Annotation of embedaddon/sqlite3/src/wal.c, revision 1.1.1.1
1.1 misho 1: /*
2: ** 2010 February 1
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: **
13: ** This file contains the implementation of a write-ahead log (WAL) used in
14: ** "journal_mode=WAL" mode.
15: **
16: ** WRITE-AHEAD LOG (WAL) FILE FORMAT
17: **
18: ** A WAL file consists of a header followed by zero or more "frames".
19: ** Each frame records the revised content of a single page from the
20: ** database file. All changes to the database are recorded by writing
21: ** frames into the WAL. Transactions commit when a frame is written that
22: ** contains a commit marker. A single WAL can and usually does record
23: ** multiple transactions. Periodically, the content of the WAL is
24: ** transferred back into the database file in an operation called a
25: ** "checkpoint".
26: **
27: ** A single WAL file can be used multiple times. In other words, the
28: ** WAL can fill up with frames and then be checkpointed and then new
29: ** frames can overwrite the old ones. A WAL always grows from beginning
30: ** toward the end. Checksums and counters attached to each frame are
31: ** used to determine which frames within the WAL are valid and which
32: ** are leftovers from prior checkpoints.
33: **
34: ** The WAL header is 32 bytes in size and consists of the following eight
35: ** big-endian 32-bit unsigned integer values:
36: **
37: ** 0: Magic number. 0x377f0682 or 0x377f0683
38: ** 4: File format version. Currently 3007000
39: ** 8: Database page size. Example: 1024
40: ** 12: Checkpoint sequence number
41: ** 16: Salt-1, random integer incremented with each checkpoint
42: ** 20: Salt-2, a different random integer changing with each ckpt
43: ** 24: Checksum-1 (first part of checksum for first 24 bytes of header).
44: ** 28: Checksum-2 (second part of checksum for first 24 bytes of header).
45: **
46: ** Immediately following the wal-header are zero or more frames. Each
47: ** frame consists of a 24-byte frame-header followed by a <page-size> bytes
48: ** of page data. The frame-header is six big-endian 32-bit unsigned
49: ** integer values, as follows:
50: **
51: ** 0: Page number.
52: ** 4: For commit records, the size of the database image in pages
53: ** after the commit. For all other records, zero.
54: ** 8: Salt-1 (copied from the header)
55: ** 12: Salt-2 (copied from the header)
56: ** 16: Checksum-1.
57: ** 20: Checksum-2.
58: **
59: ** A frame is considered valid if and only if the following conditions are
60: ** true:
61: **
62: ** (1) The salt-1 and salt-2 values in the frame-header match
63: ** salt values in the wal-header
64: **
65: ** (2) The checksum values in the final 8 bytes of the frame-header
66: ** exactly match the checksum computed consecutively on the
67: ** WAL header and the first 8 bytes and the content of all frames
68: ** up to and including the current frame.
69: **
70: ** The checksum is computed using 32-bit big-endian integers if the
71: ** magic number in the first 4 bytes of the WAL is 0x377f0683 and it
72: ** is computed using little-endian if the magic number is 0x377f0682.
73: ** The checksum values are always stored in the frame header in a
74: ** big-endian format regardless of which byte order is used to compute
75: ** the checksum. The checksum is computed by interpreting the input as
76: ** an even number of unsigned 32-bit integers: x[0] through x[N]. The
77: ** algorithm used for the checksum is as follows:
78: **
79: ** for i from 0 to n-1 step 2:
80: ** s0 += x[i] + s1;
81: ** s1 += x[i+1] + s0;
82: ** endfor
83: **
84: ** Note that s0 and s1 are both weighted checksums using fibonacci weights
85: ** in reverse order (the largest fibonacci weight occurs on the first element
86: ** of the sequence being summed.) The s1 value spans all 32-bit
87: ** terms of the sequence whereas s0 omits the final term.
88: **
89: ** On a checkpoint, the WAL is first VFS.xSync-ed, then valid content of the
90: ** WAL is transferred into the database, then the database is VFS.xSync-ed.
91: ** The VFS.xSync operations serve as write barriers - all writes launched
92: ** before the xSync must complete before any write that launches after the
93: ** xSync begins.
94: **
95: ** After each checkpoint, the salt-1 value is incremented and the salt-2
96: ** value is randomized. This prevents old and new frames in the WAL from
97: ** being considered valid at the same time and being checkpointing together
98: ** following a crash.
99: **
100: ** READER ALGORITHM
101: **
102: ** To read a page from the database (call it page number P), a reader
103: ** first checks the WAL to see if it contains page P. If so, then the
104: ** last valid instance of page P that is a followed by a commit frame
105: ** or is a commit frame itself becomes the value read. If the WAL
106: ** contains no copies of page P that are valid and which are a commit
107: ** frame or are followed by a commit frame, then page P is read from
108: ** the database file.
109: **
110: ** To start a read transaction, the reader records the index of the last
111: ** valid frame in the WAL. The reader uses this recorded "mxFrame" value
112: ** for all subsequent read operations. New transactions can be appended
113: ** to the WAL, but as long as the reader uses its original mxFrame value
114: ** and ignores the newly appended content, it will see a consistent snapshot
115: ** of the database from a single point in time. This technique allows
116: ** multiple concurrent readers to view different versions of the database
117: ** content simultaneously.
118: **
119: ** The reader algorithm in the previous paragraphs works correctly, but
120: ** because frames for page P can appear anywhere within the WAL, the
121: ** reader has to scan the entire WAL looking for page P frames. If the
122: ** WAL is large (multiple megabytes is typical) that scan can be slow,
123: ** and read performance suffers. To overcome this problem, a separate
124: ** data structure called the wal-index is maintained to expedite the
125: ** search for frames of a particular page.
126: **
127: ** WAL-INDEX FORMAT
128: **
129: ** Conceptually, the wal-index is shared memory, though VFS implementations
130: ** might choose to implement the wal-index using a mmapped file. Because
131: ** the wal-index is shared memory, SQLite does not support journal_mode=WAL
132: ** on a network filesystem. All users of the database must be able to
133: ** share memory.
134: **
135: ** The wal-index is transient. After a crash, the wal-index can (and should
136: ** be) reconstructed from the original WAL file. In fact, the VFS is required
137: ** to either truncate or zero the header of the wal-index when the last
138: ** connection to it closes. Because the wal-index is transient, it can
139: ** use an architecture-specific format; it does not have to be cross-platform.
140: ** Hence, unlike the database and WAL file formats which store all values
141: ** as big endian, the wal-index can store multi-byte values in the native
142: ** byte order of the host computer.
143: **
144: ** The purpose of the wal-index is to answer this question quickly: Given
145: ** a page number P, return the index of the last frame for page P in the WAL,
146: ** or return NULL if there are no frames for page P in the WAL.
147: **
148: ** The wal-index consists of a header region, followed by an one or
149: ** more index blocks.
150: **
151: ** The wal-index header contains the total number of frames within the WAL
152: ** in the the mxFrame field.
153: **
154: ** Each index block except for the first contains information on
155: ** HASHTABLE_NPAGE frames. The first index block contains information on
156: ** HASHTABLE_NPAGE_ONE frames. The values of HASHTABLE_NPAGE_ONE and
157: ** HASHTABLE_NPAGE are selected so that together the wal-index header and
158: ** first index block are the same size as all other index blocks in the
159: ** wal-index.
160: **
161: ** Each index block contains two sections, a page-mapping that contains the
162: ** database page number associated with each wal frame, and a hash-table
163: ** that allows readers to query an index block for a specific page number.
164: ** The page-mapping is an array of HASHTABLE_NPAGE (or HASHTABLE_NPAGE_ONE
165: ** for the first index block) 32-bit page numbers. The first entry in the
166: ** first index-block contains the database page number corresponding to the
167: ** first frame in the WAL file. The first entry in the second index block
168: ** in the WAL file corresponds to the (HASHTABLE_NPAGE_ONE+1)th frame in
169: ** the log, and so on.
170: **
171: ** The last index block in a wal-index usually contains less than the full
172: ** complement of HASHTABLE_NPAGE (or HASHTABLE_NPAGE_ONE) page-numbers,
173: ** depending on the contents of the WAL file. This does not change the
174: ** allocated size of the page-mapping array - the page-mapping array merely
175: ** contains unused entries.
176: **
177: ** Even without using the hash table, the last frame for page P
178: ** can be found by scanning the page-mapping sections of each index block
179: ** starting with the last index block and moving toward the first, and
180: ** within each index block, starting at the end and moving toward the
181: ** beginning. The first entry that equals P corresponds to the frame
182: ** holding the content for that page.
183: **
184: ** The hash table consists of HASHTABLE_NSLOT 16-bit unsigned integers.
185: ** HASHTABLE_NSLOT = 2*HASHTABLE_NPAGE, and there is one entry in the
186: ** hash table for each page number in the mapping section, so the hash
187: ** table is never more than half full. The expected number of collisions
188: ** prior to finding a match is 1. Each entry of the hash table is an
189: ** 1-based index of an entry in the mapping section of the same
190: ** index block. Let K be the 1-based index of the largest entry in
191: ** the mapping section. (For index blocks other than the last, K will
192: ** always be exactly HASHTABLE_NPAGE (4096) and for the last index block
193: ** K will be (mxFrame%HASHTABLE_NPAGE).) Unused slots of the hash table
194: ** contain a value of 0.
195: **
196: ** To look for page P in the hash table, first compute a hash iKey on
197: ** P as follows:
198: **
199: ** iKey = (P * 383) % HASHTABLE_NSLOT
200: **
201: ** Then start scanning entries of the hash table, starting with iKey
202: ** (wrapping around to the beginning when the end of the hash table is
203: ** reached) until an unused hash slot is found. Let the first unused slot
204: ** be at index iUnused. (iUnused might be less than iKey if there was
205: ** wrap-around.) Because the hash table is never more than half full,
206: ** the search is guaranteed to eventually hit an unused entry. Let
207: ** iMax be the value between iKey and iUnused, closest to iUnused,
208: ** where aHash[iMax]==P. If there is no iMax entry (if there exists
209: ** no hash slot such that aHash[i]==p) then page P is not in the
210: ** current index block. Otherwise the iMax-th mapping entry of the
211: ** current index block corresponds to the last entry that references
212: ** page P.
213: **
214: ** A hash search begins with the last index block and moves toward the
215: ** first index block, looking for entries corresponding to page P. On
216: ** average, only two or three slots in each index block need to be
217: ** examined in order to either find the last entry for page P, or to
218: ** establish that no such entry exists in the block. Each index block
219: ** holds over 4000 entries. So two or three index blocks are sufficient
220: ** to cover a typical 10 megabyte WAL file, assuming 1K pages. 8 or 10
221: ** comparisons (on average) suffice to either locate a frame in the
222: ** WAL or to establish that the frame does not exist in the WAL. This
223: ** is much faster than scanning the entire 10MB WAL.
224: **
225: ** Note that entries are added in order of increasing K. Hence, one
226: ** reader might be using some value K0 and a second reader that started
227: ** at a later time (after additional transactions were added to the WAL
228: ** and to the wal-index) might be using a different value K1, where K1>K0.
229: ** Both readers can use the same hash table and mapping section to get
230: ** the correct result. There may be entries in the hash table with
231: ** K>K0 but to the first reader, those entries will appear to be unused
232: ** slots in the hash table and so the first reader will get an answer as
233: ** if no values greater than K0 had ever been inserted into the hash table
234: ** in the first place - which is what reader one wants. Meanwhile, the
235: ** second reader using K1 will see additional values that were inserted
236: ** later, which is exactly what reader two wants.
237: **
238: ** When a rollback occurs, the value of K is decreased. Hash table entries
239: ** that correspond to frames greater than the new K value are removed
240: ** from the hash table at this point.
241: */
242: #ifndef SQLITE_OMIT_WAL
243:
244: #include "wal.h"
245:
246: /*
247: ** Trace output macros
248: */
249: #if defined(SQLITE_TEST) && defined(SQLITE_DEBUG)
250: int sqlite3WalTrace = 0;
251: # define WALTRACE(X) if(sqlite3WalTrace) sqlite3DebugPrintf X
252: #else
253: # define WALTRACE(X)
254: #endif
255:
256: /*
257: ** The maximum (and only) versions of the wal and wal-index formats
258: ** that may be interpreted by this version of SQLite.
259: **
260: ** If a client begins recovering a WAL file and finds that (a) the checksum
261: ** values in the wal-header are correct and (b) the version field is not
262: ** WAL_MAX_VERSION, recovery fails and SQLite returns SQLITE_CANTOPEN.
263: **
264: ** Similarly, if a client successfully reads a wal-index header (i.e. the
265: ** checksum test is successful) and finds that the version field is not
266: ** WALINDEX_MAX_VERSION, then no read-transaction is opened and SQLite
267: ** returns SQLITE_CANTOPEN.
268: */
269: #define WAL_MAX_VERSION 3007000
270: #define WALINDEX_MAX_VERSION 3007000
271:
272: /*
273: ** Indices of various locking bytes. WAL_NREADER is the number
274: ** of available reader locks and should be at least 3.
275: */
276: #define WAL_WRITE_LOCK 0
277: #define WAL_ALL_BUT_WRITE 1
278: #define WAL_CKPT_LOCK 1
279: #define WAL_RECOVER_LOCK 2
280: #define WAL_READ_LOCK(I) (3+(I))
281: #define WAL_NREADER (SQLITE_SHM_NLOCK-3)
282:
283:
284: /* Object declarations */
285: typedef struct WalIndexHdr WalIndexHdr;
286: typedef struct WalIterator WalIterator;
287: typedef struct WalCkptInfo WalCkptInfo;
288:
289:
290: /*
291: ** The following object holds a copy of the wal-index header content.
292: **
293: ** The actual header in the wal-index consists of two copies of this
294: ** object.
295: **
296: ** The szPage value can be any power of 2 between 512 and 32768, inclusive.
297: ** Or it can be 1 to represent a 65536-byte page. The latter case was
298: ** added in 3.7.1 when support for 64K pages was added.
299: */
300: struct WalIndexHdr {
301: u32 iVersion; /* Wal-index version */
302: u32 unused; /* Unused (padding) field */
303: u32 iChange; /* Counter incremented each transaction */
304: u8 isInit; /* 1 when initialized */
305: u8 bigEndCksum; /* True if checksums in WAL are big-endian */
306: u16 szPage; /* Database page size in bytes. 1==64K */
307: u32 mxFrame; /* Index of last valid frame in the WAL */
308: u32 nPage; /* Size of database in pages */
309: u32 aFrameCksum[2]; /* Checksum of last frame in log */
310: u32 aSalt[2]; /* Two salt values copied from WAL header */
311: u32 aCksum[2]; /* Checksum over all prior fields */
312: };
313:
314: /*
315: ** A copy of the following object occurs in the wal-index immediately
316: ** following the second copy of the WalIndexHdr. This object stores
317: ** information used by checkpoint.
318: **
319: ** nBackfill is the number of frames in the WAL that have been written
320: ** back into the database. (We call the act of moving content from WAL to
321: ** database "backfilling".) The nBackfill number is never greater than
322: ** WalIndexHdr.mxFrame. nBackfill can only be increased by threads
323: ** holding the WAL_CKPT_LOCK lock (which includes a recovery thread).
324: ** However, a WAL_WRITE_LOCK thread can move the value of nBackfill from
325: ** mxFrame back to zero when the WAL is reset.
326: **
327: ** There is one entry in aReadMark[] for each reader lock. If a reader
328: ** holds read-lock K, then the value in aReadMark[K] is no greater than
329: ** the mxFrame for that reader. The value READMARK_NOT_USED (0xffffffff)
330: ** for any aReadMark[] means that entry is unused. aReadMark[0] is
331: ** a special case; its value is never used and it exists as a place-holder
332: ** to avoid having to offset aReadMark[] indexs by one. Readers holding
333: ** WAL_READ_LOCK(0) always ignore the entire WAL and read all content
334: ** directly from the database.
335: **
336: ** The value of aReadMark[K] may only be changed by a thread that
337: ** is holding an exclusive lock on WAL_READ_LOCK(K). Thus, the value of
338: ** aReadMark[K] cannot changed while there is a reader is using that mark
339: ** since the reader will be holding a shared lock on WAL_READ_LOCK(K).
340: **
341: ** The checkpointer may only transfer frames from WAL to database where
342: ** the frame numbers are less than or equal to every aReadMark[] that is
343: ** in use (that is, every aReadMark[j] for which there is a corresponding
344: ** WAL_READ_LOCK(j)). New readers (usually) pick the aReadMark[] with the
345: ** largest value and will increase an unused aReadMark[] to mxFrame if there
346: ** is not already an aReadMark[] equal to mxFrame. The exception to the
347: ** previous sentence is when nBackfill equals mxFrame (meaning that everything
348: ** in the WAL has been backfilled into the database) then new readers
349: ** will choose aReadMark[0] which has value 0 and hence such reader will
350: ** get all their all content directly from the database file and ignore
351: ** the WAL.
352: **
353: ** Writers normally append new frames to the end of the WAL. However,
354: ** if nBackfill equals mxFrame (meaning that all WAL content has been
355: ** written back into the database) and if no readers are using the WAL
356: ** (in other words, if there are no WAL_READ_LOCK(i) where i>0) then
357: ** the writer will first "reset" the WAL back to the beginning and start
358: ** writing new content beginning at frame 1.
359: **
360: ** We assume that 32-bit loads are atomic and so no locks are needed in
361: ** order to read from any aReadMark[] entries.
362: */
363: struct WalCkptInfo {
364: u32 nBackfill; /* Number of WAL frames backfilled into DB */
365: u32 aReadMark[WAL_NREADER]; /* Reader marks */
366: };
367: #define READMARK_NOT_USED 0xffffffff
368:
369:
370: /* A block of WALINDEX_LOCK_RESERVED bytes beginning at
371: ** WALINDEX_LOCK_OFFSET is reserved for locks. Since some systems
372: ** only support mandatory file-locks, we do not read or write data
373: ** from the region of the file on which locks are applied.
374: */
375: #define WALINDEX_LOCK_OFFSET (sizeof(WalIndexHdr)*2 + sizeof(WalCkptInfo))
376: #define WALINDEX_LOCK_RESERVED 16
377: #define WALINDEX_HDR_SIZE (WALINDEX_LOCK_OFFSET+WALINDEX_LOCK_RESERVED)
378:
379: /* Size of header before each frame in wal */
380: #define WAL_FRAME_HDRSIZE 24
381:
382: /* Size of write ahead log header, including checksum. */
383: /* #define WAL_HDRSIZE 24 */
384: #define WAL_HDRSIZE 32
385:
386: /* WAL magic value. Either this value, or the same value with the least
387: ** significant bit also set (WAL_MAGIC | 0x00000001) is stored in 32-bit
388: ** big-endian format in the first 4 bytes of a WAL file.
389: **
390: ** If the LSB is set, then the checksums for each frame within the WAL
391: ** file are calculated by treating all data as an array of 32-bit
392: ** big-endian words. Otherwise, they are calculated by interpreting
393: ** all data as 32-bit little-endian words.
394: */
395: #define WAL_MAGIC 0x377f0682
396:
397: /*
398: ** Return the offset of frame iFrame in the write-ahead log file,
399: ** assuming a database page size of szPage bytes. The offset returned
400: ** is to the start of the write-ahead log frame-header.
401: */
402: #define walFrameOffset(iFrame, szPage) ( \
403: WAL_HDRSIZE + ((iFrame)-1)*(i64)((szPage)+WAL_FRAME_HDRSIZE) \
404: )
405:
406: /*
407: ** An open write-ahead log file is represented by an instance of the
408: ** following object.
409: */
410: struct Wal {
411: sqlite3_vfs *pVfs; /* The VFS used to create pDbFd */
412: sqlite3_file *pDbFd; /* File handle for the database file */
413: sqlite3_file *pWalFd; /* File handle for WAL file */
414: u32 iCallback; /* Value to pass to log callback (or 0) */
415: i64 mxWalSize; /* Truncate WAL to this size upon reset */
416: int nWiData; /* Size of array apWiData */
417: int szFirstBlock; /* Size of first block written to WAL file */
418: volatile u32 **apWiData; /* Pointer to wal-index content in memory */
419: u32 szPage; /* Database page size */
420: i16 readLock; /* Which read lock is being held. -1 for none */
421: u8 syncFlags; /* Flags to use to sync header writes */
422: u8 exclusiveMode; /* Non-zero if connection is in exclusive mode */
423: u8 writeLock; /* True if in a write transaction */
424: u8 ckptLock; /* True if holding a checkpoint lock */
425: u8 readOnly; /* WAL_RDWR, WAL_RDONLY, or WAL_SHM_RDONLY */
426: u8 truncateOnCommit; /* True to truncate WAL file on commit */
427: u8 syncHeader; /* Fsync the WAL header if true */
428: u8 padToSectorBoundary; /* Pad transactions out to the next sector */
429: WalIndexHdr hdr; /* Wal-index header for current transaction */
430: const char *zWalName; /* Name of WAL file */
431: u32 nCkpt; /* Checkpoint sequence counter in the wal-header */
432: #ifdef SQLITE_DEBUG
433: u8 lockError; /* True if a locking error has occurred */
434: #endif
435: };
436:
437: /*
438: ** Candidate values for Wal.exclusiveMode.
439: */
440: #define WAL_NORMAL_MODE 0
441: #define WAL_EXCLUSIVE_MODE 1
442: #define WAL_HEAPMEMORY_MODE 2
443:
444: /*
445: ** Possible values for WAL.readOnly
446: */
447: #define WAL_RDWR 0 /* Normal read/write connection */
448: #define WAL_RDONLY 1 /* The WAL file is readonly */
449: #define WAL_SHM_RDONLY 2 /* The SHM file is readonly */
450:
451: /*
452: ** Each page of the wal-index mapping contains a hash-table made up of
453: ** an array of HASHTABLE_NSLOT elements of the following type.
454: */
455: typedef u16 ht_slot;
456:
457: /*
458: ** This structure is used to implement an iterator that loops through
459: ** all frames in the WAL in database page order. Where two or more frames
460: ** correspond to the same database page, the iterator visits only the
461: ** frame most recently written to the WAL (in other words, the frame with
462: ** the largest index).
463: **
464: ** The internals of this structure are only accessed by:
465: **
466: ** walIteratorInit() - Create a new iterator,
467: ** walIteratorNext() - Step an iterator,
468: ** walIteratorFree() - Free an iterator.
469: **
470: ** This functionality is used by the checkpoint code (see walCheckpoint()).
471: */
472: struct WalIterator {
473: int iPrior; /* Last result returned from the iterator */
474: int nSegment; /* Number of entries in aSegment[] */
475: struct WalSegment {
476: int iNext; /* Next slot in aIndex[] not yet returned */
477: ht_slot *aIndex; /* i0, i1, i2... such that aPgno[iN] ascend */
478: u32 *aPgno; /* Array of page numbers. */
479: int nEntry; /* Nr. of entries in aPgno[] and aIndex[] */
480: int iZero; /* Frame number associated with aPgno[0] */
481: } aSegment[1]; /* One for every 32KB page in the wal-index */
482: };
483:
484: /*
485: ** Define the parameters of the hash tables in the wal-index file. There
486: ** is a hash-table following every HASHTABLE_NPAGE page numbers in the
487: ** wal-index.
488: **
489: ** Changing any of these constants will alter the wal-index format and
490: ** create incompatibilities.
491: */
492: #define HASHTABLE_NPAGE 4096 /* Must be power of 2 */
493: #define HASHTABLE_HASH_1 383 /* Should be prime */
494: #define HASHTABLE_NSLOT (HASHTABLE_NPAGE*2) /* Must be a power of 2 */
495:
496: /*
497: ** The block of page numbers associated with the first hash-table in a
498: ** wal-index is smaller than usual. This is so that there is a complete
499: ** hash-table on each aligned 32KB page of the wal-index.
500: */
501: #define HASHTABLE_NPAGE_ONE (HASHTABLE_NPAGE - (WALINDEX_HDR_SIZE/sizeof(u32)))
502:
503: /* The wal-index is divided into pages of WALINDEX_PGSZ bytes each. */
504: #define WALINDEX_PGSZ ( \
505: sizeof(ht_slot)*HASHTABLE_NSLOT + HASHTABLE_NPAGE*sizeof(u32) \
506: )
507:
508: /*
509: ** Obtain a pointer to the iPage'th page of the wal-index. The wal-index
510: ** is broken into pages of WALINDEX_PGSZ bytes. Wal-index pages are
511: ** numbered from zero.
512: **
513: ** If this call is successful, *ppPage is set to point to the wal-index
514: ** page and SQLITE_OK is returned. If an error (an OOM or VFS error) occurs,
515: ** then an SQLite error code is returned and *ppPage is set to 0.
516: */
517: static int walIndexPage(Wal *pWal, int iPage, volatile u32 **ppPage){
518: int rc = SQLITE_OK;
519:
520: /* Enlarge the pWal->apWiData[] array if required */
521: if( pWal->nWiData<=iPage ){
522: int nByte = sizeof(u32*)*(iPage+1);
523: volatile u32 **apNew;
524: apNew = (volatile u32 **)sqlite3_realloc((void *)pWal->apWiData, nByte);
525: if( !apNew ){
526: *ppPage = 0;
527: return SQLITE_NOMEM;
528: }
529: memset((void*)&apNew[pWal->nWiData], 0,
530: sizeof(u32*)*(iPage+1-pWal->nWiData));
531: pWal->apWiData = apNew;
532: pWal->nWiData = iPage+1;
533: }
534:
535: /* Request a pointer to the required page from the VFS */
536: if( pWal->apWiData[iPage]==0 ){
537: if( pWal->exclusiveMode==WAL_HEAPMEMORY_MODE ){
538: pWal->apWiData[iPage] = (u32 volatile *)sqlite3MallocZero(WALINDEX_PGSZ);
539: if( !pWal->apWiData[iPage] ) rc = SQLITE_NOMEM;
540: }else{
541: rc = sqlite3OsShmMap(pWal->pDbFd, iPage, WALINDEX_PGSZ,
542: pWal->writeLock, (void volatile **)&pWal->apWiData[iPage]
543: );
544: if( rc==SQLITE_READONLY ){
545: pWal->readOnly |= WAL_SHM_RDONLY;
546: rc = SQLITE_OK;
547: }
548: }
549: }
550:
551: *ppPage = pWal->apWiData[iPage];
552: assert( iPage==0 || *ppPage || rc!=SQLITE_OK );
553: return rc;
554: }
555:
556: /*
557: ** Return a pointer to the WalCkptInfo structure in the wal-index.
558: */
559: static volatile WalCkptInfo *walCkptInfo(Wal *pWal){
560: assert( pWal->nWiData>0 && pWal->apWiData[0] );
561: return (volatile WalCkptInfo*)&(pWal->apWiData[0][sizeof(WalIndexHdr)/2]);
562: }
563:
564: /*
565: ** Return a pointer to the WalIndexHdr structure in the wal-index.
566: */
567: static volatile WalIndexHdr *walIndexHdr(Wal *pWal){
568: assert( pWal->nWiData>0 && pWal->apWiData[0] );
569: return (volatile WalIndexHdr*)pWal->apWiData[0];
570: }
571:
572: /*
573: ** The argument to this macro must be of type u32. On a little-endian
574: ** architecture, it returns the u32 value that results from interpreting
575: ** the 4 bytes as a big-endian value. On a big-endian architecture, it
576: ** returns the value that would be produced by intepreting the 4 bytes
577: ** of the input value as a little-endian integer.
578: */
579: #define BYTESWAP32(x) ( \
580: (((x)&0x000000FF)<<24) + (((x)&0x0000FF00)<<8) \
581: + (((x)&0x00FF0000)>>8) + (((x)&0xFF000000)>>24) \
582: )
583:
584: /*
585: ** Generate or extend an 8 byte checksum based on the data in
586: ** array aByte[] and the initial values of aIn[0] and aIn[1] (or
587: ** initial values of 0 and 0 if aIn==NULL).
588: **
589: ** The checksum is written back into aOut[] before returning.
590: **
591: ** nByte must be a positive multiple of 8.
592: */
593: static void walChecksumBytes(
594: int nativeCksum, /* True for native byte-order, false for non-native */
595: u8 *a, /* Content to be checksummed */
596: int nByte, /* Bytes of content in a[]. Must be a multiple of 8. */
597: const u32 *aIn, /* Initial checksum value input */
598: u32 *aOut /* OUT: Final checksum value output */
599: ){
600: u32 s1, s2;
601: u32 *aData = (u32 *)a;
602: u32 *aEnd = (u32 *)&a[nByte];
603:
604: if( aIn ){
605: s1 = aIn[0];
606: s2 = aIn[1];
607: }else{
608: s1 = s2 = 0;
609: }
610:
611: assert( nByte>=8 );
612: assert( (nByte&0x00000007)==0 );
613:
614: if( nativeCksum ){
615: do {
616: s1 += *aData++ + s2;
617: s2 += *aData++ + s1;
618: }while( aData<aEnd );
619: }else{
620: do {
621: s1 += BYTESWAP32(aData[0]) + s2;
622: s2 += BYTESWAP32(aData[1]) + s1;
623: aData += 2;
624: }while( aData<aEnd );
625: }
626:
627: aOut[0] = s1;
628: aOut[1] = s2;
629: }
630:
631: static void walShmBarrier(Wal *pWal){
632: if( pWal->exclusiveMode!=WAL_HEAPMEMORY_MODE ){
633: sqlite3OsShmBarrier(pWal->pDbFd);
634: }
635: }
636:
637: /*
638: ** Write the header information in pWal->hdr into the wal-index.
639: **
640: ** The checksum on pWal->hdr is updated before it is written.
641: */
642: static void walIndexWriteHdr(Wal *pWal){
643: volatile WalIndexHdr *aHdr = walIndexHdr(pWal);
644: const int nCksum = offsetof(WalIndexHdr, aCksum);
645:
646: assert( pWal->writeLock );
647: pWal->hdr.isInit = 1;
648: pWal->hdr.iVersion = WALINDEX_MAX_VERSION;
649: walChecksumBytes(1, (u8*)&pWal->hdr, nCksum, 0, pWal->hdr.aCksum);
650: memcpy((void *)&aHdr[1], (void *)&pWal->hdr, sizeof(WalIndexHdr));
651: walShmBarrier(pWal);
652: memcpy((void *)&aHdr[0], (void *)&pWal->hdr, sizeof(WalIndexHdr));
653: }
654:
655: /*
656: ** This function encodes a single frame header and writes it to a buffer
657: ** supplied by the caller. A frame-header is made up of a series of
658: ** 4-byte big-endian integers, as follows:
659: **
660: ** 0: Page number.
661: ** 4: For commit records, the size of the database image in pages
662: ** after the commit. For all other records, zero.
663: ** 8: Salt-1 (copied from the wal-header)
664: ** 12: Salt-2 (copied from the wal-header)
665: ** 16: Checksum-1.
666: ** 20: Checksum-2.
667: */
668: static void walEncodeFrame(
669: Wal *pWal, /* The write-ahead log */
670: u32 iPage, /* Database page number for frame */
671: u32 nTruncate, /* New db size (or 0 for non-commit frames) */
672: u8 *aData, /* Pointer to page data */
673: u8 *aFrame /* OUT: Write encoded frame here */
674: ){
675: int nativeCksum; /* True for native byte-order checksums */
676: u32 *aCksum = pWal->hdr.aFrameCksum;
677: assert( WAL_FRAME_HDRSIZE==24 );
678: sqlite3Put4byte(&aFrame[0], iPage);
679: sqlite3Put4byte(&aFrame[4], nTruncate);
680: memcpy(&aFrame[8], pWal->hdr.aSalt, 8);
681:
682: nativeCksum = (pWal->hdr.bigEndCksum==SQLITE_BIGENDIAN);
683: walChecksumBytes(nativeCksum, aFrame, 8, aCksum, aCksum);
684: walChecksumBytes(nativeCksum, aData, pWal->szPage, aCksum, aCksum);
685:
686: sqlite3Put4byte(&aFrame[16], aCksum[0]);
687: sqlite3Put4byte(&aFrame[20], aCksum[1]);
688: }
689:
690: /*
691: ** Check to see if the frame with header in aFrame[] and content
692: ** in aData[] is valid. If it is a valid frame, fill *piPage and
693: ** *pnTruncate and return true. Return if the frame is not valid.
694: */
695: static int walDecodeFrame(
696: Wal *pWal, /* The write-ahead log */
697: u32 *piPage, /* OUT: Database page number for frame */
698: u32 *pnTruncate, /* OUT: New db size (or 0 if not commit) */
699: u8 *aData, /* Pointer to page data (for checksum) */
700: u8 *aFrame /* Frame data */
701: ){
702: int nativeCksum; /* True for native byte-order checksums */
703: u32 *aCksum = pWal->hdr.aFrameCksum;
704: u32 pgno; /* Page number of the frame */
705: assert( WAL_FRAME_HDRSIZE==24 );
706:
707: /* A frame is only valid if the salt values in the frame-header
708: ** match the salt values in the wal-header.
709: */
710: if( memcmp(&pWal->hdr.aSalt, &aFrame[8], 8)!=0 ){
711: return 0;
712: }
713:
714: /* A frame is only valid if the page number is creater than zero.
715: */
716: pgno = sqlite3Get4byte(&aFrame[0]);
717: if( pgno==0 ){
718: return 0;
719: }
720:
721: /* A frame is only valid if a checksum of the WAL header,
722: ** all prior frams, the first 16 bytes of this frame-header,
723: ** and the frame-data matches the checksum in the last 8
724: ** bytes of this frame-header.
725: */
726: nativeCksum = (pWal->hdr.bigEndCksum==SQLITE_BIGENDIAN);
727: walChecksumBytes(nativeCksum, aFrame, 8, aCksum, aCksum);
728: walChecksumBytes(nativeCksum, aData, pWal->szPage, aCksum, aCksum);
729: if( aCksum[0]!=sqlite3Get4byte(&aFrame[16])
730: || aCksum[1]!=sqlite3Get4byte(&aFrame[20])
731: ){
732: /* Checksum failed. */
733: return 0;
734: }
735:
736: /* If we reach this point, the frame is valid. Return the page number
737: ** and the new database size.
738: */
739: *piPage = pgno;
740: *pnTruncate = sqlite3Get4byte(&aFrame[4]);
741: return 1;
742: }
743:
744:
745: #if defined(SQLITE_TEST) && defined(SQLITE_DEBUG)
746: /*
747: ** Names of locks. This routine is used to provide debugging output and is not
748: ** a part of an ordinary build.
749: */
750: static const char *walLockName(int lockIdx){
751: if( lockIdx==WAL_WRITE_LOCK ){
752: return "WRITE-LOCK";
753: }else if( lockIdx==WAL_CKPT_LOCK ){
754: return "CKPT-LOCK";
755: }else if( lockIdx==WAL_RECOVER_LOCK ){
756: return "RECOVER-LOCK";
757: }else{
758: static char zName[15];
759: sqlite3_snprintf(sizeof(zName), zName, "READ-LOCK[%d]",
760: lockIdx-WAL_READ_LOCK(0));
761: return zName;
762: }
763: }
764: #endif /*defined(SQLITE_TEST) || defined(SQLITE_DEBUG) */
765:
766:
767: /*
768: ** Set or release locks on the WAL. Locks are either shared or exclusive.
769: ** A lock cannot be moved directly between shared and exclusive - it must go
770: ** through the unlocked state first.
771: **
772: ** In locking_mode=EXCLUSIVE, all of these routines become no-ops.
773: */
774: static int walLockShared(Wal *pWal, int lockIdx){
775: int rc;
776: if( pWal->exclusiveMode ) return SQLITE_OK;
777: rc = sqlite3OsShmLock(pWal->pDbFd, lockIdx, 1,
778: SQLITE_SHM_LOCK | SQLITE_SHM_SHARED);
779: WALTRACE(("WAL%p: acquire SHARED-%s %s\n", pWal,
780: walLockName(lockIdx), rc ? "failed" : "ok"));
781: VVA_ONLY( pWal->lockError = (u8)(rc!=SQLITE_OK && rc!=SQLITE_BUSY); )
782: return rc;
783: }
784: static void walUnlockShared(Wal *pWal, int lockIdx){
785: if( pWal->exclusiveMode ) return;
786: (void)sqlite3OsShmLock(pWal->pDbFd, lockIdx, 1,
787: SQLITE_SHM_UNLOCK | SQLITE_SHM_SHARED);
788: WALTRACE(("WAL%p: release SHARED-%s\n", pWal, walLockName(lockIdx)));
789: }
790: static int walLockExclusive(Wal *pWal, int lockIdx, int n){
791: int rc;
792: if( pWal->exclusiveMode ) return SQLITE_OK;
793: rc = sqlite3OsShmLock(pWal->pDbFd, lockIdx, n,
794: SQLITE_SHM_LOCK | SQLITE_SHM_EXCLUSIVE);
795: WALTRACE(("WAL%p: acquire EXCLUSIVE-%s cnt=%d %s\n", pWal,
796: walLockName(lockIdx), n, rc ? "failed" : "ok"));
797: VVA_ONLY( pWal->lockError = (u8)(rc!=SQLITE_OK && rc!=SQLITE_BUSY); )
798: return rc;
799: }
800: static void walUnlockExclusive(Wal *pWal, int lockIdx, int n){
801: if( pWal->exclusiveMode ) return;
802: (void)sqlite3OsShmLock(pWal->pDbFd, lockIdx, n,
803: SQLITE_SHM_UNLOCK | SQLITE_SHM_EXCLUSIVE);
804: WALTRACE(("WAL%p: release EXCLUSIVE-%s cnt=%d\n", pWal,
805: walLockName(lockIdx), n));
806: }
807:
808: /*
809: ** Compute a hash on a page number. The resulting hash value must land
810: ** between 0 and (HASHTABLE_NSLOT-1). The walHashNext() function advances
811: ** the hash to the next value in the event of a collision.
812: */
813: static int walHash(u32 iPage){
814: assert( iPage>0 );
815: assert( (HASHTABLE_NSLOT & (HASHTABLE_NSLOT-1))==0 );
816: return (iPage*HASHTABLE_HASH_1) & (HASHTABLE_NSLOT-1);
817: }
818: static int walNextHash(int iPriorHash){
819: return (iPriorHash+1)&(HASHTABLE_NSLOT-1);
820: }
821:
822: /*
823: ** Return pointers to the hash table and page number array stored on
824: ** page iHash of the wal-index. The wal-index is broken into 32KB pages
825: ** numbered starting from 0.
826: **
827: ** Set output variable *paHash to point to the start of the hash table
828: ** in the wal-index file. Set *piZero to one less than the frame
829: ** number of the first frame indexed by this hash table. If a
830: ** slot in the hash table is set to N, it refers to frame number
831: ** (*piZero+N) in the log.
832: **
833: ** Finally, set *paPgno so that *paPgno[1] is the page number of the
834: ** first frame indexed by the hash table, frame (*piZero+1).
835: */
836: static int walHashGet(
837: Wal *pWal, /* WAL handle */
838: int iHash, /* Find the iHash'th table */
839: volatile ht_slot **paHash, /* OUT: Pointer to hash index */
840: volatile u32 **paPgno, /* OUT: Pointer to page number array */
841: u32 *piZero /* OUT: Frame associated with *paPgno[0] */
842: ){
843: int rc; /* Return code */
844: volatile u32 *aPgno;
845:
846: rc = walIndexPage(pWal, iHash, &aPgno);
847: assert( rc==SQLITE_OK || iHash>0 );
848:
849: if( rc==SQLITE_OK ){
850: u32 iZero;
851: volatile ht_slot *aHash;
852:
853: aHash = (volatile ht_slot *)&aPgno[HASHTABLE_NPAGE];
854: if( iHash==0 ){
855: aPgno = &aPgno[WALINDEX_HDR_SIZE/sizeof(u32)];
856: iZero = 0;
857: }else{
858: iZero = HASHTABLE_NPAGE_ONE + (iHash-1)*HASHTABLE_NPAGE;
859: }
860:
861: *paPgno = &aPgno[-1];
862: *paHash = aHash;
863: *piZero = iZero;
864: }
865: return rc;
866: }
867:
868: /*
869: ** Return the number of the wal-index page that contains the hash-table
870: ** and page-number array that contain entries corresponding to WAL frame
871: ** iFrame. The wal-index is broken up into 32KB pages. Wal-index pages
872: ** are numbered starting from 0.
873: */
874: static int walFramePage(u32 iFrame){
875: int iHash = (iFrame+HASHTABLE_NPAGE-HASHTABLE_NPAGE_ONE-1) / HASHTABLE_NPAGE;
876: assert( (iHash==0 || iFrame>HASHTABLE_NPAGE_ONE)
877: && (iHash>=1 || iFrame<=HASHTABLE_NPAGE_ONE)
878: && (iHash<=1 || iFrame>(HASHTABLE_NPAGE_ONE+HASHTABLE_NPAGE))
879: && (iHash>=2 || iFrame<=HASHTABLE_NPAGE_ONE+HASHTABLE_NPAGE)
880: && (iHash<=2 || iFrame>(HASHTABLE_NPAGE_ONE+2*HASHTABLE_NPAGE))
881: );
882: return iHash;
883: }
884:
885: /*
886: ** Return the page number associated with frame iFrame in this WAL.
887: */
888: static u32 walFramePgno(Wal *pWal, u32 iFrame){
889: int iHash = walFramePage(iFrame);
890: if( iHash==0 ){
891: return pWal->apWiData[0][WALINDEX_HDR_SIZE/sizeof(u32) + iFrame - 1];
892: }
893: return pWal->apWiData[iHash][(iFrame-1-HASHTABLE_NPAGE_ONE)%HASHTABLE_NPAGE];
894: }
895:
896: /*
897: ** Remove entries from the hash table that point to WAL slots greater
898: ** than pWal->hdr.mxFrame.
899: **
900: ** This function is called whenever pWal->hdr.mxFrame is decreased due
901: ** to a rollback or savepoint.
902: **
903: ** At most only the hash table containing pWal->hdr.mxFrame needs to be
904: ** updated. Any later hash tables will be automatically cleared when
905: ** pWal->hdr.mxFrame advances to the point where those hash tables are
906: ** actually needed.
907: */
908: static void walCleanupHash(Wal *pWal){
909: volatile ht_slot *aHash = 0; /* Pointer to hash table to clear */
910: volatile u32 *aPgno = 0; /* Page number array for hash table */
911: u32 iZero = 0; /* frame == (aHash[x]+iZero) */
912: int iLimit = 0; /* Zero values greater than this */
913: int nByte; /* Number of bytes to zero in aPgno[] */
914: int i; /* Used to iterate through aHash[] */
915:
916: assert( pWal->writeLock );
917: testcase( pWal->hdr.mxFrame==HASHTABLE_NPAGE_ONE-1 );
918: testcase( pWal->hdr.mxFrame==HASHTABLE_NPAGE_ONE );
919: testcase( pWal->hdr.mxFrame==HASHTABLE_NPAGE_ONE+1 );
920:
921: if( pWal->hdr.mxFrame==0 ) return;
922:
923: /* Obtain pointers to the hash-table and page-number array containing
924: ** the entry that corresponds to frame pWal->hdr.mxFrame. It is guaranteed
925: ** that the page said hash-table and array reside on is already mapped.
926: */
927: assert( pWal->nWiData>walFramePage(pWal->hdr.mxFrame) );
928: assert( pWal->apWiData[walFramePage(pWal->hdr.mxFrame)] );
929: walHashGet(pWal, walFramePage(pWal->hdr.mxFrame), &aHash, &aPgno, &iZero);
930:
931: /* Zero all hash-table entries that correspond to frame numbers greater
932: ** than pWal->hdr.mxFrame.
933: */
934: iLimit = pWal->hdr.mxFrame - iZero;
935: assert( iLimit>0 );
936: for(i=0; i<HASHTABLE_NSLOT; i++){
937: if( aHash[i]>iLimit ){
938: aHash[i] = 0;
939: }
940: }
941:
942: /* Zero the entries in the aPgno array that correspond to frames with
943: ** frame numbers greater than pWal->hdr.mxFrame.
944: */
945: nByte = (int)((char *)aHash - (char *)&aPgno[iLimit+1]);
946: memset((void *)&aPgno[iLimit+1], 0, nByte);
947:
948: #ifdef SQLITE_ENABLE_EXPENSIVE_ASSERT
949: /* Verify that the every entry in the mapping region is still reachable
950: ** via the hash table even after the cleanup.
951: */
952: if( iLimit ){
953: int i; /* Loop counter */
954: int iKey; /* Hash key */
955: for(i=1; i<=iLimit; i++){
956: for(iKey=walHash(aPgno[i]); aHash[iKey]; iKey=walNextHash(iKey)){
957: if( aHash[iKey]==i ) break;
958: }
959: assert( aHash[iKey]==i );
960: }
961: }
962: #endif /* SQLITE_ENABLE_EXPENSIVE_ASSERT */
963: }
964:
965:
966: /*
967: ** Set an entry in the wal-index that will map database page number
968: ** pPage into WAL frame iFrame.
969: */
970: static int walIndexAppend(Wal *pWal, u32 iFrame, u32 iPage){
971: int rc; /* Return code */
972: u32 iZero = 0; /* One less than frame number of aPgno[1] */
973: volatile u32 *aPgno = 0; /* Page number array */
974: volatile ht_slot *aHash = 0; /* Hash table */
975:
976: rc = walHashGet(pWal, walFramePage(iFrame), &aHash, &aPgno, &iZero);
977:
978: /* Assuming the wal-index file was successfully mapped, populate the
979: ** page number array and hash table entry.
980: */
981: if( rc==SQLITE_OK ){
982: int iKey; /* Hash table key */
983: int idx; /* Value to write to hash-table slot */
984: int nCollide; /* Number of hash collisions */
985:
986: idx = iFrame - iZero;
987: assert( idx <= HASHTABLE_NSLOT/2 + 1 );
988:
989: /* If this is the first entry to be added to this hash-table, zero the
990: ** entire hash table and aPgno[] array before proceding.
991: */
992: if( idx==1 ){
993: int nByte = (int)((u8 *)&aHash[HASHTABLE_NSLOT] - (u8 *)&aPgno[1]);
994: memset((void*)&aPgno[1], 0, nByte);
995: }
996:
997: /* If the entry in aPgno[] is already set, then the previous writer
998: ** must have exited unexpectedly in the middle of a transaction (after
999: ** writing one or more dirty pages to the WAL to free up memory).
1000: ** Remove the remnants of that writers uncommitted transaction from
1001: ** the hash-table before writing any new entries.
1002: */
1003: if( aPgno[idx] ){
1004: walCleanupHash(pWal);
1005: assert( !aPgno[idx] );
1006: }
1007:
1008: /* Write the aPgno[] array entry and the hash-table slot. */
1009: nCollide = idx;
1010: for(iKey=walHash(iPage); aHash[iKey]; iKey=walNextHash(iKey)){
1011: if( (nCollide--)==0 ) return SQLITE_CORRUPT_BKPT;
1012: }
1013: aPgno[idx] = iPage;
1014: aHash[iKey] = (ht_slot)idx;
1015:
1016: #ifdef SQLITE_ENABLE_EXPENSIVE_ASSERT
1017: /* Verify that the number of entries in the hash table exactly equals
1018: ** the number of entries in the mapping region.
1019: */
1020: {
1021: int i; /* Loop counter */
1022: int nEntry = 0; /* Number of entries in the hash table */
1023: for(i=0; i<HASHTABLE_NSLOT; i++){ if( aHash[i] ) nEntry++; }
1024: assert( nEntry==idx );
1025: }
1026:
1027: /* Verify that the every entry in the mapping region is reachable
1028: ** via the hash table. This turns out to be a really, really expensive
1029: ** thing to check, so only do this occasionally - not on every
1030: ** iteration.
1031: */
1032: if( (idx&0x3ff)==0 ){
1033: int i; /* Loop counter */
1034: for(i=1; i<=idx; i++){
1035: for(iKey=walHash(aPgno[i]); aHash[iKey]; iKey=walNextHash(iKey)){
1036: if( aHash[iKey]==i ) break;
1037: }
1038: assert( aHash[iKey]==i );
1039: }
1040: }
1041: #endif /* SQLITE_ENABLE_EXPENSIVE_ASSERT */
1042: }
1043:
1044:
1045: return rc;
1046: }
1047:
1048:
1049: /*
1050: ** Recover the wal-index by reading the write-ahead log file.
1051: **
1052: ** This routine first tries to establish an exclusive lock on the
1053: ** wal-index to prevent other threads/processes from doing anything
1054: ** with the WAL or wal-index while recovery is running. The
1055: ** WAL_RECOVER_LOCK is also held so that other threads will know
1056: ** that this thread is running recovery. If unable to establish
1057: ** the necessary locks, this routine returns SQLITE_BUSY.
1058: */
1059: static int walIndexRecover(Wal *pWal){
1060: int rc; /* Return Code */
1061: i64 nSize; /* Size of log file */
1062: u32 aFrameCksum[2] = {0, 0};
1063: int iLock; /* Lock offset to lock for checkpoint */
1064: int nLock; /* Number of locks to hold */
1065:
1066: /* Obtain an exclusive lock on all byte in the locking range not already
1067: ** locked by the caller. The caller is guaranteed to have locked the
1068: ** WAL_WRITE_LOCK byte, and may have also locked the WAL_CKPT_LOCK byte.
1069: ** If successful, the same bytes that are locked here are unlocked before
1070: ** this function returns.
1071: */
1072: assert( pWal->ckptLock==1 || pWal->ckptLock==0 );
1073: assert( WAL_ALL_BUT_WRITE==WAL_WRITE_LOCK+1 );
1074: assert( WAL_CKPT_LOCK==WAL_ALL_BUT_WRITE );
1075: assert( pWal->writeLock );
1076: iLock = WAL_ALL_BUT_WRITE + pWal->ckptLock;
1077: nLock = SQLITE_SHM_NLOCK - iLock;
1078: rc = walLockExclusive(pWal, iLock, nLock);
1079: if( rc ){
1080: return rc;
1081: }
1082: WALTRACE(("WAL%p: recovery begin...\n", pWal));
1083:
1084: memset(&pWal->hdr, 0, sizeof(WalIndexHdr));
1085:
1086: rc = sqlite3OsFileSize(pWal->pWalFd, &nSize);
1087: if( rc!=SQLITE_OK ){
1088: goto recovery_error;
1089: }
1090:
1091: if( nSize>WAL_HDRSIZE ){
1092: u8 aBuf[WAL_HDRSIZE]; /* Buffer to load WAL header into */
1093: u8 *aFrame = 0; /* Malloc'd buffer to load entire frame */
1094: int szFrame; /* Number of bytes in buffer aFrame[] */
1095: u8 *aData; /* Pointer to data part of aFrame buffer */
1096: int iFrame; /* Index of last frame read */
1097: i64 iOffset; /* Next offset to read from log file */
1098: int szPage; /* Page size according to the log */
1099: u32 magic; /* Magic value read from WAL header */
1100: u32 version; /* Magic value read from WAL header */
1101: int isValid; /* True if this frame is valid */
1102:
1103: /* Read in the WAL header. */
1104: rc = sqlite3OsRead(pWal->pWalFd, aBuf, WAL_HDRSIZE, 0);
1105: if( rc!=SQLITE_OK ){
1106: goto recovery_error;
1107: }
1108:
1109: /* If the database page size is not a power of two, or is greater than
1110: ** SQLITE_MAX_PAGE_SIZE, conclude that the WAL file contains no valid
1111: ** data. Similarly, if the 'magic' value is invalid, ignore the whole
1112: ** WAL file.
1113: */
1114: magic = sqlite3Get4byte(&aBuf[0]);
1115: szPage = sqlite3Get4byte(&aBuf[8]);
1116: if( (magic&0xFFFFFFFE)!=WAL_MAGIC
1117: || szPage&(szPage-1)
1118: || szPage>SQLITE_MAX_PAGE_SIZE
1119: || szPage<512
1120: ){
1121: goto finished;
1122: }
1123: pWal->hdr.bigEndCksum = (u8)(magic&0x00000001);
1124: pWal->szPage = szPage;
1125: pWal->nCkpt = sqlite3Get4byte(&aBuf[12]);
1126: memcpy(&pWal->hdr.aSalt, &aBuf[16], 8);
1127:
1128: /* Verify that the WAL header checksum is correct */
1129: walChecksumBytes(pWal->hdr.bigEndCksum==SQLITE_BIGENDIAN,
1130: aBuf, WAL_HDRSIZE-2*4, 0, pWal->hdr.aFrameCksum
1131: );
1132: if( pWal->hdr.aFrameCksum[0]!=sqlite3Get4byte(&aBuf[24])
1133: || pWal->hdr.aFrameCksum[1]!=sqlite3Get4byte(&aBuf[28])
1134: ){
1135: goto finished;
1136: }
1137:
1138: /* Verify that the version number on the WAL format is one that
1139: ** are able to understand */
1140: version = sqlite3Get4byte(&aBuf[4]);
1141: if( version!=WAL_MAX_VERSION ){
1142: rc = SQLITE_CANTOPEN_BKPT;
1143: goto finished;
1144: }
1145:
1146: /* Malloc a buffer to read frames into. */
1147: szFrame = szPage + WAL_FRAME_HDRSIZE;
1148: aFrame = (u8 *)sqlite3_malloc(szFrame);
1149: if( !aFrame ){
1150: rc = SQLITE_NOMEM;
1151: goto recovery_error;
1152: }
1153: aData = &aFrame[WAL_FRAME_HDRSIZE];
1154:
1155: /* Read all frames from the log file. */
1156: iFrame = 0;
1157: for(iOffset=WAL_HDRSIZE; (iOffset+szFrame)<=nSize; iOffset+=szFrame){
1158: u32 pgno; /* Database page number for frame */
1159: u32 nTruncate; /* dbsize field from frame header */
1160:
1161: /* Read and decode the next log frame. */
1162: iFrame++;
1163: rc = sqlite3OsRead(pWal->pWalFd, aFrame, szFrame, iOffset);
1164: if( rc!=SQLITE_OK ) break;
1165: isValid = walDecodeFrame(pWal, &pgno, &nTruncate, aData, aFrame);
1166: if( !isValid ) break;
1167: rc = walIndexAppend(pWal, iFrame, pgno);
1168: if( rc!=SQLITE_OK ) break;
1169:
1170: /* If nTruncate is non-zero, this is a commit record. */
1171: if( nTruncate ){
1172: pWal->hdr.mxFrame = iFrame;
1173: pWal->hdr.nPage = nTruncate;
1174: pWal->hdr.szPage = (u16)((szPage&0xff00) | (szPage>>16));
1175: testcase( szPage<=32768 );
1176: testcase( szPage>=65536 );
1177: aFrameCksum[0] = pWal->hdr.aFrameCksum[0];
1178: aFrameCksum[1] = pWal->hdr.aFrameCksum[1];
1179: }
1180: }
1181:
1182: sqlite3_free(aFrame);
1183: }
1184:
1185: finished:
1186: if( rc==SQLITE_OK ){
1187: volatile WalCkptInfo *pInfo;
1188: int i;
1189: pWal->hdr.aFrameCksum[0] = aFrameCksum[0];
1190: pWal->hdr.aFrameCksum[1] = aFrameCksum[1];
1191: walIndexWriteHdr(pWal);
1192:
1193: /* Reset the checkpoint-header. This is safe because this thread is
1194: ** currently holding locks that exclude all other readers, writers and
1195: ** checkpointers.
1196: */
1197: pInfo = walCkptInfo(pWal);
1198: pInfo->nBackfill = 0;
1199: pInfo->aReadMark[0] = 0;
1200: for(i=1; i<WAL_NREADER; i++) pInfo->aReadMark[i] = READMARK_NOT_USED;
1201:
1202: /* If more than one frame was recovered from the log file, report an
1203: ** event via sqlite3_log(). This is to help with identifying performance
1204: ** problems caused by applications routinely shutting down without
1205: ** checkpointing the log file.
1206: */
1207: if( pWal->hdr.nPage ){
1208: sqlite3_log(SQLITE_OK, "Recovered %d frames from WAL file %s",
1209: pWal->hdr.nPage, pWal->zWalName
1210: );
1211: }
1212: }
1213:
1214: recovery_error:
1215: WALTRACE(("WAL%p: recovery %s\n", pWal, rc ? "failed" : "ok"));
1216: walUnlockExclusive(pWal, iLock, nLock);
1217: return rc;
1218: }
1219:
1220: /*
1221: ** Close an open wal-index.
1222: */
1223: static void walIndexClose(Wal *pWal, int isDelete){
1224: if( pWal->exclusiveMode==WAL_HEAPMEMORY_MODE ){
1225: int i;
1226: for(i=0; i<pWal->nWiData; i++){
1227: sqlite3_free((void *)pWal->apWiData[i]);
1228: pWal->apWiData[i] = 0;
1229: }
1230: }else{
1231: sqlite3OsShmUnmap(pWal->pDbFd, isDelete);
1232: }
1233: }
1234:
1235: /*
1236: ** Open a connection to the WAL file zWalName. The database file must
1237: ** already be opened on connection pDbFd. The buffer that zWalName points
1238: ** to must remain valid for the lifetime of the returned Wal* handle.
1239: **
1240: ** A SHARED lock should be held on the database file when this function
1241: ** is called. The purpose of this SHARED lock is to prevent any other
1242: ** client from unlinking the WAL or wal-index file. If another process
1243: ** were to do this just after this client opened one of these files, the
1244: ** system would be badly broken.
1245: **
1246: ** If the log file is successfully opened, SQLITE_OK is returned and
1247: ** *ppWal is set to point to a new WAL handle. If an error occurs,
1248: ** an SQLite error code is returned and *ppWal is left unmodified.
1249: */
1250: int sqlite3WalOpen(
1251: sqlite3_vfs *pVfs, /* vfs module to open wal and wal-index */
1252: sqlite3_file *pDbFd, /* The open database file */
1253: const char *zWalName, /* Name of the WAL file */
1254: int bNoShm, /* True to run in heap-memory mode */
1255: i64 mxWalSize, /* Truncate WAL to this size on reset */
1256: Wal **ppWal /* OUT: Allocated Wal handle */
1257: ){
1258: int rc; /* Return Code */
1259: Wal *pRet; /* Object to allocate and return */
1260: int flags; /* Flags passed to OsOpen() */
1261:
1262: assert( zWalName && zWalName[0] );
1263: assert( pDbFd );
1264:
1265: /* In the amalgamation, the os_unix.c and os_win.c source files come before
1266: ** this source file. Verify that the #defines of the locking byte offsets
1267: ** in os_unix.c and os_win.c agree with the WALINDEX_LOCK_OFFSET value.
1268: */
1269: #ifdef WIN_SHM_BASE
1270: assert( WIN_SHM_BASE==WALINDEX_LOCK_OFFSET );
1271: #endif
1272: #ifdef UNIX_SHM_BASE
1273: assert( UNIX_SHM_BASE==WALINDEX_LOCK_OFFSET );
1274: #endif
1275:
1276:
1277: /* Allocate an instance of struct Wal to return. */
1278: *ppWal = 0;
1279: pRet = (Wal*)sqlite3MallocZero(sizeof(Wal) + pVfs->szOsFile);
1280: if( !pRet ){
1281: return SQLITE_NOMEM;
1282: }
1283:
1284: pRet->pVfs = pVfs;
1285: pRet->pWalFd = (sqlite3_file *)&pRet[1];
1286: pRet->pDbFd = pDbFd;
1287: pRet->readLock = -1;
1288: pRet->mxWalSize = mxWalSize;
1289: pRet->zWalName = zWalName;
1290: pRet->syncHeader = 1;
1291: pRet->padToSectorBoundary = 1;
1292: pRet->exclusiveMode = (bNoShm ? WAL_HEAPMEMORY_MODE: WAL_NORMAL_MODE);
1293:
1294: /* Open file handle on the write-ahead log file. */
1295: flags = (SQLITE_OPEN_READWRITE|SQLITE_OPEN_CREATE|SQLITE_OPEN_WAL);
1296: rc = sqlite3OsOpen(pVfs, zWalName, pRet->pWalFd, flags, &flags);
1297: if( rc==SQLITE_OK && flags&SQLITE_OPEN_READONLY ){
1298: pRet->readOnly = WAL_RDONLY;
1299: }
1300:
1301: if( rc!=SQLITE_OK ){
1302: walIndexClose(pRet, 0);
1303: sqlite3OsClose(pRet->pWalFd);
1304: sqlite3_free(pRet);
1305: }else{
1306: int iDC = sqlite3OsDeviceCharacteristics(pRet->pWalFd);
1307: if( iDC & SQLITE_IOCAP_SEQUENTIAL ){ pRet->syncHeader = 0; }
1308: if( iDC & SQLITE_IOCAP_POWERSAFE_OVERWRITE ){
1309: pRet->padToSectorBoundary = 0;
1310: }
1311: *ppWal = pRet;
1312: WALTRACE(("WAL%d: opened\n", pRet));
1313: }
1314: return rc;
1315: }
1316:
1317: /*
1318: ** Change the size to which the WAL file is trucated on each reset.
1319: */
1320: void sqlite3WalLimit(Wal *pWal, i64 iLimit){
1321: if( pWal ) pWal->mxWalSize = iLimit;
1322: }
1323:
1324: /*
1325: ** Find the smallest page number out of all pages held in the WAL that
1326: ** has not been returned by any prior invocation of this method on the
1327: ** same WalIterator object. Write into *piFrame the frame index where
1328: ** that page was last written into the WAL. Write into *piPage the page
1329: ** number.
1330: **
1331: ** Return 0 on success. If there are no pages in the WAL with a page
1332: ** number larger than *piPage, then return 1.
1333: */
1334: static int walIteratorNext(
1335: WalIterator *p, /* Iterator */
1336: u32 *piPage, /* OUT: The page number of the next page */
1337: u32 *piFrame /* OUT: Wal frame index of next page */
1338: ){
1339: u32 iMin; /* Result pgno must be greater than iMin */
1340: u32 iRet = 0xFFFFFFFF; /* 0xffffffff is never a valid page number */
1341: int i; /* For looping through segments */
1342:
1343: iMin = p->iPrior;
1344: assert( iMin<0xffffffff );
1345: for(i=p->nSegment-1; i>=0; i--){
1346: struct WalSegment *pSegment = &p->aSegment[i];
1347: while( pSegment->iNext<pSegment->nEntry ){
1348: u32 iPg = pSegment->aPgno[pSegment->aIndex[pSegment->iNext]];
1349: if( iPg>iMin ){
1350: if( iPg<iRet ){
1351: iRet = iPg;
1352: *piFrame = pSegment->iZero + pSegment->aIndex[pSegment->iNext];
1353: }
1354: break;
1355: }
1356: pSegment->iNext++;
1357: }
1358: }
1359:
1360: *piPage = p->iPrior = iRet;
1361: return (iRet==0xFFFFFFFF);
1362: }
1363:
1364: /*
1365: ** This function merges two sorted lists into a single sorted list.
1366: **
1367: ** aLeft[] and aRight[] are arrays of indices. The sort key is
1368: ** aContent[aLeft[]] and aContent[aRight[]]. Upon entry, the following
1369: ** is guaranteed for all J<K:
1370: **
1371: ** aContent[aLeft[J]] < aContent[aLeft[K]]
1372: ** aContent[aRight[J]] < aContent[aRight[K]]
1373: **
1374: ** This routine overwrites aRight[] with a new (probably longer) sequence
1375: ** of indices such that the aRight[] contains every index that appears in
1376: ** either aLeft[] or the old aRight[] and such that the second condition
1377: ** above is still met.
1378: **
1379: ** The aContent[aLeft[X]] values will be unique for all X. And the
1380: ** aContent[aRight[X]] values will be unique too. But there might be
1381: ** one or more combinations of X and Y such that
1382: **
1383: ** aLeft[X]!=aRight[Y] && aContent[aLeft[X]] == aContent[aRight[Y]]
1384: **
1385: ** When that happens, omit the aLeft[X] and use the aRight[Y] index.
1386: */
1387: static void walMerge(
1388: const u32 *aContent, /* Pages in wal - keys for the sort */
1389: ht_slot *aLeft, /* IN: Left hand input list */
1390: int nLeft, /* IN: Elements in array *paLeft */
1391: ht_slot **paRight, /* IN/OUT: Right hand input list */
1392: int *pnRight, /* IN/OUT: Elements in *paRight */
1393: ht_slot *aTmp /* Temporary buffer */
1394: ){
1395: int iLeft = 0; /* Current index in aLeft */
1396: int iRight = 0; /* Current index in aRight */
1397: int iOut = 0; /* Current index in output buffer */
1398: int nRight = *pnRight;
1399: ht_slot *aRight = *paRight;
1400:
1401: assert( nLeft>0 && nRight>0 );
1402: while( iRight<nRight || iLeft<nLeft ){
1403: ht_slot logpage;
1404: Pgno dbpage;
1405:
1406: if( (iLeft<nLeft)
1407: && (iRight>=nRight || aContent[aLeft[iLeft]]<aContent[aRight[iRight]])
1408: ){
1409: logpage = aLeft[iLeft++];
1410: }else{
1411: logpage = aRight[iRight++];
1412: }
1413: dbpage = aContent[logpage];
1414:
1415: aTmp[iOut++] = logpage;
1416: if( iLeft<nLeft && aContent[aLeft[iLeft]]==dbpage ) iLeft++;
1417:
1418: assert( iLeft>=nLeft || aContent[aLeft[iLeft]]>dbpage );
1419: assert( iRight>=nRight || aContent[aRight[iRight]]>dbpage );
1420: }
1421:
1422: *paRight = aLeft;
1423: *pnRight = iOut;
1424: memcpy(aLeft, aTmp, sizeof(aTmp[0])*iOut);
1425: }
1426:
1427: /*
1428: ** Sort the elements in list aList using aContent[] as the sort key.
1429: ** Remove elements with duplicate keys, preferring to keep the
1430: ** larger aList[] values.
1431: **
1432: ** The aList[] entries are indices into aContent[]. The values in
1433: ** aList[] are to be sorted so that for all J<K:
1434: **
1435: ** aContent[aList[J]] < aContent[aList[K]]
1436: **
1437: ** For any X and Y such that
1438: **
1439: ** aContent[aList[X]] == aContent[aList[Y]]
1440: **
1441: ** Keep the larger of the two values aList[X] and aList[Y] and discard
1442: ** the smaller.
1443: */
1444: static void walMergesort(
1445: const u32 *aContent, /* Pages in wal */
1446: ht_slot *aBuffer, /* Buffer of at least *pnList items to use */
1447: ht_slot *aList, /* IN/OUT: List to sort */
1448: int *pnList /* IN/OUT: Number of elements in aList[] */
1449: ){
1450: struct Sublist {
1451: int nList; /* Number of elements in aList */
1452: ht_slot *aList; /* Pointer to sub-list content */
1453: };
1454:
1455: const int nList = *pnList; /* Size of input list */
1456: int nMerge = 0; /* Number of elements in list aMerge */
1457: ht_slot *aMerge = 0; /* List to be merged */
1458: int iList; /* Index into input list */
1459: int iSub = 0; /* Index into aSub array */
1460: struct Sublist aSub[13]; /* Array of sub-lists */
1461:
1462: memset(aSub, 0, sizeof(aSub));
1463: assert( nList<=HASHTABLE_NPAGE && nList>0 );
1464: assert( HASHTABLE_NPAGE==(1<<(ArraySize(aSub)-1)) );
1465:
1466: for(iList=0; iList<nList; iList++){
1467: nMerge = 1;
1468: aMerge = &aList[iList];
1469: for(iSub=0; iList & (1<<iSub); iSub++){
1470: struct Sublist *p = &aSub[iSub];
1471: assert( p->aList && p->nList<=(1<<iSub) );
1472: assert( p->aList==&aList[iList&~((2<<iSub)-1)] );
1473: walMerge(aContent, p->aList, p->nList, &aMerge, &nMerge, aBuffer);
1474: }
1475: aSub[iSub].aList = aMerge;
1476: aSub[iSub].nList = nMerge;
1477: }
1478:
1479: for(iSub++; iSub<ArraySize(aSub); iSub++){
1480: if( nList & (1<<iSub) ){
1481: struct Sublist *p = &aSub[iSub];
1482: assert( p->nList<=(1<<iSub) );
1483: assert( p->aList==&aList[nList&~((2<<iSub)-1)] );
1484: walMerge(aContent, p->aList, p->nList, &aMerge, &nMerge, aBuffer);
1485: }
1486: }
1487: assert( aMerge==aList );
1488: *pnList = nMerge;
1489:
1490: #ifdef SQLITE_DEBUG
1491: {
1492: int i;
1493: for(i=1; i<*pnList; i++){
1494: assert( aContent[aList[i]] > aContent[aList[i-1]] );
1495: }
1496: }
1497: #endif
1498: }
1499:
1500: /*
1501: ** Free an iterator allocated by walIteratorInit().
1502: */
1503: static void walIteratorFree(WalIterator *p){
1504: sqlite3ScratchFree(p);
1505: }
1506:
1507: /*
1508: ** Construct a WalInterator object that can be used to loop over all
1509: ** pages in the WAL in ascending order. The caller must hold the checkpoint
1510: ** lock.
1511: **
1512: ** On success, make *pp point to the newly allocated WalInterator object
1513: ** return SQLITE_OK. Otherwise, return an error code. If this routine
1514: ** returns an error, the value of *pp is undefined.
1515: **
1516: ** The calling routine should invoke walIteratorFree() to destroy the
1517: ** WalIterator object when it has finished with it.
1518: */
1519: static int walIteratorInit(Wal *pWal, WalIterator **pp){
1520: WalIterator *p; /* Return value */
1521: int nSegment; /* Number of segments to merge */
1522: u32 iLast; /* Last frame in log */
1523: int nByte; /* Number of bytes to allocate */
1524: int i; /* Iterator variable */
1525: ht_slot *aTmp; /* Temp space used by merge-sort */
1526: int rc = SQLITE_OK; /* Return Code */
1527:
1528: /* This routine only runs while holding the checkpoint lock. And
1529: ** it only runs if there is actually content in the log (mxFrame>0).
1530: */
1531: assert( pWal->ckptLock && pWal->hdr.mxFrame>0 );
1532: iLast = pWal->hdr.mxFrame;
1533:
1534: /* Allocate space for the WalIterator object. */
1535: nSegment = walFramePage(iLast) + 1;
1536: nByte = sizeof(WalIterator)
1537: + (nSegment-1)*sizeof(struct WalSegment)
1538: + iLast*sizeof(ht_slot);
1539: p = (WalIterator *)sqlite3ScratchMalloc(nByte);
1540: if( !p ){
1541: return SQLITE_NOMEM;
1542: }
1543: memset(p, 0, nByte);
1544: p->nSegment = nSegment;
1545:
1546: /* Allocate temporary space used by the merge-sort routine. This block
1547: ** of memory will be freed before this function returns.
1548: */
1549: aTmp = (ht_slot *)sqlite3ScratchMalloc(
1550: sizeof(ht_slot) * (iLast>HASHTABLE_NPAGE?HASHTABLE_NPAGE:iLast)
1551: );
1552: if( !aTmp ){
1553: rc = SQLITE_NOMEM;
1554: }
1555:
1556: for(i=0; rc==SQLITE_OK && i<nSegment; i++){
1557: volatile ht_slot *aHash;
1558: u32 iZero;
1559: volatile u32 *aPgno;
1560:
1561: rc = walHashGet(pWal, i, &aHash, &aPgno, &iZero);
1562: if( rc==SQLITE_OK ){
1563: int j; /* Counter variable */
1564: int nEntry; /* Number of entries in this segment */
1565: ht_slot *aIndex; /* Sorted index for this segment */
1566:
1567: aPgno++;
1568: if( (i+1)==nSegment ){
1569: nEntry = (int)(iLast - iZero);
1570: }else{
1571: nEntry = (int)((u32*)aHash - (u32*)aPgno);
1572: }
1573: aIndex = &((ht_slot *)&p->aSegment[p->nSegment])[iZero];
1574: iZero++;
1575:
1576: for(j=0; j<nEntry; j++){
1577: aIndex[j] = (ht_slot)j;
1578: }
1579: walMergesort((u32 *)aPgno, aTmp, aIndex, &nEntry);
1580: p->aSegment[i].iZero = iZero;
1581: p->aSegment[i].nEntry = nEntry;
1582: p->aSegment[i].aIndex = aIndex;
1583: p->aSegment[i].aPgno = (u32 *)aPgno;
1584: }
1585: }
1586: sqlite3ScratchFree(aTmp);
1587:
1588: if( rc!=SQLITE_OK ){
1589: walIteratorFree(p);
1590: }
1591: *pp = p;
1592: return rc;
1593: }
1594:
1595: /*
1596: ** Attempt to obtain the exclusive WAL lock defined by parameters lockIdx and
1597: ** n. If the attempt fails and parameter xBusy is not NULL, then it is a
1598: ** busy-handler function. Invoke it and retry the lock until either the
1599: ** lock is successfully obtained or the busy-handler returns 0.
1600: */
1601: static int walBusyLock(
1602: Wal *pWal, /* WAL connection */
1603: int (*xBusy)(void*), /* Function to call when busy */
1604: void *pBusyArg, /* Context argument for xBusyHandler */
1605: int lockIdx, /* Offset of first byte to lock */
1606: int n /* Number of bytes to lock */
1607: ){
1608: int rc;
1609: do {
1610: rc = walLockExclusive(pWal, lockIdx, n);
1611: }while( xBusy && rc==SQLITE_BUSY && xBusy(pBusyArg) );
1612: return rc;
1613: }
1614:
1615: /*
1616: ** The cache of the wal-index header must be valid to call this function.
1617: ** Return the page-size in bytes used by the database.
1618: */
1619: static int walPagesize(Wal *pWal){
1620: return (pWal->hdr.szPage&0xfe00) + ((pWal->hdr.szPage&0x0001)<<16);
1621: }
1622:
1623: /*
1624: ** Copy as much content as we can from the WAL back into the database file
1625: ** in response to an sqlite3_wal_checkpoint() request or the equivalent.
1626: **
1627: ** The amount of information copies from WAL to database might be limited
1628: ** by active readers. This routine will never overwrite a database page
1629: ** that a concurrent reader might be using.
1630: **
1631: ** All I/O barrier operations (a.k.a fsyncs) occur in this routine when
1632: ** SQLite is in WAL-mode in synchronous=NORMAL. That means that if
1633: ** checkpoints are always run by a background thread or background
1634: ** process, foreground threads will never block on a lengthy fsync call.
1635: **
1636: ** Fsync is called on the WAL before writing content out of the WAL and
1637: ** into the database. This ensures that if the new content is persistent
1638: ** in the WAL and can be recovered following a power-loss or hard reset.
1639: **
1640: ** Fsync is also called on the database file if (and only if) the entire
1641: ** WAL content is copied into the database file. This second fsync makes
1642: ** it safe to delete the WAL since the new content will persist in the
1643: ** database file.
1644: **
1645: ** This routine uses and updates the nBackfill field of the wal-index header.
1646: ** This is the only routine tha will increase the value of nBackfill.
1647: ** (A WAL reset or recovery will revert nBackfill to zero, but not increase
1648: ** its value.)
1649: **
1650: ** The caller must be holding sufficient locks to ensure that no other
1651: ** checkpoint is running (in any other thread or process) at the same
1652: ** time.
1653: */
1654: static int walCheckpoint(
1655: Wal *pWal, /* Wal connection */
1656: int eMode, /* One of PASSIVE, FULL or RESTART */
1657: int (*xBusyCall)(void*), /* Function to call when busy */
1658: void *pBusyArg, /* Context argument for xBusyHandler */
1659: int sync_flags, /* Flags for OsSync() (or 0) */
1660: u8 *zBuf /* Temporary buffer to use */
1661: ){
1662: int rc; /* Return code */
1663: int szPage; /* Database page-size */
1664: WalIterator *pIter = 0; /* Wal iterator context */
1665: u32 iDbpage = 0; /* Next database page to write */
1666: u32 iFrame = 0; /* Wal frame containing data for iDbpage */
1667: u32 mxSafeFrame; /* Max frame that can be backfilled */
1668: u32 mxPage; /* Max database page to write */
1669: int i; /* Loop counter */
1670: volatile WalCkptInfo *pInfo; /* The checkpoint status information */
1671: int (*xBusy)(void*) = 0; /* Function to call when waiting for locks */
1672:
1673: szPage = walPagesize(pWal);
1674: testcase( szPage<=32768 );
1675: testcase( szPage>=65536 );
1676: pInfo = walCkptInfo(pWal);
1677: if( pInfo->nBackfill>=pWal->hdr.mxFrame ) return SQLITE_OK;
1678:
1679: /* Allocate the iterator */
1680: rc = walIteratorInit(pWal, &pIter);
1681: if( rc!=SQLITE_OK ){
1682: return rc;
1683: }
1684: assert( pIter );
1685:
1686: if( eMode!=SQLITE_CHECKPOINT_PASSIVE ) xBusy = xBusyCall;
1687:
1688: /* Compute in mxSafeFrame the index of the last frame of the WAL that is
1689: ** safe to write into the database. Frames beyond mxSafeFrame might
1690: ** overwrite database pages that are in use by active readers and thus
1691: ** cannot be backfilled from the WAL.
1692: */
1693: mxSafeFrame = pWal->hdr.mxFrame;
1694: mxPage = pWal->hdr.nPage;
1695: for(i=1; i<WAL_NREADER; i++){
1696: u32 y = pInfo->aReadMark[i];
1697: if( mxSafeFrame>y ){
1698: assert( y<=pWal->hdr.mxFrame );
1699: rc = walBusyLock(pWal, xBusy, pBusyArg, WAL_READ_LOCK(i), 1);
1700: if( rc==SQLITE_OK ){
1701: pInfo->aReadMark[i] = READMARK_NOT_USED;
1702: walUnlockExclusive(pWal, WAL_READ_LOCK(i), 1);
1703: }else if( rc==SQLITE_BUSY ){
1704: mxSafeFrame = y;
1705: xBusy = 0;
1706: }else{
1707: goto walcheckpoint_out;
1708: }
1709: }
1710: }
1711:
1712: if( pInfo->nBackfill<mxSafeFrame
1713: && (rc = walBusyLock(pWal, xBusy, pBusyArg, WAL_READ_LOCK(0), 1))==SQLITE_OK
1714: ){
1715: i64 nSize; /* Current size of database file */
1716: u32 nBackfill = pInfo->nBackfill;
1717:
1718: /* Sync the WAL to disk */
1719: if( sync_flags ){
1720: rc = sqlite3OsSync(pWal->pWalFd, sync_flags);
1721: }
1722:
1723: /* If the database file may grow as a result of this checkpoint, hint
1724: ** about the eventual size of the db file to the VFS layer.
1725: */
1726: if( rc==SQLITE_OK ){
1727: i64 nReq = ((i64)mxPage * szPage);
1728: rc = sqlite3OsFileSize(pWal->pDbFd, &nSize);
1729: if( rc==SQLITE_OK && nSize<nReq ){
1730: sqlite3OsFileControlHint(pWal->pDbFd, SQLITE_FCNTL_SIZE_HINT, &nReq);
1731: }
1732: }
1733:
1734: /* Iterate through the contents of the WAL, copying data to the db file. */
1735: while( rc==SQLITE_OK && 0==walIteratorNext(pIter, &iDbpage, &iFrame) ){
1736: i64 iOffset;
1737: assert( walFramePgno(pWal, iFrame)==iDbpage );
1738: if( iFrame<=nBackfill || iFrame>mxSafeFrame || iDbpage>mxPage ) continue;
1739: iOffset = walFrameOffset(iFrame, szPage) + WAL_FRAME_HDRSIZE;
1740: /* testcase( IS_BIG_INT(iOffset) ); // requires a 4GiB WAL file */
1741: rc = sqlite3OsRead(pWal->pWalFd, zBuf, szPage, iOffset);
1742: if( rc!=SQLITE_OK ) break;
1743: iOffset = (iDbpage-1)*(i64)szPage;
1744: testcase( IS_BIG_INT(iOffset) );
1745: rc = sqlite3OsWrite(pWal->pDbFd, zBuf, szPage, iOffset);
1746: if( rc!=SQLITE_OK ) break;
1747: }
1748:
1749: /* If work was actually accomplished... */
1750: if( rc==SQLITE_OK ){
1751: if( mxSafeFrame==walIndexHdr(pWal)->mxFrame ){
1752: i64 szDb = pWal->hdr.nPage*(i64)szPage;
1753: testcase( IS_BIG_INT(szDb) );
1754: rc = sqlite3OsTruncate(pWal->pDbFd, szDb);
1755: if( rc==SQLITE_OK && sync_flags ){
1756: rc = sqlite3OsSync(pWal->pDbFd, sync_flags);
1757: }
1758: }
1759: if( rc==SQLITE_OK ){
1760: pInfo->nBackfill = mxSafeFrame;
1761: }
1762: }
1763:
1764: /* Release the reader lock held while backfilling */
1765: walUnlockExclusive(pWal, WAL_READ_LOCK(0), 1);
1766: }
1767:
1768: if( rc==SQLITE_BUSY ){
1769: /* Reset the return code so as not to report a checkpoint failure
1770: ** just because there are active readers. */
1771: rc = SQLITE_OK;
1772: }
1773:
1774: /* If this is an SQLITE_CHECKPOINT_RESTART operation, and the entire wal
1775: ** file has been copied into the database file, then block until all
1776: ** readers have finished using the wal file. This ensures that the next
1777: ** process to write to the database restarts the wal file.
1778: */
1779: if( rc==SQLITE_OK && eMode!=SQLITE_CHECKPOINT_PASSIVE ){
1780: assert( pWal->writeLock );
1781: if( pInfo->nBackfill<pWal->hdr.mxFrame ){
1782: rc = SQLITE_BUSY;
1783: }else if( eMode==SQLITE_CHECKPOINT_RESTART ){
1784: assert( mxSafeFrame==pWal->hdr.mxFrame );
1785: rc = walBusyLock(pWal, xBusy, pBusyArg, WAL_READ_LOCK(1), WAL_NREADER-1);
1786: if( rc==SQLITE_OK ){
1787: walUnlockExclusive(pWal, WAL_READ_LOCK(1), WAL_NREADER-1);
1788: }
1789: }
1790: }
1791:
1792: walcheckpoint_out:
1793: walIteratorFree(pIter);
1794: return rc;
1795: }
1796:
1797: /*
1798: ** If the WAL file is currently larger than nMax bytes in size, truncate
1799: ** it to exactly nMax bytes. If an error occurs while doing so, ignore it.
1800: */
1801: static void walLimitSize(Wal *pWal, i64 nMax){
1802: i64 sz;
1803: int rx;
1804: sqlite3BeginBenignMalloc();
1805: rx = sqlite3OsFileSize(pWal->pWalFd, &sz);
1806: if( rx==SQLITE_OK && (sz > nMax ) ){
1807: rx = sqlite3OsTruncate(pWal->pWalFd, nMax);
1808: }
1809: sqlite3EndBenignMalloc();
1810: if( rx ){
1811: sqlite3_log(rx, "cannot limit WAL size: %s", pWal->zWalName);
1812: }
1813: }
1814:
1815: /*
1816: ** Close a connection to a log file.
1817: */
1818: int sqlite3WalClose(
1819: Wal *pWal, /* Wal to close */
1820: int sync_flags, /* Flags to pass to OsSync() (or 0) */
1821: int nBuf,
1822: u8 *zBuf /* Buffer of at least nBuf bytes */
1823: ){
1824: int rc = SQLITE_OK;
1825: if( pWal ){
1826: int isDelete = 0; /* True to unlink wal and wal-index files */
1827:
1828: /* If an EXCLUSIVE lock can be obtained on the database file (using the
1829: ** ordinary, rollback-mode locking methods, this guarantees that the
1830: ** connection associated with this log file is the only connection to
1831: ** the database. In this case checkpoint the database and unlink both
1832: ** the wal and wal-index files.
1833: **
1834: ** The EXCLUSIVE lock is not released before returning.
1835: */
1836: rc = sqlite3OsLock(pWal->pDbFd, SQLITE_LOCK_EXCLUSIVE);
1837: if( rc==SQLITE_OK ){
1838: if( pWal->exclusiveMode==WAL_NORMAL_MODE ){
1839: pWal->exclusiveMode = WAL_EXCLUSIVE_MODE;
1840: }
1841: rc = sqlite3WalCheckpoint(
1842: pWal, SQLITE_CHECKPOINT_PASSIVE, 0, 0, sync_flags, nBuf, zBuf, 0, 0
1843: );
1844: if( rc==SQLITE_OK ){
1845: int bPersist = -1;
1846: sqlite3OsFileControlHint(
1847: pWal->pDbFd, SQLITE_FCNTL_PERSIST_WAL, &bPersist
1848: );
1849: if( bPersist!=1 ){
1850: /* Try to delete the WAL file if the checkpoint completed and
1851: ** fsyned (rc==SQLITE_OK) and if we are not in persistent-wal
1852: ** mode (!bPersist) */
1853: isDelete = 1;
1854: }else if( pWal->mxWalSize>=0 ){
1855: /* Try to truncate the WAL file to zero bytes if the checkpoint
1856: ** completed and fsynced (rc==SQLITE_OK) and we are in persistent
1857: ** WAL mode (bPersist) and if the PRAGMA journal_size_limit is a
1858: ** non-negative value (pWal->mxWalSize>=0). Note that we truncate
1859: ** to zero bytes as truncating to the journal_size_limit might
1860: ** leave a corrupt WAL file on disk. */
1861: walLimitSize(pWal, 0);
1862: }
1863: }
1864: }
1865:
1866: walIndexClose(pWal, isDelete);
1867: sqlite3OsClose(pWal->pWalFd);
1868: if( isDelete ){
1869: sqlite3BeginBenignMalloc();
1870: sqlite3OsDelete(pWal->pVfs, pWal->zWalName, 0);
1871: sqlite3EndBenignMalloc();
1872: }
1873: WALTRACE(("WAL%p: closed\n", pWal));
1874: sqlite3_free((void *)pWal->apWiData);
1875: sqlite3_free(pWal);
1876: }
1877: return rc;
1878: }
1879:
1880: /*
1881: ** Try to read the wal-index header. Return 0 on success and 1 if
1882: ** there is a problem.
1883: **
1884: ** The wal-index is in shared memory. Another thread or process might
1885: ** be writing the header at the same time this procedure is trying to
1886: ** read it, which might result in inconsistency. A dirty read is detected
1887: ** by verifying that both copies of the header are the same and also by
1888: ** a checksum on the header.
1889: **
1890: ** If and only if the read is consistent and the header is different from
1891: ** pWal->hdr, then pWal->hdr is updated to the content of the new header
1892: ** and *pChanged is set to 1.
1893: **
1894: ** If the checksum cannot be verified return non-zero. If the header
1895: ** is read successfully and the checksum verified, return zero.
1896: */
1897: static int walIndexTryHdr(Wal *pWal, int *pChanged){
1898: u32 aCksum[2]; /* Checksum on the header content */
1899: WalIndexHdr h1, h2; /* Two copies of the header content */
1900: WalIndexHdr volatile *aHdr; /* Header in shared memory */
1901:
1902: /* The first page of the wal-index must be mapped at this point. */
1903: assert( pWal->nWiData>0 && pWal->apWiData[0] );
1904:
1905: /* Read the header. This might happen concurrently with a write to the
1906: ** same area of shared memory on a different CPU in a SMP,
1907: ** meaning it is possible that an inconsistent snapshot is read
1908: ** from the file. If this happens, return non-zero.
1909: **
1910: ** There are two copies of the header at the beginning of the wal-index.
1911: ** When reading, read [0] first then [1]. Writes are in the reverse order.
1912: ** Memory barriers are used to prevent the compiler or the hardware from
1913: ** reordering the reads and writes.
1914: */
1915: aHdr = walIndexHdr(pWal);
1916: memcpy(&h1, (void *)&aHdr[0], sizeof(h1));
1917: walShmBarrier(pWal);
1918: memcpy(&h2, (void *)&aHdr[1], sizeof(h2));
1919:
1920: if( memcmp(&h1, &h2, sizeof(h1))!=0 ){
1921: return 1; /* Dirty read */
1922: }
1923: if( h1.isInit==0 ){
1924: return 1; /* Malformed header - probably all zeros */
1925: }
1926: walChecksumBytes(1, (u8*)&h1, sizeof(h1)-sizeof(h1.aCksum), 0, aCksum);
1927: if( aCksum[0]!=h1.aCksum[0] || aCksum[1]!=h1.aCksum[1] ){
1928: return 1; /* Checksum does not match */
1929: }
1930:
1931: if( memcmp(&pWal->hdr, &h1, sizeof(WalIndexHdr)) ){
1932: *pChanged = 1;
1933: memcpy(&pWal->hdr, &h1, sizeof(WalIndexHdr));
1934: pWal->szPage = (pWal->hdr.szPage&0xfe00) + ((pWal->hdr.szPage&0x0001)<<16);
1935: testcase( pWal->szPage<=32768 );
1936: testcase( pWal->szPage>=65536 );
1937: }
1938:
1939: /* The header was successfully read. Return zero. */
1940: return 0;
1941: }
1942:
1943: /*
1944: ** Read the wal-index header from the wal-index and into pWal->hdr.
1945: ** If the wal-header appears to be corrupt, try to reconstruct the
1946: ** wal-index from the WAL before returning.
1947: **
1948: ** Set *pChanged to 1 if the wal-index header value in pWal->hdr is
1949: ** changed by this opertion. If pWal->hdr is unchanged, set *pChanged
1950: ** to 0.
1951: **
1952: ** If the wal-index header is successfully read, return SQLITE_OK.
1953: ** Otherwise an SQLite error code.
1954: */
1955: static int walIndexReadHdr(Wal *pWal, int *pChanged){
1956: int rc; /* Return code */
1957: int badHdr; /* True if a header read failed */
1958: volatile u32 *page0; /* Chunk of wal-index containing header */
1959:
1960: /* Ensure that page 0 of the wal-index (the page that contains the
1961: ** wal-index header) is mapped. Return early if an error occurs here.
1962: */
1963: assert( pChanged );
1964: rc = walIndexPage(pWal, 0, &page0);
1965: if( rc!=SQLITE_OK ){
1966: return rc;
1967: };
1968: assert( page0 || pWal->writeLock==0 );
1969:
1970: /* If the first page of the wal-index has been mapped, try to read the
1971: ** wal-index header immediately, without holding any lock. This usually
1972: ** works, but may fail if the wal-index header is corrupt or currently
1973: ** being modified by another thread or process.
1974: */
1975: badHdr = (page0 ? walIndexTryHdr(pWal, pChanged) : 1);
1976:
1977: /* If the first attempt failed, it might have been due to a race
1978: ** with a writer. So get a WRITE lock and try again.
1979: */
1980: assert( badHdr==0 || pWal->writeLock==0 );
1981: if( badHdr ){
1982: if( pWal->readOnly & WAL_SHM_RDONLY ){
1983: if( SQLITE_OK==(rc = walLockShared(pWal, WAL_WRITE_LOCK)) ){
1984: walUnlockShared(pWal, WAL_WRITE_LOCK);
1985: rc = SQLITE_READONLY_RECOVERY;
1986: }
1987: }else if( SQLITE_OK==(rc = walLockExclusive(pWal, WAL_WRITE_LOCK, 1)) ){
1988: pWal->writeLock = 1;
1989: if( SQLITE_OK==(rc = walIndexPage(pWal, 0, &page0)) ){
1990: badHdr = walIndexTryHdr(pWal, pChanged);
1991: if( badHdr ){
1992: /* If the wal-index header is still malformed even while holding
1993: ** a WRITE lock, it can only mean that the header is corrupted and
1994: ** needs to be reconstructed. So run recovery to do exactly that.
1995: */
1996: rc = walIndexRecover(pWal);
1997: *pChanged = 1;
1998: }
1999: }
2000: pWal->writeLock = 0;
2001: walUnlockExclusive(pWal, WAL_WRITE_LOCK, 1);
2002: }
2003: }
2004:
2005: /* If the header is read successfully, check the version number to make
2006: ** sure the wal-index was not constructed with some future format that
2007: ** this version of SQLite cannot understand.
2008: */
2009: if( badHdr==0 && pWal->hdr.iVersion!=WALINDEX_MAX_VERSION ){
2010: rc = SQLITE_CANTOPEN_BKPT;
2011: }
2012:
2013: return rc;
2014: }
2015:
2016: /*
2017: ** This is the value that walTryBeginRead returns when it needs to
2018: ** be retried.
2019: */
2020: #define WAL_RETRY (-1)
2021:
2022: /*
2023: ** Attempt to start a read transaction. This might fail due to a race or
2024: ** other transient condition. When that happens, it returns WAL_RETRY to
2025: ** indicate to the caller that it is safe to retry immediately.
2026: **
2027: ** On success return SQLITE_OK. On a permanent failure (such an
2028: ** I/O error or an SQLITE_BUSY because another process is running
2029: ** recovery) return a positive error code.
2030: **
2031: ** The useWal parameter is true to force the use of the WAL and disable
2032: ** the case where the WAL is bypassed because it has been completely
2033: ** checkpointed. If useWal==0 then this routine calls walIndexReadHdr()
2034: ** to make a copy of the wal-index header into pWal->hdr. If the
2035: ** wal-index header has changed, *pChanged is set to 1 (as an indication
2036: ** to the caller that the local paget cache is obsolete and needs to be
2037: ** flushed.) When useWal==1, the wal-index header is assumed to already
2038: ** be loaded and the pChanged parameter is unused.
2039: **
2040: ** The caller must set the cnt parameter to the number of prior calls to
2041: ** this routine during the current read attempt that returned WAL_RETRY.
2042: ** This routine will start taking more aggressive measures to clear the
2043: ** race conditions after multiple WAL_RETRY returns, and after an excessive
2044: ** number of errors will ultimately return SQLITE_PROTOCOL. The
2045: ** SQLITE_PROTOCOL return indicates that some other process has gone rogue
2046: ** and is not honoring the locking protocol. There is a vanishingly small
2047: ** chance that SQLITE_PROTOCOL could be returned because of a run of really
2048: ** bad luck when there is lots of contention for the wal-index, but that
2049: ** possibility is so small that it can be safely neglected, we believe.
2050: **
2051: ** On success, this routine obtains a read lock on
2052: ** WAL_READ_LOCK(pWal->readLock). The pWal->readLock integer is
2053: ** in the range 0 <= pWal->readLock < WAL_NREADER. If pWal->readLock==(-1)
2054: ** that means the Wal does not hold any read lock. The reader must not
2055: ** access any database page that is modified by a WAL frame up to and
2056: ** including frame number aReadMark[pWal->readLock]. The reader will
2057: ** use WAL frames up to and including pWal->hdr.mxFrame if pWal->readLock>0
2058: ** Or if pWal->readLock==0, then the reader will ignore the WAL
2059: ** completely and get all content directly from the database file.
2060: ** If the useWal parameter is 1 then the WAL will never be ignored and
2061: ** this routine will always set pWal->readLock>0 on success.
2062: ** When the read transaction is completed, the caller must release the
2063: ** lock on WAL_READ_LOCK(pWal->readLock) and set pWal->readLock to -1.
2064: **
2065: ** This routine uses the nBackfill and aReadMark[] fields of the header
2066: ** to select a particular WAL_READ_LOCK() that strives to let the
2067: ** checkpoint process do as much work as possible. This routine might
2068: ** update values of the aReadMark[] array in the header, but if it does
2069: ** so it takes care to hold an exclusive lock on the corresponding
2070: ** WAL_READ_LOCK() while changing values.
2071: */
2072: static int walTryBeginRead(Wal *pWal, int *pChanged, int useWal, int cnt){
2073: volatile WalCkptInfo *pInfo; /* Checkpoint information in wal-index */
2074: u32 mxReadMark; /* Largest aReadMark[] value */
2075: int mxI; /* Index of largest aReadMark[] value */
2076: int i; /* Loop counter */
2077: int rc = SQLITE_OK; /* Return code */
2078:
2079: assert( pWal->readLock<0 ); /* Not currently locked */
2080:
2081: /* Take steps to avoid spinning forever if there is a protocol error.
2082: **
2083: ** Circumstances that cause a RETRY should only last for the briefest
2084: ** instances of time. No I/O or other system calls are done while the
2085: ** locks are held, so the locks should not be held for very long. But
2086: ** if we are unlucky, another process that is holding a lock might get
2087: ** paged out or take a page-fault that is time-consuming to resolve,
2088: ** during the few nanoseconds that it is holding the lock. In that case,
2089: ** it might take longer than normal for the lock to free.
2090: **
2091: ** After 5 RETRYs, we begin calling sqlite3OsSleep(). The first few
2092: ** calls to sqlite3OsSleep() have a delay of 1 microsecond. Really this
2093: ** is more of a scheduler yield than an actual delay. But on the 10th
2094: ** an subsequent retries, the delays start becoming longer and longer,
2095: ** so that on the 100th (and last) RETRY we delay for 21 milliseconds.
2096: ** The total delay time before giving up is less than 1 second.
2097: */
2098: if( cnt>5 ){
2099: int nDelay = 1; /* Pause time in microseconds */
2100: if( cnt>100 ){
2101: VVA_ONLY( pWal->lockError = 1; )
2102: return SQLITE_PROTOCOL;
2103: }
2104: if( cnt>=10 ) nDelay = (cnt-9)*238; /* Max delay 21ms. Total delay 996ms */
2105: sqlite3OsSleep(pWal->pVfs, nDelay);
2106: }
2107:
2108: if( !useWal ){
2109: rc = walIndexReadHdr(pWal, pChanged);
2110: if( rc==SQLITE_BUSY ){
2111: /* If there is not a recovery running in another thread or process
2112: ** then convert BUSY errors to WAL_RETRY. If recovery is known to
2113: ** be running, convert BUSY to BUSY_RECOVERY. There is a race here
2114: ** which might cause WAL_RETRY to be returned even if BUSY_RECOVERY
2115: ** would be technically correct. But the race is benign since with
2116: ** WAL_RETRY this routine will be called again and will probably be
2117: ** right on the second iteration.
2118: */
2119: if( pWal->apWiData[0]==0 ){
2120: /* This branch is taken when the xShmMap() method returns SQLITE_BUSY.
2121: ** We assume this is a transient condition, so return WAL_RETRY. The
2122: ** xShmMap() implementation used by the default unix and win32 VFS
2123: ** modules may return SQLITE_BUSY due to a race condition in the
2124: ** code that determines whether or not the shared-memory region
2125: ** must be zeroed before the requested page is returned.
2126: */
2127: rc = WAL_RETRY;
2128: }else if( SQLITE_OK==(rc = walLockShared(pWal, WAL_RECOVER_LOCK)) ){
2129: walUnlockShared(pWal, WAL_RECOVER_LOCK);
2130: rc = WAL_RETRY;
2131: }else if( rc==SQLITE_BUSY ){
2132: rc = SQLITE_BUSY_RECOVERY;
2133: }
2134: }
2135: if( rc!=SQLITE_OK ){
2136: return rc;
2137: }
2138: }
2139:
2140: pInfo = walCkptInfo(pWal);
2141: if( !useWal && pInfo->nBackfill==pWal->hdr.mxFrame ){
2142: /* The WAL has been completely backfilled (or it is empty).
2143: ** and can be safely ignored.
2144: */
2145: rc = walLockShared(pWal, WAL_READ_LOCK(0));
2146: walShmBarrier(pWal);
2147: if( rc==SQLITE_OK ){
2148: if( memcmp((void *)walIndexHdr(pWal), &pWal->hdr, sizeof(WalIndexHdr)) ){
2149: /* It is not safe to allow the reader to continue here if frames
2150: ** may have been appended to the log before READ_LOCK(0) was obtained.
2151: ** When holding READ_LOCK(0), the reader ignores the entire log file,
2152: ** which implies that the database file contains a trustworthy
2153: ** snapshoT. Since holding READ_LOCK(0) prevents a checkpoint from
2154: ** happening, this is usually correct.
2155: **
2156: ** However, if frames have been appended to the log (or if the log
2157: ** is wrapped and written for that matter) before the READ_LOCK(0)
2158: ** is obtained, that is not necessarily true. A checkpointer may
2159: ** have started to backfill the appended frames but crashed before
2160: ** it finished. Leaving a corrupt image in the database file.
2161: */
2162: walUnlockShared(pWal, WAL_READ_LOCK(0));
2163: return WAL_RETRY;
2164: }
2165: pWal->readLock = 0;
2166: return SQLITE_OK;
2167: }else if( rc!=SQLITE_BUSY ){
2168: return rc;
2169: }
2170: }
2171:
2172: /* If we get this far, it means that the reader will want to use
2173: ** the WAL to get at content from recent commits. The job now is
2174: ** to select one of the aReadMark[] entries that is closest to
2175: ** but not exceeding pWal->hdr.mxFrame and lock that entry.
2176: */
2177: mxReadMark = 0;
2178: mxI = 0;
2179: for(i=1; i<WAL_NREADER; i++){
2180: u32 thisMark = pInfo->aReadMark[i];
2181: if( mxReadMark<=thisMark && thisMark<=pWal->hdr.mxFrame ){
2182: assert( thisMark!=READMARK_NOT_USED );
2183: mxReadMark = thisMark;
2184: mxI = i;
2185: }
2186: }
2187: /* There was once an "if" here. The extra "{" is to preserve indentation. */
2188: {
2189: if( (pWal->readOnly & WAL_SHM_RDONLY)==0
2190: && (mxReadMark<pWal->hdr.mxFrame || mxI==0)
2191: ){
2192: for(i=1; i<WAL_NREADER; i++){
2193: rc = walLockExclusive(pWal, WAL_READ_LOCK(i), 1);
2194: if( rc==SQLITE_OK ){
2195: mxReadMark = pInfo->aReadMark[i] = pWal->hdr.mxFrame;
2196: mxI = i;
2197: walUnlockExclusive(pWal, WAL_READ_LOCK(i), 1);
2198: break;
2199: }else if( rc!=SQLITE_BUSY ){
2200: return rc;
2201: }
2202: }
2203: }
2204: if( mxI==0 ){
2205: assert( rc==SQLITE_BUSY || (pWal->readOnly & WAL_SHM_RDONLY)!=0 );
2206: return rc==SQLITE_BUSY ? WAL_RETRY : SQLITE_READONLY_CANTLOCK;
2207: }
2208:
2209: rc = walLockShared(pWal, WAL_READ_LOCK(mxI));
2210: if( rc ){
2211: return rc==SQLITE_BUSY ? WAL_RETRY : rc;
2212: }
2213: /* Now that the read-lock has been obtained, check that neither the
2214: ** value in the aReadMark[] array or the contents of the wal-index
2215: ** header have changed.
2216: **
2217: ** It is necessary to check that the wal-index header did not change
2218: ** between the time it was read and when the shared-lock was obtained
2219: ** on WAL_READ_LOCK(mxI) was obtained to account for the possibility
2220: ** that the log file may have been wrapped by a writer, or that frames
2221: ** that occur later in the log than pWal->hdr.mxFrame may have been
2222: ** copied into the database by a checkpointer. If either of these things
2223: ** happened, then reading the database with the current value of
2224: ** pWal->hdr.mxFrame risks reading a corrupted snapshot. So, retry
2225: ** instead.
2226: **
2227: ** This does not guarantee that the copy of the wal-index header is up to
2228: ** date before proceeding. That would not be possible without somehow
2229: ** blocking writers. It only guarantees that a dangerous checkpoint or
2230: ** log-wrap (either of which would require an exclusive lock on
2231: ** WAL_READ_LOCK(mxI)) has not occurred since the snapshot was valid.
2232: */
2233: walShmBarrier(pWal);
2234: if( pInfo->aReadMark[mxI]!=mxReadMark
2235: || memcmp((void *)walIndexHdr(pWal), &pWal->hdr, sizeof(WalIndexHdr))
2236: ){
2237: walUnlockShared(pWal, WAL_READ_LOCK(mxI));
2238: return WAL_RETRY;
2239: }else{
2240: assert( mxReadMark<=pWal->hdr.mxFrame );
2241: pWal->readLock = (i16)mxI;
2242: }
2243: }
2244: return rc;
2245: }
2246:
2247: /*
2248: ** Begin a read transaction on the database.
2249: **
2250: ** This routine used to be called sqlite3OpenSnapshot() and with good reason:
2251: ** it takes a snapshot of the state of the WAL and wal-index for the current
2252: ** instant in time. The current thread will continue to use this snapshot.
2253: ** Other threads might append new content to the WAL and wal-index but
2254: ** that extra content is ignored by the current thread.
2255: **
2256: ** If the database contents have changes since the previous read
2257: ** transaction, then *pChanged is set to 1 before returning. The
2258: ** Pager layer will use this to know that is cache is stale and
2259: ** needs to be flushed.
2260: */
2261: int sqlite3WalBeginReadTransaction(Wal *pWal, int *pChanged){
2262: int rc; /* Return code */
2263: int cnt = 0; /* Number of TryBeginRead attempts */
2264:
2265: do{
2266: rc = walTryBeginRead(pWal, pChanged, 0, ++cnt);
2267: }while( rc==WAL_RETRY );
2268: testcase( (rc&0xff)==SQLITE_BUSY );
2269: testcase( (rc&0xff)==SQLITE_IOERR );
2270: testcase( rc==SQLITE_PROTOCOL );
2271: testcase( rc==SQLITE_OK );
2272: return rc;
2273: }
2274:
2275: /*
2276: ** Finish with a read transaction. All this does is release the
2277: ** read-lock.
2278: */
2279: void sqlite3WalEndReadTransaction(Wal *pWal){
2280: sqlite3WalEndWriteTransaction(pWal);
2281: if( pWal->readLock>=0 ){
2282: walUnlockShared(pWal, WAL_READ_LOCK(pWal->readLock));
2283: pWal->readLock = -1;
2284: }
2285: }
2286:
2287: /*
2288: ** Read a page from the WAL, if it is present in the WAL and if the
2289: ** current read transaction is configured to use the WAL.
2290: **
2291: ** The *pInWal is set to 1 if the requested page is in the WAL and
2292: ** has been loaded. Or *pInWal is set to 0 if the page was not in
2293: ** the WAL and needs to be read out of the database.
2294: */
2295: int sqlite3WalRead(
2296: Wal *pWal, /* WAL handle */
2297: Pgno pgno, /* Database page number to read data for */
2298: int *pInWal, /* OUT: True if data is read from WAL */
2299: int nOut, /* Size of buffer pOut in bytes */
2300: u8 *pOut /* Buffer to write page data to */
2301: ){
2302: u32 iRead = 0; /* If !=0, WAL frame to return data from */
2303: u32 iLast = pWal->hdr.mxFrame; /* Last page in WAL for this reader */
2304: int iHash; /* Used to loop through N hash tables */
2305:
2306: /* This routine is only be called from within a read transaction. */
2307: assert( pWal->readLock>=0 || pWal->lockError );
2308:
2309: /* If the "last page" field of the wal-index header snapshot is 0, then
2310: ** no data will be read from the wal under any circumstances. Return early
2311: ** in this case as an optimization. Likewise, if pWal->readLock==0,
2312: ** then the WAL is ignored by the reader so return early, as if the
2313: ** WAL were empty.
2314: */
2315: if( iLast==0 || pWal->readLock==0 ){
2316: *pInWal = 0;
2317: return SQLITE_OK;
2318: }
2319:
2320: /* Search the hash table or tables for an entry matching page number
2321: ** pgno. Each iteration of the following for() loop searches one
2322: ** hash table (each hash table indexes up to HASHTABLE_NPAGE frames).
2323: **
2324: ** This code might run concurrently to the code in walIndexAppend()
2325: ** that adds entries to the wal-index (and possibly to this hash
2326: ** table). This means the value just read from the hash
2327: ** slot (aHash[iKey]) may have been added before or after the
2328: ** current read transaction was opened. Values added after the
2329: ** read transaction was opened may have been written incorrectly -
2330: ** i.e. these slots may contain garbage data. However, we assume
2331: ** that any slots written before the current read transaction was
2332: ** opened remain unmodified.
2333: **
2334: ** For the reasons above, the if(...) condition featured in the inner
2335: ** loop of the following block is more stringent that would be required
2336: ** if we had exclusive access to the hash-table:
2337: **
2338: ** (aPgno[iFrame]==pgno):
2339: ** This condition filters out normal hash-table collisions.
2340: **
2341: ** (iFrame<=iLast):
2342: ** This condition filters out entries that were added to the hash
2343: ** table after the current read-transaction had started.
2344: */
2345: for(iHash=walFramePage(iLast); iHash>=0 && iRead==0; iHash--){
2346: volatile ht_slot *aHash; /* Pointer to hash table */
2347: volatile u32 *aPgno; /* Pointer to array of page numbers */
2348: u32 iZero; /* Frame number corresponding to aPgno[0] */
2349: int iKey; /* Hash slot index */
2350: int nCollide; /* Number of hash collisions remaining */
2351: int rc; /* Error code */
2352:
2353: rc = walHashGet(pWal, iHash, &aHash, &aPgno, &iZero);
2354: if( rc!=SQLITE_OK ){
2355: return rc;
2356: }
2357: nCollide = HASHTABLE_NSLOT;
2358: for(iKey=walHash(pgno); aHash[iKey]; iKey=walNextHash(iKey)){
2359: u32 iFrame = aHash[iKey] + iZero;
2360: if( iFrame<=iLast && aPgno[aHash[iKey]]==pgno ){
2361: /* assert( iFrame>iRead ); -- not true if there is corruption */
2362: iRead = iFrame;
2363: }
2364: if( (nCollide--)==0 ){
2365: return SQLITE_CORRUPT_BKPT;
2366: }
2367: }
2368: }
2369:
2370: #ifdef SQLITE_ENABLE_EXPENSIVE_ASSERT
2371: /* If expensive assert() statements are available, do a linear search
2372: ** of the wal-index file content. Make sure the results agree with the
2373: ** result obtained using the hash indexes above. */
2374: {
2375: u32 iRead2 = 0;
2376: u32 iTest;
2377: for(iTest=iLast; iTest>0; iTest--){
2378: if( walFramePgno(pWal, iTest)==pgno ){
2379: iRead2 = iTest;
2380: break;
2381: }
2382: }
2383: assert( iRead==iRead2 );
2384: }
2385: #endif
2386:
2387: /* If iRead is non-zero, then it is the log frame number that contains the
2388: ** required page. Read and return data from the log file.
2389: */
2390: if( iRead ){
2391: int sz;
2392: i64 iOffset;
2393: sz = pWal->hdr.szPage;
2394: sz = (sz&0xfe00) + ((sz&0x0001)<<16);
2395: testcase( sz<=32768 );
2396: testcase( sz>=65536 );
2397: iOffset = walFrameOffset(iRead, sz) + WAL_FRAME_HDRSIZE;
2398: *pInWal = 1;
2399: /* testcase( IS_BIG_INT(iOffset) ); // requires a 4GiB WAL */
2400: return sqlite3OsRead(pWal->pWalFd, pOut, nOut, iOffset);
2401: }
2402:
2403: *pInWal = 0;
2404: return SQLITE_OK;
2405: }
2406:
2407:
2408: /*
2409: ** Return the size of the database in pages (or zero, if unknown).
2410: */
2411: Pgno sqlite3WalDbsize(Wal *pWal){
2412: if( pWal && ALWAYS(pWal->readLock>=0) ){
2413: return pWal->hdr.nPage;
2414: }
2415: return 0;
2416: }
2417:
2418:
2419: /*
2420: ** This function starts a write transaction on the WAL.
2421: **
2422: ** A read transaction must have already been started by a prior call
2423: ** to sqlite3WalBeginReadTransaction().
2424: **
2425: ** If another thread or process has written into the database since
2426: ** the read transaction was started, then it is not possible for this
2427: ** thread to write as doing so would cause a fork. So this routine
2428: ** returns SQLITE_BUSY in that case and no write transaction is started.
2429: **
2430: ** There can only be a single writer active at a time.
2431: */
2432: int sqlite3WalBeginWriteTransaction(Wal *pWal){
2433: int rc;
2434:
2435: /* Cannot start a write transaction without first holding a read
2436: ** transaction. */
2437: assert( pWal->readLock>=0 );
2438:
2439: if( pWal->readOnly ){
2440: return SQLITE_READONLY;
2441: }
2442:
2443: /* Only one writer allowed at a time. Get the write lock. Return
2444: ** SQLITE_BUSY if unable.
2445: */
2446: rc = walLockExclusive(pWal, WAL_WRITE_LOCK, 1);
2447: if( rc ){
2448: return rc;
2449: }
2450: pWal->writeLock = 1;
2451:
2452: /* If another connection has written to the database file since the
2453: ** time the read transaction on this connection was started, then
2454: ** the write is disallowed.
2455: */
2456: if( memcmp(&pWal->hdr, (void *)walIndexHdr(pWal), sizeof(WalIndexHdr))!=0 ){
2457: walUnlockExclusive(pWal, WAL_WRITE_LOCK, 1);
2458: pWal->writeLock = 0;
2459: rc = SQLITE_BUSY;
2460: }
2461:
2462: return rc;
2463: }
2464:
2465: /*
2466: ** End a write transaction. The commit has already been done. This
2467: ** routine merely releases the lock.
2468: */
2469: int sqlite3WalEndWriteTransaction(Wal *pWal){
2470: if( pWal->writeLock ){
2471: walUnlockExclusive(pWal, WAL_WRITE_LOCK, 1);
2472: pWal->writeLock = 0;
2473: pWal->truncateOnCommit = 0;
2474: }
2475: return SQLITE_OK;
2476: }
2477:
2478: /*
2479: ** If any data has been written (but not committed) to the log file, this
2480: ** function moves the write-pointer back to the start of the transaction.
2481: **
2482: ** Additionally, the callback function is invoked for each frame written
2483: ** to the WAL since the start of the transaction. If the callback returns
2484: ** other than SQLITE_OK, it is not invoked again and the error code is
2485: ** returned to the caller.
2486: **
2487: ** Otherwise, if the callback function does not return an error, this
2488: ** function returns SQLITE_OK.
2489: */
2490: int sqlite3WalUndo(Wal *pWal, int (*xUndo)(void *, Pgno), void *pUndoCtx){
2491: int rc = SQLITE_OK;
2492: if( ALWAYS(pWal->writeLock) ){
2493: Pgno iMax = pWal->hdr.mxFrame;
2494: Pgno iFrame;
2495:
2496: /* Restore the clients cache of the wal-index header to the state it
2497: ** was in before the client began writing to the database.
2498: */
2499: memcpy(&pWal->hdr, (void *)walIndexHdr(pWal), sizeof(WalIndexHdr));
2500:
2501: for(iFrame=pWal->hdr.mxFrame+1;
2502: ALWAYS(rc==SQLITE_OK) && iFrame<=iMax;
2503: iFrame++
2504: ){
2505: /* This call cannot fail. Unless the page for which the page number
2506: ** is passed as the second argument is (a) in the cache and
2507: ** (b) has an outstanding reference, then xUndo is either a no-op
2508: ** (if (a) is false) or simply expels the page from the cache (if (b)
2509: ** is false).
2510: **
2511: ** If the upper layer is doing a rollback, it is guaranteed that there
2512: ** are no outstanding references to any page other than page 1. And
2513: ** page 1 is never written to the log until the transaction is
2514: ** committed. As a result, the call to xUndo may not fail.
2515: */
2516: assert( walFramePgno(pWal, iFrame)!=1 );
2517: rc = xUndo(pUndoCtx, walFramePgno(pWal, iFrame));
2518: }
2519: walCleanupHash(pWal);
2520: }
2521: assert( rc==SQLITE_OK );
2522: return rc;
2523: }
2524:
2525: /*
2526: ** Argument aWalData must point to an array of WAL_SAVEPOINT_NDATA u32
2527: ** values. This function populates the array with values required to
2528: ** "rollback" the write position of the WAL handle back to the current
2529: ** point in the event of a savepoint rollback (via WalSavepointUndo()).
2530: */
2531: void sqlite3WalSavepoint(Wal *pWal, u32 *aWalData){
2532: assert( pWal->writeLock );
2533: aWalData[0] = pWal->hdr.mxFrame;
2534: aWalData[1] = pWal->hdr.aFrameCksum[0];
2535: aWalData[2] = pWal->hdr.aFrameCksum[1];
2536: aWalData[3] = pWal->nCkpt;
2537: }
2538:
2539: /*
2540: ** Move the write position of the WAL back to the point identified by
2541: ** the values in the aWalData[] array. aWalData must point to an array
2542: ** of WAL_SAVEPOINT_NDATA u32 values that has been previously populated
2543: ** by a call to WalSavepoint().
2544: */
2545: int sqlite3WalSavepointUndo(Wal *pWal, u32 *aWalData){
2546: int rc = SQLITE_OK;
2547:
2548: assert( pWal->writeLock );
2549: assert( aWalData[3]!=pWal->nCkpt || aWalData[0]<=pWal->hdr.mxFrame );
2550:
2551: if( aWalData[3]!=pWal->nCkpt ){
2552: /* This savepoint was opened immediately after the write-transaction
2553: ** was started. Right after that, the writer decided to wrap around
2554: ** to the start of the log. Update the savepoint values to match.
2555: */
2556: aWalData[0] = 0;
2557: aWalData[3] = pWal->nCkpt;
2558: }
2559:
2560: if( aWalData[0]<pWal->hdr.mxFrame ){
2561: pWal->hdr.mxFrame = aWalData[0];
2562: pWal->hdr.aFrameCksum[0] = aWalData[1];
2563: pWal->hdr.aFrameCksum[1] = aWalData[2];
2564: walCleanupHash(pWal);
2565: }
2566:
2567: return rc;
2568: }
2569:
2570:
2571: /*
2572: ** This function is called just before writing a set of frames to the log
2573: ** file (see sqlite3WalFrames()). It checks to see if, instead of appending
2574: ** to the current log file, it is possible to overwrite the start of the
2575: ** existing log file with the new frames (i.e. "reset" the log). If so,
2576: ** it sets pWal->hdr.mxFrame to 0. Otherwise, pWal->hdr.mxFrame is left
2577: ** unchanged.
2578: **
2579: ** SQLITE_OK is returned if no error is encountered (regardless of whether
2580: ** or not pWal->hdr.mxFrame is modified). An SQLite error code is returned
2581: ** if an error occurs.
2582: */
2583: static int walRestartLog(Wal *pWal){
2584: int rc = SQLITE_OK;
2585: int cnt;
2586:
2587: if( pWal->readLock==0 ){
2588: volatile WalCkptInfo *pInfo = walCkptInfo(pWal);
2589: assert( pInfo->nBackfill==pWal->hdr.mxFrame );
2590: if( pInfo->nBackfill>0 ){
2591: u32 salt1;
2592: sqlite3_randomness(4, &salt1);
2593: rc = walLockExclusive(pWal, WAL_READ_LOCK(1), WAL_NREADER-1);
2594: if( rc==SQLITE_OK ){
2595: /* If all readers are using WAL_READ_LOCK(0) (in other words if no
2596: ** readers are currently using the WAL), then the transactions
2597: ** frames will overwrite the start of the existing log. Update the
2598: ** wal-index header to reflect this.
2599: **
2600: ** In theory it would be Ok to update the cache of the header only
2601: ** at this point. But updating the actual wal-index header is also
2602: ** safe and means there is no special case for sqlite3WalUndo()
2603: ** to handle if this transaction is rolled back.
2604: */
2605: int i; /* Loop counter */
2606: u32 *aSalt = pWal->hdr.aSalt; /* Big-endian salt values */
2607:
2608: pWal->nCkpt++;
2609: pWal->hdr.mxFrame = 0;
2610: sqlite3Put4byte((u8*)&aSalt[0], 1 + sqlite3Get4byte((u8*)&aSalt[0]));
2611: aSalt[1] = salt1;
2612: walIndexWriteHdr(pWal);
2613: pInfo->nBackfill = 0;
2614: for(i=1; i<WAL_NREADER; i++) pInfo->aReadMark[i] = READMARK_NOT_USED;
2615: assert( pInfo->aReadMark[0]==0 );
2616: walUnlockExclusive(pWal, WAL_READ_LOCK(1), WAL_NREADER-1);
2617: }else if( rc!=SQLITE_BUSY ){
2618: return rc;
2619: }
2620: }
2621: walUnlockShared(pWal, WAL_READ_LOCK(0));
2622: pWal->readLock = -1;
2623: cnt = 0;
2624: do{
2625: int notUsed;
2626: rc = walTryBeginRead(pWal, ¬Used, 1, ++cnt);
2627: }while( rc==WAL_RETRY );
2628: assert( (rc&0xff)!=SQLITE_BUSY ); /* BUSY not possible when useWal==1 */
2629: testcase( (rc&0xff)==SQLITE_IOERR );
2630: testcase( rc==SQLITE_PROTOCOL );
2631: testcase( rc==SQLITE_OK );
2632: }
2633: return rc;
2634: }
2635:
2636: /*
2637: ** Information about the current state of the WAL file and where
2638: ** the next fsync should occur - passed from sqlite3WalFrames() into
2639: ** walWriteToLog().
2640: */
2641: typedef struct WalWriter {
2642: Wal *pWal; /* The complete WAL information */
2643: sqlite3_file *pFd; /* The WAL file to which we write */
2644: sqlite3_int64 iSyncPoint; /* Fsync at this offset */
2645: int syncFlags; /* Flags for the fsync */
2646: int szPage; /* Size of one page */
2647: } WalWriter;
2648:
2649: /*
2650: ** Write iAmt bytes of content into the WAL file beginning at iOffset.
2651: ** Do a sync when crossing the p->iSyncPoint boundary.
2652: **
2653: ** In other words, if iSyncPoint is in between iOffset and iOffset+iAmt,
2654: ** first write the part before iSyncPoint, then sync, then write the
2655: ** rest.
2656: */
2657: static int walWriteToLog(
2658: WalWriter *p, /* WAL to write to */
2659: void *pContent, /* Content to be written */
2660: int iAmt, /* Number of bytes to write */
2661: sqlite3_int64 iOffset /* Start writing at this offset */
2662: ){
2663: int rc;
2664: if( iOffset<p->iSyncPoint && iOffset+iAmt>=p->iSyncPoint ){
2665: int iFirstAmt = (int)(p->iSyncPoint - iOffset);
2666: rc = sqlite3OsWrite(p->pFd, pContent, iFirstAmt, iOffset);
2667: if( rc ) return rc;
2668: iOffset += iFirstAmt;
2669: iAmt -= iFirstAmt;
2670: pContent = (void*)(iFirstAmt + (char*)pContent);
2671: assert( p->syncFlags & (SQLITE_SYNC_NORMAL|SQLITE_SYNC_FULL) );
2672: rc = sqlite3OsSync(p->pFd, p->syncFlags);
2673: if( iAmt==0 || rc ) return rc;
2674: }
2675: rc = sqlite3OsWrite(p->pFd, pContent, iAmt, iOffset);
2676: return rc;
2677: }
2678:
2679: /*
2680: ** Write out a single frame of the WAL
2681: */
2682: static int walWriteOneFrame(
2683: WalWriter *p, /* Where to write the frame */
2684: PgHdr *pPage, /* The page of the frame to be written */
2685: int nTruncate, /* The commit flag. Usually 0. >0 for commit */
2686: sqlite3_int64 iOffset /* Byte offset at which to write */
2687: ){
2688: int rc; /* Result code from subfunctions */
2689: void *pData; /* Data actually written */
2690: u8 aFrame[WAL_FRAME_HDRSIZE]; /* Buffer to assemble frame-header in */
2691: #if defined(SQLITE_HAS_CODEC)
2692: if( (pData = sqlite3PagerCodec(pPage))==0 ) return SQLITE_NOMEM;
2693: #else
2694: pData = pPage->pData;
2695: #endif
2696: walEncodeFrame(p->pWal, pPage->pgno, nTruncate, pData, aFrame);
2697: rc = walWriteToLog(p, aFrame, sizeof(aFrame), iOffset);
2698: if( rc ) return rc;
2699: /* Write the page data */
2700: rc = walWriteToLog(p, pData, p->szPage, iOffset+sizeof(aFrame));
2701: return rc;
2702: }
2703:
2704: /*
2705: ** Write a set of frames to the log. The caller must hold the write-lock
2706: ** on the log file (obtained using sqlite3WalBeginWriteTransaction()).
2707: */
2708: int sqlite3WalFrames(
2709: Wal *pWal, /* Wal handle to write to */
2710: int szPage, /* Database page-size in bytes */
2711: PgHdr *pList, /* List of dirty pages to write */
2712: Pgno nTruncate, /* Database size after this commit */
2713: int isCommit, /* True if this is a commit */
2714: int sync_flags /* Flags to pass to OsSync() (or 0) */
2715: ){
2716: int rc; /* Used to catch return codes */
2717: u32 iFrame; /* Next frame address */
2718: PgHdr *p; /* Iterator to run through pList with. */
2719: PgHdr *pLast = 0; /* Last frame in list */
2720: int nExtra = 0; /* Number of extra copies of last page */
2721: int szFrame; /* The size of a single frame */
2722: i64 iOffset; /* Next byte to write in WAL file */
2723: WalWriter w; /* The writer */
2724:
2725: assert( pList );
2726: assert( pWal->writeLock );
2727:
2728: /* If this frame set completes a transaction, then nTruncate>0. If
2729: ** nTruncate==0 then this frame set does not complete the transaction. */
2730: assert( (isCommit!=0)==(nTruncate!=0) );
2731:
2732: #if defined(SQLITE_TEST) && defined(SQLITE_DEBUG)
2733: { int cnt; for(cnt=0, p=pList; p; p=p->pDirty, cnt++){}
2734: WALTRACE(("WAL%p: frame write begin. %d frames. mxFrame=%d. %s\n",
2735: pWal, cnt, pWal->hdr.mxFrame, isCommit ? "Commit" : "Spill"));
2736: }
2737: #endif
2738:
2739: /* See if it is possible to write these frames into the start of the
2740: ** log file, instead of appending to it at pWal->hdr.mxFrame.
2741: */
2742: if( SQLITE_OK!=(rc = walRestartLog(pWal)) ){
2743: return rc;
2744: }
2745:
2746: /* If this is the first frame written into the log, write the WAL
2747: ** header to the start of the WAL file. See comments at the top of
2748: ** this source file for a description of the WAL header format.
2749: */
2750: iFrame = pWal->hdr.mxFrame;
2751: if( iFrame==0 ){
2752: u8 aWalHdr[WAL_HDRSIZE]; /* Buffer to assemble wal-header in */
2753: u32 aCksum[2]; /* Checksum for wal-header */
2754:
2755: sqlite3Put4byte(&aWalHdr[0], (WAL_MAGIC | SQLITE_BIGENDIAN));
2756: sqlite3Put4byte(&aWalHdr[4], WAL_MAX_VERSION);
2757: sqlite3Put4byte(&aWalHdr[8], szPage);
2758: sqlite3Put4byte(&aWalHdr[12], pWal->nCkpt);
2759: if( pWal->nCkpt==0 ) sqlite3_randomness(8, pWal->hdr.aSalt);
2760: memcpy(&aWalHdr[16], pWal->hdr.aSalt, 8);
2761: walChecksumBytes(1, aWalHdr, WAL_HDRSIZE-2*4, 0, aCksum);
2762: sqlite3Put4byte(&aWalHdr[24], aCksum[0]);
2763: sqlite3Put4byte(&aWalHdr[28], aCksum[1]);
2764:
2765: pWal->szPage = szPage;
2766: pWal->hdr.bigEndCksum = SQLITE_BIGENDIAN;
2767: pWal->hdr.aFrameCksum[0] = aCksum[0];
2768: pWal->hdr.aFrameCksum[1] = aCksum[1];
2769: pWal->truncateOnCommit = 1;
2770:
2771: rc = sqlite3OsWrite(pWal->pWalFd, aWalHdr, sizeof(aWalHdr), 0);
2772: WALTRACE(("WAL%p: wal-header write %s\n", pWal, rc ? "failed" : "ok"));
2773: if( rc!=SQLITE_OK ){
2774: return rc;
2775: }
2776:
2777: /* Sync the header (unless SQLITE_IOCAP_SEQUENTIAL is true or unless
2778: ** all syncing is turned off by PRAGMA synchronous=OFF). Otherwise
2779: ** an out-of-order write following a WAL restart could result in
2780: ** database corruption. See the ticket:
2781: **
2782: ** http://localhost:591/sqlite/info/ff5be73dee
2783: */
2784: if( pWal->syncHeader && sync_flags ){
2785: rc = sqlite3OsSync(pWal->pWalFd, sync_flags & SQLITE_SYNC_MASK);
2786: if( rc ) return rc;
2787: }
2788: }
2789: assert( (int)pWal->szPage==szPage );
2790:
2791: /* Setup information needed to write frames into the WAL */
2792: w.pWal = pWal;
2793: w.pFd = pWal->pWalFd;
2794: w.iSyncPoint = 0;
2795: w.syncFlags = sync_flags;
2796: w.szPage = szPage;
2797: iOffset = walFrameOffset(iFrame+1, szPage);
2798: szFrame = szPage + WAL_FRAME_HDRSIZE;
2799:
2800: /* Write all frames into the log file exactly once */
2801: for(p=pList; p; p=p->pDirty){
2802: int nDbSize; /* 0 normally. Positive == commit flag */
2803: iFrame++;
2804: assert( iOffset==walFrameOffset(iFrame, szPage) );
2805: nDbSize = (isCommit && p->pDirty==0) ? nTruncate : 0;
2806: rc = walWriteOneFrame(&w, p, nDbSize, iOffset);
2807: if( rc ) return rc;
2808: pLast = p;
2809: iOffset += szFrame;
2810: }
2811:
2812: /* If this is the end of a transaction, then we might need to pad
2813: ** the transaction and/or sync the WAL file.
2814: **
2815: ** Padding and syncing only occur if this set of frames complete a
2816: ** transaction and if PRAGMA synchronous=FULL. If synchronous==NORMAL
2817: ** or synchonous==OFF, then no padding or syncing are needed.
2818: **
2819: ** If SQLITE_IOCAP_POWERSAFE_OVERWRITE is defined, then padding is not
2820: ** needed and only the sync is done. If padding is needed, then the
2821: ** final frame is repeated (with its commit mark) until the next sector
2822: ** boundary is crossed. Only the part of the WAL prior to the last
2823: ** sector boundary is synced; the part of the last frame that extends
2824: ** past the sector boundary is written after the sync.
2825: */
2826: if( isCommit && (sync_flags & WAL_SYNC_TRANSACTIONS)!=0 ){
2827: if( pWal->padToSectorBoundary ){
2828: int sectorSize = sqlite3OsSectorSize(pWal->pWalFd);
2829: w.iSyncPoint = ((iOffset+sectorSize-1)/sectorSize)*sectorSize;
2830: while( iOffset<w.iSyncPoint ){
2831: rc = walWriteOneFrame(&w, pLast, nTruncate, iOffset);
2832: if( rc ) return rc;
2833: iOffset += szFrame;
2834: nExtra++;
2835: }
2836: }else{
2837: rc = sqlite3OsSync(w.pFd, sync_flags & SQLITE_SYNC_MASK);
2838: }
2839: }
2840:
2841: /* If this frame set completes the first transaction in the WAL and
2842: ** if PRAGMA journal_size_limit is set, then truncate the WAL to the
2843: ** journal size limit, if possible.
2844: */
2845: if( isCommit && pWal->truncateOnCommit && pWal->mxWalSize>=0 ){
2846: i64 sz = pWal->mxWalSize;
2847: if( walFrameOffset(iFrame+nExtra+1, szPage)>pWal->mxWalSize ){
2848: sz = walFrameOffset(iFrame+nExtra+1, szPage);
2849: }
2850: walLimitSize(pWal, sz);
2851: pWal->truncateOnCommit = 0;
2852: }
2853:
2854: /* Append data to the wal-index. It is not necessary to lock the
2855: ** wal-index to do this as the SQLITE_SHM_WRITE lock held on the wal-index
2856: ** guarantees that there are no other writers, and no data that may
2857: ** be in use by existing readers is being overwritten.
2858: */
2859: iFrame = pWal->hdr.mxFrame;
2860: for(p=pList; p && rc==SQLITE_OK; p=p->pDirty){
2861: iFrame++;
2862: rc = walIndexAppend(pWal, iFrame, p->pgno);
2863: }
2864: while( rc==SQLITE_OK && nExtra>0 ){
2865: iFrame++;
2866: nExtra--;
2867: rc = walIndexAppend(pWal, iFrame, pLast->pgno);
2868: }
2869:
2870: if( rc==SQLITE_OK ){
2871: /* Update the private copy of the header. */
2872: pWal->hdr.szPage = (u16)((szPage&0xff00) | (szPage>>16));
2873: testcase( szPage<=32768 );
2874: testcase( szPage>=65536 );
2875: pWal->hdr.mxFrame = iFrame;
2876: if( isCommit ){
2877: pWal->hdr.iChange++;
2878: pWal->hdr.nPage = nTruncate;
2879: }
2880: /* If this is a commit, update the wal-index header too. */
2881: if( isCommit ){
2882: walIndexWriteHdr(pWal);
2883: pWal->iCallback = iFrame;
2884: }
2885: }
2886:
2887: WALTRACE(("WAL%p: frame write %s\n", pWal, rc ? "failed" : "ok"));
2888: return rc;
2889: }
2890:
2891: /*
2892: ** This routine is called to implement sqlite3_wal_checkpoint() and
2893: ** related interfaces.
2894: **
2895: ** Obtain a CHECKPOINT lock and then backfill as much information as
2896: ** we can from WAL into the database.
2897: **
2898: ** If parameter xBusy is not NULL, it is a pointer to a busy-handler
2899: ** callback. In this case this function runs a blocking checkpoint.
2900: */
2901: int sqlite3WalCheckpoint(
2902: Wal *pWal, /* Wal connection */
2903: int eMode, /* PASSIVE, FULL or RESTART */
2904: int (*xBusy)(void*), /* Function to call when busy */
2905: void *pBusyArg, /* Context argument for xBusyHandler */
2906: int sync_flags, /* Flags to sync db file with (or 0) */
2907: int nBuf, /* Size of temporary buffer */
2908: u8 *zBuf, /* Temporary buffer to use */
2909: int *pnLog, /* OUT: Number of frames in WAL */
2910: int *pnCkpt /* OUT: Number of backfilled frames in WAL */
2911: ){
2912: int rc; /* Return code */
2913: int isChanged = 0; /* True if a new wal-index header is loaded */
2914: int eMode2 = eMode; /* Mode to pass to walCheckpoint() */
2915:
2916: assert( pWal->ckptLock==0 );
2917: assert( pWal->writeLock==0 );
2918:
2919: if( pWal->readOnly ) return SQLITE_READONLY;
2920: WALTRACE(("WAL%p: checkpoint begins\n", pWal));
2921: rc = walLockExclusive(pWal, WAL_CKPT_LOCK, 1);
2922: if( rc ){
2923: /* Usually this is SQLITE_BUSY meaning that another thread or process
2924: ** is already running a checkpoint, or maybe a recovery. But it might
2925: ** also be SQLITE_IOERR. */
2926: return rc;
2927: }
2928: pWal->ckptLock = 1;
2929:
2930: /* If this is a blocking-checkpoint, then obtain the write-lock as well
2931: ** to prevent any writers from running while the checkpoint is underway.
2932: ** This has to be done before the call to walIndexReadHdr() below.
2933: **
2934: ** If the writer lock cannot be obtained, then a passive checkpoint is
2935: ** run instead. Since the checkpointer is not holding the writer lock,
2936: ** there is no point in blocking waiting for any readers. Assuming no
2937: ** other error occurs, this function will return SQLITE_BUSY to the caller.
2938: */
2939: if( eMode!=SQLITE_CHECKPOINT_PASSIVE ){
2940: rc = walBusyLock(pWal, xBusy, pBusyArg, WAL_WRITE_LOCK, 1);
2941: if( rc==SQLITE_OK ){
2942: pWal->writeLock = 1;
2943: }else if( rc==SQLITE_BUSY ){
2944: eMode2 = SQLITE_CHECKPOINT_PASSIVE;
2945: rc = SQLITE_OK;
2946: }
2947: }
2948:
2949: /* Read the wal-index header. */
2950: if( rc==SQLITE_OK ){
2951: rc = walIndexReadHdr(pWal, &isChanged);
2952: }
2953:
2954: /* Copy data from the log to the database file. */
2955: if( rc==SQLITE_OK ){
2956: if( pWal->hdr.mxFrame && walPagesize(pWal)!=nBuf ){
2957: rc = SQLITE_CORRUPT_BKPT;
2958: }else{
2959: rc = walCheckpoint(pWal, eMode2, xBusy, pBusyArg, sync_flags, zBuf);
2960: }
2961:
2962: /* If no error occurred, set the output variables. */
2963: if( rc==SQLITE_OK || rc==SQLITE_BUSY ){
2964: if( pnLog ) *pnLog = (int)pWal->hdr.mxFrame;
2965: if( pnCkpt ) *pnCkpt = (int)(walCkptInfo(pWal)->nBackfill);
2966: }
2967: }
2968:
2969: if( isChanged ){
2970: /* If a new wal-index header was loaded before the checkpoint was
2971: ** performed, then the pager-cache associated with pWal is now
2972: ** out of date. So zero the cached wal-index header to ensure that
2973: ** next time the pager opens a snapshot on this database it knows that
2974: ** the cache needs to be reset.
2975: */
2976: memset(&pWal->hdr, 0, sizeof(WalIndexHdr));
2977: }
2978:
2979: /* Release the locks. */
2980: sqlite3WalEndWriteTransaction(pWal);
2981: walUnlockExclusive(pWal, WAL_CKPT_LOCK, 1);
2982: pWal->ckptLock = 0;
2983: WALTRACE(("WAL%p: checkpoint %s\n", pWal, rc ? "failed" : "ok"));
2984: return (rc==SQLITE_OK && eMode!=eMode2 ? SQLITE_BUSY : rc);
2985: }
2986:
2987: /* Return the value to pass to a sqlite3_wal_hook callback, the
2988: ** number of frames in the WAL at the point of the last commit since
2989: ** sqlite3WalCallback() was called. If no commits have occurred since
2990: ** the last call, then return 0.
2991: */
2992: int sqlite3WalCallback(Wal *pWal){
2993: u32 ret = 0;
2994: if( pWal ){
2995: ret = pWal->iCallback;
2996: pWal->iCallback = 0;
2997: }
2998: return (int)ret;
2999: }
3000:
3001: /*
3002: ** This function is called to change the WAL subsystem into or out
3003: ** of locking_mode=EXCLUSIVE.
3004: **
3005: ** If op is zero, then attempt to change from locking_mode=EXCLUSIVE
3006: ** into locking_mode=NORMAL. This means that we must acquire a lock
3007: ** on the pWal->readLock byte. If the WAL is already in locking_mode=NORMAL
3008: ** or if the acquisition of the lock fails, then return 0. If the
3009: ** transition out of exclusive-mode is successful, return 1. This
3010: ** operation must occur while the pager is still holding the exclusive
3011: ** lock on the main database file.
3012: **
3013: ** If op is one, then change from locking_mode=NORMAL into
3014: ** locking_mode=EXCLUSIVE. This means that the pWal->readLock must
3015: ** be released. Return 1 if the transition is made and 0 if the
3016: ** WAL is already in exclusive-locking mode - meaning that this
3017: ** routine is a no-op. The pager must already hold the exclusive lock
3018: ** on the main database file before invoking this operation.
3019: **
3020: ** If op is negative, then do a dry-run of the op==1 case but do
3021: ** not actually change anything. The pager uses this to see if it
3022: ** should acquire the database exclusive lock prior to invoking
3023: ** the op==1 case.
3024: */
3025: int sqlite3WalExclusiveMode(Wal *pWal, int op){
3026: int rc;
3027: assert( pWal->writeLock==0 );
3028: assert( pWal->exclusiveMode!=WAL_HEAPMEMORY_MODE || op==-1 );
3029:
3030: /* pWal->readLock is usually set, but might be -1 if there was a
3031: ** prior error while attempting to acquire are read-lock. This cannot
3032: ** happen if the connection is actually in exclusive mode (as no xShmLock
3033: ** locks are taken in this case). Nor should the pager attempt to
3034: ** upgrade to exclusive-mode following such an error.
3035: */
3036: assert( pWal->readLock>=0 || pWal->lockError );
3037: assert( pWal->readLock>=0 || (op<=0 && pWal->exclusiveMode==0) );
3038:
3039: if( op==0 ){
3040: if( pWal->exclusiveMode ){
3041: pWal->exclusiveMode = 0;
3042: if( walLockShared(pWal, WAL_READ_LOCK(pWal->readLock))!=SQLITE_OK ){
3043: pWal->exclusiveMode = 1;
3044: }
3045: rc = pWal->exclusiveMode==0;
3046: }else{
3047: /* Already in locking_mode=NORMAL */
3048: rc = 0;
3049: }
3050: }else if( op>0 ){
3051: assert( pWal->exclusiveMode==0 );
3052: assert( pWal->readLock>=0 );
3053: walUnlockShared(pWal, WAL_READ_LOCK(pWal->readLock));
3054: pWal->exclusiveMode = 1;
3055: rc = 1;
3056: }else{
3057: rc = pWal->exclusiveMode==0;
3058: }
3059: return rc;
3060: }
3061:
3062: /*
3063: ** Return true if the argument is non-NULL and the WAL module is using
3064: ** heap-memory for the wal-index. Otherwise, if the argument is NULL or the
3065: ** WAL module is using shared-memory, return false.
3066: */
3067: int sqlite3WalHeapMemory(Wal *pWal){
3068: return (pWal && pWal->exclusiveMode==WAL_HEAPMEMORY_MODE );
3069: }
3070:
3071: #endif /* #ifndef SQLITE_OMIT_WAL */
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>
![[BACK]](/icons/cvsweb/back.gif){kind=icon}
Return to wal.c CVS log ![[TXT]](/icons/cvsweb/text.gif){kind=icon}
![[DIR]](/icons/cvsweb/dir.gif){kind=icon}
Up to [ELWIX - Embedded LightWeight unIX -] / embedaddon / sqlite3 / src