xref: /sqlite-3.40.0/src/btreeInt.h (revision 7aa3ebee)
1 /*
2 ** 2004 April 6
3 **
4 ** The author disclaims copyright to this source code.  In place of
5 ** a legal notice, here is a blessing:
6 **
7 **    May you do good and not evil.
8 **    May you find forgiveness for yourself and forgive others.
9 **    May you share freely, never taking more than you give.
10 **
11 *************************************************************************
12 ** This file implements an external (disk-based) database using BTrees.
13 ** For a detailed discussion of BTrees, refer to
14 **
15 **     Donald E. Knuth, THE ART OF COMPUTER PROGRAMMING, Volume 3:
16 **     "Sorting And Searching", pages 473-480. Addison-Wesley
17 **     Publishing Company, Reading, Massachusetts.
18 **
19 ** The basic idea is that each page of the file contains N database
20 ** entries and N+1 pointers to subpages.
21 **
22 **   ----------------------------------------------------------------
23 **   |  Ptr(0) | Key(0) | Ptr(1) | Key(1) | ... | Key(N-1) | Ptr(N) |
24 **   ----------------------------------------------------------------
25 **
26 ** All of the keys on the page that Ptr(0) points to have values less
27 ** than Key(0).  All of the keys on page Ptr(1) and its subpages have
28 ** values greater than Key(0) and less than Key(1).  All of the keys
29 ** on Ptr(N) and its subpages have values greater than Key(N-1).  And
30 ** so forth.
31 **
32 ** Finding a particular key requires reading O(log(M)) pages from the
33 ** disk where M is the number of entries in the tree.
34 **
35 ** In this implementation, a single file can hold one or more separate
36 ** BTrees.  Each BTree is identified by the index of its root page.  The
37 ** key and data for any entry are combined to form the "payload".  A
38 ** fixed amount of payload can be carried directly on the database
39 ** page.  If the payload is larger than the preset amount then surplus
40 ** bytes are stored on overflow pages.  The payload for an entry
41 ** and the preceding pointer are combined to form a "Cell".  Each
42 ** page has a small header which contains the Ptr(N) pointer and other
43 ** information such as the size of key and data.
44 **
45 ** FORMAT DETAILS
46 **
47 ** The file is divided into pages.  The first page is called page 1,
48 ** the second is page 2, and so forth.  A page number of zero indicates
49 ** "no such page".  The page size can be any power of 2 between 512 and 65536.
50 ** Each page can be either a btree page, a freelist page, an overflow
51 ** page, or a pointer-map page.
52 **
53 ** The first page is always a btree page.  The first 100 bytes of the first
54 ** page contain a special header (the "file header") that describes the file.
55 ** The format of the file header is as follows:
56 **
57 **   OFFSET   SIZE    DESCRIPTION
58 **      0      16     Header string: "SQLite format 3\000"
59 **     16       2     Page size in bytes.  (1 means 65536)
60 **     18       1     File format write version
61 **     19       1     File format read version
62 **     20       1     Bytes of unused space at the end of each page
63 **     21       1     Max embedded payload fraction (must be 64)
64 **     22       1     Min embedded payload fraction (must be 32)
65 **     23       1     Min leaf payload fraction (must be 32)
66 **     24       4     File change counter
67 **     28       4     Reserved for future use
68 **     32       4     First freelist page
69 **     36       4     Number of freelist pages in the file
70 **     40      60     15 4-byte meta values passed to higher layers
71 **
72 **     40       4     Schema cookie
73 **     44       4     File format of schema layer
74 **     48       4     Size of page cache
75 **     52       4     Largest root-page (auto/incr_vacuum)
76 **     56       4     1=UTF-8 2=UTF16le 3=UTF16be
77 **     60       4     User version
78 **     64       4     Incremental vacuum mode
79 **     68       4     Application-ID
80 **     72      20     unused
81 **     92       4     The version-valid-for number
82 **     96       4     SQLITE_VERSION_NUMBER
83 **
84 ** All of the integer values are big-endian (most significant byte first).
85 **
86 ** The file change counter is incremented when the database is changed
87 ** This counter allows other processes to know when the file has changed
88 ** and thus when they need to flush their cache.
89 **
90 ** The max embedded payload fraction is the amount of the total usable
91 ** space in a page that can be consumed by a single cell for standard
92 ** B-tree (non-LEAFDATA) tables.  A value of 255 means 100%.  The default
93 ** is to limit the maximum cell size so that at least 4 cells will fit
94 ** on one page.  Thus the default max embedded payload fraction is 64.
95 **
96 ** If the payload for a cell is larger than the max payload, then extra
97 ** payload is spilled to overflow pages.  Once an overflow page is allocated,
98 ** as many bytes as possible are moved into the overflow pages without letting
99 ** the cell size drop below the min embedded payload fraction.
100 **
101 ** The min leaf payload fraction is like the min embedded payload fraction
102 ** except that it applies to leaf nodes in a LEAFDATA tree.  The maximum
103 ** payload fraction for a LEAFDATA tree is always 100% (or 255) and it
104 ** not specified in the header.
105 **
106 ** Each btree pages is divided into three sections:  The header, the
107 ** cell pointer array, and the cell content area.  Page 1 also has a 100-byte
108 ** file header that occurs before the page header.
109 **
110 **      |----------------|
111 **      | file header    |   100 bytes.  Page 1 only.
112 **      |----------------|
113 **      | page header    |   8 bytes for leaves.  12 bytes for interior nodes
114 **      |----------------|
115 **      | cell pointer   |   |  2 bytes per cell.  Sorted order.
116 **      | array          |   |  Grows downward
117 **      |                |   v
118 **      |----------------|
119 **      | unallocated    |
120 **      | space          |
121 **      |----------------|   ^  Grows upwards
122 **      | cell content   |   |  Arbitrary order interspersed with freeblocks.
123 **      | area           |   |  and free space fragments.
124 **      |----------------|
125 **
126 ** The page headers looks like this:
127 **
128 **   OFFSET   SIZE     DESCRIPTION
129 **      0       1      Flags. 1: intkey, 2: zerodata, 4: leafdata, 8: leaf
130 **      1       2      byte offset to the first freeblock
131 **      3       2      number of cells on this page
132 **      5       2      first byte of the cell content area
133 **      7       1      number of fragmented free bytes
134 **      8       4      Right child (the Ptr(N) value).  Omitted on leaves.
135 **
136 ** The flags define the format of this btree page.  The leaf flag means that
137 ** this page has no children.  The zerodata flag means that this page carries
138 ** only keys and no data.  The intkey flag means that the key is an integer
139 ** which is stored in the key size entry of the cell header rather than in
140 ** the payload area.
141 **
142 ** The cell pointer array begins on the first byte after the page header.
143 ** The cell pointer array contains zero or more 2-byte numbers which are
144 ** offsets from the beginning of the page to the cell content in the cell
145 ** content area.  The cell pointers occur in sorted order.  The system strives
146 ** to keep free space after the last cell pointer so that new cells can
147 ** be easily added without having to defragment the page.
148 **
149 ** Cell content is stored at the very end of the page and grows toward the
150 ** beginning of the page.
151 **
152 ** Unused space within the cell content area is collected into a linked list of
153 ** freeblocks.  Each freeblock is at least 4 bytes in size.  The byte offset
154 ** to the first freeblock is given in the header.  Freeblocks occur in
155 ** increasing order.  Because a freeblock must be at least 4 bytes in size,
156 ** any group of 3 or fewer unused bytes in the cell content area cannot
157 ** exist on the freeblock chain.  A group of 3 or fewer free bytes is called
158 ** a fragment.  The total number of bytes in all fragments is recorded.
159 ** in the page header at offset 7.
160 **
161 **    SIZE    DESCRIPTION
162 **      2     Byte offset of the next freeblock
163 **      2     Bytes in this freeblock
164 **
165 ** Cells are of variable length.  Cells are stored in the cell content area at
166 ** the end of the page.  Pointers to the cells are in the cell pointer array
167 ** that immediately follows the page header.  Cells is not necessarily
168 ** contiguous or in order, but cell pointers are contiguous and in order.
169 **
170 ** Cell content makes use of variable length integers.  A variable
171 ** length integer is 1 to 9 bytes where the lower 7 bits of each
172 ** byte are used.  The integer consists of all bytes that have bit 8 set and
173 ** the first byte with bit 8 clear.  The most significant byte of the integer
174 ** appears first.  A variable-length integer may not be more than 9 bytes long.
175 ** As a special case, all 8 bytes of the 9th byte are used as data.  This
176 ** allows a 64-bit integer to be encoded in 9 bytes.
177 **
178 **    0x00                      becomes  0x00000000
179 **    0x7f                      becomes  0x0000007f
180 **    0x81 0x00                 becomes  0x00000080
181 **    0x82 0x00                 becomes  0x00000100
182 **    0x80 0x7f                 becomes  0x0000007f
183 **    0x8a 0x91 0xd1 0xac 0x78  becomes  0x12345678
184 **    0x81 0x81 0x81 0x81 0x01  becomes  0x10204081
185 **
186 ** Variable length integers are used for rowids and to hold the number of
187 ** bytes of key and data in a btree cell.
188 **
189 ** The content of a cell looks like this:
190 **
191 **    SIZE    DESCRIPTION
192 **      4     Page number of the left child. Omitted if leaf flag is set.
193 **     var    Number of bytes of data. Omitted if the zerodata flag is set.
194 **     var    Number of bytes of key. Or the key itself if intkey flag is set.
195 **      *     Payload
196 **      4     First page of the overflow chain.  Omitted if no overflow
197 **
198 ** Overflow pages form a linked list.  Each page except the last is completely
199 ** filled with data (pagesize - 4 bytes).  The last page can have as little
200 ** as 1 byte of data.
201 **
202 **    SIZE    DESCRIPTION
203 **      4     Page number of next overflow page
204 **      *     Data
205 **
206 ** Freelist pages come in two subtypes: trunk pages and leaf pages.  The
207 ** file header points to the first in a linked list of trunk page.  Each trunk
208 ** page points to multiple leaf pages.  The content of a leaf page is
209 ** unspecified.  A trunk page looks like this:
210 **
211 **    SIZE    DESCRIPTION
212 **      4     Page number of next trunk page
213 **      4     Number of leaf pointers on this page
214 **      *     zero or more pages numbers of leaves
215 */
216 #include "sqliteInt.h"
217 
218 
219 /* The following value is the maximum cell size assuming a maximum page
220 ** size give above.
221 */
222 #define MX_CELL_SIZE(pBt)  ((int)(pBt->pageSize-8))
223 
224 /* The maximum number of cells on a single page of the database.  This
225 ** assumes a minimum cell size of 6 bytes  (4 bytes for the cell itself
226 ** plus 2 bytes for the index to the cell in the page header).  Such
227 ** small cells will be rare, but they are possible.
228 */
229 #define MX_CELL(pBt) ((pBt->pageSize-8)/6)
230 
231 /* Forward declarations */
232 typedef struct MemPage MemPage;
233 typedef struct BtLock BtLock;
234 typedef struct CellInfo CellInfo;
235 
236 /*
237 ** This is a magic string that appears at the beginning of every
238 ** SQLite database in order to identify the file as a real database.
239 **
240 ** You can change this value at compile-time by specifying a
241 ** -DSQLITE_FILE_HEADER="..." on the compiler command-line.  The
242 ** header must be exactly 16 bytes including the zero-terminator so
243 ** the string itself should be 15 characters long.  If you change
244 ** the header, then your custom library will not be able to read
245 ** databases generated by the standard tools and the standard tools
246 ** will not be able to read databases created by your custom library.
247 */
248 #ifndef SQLITE_FILE_HEADER /* 123456789 123456 */
249 #  define SQLITE_FILE_HEADER "SQLite format 3"
250 #endif
251 
252 /*
253 ** Page type flags.  An ORed combination of these flags appear as the
254 ** first byte of on-disk image of every BTree page.
255 */
256 #define PTF_INTKEY    0x01
257 #define PTF_ZERODATA  0x02
258 #define PTF_LEAFDATA  0x04
259 #define PTF_LEAF      0x08
260 
261 /*
262 ** As each page of the file is loaded into memory, an instance of the following
263 ** structure is appended and initialized to zero.  This structure stores
264 ** information about the page that is decoded from the raw file page.
265 **
266 ** The pParent field points back to the parent page.  This allows us to
267 ** walk up the BTree from any leaf to the root.  Care must be taken to
268 ** unref() the parent page pointer when this page is no longer referenced.
269 ** The pageDestructor() routine handles that chore.
270 **
271 ** Access to all fields of this structure is controlled by the mutex
272 ** stored in MemPage.pBt->mutex.
273 */
274 struct MemPage {
275   u8 isInit;           /* True if previously initialized. MUST BE FIRST! */
276   u8 nOverflow;        /* Number of overflow cell bodies in aCell[] */
277   u8 intKey;           /* True if table b-trees.  False for index b-trees */
278   u8 intKeyLeaf;       /* True if the leaf of an intKey table */
279   u8 leaf;             /* True if a leaf page */
280   u8 hdrOffset;        /* 100 for page 1.  0 otherwise */
281   u8 childPtrSize;     /* 0 if leaf==1.  4 if leaf==0 */
282   u8 max1bytePayload;  /* min(maxLocal,127) */
283   u8 bBusy;            /* Prevent endless loops on corrupt database files */
284   u16 maxLocal;        /* Copy of BtShared.maxLocal or BtShared.maxLeaf */
285   u16 minLocal;        /* Copy of BtShared.minLocal or BtShared.minLeaf */
286   u16 cellOffset;      /* Index in aData of first cell pointer */
287   u16 nFree;           /* Number of free bytes on the page */
288   u16 nCell;           /* Number of cells on this page, local and ovfl */
289   u16 maskPage;        /* Mask for page offset */
290   u16 aiOvfl[5];       /* Insert the i-th overflow cell before the aiOvfl-th
291                        ** non-overflow cell */
292   u8 *apOvfl[5];       /* Pointers to the body of overflow cells */
293   BtShared *pBt;       /* Pointer to BtShared that this page is part of */
294   u8 *aData;           /* Pointer to disk image of the page data */
295   u8 *aDataEnd;        /* One byte past the end of usable data */
296   u8 *aCellIdx;        /* The cell index area */
297   u8 *aDataOfst;       /* Same as aData for leaves.  aData+4 for interior */
298   DbPage *pDbPage;     /* Pager page handle */
299   u16 (*xCellSize)(MemPage*,u8*);             /* cellSizePtr method */
300   void (*xParseCell)(MemPage*,u8*,CellInfo*); /* btreeParseCell method */
301   Pgno pgno;           /* Page number for this page */
302 };
303 
304 /*
305 ** The in-memory image of a disk page has the auxiliary information appended
306 ** to the end.  EXTRA_SIZE is the number of bytes of space needed to hold
307 ** that extra information.
308 */
309 #define EXTRA_SIZE sizeof(MemPage)
310 
311 /*
312 ** A linked list of the following structures is stored at BtShared.pLock.
313 ** Locks are added (or upgraded from READ_LOCK to WRITE_LOCK) when a cursor
314 ** is opened on the table with root page BtShared.iTable. Locks are removed
315 ** from this list when a transaction is committed or rolled back, or when
316 ** a btree handle is closed.
317 */
318 struct BtLock {
319   Btree *pBtree;        /* Btree handle holding this lock */
320   Pgno iTable;          /* Root page of table */
321   u8 eLock;             /* READ_LOCK or WRITE_LOCK */
322   BtLock *pNext;        /* Next in BtShared.pLock list */
323 };
324 
325 /* Candidate values for BtLock.eLock */
326 #define READ_LOCK     1
327 #define WRITE_LOCK    2
328 
329 /* A Btree handle
330 **
331 ** A database connection contains a pointer to an instance of
332 ** this object for every database file that it has open.  This structure
333 ** is opaque to the database connection.  The database connection cannot
334 ** see the internals of this structure and only deals with pointers to
335 ** this structure.
336 **
337 ** For some database files, the same underlying database cache might be
338 ** shared between multiple connections.  In that case, each connection
339 ** has it own instance of this object.  But each instance of this object
340 ** points to the same BtShared object.  The database cache and the
341 ** schema associated with the database file are all contained within
342 ** the BtShared object.
343 **
344 ** All fields in this structure are accessed under sqlite3.mutex.
345 ** The pBt pointer itself may not be changed while there exists cursors
346 ** in the referenced BtShared that point back to this Btree since those
347 ** cursors have to go through this Btree to find their BtShared and
348 ** they often do so without holding sqlite3.mutex.
349 */
350 struct Btree {
351   sqlite3 *db;       /* The database connection holding this btree */
352   BtShared *pBt;     /* Sharable content of this btree */
353   u8 inTrans;        /* TRANS_NONE, TRANS_READ or TRANS_WRITE */
354   u8 sharable;       /* True if we can share pBt with another db */
355   u8 locked;         /* True if db currently has pBt locked */
356   u8 hasIncrblobCur; /* True if there are one or more Incrblob cursors */
357   int wantToLock;    /* Number of nested calls to sqlite3BtreeEnter() */
358   int nBackup;       /* Number of backup operations reading this btree */
359   u32 iDataVersion;  /* Combines with pBt->pPager->iDataVersion */
360   Btree *pNext;      /* List of other sharable Btrees from the same db */
361   Btree *pPrev;      /* Back pointer of the same list */
362 #ifndef SQLITE_OMIT_SHARED_CACHE
363   BtLock lock;       /* Object used to lock page 1 */
364 #endif
365 };
366 
367 /*
368 ** Btree.inTrans may take one of the following values.
369 **
370 ** If the shared-data extension is enabled, there may be multiple users
371 ** of the Btree structure. At most one of these may open a write transaction,
372 ** but any number may have active read transactions.
373 */
374 #define TRANS_NONE  0
375 #define TRANS_READ  1
376 #define TRANS_WRITE 2
377 
378 /*
379 ** An instance of this object represents a single database file.
380 **
381 ** A single database file can be in use at the same time by two
382 ** or more database connections.  When two or more connections are
383 ** sharing the same database file, each connection has it own
384 ** private Btree object for the file and each of those Btrees points
385 ** to this one BtShared object.  BtShared.nRef is the number of
386 ** connections currently sharing this database file.
387 **
388 ** Fields in this structure are accessed under the BtShared.mutex
389 ** mutex, except for nRef and pNext which are accessed under the
390 ** global SQLITE_MUTEX_STATIC_MASTER mutex.  The pPager field
391 ** may not be modified once it is initially set as long as nRef>0.
392 ** The pSchema field may be set once under BtShared.mutex and
393 ** thereafter is unchanged as long as nRef>0.
394 **
395 ** isPending:
396 **
397 **   If a BtShared client fails to obtain a write-lock on a database
398 **   table (because there exists one or more read-locks on the table),
399 **   the shared-cache enters 'pending-lock' state and isPending is
400 **   set to true.
401 **
402 **   The shared-cache leaves the 'pending lock' state when either of
403 **   the following occur:
404 **
405 **     1) The current writer (BtShared.pWriter) concludes its transaction, OR
406 **     2) The number of locks held by other connections drops to zero.
407 **
408 **   while in the 'pending-lock' state, no connection may start a new
409 **   transaction.
410 **
411 **   This feature is included to help prevent writer-starvation.
412 */
413 struct BtShared {
414   Pager *pPager;        /* The page cache */
415   sqlite3 *db;          /* Database connection currently using this Btree */
416   BtCursor *pCursor;    /* A list of all open cursors */
417   MemPage *pPage1;      /* First page of the database */
418   u8 openFlags;         /* Flags to sqlite3BtreeOpen() */
419 #ifndef SQLITE_OMIT_AUTOVACUUM
420   u8 autoVacuum;        /* True if auto-vacuum is enabled */
421   u8 incrVacuum;        /* True if incr-vacuum is enabled */
422   u8 bDoTruncate;       /* True to truncate db on commit */
423 #endif
424   u8 inTransaction;     /* Transaction state */
425   u8 max1bytePayload;   /* Maximum first byte of cell for a 1-byte payload */
426 #ifdef SQLITE_HAS_CODEC
427   u8 optimalReserve;    /* Desired amount of reserved space per page */
428 #endif
429   u16 btsFlags;         /* Boolean parameters.  See BTS_* macros below */
430   u16 maxLocal;         /* Maximum local payload in non-LEAFDATA tables */
431   u16 minLocal;         /* Minimum local payload in non-LEAFDATA tables */
432   u16 maxLeaf;          /* Maximum local payload in a LEAFDATA table */
433   u16 minLeaf;          /* Minimum local payload in a LEAFDATA table */
434   u32 pageSize;         /* Total number of bytes on a page */
435   u32 usableSize;       /* Number of usable bytes on each page */
436   int nTransaction;     /* Number of open transactions (read + write) */
437   u32 nPage;            /* Number of pages in the database */
438   void *pSchema;        /* Pointer to space allocated by sqlite3BtreeSchema() */
439   void (*xFreeSchema)(void*);  /* Destructor for BtShared.pSchema */
440   sqlite3_mutex *mutex; /* Non-recursive mutex required to access this object */
441   Bitvec *pHasContent;  /* Set of pages moved to free-list this transaction */
442 #ifndef SQLITE_OMIT_SHARED_CACHE
443   int nRef;             /* Number of references to this structure */
444   BtShared *pNext;      /* Next on a list of sharable BtShared structs */
445   BtLock *pLock;        /* List of locks held on this shared-btree struct */
446   Btree *pWriter;       /* Btree with currently open write transaction */
447 #endif
448   u8 *pTmpSpace;        /* Temp space sufficient to hold a single cell */
449 };
450 
451 /*
452 ** Allowed values for BtShared.btsFlags
453 */
454 #define BTS_READ_ONLY        0x0001   /* Underlying file is readonly */
455 #define BTS_PAGESIZE_FIXED   0x0002   /* Page size can no longer be changed */
456 #define BTS_SECURE_DELETE    0x0004   /* PRAGMA secure_delete is enabled */
457 #define BTS_INITIALLY_EMPTY  0x0008   /* Database was empty at trans start */
458 #define BTS_NO_WAL           0x0010   /* Do not open write-ahead-log files */
459 #define BTS_EXCLUSIVE        0x0020   /* pWriter has an exclusive lock */
460 #define BTS_PENDING          0x0040   /* Waiting for read-locks to clear */
461 
462 /*
463 ** An instance of the following structure is used to hold information
464 ** about a cell.  The parseCellPtr() function fills in this structure
465 ** based on information extract from the raw disk page.
466 */
467 struct CellInfo {
468   i64 nKey;      /* The key for INTKEY tables, or nPayload otherwise */
469   u8 *pPayload;  /* Pointer to the start of payload */
470   u32 nPayload;  /* Bytes of payload */
471   u16 nLocal;    /* Amount of payload held locally, not on overflow */
472   u16 nSize;     /* Size of the cell content on the main b-tree page */
473 };
474 
475 /*
476 ** Maximum depth of an SQLite B-Tree structure. Any B-Tree deeper than
477 ** this will be declared corrupt. This value is calculated based on a
478 ** maximum database size of 2^31 pages a minimum fanout of 2 for a
479 ** root-node and 3 for all other internal nodes.
480 **
481 ** If a tree that appears to be taller than this is encountered, it is
482 ** assumed that the database is corrupt.
483 */
484 #define BTCURSOR_MAX_DEPTH 20
485 
486 /*
487 ** A cursor is a pointer to a particular entry within a particular
488 ** b-tree within a database file.
489 **
490 ** The entry is identified by its MemPage and the index in
491 ** MemPage.aCell[] of the entry.
492 **
493 ** A single database file can be shared by two more database connections,
494 ** but cursors cannot be shared.  Each cursor is associated with a
495 ** particular database connection identified BtCursor.pBtree.db.
496 **
497 ** Fields in this structure are accessed under the BtShared.mutex
498 ** found at self->pBt->mutex.
499 **
500 ** skipNext meaning:
501 **    eState==SKIPNEXT && skipNext>0:  Next sqlite3BtreeNext() is no-op.
502 **    eState==SKIPNEXT && skipNext<0:  Next sqlite3BtreePrevious() is no-op.
503 **    eState==FAULT:                   Cursor fault with skipNext as error code.
504 */
505 struct BtCursor {
506   Btree *pBtree;            /* The Btree to which this cursor belongs */
507   BtShared *pBt;            /* The BtShared this cursor points to */
508   BtCursor *pNext;          /* Forms a linked list of all cursors */
509   Pgno *aOverflow;          /* Cache of overflow page locations */
510   CellInfo info;            /* A parse of the cell we are pointing at */
511   i64 nKey;                 /* Size of pKey, or last integer key */
512   void *pKey;               /* Saved key that was cursor last known position */
513   Pgno pgnoRoot;            /* The root page of this tree */
514   int nOvflAlloc;           /* Allocated size of aOverflow[] array */
515   int skipNext;    /* Prev() is noop if negative. Next() is noop if positive.
516                    ** Error code if eState==CURSOR_FAULT */
517   u8 curFlags;              /* zero or more BTCF_* flags defined below */
518   u8 curPagerFlags;         /* Flags to send to sqlite3PagerGet() */
519   u8 eState;                /* One of the CURSOR_XXX constants (see below) */
520   u8 hints;                 /* As configured by CursorSetHints() */
521   /* All fields above are zeroed when the cursor is allocated.  See
522   ** sqlite3BtreeCursorZero().  Fields that follow must be manually
523   ** initialized. */
524   i8 iPage;                 /* Index of current page in apPage */
525   u8 curIntKey;             /* Value of apPage[0]->intKey */
526   struct KeyInfo *pKeyInfo; /* Argument passed to comparison function */
527   void *padding1;           /* Make object size a multiple of 16 */
528   u16 aiIdx[BTCURSOR_MAX_DEPTH];        /* Current index in apPage[i] */
529   MemPage *apPage[BTCURSOR_MAX_DEPTH];  /* Pages from root to current page */
530 };
531 
532 /*
533 ** Legal values for BtCursor.curFlags
534 */
535 #define BTCF_WriteFlag    0x01   /* True if a write cursor */
536 #define BTCF_ValidNKey    0x02   /* True if info.nKey is valid */
537 #define BTCF_ValidOvfl    0x04   /* True if aOverflow is valid */
538 #define BTCF_AtLast       0x08   /* Cursor is pointing ot the last entry */
539 #define BTCF_Incrblob     0x10   /* True if an incremental I/O handle */
540 #define BTCF_Multiple     0x20   /* Maybe another cursor on the same btree */
541 
542 /*
543 ** Potential values for BtCursor.eState.
544 **
545 ** CURSOR_INVALID:
546 **   Cursor does not point to a valid entry. This can happen (for example)
547 **   because the table is empty or because BtreeCursorFirst() has not been
548 **   called.
549 **
550 ** CURSOR_VALID:
551 **   Cursor points to a valid entry. getPayload() etc. may be called.
552 **
553 ** CURSOR_SKIPNEXT:
554 **   Cursor is valid except that the Cursor.skipNext field is non-zero
555 **   indicating that the next sqlite3BtreeNext() or sqlite3BtreePrevious()
556 **   operation should be a no-op.
557 **
558 ** CURSOR_REQUIRESEEK:
559 **   The table that this cursor was opened on still exists, but has been
560 **   modified since the cursor was last used. The cursor position is saved
561 **   in variables BtCursor.pKey and BtCursor.nKey. When a cursor is in
562 **   this state, restoreCursorPosition() can be called to attempt to
563 **   seek the cursor to the saved position.
564 **
565 ** CURSOR_FAULT:
566 **   An unrecoverable error (an I/O error or a malloc failure) has occurred
567 **   on a different connection that shares the BtShared cache with this
568 **   cursor.  The error has left the cache in an inconsistent state.
569 **   Do nothing else with this cursor.  Any attempt to use the cursor
570 **   should return the error code stored in BtCursor.skipNext
571 */
572 #define CURSOR_INVALID           0
573 #define CURSOR_VALID             1
574 #define CURSOR_SKIPNEXT          2
575 #define CURSOR_REQUIRESEEK       3
576 #define CURSOR_FAULT             4
577 
578 /*
579 ** The database page the PENDING_BYTE occupies. This page is never used.
580 */
581 # define PENDING_BYTE_PAGE(pBt) PAGER_MJ_PGNO(pBt)
582 
583 /*
584 ** These macros define the location of the pointer-map entry for a
585 ** database page. The first argument to each is the number of usable
586 ** bytes on each page of the database (often 1024). The second is the
587 ** page number to look up in the pointer map.
588 **
589 ** PTRMAP_PAGENO returns the database page number of the pointer-map
590 ** page that stores the required pointer. PTRMAP_PTROFFSET returns
591 ** the offset of the requested map entry.
592 **
593 ** If the pgno argument passed to PTRMAP_PAGENO is a pointer-map page,
594 ** then pgno is returned. So (pgno==PTRMAP_PAGENO(pgsz, pgno)) can be
595 ** used to test if pgno is a pointer-map page. PTRMAP_ISPAGE implements
596 ** this test.
597 */
598 #define PTRMAP_PAGENO(pBt, pgno) ptrmapPageno(pBt, pgno)
599 #define PTRMAP_PTROFFSET(pgptrmap, pgno) (5*(pgno-pgptrmap-1))
600 #define PTRMAP_ISPAGE(pBt, pgno) (PTRMAP_PAGENO((pBt),(pgno))==(pgno))
601 
602 /*
603 ** The pointer map is a lookup table that identifies the parent page for
604 ** each child page in the database file.  The parent page is the page that
605 ** contains a pointer to the child.  Every page in the database contains
606 ** 0 or 1 parent pages.  (In this context 'database page' refers
607 ** to any page that is not part of the pointer map itself.)  Each pointer map
608 ** entry consists of a single byte 'type' and a 4 byte parent page number.
609 ** The PTRMAP_XXX identifiers below are the valid types.
610 **
611 ** The purpose of the pointer map is to facility moving pages from one
612 ** position in the file to another as part of autovacuum.  When a page
613 ** is moved, the pointer in its parent must be updated to point to the
614 ** new location.  The pointer map is used to locate the parent page quickly.
615 **
616 ** PTRMAP_ROOTPAGE: The database page is a root-page. The page-number is not
617 **                  used in this case.
618 **
619 ** PTRMAP_FREEPAGE: The database page is an unused (free) page. The page-number
620 **                  is not used in this case.
621 **
622 ** PTRMAP_OVERFLOW1: The database page is the first page in a list of
623 **                   overflow pages. The page number identifies the page that
624 **                   contains the cell with a pointer to this overflow page.
625 **
626 ** PTRMAP_OVERFLOW2: The database page is the second or later page in a list of
627 **                   overflow pages. The page-number identifies the previous
628 **                   page in the overflow page list.
629 **
630 ** PTRMAP_BTREE: The database page is a non-root btree page. The page number
631 **               identifies the parent page in the btree.
632 */
633 #define PTRMAP_ROOTPAGE 1
634 #define PTRMAP_FREEPAGE 2
635 #define PTRMAP_OVERFLOW1 3
636 #define PTRMAP_OVERFLOW2 4
637 #define PTRMAP_BTREE 5
638 
639 /* A bunch of assert() statements to check the transaction state variables
640 ** of handle p (type Btree*) are internally consistent.
641 */
642 #define btreeIntegrity(p) \
643   assert( p->pBt->inTransaction!=TRANS_NONE || p->pBt->nTransaction==0 ); \
644   assert( p->pBt->inTransaction>=p->inTrans );
645 
646 
647 /*
648 ** The ISAUTOVACUUM macro is used within balance_nonroot() to determine
649 ** if the database supports auto-vacuum or not. Because it is used
650 ** within an expression that is an argument to another macro
651 ** (sqliteMallocRaw), it is not possible to use conditional compilation.
652 ** So, this macro is defined instead.
653 */
654 #ifndef SQLITE_OMIT_AUTOVACUUM
655 #define ISAUTOVACUUM (pBt->autoVacuum)
656 #else
657 #define ISAUTOVACUUM 0
658 #endif
659 
660 
661 /*
662 ** This structure is passed around through all the sanity checking routines
663 ** in order to keep track of some global state information.
664 **
665 ** The aRef[] array is allocated so that there is 1 bit for each page in
666 ** the database. As the integrity-check proceeds, for each page used in
667 ** the database the corresponding bit is set. This allows integrity-check to
668 ** detect pages that are used twice and orphaned pages (both of which
669 ** indicate corruption).
670 */
671 typedef struct IntegrityCk IntegrityCk;
672 struct IntegrityCk {
673   BtShared *pBt;    /* The tree being checked out */
674   Pager *pPager;    /* The associated pager.  Also accessible by pBt->pPager */
675   u8 *aPgRef;       /* 1 bit per page in the db (see above) */
676   Pgno nPage;       /* Number of pages in the database */
677   int mxErr;        /* Stop accumulating errors when this reaches zero */
678   int nErr;         /* Number of messages written to zErrMsg so far */
679   int mallocFailed; /* A memory allocation error has occurred */
680   const char *zPfx; /* Error message prefix */
681   int v1, v2;       /* Values for up to two %d fields in zPfx */
682   StrAccum errMsg;  /* Accumulate the error message text here */
683   u32 *heap;        /* Min-heap used for analyzing cell coverage */
684 };
685 
686 /*
687 ** Routines to read or write a two- and four-byte big-endian integer values.
688 */
689 #define get2byte(x)   ((x)[0]<<8 | (x)[1])
690 #define put2byte(p,v) ((p)[0] = (u8)((v)>>8), (p)[1] = (u8)(v))
691 #define get4byte sqlite3Get4byte
692 #define put4byte sqlite3Put4byte
693 
694 /*
695 ** get2byteAligned(), unlike get2byte(), requires that its argument point to a
696 ** two-byte aligned address.  get2bytea() is only used for accessing the
697 ** cell addresses in a btree header.
698 */
699 #if SQLITE_BYTEORDER==4321
700 # define get2byteAligned(x)  (*(u16*)(x))
701 #elif SQLITE_BYTEORDER==1234 && !defined(SQLITE_DISABLE_INTRINSIC) \
702     && GCC_VERSION>=4008000
703 # define get2byteAligned(x)  __builtin_bswap16(*(u16*)(x))
704 #elif SQLITE_BYTEORDER==1234 && !defined(SQLITE_DISABLE_INTRINSIC) \
705     && defined(_MSC_VER) && _MSC_VER>=1300
706 # define get2byteAligned(x)  _byteswap_ushort(*(u16*)(x))
707 #else
708 # define get2byteAligned(x)  ((x)[0]<<8 | (x)[1])
709 #endif
710