1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
|
/*
** 2010 April 7
**
** The author disclaims copyright to this source code. In place of
** a legal notice, here is a blessing:
**
** May you do good and not evil.
** May you find forgiveness for yourself and forgive others.
** May you share freely, never taking more than you give.
**
*************************************************************************
**
** This file implements an example of a simple VFS implementation that
** omits complex features often not required or not possible on embedded
** platforms. Code is included to buffer writes to the journal file,
** which can be a significant performance improvement on some embedded
** platforms.
**
** OVERVIEW
**
** The code in this file implements a minimal SQLite VFS that can be
** used on Linux and other posix-like operating systems. The following
** system calls are used:
**
** File-system: access(), unlink(), getcwd()
** File IO: open(), read(), write(), fsync(), close(), fstat()
** Other: sleep(), usleep(), time()
**
** The following VFS features are omitted:
**
** 1. File locking. The user must ensure that there is at most one
** connection to each database when using this VFS. Multiple
** connections to a single shared-cache count as a single connection
** for the purposes of the previous statement.
**
** 2. The loading of dynamic extensions (shared libraries).
**
** 3. Temporary files. The user must configure SQLite to use in-memory
** temp files when using this VFS. The easiest way to do this is to
** compile with:
**
** -DSQLITE_TEMP_STORE=3
**
** 4. File truncation. As of version 3.6.24, SQLite may run without
** a working xTruncate() call, providing the user does not configure
** SQLite to use "journal_mode=truncate", or use both
** "journal_mode=persist" and ATTACHed databases.
**
** It is assumed that the system uses UNIX-like path-names. Specifically,
** that '/' characters are used to separate path components and that
** a path-name is a relative path unless it begins with a '/'. And that
** no UTF-8 encoded paths are greater than 512 bytes in length.
**
** JOURNAL WRITE-BUFFERING
**
** To commit a transaction to the database, SQLite first writes rollback
** information into the journal file. This usually consists of 4 steps:
**
** 1. The rollback information is sequentially written into the journal
** file, starting at the start of the file.
** 2. The journal file is synced to disk.
** 3. A modification is made to the first few bytes of the journal file.
** 4. The journal file is synced to disk again.
**
** Most of the data is written in step 1 using a series of calls to the
** VFS xWrite() method. The buffers passed to the xWrite() calls are of
** various sizes. For example, as of version 3.6.24, when committing a
** transaction that modifies 3 pages of a database file that uses 4096
** byte pages residing on a media with 512 byte sectors, SQLite makes
** eleven calls to the xWrite() method to create the rollback journal,
** as follows:
**
** Write offset | Bytes written
** ----------------------------
** 0 512
** 512 4
** 516 4096
** 4612 4
** 4616 4
** 4620 4096
** 8716 4
** 8720 4
** 8724 4096
** 12820 4
** ++++++++++++SYNC+++++++++++
** 0 12
** ++++++++++++SYNC+++++++++++
**
** On many operating systems, this is an efficient way to write to a file.
** However, on some embedded systems that do not cache writes in OS
** buffers it is much more efficient to write data in blocks that are
** an integer multiple of the sector-size in size and aligned at the
** start of a sector.
**
** To work around this, the code in this file allocates a fixed size
** buffer of SQLITE_DEMOVFS_BUFFERSZ using sqlite3_malloc() whenever a
** journal file is opened. It uses the buffer to coalesce sequential
** writes into aligned SQLITE_DEMOVFS_BUFFERSZ blocks. When SQLite
** invokes the xSync() method to sync the contents of the file to disk,
** all accumulated data is written out, even if it does not constitute
** a complete block. This means the actual IO to create the rollback
** journal for the example transaction above is this:
**
** Write offset | Bytes written
** ----------------------------
** 0 8192
** 8192 4632
** ++++++++++++SYNC+++++++++++
** 0 12
** ++++++++++++SYNC+++++++++++
**
** Much more efficient if the underlying OS is not caching write
** operations.
*/
#if !defined(SQLITE_TEST) || SQLITE_OS_UNIX
#include <sqlite3.h>
#include <assert.h>
#include <string.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/file.h>
#include <sys/param.h>
#include <unistd.h>
#include <time.h>
#include <errno.h>
#include <fcntl.h>
/*
** Size of the write buffer used by journal files in bytes.
*/
#ifndef SQLITE_DEMOVFS_BUFFERSZ
# define SQLITE_DEMOVFS_BUFFERSZ 8192
#endif
/*
** The maximum pathname length supported by this VFS.
*/
#define MAXPATHNAME 512
/*
** When using this VFS, the sqlite3_file* handles that SQLite uses are
** actually pointers to instances of type DemoFile.
*/
typedef struct DemoFile DemoFile;
struct DemoFile {
sqlite3_file base; /* Base class. Must be first. */
int fd; /* File descriptor */
char *aBuffer; /* Pointer to malloc'd buffer */
int nBuffer; /* Valid bytes of data in zBuffer */
sqlite3_int64 iBufferOfst; /* Offset in file of zBuffer[0] */
};
/*
** Write directly to the file passed as the first argument. Even if the
** file has a write-buffer (DemoFile.aBuffer), ignore it.
*/
static int demoDirectWrite(
DemoFile *p, /* File handle */
const void *zBuf, /* Buffer containing data to write */
int iAmt, /* Size of data to write in bytes */
sqlite_int64 iOfst /* File offset to write to */
){
off_t ofst; /* Return value from lseek() */
size_t nWrite; /* Return value from write() */
ofst = lseek(p->fd, iOfst, SEEK_SET);
if( ofst!=iOfst ){
return SQLITE_IOERR_WRITE;
}
nWrite = write(p->fd, zBuf, iAmt);
if( nWrite!=iAmt ){
return SQLITE_IOERR_WRITE;
}
return SQLITE_OK;
}
/*
** Flush the contents of the DemoFile.aBuffer buffer to disk. This is a
** no-op if this particular file does not have a buffer (i.e. it is not
** a journal file) or if the buffer is currently empty.
*/
static int demoFlushBuffer(DemoFile *p){
int rc = SQLITE_OK;
if( p->nBuffer ){
rc = demoDirectWrite(p, p->aBuffer, p->nBuffer, p->iBufferOfst);
p->nBuffer = 0;
}
return rc;
}
/*
** Close a file.
*/
static int demoClose(sqlite3_file *pFile){
int rc;
DemoFile *p = (DemoFile*)pFile;
rc = demoFlushBuffer(p);
sqlite3_free(p->aBuffer);
close(p->fd);
return rc;
}
/*
** Read data from a file.
*/
static int demoRead(
sqlite3_file *pFile,
void *zBuf,
int iAmt,
sqlite_int64 iOfst
){
DemoFile *p = (DemoFile*)pFile;
off_t ofst; /* Return value from lseek() */
int nRead; /* Return value from read() */
int rc; /* Return code from demoFlushBuffer() */
/* Flush any data in the write buffer to disk in case this operation
** is trying to read data the file-region currently cached in the buffer.
** It would be possible to detect this case and possibly save an
** unnecessary write here, but in practice SQLite will rarely read from
** a journal file when there is data cached in the write-buffer.
*/
rc = demoFlushBuffer(p);
if( rc!=SQLITE_OK ){
return rc;
}
ofst = lseek(p->fd, iOfst, SEEK_SET);
if( ofst!=iOfst ){
return SQLITE_IOERR_READ;
}
nRead = read(p->fd, zBuf, iAmt);
if( nRead==iAmt ){
return SQLITE_OK;
}else if( nRead>=0 ){
return SQLITE_IOERR_SHORT_READ;
}
return SQLITE_IOERR_READ;
}
/*
** Write data to a crash-file.
*/
static int demoWrite(
sqlite3_file *pFile,
const void *zBuf,
int iAmt,
sqlite_int64 iOfst
){
DemoFile *p = (DemoFile*)pFile;
if( p->aBuffer ){
char *z = (char *)zBuf; /* Pointer to remaining data to write */
int n = iAmt; /* Number of bytes at z */
sqlite3_int64 i = iOfst; /* File offset to write to */
while( n>0 ){
int nCopy; /* Number of bytes to copy into buffer */
/* If the buffer is full, or if this data is not being written directly
** following the data already buffered, flush the buffer. Flushing
** the buffer is a no-op if it is empty.
*/
if( p->nBuffer==SQLITE_DEMOVFS_BUFFERSZ || p->iBufferOfst+p->nBuffer!=i ){
int rc = demoFlushBuffer(p);
if( rc!=SQLITE_OK ){
return rc;
}
}
assert( p->nBuffer==0 || p->iBufferOfst+p->nBuffer==i );
p->iBufferOfst = i - p->nBuffer;
/* Copy as much data as possible into the buffer. */
nCopy = SQLITE_DEMOVFS_BUFFERSZ - p->nBuffer;
if( nCopy>n ){
nCopy = n;
}
memcpy(&p->aBuffer[p->nBuffer], z, nCopy);
p->nBuffer += nCopy;
n -= nCopy;
i += nCopy;
z += nCopy;
}
}else{
return demoDirectWrite(p, zBuf, iAmt, iOfst);
}
return SQLITE_OK;
}
/*
** Truncate a file. This is a no-op for this VFS (see header comments at
** the top of the file).
*/
static int demoTruncate(sqlite3_file *pFile, sqlite_int64 size){
#if 0
if( ftruncate(((DemoFile *)pFile)->fd, size) ) return SQLITE_IOERR_TRUNCATE;
#endif
return SQLITE_OK;
}
/*
** Sync the contents of the file to the persistent media.
*/
static int demoSync(sqlite3_file *pFile, int flags){
DemoFile *p = (DemoFile*)pFile;
int rc;
rc = demoFlushBuffer(p);
if( rc!=SQLITE_OK ){
return rc;
}
rc = fsync(p->fd);
return (rc==0 ? SQLITE_OK : SQLITE_IOERR_FSYNC);
}
/*
** Write the size of the file in bytes to *pSize.
*/
static int demoFileSize(sqlite3_file *pFile, sqlite_int64 *pSize){
DemoFile *p = (DemoFile*)pFile;
int rc; /* Return code from fstat() call */
struct stat sStat; /* Output of fstat() call */
/* Flush the contents of the buffer to disk. As with the flush in the
** demoRead() method, it would be possible to avoid this and save a write
** here and there. But in practice this comes up so infrequently it is
** not worth the trouble.
*/
rc = demoFlushBuffer(p);
if( rc!=SQLITE_OK ){
return rc;
}
rc = fstat(p->fd, &sStat);
if( rc!=0 ) return SQLITE_IOERR_FSTAT;
*pSize = sStat.st_size;
return SQLITE_OK;
}
/*
** Locking functions. The xLock() and xUnlock() methods are both no-ops.
** The xCheckReservedLock() always indicates that no other process holds
** a reserved lock on the database file. This ensures that if a hot-journal
** file is found in the file-system it is rolled back.
*/
static int demoLock(sqlite3_file *pFile, int eLock){
return SQLITE_OK;
}
static int demoUnlock(sqlite3_file *pFile, int eLock){
return SQLITE_OK;
}
static int demoCheckReservedLock(sqlite3_file *pFile, int *pResOut){
*pResOut = 0;
return SQLITE_OK;
}
/*
** No xFileControl() verbs are implemented by this VFS.
*/
static int demoFileControl(sqlite3_file *pFile, int op, void *pArg){
return SQLITE_OK;
}
/*
** The xSectorSize() and xDeviceCharacteristics() methods. These two
** may return special values allowing SQLite to optimize file-system
** access to some extent. But it is also safe to simply return 0.
*/
static int demoSectorSize(sqlite3_file *pFile){
return 0;
}
static int demoDeviceCharacteristics(sqlite3_file *pFile){
return 0;
}
/*
** Open a file handle.
*/
static int demoOpen(
sqlite3_vfs *pVfs, /* VFS */
const char *zName, /* File to open, or 0 for a temp file */
sqlite3_file *pFile, /* Pointer to DemoFile struct to populate */
int flags, /* Input SQLITE_OPEN_XXX flags */
int *pOutFlags /* Output SQLITE_OPEN_XXX flags (or NULL) */
){
static const sqlite3_io_methods demoio = {
1, /* iVersion */
demoClose, /* xClose */
demoRead, /* xRead */
demoWrite, /* xWrite */
demoTruncate, /* xTruncate */
demoSync, /* xSync */
demoFileSize, /* xFileSize */
demoLock, /* xLock */
demoUnlock, /* xUnlock */
demoCheckReservedLock, /* xCheckReservedLock */
demoFileControl, /* xFileControl */
demoSectorSize, /* xSectorSize */
demoDeviceCharacteristics /* xDeviceCharacteristics */
};
DemoFile *p = (DemoFile*)pFile; /* Populate this structure */
int oflags = 0; /* flags to pass to open() call */
char *aBuf = 0;
if( zName==0 ){
return SQLITE_IOERR;
}
if( flags&SQLITE_OPEN_MAIN_JOURNAL ){
aBuf = (char *)sqlite3_malloc(SQLITE_DEMOVFS_BUFFERSZ);
if( !aBuf ){
return SQLITE_NOMEM;
}
}
if( flags&SQLITE_OPEN_EXCLUSIVE ) oflags |= O_EXCL;
if( flags&SQLITE_OPEN_CREATE ) oflags |= O_CREAT;
if( flags&SQLITE_OPEN_READONLY ) oflags |= O_RDONLY;
if( flags&SQLITE_OPEN_READWRITE ) oflags |= O_RDWR;
memset(p, 0, sizeof(DemoFile));
p->fd = open(zName, oflags, 0600);
if( p->fd<0 ){
sqlite3_free(aBuf);
return SQLITE_CANTOPEN;
}
p->aBuffer = aBuf;
if( pOutFlags ){
*pOutFlags = flags;
}
p->base.pMethods = &demoio;
return SQLITE_OK;
}
/*
** Delete the file identified by argument zPath. If the dirSync parameter
** is non-zero, then ensure the file-system modification to delete the
** file has been synced to disk before returning.
*/
static int demoDelete(sqlite3_vfs *pVfs, const char *zPath, int dirSync){
int rc; /* Return code */
rc = unlink(zPath);
if( rc!=0 && errno==ENOENT ) return SQLITE_OK;
if( rc==0 && dirSync ){
int dfd; /* File descriptor open on directory */
int i; /* Iterator variable */
char zDir[MAXPATHNAME+1]; /* Name of directory containing file zPath */
/* Figure out the directory name from the path of the file deleted. */
sqlite3_snprintf(MAXPATHNAME, zDir, "%s", zPath);
zDir[MAXPATHNAME] = '\0';
for(i=strlen(zDir); i>1 && zDir[i]!='/'; i++);
zDir[i] = '\0';
/* Open a file-descriptor on the directory. Sync. Close. */
dfd = open(zDir, O_RDONLY, 0);
if( dfd<0 ){
rc = -1;
}else{
rc = fsync(dfd);
close(dfd);
}
}
return (rc==0 ? SQLITE_OK : SQLITE_IOERR_DELETE);
}
#ifndef F_OK
# define F_OK 0
#endif
#ifndef R_OK
# define R_OK 4
#endif
#ifndef W_OK
# define W_OK 2
#endif
/*
** Query the file-system to see if the named file exists, is readable or
** is both readable and writable.
*/
static int demoAccess(
sqlite3_vfs *pVfs,
const char *zPath,
int flags,
int *pResOut
){
int rc; /* access() return code */
int eAccess = F_OK; /* Second argument to access() */
assert( flags==SQLITE_ACCESS_EXISTS /* access(zPath, F_OK) */
|| flags==SQLITE_ACCESS_READ /* access(zPath, R_OK) */
|| flags==SQLITE_ACCESS_READWRITE /* access(zPath, R_OK|W_OK) */
);
if( flags==SQLITE_ACCESS_READWRITE ) eAccess = R_OK|W_OK;
if( flags==SQLITE_ACCESS_READ ) eAccess = R_OK;
rc = access(zPath, eAccess);
*pResOut = (rc==0);
return SQLITE_OK;
}
/*
** Argument zPath points to a nul-terminated string containing a file path.
** If zPath is an absolute path, then it is copied as is into the output
** buffer. Otherwise, if it is a relative path, then the equivalent full
** path is written to the output buffer.
**
** This function assumes that paths are UNIX style. Specifically, that:
**
** 1. Path components are separated by a '/'. and
** 2. Full paths begin with a '/' character.
*/
static int demoFullPathname(
sqlite3_vfs *pVfs, /* VFS */
const char *zPath, /* Input path (possibly a relative path) */
int nPathOut, /* Size of output buffer in bytes */
char *zPathOut /* Pointer to output buffer */
){
char zDir[MAXPATHNAME+1];
if( zPath[0]=='/' ){
zDir[0] = '\0';
}else{
if( getcwd(zDir, sizeof(zDir))==0 ) return SQLITE_IOERR;
}
zDir[MAXPATHNAME] = '\0';
sqlite3_snprintf(nPathOut, zPathOut, "%s/%s", zDir, zPath);
zPathOut[nPathOut-1] = '\0';
return SQLITE_OK;
}
/*
** The following four VFS methods:
**
** xDlOpen
** xDlError
** xDlSym
** xDlClose
**
** are supposed to implement the functionality needed by SQLite to load
** extensions compiled as shared objects. This simple VFS does not support
** this functionality, so the following functions are no-ops.
*/
static void *demoDlOpen(sqlite3_vfs *pVfs, const char *zPath){
return 0;
}
static void demoDlError(sqlite3_vfs *pVfs, int nByte, char *zErrMsg){
sqlite3_snprintf(nByte, zErrMsg, "Loadable extensions are not supported");
zErrMsg[nByte-1] = '\0';
}
static void (*demoDlSym(sqlite3_vfs *pVfs, void *pH, const char *z))(void){
return 0;
}
static void demoDlClose(sqlite3_vfs *pVfs, void *pHandle){
return;
}
/*
** Parameter zByte points to a buffer nByte bytes in size. Populate this
** buffer with pseudo-random data.
*/
static int demoRandomness(sqlite3_vfs *pVfs, int nByte, char *zByte){
return SQLITE_OK;
}
/*
** Sleep for at least nMicro microseconds. Return the (approximate) number
** of microseconds slept for.
*/
static int demoSleep(sqlite3_vfs *pVfs, int nMicro){
sleep(nMicro / 1000000);
usleep(nMicro % 1000000);
return nMicro;
}
/*
** Set *pTime to the current UTC time expressed as a Julian day. Return
** SQLITE_OK if successful, or an error code otherwise.
**
** http://en.wikipedia.org/wiki/Julian_day
**
** This implementation is not very good. The current time is rounded to
** an integer number of seconds. Also, assuming time_t is a signed 32-bit
** value, it will stop working some time in the year 2038 AD (the so-called
** "year 2038" problem that afflicts systems that store time this way).
*/
static int demoCurrentTime(sqlite3_vfs *pVfs, double *pTime){
time_t t = time(0);
*pTime = t/86400.0 + 2440587.5;
return SQLITE_OK;
}
/*
** This function returns a pointer to the VFS implemented in this file.
** To make the VFS available to SQLite:
**
** sqlite3_vfs_register(sqlite3_demovfs(), 0);
*/
sqlite3_vfs *sqlite3_demovfs(void){
static sqlite3_vfs demovfs = {
1, /* iVersion */
sizeof(DemoFile), /* szOsFile */
MAXPATHNAME, /* mxPathname */
0, /* pNext */
"demo", /* zName */
0, /* pAppData */
demoOpen, /* xOpen */
demoDelete, /* xDelete */
demoAccess, /* xAccess */
demoFullPathname, /* xFullPathname */
demoDlOpen, /* xDlOpen */
demoDlError, /* xDlError */
demoDlSym, /* xDlSym */
demoDlClose, /* xDlClose */
demoRandomness, /* xRandomness */
demoSleep, /* xSleep */
demoCurrentTime, /* xCurrentTime */
};
return &demovfs;
}
#endif /* !defined(SQLITE_TEST) || SQLITE_OS_UNIX */
#ifdef SQLITE_TEST
#include <tcl.h>
#if SQLITE_OS_UNIX
static int register_demovfs(
ClientData clientData, /* Pointer to sqlite3_enable_XXX function */
Tcl_Interp *interp, /* The TCL interpreter that invoked this command */
int objc, /* Number of arguments */
Tcl_Obj *CONST objv[] /* Command arguments */
){
sqlite3_vfs_register(sqlite3_demovfs(), 1);
return TCL_OK;
}
static int unregister_demovfs(
ClientData clientData, /* Pointer to sqlite3_enable_XXX function */
Tcl_Interp *interp, /* The TCL interpreter that invoked this command */
int objc, /* Number of arguments */
Tcl_Obj *CONST objv[] /* Command arguments */
){
sqlite3_vfs_unregister(sqlite3_demovfs());
return TCL_OK;
}
/*
** Register commands with the TCL interpreter.
*/
int Sqlitetest_demovfs_Init(Tcl_Interp *interp){
Tcl_CreateObjCommand(interp, "register_demovfs", register_demovfs, 0, 0);
Tcl_CreateObjCommand(interp, "unregister_demovfs", unregister_demovfs, 0, 0);
return TCL_OK;
}
#else
int Sqlitetest_demovfs_Init(Tcl_Interp *interp){ return TCL_OK; }
#endif
#endif /* SQLITE_TEST */
|