754 */
4755 rc = SQLITE_CANTOPEN_BKPT;
4756 }
4757 if( rc!=SQLITE_OK ){
4758 sqlite3DbFree(0, zPathname);
4759 return rc;
4760 }
4761 }
4762
4763 /* Allocate memory for the Pager structure, PCache object, the
4764 ** three file descriptors, the database file name and the journal
4765 ** file name. The layout in memory is as follows:
4766 **
4767 ** Pager object (sizeof(Pager) bytes)
4768 ** PCache object (sqlite3PcacheSize() bytes)
4769 ** Database file handle (pVfs->szOsFile bytes)
4770 ** Sub-journal file handle (journalFileSize bytes)
4771 ** Main journal file handle (journalFileSize bytes)
4772 ** Ptr back to the Pager (sizeof(Pager*) bytes)
4773 ** \0\0\0\0 database prefix (4 bytes)
4774 ** Database file name (nPathname+1 bytes)
4775 ** URI query parameters (nUriByte bytes)
4776 ** Journal filename (nPathname+8+1 bytes)
4777 ** WAL filename (nPathname+4+1 bytes)
4778 ** \0\0\0 terminator (3 bytes)
4779 **
4780 ** Some 3rd-party software, over which we have no control, depends on
4781 ** the specific order of the filenames and the \0 separators between them
4782 ** so that it can (for example) find the database filename given the WAL
4783 ** filename without using the sqlite3_filename_database() API. This is a
4784 ** misuse of SQLite and a bug in the 3rd-party software, but the 3rd-party
4785 ** software is in widespread use, so we try to avoid changing the filename
4786 ** order and formatting if possible. In particular, the details of the
4787 ** filename format expected by 3rd-party software should be as follows:
4788 **
4789 ** - Main Database Path
4790 ** - \0
4791 ** - Multiple URI components consisting of:
4792 ** - Key
4793 ** - \0
4794 ** - Value
4795 ** - \0
4796 ** - \0
4797 ** - Journal Path
4798 ** - \0
4799 ** - WAL Path (zWALName)
4800 ** - \0
4801 **
4802 ** The sqlite3_create_filename() interface and the databaseFilename() utility
4803 ** that is used by sqlite3_filename_database() and kin also depend on the
4804 ** specific formatting and order of the various filenames, so if the format
4805 ** changes here, be sure to change it there as well.
4806 */
4807 pPtr = (u8 *)sqlite3MallocZero(
4808 ROUND8(sizeof(*pPager)) + /* Pager structure */
4809 ROUND8(pcacheSize) + /* PCache object */
4810 ROUND8(pVfs->szOsFile) + /* The main db file */
4811 journalFileSize * 2 + /* The two journal files */
4812 sizeof(pPager) + /* Space to hold a pointer */
4813 4 + /* Database prefix */
4814 nPathname + 1 + /* database filename */
4815 nUriByte + /* query parameters */
4816 nPathname + 8 + 1 + /* Journal filename */
4817#ifndef SQLITE_OMIT_WAL
4818 nPathname + 4 + 1 + /* WAL filename */
4819#endif
4820 3 /* Terminator */
4821 );
4822 assert( EIGHT_BYTE_ALIGNMENT(SQLITE_INT_TO_PTR(journalFileSize)) );
4823 if( !pPtr ){
4824 sqlite3DbFree(0, zPathname);
4825 return SQLITE_NOMEM_BKPT;
4826 }
4827 pPager = (Pager*)pPtr; pPtr += ROUND8(sizeof(*pPager));
4828 pPager->pPCache = (PCache*)pPtr; pPtr += ROUND8(pcacheSize);
4829 pPager->fd = (sqlite3_file*)pPtr; pPtr += ROUND8(pVfs->szOsFile);
4830 pPager->sjfd = (sqlite3_file*)pPtr; pPtr += journalFileSize;
4831 pPager->jfd = (sqlite3_file*)pPtr; pPtr += journalFileSize;
4832 assert( EIGHT_BYTE_ALIGNMENT(pPager->jfd) );
4833 memcpy(pPtr, &pPager, sizeof(pPager)); pPtr += sizeof(pPager);
4834
4835 /* Fill in the Pager.zFilename and pPager.zQueryParam fields */
4836 pPtr += 4; /* Skip zero prefix */
4837 pPager->zFilename = (char*)pPtr;
4838 if( nPathname>0 ){
4839 memcpy(pPtr, zPathname, nPathname); pPtr += nPathname + 1;
4840 if( zUri ){
4841 memcpy(pPtr, zUri, nUriByte); pPtr += nUriByte;
4842 }else{
4843 pPtr++;
4844 }
4845 }
4846
4847
4848 /* Fill in Pager.zJournal */
4849 if( nPathname>0 ){
4850 pPager->zJournal = (char*)pPtr;
4851 memcpy(pPtr, zPathname, nPathname); pPtr += nPathname;
4852 memcpy(pPtr, "-journal",8); pPtr += 8 + 1;
4853#ifdef SQLITE_ENABLE_8_3_NAMES
4854 sqlite3FileSuffix3(zFilename,pPager->zJournal);
4855 pPtr = (u8*)(pPager->zJournal + sqlite3Strlen30(pPager->zJournal)+1);
4856#endif
4857 }else{
4858 pPager->zJournal = 0;
4859 }
4860
4861#ifndef SQLITE_OMIT_WAL
4862 /* Fill in Pager.zWal */
4863 if( nPathname>0 ){
4864 pPager->zWal = (char*)pPtr;
4865 memcpy(pPtr, zPathname, nPathname); pPtr += nPathname;
4866 memcpy(pPtr, "-wal", 4); pPtr += 4 + 1;
4867#ifdef SQLITE_ENABLE_8_3_NAMES
4868 sqlite3FileSuffix3(zFilename, pPager->zWal);
4869 pPtr = (u8*)(pPager->zWal + sqlite3Strlen30(pPager->zWal)+1);
4870#endif
4871 }else{
4872 pPager->zWal = 0;
4873 }
4874#endif
4875 (void)pPtr; /* Suppress warning about unused pPtr value */
4876
4877 if( nPathname ) sqlite3DbFree(0, zPathname);
4878 pPager->pVfs = pVfs;
4879 pPager->vfsFlags = vfsFlags;
4880
4881 /* Open the pager file.
4882 */
4883 if( zFilename && zFilename[0] ){
4884 int fout = 0; /* VFS flags returned by xOpen() */
4885 rc = sqlite3OsOpen(pVfs, pPager->zFilename, pPager->fd, vfsFlags, &fout);
4886 assert( !memDb );
4887#ifndef SQLITE_OMIT_DESERIALIZE
4888 pPager->memVfs = memJM = (fout&SQLITE_OPEN_MEMORY)!=0;
4889#endif
4890 readOnly = (fout&SQLITE_OPEN_READONLY)!=0;
4891
4892 /* If the file was successfully opened for read/write access,
4893 ** choose a default page size in case we have to create the
4894 ** database file. The default page size is the maximum of:
4895 **
4896 ** + SQLITE_DEFAULT_PAGE_SIZE,
4897 ** + The value returned by sqlite3OsSectorSize()
4898 ** + The largest page size that can be written atomically.
4899 */
4900 if( rc==SQLITE_OK ){
4901 int iDc = sqlite3OsDeviceCharacteristics(pPager->fd);
4902 if( !readOnly ){
4903 setSectorSize(pPager);
4904 assert(SQLITE_DEFAULT_PAGE_SIZE<=SQLITE_MAX_DEFAULT_PAGE_SIZE);
4905 if( szPageDflt<pPager->sectorSize ){
4906 if( pPager->sectorSize>SQLITE_MAX_DEFAULT_PAGE_SIZE ){
4907 szPageDflt = SQLITE_MAX_DEFAULT_PAGE_SIZE;
4908 }else{
4909 szPageDflt = (u32)pPager->sectorSize;
4910 }
4911 }
4912#ifdef SQLITE_ENABLE_ATOMIC_WRITE
4913 {
4914 int ii;
4915 assert(SQLITE_IOCAP_ATOMIC512==(512>>8));
4916 assert(SQLITE_IOCAP_ATOMIC64K==(65536>>8));
4917 assert(SQLITE_MAX_DEFAULT_PAGE_SIZE<=65536);
4918 for(ii=szPageDflt; ii<=SQLITE_MAX_DEFAULT_PAGE_SIZE; ii=ii*2){
4919 if( iDc&(SQLITE_IOCAP_ATOMIC|(ii>>8)) ){
4920 szPageDflt = ii;
4921 }
4922 }
4923 }
4924#endif
4925 }
4926 pPager->noLock = sqlite3_uri_boolean(pPager->zFilename, "nolock", 0);
4927 if( (iDc & SQLITE_IOCAP_IMMUTABLE)!=0
4928 || sqlite3_uri_boolean(pPager->zFilename, "immutable", 0) ){
4929 vfsFlags |= SQLITE_OPEN_READONLY;
4930 goto act_like_temp_file;
4931 }
4932 }
4933 }else{
4934 /* If a temporary file is requested, it is not opened immediately.
4935 ** In this case we accept the default page size and delay actually
4936 ** opening the file until the first call to OsWrite().
4937 **
4938 ** This branch is also run for an in-memory database. An in-memory
4939 ** database is the same as a temp-file that is never written out to
4940 ** disk and uses an in-memory rollback journal.
4941 **
4942 ** This branch also runs for files marked as immutable.
4943 */
4944act_like_temp_file:
4945 tempFile = 1;
4946 pPager->eState = PAGER_READER; /* Pretend we already have a lock */
4947 pPager->eLock = EXCLUSIVE_LOCK; /* Pretend we are in EXCLUSIVE mode */
4948 pPager->noLock = 1; /* Do no locking */
4949 readOnly = (vfsFlags&SQLITE_OPEN_READONLY);
4950 }
4951
4952 /* The following call to PagerSetPagesize() serves to set the value of
4953 ** Pager.pageSize and to allocate the Pager.pTmpSpace buffer.
4954 */
4955 if( rc==SQLITE_OK ){
4956 assert( pPager->memDb==0 );
4957 rc = sqlite3PagerSetPagesize(pPager, &szPageDflt, -1);
4958 testcase( rc!=SQLITE_OK );
4959 }
4960
4961 /* Initialize the PCache object. */
4962 if( rc==SQLITE_OK ){
4963 nExtra = ROUND8(nExtra);
4964 assert( nExtra>=8 && nExtra<1000 );
4965 rc = sqlite3PcacheOpen(szPageDflt, nExtra, !memDb,
4966 !memDb?pagerStress:0, (void *)pPager, pPager->pPCache);
4967 }
4968
4969 /* If an error occurred above, free the Pager structure and close the file.
4970 */
4971 if( rc!=SQLITE_OK ){
4972 sqlite3OsClose(pPager->fd);
4973 sqlite3PageFree(pPager->pTmpSpace);
4974 sqlite3_free(pPager);
4975 return rc;
4976 }
4977
4978 PAGERTRACE(("OPEN %d %s\n", FILEHANDLEID(pPager->fd), pPager->zFilename));
4979 IOTRACE(("OPEN %p %s\n", pPager, pPager->zFilename))
4980
4981 pPager->useJournal = (u8)useJournal;
4982 /* pPager->stmtOpen = 0; */
4983 /* pPager->stmtInUse = 0; */
4984 /* pPager->nRef = 0; */
4985 /* pPager->stmtSize = 0; */
4986 /* pPager->stmtJSize = 0; */
4987 /* pPager->nPage = 0; */
4988 pPager->mxPgno = SQLITE_MAX_PAGE_COUNT;
4989 /* pPager->state = PAGER_UNLOCK; */
4990 /* pPager->errMask = 0; */
4991 pPager->tempFile = (u8)tempFile;
4992 assert( tempFile==PAGER_LOCKINGMODE_NORMAL
4993 || tempFile==PAGER_LOCKINGMODE_EXCLUSIVE );
4994 assert( PAGER_LOCKINGMODE_EXCLUSIVE==1 );
4995 pPager->exclusiveMode = (u8)tempFile;
4996 pPager->changeCountDone = pPager->tempFile;
4997 pPager->memDb = (u8)memDb;
4998 pPager->readOnly = (u8)readOnly;
4999 assert( useJournal || pPager->tempFile );
5000 pPager->noSync = pPager->tempFile;
5001 if( pPager->noSync ){
5002 assert( pPager->fullSync==0 );
5003 assert( pPager->extraSync==0 );
5004 assert( pPager->syncFlags==0 );
5005 assert( pPager->walSyncFlags==0 );
5006 }else{
5007 pPager->fullSync = 1;
5008 pPager->extraSync = 0;
5009 pPager->syncFlags = SQLITE_SYNC_NORMAL;
5010 pPager->walSyncFlags = SQLITE_SYNC_NORMAL | (SQLITE_SYNC_NORMAL<<2);
5011 }
5012 /* pPager->pFirst = 0; */
5013 /* pPager->pFirstSynced = 0; */
5014 /* pPager->pLast = 0; */
5015 pPager->nExtra = (u16)nExtra;
5016 pPager->journalSizeLimit = SQLITE_DEFAULT_JOURNAL_SIZE_LIMIT;
5017 assert( isOpen(pPager->fd) || tempFile );
5018 setSectorSize(pPager);
5019 if( !useJournal ){
5020 pPager->journalMode = PAGER_JOURNALMODE_OFF;
5021 }else if( memDb || memJM ){
5022 pPager->journalMode = PAGER_JOURNALMODE_MEMORY;
5023 }
5024 /* pPager->xBusyHandler = 0; */
5025 /* pPager->pBusyHandlerArg = 0; */
5026 pPager->xReiniter = xReinit;
5027 setGetterMethod(pPager);
5028 /* memset(pPager->aHash, 0, sizeof(pPager->aHash)); */
5029 /* pPager->szMmap = SQLITE_DEFAULT_MMAP_SIZE // will be set by btree.c */
5030
5031 *ppPager = pPager;
5032 return SQLITE_OK;
5033}
5034
5035/*
5036** Return the sqlite3_file for the main database given the name
5037** of the corresonding WAL or Journal name as passed into
5038** xOpen.
5039*/
5040sqlite3_file *sqlite3_database_file_object(const char *zName){
5041 Pager *pPager;
5042 while( zName[-1]!=0 || zName[-2]!=0 || zName[-3]!=0 || zName[-4]!=0 ){
5043 zName--;
5044 }
5045 pPager = *(Pager**)(zName - 4 - sizeof(Pager*));
5046 return pPager->fd;
5047}
5048
5049
5050/*
5051** This function is called after transitioning from PAGER_UNLOCK to
5052** PAGER_SHARED state. It tests if there is a hot journal present in
5053** the file-system for the given pager. A hot journal is one that
5054** needs to be played back. According to this function, a hot-journal
5055** file exists if the following criteria are met:
5056**
5057** * The journal file exists in the file system, and
5058** * No process holds a RESERVED or greater lock on the database file, and
5059** * The database file itself is greater than 0 bytes in size, and
5060** * The first byte of the journal file exists and is not 0x00.
5061**
5062** If the current size of the database file is 0 but a journal file
5063** exists, that is probably an old journal left over from a prior
5064** database with the same name. In this case the journal file is
5065** just deleted using OsDelete, *pExists is set to 0 and SQLITE_OK
5066** is returned.
5067**
5068** This routine does not check if there is a super-journal filename
5069** at the end of the file. If there is, and that super-journal file
5070** does not exist, then the journal file is not really hot. In this
5071** case this routine will return a false-positive. The pager_playback()
5072** routine will discover that the journal file is not really hot and
5073** will not roll it back.
5074**
5075** If a hot-journal file is found to exist, *pExists is set to 1 and
5076** SQLITE_OK returned. If no hot-journal file is present, *pExists is
5077** set to 0 and SQLITE_OK returned. If an IO error occurs while trying
5078** to determine whether or not a hot-journal file exists, the IO error
5079** code is returned and the value of *pExists is undefined.
5080*/
5081static int hasHotJournal(Pager *pPager, int *pExists){
5082 sqlite3_vfs * const pVfs = pPager->pVfs;
5083 int rc = SQLITE_OK; /* Return code */
5084 int exists = 1; /* True if a journal file is present */
5085 int jrnlOpen = !!isOpen(pPager->jfd);
5086
5087 assert( pPager->useJournal );
5088 assert( isOpen(pPager->fd) );
5089 assert( pPager->eState==PAGER_OPEN );
5090
5091 assert( jrnlOpen==0 || ( sqlite3OsDeviceCharacteristics(pPager->jfd) &
5092 SQLITE_IOCAP_UNDELETABLE_WHEN_OPEN
5093 ));
5094
5095 *pExists = 0;
5096 if( !jrnlOpen ){
5097 rc = sqlite3OsAccess(pVfs, pPager->zJournal, SQLITE_ACCESS_EXISTS, &exists);
5098 }
5099 if( rc==SQLITE_OK && exists ){
5100 int locked = 0; /* True if some process holds a RESERVED lock */
5101
5102 /* Race condition here: Another process might have been holding the
5103 ** the RESERVED lock and have a journal open at the sqlite3OsAccess()
5104 ** call above, but then delete the journal and drop the lock before
5105 ** we get to the following sqlite3OsCheckReservedLock() call. If that
5106 ** is the case, this routine might think there is a hot journal when
5107 ** in fact there is none. This results in a false-positive which will
5108 ** be dealt with by the playback routine. Ticket #3883.
5109 */
5110 rc = sqlite3OsCheckReservedLock(pPager->fd, &locked);
5111 if( rc==SQLITE_OK && !locked ){
5112 Pgno nPage; /* Number of pages in database file */
5113
5114 assert( pPager->tempFile==0 );
5115 rc = pagerPagecount(pPager, &nPage);
5116 if( rc==SQLITE_OK ){
5117 /* If the database is zero pages in size, that means that either (1) the
5118 ** journal is a remnant from a prior database with the same name where
5119 ** the database file but not the journal was deleted, or (2) the initial
5120 ** transaction that populates a new database is being rolled back.
5121 ** In either case, the journal file can be deleted. However, take care
5122 ** not to delete the journal file if it is already open due to
5123 ** journal_mode=PERSIST.
5124 */
5125 if( nPage==0 && !jrnlOpen ){
5126 sqlite3BeginBenignMalloc();
5127 if( pagerLockDb(pPager, RESERVED_LOCK)==SQLITE_OK ){
5128 sqlite3OsDelete(pVfs, pPager->zJournal, 0);
5129 if( !pPager->exclusiveMode ) pagerUnlockDb(pPager, SHARED_LOCK);
5130 }
5131 sqlite3EndBenignMalloc();
5132 }else{
5133 /* The journal file exists and no other connection has a reserved
5134 ** or greater lock on the database file. Now check that there is
5135 ** at least one non-zero bytes at the start of the journal file.
5136 ** If there is, then we consider this journal to be hot. If not,
5137 ** it can be ignored.
5138 */
5139 if( !jrnlOpen ){
5140 int f = SQLITE_OPEN_READONLY|SQLITE_OPEN_MAIN_JOURNAL;
5141 rc = sqlite3OsOpen(pVfs, pPager->zJournal, pPager->jfd, f, &f);
5142 }
5143 if( rc==SQLITE_OK ){
5144 u8 first = 0;
5145 rc = sqlite3OsRead(pPager->jfd, (void *)&first, 1, 0);
5146 if( rc==SQLITE_IOERR_SHORT_READ ){
5147 rc = SQLITE_OK;
5148 }
5149 if( !jrnlOpen ){
5150 sqlite3OsClose(pPager->jfd);
5151 }
5152 *pExists = (first!=0);
5153 }else if( rc==SQLITE_CANTOPEN ){
5154 /* If we cannot open the rollback journal file in order to see if
5155 ** it has a zero header, that might be due to an I/O error, or
5156 ** it might be due to the race condition described above and in
5157 ** ticket #3883. Either way, assume that the journal is hot.
5158 ** This might be a false positive. But if it is, then the
5159 ** automatic journal playback and recovery mechanism will deal
5160 ** with it under an EXCLUSIVE lock where we do not need to
5161 ** worry so much with race conditions.
5162 */
5163 *pExists = 1;
5164 rc = SQLITE_OK;
5165 }
5166 }
5167 }
5168 }
5169 }
5170
5171 return rc;
5172}
5173
5174/*
5175** This function is called to obtain a shared lock on the database file.
5176** It is illegal to call sqlite3PagerGet() until after this function
5177** has been successfully called. If a shared-lock is already held when
5178** this function is called, it is a no-op.
5179**
5180** The following operations are also performed by this function.
5181**
5182** 1) If the pager is currently in PAGER_OPEN state (no lock held
5183** on the database file), then an attempt is made to obtain a
5184** SHARED lock on the database file. Immediately after obtaining
5185** the SHARED lock, the file-system is checked for a hot-journal,
5186** which is played back if present. Following any hot-journal
5187** rollback, the contents of the cache are validated by checking
5188** the 'change-counter' field of the database file header and
5189** discarded if they are found to be invalid.
5190**
5191** 2) If the pager is running in exclusive-mode, and there are currently
5192** no outstanding references to any pages, and is in the error state,
5193** then an attempt is made to clear the error state by discarding
5194** the contents of the page cache and rolling back any open journal
5195** file.
5196**
5197** If everything is successful, SQLITE_OK is returned. If an IO error
5198** occurs while locking the database, checking for a hot-journal file or
5199** rolling back a journal file, the IO error code is returned.
5200*/
5201int sqlite3PagerSharedLock(Pager *pPager){
5202 int rc = SQLITE_OK; /* Return code */
5203
5204 /* This routine is only called from b-tree and only when there are no
5205 ** outstanding pages. This implies that the pager state should either
5206 ** be OPEN or READER. READER is only possible if the pager is or was in
5207 ** exclusive access mode. */
5208 assert( sqlite3PcacheRefCount(pPager->pPCache)==0 );
5209 assert( assert_pager_state(pPager) );
5210 assert( pPager->eState==PAGER_OPEN || pPager->eState==PAGER_READER );
5211 assert( pPager->errCode==SQLITE_OK );
5212
5213 if( !pagerUseWal(pPager) && pPager->eState==PAGER_OPEN ){
5214 int bHotJournal = 1; /* True if there exists a hot journal-file */
5215
5216 assert( !MEMDB );
5217 assert( pPager->tempFile==0 || pPager->eLock==EXCLUSIVE_LOCK );
5218
5219 rc = pager_wait_on_lock(pPager, SHARED_LOCK);
5220 if( rc!=SQLITE_OK ){
5221 assert( pPager->eLock==NO_LOCK || pPager->eLock==UNKNOWN_LOCK );
5222 goto failed;
5223 }
5224
5225 /* If a journal file exists, and there is no RESERVED lock on the
5226 ** database file, then it either needs to be played back or deleted.
5227 */
5228 if( pPager->eLock<=SHARED_LOCK ){
5229 rc = hasHotJournal(pPager, &bHotJournal);
5230 }
5231 if( rc!=SQLITE_OK ){
5232 goto failed;
5233 }
5234 if( bHotJournal ){
5235 if( pPager->readOnly ){
5236 rc = SQLITE_READONLY_ROLLBACK;
5237 goto failed;
5238 }
5239
5240 /* Get an EXCLUSIVE lock on the database file. At this point it is
5241 ** important that a RESERVED lock is not obtained on the way to the
5242 ** EXCLUSIVE lock. If it were, another process might open the
5243 ** database file, detect the RESERVED lock, and conclude that the
5244 ** database is safe to read while this process is still rolling the
5245 ** hot-journal back.
5246 **
5247 ** Because the intermediate RESERVED lock is not requested, any
5248 ** other process attempting to access the database file will get to
5249 ** this point in the code and fail to obtain its own EXCLUSIVE lock
5250 ** on the database file.
5251 **
5252 ** Unless the pager is in locking_mode=exclusive mode, the lock is
5253 ** downgraded to SHARED_LOCK before this function returns.
5254 */
5255 rc = pagerLockDb(pPager, EXCLUSIVE_LOCK);
5256 if( rc!=SQLITE_OK ){
5257 goto failed;
5258 }
5259
5260 /* If it is not already open and the file exists on disk, open the
5261 ** journal for read/write access. Write access is required because
5262 ** in exclusive-access mode the file descriptor will be kept open
5263 ** and possibly used for a transaction later on. Also, write-access
5264 ** is usually required to finalize the journal in journal_mode=persist
5265 ** mode (and also for journal_mode=truncate on some systems).
5266 **
5267 ** If the journal does not exist, it usually means that some
5268 ** other connection managed to get in and roll it back before
5269 ** this connection obtained the exclusive lock above. Or, it
5270 ** may mean that the pager was in the error-state when this
5271 ** function was called and the journal file does not exist.
5272 */
5273 if( !isOpen(pPager->jfd) && pPager->journalMode!=PAGER_JOURNALMODE_OFF ){
5274 sqlite3_vfs * const pVfs = pPager->pVfs;
5275 int bExists; /* True if journal file exists */
5276 rc = sqlite3OsAccess(
5277 pVfs, pPager->zJournal, SQLITE_ACCESS_EXISTS, &bExists);
5278 if( rc==SQLITE_OK && bExists ){
5279 int fout = 0;
5280 int f = SQLITE_OPEN_READWRITE|SQLITE_OPEN_MAIN_JOURNAL;
5281 assert( !pPager->tempFile );
5282 rc = sqlite3OsOpen(pVfs, pPager->zJournal, pPager->jfd, f, &fout);
5283 assert( rc!=SQLITE_OK || isOpen(pPager->jfd) );
5284 if( rc==SQLITE_OK && fout&SQLITE_OPEN_READONLY ){
5285 rc = SQLITE_CANTOPEN_BKPT;
5286 sqlite3OsClose(pPager->jfd);
5287 }
5288 }
5289 }
5290
5291 /* Playback and delete the journal. Drop the database write
5292 ** lock and reacquire the read lock. Purge the cache before
5293 ** playing back the hot-journal so that we don't end up with
5294 ** an inconsistent cache. Sync the hot journal before playing
5295 ** it back since the process that crashed and left the hot journal
5296 ** probably did not sync it and we are required to always sync
5297 ** the journal before playing it back.
5298 */
5299 if( isOpen(pPager->jfd) ){
5300 assert( rc==SQLITE_OK );
5301 rc = pagerSyncHotJournal(pPager);
5302 if( rc==SQLITE_OK ){
5303 rc = pager_playback(pPager, !pPager->tempFile);
5304 pPager->eState = PAGER_OPEN;
5305 }
5306 }else if( !pPager->exclusiveMode ){
5307 pagerUnlockDb(pPager, SHARED_LOCK);
5308 }
5309
5310 if( rc!=SQLITE_OK ){
5311 /* This branch is taken if an error occurs while trying to open
5312 ** or roll back a hot-journal while holding an EXCLUSIVE lock. The
5313 ** pager_unlock() routine will be called before returning to unlock
5314 ** the file. If the unlock attempt fails, then Pager.eLock must be
5315 ** set to UNKNOWN_LOCK (see the comment above the #define for
5316 ** UNKNOWN_LOCK above for an explanation).
5317 **
5318 ** In order to get pager_unlock() to do this, set Pager.eState to
5319 ** PAGER_ERROR now. This is not actually counted as a transition
5320 ** to ERROR state in the state diagram at the top of this file,
5321 ** since we know that the same call to pager_unlock() will very
5322 ** shortly transition the pager object to the OPEN state. Calling
5323 ** assert_pager_state() would fail now, as it should not be possible
5324 ** to be in ERROR state when there are zero outstanding page
5325 ** references.
5326 */
5327 pager_error(pPager, rc);
5328 goto failed;
5329 }
5330
5331 assert( pPager->eState==PAGER_OPEN );
5332 assert( (pPager->eLock==SHARED_LOCK)
5333 || (pPager->exclusiveMode && pPager->eLock>SHARED_LOCK)
5334 );
5335 }
5336
5337 if( !pPager->tempFile && pPager->hasHeldSharedLock ){
5338 /* The shared-lock has just been acquired then check to
5339 ** see if the database has been modified. If the database has changed,
5340 ** flush the cache. The hasHeldSharedLock flag prevents this from
5341 ** occurring on the very first access to a file, in order to save a
5342 ** single unnecessary sqlite3OsRead() call at the start-up.
5343 **
5344 ** Database changes are detected by looking at 15 bytes beginning
5345 ** at offset 24 into the file. The first 4 of these 16 bytes are
5346 ** a 32-bit counter that is incremented with each change. The
5347 ** other bytes change randomly with each file change when
5348 ** a codec is in use.
5349 **
5350 ** There is a vanishingly small chance that a change will not be
5351 ** detected. The chance of an undetected change is so small that
5352 ** it can be neglected.
5353 */
5354 char dbFileVers[sizeof(pPager->dbFileVers)];
5355
5356 IOTRACE(("CKVERS %p %d\n", pPager, sizeof(dbFileVers)));
5357 rc = sqlite3OsRead(pPager->fd, &dbFileVers, sizeof(dbFileVers), 24);
5358 if( rc!=SQLITE_OK ){
5359 if( rc!=SQLITE_IOERR_SHORT_READ ){
5360 goto failed;
5361 }
5362 memset(dbFileVers, 0, sizeof(dbFileVers));
5363 }
5364
5365 if( memcmp(pPager->dbFileVers, dbFileVers, sizeof(dbFileVers))!=0 ){
5366 pager_reset(pPager);
5367
5368 /* Unmap the database file. It is possible that external processes
5369 ** may have truncated the database file and then extended it back
5370 ** to its original size while this process was not holding a lock.
5371 ** In this case there may exist a Pager.pMap mapping that appears
5372 ** to be the right size but is not actually valid. Avoid this
5373 ** possibility by unmapping the db here. */
5374 if( USEFETCH(pPager) ){
5375 sqlite3OsUnfetch(pPager->fd, 0, 0);
5376 }
5377 }
5378 }
5379
5380 /* If there is a WAL file in the file-system, open this database in WAL
5381 ** mode. Otherwise, the following function call is a no-op.
5382 */
5383 rc = pagerOpenWalIfPresent(pPager);
5384#ifndef SQLITE_OMIT_WAL
5385 assert( pPager->pWal==0 || rc==SQLITE_OK );
5386#endif
5387 }
5388
5389 if( pagerUseWal(pPager) ){
5390 assert( rc==SQLITE_OK );
5391 rc = pagerBeginReadTransaction(pPager);
5392 }
5393
5394 if( pPager->tempFile==0 && pPager->eState==PAGER_OPEN && rc==SQLITE_OK ){
5395 rc = pagerPagecount(pPager, &pPager->dbSize);
5396 }
5397
5398 failed:
5399 if( rc!=SQLITE_OK ){
5400 assert( !MEMDB );
5401 pager_unlock(pPager);
5402 assert( pPager->eState==PAGER_OPEN );
5403 }else{
5404 pPager->eState = PAGER_READER;
5405 pPager->hasHeldSharedLock = 1;
5406 }
5407 return rc;
5408}
5409
5410/*
5411** If the reference count has reached zero, rollback any active
5412** transaction and unlock the pager.
5413**
5414** Except, in locking_mode=EXCLUSIVE when there is nothing to in
5415** the rollback journal, the unlock is not performed and there is
5416** nothing to rollback, so this routine is a no-op.
5417*/
5418static void pagerUnlockIfUnused(Pager *pPager){
5419 if( sqlite3PcacheRefCount(pPager->pPCache)==0 ){
5420 assert( pPager->nMmapOut==0 ); /* because page1 is never memory mapped */
5421 pagerUnlockAndRollback(pPager);
5422 }
5423}
5424
5425/*
5426** The page getter methods each try to acquire a reference to a
5427** page with page number pgno. If the requested reference is
5428** successfully obtained, it is copied to *ppPage and SQLITE_OK returned.
5429**
5430** There are different implementations of the getter method depending
5431** on the current state of the pager.
5432**
5433** getPageNormal() -- The normal getter
5434** getPageError() -- Used if the pager is in an error state
5435** getPageMmap() -- Used if memory-mapped I/O is enabled
5436**
5437** If the requested page is already in the cache, it is returned.
5438** Otherwise, a new page object is allocated and populated with data
5439** read from the database file. In some cases, the pcache module may
5440** choose not to allocate a new page object and may reuse an existing
5441** object with no outstanding references.
5442**
5443** The extra data appended to a page is always initialized to zeros the
5444** first time a page is loaded into memory. If the page requested is
5445** already in the cache when this function is called, then the extra
5446** data is left as it was when the page object was last used.
5447**
5448** If the database image is smaller than the requested page or if
5449** the flags parameter contains the PAGER_GET_NOCONTENT bit and the
5450** requested page is not already stored in the cache, then no
5451** actual disk read occurs. In this case the memory image of the
5452** page is initialized to all zeros.
5453**
5454** If PAGER_GET_NOCONTENT is true, it means that we do not care about
5455** the contents of the page. This occurs in two scenarios:
5456**
5457** a) When reading a free-list leaf page from the database, and
5458**
5459** b) When a savepoint is being rolled back and we need to load
5460** a new page into the cache to be filled with the data read
5461** from the savepoint journal.
5462**
5463** If PAGER_GET_NOCONTENT is true, then the data returned is zeroed instead
5464** of being read from the database. Additionally, the bits corresponding
5465** to pgno in Pager.pInJournal (bitvec of pages already written to the
5466** journal file) and the PagerSavepoint.pInSavepoint bitvecs of any open
5467** savepoints are set. This means if the page is made writable at any
5468** point in the future, using a call to sqlite3PagerWrite(), its contents
5469** will not be journaled. This saves IO.
5470**
5471** The acquisition might fail for several reasons. In all cases,
5472** an appropriate error code is returned and *ppPage is set to NULL.
5473**
5474** See also sqlite3PagerLookup(). Both this routine and Lookup() attempt
5475** to find a page in the in-memory cache first. If the page is not already
5476** in memory, this routine goes to disk to read it in whereas Lookup()
5477** just returns 0. This routine acquires a read-lock the first time it
5478** has to go to disk, and could also playback an old journal if necessary.
5479** Since Lookup() never goes to disk, it never has to deal with locks
5480** or journal files.
5481*/
5482static int getPageNormal(
5483 Pager *pPager, /* The pager open on the database file */
5484 Pgno pgno, /* Page number to fetch */
5485 DbPage **ppPage, /* Write a pointer to the page here */
5486 int flags /* PAGER_GET_XXX flags */
5487){
5488 int rc = SQLITE_OK;
5489 PgHdr *pPg;
5490 u8 noContent; /* True if PAGER_GET_NOCONTENT is set */
5491 sqlite3_pcache_page *pBase;
5492
5493 assert( pPager->errCode==SQLITE_OK );
5494 assert( pPager->eState>=PAGER_READER );
5495 assert( assert_pager_state(pPager) );
5496 assert( pPager->hasHeldSharedLock==1 );
5497
5498 if( pgno==0 ) return SQLITE_CORRUPT_BKPT;
5499 pBase = sqlite3PcacheFetch(pPager->pPCache, pgno, 3);
5500 if( pBase==0 ){
5501 pPg = 0;
5502 rc = sqlite3PcacheFetchStress(pPager->pPCache, pgno, &pBase);
5503 if( rc!=SQLITE_OK ) goto pager_acquire_err;
5504 if( pBase==0 ){
5505 rc = SQLITE_NOMEM_BKPT;
5506 goto pager_acquire_err;
5507 }
5508 }
5509 pPg = *ppPage = sqlite3PcacheFetchFinish(pPager->pPCache, pgno, pBase);
5510 assert( pPg==(*ppPage) );
5511 assert( pPg->pgno==pgno );
5512 assert( pPg->pPager==pPager || pPg->pPager==0 );
5513
5514 noContent = (flags & PAGER_GET_NOCONTENT)!=0;
5515 if( pPg->pPager && !noContent ){
5516 /* In this case the pcache already contains an initialized copy of
5517 ** the page. Return without further ado. */
5518 assert( pgno!=PAGER_SJ_PGNO(pPager) );
5519 pPager->aStat[PAGER_STAT_HIT]++;
5520 return SQLITE_OK;
5521
5522 }else{
5523 /* The pager cache has created a new page. Its content needs to
5524 ** be initialized. But first some error checks:
5525 **
5526 ** (*) obsolete. Was: maximum page number is 2^31
5527 ** (2) Never try to fetch the locking page
5528 */
5529 if( pgno==PAGER_SJ_PGNO(pPager) ){
5530 rc = SQLITE_CORRUPT_BKPT;
5531 goto pager_acquire_err;
5532 }
5533
5534 pPg->pPager = pPager;
5535
5536 assert( !isOpen(pPager->fd) || !MEMDB );
5537 if( !isOpen(pPager->fd) || pPager->dbSize<pgno || noContent ){
5538 if( pgno>pPager->mxPgno ){
5539 rc = SQLITE_FULL;
5540 goto pager_acquire_err;
5541 }
5542 if( noContent ){
5543 /* Failure to set the bits in the InJournal bit-vectors is benign.
5544 ** It merely means that we might do some extra work to journal a
5545 ** page that does not need to be journaled. Nevertheless, be sure
5546 ** to test the case where a malloc error occurs while trying to set
5547 ** a bit in a bit vector.
5548 */
5549 sqlite3BeginBenignMalloc();
5550 if( pgno<=pPager->dbOrigSize ){
5551 TESTONLY( rc = ) sqlite3BitvecSet(pPager->pInJournal, pgno);
5552 testcase( rc==SQLITE_NOMEM );
5553 }
5554 TESTONLY( rc = ) addToSavepointBitvecs(pPager, pgno);
5555 testcase( rc==SQLITE_NOMEM );
5556 sqlite3EndBenignMalloc();
5557 }
5558 memset(pPg->pData, 0, pPager->pageSize);
5559 IOTRACE(("ZERO %p %d\n", pPager, pgno));
5560 }else{
5561 assert( pPg->pPager==pPager );
5562 pPager->aStat[PAGER_STAT_MISS]++;
5563 rc = readDbPage(pPg);
5564 if( rc!=SQLITE_OK ){
5565 goto pager_acquire_err;
5566 }
5567 }
5568 pager_set_pagehash(pPg);
5569 }
5570 return SQLITE_OK;
5571
5572pager_acquire_err:
5573 assert( rc!=SQLITE_OK );
5574 if( pPg ){
5575 sqlite3PcacheDrop(pPg);
5576 }
5577 pagerUnlockIfUnused(pPager);
5578 *ppPage = 0;
5579 return rc;
5580}
5581
5582#if SQLITE_MAX_MMAP_SIZE>0
5583/* The page getter for when memory-mapped I/O is enabled */
5584static int getPageMMap(
5585 Pager *pPager, /* The pager open on the database file */
5586 Pgno pgno, /* Page number to fetch */
5587 DbPage **ppPage, /* Write a pointer to the page here */
5588 int flags /* PAGER_GET_XXX flags */
5589){
5590 int rc = SQLITE_OK;
5591 PgHdr *pPg = 0;
5592 u32 iFrame = 0; /* Frame to read from WAL file */
5593
5594 /* It is acceptable to use a read-only (mmap) page for any page except
5595 ** page 1 if there is no write-transaction open or the ACQUIRE_READONLY
5596 ** flag was specified by the caller. And so long as the db is not a
5597 ** temporary or in-memory database. */
5598 const int bMmapOk = (pgno>1
5599 && (pPager->eState==PAGER_READER || (flags & PAGER_GET_READONLY))
5600 );
5601
5602 assert( USEFETCH(pPager) );
5603
5604 /* Optimization note: Adding the "pgno<=1" term before "pgno==0" here
5605 ** allows the compiler optimizer to reuse the results of the "pgno>1"
5606 ** test in the previous statement, and avoid testing pgno==0 in the
5607 ** common case where pgno is large. */
5608 if( pgno<=1 && pgno==0 ){
5609 return SQLITE_CORRUPT_BKPT;
5610 }
5611 assert( pPager->eState>=PAGER_READER );
5612 assert( assert_pager_state(pPager) );
5613 assert( pPager->hasHeldSharedLock==1 );
5614 assert( pPager->errCode==SQLITE_OK );
5615
5616 if( bMmapOk && pagerUseWal(pPager) ){
5617 rc = sqlite3WalFindFrame(pPager->pWal, pgno, &iFrame);
5618 if( rc!=SQLITE_OK ){
5619 *ppPage = 0;
5620 return rc;
5621 }
5622 }
5623 if( bMmapOk && iFrame==0 ){
5624 void *pData = 0;
5625 rc = sqlite3OsFetch(pPager->fd,
5626 (i64)(pgno-1) * pPager->pageSize, pPager->pageSize, &pData
5627 );
5628 if( rc==SQLITE_OK && pData ){
5629 if( pPager->eState>PAGER_READER || pPager->tempFile ){
5630 pPg = sqlite3PagerLookup(pPager, pgno);
5631 }
5632 if( pPg==0 ){
5633 rc = pagerAcquireMapPage(pPager, pgno, pData, &pPg);
5634 }else{
5635 sqlite3OsUnfetch(pPager->fd, (i64)(pgno-1)*pPager->pageSize, pData);
5636 }
5637 if( pPg ){
5638 assert( rc==SQLITE_OK );
5639 *ppPage = pPg;
5640 return SQLITE_OK;
5641 }
5642 }
5643 if( rc!=SQLITE_OK ){
5644 *ppPage = 0;
5645 return rc;
5646 }
5647 }
5648 return getPageNormal(pPager, pgno, ppPage, flags);
5649}
5650#endif /* SQLITE_MAX_MMAP_SIZE>0 */
5651
5652/* The page getter method for when the pager is an error state */
5653static int getPageError(
5654 Pager *pPager, /* The pager open on the database file */
5655 Pgno pgno, /* Page number to fetch */
5656 DbPage **ppPage, /* Write a pointer to the page here */
5657 int flags /* PAGER_GET_XXX flags */
5658){
5659 UNUSED_PARAMETER(pgno);
5660 UNUSED_PARAMETER(flags);
5661 assert( pPager->errCode!=SQLITE_OK );
5662 *ppPage = 0;
5663 return pPager->errCode;
5664}
5665
5666
5667/* Dispatch all page fetch requests to the appropriate getter method.
5668*/
5669int sqlite3PagerGet(
5670 Pager *pPager, /* The pager open on the database file */
5671 Pgno pgno, /* Page number to fetch */
5672 DbPage **ppPage, /* Write a pointer to the page here */
5673 int flags /* PAGER_GET_XXX flags */
5674){
5675 /* printf("PAGE %u\n", pgno); fflush(stdout); */
5676 return pPager->xGet(pPager, pgno, ppPage, flags);
5677}
5678
5679/*
5680** Acquire a page if it is already in the in-memory cache. Do
5681** not read the page from disk. Return a pointer to the page,
5682** or 0 if the page is not in cache.
5683**
5684** See also sqlite3PagerGet(). The difference between this routine
5685** and sqlite3PagerGet() is that _get() will go to the disk and read
5686** in the page if the page is not already in cache. This routine
5687** returns NULL if the page is not in cache or if a disk I/O error
5688** has ever happened.
5689*/
5690DbPage *sqlite3PagerLookup(Pager *pPager, Pgno pgno){
5691 sqlite3_pcache_page *pPage;
5692 assert( pPager!=0 );
5693 assert( pgno!=0 );
5694 assert( pPager->pPCache!=0 );
5695 pPage = sqlite3PcacheFetch(pPager->pPCache, pgno, 0);
5696 assert( pPage==0 || pPager->hasHeldSharedLock );
5697 if( pPage==0 ) return 0;
5698 return sqlite3PcacheFetchFinish(pPager->pPCache, pgno, pPage);
5699}
5700
5701/*
5702** Release a page reference.
5703**
5704** The sqlite3PagerUnref() and sqlite3PagerUnrefNotNull() may only be
5705** used if we know that the page being released is not the last page.
5706** The btree layer always holds page1 open until the end, so these first
5707** to routines can be used to release any page other than BtShared.pPage1.
5708**
5709** Use sqlite3PagerUnrefPageOne() to release page1. This latter routine
5710** checks the total number of outstanding pages and if the number of
5711** pages reaches zero it drops the database lock.
5712*/
5713void sqlite3PagerUnrefNotNull(DbPage *pPg){
5714 TESTONLY( Pager *pPager = pPg->pPager; )
5715 assert( pPg!=0 );
5716 if( pPg->flags & PGHDR_MMAP ){
5717 assert( pPg->pgno!=1 ); /* Page1 is never memory mapped */
5718 pagerReleaseMapPage(pPg);
5719 }else{
5720 sqlite3PcacheRelease(pPg);
5721 }
5722 /* Do not use this routine to release the last reference to page1 */
5723 assert( sqlite3PcacheRefCount(pPager->pPCache)>0 );
5724}
5725void sqlite3PagerUnref(DbPage *pPg){
5726 if( pPg ) sqlite3PagerUnrefNotNull(pPg);
5727}
5728void sqlite3PagerUnrefPageOne(DbPage *pPg){
5729 Pager *pPager;
5730 assert( pPg!=0 );
5731 assert( pPg->pgno==1 );
5732 assert( (pPg->flags & PGHDR_MMAP)==0 ); /* Page1 is never memory mapped */
5733 pPager = pPg->pPager;
5734 sqlite3PcacheRelease(pPg);
5735 pagerUnlockIfUnused(pPager);
5736}
5737
5738/*
5739** This function is called at the start of every write transaction.
5740** There must already be a RESERVED or EXCLUSIVE lock on the database
5741** file when this routine is called.
5742**
5743** Open the journal file for pager pPager and write a journal header
5744** to the start of it. If there are active savepoints, open the sub-journal
5745** as well. This function is only used when the journal file is being
5746** opened to write a rollback log for a transaction. It is not used
5747** when opening a hot journal file to roll it back.
5748**
5749** If the journal file is already open (as it may be in exclusive mode),
5750** then this function just writes a journal header to the start of the
5751** already open file.
5752**
5753** Whether or not the journal file is opened by this function, the
5754** Pager.pInJournal bitvec structure is allocated.
5755**
5756** Return SQLITE_OK if everything is successful. Otherwise, return
5757** SQLITE_NOMEM if the attempt to allocate Pager.pInJournal fails, or
5758** an IO error code if opening or writing the journal file fails.
5759*/
5760static int pager_open_journal(Pager *pPager){
5761 int rc = SQLITE_OK; /* Return code */
5762 sqlite3_vfs * const pVfs = pPager->pVfs; /* Local cache of vfs pointer */
5763
5764 assert( pPager->eState==PAGER_WRITER_LOCKED );
5765 assert( assert_pager_state(pPager) );
5766 assert( pPager->pInJournal==0 );
5767
5768 /* If already in the error state, this function is a no-op. But on
5769 ** the other hand, this routine is never called if we are already in
5770 ** an error state. */
5771 if( NEVER(pPager->errCode) ) return pPager->errCode;
5772
5773 if( !pagerUseWal(pPager) && pPager->journalMode!=PAGER_JOURNALMODE_OFF ){
5774 pPager->pInJournal = sqlite3BitvecCreate(pPager->dbSize);
5775 if( pPager->pInJournal==0 ){
5776 return SQLITE_NOMEM_BKPT;
5777 }
5778
5779 /* Open the journal file if it is not already open. */
5780 if( !isOpen(pPager->jfd) ){
5781 if( pPager->journalMode==PAGER_JOURNALMODE_MEMORY ){
5782 sqlite3MemJournalOpen(pPager->jfd);
5783 }else{
5784 int flags = SQLITE_OPEN_READWRITE|SQLITE_OPEN_CREATE;
5785 int nSpill;
5786
5787 if( pPager->tempFile ){
5788 flags |= (SQLITE_OPEN_DELETEONCLOSE|SQLITE_OPEN_TEMP_JOURNAL);
5789 flags |= SQLITE_OPEN_EXCLUSIVE;
5790 nSpill = sqlite3Config.nStmtSpill;
5791 }else{
5792 flags |= SQLITE_OPEN_MAIN_JOURNAL;
5793 nSpill = jrnlBufferSize(pPager);
5794 }
5795
5796 /* Verify that the database still has the same name as it did when
5797 ** it was originally opened. */
5798 rc = databaseIsUnmoved(pPager);
5799 if( rc==SQLITE_OK ){
5800 rc = sqlite3JournalOpen (
5801 pVfs, pPager->zJournal, pPager->jfd, flags, nSpill
5802 );
5803 }
5804 }
5805 assert( rc!=SQLITE_OK || isOpen(pPager->jfd) );
5806 }
5807
5808
5809 /* Write the first journal header to the journal file and open
5810 ** the sub-journal if necessary.
5811 */
5812 if( rc==SQLITE_OK ){
5813 /* TODO: Check if all of these are really required. */
5814 pPager->nRec = 0;
5815 pPager->journalOff = 0;
5816 pPager->setSuper = 0;
5817 pPager->journalHdr = 0;
5818 rc = writeJournalHdr(pPager);
5819 }
5820 }
5821
5822 if( rc!=SQLITE_OK ){
5823 sqlite3BitvecDestroy(pPager->pInJournal);
5824 pPager->pInJournal = 0;
5825 pPager->journalOff = 0;
5826 }else{
5827 assert( pPager->eState==PAGER_WRITER_LOCKED );
5828 pPager->eState = PAGER_WRITER_CACHEMOD;
5829 }
5830
5831 return rc;
5832}
5833
5834/*
5835** Begin a write-transaction on the specified pager object. If a
5836** write-transaction has already been opened, this function is a no-op.
5837**
5838** If the exFlag argument is false, then acquire at least a RESERVED
5839** lock on the database file. If exFlag is true, then acquire at least
5840** an EXCLUSIVE lock. If such a lock is already held, no locking
5841** functions need be called.
5842**
5843** If the subjInMemory argument is non-zero, then any sub-journal opened
5844** within this transaction will be opened as an in-memory file. This
5845** has no effect if the sub-journal is already opened (as it may be when
5846** running in exclusive mode) or if the transaction does not require a
5847** sub-journal. If the subjInMemory argument is zero, then any required
5848** sub-journal is implemented in-memory if pPager is an in-memory database,
5849** or using a temporary file otherwise.
5850*/
5851int sqlite3PagerBegin(Pager *pPager, int exFlag, int subjInMemory){
5852 int rc = SQLITE_OK;
5853
5854 if( pPager->errCode ) return pPager->errCode;
5855 assert( pPager->eState>=PAGER_READER && pPager->eState<PAGER_ERROR );
5856 pPager->subjInMemory = (u8)subjInMemory;
5857
5858 if( pPager->eState==PAGER_READER ){
5859 assert( pPager->pInJournal==0 );
5860
5861 if( pagerUseWal(pPager) ){
5862 /* If the pager is configured to use locking_mode=exclusive, and an
5863 ** exclusive lock on the database is not already held, obtain it now.
5864 */
5865 if( pPager->exclusiveMode && sqlite3WalExclusiveMode(pPager->pWal, -1) ){
5866 rc = pagerLockDb(pPager, EXCLUSIVE_LOCK);
5867 if( rc!=SQLITE_OK ){
5868 return rc;
5869 }
5870 (void)sqlite3WalExclusiveMode(pPager->pWal, 1);
5871 }
5872
5873 /* Grab the write lock on the log file. If successful, upgrade to
5874 ** PAGER_RESERVED state. Otherwise, return an error code to the caller.
5875 ** The busy-handler is not invoked if another connection already
5876 ** holds the write-lock. If possible, the upper layer will call it.
5877 */
5878 rc = sqlite3WalBeginWriteTransaction(pPager->pWal);
5879 }else{
5880 /* Obtain a RESERVED lock on the database file. If the exFlag parameter
5881 ** is true, then immediately upgrade this to an EXCLUSIVE lock. The
5882 ** busy-handler callback can be used when upgrading to the EXCLUSIVE
5883 ** lock, but not when obtaining the RESERVED lock.
5884 */
5885 rc = pagerLockDb(pPager, RESERVED_LOCK);
5886 if( rc==SQLITE_OK && exFlag ){
5887 rc = pager_wait_on_lock(pPager, EXCLUSIVE_LOCK);
5888 }
5889 }
5890
5891 if( rc==SQLITE_OK ){
5892 /* Change to WRITER_LOCKED state.
5893 **
5894 ** WAL mode sets Pager.eState to PAGER_WRITER_LOCKED or CACHEMOD
5895 ** when it has an open transaction, but never to DBMOD or FINISHED.
5896 ** This is because in those states the code to roll back savepoint
5897 ** transactions may copy data from the sub-journal into the database
5898 ** file as well as into the page cache. Which would be incorrect in
5899 ** WAL mode.
5900 */
5901 pPager->eState = PAGER_WRITER_LOCKED;
5902 pPager->dbHintSize = pPager->dbSize;
5903 pPager->dbFileSize = pPager->dbSize;
5904 pPager->dbOrigSize = pPager->dbSize;
5905 pPager->journalOff = 0;
5906 }
5907
5908 assert( rc==SQLITE_OK || pPager->eState==PAGER_READER );
5909 assert( rc!=SQLITE_OK || pPager->eState==PAGER_WRITER_LOCKED );
5910 assert( assert_pager_state(pPager) );
5911 }
5912
5913 PAGERTRACE(("TRANSACTION %d\n", PAGERID(pPager)));
5914 return rc;
5915}
5916
5917/*
5918** Write page pPg onto the end of the rollback journal.
5919*/
5920static SQLITE_NOINLINE int pagerAddPageToRollbackJournal(PgHdr *pPg){
5921 Pager *pPager = pPg->pPager;
5922 int rc;
5923 u32 cksum;
5924 char *pData2;
5925 i64 iOff = pPager->journalOff;
5926
5927 /* We should never write to the journal file the page that
5928 ** contains the database locks. The following assert verifies
5929 ** that we do not. */
5930 assert( pPg->pgno!=PAGER_SJ_PGNO(pPager) );
5931
5932 assert( pPager->journalHdr<=pPager->journalOff );
5933 pData2 = pPg->pData;
5934 cksum = pager_cksum(pPager, (u8*)pData2);
5935
5936 /* Even if an IO or diskfull error occurs while journalling the
5937 ** page in the block above, set the need-sync flag for the page.
5938 ** Otherwise, when the transaction is rolled back, the logic in
5939 ** playback_one_page() will think that the page needs to be restored
5940 ** in the database file. And if an IO error occurs while doing so,
5941 ** then corruption may follow.
5942 */
5943 pPg->flags |= PGHDR_NEED_SYNC;
5944
5945 rc = write32bits(pPager->jfd, iOff, pPg->pgno);
5946 if( rc!=SQLITE_OK ) return rc;
5947 rc = sqlite3OsWrite(pPager->jfd, pData2, pPager->pageSize, iOff+4);
5948 if( rc!=SQLITE_OK ) return rc;
5949 rc = write32bits(pPager->jfd, iOff+pPager->pageSize+4, cksum);
5950 if( rc!=SQLITE_OK ) return rc;
5951
5952 IOTRACE(("JOUT %p %d %lld %d\n", pPager, pPg->pgno,
5953 pPager->journalOff, pPager->pageSize));
5954 PAGER_INCR(sqlite3_pager_writej_count);
5955 PAGERTRACE(("JOURNAL %d page %d needSync=%d hash(%08x)\n",
5956 PAGERID(pPager), pPg->pgno,
5957 ((pPg->flags&PGHDR_NEED_SYNC)?1:0), pager_pagehash(pPg)));
5958
5959 pPager->journalOff += 8 + pPager->pageSize;
5960 pPager->nRec++;
5961 assert( pPager->pInJournal!=0 );
5962 rc = sqlite3BitvecSet(pPager->pInJournal, pPg->pgno);
5963 testcase( rc==SQLITE_NOMEM );
5964 assert( rc==SQLITE_OK || rc==SQLITE_NOMEM );
5965 rc |= addToSavepointBitvecs(pPager, pPg->pgno);
5966 assert( rc==SQLITE_OK || rc==SQLITE_NOMEM );
5967 return rc;
5968}
5969
5970/*
5971** Mark a single data page as writeable. The page is written into the
5972** main journal or sub-journal as required. If the page is written into
5973** one of the journals, the corresponding bit is set in the
5974** Pager.pInJournal bitvec and the PagerSavepoint.pInSavepoint bitvecs
5975** of any open savepoints as appropriate.
5976*/
5977static int pager_write(PgHdr *pPg){
5978 Pager *pPager = pPg->pPager;
5979 int rc = SQLITE_OK;
5980
5981 /* This routine is not called unless a write-transaction has already
5982 ** been started. The journal file may or may not be open at this point.
5983 ** It is never called in the ERROR state.
5984 */
5985 assert( pPager->eState==PAGER_WRITER_LOCKED
5986 || pPager->eState==PAGER_WRITER_CACHEMOD
5987 || pPager->eState==PAGER_WRITER_DBMOD
5988 );
5989 assert( assert_pager_state(pPager) );
5990 assert( pPager->errCode==0 );
5991 assert( pPager->readOnly==0 );
5992 CHECK_PAGE(pPg);
5993
5994 /* The journal file needs to be opened. Higher level routines have already
5995 ** obtained the necessary locks to begin the write-transaction, but the
5996 ** rollback journal might not yet be open. Open it now if this is the case.
5997 **
5998 ** This is done before calling sqlite3PcacheMakeDirty() on the page.
5999 ** Otherwise, if it were done after calling sqlite3PcacheMakeDirty(), then
6000 ** an error might occur and the pager would end up in WRITER_LOCKED state
6001 ** with pages marked as dirty in the cache.
6002 */
6003 if( pPager->eState==PAGER_WRITER_LOCKED ){
6004 rc = pager_open_journal(pPager);
6005 if( rc!=SQLITE_OK ) return rc;
6006 }
6007 assert( pPager->eState>=PAGER_WRITER_CACHEMOD );
6008 assert( assert_pager_state(pPager) );
6009
6010 /* Mark the page that is about to be modified as dirty. */
6011 sqlite3PcacheMakeDirty(pPg);
6012
6013 /* If a rollback journal is in use, them make sure the page that is about
6014 ** to change is in the rollback journal, or if the page is a new page off
6015 ** then end of the file, make sure it is marked as PGHDR_NEED_SYNC.
6016 */
6017 assert( (pPager->pInJournal!=0) == isOpen(pPager->jfd) );
6018 if( pPager->pInJournal!=0
6019 && sqlite3BitvecTestNotNull(pPager->pInJournal, pPg->pgno)==0
6020 ){
6021 assert( pagerUseWal(pPager)==0 );
6022 if( pPg->pgno<=pPager->dbOrigSize ){
6023 rc = pagerAddPageToRollbackJournal(pPg);
6024 if( rc!=SQLITE_OK ){
6025 return rc;
6026 }
6027 }else{
6028 if( pPager->eState!=PAGER_WRITER_DBMOD ){
6029 pPg->flags |= PGHDR_NEED_SYNC;
6030 }
6031 PAGERTRACE(("APPEND %d page %d needSync=%d\n",
6032 PAGERID(pPager), pPg->pgno,
6033 ((pPg->flags&PGHDR_NEED_SYNC)?1:0)));
6034 }
6035 }
6036
6037 /* The PGHDR_DIRTY bit is set above when the page was added to the dirty-list
6038 ** and before writing the page into the rollback journal. Wait until now,
6039 ** after the page has been successfully journalled, before setting the
6040 ** PGHDR_WRITEABLE bit that indicates that the page can be safely modified.
6041 */
6042 pPg->flags |= PGHDR_WRITEABLE;
6043
6044 /* If the statement journal is open and the page is not in it,
6045 ** then write the page into the statement journal.
6046 */
6047 if( pPager->nSavepoint>0 ){
6048 rc = subjournalPageIfRequired(pPg);
6049 }
6050
6051 /* Update the database size and return. */
6052 if( pPager->dbSize<pPg->pgno ){
6053 pPager->dbSize = pPg->pgno;
6054 }
6055 return rc;
6056}
6057
6058/*
6059** This is a variant of sqlite3PagerWrite() that runs when the sector size
6060** is larger than the page size. SQLite makes the (reasonable) assumption that
6061** all bytes of a sector are written together by hardware. Hence, all bytes of
6062** a sector need to be journalled in case of a power loss in the middle of
6063** a write.
6064**
6065** Usually, the sector size is less than or equal to the page size, in which
6066** case pages can be individually written. This routine only runs in the
6067** exceptional case where the page size is smaller than the sector size.
6068*/
6069static SQLITE_NOINLINE int pagerWriteLargeSector(PgHdr *pPg){
6070 int rc = SQLITE_OK; /* Return code */
6071 Pgno nPageCount; /* Total number of pages in database file */
6072 Pgno pg1; /* First page of the sector pPg is located on. */
6073 int nPage = 0; /* Number of pages starting at pg1 to journal */
6074 int ii; /* Loop counter */
6075 int needSync = 0; /* True if any page has PGHDR_NEED_SYNC */
6076 Pager *pPager = pPg->pPager; /* The pager that owns pPg */
6077 Pgno nPagePerSector = (pPager->sectorSize/pPager->pageSize);
6078
6079 /* Set the doNotSpill NOSYNC bit to 1. This is because we cannot allow
6080 ** a journal header to be written between the pages journaled by
6081 ** this function.
6082 */
6083 assert( !MEMDB );
6084 assert( (pPager->doNotSpill & SPILLFLAG_NOSYNC)==0 );
6085 pPager->doNotSpill |= SPILLFLAG_NOSYNC;
6086
6087 /* This trick assumes that both the page-size and sector-size are
6088 ** an integer power of 2. It sets variable pg1 to the identifier
6089 ** of the first page of the sector pPg is located on.
6090 */
6091 pg1 = ((pPg->pgno-1) & ~(nPagePerSector-1)) + 1;
6092
6093 nPageCount = pPager->dbSize;
6094 if( pPg->pgno>nPageCount ){
6095 nPage = (pPg->pgno - pg1)+1;
6096 }else if( (pg1+nPagePerSector-1)>nPageCount ){
6097 nPage = nPageCount+1-pg1;
6098 }else{
6099 nPage = nPagePerSector;
6100 }
6101 assert(nPage>0);
6102 assert(pg1<=pPg->pgno);
6103 assert((pg1+nPage)>pPg->pgno);
6104
6105 for(ii=0; ii<nPage && rc==SQLITE_OK; ii++){
6106 Pgno pg = pg1+ii;
6107 PgHdr *pPage;
6108 if( pg==pPg->pgno || !sqlite3BitvecTest(pPager->pInJournal, pg) ){
6109 if( pg!=PAGER_SJ_PGNO(pPager) ){
6110 rc = sqlite3PagerGet(pPager, pg, &pPage, 0);
6111 if( rc==SQLITE_OK ){
6112 rc = pager_write(pPage);
6113 if( pPage->flags&PGHDR_NEED_SYNC ){
6114 needSync = 1;
6115 }
6116 sqlite3PagerUnrefNotNull(pPage);
6117 }
6118 }
6119 }else if( (pPage = sqlite3PagerLookup(pPager, pg))!=0 ){
6120 if( pPage->flags&PGHDR_NEED_SYNC ){
6121 needSync = 1;
6122 }
6123 sqlite3PagerUnrefNotNull(pPage);
6124 }
6125 }
6126
6127 /* If the PGHDR_NEED_SYNC flag is set for any of the nPage pages
6128 ** starting at pg1, then it needs to be set for all of them. Because
6129 ** writing to any of these nPage pages may damage the others, the
6130 ** journal file must contain sync()ed copies of all of them
6131 ** before any of them can be written out to the database file.
6132 */
6133 if( rc==SQLITE_OK && needSync ){
6134 assert( !MEMDB );
6135 for(ii=0; ii<nPage; ii++){
6136 PgHdr *pPage = sqlite3PagerLookup(pPager, pg1+ii);
6137 if( pPage ){
6138 pPage->flags |= PGHDR_NEED_SYNC;
6139 sqlite3PagerUnrefNotNull(pPage);
6140 }
6141 }
6142 }
6143
6144 assert( (pPager->doNotSpill & SPILLFLAG_NOSYNC)!=0 );
6145 pPager->doNotSpill &= ~SPILLFLAG_NOSYNC;
6146 return rc;
6147}
6148
6149/*
6150** Mark a data page as writeable. This routine must be called before
6151** making changes to a page. The caller must check the return value
6152** of this function and be careful not to change any page data unless
6153** this routine returns SQLITE_OK.
6154**
6155** The difference between this function and pager_write() is that this
6156** function also deals with the special case where 2 or more pages
6157** fit on a single disk sector. In this case all co-resident pages
6158** must have been written to the journal file before returning.
6159**
6160** If an error occurs, SQLITE_NOMEM or an IO error code is returned
6161** as appropriate. Otherwise, SQLITE_OK.
6162*/
6163int sqlite3PagerWrite(PgHdr *pPg){
6164 Pager *pPager = pPg->pPager;
6165 assert( (pPg->flags & PGHDR_MMAP)==0 );
6166 assert( pPager->eState>=PAGER_WRITER_LOCKED );
6167 assert( assert_pager_state(pPager) );
6168 if( (pPg->flags & PGHDR_WRITEABLE)!=0 && pPager->dbSize>=pPg->pgno ){
6169 if( pPager->nSavepoint ) return subjournalPageIfRequired(pPg);
6170 return SQLITE_OK;
6171 }else if( pPager->errCode ){
6172 return pPager->errCode;
6173 }else if( pPager->sectorSize > (u32)pPager->pageSize ){
6174 assert( pPager->tempFile==0 );
6175 return pagerWriteLargeSector(pPg);
6176 }else{
6177 return pager_write(pPg);
6178 }
6179}
6180
6181/*
6182** Return TRUE if the page given in the argument was previously passed
6183** to sqlite3PagerWrite(). In other words, return TRUE if it is ok
6184** to change the content of the page.
6185*/
6186#ifndef NDEBUG
6187int sqlite3PagerIswriteable(DbPage *pPg){
6188 return pPg->flags & PGHDR_WRITEABLE;
6189}
6190#endif
6191
6192/*
6193** A call to this routine tells the pager that it is not necessary to
6194** write the information on page pPg back to the disk, even though
6195** that page might be marked as dirty. This happens, for example, when
6196** the page has been added as a leaf of the freelist and so its
6197** content no longer matters.
6198**
6199** The overlying software layer calls this routine when all of the data
6200** on the given page is unused. The pager marks the page as clean so
6201** that it does not get written to disk.
6202**
6203** Tests show that this optimization can quadruple the speed of large
6204** DELETE operations.
6205**
6206** This optimization cannot be used with a temp-file, as the page may
6207** have been dirty at the start of the transaction. In that case, if
6208** memory pressure forces page pPg out of the cache, the data does need
6209** to be written out to disk so that it may be read back in if the
6210** current transaction is rolled back.
6211*/
6212void sqlite3PagerDontWrite(PgHdr *pPg){
6213 Pager *pPager = pPg->pPager;
6214 if( !pPager->tempFile && (pPg->flags&PGHDR_DIRTY) && pPager->nSavepoint==0 ){
6215 PAGERTRACE(("DONT_WRITE page %d of %d\n", pPg->pgno, PAGERID(pPager)));
6216 IOTRACE(("CLEAN %p %d\n", pPager, pPg->pgno))
6217 pPg->flags |= PGHDR_DONT_WRITE;
6218 pPg->flags &= ~PGHDR_WRITEABLE;
6219 testcase( pPg->flags & PGHDR_NEED_SYNC );
6220 pager_set_pagehash(pPg);
6221 }
6222}
6223
6224/*
6225** This routine is called to increment the value of the database file
6226** change-counter, stored as a 4-byte big-endian integer starting at
6227** byte offset 24 of the pager file. The secondary change counter at
6228** 92 is also updated, as is the SQLite version number at offset 96.
6229**
6230** But this only happens if the pPager->changeCountDone flag is false.
6231** To avoid excess churning of page 1, the update only happens once.
6232** See also the pager_write_changecounter() routine that does an
6233** unconditional update of the change counters.
6234**
6235** If the isDirectMode flag is zero, then this is done by calling
6236** sqlite3PagerWrite() on page 1, then modifying the contents of the
6237** page data. In this case the file will be updated when the current
6238** transaction is committed.
6239**
6240** The isDirectMode flag may only be non-zero if the library was compiled
6241** with the SQLITE_ENABLE_ATOMIC_WRITE macro defined. In this case,
6242** if isDirect is non-zero, then the database file is updated directly
6243** by writing an updated version of page 1 using a call to the
6244** sqlite3OsWrite() function.
6245*/
6246static int pager_incr_changecounter(Pager *pPager, int isDirectMode){
6247 int rc = SQLITE_OK;
6248
6249 assert( pPager->eState==PAGER_WRITER_CACHEMOD
6250 || pPager->eState==PAGER_WRITER_DBMOD
6251 );
6252 assert( assert_pager_state(pPager) );
6253
6254 /* Declare and initialize constant integer 'isDirect'. If the
6255 ** atomic-write optimization is enabled in this build, then isDirect
6256 ** is initialized to the value passed as the isDirectMode parameter
6257 ** to this function. Otherwise, it is always set to zero.
6258 **
6259 ** The idea is that if the atomic-write optimization is not
6260 ** enabled at compile time, the compiler can omit the tests of
6261 ** 'isDirect' below, as well as the block enclosed in the
6262 ** "if( isDirect )" condition.
6263 */
6264#ifndef SQLITE_ENABLE_ATOMIC_WRITE
6265# define DIRECT_MODE 0
6266 assert( isDirectMode==0 );
6267 UNUSED_PARAMETER(isDirectMode);
6268#else
6269# define DIRECT_MODE isDirectMode
6270#endif
6271
6272 if( !pPager->changeCountDone && ALWAYS(pPager->dbSize>0) ){
6273 PgHdr *pPgHdr; /* Reference to page 1 */
6274
6275 assert( !pPager->tempFile && isOpen(pPager->fd) );
6276
6277 /* Open page 1 of the file for writing. */
6278 rc = sqlite3PagerGet(pPager, 1, &pPgHdr, 0);
6279 assert( pPgHdr==0 || rc==SQLITE_OK );
6280
6281 /* If page one was fetched successfully, and this function is not
6282 ** operating in direct-mode, make page 1 writable. When not in
6283 ** direct mode, page 1 is always held in cache and hence the PagerGet()
6284 ** above is always successful - hence the ALWAYS on rc==SQLITE_OK.
6285 */
6286 if( !DIRECT_MODE && ALWAYS(rc==SQLITE_OK) ){
6287 rc = sqlite3PagerWrite(pPgHdr);
6288 }
6289
6290 if( rc==SQLITE_OK ){
6291 /* Actually do the update of the change counter */
6292 pager_write_changecounter(pPgHdr);
6293
6294 /* If running in direct mode, write the contents of page 1 to the file. */
6295 if( DIRECT_MODE ){
6296 const void *zBuf;
6297 assert( pPager->dbFileSize>0 );
6298 zBuf = pPgHdr->pData;
6299 if( rc==SQLITE_OK ){
6300 rc = sqlite3OsWrite(pPager->fd, zBuf, pPager->pageSize, 0);
6301 pPager->aStat[PAGER_STAT_WRITE]++;
6302 }
6303 if( rc==SQLITE_OK ){
6304 /* Update the pager's copy of the change-counter. Otherwise, the
6305 ** next time a read transaction is opened the cache will be
6306 ** flushed (as the change-counter values will not match). */
6307 const void *pCopy = (const void *)&((const char *)zBuf)[24];
6308 memcpy(&pPager->dbFileVers, pCopy, sizeof(pPager->dbFileVers));
6309 pPager->changeCountDone = 1;
6310 }
6311 }else{
6312 pPager->changeCountDone = 1;
6313 }
6314 }
6315
6316 /* Release the page reference. */
6317 sqlite3PagerUnref(pPgHdr);
6318 }
6319 return rc;
6320}
6321
6322/*
6323** Sync the database file to disk. This is a no-op for in-memory databases
6324** or pages with the Pager.noSync flag set.
6325**
6326** If successful, or if called on a pager for which it is a no-op, this
6327** function returns SQLITE_OK. Otherwise, an IO error code is returned.
6328*/
6329int sqlite3PagerSync(Pager *pPager, const char *zSuper){
6330 int rc = SQLITE_OK;
6331 void *pArg = (void*)zSuper;
6332 rc = sqlite3OsFileControl(pPager->fd, SQLITE_FCNTL_SYNC, pArg);
6333 if( rc==SQLITE_NOTFOUND ) rc = SQLITE_OK;
6334 if( rc==SQLITE_OK && !pPager->noSync ){
6335 assert( !MEMDB );
6336 rc = sqlite3OsSync(pPager->fd, pPager->syncFlags);
6337 }
6338 return rc;
6339}
6340
6341/*
6342** This function may only be called while a write-transaction is active in
6343** rollback. If the connection is in WAL mode, this call is a no-op.
6344** Otherwise, if the connection does not already have an EXCLUSIVE lock on
6345** the database file, an attempt is made to obtain one.
6346**
6347** If the EXCLUSIVE lock is already held or the attempt to obtain it is
6348** successful, or the connection is in WAL mode, SQLITE_OK is returned.
6349** Otherwise, either SQLITE_BUSY or an SQLITE_IOERR_XXX error code is
6350** returned.
6351*/
6352int sqlite3PagerExclusiveLock(Pager *pPager){
6353 int rc = pPager->errCode;
6354 assert( assert_pager_state(pPager) );
6355 if( rc==SQLITE_OK ){
6356 assert( pPager->eState==PAGER_WRITER_CACHEMOD
6357 || pPager->eState==PAGER_WRITER_DBMOD
6358 || pPager->eState==PAGER_WRITER_LOCKED
6359 );
6360 assert( assert_pager_state(pPager) );
6361 if( 0==pagerUseWal(pPager) ){
6362 rc = pager_wait_on_lock(pPager, EXCLUSIVE_LOCK);
6363 }
6364 }
6365 return rc;
6366}
6367
6368/*
6369** Sync the database file for the pager pPager. zSuper points to the name
6370** of a super-journal file that should be written into the individual
6371** journal file. zSuper may be NULL, which is interpreted as no
6372** super-journal (a single database transaction).
6373**
6374** This routine ensures that:
6375**
6376** * The database file change-counter is updated,
6377** * the journal is synced (unless the atomic-write optimization is used),
6378** * all dirty pages are written to the database file,
6379** * the database file is truncated (if required), and
6380** * the database file synced.
6381**
6382** The only thing that remains to commit the transaction is to finalize
6383** (delete, truncate or zero the first part of) the journal file (or
6384** delete the super-journal file if specified).
6385**
6386** Note that if zSuper==NULL, this does not overwrite a previous value
6387** passed to an sqlite3PagerCommitPhaseOne() call.
6388**
6389** If the final parameter - noSync - is true, then the database file itself
6390** is not synced. The caller must call sqlite3PagerSync() directly to
6391** sync the database file before calling CommitPhaseTwo() to delete the
6392** journal file in this case.
6393*/
6394int sqlite3PagerCommitPhaseOne(
6395 Pager *pPager, /* Pager object */
6396 const char *zSuper, /* If not NULL, the super-journal name */
6397 int noSync /* True to omit the xSync on the db file */
6398){
6399 int rc = SQLITE_OK; /* Return code */
6400
6401 assert( pPager->eState==PAGER_WRITER_LOCKED
6402 || pPager->eState==PAGER_WRITER_CACHEMOD
6403 || pPager->eState==PAGER_WRITER_DBMOD
6404 || pPager->eState==PAGER_ERROR
6405 );
6406 assert( assert_pager_state(pPager) );
6407
6408 /* If a prior error occurred, report that error again. */
6409 if( NEVER(pPager->errCode) ) return pPager->errCode;
6410
6411 /* Provide the ability to easily simulate an I/O error during testing */
6412 if( sqlite3FaultSim(400) ) return SQLITE_IOERR;
6413
6414 PAGERTRACE(("DATABASE SYNC: File=%s zSuper=%s nSize=%d\n",
6415 pPager->zFilename, zSuper, pPager->dbSize));
6416
6417 /* If no database changes have been made, return early. */
6418 if( pPager->eState<PAGER_WRITER_CACHEMOD ) return SQLITE_OK;
6419
6420 assert( MEMDB==0 || pPager->tempFile );
6421 assert( isOpen(pPager->fd) || pPager->tempFile );
6422 if( 0==pagerFlushOnCommit(pPager, 1) ){
6423 /* If this is an in-memory db, or no pages have been written to, or this
6424 ** function has already been called, it is mostly a no-op. However, any
6425 ** backup in progress needs to be restarted. */
6426 sqlite3BackupRestart(pPager->pBackup);
6427 }else{
6428 PgHdr *pList;
6429 if( pagerUseWal(pPager) ){
6430 PgHdr *pPageOne = 0;
6431 pList = sqlite3PcacheDirtyList(pPager->pPCache);
6432 if( pList==0 ){
6433 /* Must have at least one page for the WAL commit flag.
6434 ** Ticket [2d1a5c67dfc2363e44f29d9bbd57f] 2011-05-18 */
6435 rc = sqlite3PagerGet(pPager, 1, &pPageOne, 0);
6436 pList = pPageOne;
6437 pList->pDirty = 0;
6438 }
6439 assert( rc==SQLITE_OK );
6440 if( ALWAYS(pList) ){
6441 rc = pagerWalFrames(pPager, pList, pPager->dbSize, 1);
6442 }
6443 sqlite3PagerUnref(pPageOne);
6444 if( rc==SQLITE_OK ){
6445 sqlite3PcacheCleanAll(pPager->pPCache);
6446 }
6447 }else{
6448 /* The bBatch boolean is true if the batch-atomic-write commit method
6449 ** should be used. No rollback journal is created if batch-atomic-write
6450 ** is enabled.
6451 */
6452#ifdef SQLITE_ENABLE_BATCH_ATOMIC_WRITE
6453 sqlite3_file *fd = pPager->fd;
6454 int bBatch = zSuper==0 /* An SQLITE_IOCAP_BATCH_ATOMIC commit */
6455 && (sqlite3OsDeviceCharacteristics(fd) & SQLITE_IOCAP_BATCH_ATOMIC)
6456 && !pPager->noSync
6457 && sqlite3JournalIsInMemory(pPager->jfd);
6458#else
6459# define bBatch 0
6460#endif
6461
6462#ifdef SQLITE_ENABLE_ATOMIC_WRITE
6463 /* The following block updates the change-counter. Exactly how it
6464 ** does this depends on whether or not the atomic-update optimization
6465 ** was enabled at compile time, and if this transaction meets the
6466 ** runtime criteria to use the operation:
6467 **
6468 ** * The file-system supports the atomic-write property for
6469 ** blocks of size page-size, and
6470 ** * This commit is not part of a multi-file transaction, and
6471 ** * Exactly one page has been modified and store in the journal file.
6472 **
6473 ** If the optimization was not enabled at compile time, then the
6474 ** pager_incr_changecounter() function is called to update the change
6475 ** counter in 'indirect-mode'. If the optimization is compiled in but
6476 ** is not applicable to this transaction, call sqlite3JournalCreate()
6477 ** to make sure the journal file has actually been created, then call
6478 ** pager_incr_changecounter() to update the change-counter in indirect
6479 ** mode.
6480 **
6481 ** Otherwise, if the optimization is both enabled and applicable,
6482 ** then call pager_incr_changecounter() to update the change-counter
6483 ** in 'direct' mode. In this case the journal file will never be
6484 ** created for this transaction.
6485 */
6486 if( bBatch==0 ){
6487 PgHdr *pPg;
6488 assert( isOpen(pPager->jfd)
6489 || pPager->journalMode==PAGER_JOURNALMODE_OFF
6490 || pPager->journalMode==PAGER_JOURNALMODE_WAL
6491 );
6492 if( !zSuper && isOpen(pPager->jfd)
6493 && pPager->journalOff==jrnlBufferSize(pPager)
6494 && pPager->dbSize>=pPager->dbOrigSize
6495 && (!(pPg = sqlite3PcacheDirtyList(pPager->pPCache)) || 0==pPg->pDirty)
6496 ){
6497 /* Update the db file change counter via the direct-write method. The
6498 ** following call will modify the in-memory representation of page 1
6499 ** to include the updated change counter and then write page 1
6500 ** directly to the database file. Because of the atomic-write
6501 ** property of the host file-system, this is safe.
6502 */
6503 rc = pager_incr_changecounter(pPager, 1);
6504 }else{
6505 rc = sqlite3JournalCreate(pPager->jfd);
6506 if( rc==SQLITE_OK ){
6507 rc = pager_incr_changecounter(pPager, 0);
6508 }
6509 }
6510 }
6511#else /* SQLITE_ENABLE_ATOMIC_WRITE */
6512#ifdef SQLITE_ENABLE_BATCH_ATOMIC_WRITE
6513 if( zSuper ){
6514 rc = sqlite3JournalCreate(pPager->jfd);
6515 if( rc!=SQLITE_OK ) goto commit_phase_one_exit;
6516 assert( bBatch==0 );
6517 }
6518#endif
6519 rc = pager_incr_changecounter(pPager, 0);
6520#endif /* !SQLITE_ENABLE_ATOMIC_WRITE */
6521 if( rc!=SQLITE_OK ) goto commit_phase_one_exit;
6522
6523 /* Write the super-journal name into the journal file. If a
6524 ** super-journal file name has already been written to the journal file,
6525 ** or if zSuper is NULL (no super-journal), then this call is a no-op.
6526 */
6527 rc = writeSuperJournal(pPager, zSuper);
6528 if( rc!=SQLITE_OK ) goto commit_phase_one_exit;
6529
6530 /* Sync the journal file and write all dirty pages to the database.
6531 ** If the atomic-update optimization is being used, this sync will not
6532 ** create the journal file or perform any real IO.
6533 **
6534 ** Because the change-counter page was just modified, unless the
6535 ** atomic-update optimization is used it is almost certain that the
6536 ** journal requires a sync here. However, in locking_mode=exclusive
6537 ** on a system under memory pressure it is just possible that this is
6538 ** not the case. In this case it is likely enough that the redundant
6539 ** xSync() call will be changed to a no-op by the OS anyhow.
6540 */
6541 rc = syncJournal(pPager, 0);
6542 if( rc!=SQLITE_OK ) goto commit_phase_one_exit;
6543
6544 pList = sqlite3PcacheDirtyList(pPager->pPCache);
6545#ifdef SQLITE_ENABLE_BATCH_ATOMIC_WRITE
6546 if( bBatch ){
6547 rc = sqlite3OsFileControl(fd, SQLITE_FCNTL_BEGIN_ATOMIC_WRITE, 0);
6548 if( rc==SQLITE_OK ){
6549 rc = pager_write_pagelist(pPager, pList);
6550 if( rc==SQLITE_OK ){
6551 rc = sqlite3OsFileControl(fd, SQLITE_FCNTL_COMMIT_ATOMIC_WRITE, 0);
6552 }
6553 if( rc!=SQLITE_OK ){
6554 sqlite3OsFileControlHint(fd, SQLITE_FCNTL_ROLLBACK_ATOMIC_WRITE, 0);
6555 }
6556 }
6557
6558 if( (rc&0xFF)==SQLITE_IOERR && rc!=SQLITE_IOERR_NOMEM ){
6559 rc = sqlite3JournalCreate(pPager->jfd);
6560 if( rc!=SQLITE_OK ){
6561 sqlite3OsClose(pPager->jfd);
6562 goto commit_phase_one_exit;
6563 }
6564 bBatch = 0;
6565 }else{
6566 sqlite3OsClose(pPager->jfd);
6567 }
6568 }
6569#endif /* SQLITE_ENABLE_BATCH_ATOMIC_WRITE */
6570
6571 if( bBatch==0 ){
6572 rc = pager_write_pagelist(pPager, pList);
6573 }
6574 if( rc!=SQLITE_OK ){
6575 assert( rc!=SQLITE_IOERR_BLOCKED );
6576 goto commit_phase_one_exit;
6577 }
6578 sqlite3PcacheCleanAll(pPager->pPCache);
6579
6580 /* If the file on disk is smaller than the database image, use
6581 ** pager_truncate to grow the file here. This can happen if the database
6582 ** image was extended as part of the current transaction and then the
6583 ** last page in the db image moved to the free-list. In this case the
6584 ** last page is never written out to disk, leaving the database file
6585 ** undersized. Fix this now if it is the case. */
6586 if( pPager->dbSize>pPager->dbFileSize ){
6587 Pgno nNew = pPager->dbSize - (pPager->dbSize==PAGER_SJ_PGNO(pPager));
6588 assert( pPager->eState==PAGER_WRITER_DBMOD );
6589 rc = pager_truncate(pPager, nNew);
6590 if( rc!=SQLITE_OK ) goto commit_phase_one_exit;
6591 }
6592
6593 /* Finally, sync the database file. */
6594 if( !noSync ){
6595 rc = sqlite3PagerSync(pPager, zSuper);
6596 }
6597 IOTRACE(("DBSYNC %p\n", pPager))
6598 }
6599 }
6600
6601commit_phase_one_exit:
6602 if( rc==SQLITE_OK && !pagerUseWal(pPager) ){
6603 pPager->eState = PAGER_WRITER_FINISHED;
6604 }
6605 return rc;
6606}
6607
6608
6609/*
6610** When this function is called, the database file has been completely
6611** updated to reflect the changes made by the current transaction and
6612** synced to disk. The journal file still exists in the file-system
6613** though, and if a failure occurs at this point it will eventually
6614** be used as a hot-journal and the current transaction rolled back.
6615**
6616** This function finalizes the journal file, either by deleting,
6617** truncating or partially zeroing it, so that it cannot be used
6618** for hot-journal rollback. Once this is done the transaction is
6619** irrevocably committed.
6620**
6621** If an error occurs, an IO error code is returned and the pager
6622** moves into the error state. Otherwise, SQLITE_OK is returned.
6623*/
6624int sqlite3PagerCommitPhaseTwo(Pager *pPager){
6625 int rc = SQLITE_OK; /* Return code */
6626
6627 /* This routine should not be called if a prior error has occurred.
6628 ** But if (due to a coding error elsewhere in the system) it does get
6629 ** called, just return the same error code without doing anything. */
6630 if( NEVER(pPager->errCode) ) return pPager->errCode;
6631 pPager->iDataVersion++;
6632
6633 assert( pPager->eState==PAGER_WRITER_LOCKED
6634 || pPager->eState==PAGER_WRITER_FINISHED
6635 || (pagerUseWal(pPager) && pPager->eState==PAGER_WRITER_CACHEMOD)
6636 );
6637 assert( assert_pager_state(pPager) );
6638
6639 /* An optimization. If the database was not actually modified during
6640 ** this transaction, the pager is running in exclusive-mode and is
6641 ** using persistent journals, then this function is a no-op.
6642 **
6643 ** The start of the journal file currently contains a single journal
6644 ** header with the nRec field set to 0. If such a journal is used as
6645 ** a hot-journal during hot-journal rollback, 0 changes will be made
6646 ** to the database file. So there is no need to zero the journal
6647 ** header. Since the pager is in exclusive mode, there is no need
6648 ** to drop any locks either.
6649 */
6650 if( pPager->eState==PAGER_WRITER_LOCKED
6651 && pPager->exclusiveMode
6652 && pPager->journalMode==PAGER_JOURNALMODE_PERSIST
6653 ){
6654 assert( pPager->journalOff==JOURNAL_HDR_SZ(pPager) || !pPager->journalOff );
6655 pPager->eState = PAGER_READER;
6656 return SQLITE_OK;
6657 }
6658
6659 PAGERTRACE(("COMMIT %d\n", PAGERID(pPager)));
6660 rc = pager_end_transaction(pPager, pPager->setSuper, 1);
6661 return pager_error(pPager, rc);
6662}
6663
6664/*
6665** If a write transaction is open, then all changes made within the
6666** transaction are reverted and the current write-transaction is closed.
6667** The pager falls back to PAGER_READER state if successful, or PAGER_ERROR
6668** state if an error occurs.
6669**
6670** If the pager is already in PAGER_ERROR state when this function is called,
6671** it returns Pager.errCode immediately. No work is performed in this case.
6672**
6673** Otherwise, in rollback mode, this function performs two functions:
6674**
6675** 1) It rolls back the journal file, restoring all database file and
6676** in-memory cache pages to the state they were in when the transaction
6677** was opened, and
6678**
6679** 2) It finalizes the journal file, so that it is not used for hot
6680** rollback at any point in the future.
6681**
6682** Finalization of the journal file (task 2) is only performed if the
6683** rollback is successful.
6684**
6685** In WAL mode, all cache-entries containing data modified within the
6686** current transaction are either expelled from the cache or reverted to
6687** their pre-transaction state by re-reading data from the database or
6688** WAL files. The WAL transaction is then closed.
6689*/
6690int sqlite3PagerRollback(Pager *pPager){
6691 int rc = SQLITE_OK; /* Return code */
6692 PAGERTRACE(("ROLLBACK %d\n", PAGERID(pPager)));
6693
6694 /* PagerRollback() is a no-op if called in READER or OPEN state. If
6695 ** the pager is already in the ERROR state, the rollback is not
6696 ** attempted here. Instead, the error code is returned to the caller.
6697 */
6698 assert( assert_pager_state(pPager) );
6699 if( pPager->eState==PAGER_ERROR ) return pPager->errCode;
6700 if( pPager->eState<=PAGER_READER ) return SQLITE_OK;
6701
6702 if( pagerUseWal(pPager) ){
6703 int rc2;
6704 rc = sqlite3PagerSavepoint(pPager, SAVEPOINT_ROLLBACK, -1);
6705 rc2 = pager_end_transaction(pPager, pPager->setSuper, 0);
6706 if( rc==SQLITE_OK ) rc = rc2;
6707 }else if( !isOpen(pPager->jfd) || pPager->eState==PAGER_WRITER_LOCKED ){
6708 int eState = pPager->eState;
6709 rc = pager_end_transaction(pPager, 0, 0);
6710 if( !MEMDB && eState>PAGER_WRITER_LOCKED ){
6711 /* This can happen using journal_mode=off. Move the pager to the error
6712 ** state to indicate that the contents of the cache may not be trusted.
6713 ** Any active readers will get SQLITE_ABORT.
6714 */
6715 pPager->errCode = SQLITE_ABORT;
6716 pPager->eState = PAGER_ERROR;
6717 setGetterMethod(pPager);
6718 return rc;
6719 }
6720 }else{
6721 rc = pager_playback(pPager, 0);
6722 }
6723
6724 assert( pPager->eState==PAGER_READER || rc!=SQLITE_OK );
6725 assert( rc==SQLITE_OK || rc==SQLITE_FULL || rc==SQLITE_CORRUPT
6726 || rc==SQLITE_NOMEM || (rc&0xFF)==SQLITE_IOERR
6727 || rc==SQLITE_CANTOPEN
6728 );
6729
6730 /* If an error occurs during a ROLLBACK, we can no longer trust the pager
6731 ** cache. So call pager_error() on the way out to make any error persistent.
6732 */
6733 return pager_error(pPager, rc);
6734}
6735
6736/*
6737** Return TRUE if the database file is opened read-only. Return FALSE
6738** if the database is (in theory) writable.
6739*/
6740u8 sqlite3PagerIsreadonly(Pager *pPager){
6741 return pPager->readOnly;
6742}
6743
6744#ifdef SQLITE_DEBUG
6745/*
6746** Return the sum of the reference counts for all pages held by pPager.
6747*/
6748int sqlite3PagerRefcount(Pager *pPager){
6749 return sqlite3PcacheRefCount(pPager->pPCache);
6750}
6751#endif
6752
6753/*
6754** Return the approximate number of bytes of memory currently
6755** used by the pager and its associated cache.
6756*/
6757int sqlite3PagerMemUsed(Pager *pPager){
6758 int perPageSize = pPager->pageSize + pPager->nExtra
6759 + (int)(sizeof(PgHdr) + 5*sizeof(void*));
6760 return perPageSize*sqlite3PcachePagecount(pPager->pPCache)
6761 + sqlite3MallocSize(pPager)
6762 + pPager->pageSize;
6763}
6764
6765/*
6766** Return the number of references to the specified page.
6767*/
6768int sqlite3PagerPageRefcount(DbPage *pPage){
6769 return sqlite3PcachePageRefcount(pPage);
6770}
6771
6772#ifdef SQLITE_TEST
6773/*
6774** This routine is used for testing and analysis only.
6775*/
6776int *sqlite3PagerStats(Pager *pPager){
6777 static int a[11];
6778 a[0] = sqlite3PcacheRefCount(pPager->pPCache);
6779 a[1] = sqlite3PcachePagecount(pPager->pPCache);
6780 a[2] = sqlite3PcacheGetCachesize(pPager->pPCache);
6781 a[3] = pPager->eState==PAGER_OPEN ? -1 : (int) pPager->dbSize;
6782 a[4] = pPager->eState;
6783 a[5] = pPager->errCode;
6784 a[6] = pPager->aStat[PAGER_STAT_HIT];
6785 a[7] = pPager->aStat[PAGER_STAT_MISS];
6786 a[8] = 0; /* Used to be pPager->nOvfl */
6787 a[9] = pPager->nRead;
6788 a[10] = pPager->aStat[PAGER_STAT_WRITE];
6789 return a;
6790}
6791#endif
6792
6793/*
6794** Parameter eStat must be one of SQLITE_DBSTATUS_CACHE_HIT, _MISS, _WRITE,
6795** or _WRITE+1. The SQLITE_DBSTATUS_CACHE_WRITE+1 case is a translation
6796** of SQLITE_DBSTATUS_CACHE_SPILL. The _SPILL case is not contiguous because
6797** it was added later.
6798**
6799** Before returning, *pnVal is incremented by the
6800** current cache hit or miss count, according to the value of eStat. If the
6801** reset parameter is non-zero, the cache hit or miss count is zeroed before
6802** returning.
6803*/
6804void sqlite3PagerCacheStat(Pager *pPager, int eStat, int reset, int *pnVal){
6805
6806 assert( eStat==SQLITE_DBSTATUS_CACHE_HIT
6807 || eStat==SQLITE_DBSTATUS_CACHE_MISS
6808 || eStat==SQLITE_DBSTATUS_CACHE_WRITE
6809 || eStat==SQLITE_DBSTATUS_CACHE_WRITE+1
6810 );
6811
6812 assert( SQLITE_DBSTATUS_CACHE_HIT+1==SQLITE_DBSTATUS_CACHE_MISS );
6813 assert( SQLITE_DBSTATUS_CACHE_HIT+2==SQLITE_DBSTATUS_CACHE_WRITE );
6814 assert( PAGER_STAT_HIT==0 && PAGER_STAT_MISS==1
6815 && PAGER_STAT_WRITE==2 && PAGER_STAT_SPILL==3 );
6816
6817 eStat -= SQLITE_DBSTATUS_CACHE_HIT;
6818 *pnVal += pPager->aStat[eStat];
6819 if( reset ){
6820 pPager->aStat[eStat] = 0;
6821 }
6822}
6823
6824/*
6825** Return true if this is an in-memory or temp-file backed pager.
6826*/
6827int sqlite3PagerIsMemdb(Pager *pPager){
6828 return pPager->tempFile || pPager->memVfs;
6829}
6830
6831/*
6832** Check that there are at least nSavepoint savepoints open. If there are
6833** currently less than nSavepoints open, then open one or more savepoints
6834** to make up the difference. If the number of savepoints is already
6835** equal to nSavepoint, then this function is a no-op.
6836**
6837** If a memory allocation fails, SQLITE_NOMEM is returned. If an error
6838** occurs while opening the sub-journal file, then an IO error code is
6839** returned. Otherwise, SQLITE_OK.
6840*/
6841static SQLITE_NOINLINE int pagerOpenSavepoint(Pager *pPager, int nSavepoint){
6842 int rc = SQLITE_OK; /* Return code */
6843 int nCurrent = pPager->nSavepoint; /* Current number of savepoints */
6844 int ii; /* Iterator variable */
6845 PagerSavepoint *aNew; /* New Pager.aSavepoint array */
6846
6847 assert( pPager->eState>=PAGER_WRITER_LOCKED );
6848 assert( assert_pager_state(pPager) );
6849 assert( nSavepoint>nCurrent && pPager->useJournal );
6850
6851 /* Grow the Pager.aSavepoint array using realloc(). Return SQLITE_NOMEM
6852 ** if the allocation fails. Otherwise, zero the new portion in case a
6853 ** malloc failure occurs while populating it in the for(...) loop below.
6854 */
6855 aNew = (PagerSavepoint *)sqlite3Realloc(
6856 pPager->aSavepoint, sizeof(PagerSavepoint)*nSavepoint
6857 );
6858 if( !aNew ){
6859 return SQLITE_NOMEM_BKPT;
6860 }
6861 memset(&aNew[nCurrent], 0, (nSavepoint-nCurrent) * sizeof(PagerSavepoint));
6862 pPager->aSavepoint = aNew;
6863
6864 /* Populate the PagerSavepoint structures just allocated. */
6865 for(ii=nCurrent; ii<nSavepoint; ii++){
6866 aNew[ii].nOrig = pPager->dbSize;
6867 if( isOpen(pPager->jfd) && pPager->journalOff>0 ){
6868 aNew[ii].iOffset = pPager->journalOff;
6869 }else{
6870 aNew[ii].iOffset = JOURNAL_HDR_SZ(pPager);
6871 }
6872 aNew[ii].iSubRec = pPager->nSubRec;
6873 aNew[ii].pInSavepoint = sqlite3BitvecCreate(pPager->dbSize);
6874 aNew[ii].bTruncateOnRelease = 1;
6875 if( !aNew[ii].pInSavepoint ){
6876 return SQLITE_NOMEM_BKPT;
6877 }
6878 if( pagerUseWal(pPager) ){
6879 sqlite3WalSavepoint(pPager->pWal, aNew[ii].aWalData);
6880 }
6881 pPager->nSavepoint = ii+1;
6882 }
6883 assert( pPager->nSavepoint==nSavepoint );
6884 assertTruncateConstraint(pPager);
6885 return rc;
6886}
6887int sqlite3PagerOpenSavepoint(Pager *pPager, int nSavepoint){
6888 assert( pPager->eState>=PAGER_WRITER_LOCKED );
6889 assert( assert_pager_state(pPager) );
6890
6891 if( nSavepoint>pPager->nSavepoint && pPager->useJournal ){
6892 return pagerOpenSavepoint(pPager, nSavepoint);
6893 }else{
6894 return SQLITE_OK;
6895 }
6896}
6897
6898
6899/*
6900** This function is called to rollback or release (commit) a savepoint.
6901** The savepoint to release or rollback need not be the most recently
6902** created savepoint.
6903**
6904** Parameter op is always either SAVEPOINT_ROLLBACK or SAVEPOINT_RELEASE.
6905** If it is SAVEPOINT_RELEASE, then release and destroy the savepoint with
6906** index iSavepoint. If it is SAVEPOINT_ROLLBACK, then rollback all changes
6907** that have occurred since the specified savepoint was created.
6908**
6909** The savepoint to rollback or release is identified by parameter
6910** iSavepoint. A value of 0 means to operate on the outermost savepoint
6911** (the first created). A value of (Pager.nSavepoint-1) means operate
6912** on the most recently created savepoint. If iSavepoint is greater than
6913** (Pager.nSavepoint-1), then this function is a no-op.
6914**
6915** If a negative value is passed to this function, then the current
6916** transaction is rolled back. This is different to calling
6917** sqlite3PagerRollback() because this function does not terminate
6918** the transaction or unlock the database, it just restores the
6919** contents of the database to its original state.
6920**
6921** In any case, all savepoints with an index greater than iSavepoint
6922** are destroyed. If this is a release operation (op==SAVEPOINT_RELEASE),
6923** then savepoint iSavepoint is also destroyed.
6924**
6925** This function may return SQLITE_NOMEM if a memory allocation fails,
6926** or an IO error code if an IO error occurs while rolling back a
6927** savepoint. If no errors occur, SQLITE_OK is returned.
6928*/
6929int sqlite3PagerSavepoint(Pager *pPager, int op, int iSavepoint){
6930 int rc = pPager->errCode;
6931
6932#ifdef SQLITE_ENABLE_ZIPVFS
6933 if( op==SAVEPOINT_RELEASE ) rc = SQLITE_OK;
6934#endif
6935
6936 assert( op==SAVEPOINT_RELEASE || op==SAVEPOINT_ROLLBACK );
6937 assert( iSavepoint>=0 || op==SAVEPOINT_ROLLBACK );
6938
6939 if( rc==SQLITE_OK && iSavepoint<pPager->nSavepoint ){
6940 int ii; /* Iterator variable */
6941 int nNew; /* Number of remaining savepoints after this op. */
6942
6943 /* Figure out how many savepoints will still be active after this
6944 ** operation. Store this value in nNew. Then free resources associated
6945 ** with any savepoints that are destroyed by this operation.
6946 */
6947 nNew = iSavepoint + (( op==SAVEPOINT_RELEASE ) ? 0 : 1);
6948 for(ii=nNew; ii<pPager->nSavepoint; ii++){
6949 sqlite3BitvecDestroy(pPager->aSavepoint[ii].pInSavepoint);
6950 }
6951 pPager->nSavepoint = nNew;
6952
6953 /* Truncate the sub-journal so that it only includes the parts
6954 ** that are still in use. */
6955 if( op==SAVEPOINT_RELEASE ){
6956 PagerSavepoint *pRel = &pPager->aSavepoint[nNew];
6957 if( pRel->bTruncateOnRelease && isOpen(pPager->sjfd) ){
6958 /* Only truncate if it is an in-memory sub-journal. */
6959 if( sqlite3JournalIsInMemory(pPager->sjfd) ){
6960 i64 sz = (pPager->pageSize+4)*(i64)pRel->iSubRec;
6961 rc = sqlite3OsTruncate(pPager->sjfd, sz);
6962 assert( rc==SQLITE_OK );
6963 }
6964 pPager->nSubRec = pRel->iSubRec;
6965 }
6966 }
6967 /* Else this is a rollback operation, playback the specified savepoint.
6968 ** If this is a temp-file, it is possible that the journal file has
6969 ** not yet been opened. In this case there have been no changes to
6970 ** the database file, so the playback operation can be skipped.
6971 */
6972 else if( pagerUseWal(pPager) || isOpen(pPager->jfd) ){
6973 PagerSavepoint *pSavepoint = (nNew==0)?0:&pPager->aSavepoint[nNew-1];
6974 rc = pagerPlaybackSavepoint(pPager, pSavepoint);
6975 assert(rc!=SQLITE_DONE);
6976 }
6977
6978#ifdef SQLITE_ENABLE_ZIPVFS
6979 /* If the cache has been modified but the savepoint cannot be rolled
6980 ** back journal_mode=off, put the pager in the error state. This way,
6981 ** if the VFS used by this pager includes ZipVFS, the entire transaction
6982 ** can be rolled back at the ZipVFS level. */
6983 else if(
6984 pPager->journalMode==PAGER_JOURNALMODE_OFF
6985 && pPager->eState>=PAGER_WRITER_CACHEMOD
6986 ){
6987 pPager->errCode = SQLITE_ABORT;
6988 pPager->eState = PAGER_ERROR;
6989 setGetterMethod(pPager);
6990 }
6991#endif
6992 }
6993
6994 return rc;
6995}
6996
6997/*
6998** Return the full pathname of the database file.
6999**
7000** Except, if the pager is in-memory only, then return an empty string if
7001** nullIfMemDb is true. This routine is called with nullIfMemDb==1 when
7002** used to report the filename to the user, for compatibility with legacy
7003** behavior. But when the Btree needs to know the filename for matching to
7004** shared cache, it uses nullIfMemDb==0 so that in-memory databases can
7005** participate in shared-cache.
7006**
7007** The return value to this routine is always safe to use with
7008** sqlite3_uri_parameter() and sqlite3_filename_database() and friends.
7009*/
7010const char *sqlite3PagerFilename(const Pager *pPager, int nullIfMemDb){
7011 static const char zFake[8] = { 0, 0, 0, 0, 0, 0, 0, 0 };
7012 return (nullIfMemDb && pPager->memDb) ? &zFake[4] : pPager->zFilename;
7013}
7014
7015/*
7016** Return the VFS structure for the pager.
7017*/
7018sqlite3_vfs *sqlite3PagerVfs(Pager *pPager){
7019 return pPager->pVfs;
7020}
7021
7022/*
7023** Return the file handle for the database file associated
7024** with the pager. This might return NULL if the file has
7025** not yet been opened.
7026*/
7027sqlite3_file *sqlite3PagerFile(Pager *pPager){
7028 return pPager->fd;
7029}
7030
7031/*
7032** Return the file handle for the journal file (if it exists).
7033** This will be either the rollback journal or the WAL file.
7034*/
7035sqlite3_file *sqlite3PagerJrnlFile(Pager *pPager){
7036#if SQLITE_OMIT_WAL
7037 return pPager->jfd;
7038#else
7039 return pPager->pWal ? sqlite3WalFile(pPager->pWal) : pPager->jfd;
7040#endif
7041}
7042
7043/*
7044** Return the full pathname of the journal file.
7045*/
7046const char *sqlite3PagerJournalname(Pager *pPager){
7047 return pPager->zJournal;
7048}
7049
7050#ifndef SQLITE_OMIT_AUTOVACUUM
7051/*
7052** Move the page pPg to location pgno in the file.
7053**
7054** There must be no references to the page previously located at
7055** pgno (which we call pPgOld) though that page is allowed to be
7056** in cache. If the page previously located at pgno is not already
7057** in the rollback journal, it is not put there by by this routine.
7058**
7059** References to the page pPg remain valid. Updating any
7060** meta-data associated with pPg (i.e. data stored in the nExtra bytes
7061** allocated along with the page) is the responsibility of the caller.
7062**
7063** A transaction must be active when this routine is called. It used to be
7064** required that a statement transaction was not active, but this restriction
7065** has been removed (CREATE INDEX needs to move a page when a statement
7066** transaction is active).
7067**
7068** If the fourth argument, isCommit, is non-zero, then this page is being
7069** moved as part of a database reorganization just before the transaction
7070** is being committed. In this case, it is guaranteed that the database page
7071** pPg refers to will not be written to again within this transaction.
7072**
7073** This function may return SQLITE_NOMEM or an IO error code if an error
7074** occurs. Otherwise, it returns SQLITE_OK.
7075*/
7076int sqlite3PagerMovepage(Pager *pPager, DbPage *pPg, Pgno pgno, int isCommit){
7077 PgHdr *pPgOld; /* The page being overwritten. */
7078 Pgno needSyncPgno = 0; /* Old value of pPg->pgno, if sync is required */
7079 int rc; /* Return code */
7080 Pgno origPgno; /* The original page number */
7081
7082 assert( pPg->nRef>0 );
7083 assert( pPager->eState==PAGER_WRITER_CACHEMOD
7084 || pPager->eState==PAGER_WRITER_DBMOD
7085 );
7086 assert( assert_pager_state(pPager) );
7087
7088 /* In order to be able to rollback, an in-memory database must journal
7089 ** the page we are moving from.
7090 */
7091 assert( pPager->tempFile || !MEMDB );
7092 if( pPager->tempFile ){
7093 rc = sqlite3PagerWrite(pPg);
7094 if( rc ) return rc;
7095 }
7096
7097 /* If the page being moved is dirty and has not been saved by the latest
7098 ** savepoint, then save the current contents of the page into the
7099 ** sub-journal now. This is required to handle the following scenario:
7100 **
7101 ** BEGIN;
7102 ** <journal page X, then modify it in memory>
7103 ** SAVEPOINT one;
7104 ** <Move page X to location Y>
7105 ** ROLLBACK TO one;
7106 **
7107 ** If page X were not written to the sub-journal here, it would not
7108 ** be possible to restore its contents when the "ROLLBACK TO one"
7109 ** statement were is processed.
7110 **
7111 ** subjournalPage() may need to allocate space to store pPg->pgno into
7112 ** one or more savepoint bitvecs. This is the reason this function
7113 ** may return SQLITE_NOMEM.
7114 */
7115 if( (pPg->flags & PGHDR_DIRTY)!=0
7116 && SQLITE_OK!=(rc = subjournalPageIfRequired(pPg))
7117 ){
7118 return rc;
7119 }
7120
7121 PAGERTRACE(("MOVE %d page %d (needSync=%d) moves to %d\n",
7122 PAGERID(pPager), pPg->pgno, (pPg->flags&PGHDR_NEED_SYNC)?1:0, pgno));
7123 IOTRACE(("MOVE %p %d %d\n", pPager, pPg->pgno, pgno))
7124
7125 /* If the journal needs to be sync()ed before page pPg->pgno can
7126 ** be written to, store pPg->pgno in local variable needSyncPgno.
7127 **
7128 ** If the isCommit flag is set, there is no need to remember that
7129 ** the journal needs to be sync()ed before database page pPg->pgno
7130 ** can be written to. The caller has already promised not to write to it.
7131 */
7132 if( (pPg->flags&PGHDR_NEED_SYNC) && !isCommit ){
7133 needSyncPgno = pPg->pgno;
7134 assert( pPager->journalMode==PAGER_JOURNALMODE_OFF ||
7135 pageInJournal(pPager, pPg) || pPg->pgno>pPager->dbOrigSize );
7136 assert( pPg->flags&PGHDR_DIRTY );
7137 }
7138
7139 /* If the cache contains a page with page-number pgno, remove it
7140 ** from its hash chain. Also, if the PGHDR_NEED_SYNC flag was set for
7141 ** page pgno before the 'move' operation, it needs to be retained
7142 ** for the page moved there.
7143 */
7144 pPg->flags &= ~PGHDR_NEED_SYNC;
7145 pPgOld = sqlite3PagerLookup(pPager, pgno);
7146 assert( !pPgOld || pPgOld->nRef==1 || CORRUPT_DB );
7147 if( pPgOld ){
7148 if( NEVER(pPgOld->nRef>1) ){
7149 sqlite3PagerUnrefNotNull(pPgOld);
7150 return SQLITE_CORRUPT_BKPT;
7151 }
7152 pPg->flags |= (pPgOld->flags&PGHDR_NEED_SYNC);
7153 if( pPager->tempFile ){
7154 /* Do not discard pages from an in-memory database since we might
7155 ** need to rollback later. Just move the page out of the way. */
7156 sqlite3PcacheMove(pPgOld, pPager->dbSize+1);
7157 }else{
7158 sqlite3PcacheDrop(pPgOld);
7159 }
7160 }
7161
7162 origPgno = pPg->pgno;
7163 sqlite3PcacheMove(pPg, pgno);
7164 sqlite3PcacheMakeDirty(pPg);
7165
7166 /* For an in-memory database, make sure the original page continues
7167 ** to exist, in case the transaction needs to roll back. Use pPgOld
7168 ** as the original page since it has already been allocated.
7169 */
7170 if( pPager->tempFile && pPgOld ){
7171 sqlite3PcacheMove(pPgOld, origPgno);
7172 sqlite3PagerUnrefNotNull(pPgOld);
7173 }
7174
7175 if( needSyncPgno ){
7176 /* If needSyncPgno is non-zero, then the journal file needs to be
7177 ** sync()ed before any data is written to database file page needSyncPgno.
7178 ** Currently, no such page exists in the page-cache and the
7179 ** "is journaled" bitvec flag has been set. This needs to be remedied by
7180 ** loading the page into the pager-cache and setting the PGHDR_NEED_SYNC
7181 ** flag.
7182 **
7183 ** If the attempt to load the page into the page-cache fails, (due
7184 ** to a malloc() or IO failure), clear the bit in the pInJournal[]
7185 ** array. Otherwise, if the page is loaded and written again in
7186 ** this transaction, it may be written to the database file before
7187 ** it is synced into the journal file. This way, it may end up in
7188 ** the journal file twice, but that is not a problem.
7189 */
7190 PgHdr *pPgHdr;
7191 rc = sqlite3PagerGet(pPager, needSyncPgno, &pPgHdr, 0);
7192 if( rc!=SQLITE_OK ){
7193 if( needSyncPgno<=pPager->dbOrigSize ){
7194 assert( pPager->pTmpSpace!=0 );
7195 sqlite3BitvecClear(pPager->pInJournal, needSyncPgno, pPager->pTmpSpace);
7196 }
7197 return rc;
7198 }
7199 pPgHdr->flags |= PGHDR_NEED_SYNC;
7200 sqlite3PcacheMakeDirty(pPgHdr);
7201 sqlite3PagerUnrefNotNull(pPgHdr);
7202 }
7203
7204 return SQLITE_OK;
7205}
7206#endif
7207
7208/*
7209** The page handle passed as the first argument refers to a dirty page
7210** with a page number other than iNew. This function changes the page's
7211** page number to iNew and sets the value of the PgHdr.flags field to
7212** the value passed as the third parameter.
7213*/
7214void sqlite3PagerRekey(DbPage *pPg, Pgno iNew, u16 flags){
7215 assert( pPg->pgno!=iNew );
7216 pPg->flags = flags;
7217 sqlite3PcacheMove(pPg, iNew);
7218}
7219
7220/*
7221** Return a pointer to the data for the specified page.
7222*/
7223void *sqlite3PagerGetData(DbPage *pPg){
7224 assert( pPg->nRef>0 || pPg->pPager->memDb );
7225 return pPg->pData;
7226}
7227
7228/*
7229** Return a pointer to the Pager.nExtra bytes of "extra" space
7230** allocated along with the specified page.
7231*/
7232void *sqlite3PagerGetExtra(DbPage *pPg){
7233 return pPg->pExtra;
7234}
7235
7236/*
7237** Get/set the locking-mode for this pager. Parameter eMode must be one
7238** of PAGER_LOCKINGMODE_QUERY, PAGER_LOCKINGMODE_NORMAL or
7239** PAGER_LOCKINGMODE_EXCLUSIVE. If the parameter is not _QUERY, then
7240** the locking-mode is set to the value specified.
7241**
7242** The returned value is either PAGER_LOCKINGMODE_NORMAL or
7243** PAGER_LOCKINGMODE_EXCLUSIVE, indicating the current (possibly updated)
7244** locking-mode.
7245*/
7246int sqlite3PagerLockingMode(Pager *pPager, int eMode){
7247 assert( eMode==PAGER_LOCKINGMODE_QUERY
7248 || eMode==PAGER_LOCKINGMODE_NORMAL
7249 || eMode==PAGER_LOCKINGMODE_EXCLUSIVE );
7250 assert( PAGER_LOCKINGMODE_QUERY<0 );
7251 assert( PAGER_LOCKINGMODE_NORMAL>=0 && PAGER_LOCKINGMODE_EXCLUSIVE>=0 );
7252 assert( pPager->exclusiveMode || 0==sqlite3WalHeapMemory(pPager->pWal) );
7253 if( eMode>=0 && !pPager->tempFile && !sqlite3WalHeapMemory(pPager->pWal) ){
7254 pPager->exclusiveMode = (u8)eMode;
7255 }
7256 return (int)pPager->exclusiveMode;
7257}
7258
7259/*
7260** Set the journal-mode for this pager. Parameter eMode must be one of:
7261**
7262** PAGER_JOURNALMODE_DELETE
7263** PAGER_JOURNALMODE_TRUNCATE
7264** PAGER_JOURNALMODE_PERSIST
7265** PAGER_JOURNALMODE_OFF
7266** PAGER_JOURNALMODE_MEMORY
7267** PAGER_JOURNALMODE_WAL
7268**
7269** The journalmode is set to the value specified if the change is allowed.
7270** The change may be disallowed for the following reasons:
7271**
7272** * An in-memory database can only have its journal_mode set to _OFF
7273** or _MEMORY.
7274**
7275** * Temporary databases cannot have _WAL journalmode.
7276**
7277** The returned indicate the current (possibly updated) journal-mode.
7278*/
7279int sqlite3PagerSetJournalMode(Pager *pPager, int eMode){
7280 u8 eOld = pPager->journalMode; /* Prior journalmode */
7281
7282 /* The eMode parameter is always valid */
7283 assert( eMode==PAGER_JOURNALMODE_DELETE /* 0 */
7284 || eMode==PAGER_JOURNALMODE_PERSIST /* 1 */
7285 || eMode==PAGER_JOURNALMODE_OFF /* 2 */
7286 || eMode==PAGER_JOURNALMODE_TRUNCATE /* 3 */
7287 || eMode==PAGER_JOURNALMODE_MEMORY /* 4 */
7288 || eMode==PAGER_JOURNALMODE_WAL /* 5 */ );
7289
7290 /* This routine is only called from the OP_JournalMode opcode, and
7291 ** the logic there will never allow a temporary file to be changed
7292 ** to WAL mode.
7293 */
7294 assert( pPager->tempFile==0 || eMode!=PAGER_JOURNALMODE_WAL );
7295
7296 /* Do allow the journalmode of an in-memory database to be set to
7297 ** anything other than MEMORY or OFF
7298 */
7299 if( MEMDB ){
7300 assert( eOld==PAGER_JOURNALMODE_MEMORY || eOld==PAGER_JOURNALMODE_OFF );
7301 if( eMode!=PAGER_JOURNALMODE_MEMORY && eMode!=PAGER_JOURNALMODE_OFF ){
7302 eMode = eOld;
7303 }
7304 }
7305
7306 if( eMode!=eOld ){
7307
7308 /* Change the journal mode. */
7309 assert( pPager->eState!=PAGER_ERROR );
7310 pPager->journalMode = (u8)eMode;
7311
7312 /* When transistioning from TRUNCATE or PERSIST to any other journal
7313 ** mode except WAL, unless the pager is in locking_mode=exclusive mode,
7314 ** delete the journal file.
7315 */
7316 assert( (PAGER_JOURNALMODE_TRUNCATE & 5)==1 );
7317 assert( (PAGER_JOURNALMODE_PERSIST & 5)==1 );
7318 assert( (PAGER_JOURNALMODE_DELETE & 5)==0 );
7319 assert( (PAGER_JOURNALMODE_MEMORY & 5)==4 );
7320 assert( (PAGER_JOURNALMODE_OFF & 5)==0 );
7321 assert( (PAGER_JOURNALMODE_WAL & 5)==5 );
7322
7323 assert( isOpen(pPager->fd) || pPager->exclusiveMode );
7324 if( !pPager->exclusiveMode && (eOld & 5)==1 && (eMode & 1)==0 ){
7325 /* In this case we would like to delete the journal file. If it is
7326 ** not possible, then that is not a problem. Deleting the journal file
7327 ** here is an optimization only.
7328 **
7329 ** Before deleting the journal file, obtain a RESERVED lock on the
7330 ** database file. This ensures that the journal file is not deleted
7331 ** while it is in use by some other client.
7332 */
7333 sqlite3OsClose(pPager->jfd);
7334 if( pPager->eLock>=RESERVED_LOCK ){
7335 sqlite3OsDelete(pPager->pVfs, pPager->zJournal, 0);
7336 }else{
7337 int rc = SQLITE_OK;
7338 int state = pPager->eState;
7339 assert( state==PAGER_OPEN || state==PAGER_READER );
7340 if( state==PAGER_OPEN ){
7341 rc = sqlite3PagerSharedLock(pPager);
7342 }
7343 if( pPager->eState==PAGER_READER ){
7344 assert( rc==SQLITE_OK );
7345 rc = pagerLockDb(pPager, RESERVED_LOCK);
7346 }
7347 if( rc==SQLITE_OK ){
7348 sqlite3OsDelete(pPager->pVfs, pPager->zJournal, 0);
7349 }
7350 if( rc==SQLITE_OK && state==PAGER_READER ){
7351 pagerUnlockDb(pPager, SHARED_LOCK);
7352 }else if( state==PAGER_OPEN ){
7353 pager_unlock(pPager);
7354 }
7355 assert( state==pPager->eState );
7356 }
7357 }else if( eMode==PAGER_JOURNALMODE_OFF ){
7358 sqlite3OsClose(pPager->jfd);
7359 }
7360 }
7361
7362 /* Return the new journal mode */
7363 return (int)pPager->journalMode;
7364}
7365
7366/*
7367** Return the current journal mode.
7368*/
7369int sqlite3PagerGetJournalMode(Pager *pPager){
7370 return (int)pPager->journalMode;
7371}
7372
7373/*
7374** Return TRUE if the pager is in a state where it is OK to change the
7375** journalmode. Journalmode changes can only happen when the database
7376** is unmodified.
7377*/
7378int sqlite3PagerOkToChangeJournalMode(Pager *pPager){
7379 assert( assert_pager_state(pPager) );
7380 if( pPager->eState>=PAGER_WRITER_CACHEMOD ) return 0;
7381 if( NEVER(isOpen(pPager->jfd) && pPager->journalOff>0) ) return 0;
7382 return 1;
7383}
7384
7385/*
7386** Get/set the size-limit used for persistent journal files.
7387**
7388** Setting the size limit to -1 means no limit is enforced.
7389** An attempt to set a limit smaller than -1 is a no-op.
7390*/
7391i64 sqlite3PagerJournalSizeLimit(Pager *pPager, i64 iLimit){
7392 if( iLimit>=-1 ){
7393 pPager->journalSizeLimit = iLimit;
7394 sqlite3WalLimit(pPager->pWal, iLimit);
7395 }
7396 return pPager->journalSizeLimit;
7397}
7398
7399/*
7400** Return a pointer to the pPager->pBackup variable. The backup module
7401** in backup.c maintains the content of this variable. This module
7402** uses it opaquely as an argument to sqlite3BackupRestart() and
7403** sqlite3BackupUpdate() only.
7404*/
7405sqlite3_backup **sqlite3PagerBackupPtr(Pager *pPager){
7406 return &pPager->pBackup;
7407}
7408
7409#ifndef SQLITE_OMIT_VACUUM
7410/*
7411** Unless this is an in-memory or temporary database, clear the pager cache.
7412*/
7413void sqlite3PagerClearCache(Pager *pPager){
7414 assert( MEMDB==0 || pPager->tempFile );
7415 if( pPager->tempFile==0 ) pager_reset(pPager);
7416}
7417#endif
7418
7419
7420#ifndef SQLITE_OMIT_WAL
7421/*
7422** This function is called when the user invokes "PRAGMA wal_checkpoint",
7423** "PRAGMA wal_blocking_checkpoint" or calls the sqlite3_wal_checkpoint()
7424** or wal_blocking_checkpoint() API functions.
7425**
7426** Parameter eMode is one of SQLITE_CHECKPOINT_PASSIVE, FULL or RESTART.
7427*/
7428int sqlite3PagerCheckpoint(
7429 Pager *pPager, /* Checkpoint on this pager */
7430 sqlite3 *db, /* Db handle used to check for interrupts */
7431 int eMode, /* Type of checkpoint */
7432 int *pnLog, /* OUT: Final number of frames in log */
7433 int *pnCkpt /* OUT: Final number of checkpointed frames */
7434){
7435 int rc = SQLITE_OK;
7436 if( pPager->pWal==0 && pPager->journalMode==PAGER_JOURNALMODE_WAL ){
7437 /* This only happens when a database file is zero bytes in size opened and
7438 ** then "PRAGMA journal_mode=WAL" is run and then sqlite3_wal_checkpoint()
7439 ** is invoked without any intervening transactions. We need to start
7440 ** a transaction to initialize pWal. The PRAGMA table_list statement is
7441 ** used for this since it starts transactions on every database file,
7442 ** including all ATTACHed databases. This seems expensive for a single
7443 ** sqlite3_wal_checkpoint() call, but it happens very rarely.
7444 ** https://sqlite.org/forum/forumpost/fd0f19d229156939
7445 */
7446 sqlite3_exec(db, "PRAGMA table_list",0,0,0);
7447 }
7448 if( pPager->pWal ){
7449 rc = sqlite3WalCheckpoint(pPager->pWal, db, eMode,
7450 (eMode==SQLITE_CHECKPOINT_PASSIVE ? 0 : pPager->xBusyHandler),
7451 pPager->pBusyHandlerArg,
7452 pPager->walSyncFlags, pPager->pageSize, (u8 *)pPager->pTmpSpace,
7453 pnLog, pnCkpt
7454 );
7455 }
7456 return rc;
7457}
7458
7459int sqlite3PagerWalCallback(Pager *pPager){
7460 return sqlite3WalCallback(pPager->pWal);
7461}
7462
7463/*
7464** Return true if the underlying VFS for the given pager supports the
7465** primitives necessary for write-ahead logging.
7466*/
7467int sqlite3PagerWalSupported(Pager *pPager){
7468 const sqlite3_io_methods *pMethods = pPager->fd->pMethods;
7469 if( pPager->noLock ) return 0;
7470 return pPager->exclusiveMode || (pMethods->iVersion>=2 && pMethods->xShmMap);
7471}
7472
7473/*
7474** Attempt to take an exclusive lock on the database file. If a PENDING lock
7475** is obtained instead, immediately release it.
7476*/
7477static int pagerExclusiveLock(Pager *pPager){
7478 int rc; /* Return code */
7479
7480 assert( pPager->eLock==SHARED_LOCK || pPager->eLock==EXCLUSIVE_LOCK );
7481 rc = pagerLockDb(pPager, EXCLUSIVE_LOCK);
7482 if( rc!=SQLITE_OK ){
7483 /* If the attempt to grab the exclusive lock failed, release the
7484 ** pending lock that may have been obtained instead. */
7485 pagerUnlockDb(pPager, SHARED_LOCK);
7486 }
7487
7488 return rc;
7489}
7490
7491/*
7492** Call sqlite3WalOpen() to open the WAL handle. If the pager is in
7493** exclusive-locking mode when this function is called, take an EXCLUSIVE
7494** lock on the database file and use heap-memory to store the wal-index
7495** in. Otherwise, use the normal shared-memory.
7496*/
7497static int pagerOpenWal(Pager *pPager){
7498 int rc = SQLITE_OK;
7499
7500 assert( pPager->pWal==0 && pPager->tempFile==0 );
7501 assert( pPager->eLock==SHARED_LOCK || pPager->eLock==EXCLUSIVE_LOCK );
7502
7503 /* If the pager is already in exclusive-mode, the WAL module will use
7504 ** heap-memory for the wal-index instead of the VFS shared-memory
7505 ** implementation. Take the exclusive lock now, before opening the WAL
7506 ** file, to make sure this is safe.
7507 */
7508 if( pPager->exclusiveMode ){
7509 rc = pagerExclusiveLock(pPager);
7510 }
7511
7512 /* Open the connection to the log file. If this operation fails,
7513 ** (e.g. due to malloc() failure), return an error code.
7514 */
7515 if( rc==SQLITE_OK ){
7516 rc = sqlite3WalOpen(pPager->pVfs,
7517 pPager->fd, pPager->zWal, pPager->exclusiveMode,
7518 pPager->journalSizeLimit, &pPager->pWal
7519 );
7520 }
7521 pagerFixMaplimit(pPager);
7522
7523 return rc;
7524}
7525
7526
7527/*
7528** The caller must be holding a SHARED lock on the database file to call
7529** this function.
7530**
7531** If the pager passed as the first argument is open on a real database
7532** file (not a temp file or an in-memory database), and the WAL file
7533** is not already open, make an attempt to open it now. If successful,
7534** return SQLITE_OK. If an error occurs or the VFS used by the pager does
7535** not support the xShmXXX() methods, return an error code. *pbOpen is
7536** not modified in either case.
7537**
7538** If the pager is open on a temp-file (or in-memory database), or if
7539** the WAL file is already open, set *pbOpen to 1 and return SQLITE_OK
7540** without doing anything.
7541*/
7542int sqlite3PagerOpenWal(
7543 Pager *pPager, /* Pager object */
7544 int *pbOpen /* OUT: Set to true if call is a no-op */
7545){
7546 int rc = SQLITE_OK; /* Return code */
7547
7548 assert( assert_pager_state(pPager) );
7549 assert( pPager->eState==PAGER_OPEN || pbOpen );
7550 assert( pPager->eState==PAGER_READER || !pbOpen );
7551 assert( pbOpen==0 || *pbOpen==0 );
7552 assert( pbOpen!=0 || (!pPager->tempFile && !pPager->pWal) );
7553
7554 if( !pPager->tempFile && !pPager->pWal ){
7555 if( !sqlite3PagerWalSupported(pPager) ) return SQLITE_CANTOPEN;
7556
7557 /* Close any rollback journal previously open */
7558 sqlite3OsClose(pPager->jfd);
7559
7560 rc = pagerOpenWal(pPager);
7561 if( rc==SQLITE_OK ){
7562 pPager->journalMode = PAGER_JOURNALMODE_WAL;
7563 pPager->eState = PAGER_OPEN;
7564 }
7565 }else{
7566 *pbOpen = 1;
7567 }
7568
7569 return rc;
7570}
7571
7572/*
7573** This function is called to close the connection to the log file prior
7574** to switching from WAL to rollback mode.
7575**
7576** Before closing the log file, this function attempts to take an
7577** EXCLUSIVE lock on the database file. If this cannot be obtained, an
7578** error (SQLITE_BUSY) is returned and the log connection is not closed.
7579** If successful, the EXCLUSIVE lock is not released before returning.
7580*/
7581int sqlite3PagerCloseWal(Pager *pPager, sqlite3 *db){
7582 int rc = SQLITE_OK;
7583
7584 assert( pPager->journalMode==PAGER_JOURNALMODE_WAL );
7585
7586 /* If the log file is not already open, but does exist in the file-system,
7587 ** it may need to be checkpointed before the connection can switch to
7588 ** rollback mode. Open it now so this can happen.
7589 */
7590 if( !pPager->pWal ){
7591 int logexists = 0;
7592 rc = pagerLockDb(pPager, SHARED_LOCK);
7593 if( rc==SQLITE_OK ){
7594 rc = sqlite3OsAccess(
7595 pPager->pVfs, pPager->zWal, SQLITE_ACCESS_EXISTS, &logexists
7596 );
7597 }
7598 if( rc==SQLITE_OK && logexists ){
7599 rc = pagerOpenWal(pPager);
7600 }
7601 }
7602
7603 /* Checkpoint and close the log. Because an EXCLUSIVE lock is held on
7604 ** the database file, the log and log-summary files will be deleted.
7605 */
7606 if( rc==SQLITE_OK && pPager->pWal ){
7607 rc = pagerExclusiveLock(pPager);
7608 if( rc==SQLITE_OK ){
7609 rc = sqlite3WalClose(pPager->pWal, db, pPager->walSyncFlags,
7610 pPager->pageSize, (u8*)pPager->pTmpSpace);
7611 pPager->pWal = 0;
7612 pagerFixMaplimit(pPager);
7613 if( rc && !pPager->exclusiveMode ) pagerUnlockDb(pPager, SHARED_LOCK);
7614 }
7615 }
7616 return rc;
7617}
7618
7619#ifdef SQLITE_ENABLE_SETLK_TIMEOUT
7620/*
7621** If pager pPager is a wal-mode database not in exclusive locking mode,
7622** invoke the sqlite3WalWriteLock() function on the associated Wal object
7623** with the same db and bLock parameters as were passed to this function.
7624** Return an SQLite error code if an error occurs, or SQLITE_OK otherwise.
7625*/
7626int sqlite3PagerWalWriteLock(Pager *pPager, int bLock){
7627 int rc = SQLITE_OK;
7628 if( pagerUseWal(pPager) && pPager->exclusiveMode==0 ){
7629 rc = sqlite3WalWriteLock(pPager->pWal, bLock);
7630 }
7631 return rc;
7632}
7633
7634/*
7635** Set the database handle used by the wal layer to determine if
7636** blocking locks are required.
7637*/
7638void sqlite3PagerWalDb(Pager *pPager, sqlite3 *db){
7639 if( pagerUseWal(pPager) ){
7640 sqlite3WalDb(pPager->pWal, db);
7641 }
7642}
7643#endif
7644
7645#ifdef SQLITE_ENABLE_SNAPSHOT
7646/*
7647** If this is a WAL database, obtain a snapshot handle for the snapshot
7648** currently open. Otherwise, return an error.
7649*/
7650int sqlite3PagerSnapshotGet(Pager *pPager, sqlite3_snapshot **ppSnapshot){
7651 int rc = SQLITE_ERROR;
7652 if( pPager->pWal ){
7653 rc = sqlite3WalSnapshotGet(pPager->pWal, ppSnapshot);
7654 }
7655 return rc;
7656}
7657
7658/*
7659** If this is a WAL database, store a pointer to pSnapshot. Next time a
7660** read transaction is opened, attempt to read from the snapshot it
7661** identifies. If this is not a WAL database, return an error.
7662*/
7663int sqlite3PagerSnapshotOpen(
7664 Pager *pPager,
7665 sqlite3_snapshot *pSnapshot
7666){
7667 int rc = SQLITE_OK;
7668 if( pPager->pWal ){
7669 sqlite3WalSnapshotOpen(pPager->pWal, pSnapshot);
7670 }else{
7671 rc = SQLITE_ERROR;
7672 }
7673 return rc;
7674}
7675
7676/*
7677** If this is a WAL database, call sqlite3WalSnapshotRecover(). If this
7678** is not a WAL database, return an error.
7679*/
7680int sqlite3PagerSnapshotRecover(Pager *pPager){
7681 int rc;
7682 if( pPager->pWal ){
7683 rc = sqlite3WalSnapshotRecover(pPager->pWal);
7684 }else{
7685 rc = SQLITE_ERROR;
7686 }
7687 return rc;
7688}
7689
7690/*
7691** The caller currently has a read transaction open on the database.
7692** If this is not a WAL database, SQLITE_ERROR is returned. Otherwise,
7693** this function takes a SHARED lock on the CHECKPOINTER slot and then
7694** checks if the snapshot passed as the second argument is still
7695** available. If so, SQLITE_OK is returned.
7696**
7697** If the snapshot is not available, SQLITE_ERROR is returned. Or, if
7698** the CHECKPOINTER lock cannot be obtained, SQLITE_BUSY. If any error
7699** occurs (any value other than SQLITE_OK is returned), the CHECKPOINTER
7700** lock is released before returning.
7701*/
7702int sqlite3PagerSnapshotCheck(Pager *pPager, sqlite3_snapshot *pSnapshot){
7703 int rc;
7704 if( pPager->pWal ){
7705 rc = sqlite3WalSnapshotCheck(pPager->pWal, pSnapshot);
7706 }else{
7707 rc = SQLITE_ERROR;
7708 }
7709 return rc;
7710}
7711
7712/*
7713** Release a lock obtained by an earlier successful call to
7714** sqlite3PagerSnapshotCheck().
7715*/
7716void sqlite3PagerSnapshotUnlock(Pager *pPager){
7717 assert( pPager->pWal );
7718 sqlite3WalSnapshotUnlock(pPager->pWal);
7719}
7720
7721#endif /* SQLITE_ENABLE_SNAPSHOT */
7722#endif /* !SQLITE_OMIT_WAL */
7723
7724#ifdef SQLITE_ENABLE_ZIPVFS
7725/*
7726** A read-lock must be held on the pager when this function is called. If
7727** the pager is in WAL mode and the WAL file currently contains one or more
7728** frames, return the size in bytes of the page images stored within the
7729** WAL frames. Otherwise, if this is not a WAL database or the WAL file
7730** is empty, return 0.
7731*/
7732int sqlite3PagerWalFramesize(Pager *pPager){
7733 assert( pPager->eState>=PAGER_READER );
7734 return sqlite3WalFramesize(pPager->pWal);
7735}
7736#endif
7737
7738#endif /* SQLITE_OMIT_DISKIO */
7739