Age Owner Branch data TLA Line data Source code
1 : : /*-------------------------------------------------------------------------
2 : : *
3 : : * xlogreader.c
4 : : * Generic XLog reading facility
5 : : *
6 : : * Portions Copyright (c) 2013-2024, PostgreSQL Global Development Group
7 : : *
8 : : * IDENTIFICATION
9 : : * src/backend/access/transam/xlogreader.c
10 : : *
11 : : * NOTES
12 : : * See xlogreader.h for more notes on this facility.
13 : : *
14 : : * This file is compiled as both front-end and backend code, so it
15 : : * may not use ereport, server-defined static variables, etc.
16 : : *-------------------------------------------------------------------------
17 : : */
18 : : #include "postgres.h"
19 : :
20 : : #include <unistd.h>
21 : : #ifdef USE_LZ4
22 : : #include <lz4.h>
23 : : #endif
24 : : #ifdef USE_ZSTD
25 : : #include <zstd.h>
26 : : #endif
27 : :
28 : : #include "access/transam.h"
29 : : #include "access/xlog_internal.h"
30 : : #include "access/xlogreader.h"
31 : : #include "access/xlogrecord.h"
32 : : #include "catalog/pg_control.h"
33 : : #include "common/pg_lzcompress.h"
34 : : #include "replication/origin.h"
35 : :
36 : : #ifndef FRONTEND
37 : : #include "pgstat.h"
38 : : #else
39 : : #include "common/logging.h"
40 : : #endif
41 : :
42 : : static void report_invalid_record(XLogReaderState *state, const char *fmt,...)
43 : : pg_attribute_printf(2, 3);
44 : : static void allocate_recordbuf(XLogReaderState *state, uint32 reclength);
45 : : static int ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr,
46 : : int reqLen);
47 : : static void XLogReaderInvalReadState(XLogReaderState *state);
48 : : static XLogPageReadResult XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking);
49 : : static bool ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr,
50 : : XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess);
51 : : static bool ValidXLogRecord(XLogReaderState *state, XLogRecord *record,
52 : : XLogRecPtr recptr);
53 : : static void ResetDecoder(XLogReaderState *state);
54 : : static void WALOpenSegmentInit(WALOpenSegment *seg, WALSegmentContext *segcxt,
55 : : int segsize, const char *waldir);
56 : :
57 : : /* size of the buffer allocated for error message. */
58 : : #define MAX_ERRORMSG_LEN 1000
59 : :
60 : : /*
61 : : * Default size; large enough that typical users of XLogReader won't often need
62 : : * to use the 'oversized' memory allocation code path.
63 : : */
64 : : #define DEFAULT_DECODE_BUFFER_SIZE (64 * 1024)
65 : :
66 : : /*
67 : : * Construct a string in state->errormsg_buf explaining what's wrong with
68 : : * the current record being read.
69 : : */
70 : : static void
4106 alvherre@alvh.no-ip. 71 :CBC 348 : report_invalid_record(XLogReaderState *state, const char *fmt,...)
72 : : {
73 : : va_list args;
74 : :
75 : 348 : fmt = _(fmt);
76 : :
77 : 348 : va_start(args, fmt);
78 : 348 : vsnprintf(state->errormsg_buf, MAX_ERRORMSG_LEN, fmt, args);
79 : 348 : va_end(args);
80 : :
758 tmunro@postgresql.or 81 : 348 : state->errormsg_deferred = true;
82 : 348 : }
83 : :
84 : : /*
85 : : * Set the size of the decoding buffer. A pointer to a caller supplied memory
86 : : * region may also be passed in, in which case non-oversized records will be
87 : : * decoded there.
88 : : */
89 : : void
90 : 823 : XLogReaderSetDecodeBuffer(XLogReaderState *state, void *buffer, size_t size)
91 : : {
92 [ - + ]: 823 : Assert(state->decode_buffer == NULL);
93 : :
94 : 823 : state->decode_buffer = buffer;
95 : 823 : state->decode_buffer_size = size;
96 : 823 : state->decode_buffer_tail = buffer;
97 : 823 : state->decode_buffer_head = buffer;
4106 alvherre@alvh.no-ip. 98 : 823 : }
99 : :
100 : : /*
101 : : * Allocate and initialize a new XLogReader.
102 : : *
103 : : * Returns NULL if the xlogreader couldn't be allocated.
104 : : */
105 : : XLogReaderState *
1664 106 : 2721 : XLogReaderAllocate(int wal_segment_size, const char *waldir,
107 : : XLogReaderRoutine *routine, void *private_data)
108 : : {
109 : : XLogReaderState *state;
110 : :
111 : : state = (XLogReaderState *)
3299 fujii@postgresql.org 112 : 2721 : palloc_extended(sizeof(XLogReaderState),
113 : : MCXT_ALLOC_NO_OOM | MCXT_ALLOC_ZERO);
114 [ - + ]: 2721 : if (!state)
3299 fujii@postgresql.org 115 :UBC 0 : return NULL;
116 : :
117 : : /* initialize caller-provided support functions */
1070 tmunro@postgresql.or 118 :CBC 2721 : state->routine = *routine;
119 : :
120 : : /*
121 : : * Permanently allocate readBuf. We do it this way, rather than just
122 : : * making a static array, for two reasons: (1) no need to waste the
123 : : * storage in most instantiations of the backend; (2) a static char array
124 : : * isn't guaranteed to have any particular alignment, whereas
125 : : * palloc_extended() will provide MAXALIGN'd storage.
126 : : */
3299 fujii@postgresql.org 127 : 2721 : state->readBuf = (char *) palloc_extended(XLOG_BLCKSZ,
128 : : MCXT_ALLOC_NO_OOM);
129 [ - + ]: 2721 : if (!state->readBuf)
130 : : {
3299 fujii@postgresql.org 131 :UBC 0 : pfree(state);
132 : 0 : return NULL;
133 : : }
134 : :
135 : : /* Initialize segment info. */
1664 alvherre@alvh.no-ip. 136 :CBC 2721 : WALOpenSegmentInit(&state->seg, &state->segcxt, wal_segment_size,
137 : : waldir);
138 : :
139 : : /* system_identifier initialized to zeroes above */
1070 tmunro@postgresql.or 140 : 2721 : state->private_data = private_data;
141 : : /* ReadRecPtr, EndRecPtr and readLen initialized to zeroes above */
3299 fujii@postgresql.org 142 : 2721 : state->errormsg_buf = palloc_extended(MAX_ERRORMSG_LEN + 1,
143 : : MCXT_ALLOC_NO_OOM);
144 [ - + ]: 2721 : if (!state->errormsg_buf)
145 : : {
3299 fujii@postgresql.org 146 :UBC 0 : pfree(state->readBuf);
147 : 0 : pfree(state);
148 : 0 : return NULL;
149 : : }
4106 alvherre@alvh.no-ip. 150 :CBC 2721 : state->errormsg_buf[0] = '\0';
151 : :
152 : : /*
153 : : * Allocate an initial readRecordBuf of minimal size, which can later be
154 : : * enlarged if necessary.
155 : : */
194 michael@paquier.xyz 156 : 2721 : allocate_recordbuf(state, 0);
4106 alvherre@alvh.no-ip. 157 : 2721 : return state;
158 : : }
159 : :
160 : : void
161 : 2162 : XLogReaderFree(XLogReaderState *state)
162 : : {
1070 tmunro@postgresql.or 163 [ + + ]: 2162 : if (state->seg.ws_file != -1)
164 : 1339 : state->routine.segment_close(state);
165 : :
758 166 [ + + + - ]: 2162 : if (state->decode_buffer && state->free_decode_buffer)
167 : 2107 : pfree(state->decode_buffer);
168 : :
3433 heikki.linnakangas@i 169 : 2162 : pfree(state->errormsg_buf);
4106 alvherre@alvh.no-ip. 170 [ + - ]: 2162 : if (state->readRecordBuf)
3433 heikki.linnakangas@i 171 : 2162 : pfree(state->readRecordBuf);
172 : 2162 : pfree(state->readBuf);
173 : 2162 : pfree(state);
4106 alvherre@alvh.no-ip. 174 : 2162 : }
175 : :
176 : : /*
177 : : * Allocate readRecordBuf to fit a record of at least the given length.
178 : : *
179 : : * readRecordBufSize is set to the new buffer size.
180 : : *
181 : : * To avoid useless small increases, round its size to a multiple of
182 : : * XLOG_BLCKSZ, and make sure it's at least 5*Max(BLCKSZ, XLOG_BLCKSZ) to start
183 : : * with. (That is enough for all "normal" records, but very large commit or
184 : : * abort records might need more space.)
185 : : *
186 : : * Note: This routine should *never* be called for xl_tot_len until the header
187 : : * of the record has been fully validated.
188 : : */
189 : : static void
190 : 2773 : allocate_recordbuf(XLogReaderState *state, uint32 reclength)
191 : : {
192 : 2773 : uint32 newSize = reclength;
193 : :
194 : 2773 : newSize += XLOG_BLCKSZ - (newSize % XLOG_BLCKSZ);
195 : 2773 : newSize = Max(newSize, 5 * Max(BLCKSZ, XLOG_BLCKSZ));
196 : :
197 [ + + ]: 2773 : if (state->readRecordBuf)
3433 heikki.linnakangas@i 198 : 52 : pfree(state->readRecordBuf);
194 michael@paquier.xyz 199 : 2773 : state->readRecordBuf = (char *) palloc(newSize);
4106 alvherre@alvh.no-ip. 200 : 2773 : state->readRecordBufSize = newSize;
201 : 2773 : }
202 : :
203 : : /*
204 : : * Initialize the passed segment structs.
205 : : */
206 : : static void
1664 207 : 2721 : WALOpenSegmentInit(WALOpenSegment *seg, WALSegmentContext *segcxt,
208 : : int segsize, const char *waldir)
209 : : {
210 : 2721 : seg->ws_file = -1;
211 : 2721 : seg->ws_segno = 0;
212 : 2721 : seg->ws_tli = 0;
213 : :
214 : 2721 : segcxt->ws_segsize = segsize;
215 [ + + ]: 2721 : if (waldir)
216 : 105 : snprintf(segcxt->ws_dir, MAXPGPATH, "%s", waldir);
217 : 2721 : }
218 : :
219 : : /*
220 : : * Begin reading WAL at 'RecPtr'.
221 : : *
222 : : * 'RecPtr' should point to the beginning of a valid WAL record. Pointing at
223 : : * the beginning of a page is also OK, if there is a new record right after
224 : : * the page header, i.e. not a continuation.
225 : : *
226 : : * This does not make any attempt to read the WAL yet, and hence cannot fail.
227 : : * If the starting address is not correct, the first call to XLogReadRecord()
228 : : * will error out.
229 : : */
230 : : void
1540 heikki.linnakangas@i 231 : 6087 : XLogBeginRead(XLogReaderState *state, XLogRecPtr RecPtr)
232 : : {
233 [ - + ]: 6087 : Assert(!XLogRecPtrIsInvalid(RecPtr));
234 : :
235 : 6087 : ResetDecoder(state);
236 : :
237 : : /* Begin at the passed-in record pointer. */
238 : 6087 : state->EndRecPtr = RecPtr;
758 tmunro@postgresql.or 239 : 6087 : state->NextRecPtr = RecPtr;
1540 heikki.linnakangas@i 240 : 6087 : state->ReadRecPtr = InvalidXLogRecPtr;
758 tmunro@postgresql.or 241 : 6087 : state->DecodeRecPtr = InvalidXLogRecPtr;
242 : 6087 : }
243 : :
244 : : /*
245 : : * Release the last record that was returned by XLogNextRecord(), if any, to
246 : : * free up space. Returns the LSN past the end of the record.
247 : : */
248 : : XLogRecPtr
249 : 12189213 : XLogReleasePreviousRecord(XLogReaderState *state)
250 : : {
251 : : DecodedXLogRecord *record;
252 : : XLogRecPtr next_lsn;
253 : :
254 [ + + ]: 12189213 : if (!state->record)
584 255 : 6109173 : return InvalidXLogRecPtr;
256 : :
257 : : /*
258 : : * Remove it from the decoded record queue. It must be the oldest item
259 : : * decoded, decode_queue_head.
260 : : */
758 261 : 6080040 : record = state->record;
584 262 : 6080040 : next_lsn = record->next_lsn;
758 263 [ - + ]: 6080040 : Assert(record == state->decode_queue_head);
264 : 6080040 : state->record = NULL;
265 : 6080040 : state->decode_queue_head = record->next;
266 : :
267 : : /* It might also be the newest item decoded, decode_queue_tail. */
268 [ + + ]: 6080040 : if (state->decode_queue_tail == record)
269 : 3124602 : state->decode_queue_tail = NULL;
270 : :
271 : : /* Release the space. */
272 [ + + ]: 6080040 : if (unlikely(record->oversized))
273 : : {
274 : : /* It's not in the decode buffer, so free it to release space. */
275 : 72 : pfree(record);
276 : : }
277 : : else
278 : : {
279 : : /* It must be the head (oldest) record in the decode buffer. */
280 [ - + ]: 6079968 : Assert(state->decode_buffer_head == (char *) record);
281 : :
282 : : /*
283 : : * We need to update head to point to the next record that is in the
284 : : * decode buffer, if any, being careful to skip oversized ones
285 : : * (they're not in the decode buffer).
286 : : */
287 : 6079968 : record = record->next;
288 [ + + - + : 6079968 : while (unlikely(record && record->oversized))
- + ]
758 tmunro@postgresql.or 289 :UBC 0 : record = record->next;
290 : :
758 tmunro@postgresql.or 291 [ + + ]:CBC 6079968 : if (record)
292 : : {
293 : : /* Adjust head to release space up to the next record. */
294 : 2955438 : state->decode_buffer_head = (char *) record;
295 : : }
296 : : else
297 : : {
298 : : /*
299 : : * Otherwise we might as well just reset head and tail to the
300 : : * start of the buffer space, because we're empty. This means
301 : : * we'll keep overwriting the same piece of memory if we're not
302 : : * doing any prefetching.
303 : : */
304 : 3124530 : state->decode_buffer_head = state->decode_buffer;
305 : 3124530 : state->decode_buffer_tail = state->decode_buffer;
306 : : }
307 : : }
308 : :
584 309 : 6080040 : return next_lsn;
310 : : }
311 : :
312 : : /*
313 : : * Attempt to read an XLOG record.
314 : : *
315 : : * XLogBeginRead() or XLogFindNextRecord() and then XLogReadAhead() must be
316 : : * called before the first call to XLogNextRecord(). This functions returns
317 : : * records and errors that were put into an internal queue by XLogReadAhead().
318 : : *
319 : : * On success, a record is returned.
320 : : *
321 : : * The returned record (or *errormsg) points to an internal buffer that's
322 : : * valid until the next call to XLogNextRecord.
323 : : */
324 : : DecodedXLogRecord *
758 325 : 6094478 : XLogNextRecord(XLogReaderState *state, char **errormsg)
326 : : {
327 : : /* Release the last record returned by XLogNextRecord(). */
328 : 6094478 : XLogReleasePreviousRecord(state);
329 : :
330 [ + + ]: 6094478 : if (state->decode_queue_head == NULL)
331 : : {
332 : 8996 : *errormsg = NULL;
333 [ + + ]: 8996 : if (state->errormsg_deferred)
334 : : {
335 [ + + ]: 343 : if (state->errormsg_buf[0] != '\0')
336 : 338 : *errormsg = state->errormsg_buf;
337 : 343 : state->errormsg_deferred = false;
338 : : }
339 : :
340 : : /*
341 : : * state->EndRecPtr is expected to have been set by the last call to
342 : : * XLogBeginRead() or XLogNextRecord(), and is the location of the
343 : : * error.
344 : : */
345 [ - + ]: 8996 : Assert(!XLogRecPtrIsInvalid(state->EndRecPtr));
346 : :
347 : 8996 : return NULL;
348 : : }
349 : :
350 : : /*
351 : : * Record this as the most recent record returned, so that we'll release
352 : : * it next time. This also exposes it to the traditional
353 : : * XLogRecXXX(xlogreader) macros, which work with the decoder rather than
354 : : * the record for historical reasons.
355 : : */
356 : 6085482 : state->record = state->decode_queue_head;
357 : :
358 : : /*
359 : : * Update the pointers to the beginning and one-past-the-end of this
360 : : * record, again for the benefit of historical code that expected the
361 : : * decoder to track this rather than accessing these fields of the record
362 : : * itself.
363 : : */
364 : 6085482 : state->ReadRecPtr = state->record->lsn;
365 : 6085482 : state->EndRecPtr = state->record->next_lsn;
366 : :
367 : 6085482 : *errormsg = NULL;
368 : :
369 : 6085482 : return state->record;
370 : : }
371 : :
372 : : /*
373 : : * Attempt to read an XLOG record.
374 : : *
375 : : * XLogBeginRead() or XLogFindNextRecord() must be called before the first call
376 : : * to XLogReadRecord().
377 : : *
378 : : * If the page_read callback fails to read the requested data, NULL is
379 : : * returned. The callback is expected to have reported the error; errormsg
380 : : * is set to NULL.
381 : : *
382 : : * If the reading fails for some other reason, NULL is also returned, and
383 : : * *errormsg is set to a string with details of the failure.
384 : : *
385 : : * The returned pointer (or *errormsg) points to an internal buffer that's
386 : : * valid until the next call to XLogReadRecord.
387 : : */
388 : : XLogRecord *
1070 389 : 3110620 : XLogReadRecord(XLogReaderState *state, char **errormsg)
390 : : {
391 : : DecodedXLogRecord *decoded;
392 : :
393 : : /*
394 : : * Release last returned record, if there is one. We need to do this so
395 : : * that we can check for empty decode queue accurately.
396 : : */
758 397 : 3110620 : XLogReleasePreviousRecord(state);
398 : :
399 : : /*
400 : : * Call XLogReadAhead() in blocking mode to make sure there is something
401 : : * in the queue, though we don't use the result.
402 : : */
403 [ + - ]: 3110620 : if (!XLogReaderHasQueuedRecordOrError(state))
404 : 3110620 : XLogReadAhead(state, false /* nonblocking */ );
405 : :
406 : : /* Consume the head record or error. */
407 : 3110446 : decoded = XLogNextRecord(state, errormsg);
408 [ + + ]: 3110446 : if (decoded)
409 : : {
410 : : /*
411 : : * This function returns a pointer to the record's header, not the
412 : : * actual decoded record. The caller will access the decoded record
413 : : * through the XLogRecGetXXX() macros, which reach the decoded
414 : : * recorded as xlogreader->record.
415 : : */
416 [ - + ]: 3101807 : Assert(state->record == decoded);
417 : 3101807 : return &decoded->header;
418 : : }
419 : :
420 : 8639 : return NULL;
421 : : }
422 : :
423 : : /*
424 : : * Allocate space for a decoded record. The only member of the returned
425 : : * object that is initialized is the 'oversized' flag, indicating that the
426 : : * decoded record wouldn't fit in the decode buffer and must eventually be
427 : : * freed explicitly.
428 : : *
429 : : * The caller is responsible for adjusting decode_buffer_tail with the real
430 : : * size after successfully decoding a record into this space. This way, if
431 : : * decoding fails, then there is nothing to undo unless the 'oversized' flag
432 : : * was set and pfree() must be called.
433 : : *
434 : : * Return NULL if there is no space in the decode buffer and allow_oversized
435 : : * is false, or if memory allocation fails for an oversized buffer.
436 : : */
437 : : static DecodedXLogRecord *
438 : 6086199 : XLogReadRecordAlloc(XLogReaderState *state, size_t xl_tot_len, bool allow_oversized)
439 : : {
440 : 6086199 : size_t required_space = DecodeXLogRecordRequiredSpace(xl_tot_len);
441 : 6086199 : DecodedXLogRecord *decoded = NULL;
442 : :
443 : : /* Allocate a circular decode buffer if we don't have one already. */
444 [ + + ]: 6086199 : if (unlikely(state->decode_buffer == NULL))
445 : : {
446 [ + + ]: 2396 : if (state->decode_buffer_size == 0)
447 : 1573 : state->decode_buffer_size = DEFAULT_DECODE_BUFFER_SIZE;
448 : 2396 : state->decode_buffer = palloc(state->decode_buffer_size);
449 : 2396 : state->decode_buffer_head = state->decode_buffer;
450 : 2396 : state->decode_buffer_tail = state->decode_buffer;
451 : 2396 : state->free_decode_buffer = true;
452 : : }
453 : :
454 : : /* Try to allocate space in the circular decode buffer. */
455 [ + + ]: 6086199 : if (state->decode_buffer_tail >= state->decode_buffer_head)
456 : : {
457 : : /* Empty, or tail is to the right of head. */
128 458 : 6051760 : if (required_space <=
459 : 6051760 : state->decode_buffer_size -
460 [ + + ]: 6051760 : (state->decode_buffer_tail - state->decode_buffer))
461 : : {
462 : : /*-
463 : : * There is space between tail and end.
464 : : *
465 : : * +-----+--------------------+-----+
466 : : * | |////////////////////|here!|
467 : : * +-----+--------------------+-----+
468 : : * ^ ^
469 : : * | |
470 : : * h t
471 : : */
758 472 : 6050639 : decoded = (DecodedXLogRecord *) state->decode_buffer_tail;
473 : 6050639 : decoded->oversized = false;
474 : 6050639 : return decoded;
475 : : }
128 476 : 1121 : else if (required_space <
477 [ + + ]: 1121 : state->decode_buffer_head - state->decode_buffer)
478 : : {
479 : : /*-
480 : : * There is space between start and head.
481 : : *
482 : : * +-----+--------------------+-----+
483 : : * |here!|////////////////////| |
484 : : * +-----+--------------------+-----+
485 : : * ^ ^
486 : : * | |
487 : : * h t
488 : : */
758 489 : 943 : decoded = (DecodedXLogRecord *) state->decode_buffer;
490 : 943 : decoded->oversized = false;
491 : 943 : return decoded;
492 : : }
493 : : }
494 : : else
495 : : {
496 : : /* Tail is to the left of head. */
128 497 : 34439 : if (required_space <
498 [ + - ]: 34439 : state->decode_buffer_head - state->decode_buffer_tail)
499 : : {
500 : : /*-
501 : : * There is space between tail and head.
502 : : *
503 : : * +-----+--------------------+-----+
504 : : * |/////|here! |/////|
505 : : * +-----+--------------------+-----+
506 : : * ^ ^
507 : : * | |
508 : : * t h
509 : : */
758 510 : 34439 : decoded = (DecodedXLogRecord *) state->decode_buffer_tail;
511 : 34439 : decoded->oversized = false;
512 : 34439 : return decoded;
513 : : }
514 : : }
515 : :
516 : : /* Not enough space in the decode buffer. Are we allowed to allocate? */
517 [ + + ]: 178 : if (allow_oversized)
518 : : {
194 michael@paquier.xyz 519 : 72 : decoded = palloc(required_space);
758 tmunro@postgresql.or 520 : 72 : decoded->oversized = true;
521 : 72 : return decoded;
522 : : }
523 : :
524 : 106 : return NULL;
525 : : }
526 : :
527 : : static XLogPageReadResult
528 : 6108156 : XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking)
529 : : {
530 : : XLogRecPtr RecPtr;
531 : : XLogRecord *record;
532 : : XLogRecPtr targetPagePtr;
533 : : bool randAccess;
534 : : uint32 len,
535 : : total_len;
536 : : uint32 targetRecOff;
537 : : uint32 pageHeaderSize;
538 : : bool assembled;
539 : : bool gotheader;
540 : : int readOff;
541 : : DecodedXLogRecord *decoded;
542 : : char *errormsg; /* not used */
543 : :
544 : : /*
545 : : * randAccess indicates whether to verify the previous-record pointer of
546 : : * the record we're reading. We only do this if we're reading
547 : : * sequentially, which is what we initially assume.
548 : : */
1070 549 : 6108156 : randAccess = false;
550 : :
551 : : /* reset error state */
552 : 6108156 : state->errormsg_buf[0] = '\0';
758 553 : 6108156 : decoded = NULL;
554 : :
928 alvherre@alvh.no-ip. 555 : 6108156 : state->abortedRecPtr = InvalidXLogRecPtr;
556 : 6108156 : state->missingContrecPtr = InvalidXLogRecPtr;
557 : :
758 tmunro@postgresql.or 558 : 6108156 : RecPtr = state->NextRecPtr;
559 : :
560 [ + + ]: 6108156 : if (state->DecodeRecPtr != InvalidXLogRecPtr)
561 : : {
562 : : /* read the record after the one we just read */
563 : :
564 : : /*
565 : : * NextRecPtr is pointing to end+1 of the previous WAL record. If
566 : : * we're at a page boundary, no more records can fit on the current
567 : : * page. We must skip over the page header, but we can't do that until
568 : : * we've read in the page, since the header size is variable.
569 : : */
570 : : }
571 : : else
572 : : {
573 : : /*
574 : : * Caller supplied a position to start at.
575 : : *
576 : : * In this case, NextRecPtr should already be pointing either to a
577 : : * valid record starting position or alternatively to the beginning of
578 : : * a page. See the header comments for XLogBeginRead.
579 : : */
605 rhaas@postgresql.org 580 [ + - - + ]: 6038 : Assert(RecPtr % XLOG_BLCKSZ == 0 || XRecOffIsValid(RecPtr));
1070 tmunro@postgresql.or 581 : 6038 : randAccess = true;
582 : : }
583 : :
928 alvherre@alvh.no-ip. 584 : 6108156 : restart:
758 tmunro@postgresql.or 585 : 6108157 : state->nonblocking = nonblocking;
1070 586 : 6108157 : state->currRecPtr = RecPtr;
928 alvherre@alvh.no-ip. 587 : 6108157 : assembled = false;
588 : :
1070 tmunro@postgresql.or 589 : 6108157 : targetPagePtr = RecPtr - (RecPtr % XLOG_BLCKSZ);
590 : 6108157 : targetRecOff = RecPtr % XLOG_BLCKSZ;
591 : :
592 : : /*
593 : : * Read the page containing the record into state->readBuf. Request enough
594 : : * byte to cover the whole record header, or at least the part of it that
595 : : * fits on the same page.
596 : : */
597 : 6108157 : readOff = ReadPageInternal(state, targetPagePtr,
598 : 6108157 : Min(targetRecOff + SizeOfXLogRecord, XLOG_BLCKSZ));
758 599 [ + + ]: 6107902 : if (readOff == XLREAD_WOULDBLOCK)
600 : 12776 : return XLREAD_WOULDBLOCK;
601 [ + + ]: 6095126 : else if (readOff < 0)
1070 602 : 8670 : goto err;
603 : :
604 : : /*
605 : : * ReadPageInternal always returns at least the page header, so we can
606 : : * examine it now.
607 : : */
608 [ + + ]: 6086456 : pageHeaderSize = XLogPageHeaderSize((XLogPageHeader) state->readBuf);
609 [ + + ]: 6086456 : if (targetRecOff == 0)
610 : : {
611 : : /*
612 : : * At page start, so skip over page header.
613 : : */
614 : 6187 : RecPtr += pageHeaderSize;
615 : 6187 : targetRecOff = pageHeaderSize;
616 : : }
617 [ - + ]: 6080269 : else if (targetRecOff < pageHeaderSize)
618 : : {
409 peter@eisentraut.org 619 :UBC 0 : report_invalid_record(state, "invalid record offset at %X/%X: expected at least %u, got %u",
620 : 0 : LSN_FORMAT_ARGS(RecPtr),
621 : : pageHeaderSize, targetRecOff);
1070 tmunro@postgresql.or 622 : 0 : goto err;
623 : : }
624 : :
1070 tmunro@postgresql.or 625 [ + + - + ]:CBC 6086456 : if ((((XLogPageHeader) state->readBuf)->xlp_info & XLP_FIRST_IS_CONTRECORD) &&
626 : : targetRecOff == pageHeaderSize)
627 : : {
1070 tmunro@postgresql.or 628 :UBC 0 : report_invalid_record(state, "contrecord is requested by %X/%X",
629 : 0 : LSN_FORMAT_ARGS(RecPtr));
630 : 0 : goto err;
631 : : }
632 : :
633 : : /* ReadPageInternal has verified the page header */
1070 tmunro@postgresql.or 634 [ - + ]:CBC 6086456 : Assert(pageHeaderSize <= readOff);
635 : :
636 : : /*
637 : : * Read the record length.
638 : : *
639 : : * NB: Even though we use an XLogRecord pointer here, the whole record
640 : : * header might not fit on this page. xl_tot_len is the first field of the
641 : : * struct, so it must be on this page (the records are MAXALIGNed), but we
642 : : * cannot access any other fields until we've verified that we got the
643 : : * whole header.
644 : : */
645 : 6086456 : record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ);
646 : 6086456 : total_len = record->xl_tot_len;
647 : :
648 : : /*
649 : : * If the whole record header is on this page, validate it immediately.
650 : : * Otherwise do just a basic sanity check on xl_tot_len, and validate the
651 : : * rest of the header after reading it from the next page. The xl_tot_len
652 : : * check is necessary here to ensure that we enter the "Need to reassemble
653 : : * record" code path below; otherwise we might fail to apply
654 : : * ValidXLogRecordHeader at all.
655 : : */
656 [ + + ]: 6086456 : if (targetRecOff <= XLOG_BLCKSZ - SizeOfXLogRecord)
657 : : {
758 658 [ + + ]: 6074426 : if (!ValidXLogRecordHeader(state, RecPtr, state->DecodeRecPtr, record,
659 : : randAccess))
1070 660 : 328 : goto err;
661 : 6074098 : gotheader = true;
662 : : }
663 : : else
664 : : {
665 : : /* There may be no next page if it's too small. */
201 666 [ + + ]: 12030 : if (total_len < SizeOfXLogRecord)
667 : : {
668 : 1 : report_invalid_record(state,
669 : : "invalid record length at %X/%X: expected at least %u, got %u",
670 : 1 : LSN_FORMAT_ARGS(RecPtr),
671 : : (uint32) SizeOfXLogRecord, total_len);
672 : 1 : goto err;
673 : : }
674 : : /* We'll validate the header once we have the next page. */
1070 675 : 12029 : gotheader = false;
676 : : }
677 : :
678 : : /*
679 : : * Try to find space to decode this record, if we can do so without
680 : : * calling palloc. If we can't, we'll try again below after we've
681 : : * validated that total_len isn't garbage bytes from a recycled WAL page.
682 : : */
758 683 : 6086127 : decoded = XLogReadRecordAlloc(state,
684 : : total_len,
685 : : false /* allow_oversized */ );
204 686 [ + + + + ]: 6086127 : if (decoded == NULL && nonblocking)
687 : : {
688 : : /*
689 : : * There is no space in the circular decode buffer, and the caller is
690 : : * only reading ahead. The caller should consume existing records to
691 : : * make space.
692 : : */
693 : 22 : return XLREAD_WOULDBLOCK;
694 : : }
695 : :
1070 696 : 6086105 : len = XLOG_BLCKSZ - RecPtr % XLOG_BLCKSZ;
697 [ + + ]: 6086105 : if (total_len > len)
698 : : {
699 : : /* Need to reassemble record */
700 : : char *contdata;
701 : : XLogPageHeader pageHeader;
702 : : char *buffer;
703 : : uint32 gotlen;
704 : :
928 alvherre@alvh.no-ip. 705 : 117289 : assembled = true;
706 : :
707 : : /*
708 : : * We always have space for a couple of pages, enough to validate a
709 : : * boundary-spanning record header.
710 : : */
204 tmunro@postgresql.or 711 [ - + ]: 117289 : Assert(state->readRecordBufSize >= XLOG_BLCKSZ * 2);
712 [ - + ]: 117289 : Assert(state->readRecordBufSize >= len);
713 : :
714 : : /* Copy the first fragment of the record from the first page. */
1070 715 : 117289 : memcpy(state->readRecordBuf,
716 : 117289 : state->readBuf + RecPtr % XLOG_BLCKSZ, len);
717 : 117289 : buffer = state->readRecordBuf + len;
718 : 117289 : gotlen = len;
719 : :
720 : : do
721 : : {
722 : : /* Calculate pointer to beginning of next page */
723 : 124423 : targetPagePtr += XLOG_BLCKSZ;
724 : :
725 : : /* Wait for the next page to become available */
726 : 124423 : readOff = ReadPageInternal(state, targetPagePtr,
727 : 124423 : Min(total_len - gotlen + SizeOfXLogShortPHD,
728 : : XLOG_BLCKSZ));
729 : :
758 730 [ + + ]: 124421 : if (readOff == XLREAD_WOULDBLOCK)
731 : 320 : return XLREAD_WOULDBLOCK;
732 [ + + ]: 124101 : else if (readOff < 0)
1070 733 : 11 : goto err;
734 : :
735 [ - + ]: 124090 : Assert(SizeOfXLogShortPHD <= readOff);
736 : :
737 : 124090 : pageHeader = (XLogPageHeader) state->readBuf;
738 : :
739 : : /*
740 : : * If we were expecting a continuation record and got an
741 : : * "overwrite contrecord" flag, that means the continuation record
742 : : * was overwritten with a different record. Restart the read by
743 : : * assuming the address to read is the location where we found
744 : : * this flag; but keep track of the LSN of the record we were
745 : : * reading, for later verification.
746 : : */
928 alvherre@alvh.no-ip. 747 [ + + ]: 124090 : if (pageHeader->xlp_info & XLP_FIRST_IS_OVERWRITE_CONTRECORD)
748 : : {
870 749 : 1 : state->overwrittenRecPtr = RecPtr;
928 750 : 1 : RecPtr = targetPagePtr;
751 : 1 : goto restart;
752 : : }
753 : :
754 : : /* Check that the continuation on next page looks valid */
1070 tmunro@postgresql.or 755 [ + + ]: 124089 : if (!(pageHeader->xlp_info & XLP_FIRST_IS_CONTRECORD))
756 : : {
757 : 1 : report_invalid_record(state,
758 : : "there is no contrecord flag at %X/%X",
759 : 1 : LSN_FORMAT_ARGS(RecPtr));
760 : 1 : goto err;
761 : : }
762 : :
763 : : /*
764 : : * Cross-check that xlp_rem_len agrees with how much of the record
765 : : * we expect there to be left.
766 : : */
767 [ + - ]: 124088 : if (pageHeader->xlp_rem_len == 0 ||
768 [ + + ]: 124088 : total_len != (pageHeader->xlp_rem_len + gotlen))
769 : : {
770 : 2 : report_invalid_record(state,
771 : : "invalid contrecord length %u (expected %lld) at %X/%X",
772 : : pageHeader->xlp_rem_len,
773 : 2 : ((long long) total_len) - gotlen,
774 : 2 : LSN_FORMAT_ARGS(RecPtr));
775 : 2 : goto err;
776 : : }
777 : :
778 : : /* Append the continuation from this page to the buffer */
779 [ + + ]: 124086 : pageHeaderSize = XLogPageHeaderSize(pageHeader);
780 : :
781 [ - + ]: 124086 : if (readOff < pageHeaderSize)
1070 tmunro@postgresql.or 782 :UBC 0 : readOff = ReadPageInternal(state, targetPagePtr,
783 : : pageHeaderSize);
784 : :
1070 tmunro@postgresql.or 785 [ - + ]:CBC 124086 : Assert(pageHeaderSize <= readOff);
786 : :
787 : 124086 : contdata = (char *) state->readBuf + pageHeaderSize;
788 : 124086 : len = XLOG_BLCKSZ - pageHeaderSize;
789 [ + + ]: 124086 : if (pageHeader->xlp_rem_len < len)
790 : 116952 : len = pageHeader->xlp_rem_len;
791 : :
792 [ - + ]: 124086 : if (readOff < pageHeaderSize + len)
1070 tmunro@postgresql.or 793 :UBC 0 : readOff = ReadPageInternal(state, targetPagePtr,
794 : 0 : pageHeaderSize + len);
795 : :
1070 tmunro@postgresql.or 796 :CBC 124086 : memcpy(buffer, (char *) contdata, len);
797 : 124086 : buffer += len;
798 : 124086 : gotlen += len;
799 : :
800 : : /* If we just reassembled the record header, validate it. */
801 [ + + ]: 124086 : if (!gotheader)
802 : : {
803 : 11969 : record = (XLogRecord *) state->readRecordBuf;
758 804 [ - + ]: 11969 : if (!ValidXLogRecordHeader(state, RecPtr, state->DecodeRecPtr,
805 : : record, randAccess))
4106 alvherre@alvh.no-ip. 806 :UBC 0 : goto err;
1070 tmunro@postgresql.or 807 :CBC 11969 : gotheader = true;
808 : : }
809 : :
810 : : /*
811 : : * We might need a bigger buffer. We have validated the record
812 : : * header, in the case that it split over a page boundary. We've
813 : : * also cross-checked total_len against xlp_rem_len on the second
814 : : * page, and verified xlp_pageaddr on both.
815 : : */
204 816 [ + + ]: 124086 : if (total_len > state->readRecordBufSize)
817 : : {
818 : : char save_copy[XLOG_BLCKSZ * 2];
819 : :
820 : : /*
821 : : * Save and restore the data we already had. It can't be more
822 : : * than two pages.
823 : : */
824 [ - + ]: 52 : Assert(gotlen <= lengthof(save_copy));
825 [ - + ]: 52 : Assert(gotlen <= state->readRecordBufSize);
826 : 52 : memcpy(save_copy, state->readRecordBuf, gotlen);
194 michael@paquier.xyz 827 : 52 : allocate_recordbuf(state, total_len);
204 tmunro@postgresql.or 828 : 52 : memcpy(state->readRecordBuf, save_copy, gotlen);
829 : 52 : buffer = state->readRecordBuf + gotlen;
830 : : }
831 [ + + ]: 124086 : } while (gotlen < total_len);
1070 832 [ - + ]: 116952 : Assert(gotheader);
833 : :
834 : 116952 : record = (XLogRecord *) state->readRecordBuf;
835 [ - + ]: 116952 : if (!ValidXLogRecord(state, record, RecPtr))
1070 tmunro@postgresql.or 836 :UBC 0 : goto err;
837 : :
1070 tmunro@postgresql.or 838 [ + + ]:CBC 116952 : pageHeaderSize = XLogPageHeaderSize((XLogPageHeader) state->readBuf);
758 839 : 116952 : state->DecodeRecPtr = RecPtr;
840 : 116952 : state->NextRecPtr = targetPagePtr + pageHeaderSize
1070 841 : 116952 : + MAXALIGN(pageHeader->xlp_rem_len);
842 : : }
843 : : else
844 : : {
845 : : /* Wait for the record data to become available */
846 : 5968816 : readOff = ReadPageInternal(state, targetPagePtr,
847 : 5968816 : Min(targetRecOff + total_len, XLOG_BLCKSZ));
758 848 [ - + ]: 5968816 : if (readOff == XLREAD_WOULDBLOCK)
758 tmunro@postgresql.or 849 :UBC 0 : return XLREAD_WOULDBLOCK;
758 tmunro@postgresql.or 850 [ - + ]:CBC 5968816 : else if (readOff < 0)
1070 tmunro@postgresql.or 851 :UBC 0 : goto err;
852 : :
853 : : /* Record does not cross a page boundary */
1070 tmunro@postgresql.or 854 [ + + ]:CBC 5968816 : if (!ValidXLogRecord(state, record, RecPtr))
855 : 1 : goto err;
856 : :
758 857 : 5968815 : state->NextRecPtr = RecPtr + MAXALIGN(total_len);
858 : :
859 : 5968815 : state->DecodeRecPtr = RecPtr;
860 : : }
861 : :
862 : : /*
863 : : * Special processing if it's an XLOG SWITCH record
864 : : */
1102 865 [ + + ]: 6085767 : if (record->xl_rmid == RM_XLOG_ID &&
866 [ + + ]: 75981 : (record->xl_info & ~XLR_INFO_MASK) == XLOG_SWITCH)
867 : : {
868 : : /* Pretend it extends to end of segment */
758 869 : 229 : state->NextRecPtr += state->segcxt.ws_segsize - 1;
870 : 229 : state->NextRecPtr -= XLogSegmentOffset(state->NextRecPtr, state->segcxt.ws_segsize);
871 : : }
872 : :
873 : : /*
874 : : * If we got here without a DecodedXLogRecord, it means we needed to
875 : : * validate total_len before trusting it, but by now we've done that.
876 : : */
204 877 [ + + ]: 6085767 : if (decoded == NULL)
878 : : {
879 [ - + ]: 72 : Assert(!nonblocking);
880 : 72 : decoded = XLogReadRecordAlloc(state,
881 : : total_len,
882 : : true /* allow_oversized */ );
883 : : /* allocation should always happen under allow_oversized */
194 michael@paquier.xyz 884 [ - + ]: 72 : Assert(decoded != NULL);
885 : : }
886 : :
758 tmunro@postgresql.or 887 [ + - ]: 6085767 : if (DecodeXLogRecord(state, decoded, record, RecPtr, &errormsg))
888 : : {
889 : : /* Record the location of the next record. */
890 : 6085767 : decoded->next_lsn = state->NextRecPtr;
891 : :
892 : : /*
893 : : * If it's in the decode buffer, mark the decode buffer space as
894 : : * occupied.
895 : : */
896 [ + + ]: 6085767 : if (!decoded->oversized)
897 : : {
898 : : /* The new decode buffer head must be MAXALIGNed. */
899 [ - + ]: 6085695 : Assert(decoded->size == MAXALIGN(decoded->size));
900 [ + + ]: 6085695 : if ((char *) decoded == state->decode_buffer)
901 : 3130914 : state->decode_buffer_tail = state->decode_buffer + decoded->size;
902 : : else
903 : 2954781 : state->decode_buffer_tail += decoded->size;
904 : : }
905 : :
906 : : /* Insert it into the queue of decoded records. */
907 [ - + ]: 6085767 : Assert(state->decode_queue_tail != decoded);
908 [ + + ]: 6085767 : if (state->decode_queue_tail)
909 : 2955723 : state->decode_queue_tail->next = decoded;
910 : 6085767 : state->decode_queue_tail = decoded;
911 [ + + ]: 6085767 : if (!state->decode_queue_head)
912 : 3130044 : state->decode_queue_head = decoded;
913 : 6085767 : return XLREAD_SUCCESS;
914 : : }
915 : :
4106 alvherre@alvh.no-ip. 916 :UBC 0 : err:
928 alvherre@alvh.no-ip. 917 [ + + ]:CBC 9014 : if (assembled)
918 : : {
919 : : /*
920 : : * We get here when a record that spans multiple pages needs to be
921 : : * assembled, but something went wrong -- perhaps a contrecord piece
922 : : * was lost. If caller is WAL replay, it will know where the aborted
923 : : * record was and where to direct followup WAL to be written, marking
924 : : * the next piece with XLP_FIRST_IS_OVERWRITE_CONTRECORD, which will
925 : : * in turn signal downstream WAL consumers that the broken WAL record
926 : : * is to be ignored.
927 : : */
928 : 14 : state->abortedRecPtr = RecPtr;
929 : 14 : state->missingContrecPtr = targetPagePtr;
930 : :
931 : : /*
932 : : * If we got here without reporting an error, make sure an error is
933 : : * queued so that XLogPrefetcherReadRecord() doesn't bring us back a
934 : : * second time and clobber the above state.
935 : : */
286 tmunro@postgresql.or 936 : 14 : state->errormsg_deferred = true;
937 : : }
938 : :
758 939 [ + + - + ]: 9014 : if (decoded && decoded->oversized)
758 tmunro@postgresql.or 940 :UBC 0 : pfree(decoded);
941 : :
942 : : /*
943 : : * Invalidate the read state. We might read from a different source after
944 : : * failure.
945 : : */
2937 alvherre@alvh.no-ip. 946 :CBC 9014 : XLogReaderInvalReadState(state);
947 : :
948 : : /*
949 : : * If an error was written to errmsg_buf, it'll be returned to the caller
950 : : * of XLogReadRecord() after all successfully decoded records from the
951 : : * read queue.
952 : : */
953 : :
758 tmunro@postgresql.or 954 : 9014 : return XLREAD_FAIL;
955 : : }
956 : :
957 : : /*
958 : : * Try to decode the next available record, and return it. The record will
959 : : * also be returned to XLogNextRecord(), which must be called to 'consume'
960 : : * each record.
961 : : *
962 : : * If nonblocking is true, may return NULL due to lack of data or WAL decoding
963 : : * space.
964 : : */
965 : : DecodedXLogRecord *
966 : 6108156 : XLogReadAhead(XLogReaderState *state, bool nonblocking)
967 : : {
968 : : XLogPageReadResult result;
969 : :
970 [ - + ]: 6108156 : if (state->errormsg_deferred)
758 tmunro@postgresql.or 971 :UBC 0 : return NULL;
972 : :
758 tmunro@postgresql.or 973 :CBC 6108156 : result = XLogDecodeNextRecord(state, nonblocking);
974 [ + + ]: 6107899 : if (result == XLREAD_SUCCESS)
975 : : {
976 [ - + ]: 6085767 : Assert(state->decode_queue_tail != NULL);
977 : 6085767 : return state->decode_queue_tail;
978 : : }
979 : :
1070 980 : 22132 : return NULL;
981 : : }
982 : :
983 : : /*
984 : : * Read a single xlog page including at least [pageptr, reqLen] of valid data
985 : : * via the page_read() callback.
986 : : *
987 : : * Returns XLREAD_FAIL if the required page cannot be read for some
988 : : * reason; errormsg_buf is set in that case (unless the error occurs in the
989 : : * page_read callback).
990 : : *
991 : : * Returns XLREAD_WOULDBLOCK if the requested data can't be read without
992 : : * waiting. This can be returned only if the installed page_read callback
993 : : * respects the state->nonblocking flag, and cannot read the requested data
994 : : * immediately.
995 : : *
996 : : * We fetch the page from a reader-local cache if we know we have the required
997 : : * data and if there hasn't been any error since caching the data.
998 : : */
999 : : static int
1000 : 12201586 : ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr, int reqLen)
1001 : : {
1002 : : int readLen;
1003 : : uint32 targetPageOff;
1004 : : XLogSegNo targetSegNo;
1005 : : XLogPageHeader hdr;
1006 : :
1007 [ - + ]: 12201586 : Assert((pageptr % XLOG_BLCKSZ) == 0);
1008 : :
1009 : 12201586 : XLByteToSeg(pageptr, targetSegNo, state->segcxt.ws_segsize);
1010 : 12201586 : targetPageOff = XLogSegmentOffset(pageptr, state->segcxt.ws_segsize);
1011 : :
1012 : : /* check whether we have all the requested data already */
1013 [ + + ]: 12201586 : if (targetSegNo == state->seg.ws_segno &&
1014 [ + + + + ]: 12190012 : targetPageOff == state->segoff && reqLen <= state->readLen)
1015 : 12030600 : return state->readLen;
1016 : :
1017 : : /*
1018 : : * Invalidate contents of internal buffer before read attempt. Just set
1019 : : * the length to 0, rather than a full XLogReaderInvalReadState(), so we
1020 : : * don't forget the segment we last successfully read.
1021 : : */
589 1022 : 170986 : state->readLen = 0;
1023 : :
1024 : : /*
1025 : : * Data is not in our buffer.
1026 : : *
1027 : : * Every time we actually read the segment, even if we looked at parts of
1028 : : * it before, we need to do verification as the page_read callback might
1029 : : * now be rereading data from a different source.
1030 : : *
1031 : : * Whenever switching to a new WAL segment, we read the first page of the
1032 : : * file and validate its header, even if that's not where the target
1033 : : * record is. This is so that we can check the additional identification
1034 : : * info that is present in the first page's "long" header.
1035 : : */
1070 1036 [ + + + + ]: 170986 : if (targetSegNo != state->seg.ws_segno && targetPageOff != 0)
1037 : : {
1038 : 10767 : XLogRecPtr targetSegmentPtr = pageptr - targetPageOff;
1039 : :
1040 : 10767 : readLen = state->routine.page_read(state, targetSegmentPtr, XLOG_BLCKSZ,
1041 : : state->currRecPtr,
1042 : : state->readBuf);
758 1043 [ - + ]: 10756 : if (readLen == XLREAD_WOULDBLOCK)
758 tmunro@postgresql.or 1044 :UBC 0 : return XLREAD_WOULDBLOCK;
758 tmunro@postgresql.or 1045 [ - + ]:CBC 10756 : else if (readLen < 0)
1070 tmunro@postgresql.or 1046 :UBC 0 : goto err;
1047 : :
1048 : : /* we can be sure to have enough WAL available, we scrolled back */
1070 tmunro@postgresql.or 1049 [ - + ]:CBC 10756 : Assert(readLen == XLOG_BLCKSZ);
1050 : :
1051 [ - + ]: 10756 : if (!XLogReaderValidatePageHeader(state, targetSegmentPtr,
1052 : : state->readBuf))
1070 tmunro@postgresql.or 1053 :UBC 0 : goto err;
1054 : : }
1055 : :
1056 : : /*
1057 : : * First, read the requested data length, but at least a short page header
1058 : : * so that we can validate it.
1059 : : */
1070 tmunro@postgresql.or 1060 :CBC 170975 : readLen = state->routine.page_read(state, pageptr, Max(reqLen, SizeOfXLogShortPHD),
1061 : : state->currRecPtr,
1062 : : state->readBuf);
758 1063 [ + + ]: 170729 : if (readLen == XLREAD_WOULDBLOCK)
1064 : 13096 : return XLREAD_WOULDBLOCK;
1065 [ + + ]: 157633 : else if (readLen < 0)
1070 1066 : 8675 : goto err;
1067 : :
1068 [ - + ]: 148958 : Assert(readLen <= XLOG_BLCKSZ);
1069 : :
1070 : : /* Do we have enough data to check the header length? */
1071 [ - + ]: 148958 : if (readLen <= SizeOfXLogShortPHD)
1070 tmunro@postgresql.or 1072 :UBC 0 : goto err;
1073 : :
1070 tmunro@postgresql.or 1074 [ - + ]:CBC 148958 : Assert(readLen >= reqLen);
1075 : :
1076 : 148958 : hdr = (XLogPageHeader) state->readBuf;
1077 : :
1078 : : /* still not enough */
1079 [ + + - + ]: 148958 : if (readLen < XLogPageHeaderSize(hdr))
1080 : : {
1070 tmunro@postgresql.or 1081 [ # # ]:UBC 0 : readLen = state->routine.page_read(state, pageptr, XLogPageHeaderSize(hdr),
1082 : : state->currRecPtr,
1083 : : state->readBuf);
758 1084 [ # # ]: 0 : if (readLen == XLREAD_WOULDBLOCK)
1085 : 0 : return XLREAD_WOULDBLOCK;
1086 [ # # ]: 0 : else if (readLen < 0)
1070 1087 : 0 : goto err;
1088 : : }
1089 : :
1090 : : /*
1091 : : * Now that we know we have the full header, validate it.
1092 : : */
1070 tmunro@postgresql.or 1093 [ + + ]:CBC 148958 : if (!XLogReaderValidatePageHeader(state, pageptr, (char *) hdr))
1094 : 6 : goto err;
1095 : :
1096 : : /* update read state information */
1097 : 148952 : state->seg.ws_segno = targetSegNo;
1098 : 148952 : state->segoff = targetPageOff;
1099 : 148952 : state->readLen = readLen;
1100 : :
1101 : 148952 : return readLen;
1102 : :
1103 : 8681 : err:
589 1104 : 8681 : XLogReaderInvalReadState(state);
1105 : :
758 1106 : 8681 : return XLREAD_FAIL;
1107 : : }
1108 : :
1109 : : /*
1110 : : * Invalidate the xlogreader's read state to force a re-read.
1111 : : */
1112 : : static void
2937 alvherre@alvh.no-ip. 1113 : 17695 : XLogReaderInvalReadState(XLogReaderState *state)
1114 : : {
1070 tmunro@postgresql.or 1115 : 17695 : state->seg.ws_segno = 0;
1116 : 17695 : state->segoff = 0;
1117 : 17695 : state->readLen = 0;
4106 alvherre@alvh.no-ip. 1118 : 17695 : }
1119 : :
1120 : : /*
1121 : : * Validate an XLOG record header.
1122 : : *
1123 : : * This is just a convenience subroutine to avoid duplicated code in
1124 : : * XLogReadRecord. It's not intended for use from anywhere else.
1125 : : */
1126 : : static bool
1127 : 6086395 : ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr,
1128 : : XLogRecPtr PrevRecPtr, XLogRecord *record,
1129 : : bool randAccess)
1130 : : {
3433 heikki.linnakangas@i 1131 [ + + ]: 6086395 : if (record->xl_tot_len < SizeOfXLogRecord)
1132 : : {
4106 alvherre@alvh.no-ip. 1133 : 320 : report_invalid_record(state,
1134 : : "invalid record length at %X/%X: expected at least %u, got %u",
1146 peter@eisentraut.org 1135 : 320 : LSN_FORMAT_ARGS(RecPtr),
1136 : : (uint32) SizeOfXLogRecord, record->xl_tot_len);
4106 alvherre@alvh.no-ip. 1137 : 320 : return false;
1138 : : }
738 jdavis@postgresql.or 1139 [ + + - + ]: 6086075 : if (!RmgrIdIsValid(record->xl_rmid))
1140 : : {
4106 alvherre@alvh.no-ip. 1141 :UBC 0 : report_invalid_record(state,
1142 : : "invalid resource manager ID %u at %X/%X",
1146 peter@eisentraut.org 1143 : 0 : record->xl_rmid, LSN_FORMAT_ARGS(RecPtr));
4106 alvherre@alvh.no-ip. 1144 : 0 : return false;
1145 : : }
1070 tmunro@postgresql.or 1146 [ + + ]:CBC 6086075 : if (randAccess)
1147 : : {
1148 : : /*
1149 : : * We can't exactly verify the prev-link, but surely it should be less
1150 : : * than the record's own address.
1151 : : */
4106 alvherre@alvh.no-ip. 1152 [ - + ]: 6038 : if (!(record->xl_prev < RecPtr))
1153 : : {
4106 alvherre@alvh.no-ip. 1154 :UBC 0 : report_invalid_record(state,
1155 : : "record with incorrect prev-link %X/%X at %X/%X",
1146 peter@eisentraut.org 1156 : 0 : LSN_FORMAT_ARGS(record->xl_prev),
1157 : 0 : LSN_FORMAT_ARGS(RecPtr));
4106 alvherre@alvh.no-ip. 1158 : 0 : return false;
1159 : : }
1160 : : }
1161 : : else
1162 : : {
1163 : : /*
1164 : : * Record's prev-link should exactly match our previous location. This
1165 : : * check guards against torn WAL pages where a stale but valid-looking
1166 : : * WAL record starts on a sector boundary.
1167 : : */
4106 alvherre@alvh.no-ip. 1168 [ + + ]:CBC 6080037 : if (record->xl_prev != PrevRecPtr)
1169 : : {
1170 : 8 : report_invalid_record(state,
1171 : : "record with incorrect prev-link %X/%X at %X/%X",
1146 peter@eisentraut.org 1172 : 8 : LSN_FORMAT_ARGS(record->xl_prev),
1173 : 8 : LSN_FORMAT_ARGS(RecPtr));
4106 alvherre@alvh.no-ip. 1174 : 8 : return false;
1175 : : }
1176 : : }
1177 : :
1178 : 6086067 : return true;
1179 : : }
1180 : :
1181 : :
1182 : : /*
1183 : : * CRC-check an XLOG record. We do not believe the contents of an XLOG
1184 : : * record (other than to the minimal extent of computing the amount of
1185 : : * data to read in) until we've checked the CRCs.
1186 : : *
1187 : : * We assume all of the record (that is, xl_tot_len bytes) has been read
1188 : : * into memory at *record. Also, ValidXLogRecordHeader() has accepted the
1189 : : * record's header, which means in particular that xl_tot_len is at least
1190 : : * SizeOfXLogRecord.
1191 : : */
1192 : : static bool
1193 : 6085768 : ValidXLogRecord(XLogReaderState *state, XLogRecord *record, XLogRecPtr recptr)
1194 : : {
1195 : : pg_crc32c crc;
1196 : :
201 tmunro@postgresql.or 1197 [ - + ]: 6085768 : Assert(record->xl_tot_len >= SizeOfXLogRecord);
1198 : :
1199 : : /* Calculate the CRC */
3449 heikki.linnakangas@i 1200 : 6085768 : INIT_CRC32C(crc);
3433 1201 : 6085768 : COMP_CRC32C(crc, ((char *) record) + SizeOfXLogRecord, record->xl_tot_len - SizeOfXLogRecord);
1202 : : /* include the record header last */
3449 1203 : 6085768 : COMP_CRC32C(crc, (char *) record, offsetof(XLogRecord, xl_crc));
1204 : 6085768 : FIN_CRC32C(crc);
1205 : :
1206 [ + + ]: 6085768 : if (!EQ_CRC32C(record->xl_crc, crc))
1207 : : {
4106 alvherre@alvh.no-ip. 1208 : 1 : report_invalid_record(state,
1209 : : "incorrect resource manager data checksum in record at %X/%X",
1146 peter@eisentraut.org 1210 : 1 : LSN_FORMAT_ARGS(recptr));
4106 alvherre@alvh.no-ip. 1211 : 1 : return false;
1212 : : }
1213 : :
1214 : 6085767 : return true;
1215 : : }
1216 : :
1217 : : /*
1218 : : * Validate a page header.
1219 : : *
1220 : : * Check if 'phdr' is valid as the header of the XLog page at position
1221 : : * 'recptr'.
1222 : : */
1223 : : bool
2171 heikki.linnakangas@i 1224 : 238127 : XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr,
1225 : : char *phdr)
1226 : : {
1227 : : XLogSegNo segno;
1228 : : int32 offset;
1229 : 238127 : XLogPageHeader hdr = (XLogPageHeader) phdr;
1230 : :
4106 alvherre@alvh.no-ip. 1231 [ - + ]: 238127 : Assert((recptr % XLOG_BLCKSZ) == 0);
1232 : :
1664 1233 : 238127 : XLByteToSeg(recptr, segno, state->segcxt.ws_segsize);
1234 : 238127 : offset = XLogSegmentOffset(recptr, state->segcxt.ws_segsize);
1235 : :
4106 1236 [ + + ]: 238127 : if (hdr->xlp_magic != XLOG_PAGE_MAGIC)
1237 : : {
1238 : : char fname[MAXFNAMELEN];
1239 : :
1664 1240 : 9 : XLogFileName(fname, state->seg.ws_tli, segno, state->segcxt.ws_segsize);
1241 : :
4106 1242 : 9 : report_invalid_record(state,
1243 : : "invalid magic number %04X in WAL segment %s, LSN %X/%X, offset %u",
1244 : 9 : hdr->xlp_magic,
1245 : : fname,
496 michael@paquier.xyz 1246 : 9 : LSN_FORMAT_ARGS(recptr),
1247 : : offset);
4106 alvherre@alvh.no-ip. 1248 : 9 : return false;
1249 : : }
1250 : :
1251 [ + + ]: 238118 : if ((hdr->xlp_info & ~XLP_ALL_FLAGS) != 0)
1252 : : {
1253 : : char fname[MAXFNAMELEN];
1254 : :
1664 1255 : 1 : XLogFileName(fname, state->seg.ws_tli, segno, state->segcxt.ws_segsize);
1256 : :
4106 1257 : 1 : report_invalid_record(state,
1258 : : "invalid info bits %04X in WAL segment %s, LSN %X/%X, offset %u",
1259 : 1 : hdr->xlp_info,
1260 : : fname,
496 michael@paquier.xyz 1261 : 1 : LSN_FORMAT_ARGS(recptr),
1262 : : offset);
4106 alvherre@alvh.no-ip. 1263 : 1 : return false;
1264 : : }
1265 : :
1266 [ + + ]: 238117 : if (hdr->xlp_info & XLP_LONG_HEADER)
1267 : : {
1268 : 12035 : XLogLongPageHeader longhdr = (XLogLongPageHeader) hdr;
1269 : :
1270 [ + + ]: 12035 : if (state->system_identifier &&
1271 [ - + ]: 1926 : longhdr->xlp_sysid != state->system_identifier)
1272 : : {
4106 alvherre@alvh.no-ip. 1273 :UBC 0 : report_invalid_record(state,
1274 : : "WAL file is from different database system: WAL file database system identifier is %llu, pg_control database system identifier is %llu",
1774 peter@eisentraut.org 1275 : 0 : (unsigned long long) longhdr->xlp_sysid,
1276 : 0 : (unsigned long long) state->system_identifier);
4106 alvherre@alvh.no-ip. 1277 : 0 : return false;
1278 : : }
1664 alvherre@alvh.no-ip. 1279 [ - + ]:CBC 12035 : else if (longhdr->xlp_seg_size != state->segcxt.ws_segsize)
1280 : : {
4106 alvherre@alvh.no-ip. 1281 :UBC 0 : report_invalid_record(state,
1282 : : "WAL file is from different database system: incorrect segment size in page header");
1283 : 0 : return false;
1284 : : }
4106 alvherre@alvh.no-ip. 1285 [ - + ]:CBC 12035 : else if (longhdr->xlp_xlog_blcksz != XLOG_BLCKSZ)
1286 : : {
4106 alvherre@alvh.no-ip. 1287 :UBC 0 : report_invalid_record(state,
1288 : : "WAL file is from different database system: incorrect XLOG_BLCKSZ in page header");
1289 : 0 : return false;
1290 : : }
1291 : : }
4106 alvherre@alvh.no-ip. 1292 [ - + ]:CBC 226082 : else if (offset == 0)
1293 : : {
1294 : : char fname[MAXFNAMELEN];
1295 : :
1664 alvherre@alvh.no-ip. 1296 :UBC 0 : XLogFileName(fname, state->seg.ws_tli, segno, state->segcxt.ws_segsize);
1297 : :
1298 : : /* hmm, first page of file doesn't have a long header? */
4106 1299 : 0 : report_invalid_record(state,
1300 : : "invalid info bits %04X in WAL segment %s, LSN %X/%X, offset %u",
1301 : 0 : hdr->xlp_info,
1302 : : fname,
496 michael@paquier.xyz 1303 : 0 : LSN_FORMAT_ARGS(recptr),
1304 : : offset);
4106 alvherre@alvh.no-ip. 1305 : 0 : return false;
1306 : : }
1307 : :
1308 : : /*
1309 : : * Check that the address on the page agrees with what we expected. This
1310 : : * check typically fails when an old WAL segment is recycled, and hasn't
1311 : : * yet been overwritten with new data yet.
1312 : : */
550 michael@paquier.xyz 1313 [ + + ]:CBC 238117 : if (hdr->xlp_pageaddr != recptr)
1314 : : {
1315 : : char fname[MAXFNAMELEN];
1316 : :
1664 alvherre@alvh.no-ip. 1317 : 5 : XLogFileName(fname, state->seg.ws_tli, segno, state->segcxt.ws_segsize);
1318 : :
4106 1319 : 5 : report_invalid_record(state,
1320 : : "unexpected pageaddr %X/%X in WAL segment %s, LSN %X/%X, offset %u",
1146 peter@eisentraut.org 1321 : 5 : LSN_FORMAT_ARGS(hdr->xlp_pageaddr),
1322 : : fname,
496 michael@paquier.xyz 1323 : 5 : LSN_FORMAT_ARGS(recptr),
1324 : : offset);
4106 alvherre@alvh.no-ip. 1325 : 5 : return false;
1326 : : }
1327 : :
1328 : : /*
1329 : : * Since child timelines are always assigned a TLI greater than their
1330 : : * immediate parent's TLI, we should never see TLI go backwards across
1331 : : * successive pages of a consistent WAL sequence.
1332 : : *
1333 : : * Sometimes we re-read a segment that's already been (partially) read. So
1334 : : * we only verify TLIs for pages that are later than the last remembered
1335 : : * LSN.
1336 : : */
1337 [ + + ]: 238112 : if (recptr > state->latestPagePtr)
1338 : : {
1339 [ - + ]: 134967 : if (hdr->xlp_tli < state->latestPageTLI)
1340 : : {
1341 : : char fname[MAXFNAMELEN];
1342 : :
1664 alvherre@alvh.no-ip. 1343 :UBC 0 : XLogFileName(fname, state->seg.ws_tli, segno, state->segcxt.ws_segsize);
1344 : :
4106 1345 : 0 : report_invalid_record(state,
1346 : : "out-of-sequence timeline ID %u (after %u) in WAL segment %s, LSN %X/%X, offset %u",
1347 : : hdr->xlp_tli,
1348 : : state->latestPageTLI,
1349 : : fname,
496 michael@paquier.xyz 1350 : 0 : LSN_FORMAT_ARGS(recptr),
1351 : : offset);
4106 alvherre@alvh.no-ip. 1352 : 0 : return false;
1353 : : }
1354 : : }
4106 alvherre@alvh.no-ip. 1355 :CBC 238112 : state->latestPagePtr = recptr;
1356 : 238112 : state->latestPageTLI = hdr->xlp_tli;
1357 : :
1358 : 238112 : return true;
1359 : : }
1360 : :
1361 : : /*
1362 : : * Forget about an error produced by XLogReaderValidatePageHeader().
1363 : : */
1364 : : void
589 tmunro@postgresql.or 1365 : 9 : XLogReaderResetError(XLogReaderState *state)
1366 : : {
1367 : 9 : state->errormsg_buf[0] = '\0';
1368 : 9 : state->errormsg_deferred = false;
1369 : 9 : }
1370 : :
1371 : : /*
1372 : : * Find the first record with an lsn >= RecPtr.
1373 : : *
1374 : : * This is different from XLogBeginRead() in that RecPtr doesn't need to point
1375 : : * to a valid record boundary. Useful for checking whether RecPtr is a valid
1376 : : * xlog address for reading, and to find the first valid address after some
1377 : : * address when dumping records for debugging purposes.
1378 : : *
1379 : : * This positions the reader, like XLogBeginRead(), so that the next call to
1380 : : * XLogReadRecord() will read the next valid record.
1381 : : */
1382 : : XLogRecPtr
1070 1383 : 95 : XLogFindNextRecord(XLogReaderState *state, XLogRecPtr RecPtr)
1384 : : {
1385 : : XLogRecPtr tmpRecPtr;
1386 : 95 : XLogRecPtr found = InvalidXLogRecPtr;
1387 : : XLogPageHeader header;
1388 : : char *errormsg;
1389 : :
1390 [ - + ]: 95 : Assert(!XLogRecPtrIsInvalid(RecPtr));
1391 : :
1392 : : /* Make sure ReadPageInternal() can't return XLREAD_WOULDBLOCK. */
758 1393 : 95 : state->nonblocking = false;
1394 : :
1395 : : /*
1396 : : * skip over potential continuation data, keeping in mind that it may span
1397 : : * multiple pages
1398 : : */
1070 1399 : 95 : tmpRecPtr = RecPtr;
1400 : : while (true)
2785 fujii@postgresql.org 1401 :UBC 0 : {
1402 : : XLogRecPtr targetPagePtr;
1403 : : int targetRecOff;
1404 : : uint32 pageHeaderSize;
1405 : : int readLen;
1406 : :
1407 : : /*
1408 : : * Compute targetRecOff. It should typically be equal or greater than
1409 : : * short page-header since a valid record can't start anywhere before
1410 : : * that, except when caller has explicitly specified the offset that
1411 : : * falls somewhere there or when we are skipping multi-page
1412 : : * continuation record. It doesn't matter though because
1413 : : * ReadPageInternal() is prepared to handle that and will read at
1414 : : * least short page-header worth of data
1415 : : */
1070 tmunro@postgresql.or 1416 :CBC 95 : targetRecOff = tmpRecPtr % XLOG_BLCKSZ;
1417 : :
1418 : : /* scroll back to page boundary */
1419 : 95 : targetPagePtr = tmpRecPtr - targetRecOff;
1420 : :
1421 : : /* Read the page containing the record */
1422 : 95 : readLen = ReadPageInternal(state, targetPagePtr, targetRecOff);
1423 [ - + ]: 95 : if (readLen < 0)
2785 fujii@postgresql.org 1424 :UBC 0 : goto err;
1425 : :
1070 tmunro@postgresql.or 1426 :CBC 95 : header = (XLogPageHeader) state->readBuf;
1427 : :
2785 fujii@postgresql.org 1428 [ + + ]: 95 : pageHeaderSize = XLogPageHeaderSize(header);
1429 : :
1430 : : /* make sure we have enough data for the page header */
1070 tmunro@postgresql.or 1431 : 95 : readLen = ReadPageInternal(state, targetPagePtr, pageHeaderSize);
1432 [ - + ]: 95 : if (readLen < 0)
1070 tmunro@postgresql.or 1433 :UBC 0 : goto err;
1434 : :
1435 : : /* skip over potential continuation data */
2785 fujii@postgresql.org 1436 [ + + ]:CBC 95 : if (header->xlp_info & XLP_FIRST_IS_CONTRECORD)
1437 : : {
1438 : : /*
1439 : : * If the length of the remaining continuation data is more than
1440 : : * what can fit in this page, the continuation record crosses over
1441 : : * this page. Read the next page and try again. xlp_rem_len in the
1442 : : * next page header will contain the remaining length of the
1443 : : * continuation data
1444 : : *
1445 : : * Note that record headers are MAXALIGN'ed
1446 : : */
1620 1447 [ - + ]: 30 : if (MAXALIGN(header->xlp_rem_len) >= (XLOG_BLCKSZ - pageHeaderSize))
1070 tmunro@postgresql.or 1448 :UBC 0 : tmpRecPtr = targetPagePtr + XLOG_BLCKSZ;
1449 : : else
1450 : : {
1451 : : /*
1452 : : * The previous continuation record ends in this page. Set
1453 : : * tmpRecPtr to point to the first valid record
1454 : : */
1070 tmunro@postgresql.or 1455 :CBC 30 : tmpRecPtr = targetPagePtr + pageHeaderSize
2785 fujii@postgresql.org 1456 : 30 : + MAXALIGN(header->xlp_rem_len);
1457 : 30 : break;
1458 : : }
1459 : : }
1460 : : else
1461 : : {
1070 tmunro@postgresql.or 1462 : 65 : tmpRecPtr = targetPagePtr + pageHeaderSize;
2785 fujii@postgresql.org 1463 : 65 : break;
1464 : : }
1465 : : }
1466 : :
1467 : : /*
1468 : : * we know now that tmpRecPtr is an address pointing to a valid XLogRecord
1469 : : * because either we're at the first record after the beginning of a page
1470 : : * or we just jumped over the remaining data of a continuation.
1471 : : */
1070 tmunro@postgresql.or 1472 : 95 : XLogBeginRead(state, tmpRecPtr);
1473 [ + - ]: 951 : while (XLogReadRecord(state, &errormsg) != NULL)
1474 : : {
1475 : : /* past the record we've found, break out */
1476 [ + + ]: 951 : if (RecPtr <= state->ReadRecPtr)
1477 : : {
1478 : : /* Rewind the reader to the beginning of the last record. */
1479 : 95 : found = state->ReadRecPtr;
1480 : 95 : XLogBeginRead(state, found);
1481 : 95 : return found;
1482 : : }
1483 : : }
1484 : :
4106 alvherre@alvh.no-ip. 1485 :UBC 0 : err:
1070 tmunro@postgresql.or 1486 : 0 : XLogReaderInvalReadState(state);
1487 : :
1488 : 0 : return InvalidXLogRecPtr;
1489 : : }
1490 : :
1491 : : /*
1492 : : * Helper function to ease writing of XLogReaderRoutine->page_read callbacks.
1493 : : * If this function is used, caller must supply a segment_open callback in
1494 : : * 'state', as that is used here.
1495 : : *
1496 : : * Read 'count' bytes into 'buf', starting at location 'startptr', from WAL
1497 : : * fetched from timeline 'tli'.
1498 : : *
1499 : : * Returns true if succeeded, false if an error occurs, in which case
1500 : : * 'errinfo' receives error details.
1501 : : */
1502 : : bool
1437 alvherre@alvh.no-ip. 1503 :CBC 76067 : WALRead(XLogReaderState *state,
1504 : : char *buf, XLogRecPtr startptr, Size count, TimeLineID tli,
1505 : : WALReadError *errinfo)
1506 : : {
1507 : : char *p;
1508 : : XLogRecPtr recptr;
1509 : : Size nbytes;
1510 : :
1602 1511 : 76067 : p = buf;
1512 : 76067 : recptr = startptr;
1513 : 76067 : nbytes = count;
1514 : :
1515 [ + + ]: 152139 : while (nbytes > 0)
1516 : : {
1517 : : uint32 startoff;
1518 : : int segbytes;
1519 : : int readbytes;
1520 : :
1432 1521 : 76075 : startoff = XLogSegmentOffset(recptr, state->segcxt.ws_segsize);
1522 : :
1523 : : /*
1524 : : * If the data we want is not in a segment we have open, close what we
1525 : : * have (if anything) and open the next one, using the caller's
1526 : : * provided segment_open callback.
1527 : : */
1528 [ + + ]: 76075 : if (state->seg.ws_file < 0 ||
1529 [ + + ]: 74295 : !XLByteInSeg(recptr, state->seg.ws_segno, state->segcxt.ws_segsize) ||
1530 [ + + ]: 65764 : tli != state->seg.ws_tli)
1531 : : {
1532 : : XLogSegNo nextSegNo;
1533 : :
1534 [ + + ]: 10313 : if (state->seg.ws_file >= 0)
1070 tmunro@postgresql.or 1535 : 8533 : state->routine.segment_close(state);
1536 : :
1432 alvherre@alvh.no-ip. 1537 : 10313 : XLByteToSeg(recptr, nextSegNo, state->segcxt.ws_segsize);
1070 tmunro@postgresql.or 1538 : 10313 : state->routine.segment_open(state, nextSegNo, &tli);
1539 : :
1540 : : /* This shouldn't happen -- indicates a bug in segment_open */
1432 alvherre@alvh.no-ip. 1541 [ - + ]: 10310 : Assert(state->seg.ws_file >= 0);
1542 : :
1543 : : /* Update the current segment info. */
1544 : 10310 : state->seg.ws_tli = tli;
1545 : 10310 : state->seg.ws_segno = nextSegNo;
1546 : : }
1547 : :
1548 : : /* How many bytes are within this segment? */
1549 [ + + ]: 76072 : if (nbytes > (state->segcxt.ws_segsize - startoff))
1550 : 8 : segbytes = state->segcxt.ws_segsize - startoff;
1551 : : else
1602 1552 : 76064 : segbytes = nbytes;
1553 : :
1554 : : #ifndef FRONTEND
1555 : 59354 : pgstat_report_wait_start(WAIT_EVENT_WAL_READ);
1556 : : #endif
1557 : :
1558 : : /* Reset errno first; eases reporting non-errno-affecting errors */
1559 : 76072 : errno = 0;
563 tmunro@postgresql.or 1560 : 76072 : readbytes = pg_pread(state->seg.ws_file, p, segbytes, (off_t) startoff);
1561 : :
1562 : : #ifndef FRONTEND
1602 alvherre@alvh.no-ip. 1563 : 59354 : pgstat_report_wait_end();
1564 : : #endif
1565 : :
1566 [ - + ]: 76072 : if (readbytes <= 0)
1567 : : {
1602 alvherre@alvh.no-ip. 1568 :UBC 0 : errinfo->wre_errno = errno;
1569 : 0 : errinfo->wre_req = segbytes;
1570 : 0 : errinfo->wre_read = readbytes;
1571 : 0 : errinfo->wre_off = startoff;
1432 1572 : 0 : errinfo->wre_seg = state->seg;
1602 1573 : 0 : return false;
1574 : : }
1575 : :
1576 : : /* Update state for read */
1602 alvherre@alvh.no-ip. 1577 :CBC 76072 : recptr += readbytes;
1578 : 76072 : nbytes -= readbytes;
1579 : 76072 : p += readbytes;
1580 : : }
1581 : :
1582 : 76064 : return true;
1583 : : }
1584 : :
1585 : : /* ----------------------------------------
1586 : : * Functions for decoding the data and block references in a record.
1587 : : * ----------------------------------------
1588 : : */
1589 : :
1590 : : /*
1591 : : * Private function to reset the state, forgetting all decoded records, if we
1592 : : * are asked to move to a new read position.
1593 : : */
1594 : : static void
3433 heikki.linnakangas@i 1595 : 6087 : ResetDecoder(XLogReaderState *state)
1596 : : {
1597 : : DecodedXLogRecord *r;
1598 : :
1599 : : /* Reset the decoded record queue, freeing any oversized records. */
758 tmunro@postgresql.or 1600 [ + + ]: 15717 : while ((r = state->decode_queue_head) != NULL)
1601 : : {
1602 : 3543 : state->decode_queue_head = r->next;
1603 [ + - ]: 3543 : if (r->oversized)
758 tmunro@postgresql.or 1604 :UBC 0 : pfree(r);
1605 : : }
758 tmunro@postgresql.or 1606 :CBC 6087 : state->decode_queue_tail = NULL;
1607 : 6087 : state->decode_queue_head = NULL;
1608 : 6087 : state->record = NULL;
1609 : :
1610 : : /* Reset the decode buffer to empty. */
1611 : 6087 : state->decode_buffer_tail = state->decode_buffer;
1612 : 6087 : state->decode_buffer_head = state->decode_buffer;
1613 : :
1614 : : /* Clear error state. */
1615 : 6087 : state->errormsg_buf[0] = '\0';
1616 : 6087 : state->errormsg_deferred = false;
3433 heikki.linnakangas@i 1617 : 6087 : }
1618 : :
1619 : : /*
1620 : : * Compute the maximum possible amount of padding that could be required to
1621 : : * decode a record, given xl_tot_len from the record's header. This is the
1622 : : * amount of output buffer space that we need to decode a record, though we
1623 : : * might not finish up using it all.
1624 : : *
1625 : : * This computation is pessimistic and assumes the maximum possible number of
1626 : : * blocks, due to lack of better information.
1627 : : */
1628 : : size_t
758 tmunro@postgresql.or 1629 : 12191489 : DecodeXLogRecordRequiredSpace(size_t xl_tot_len)
1630 : : {
1631 : 12191489 : size_t size = 0;
1632 : :
1633 : : /* Account for the fixed size part of the decoded record struct. */
1634 : 12191489 : size += offsetof(DecodedXLogRecord, blocks[0]);
1635 : : /* Account for the flexible blocks array of maximum possible size. */
1636 : 12191489 : size += sizeof(DecodedBkpBlock) * (XLR_MAX_BLOCK_ID + 1);
1637 : : /* Account for all the raw main and block data. */
1638 : 12191489 : size += xl_tot_len;
1639 : : /* We might insert padding before main_data. */
1640 : 12191489 : size += (MAXIMUM_ALIGNOF - 1);
1641 : : /* We might insert padding before each block's data. */
1642 : 12191489 : size += (MAXIMUM_ALIGNOF - 1) * (XLR_MAX_BLOCK_ID + 1);
1643 : : /* We might insert padding at the end. */
1644 : 12191489 : size += (MAXIMUM_ALIGNOF - 1);
1645 : :
1646 : 12191489 : return size;
1647 : : }
1648 : :
1649 : : /*
1650 : : * Decode a record. "decoded" must point to a MAXALIGNed memory area that has
1651 : : * space for at least DecodeXLogRecordRequiredSpace(record) bytes. On
1652 : : * success, decoded->size contains the actual space occupied by the decoded
1653 : : * record, which may turn out to be less.
1654 : : *
1655 : : * Only decoded->oversized member must be initialized already, and will not be
1656 : : * modified. Other members will be initialized as required.
1657 : : *
1658 : : * On error, a human-readable error message is returned in *errormsg, and
1659 : : * the return value is false.
1660 : : */
1661 : : bool
1662 : 6085767 : DecodeXLogRecord(XLogReaderState *state,
1663 : : DecodedXLogRecord *decoded,
1664 : : XLogRecord *record,
1665 : : XLogRecPtr lsn,
1666 : : char **errormsg)
1667 : : {
1668 : : /*
1669 : : * read next _size bytes from record buffer, but check for overrun first.
1670 : : */
1671 : : #define COPY_HEADER_FIELD(_dst, _size) \
1672 : : do { \
1673 : : if (remaining < _size) \
1674 : : goto shortdata_err; \
1675 : : memcpy(_dst, ptr, _size); \
1676 : : ptr += _size; \
1677 : : remaining -= _size; \
1678 : : } while(0)
1679 : :
1680 : : char *ptr;
1681 : : char *out;
1682 : : uint32 remaining;
1683 : : uint32 datatotal;
648 rhaas@postgresql.org 1684 : 6085767 : RelFileLocator *rlocator = NULL;
1685 : : uint8 block_id;
1686 : :
758 tmunro@postgresql.or 1687 : 6085767 : decoded->header = *record;
1688 : 6085767 : decoded->lsn = lsn;
1689 : 6085767 : decoded->next = NULL;
1690 : 6085767 : decoded->record_origin = InvalidRepOriginId;
1691 : 6085767 : decoded->toplevel_xid = InvalidTransactionId;
1692 : 6085767 : decoded->main_data = NULL;
1693 : 6085767 : decoded->main_data_len = 0;
1694 : 6085767 : decoded->max_block_id = -1;
3433 heikki.linnakangas@i 1695 : 6085767 : ptr = (char *) record;
1696 : 6085767 : ptr += SizeOfXLogRecord;
1697 : 6085767 : remaining = record->xl_tot_len - SizeOfXLogRecord;
1698 : :
1699 : : /* Decode the headers */
1700 : 6085767 : datatotal = 0;
1701 [ + + ]: 12499521 : while (remaining > datatotal)
1702 : : {
1703 [ - + ]: 12325672 : COPY_HEADER_FIELD(&block_id, sizeof(uint8));
1704 : :
1705 [ + + ]: 12325672 : if (block_id == XLR_BLOCK_ID_DATA_SHORT)
1706 : : {
1707 : : /* XLogRecordDataHeaderShort */
1708 : : uint8 main_data_len;
1709 : :
1710 [ - + ]: 5900413 : COPY_HEADER_FIELD(&main_data_len, sizeof(uint8));
1711 : :
758 tmunro@postgresql.or 1712 : 5900413 : decoded->main_data_len = main_data_len;
3433 heikki.linnakangas@i 1713 : 5900413 : datatotal += main_data_len;
1714 : 5900413 : break; /* by convention, the main data fragment is
1715 : : * always last */
1716 : : }
1717 [ + + ]: 6425259 : else if (block_id == XLR_BLOCK_ID_DATA_LONG)
1718 : : {
1719 : : /* XLogRecordDataHeaderLong */
1720 : : uint32 main_data_len;
1721 : :
1722 [ - + ]: 11505 : COPY_HEADER_FIELD(&main_data_len, sizeof(uint32));
758 tmunro@postgresql.or 1723 : 11505 : decoded->main_data_len = main_data_len;
3433 heikki.linnakangas@i 1724 : 11505 : datatotal += main_data_len;
1725 : 11505 : break; /* by convention, the main data fragment is
1726 : : * always last */
1727 : : }
3273 andres@anarazel.de 1728 [ + + ]: 6413754 : else if (block_id == XLR_BLOCK_ID_ORIGIN)
1729 : : {
758 tmunro@postgresql.or 1730 [ - + ]: 12931 : COPY_HEADER_FIELD(&decoded->record_origin, sizeof(RepOriginId));
1731 : : }
1364 akapila@postgresql.o 1732 [ + + ]: 6400823 : else if (block_id == XLR_BLOCK_ID_TOPLEVEL_XID)
1733 : : {
758 tmunro@postgresql.or 1734 [ - + ]: 677 : COPY_HEADER_FIELD(&decoded->toplevel_xid, sizeof(TransactionId));
1735 : : }
3433 heikki.linnakangas@i 1736 [ + - ]: 6400146 : else if (block_id <= XLR_MAX_BLOCK_ID)
1737 : : {
1738 : : /* XLogRecordBlockHeader */
1739 : : DecodedBkpBlock *blk;
1740 : : uint8 fork_flags;
1741 : :
1742 : : /* mark any intervening block IDs as not in use */
758 tmunro@postgresql.or 1743 [ + + ]: 6403035 : for (int i = decoded->max_block_id + 1; i < block_id; ++i)
1744 : 2889 : decoded->blocks[i].in_use = false;
1745 : :
1746 [ - + ]: 6400146 : if (block_id <= decoded->max_block_id)
1747 : : {
3433 heikki.linnakangas@i 1748 :UBC 0 : report_invalid_record(state,
1749 : : "out-of-order block_id %u at %X/%X",
1750 : : block_id,
1146 peter@eisentraut.org 1751 : 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
3433 heikki.linnakangas@i 1752 : 0 : goto err;
1753 : : }
758 tmunro@postgresql.or 1754 :CBC 6400146 : decoded->max_block_id = block_id;
1755 : :
1756 : 6400146 : blk = &decoded->blocks[block_id];
3433 heikki.linnakangas@i 1757 : 6400146 : blk->in_use = true;
2622 rhaas@postgresql.org 1758 : 6400146 : blk->apply_image = false;
1759 : :
3433 heikki.linnakangas@i 1760 [ - + ]: 6400146 : COPY_HEADER_FIELD(&fork_flags, sizeof(uint8));
1761 : 6400146 : blk->forknum = fork_flags & BKPBLOCK_FORK_MASK;
1762 : 6400146 : blk->flags = fork_flags;
1763 : 6400146 : blk->has_image = ((fork_flags & BKPBLOCK_HAS_IMAGE) != 0);
1764 : 6400146 : blk->has_data = ((fork_flags & BKPBLOCK_HAS_DATA) != 0);
1765 : :
738 tmunro@postgresql.or 1766 : 6400146 : blk->prefetch_buffer = InvalidBuffer;
1767 : :
3433 heikki.linnakangas@i 1768 [ - + ]: 6400146 : COPY_HEADER_FIELD(&blk->data_len, sizeof(uint16));
1769 : : /* cross-check that the HAS_DATA flag is set iff data_length > 0 */
1770 [ + + - + ]: 6400146 : if (blk->has_data && blk->data_len == 0)
1771 : : {
3433 heikki.linnakangas@i 1772 :UBC 0 : report_invalid_record(state,
1773 : : "BKPBLOCK_HAS_DATA set, but no data included at %X/%X",
1146 peter@eisentraut.org 1774 : 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
3324 fujii@postgresql.org 1775 : 0 : goto err;
1776 : : }
3433 heikki.linnakangas@i 1777 [ + + - + ]:CBC 6400146 : if (!blk->has_data && blk->data_len != 0)
1778 : : {
3433 heikki.linnakangas@i 1779 :UBC 0 : report_invalid_record(state,
1780 : : "BKPBLOCK_HAS_DATA not set, but data length is %u at %X/%X",
1781 : 0 : (unsigned int) blk->data_len,
1146 peter@eisentraut.org 1782 : 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
3324 fujii@postgresql.org 1783 : 0 : goto err;
1784 : : }
3433 heikki.linnakangas@i 1785 :CBC 6400146 : datatotal += blk->data_len;
1786 : :
1787 [ + + ]: 6400146 : if (blk->has_image)
1788 : : {
3322 fujii@postgresql.org 1789 [ - + ]: 103977 : COPY_HEADER_FIELD(&blk->bimg_len, sizeof(uint16));
3433 heikki.linnakangas@i 1790 [ - + ]: 103977 : COPY_HEADER_FIELD(&blk->hole_offset, sizeof(uint16));
3322 fujii@postgresql.org 1791 [ - + ]: 103977 : COPY_HEADER_FIELD(&blk->bimg_info, sizeof(uint8));
1792 : :
2622 rhaas@postgresql.org 1793 : 103977 : blk->apply_image = ((blk->bimg_info & BKPIMAGE_APPLY) != 0);
1794 : :
1020 michael@paquier.xyz 1795 [ - + ]: 103977 : if (BKPIMAGE_COMPRESSED(blk->bimg_info))
1796 : : {
3322 fujii@postgresql.org 1797 [ # # ]:UBC 0 : if (blk->bimg_info & BKPIMAGE_HAS_HOLE)
1798 [ # # ]: 0 : COPY_HEADER_FIELD(&blk->hole_length, sizeof(uint16));
1799 : : else
1800 : 0 : blk->hole_length = 0;
1801 : : }
1802 : : else
3322 fujii@postgresql.org 1803 :CBC 103977 : blk->hole_length = BLCKSZ - blk->bimg_len;
1804 : 103977 : datatotal += blk->bimg_len;
1805 : :
1806 : : /*
1807 : : * cross-check that hole_offset > 0, hole_length > 0 and
1808 : : * bimg_len < BLCKSZ if the HAS_HOLE flag is set.
1809 : : */
1810 [ + + ]: 103977 : if ((blk->bimg_info & BKPIMAGE_HAS_HOLE) &&
1811 [ + - ]: 97424 : (blk->hole_offset == 0 ||
1812 [ + - ]: 97424 : blk->hole_length == 0 ||
1813 [ - + ]: 97424 : blk->bimg_len == BLCKSZ))
1814 : : {
3322 fujii@postgresql.org 1815 :UBC 0 : report_invalid_record(state,
1816 : : "BKPIMAGE_HAS_HOLE set, but hole offset %u length %u block image length %u at %X/%X",
1817 : 0 : (unsigned int) blk->hole_offset,
1818 : 0 : (unsigned int) blk->hole_length,
1819 : 0 : (unsigned int) blk->bimg_len,
1146 peter@eisentraut.org 1820 : 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
3322 fujii@postgresql.org 1821 : 0 : goto err;
1822 : : }
1823 : :
1824 : : /*
1825 : : * cross-check that hole_offset == 0 and hole_length == 0 if
1826 : : * the HAS_HOLE flag is not set.
1827 : : */
3322 fujii@postgresql.org 1828 [ + + ]:CBC 103977 : if (!(blk->bimg_info & BKPIMAGE_HAS_HOLE) &&
1829 [ + - - + ]: 6553 : (blk->hole_offset != 0 || blk->hole_length != 0))
1830 : : {
3322 fujii@postgresql.org 1831 :UBC 0 : report_invalid_record(state,
1832 : : "BKPIMAGE_HAS_HOLE not set, but hole offset %u length %u at %X/%X",
1833 : 0 : (unsigned int) blk->hole_offset,
1834 : 0 : (unsigned int) blk->hole_length,
1146 peter@eisentraut.org 1835 : 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
3322 fujii@postgresql.org 1836 : 0 : goto err;
1837 : : }
1838 : :
1839 : : /*
1840 : : * Cross-check that bimg_len < BLCKSZ if it is compressed.
1841 : : */
1020 michael@paquier.xyz 1842 [ - + ]:CBC 103977 : if (BKPIMAGE_COMPRESSED(blk->bimg_info) &&
3322 fujii@postgresql.org 1843 [ # # ]:UBC 0 : blk->bimg_len == BLCKSZ)
1844 : : {
1845 : 0 : report_invalid_record(state,
1846 : : "BKPIMAGE_COMPRESSED set, but block image length %u at %X/%X",
1847 : 0 : (unsigned int) blk->bimg_len,
1146 peter@eisentraut.org 1848 : 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
3322 fujii@postgresql.org 1849 : 0 : goto err;
1850 : : }
1851 : :
1852 : : /*
1853 : : * cross-check that bimg_len = BLCKSZ if neither HAS_HOLE is
1854 : : * set nor COMPRESSED().
1855 : : */
3322 fujii@postgresql.org 1856 [ + + ]:CBC 103977 : if (!(blk->bimg_info & BKPIMAGE_HAS_HOLE) &&
1020 michael@paquier.xyz 1857 [ + - ]: 6553 : !BKPIMAGE_COMPRESSED(blk->bimg_info) &&
3322 fujii@postgresql.org 1858 [ - + ]: 6553 : blk->bimg_len != BLCKSZ)
1859 : : {
3322 fujii@postgresql.org 1860 :UBC 0 : report_invalid_record(state,
1861 : : "neither BKPIMAGE_HAS_HOLE nor BKPIMAGE_COMPRESSED set, but block image length is %u at %X/%X",
1862 : 0 : (unsigned int) blk->data_len,
1146 peter@eisentraut.org 1863 : 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
3322 fujii@postgresql.org 1864 : 0 : goto err;
1865 : : }
1866 : : }
3433 heikki.linnakangas@i 1867 [ + + ]:CBC 6400146 : if (!(fork_flags & BKPBLOCK_SAME_REL))
1868 : : {
648 rhaas@postgresql.org 1869 [ - + ]: 5948045 : COPY_HEADER_FIELD(&blk->rlocator, sizeof(RelFileLocator));
1870 : 5948045 : rlocator = &blk->rlocator;
1871 : : }
1872 : : else
1873 : : {
1874 [ - + ]: 452101 : if (rlocator == NULL)
1875 : : {
3433 heikki.linnakangas@i 1876 :UBC 0 : report_invalid_record(state,
1877 : : "BKPBLOCK_SAME_REL set but no previous rel at %X/%X",
1146 peter@eisentraut.org 1878 : 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
3433 heikki.linnakangas@i 1879 : 0 : goto err;
1880 : : }
1881 : :
648 rhaas@postgresql.org 1882 :CBC 452101 : blk->rlocator = *rlocator;
1883 : : }
3433 heikki.linnakangas@i 1884 [ - + ]: 6400146 : COPY_HEADER_FIELD(&blk->blkno, sizeof(BlockNumber));
1885 : : }
1886 : : else
1887 : : {
3433 heikki.linnakangas@i 1888 :UBC 0 : report_invalid_record(state,
1889 : : "invalid block_id %u at %X/%X",
1146 peter@eisentraut.org 1890 : 0 : block_id, LSN_FORMAT_ARGS(state->ReadRecPtr));
3433 heikki.linnakangas@i 1891 : 0 : goto err;
1892 : : }
1893 : : }
1894 : :
3433 heikki.linnakangas@i 1895 [ - + ]:CBC 6085767 : if (remaining != datatotal)
3433 heikki.linnakangas@i 1896 :UBC 0 : goto shortdata_err;
1897 : :
1898 : : /*
1899 : : * Ok, we've parsed the fragment headers, and verified that the total
1900 : : * length of the payload in the fragments is equal to the amount of data
1901 : : * left. Copy the data of each fragment to contiguous space after the
1902 : : * blocks array, inserting alignment padding before the data fragments so
1903 : : * they can be cast to struct pointers by REDO routines.
1904 : : */
758 tmunro@postgresql.or 1905 :CBC 6085767 : out = ((char *) decoded) +
1906 : 6085767 : offsetof(DecodedXLogRecord, blocks) +
1907 : 6085767 : sizeof(decoded->blocks[0]) * (decoded->max_block_id + 1);
1908 : :
1909 : : /* block data first */
1910 [ + + ]: 12488802 : for (block_id = 0; block_id <= decoded->max_block_id; block_id++)
1911 : : {
1912 : 6403035 : DecodedBkpBlock *blk = &decoded->blocks[block_id];
1913 : :
3433 heikki.linnakangas@i 1914 [ + + ]: 6403035 : if (!blk->in_use)
1915 : 2889 : continue;
1916 : :
2622 rhaas@postgresql.org 1917 [ + + - + ]: 6400146 : Assert(blk->has_image || !blk->apply_image);
1918 : :
3433 heikki.linnakangas@i 1919 [ + + ]: 6400146 : if (blk->has_image)
1920 : : {
1921 : : /* no need to align image */
758 tmunro@postgresql.or 1922 : 103977 : blk->bkp_image = out;
1923 : 103977 : memcpy(out, ptr, blk->bimg_len);
3322 fujii@postgresql.org 1924 : 103977 : ptr += blk->bimg_len;
758 tmunro@postgresql.or 1925 : 103977 : out += blk->bimg_len;
1926 : : }
3433 heikki.linnakangas@i 1927 [ + + ]: 6400146 : if (blk->has_data)
1928 : : {
758 tmunro@postgresql.or 1929 : 4917587 : out = (char *) MAXALIGN(out);
1930 : 4917587 : blk->data = out;
3433 heikki.linnakangas@i 1931 : 4917587 : memcpy(blk->data, ptr, blk->data_len);
1932 : 4917587 : ptr += blk->data_len;
758 tmunro@postgresql.or 1933 : 4917587 : out += blk->data_len;
1934 : : }
1935 : : }
1936 : :
1937 : : /* and finally, the main data */
1938 [ + + ]: 6085767 : if (decoded->main_data_len > 0)
1939 : : {
1940 : 5911918 : out = (char *) MAXALIGN(out);
1941 : 5911918 : decoded->main_data = out;
1942 : 5911918 : memcpy(decoded->main_data, ptr, decoded->main_data_len);
1943 : 5911918 : ptr += decoded->main_data_len;
1944 : 5911918 : out += decoded->main_data_len;
1945 : : }
1946 : :
1947 : : /* Report the actual size we used. */
1948 : 6085767 : decoded->size = MAXALIGN(out - (char *) decoded);
1949 [ - + ]: 6085767 : Assert(DecodeXLogRecordRequiredSpace(record->xl_tot_len) >=
1950 : : decoded->size);
1951 : :
3433 heikki.linnakangas@i 1952 : 6085767 : return true;
1953 : :
3433 heikki.linnakangas@i 1954 :UBC 0 : shortdata_err:
1955 : 0 : report_invalid_record(state,
1956 : : "record with invalid length at %X/%X",
1146 peter@eisentraut.org 1957 : 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
3433 heikki.linnakangas@i 1958 : 0 : err:
1959 : 0 : *errormsg = state->errormsg_buf;
1960 : :
1961 : 0 : return false;
1962 : : }
1963 : :
1964 : : /*
1965 : : * Returns information about the block that a block reference refers to.
1966 : : *
1967 : : * This is like XLogRecGetBlockTagExtended, except that the block reference
1968 : : * must exist and there's no access to prefetch_buffer.
1969 : : */
1970 : : void
3433 heikki.linnakangas@i 1971 :CBC 3436443 : XLogRecGetBlockTag(XLogReaderState *record, uint8 block_id,
1972 : : RelFileLocator *rlocator, ForkNumber *forknum,
1973 : : BlockNumber *blknum)
1974 : : {
648 rhaas@postgresql.org 1975 [ - + ]: 3436443 : if (!XLogRecGetBlockTagExtended(record, block_id, rlocator, forknum,
1976 : : blknum, NULL))
1977 : : {
1978 : : #ifndef FRONTEND
568 peter@eisentraut.org 1979 [ # # ]:UBC 0 : elog(ERROR, "could not locate backup block with ID %d in WAL record",
1980 : : block_id);
1981 : : #else
1982 : 0 : pg_fatal("could not locate backup block with ID %d in WAL record",
1983 : : block_id);
1984 : : #endif
1985 : : }
738 tmunro@postgresql.or 1986 :CBC 3436443 : }
1987 : :
1988 : : /*
1989 : : * Returns information about the block that a block reference refers to,
1990 : : * optionally including the buffer that the block may already be in.
1991 : : *
1992 : : * If the WAL record contains a block reference with the given ID, *rlocator,
1993 : : * *forknum, *blknum and *prefetch_buffer are filled in (if not NULL), and
1994 : : * returns true. Otherwise returns false.
1995 : : */
1996 : : bool
1997 : 7417350 : XLogRecGetBlockTagExtended(XLogReaderState *record, uint8 block_id,
1998 : : RelFileLocator *rlocator, ForkNumber *forknum,
1999 : : BlockNumber *blknum,
2000 : : Buffer *prefetch_buffer)
2001 : : {
2002 : : DecodedBkpBlock *bkpb;
2003 : :
734 tgl@sss.pgh.pa.us 2004 [ + + + + ]: 7417350 : if (!XLogRecHasBlockRef(record, block_id))
3433 heikki.linnakangas@i 2005 : 39180 : return false;
2006 : :
758 tmunro@postgresql.or 2007 : 7378170 : bkpb = &record->record->blocks[block_id];
648 rhaas@postgresql.org 2008 [ + + ]: 7378170 : if (rlocator)
2009 : 7318631 : *rlocator = bkpb->rlocator;
3433 heikki.linnakangas@i 2010 [ + + ]: 7378170 : if (forknum)
2011 : 3887259 : *forknum = bkpb->forknum;
2012 [ + + ]: 7378170 : if (blknum)
2013 : 5870643 : *blknum = bkpb->blkno;
738 tmunro@postgresql.or 2014 [ + + ]: 7378170 : if (prefetch_buffer)
2015 : 3137812 : *prefetch_buffer = bkpb->prefetch_buffer;
3433 heikki.linnakangas@i 2016 : 7378170 : return true;
2017 : : }
2018 : :
2019 : : /*
2020 : : * Returns the data associated with a block reference, or NULL if there is
2021 : : * no data (e.g. because a full-page image was taken instead). The returned
2022 : : * pointer points to a MAXALIGNed buffer.
2023 : : */
2024 : : char *
2025 : 3714312 : XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len)
2026 : : {
2027 : : DecodedBkpBlock *bkpb;
2028 : :
758 tmunro@postgresql.or 2029 [ + - ]: 3714312 : if (block_id > record->record->max_block_id ||
2030 [ - + ]: 3714312 : !record->record->blocks[block_id].in_use)
3433 heikki.linnakangas@i 2031 :UBC 0 : return NULL;
2032 : :
758 tmunro@postgresql.or 2033 :CBC 3714312 : bkpb = &record->record->blocks[block_id];
2034 : :
3433 heikki.linnakangas@i 2035 [ + + ]: 3714312 : if (!bkpb->has_data)
2036 : : {
2037 [ + - ]: 228 : if (len)
2038 : 228 : *len = 0;
2039 : 228 : return NULL;
2040 : : }
2041 : : else
2042 : : {
2043 [ + + ]: 3714084 : if (len)
2044 : 3710065 : *len = bkpb->data_len;
2045 : 3714084 : return bkpb->data;
2046 : : }
2047 : : }
2048 : :
2049 : : /*
2050 : : * Restore a full-page image from a backup block attached to an XLOG record.
2051 : : *
2052 : : * Returns true if a full-page image is restored, and false on failure with
2053 : : * an error to be consumed by the caller.
2054 : : */
2055 : : bool
2056 : 61431 : RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page)
2057 : : {
2058 : : DecodedBkpBlock *bkpb;
2059 : : char *ptr;
2060 : : PGAlignedBlock tmp;
2061 : :
758 tmunro@postgresql.or 2062 [ + - ]: 61431 : if (block_id > record->record->max_block_id ||
2063 [ - + ]: 61431 : !record->record->blocks[block_id].in_use)
2064 : : {
583 michael@paquier.xyz 2065 :UBC 0 : report_invalid_record(record,
2066 : : "could not restore image at %X/%X with invalid block %d specified",
2067 : 0 : LSN_FORMAT_ARGS(record->ReadRecPtr),
2068 : : block_id);
3433 heikki.linnakangas@i 2069 : 0 : return false;
2070 : : }
758 tmunro@postgresql.or 2071 [ - + ]:CBC 61431 : if (!record->record->blocks[block_id].has_image)
2072 : : {
583 michael@paquier.xyz 2073 :UBC 0 : report_invalid_record(record, "could not restore image at %X/%X with invalid state, block %d",
2074 : 0 : LSN_FORMAT_ARGS(record->ReadRecPtr),
2075 : : block_id);
3433 heikki.linnakangas@i 2076 : 0 : return false;
2077 : : }
2078 : :
758 tmunro@postgresql.or 2079 :CBC 61431 : bkpb = &record->record->blocks[block_id];
3322 fujii@postgresql.org 2080 : 61431 : ptr = bkpb->bkp_image;
2081 : :
1020 michael@paquier.xyz 2082 [ - + ]: 61431 : if (BKPIMAGE_COMPRESSED(bkpb->bimg_info))
2083 : : {
2084 : : /* If a backup block image is compressed, decompress it */
1020 michael@paquier.xyz 2085 :UBC 0 : bool decomp_success = true;
2086 : :
2087 [ # # ]: 0 : if ((bkpb->bimg_info & BKPIMAGE_COMPRESS_PGLZ) != 0)
2088 : : {
2089 [ # # ]: 0 : if (pglz_decompress(ptr, bkpb->bimg_len, tmp.data,
2090 : 0 : BLCKSZ - bkpb->hole_length, true) < 0)
2091 : 0 : decomp_success = false;
2092 : : }
2093 [ # # ]: 0 : else if ((bkpb->bimg_info & BKPIMAGE_COMPRESS_LZ4) != 0)
2094 : : {
2095 : : #ifdef USE_LZ4
2096 [ # # ]: 0 : if (LZ4_decompress_safe(ptr, tmp.data,
2097 : 0 : bkpb->bimg_len, BLCKSZ - bkpb->hole_length) <= 0)
2098 : 0 : decomp_success = false;
2099 : : #else
2100 : : report_invalid_record(record, "could not restore image at %X/%X compressed with %s not supported by build, block %d",
2101 : : LSN_FORMAT_ARGS(record->ReadRecPtr),
2102 : : "LZ4",
2103 : : block_id);
2104 : : return false;
2105 : : #endif
2106 : : }
765 2107 [ # # ]: 0 : else if ((bkpb->bimg_info & BKPIMAGE_COMPRESS_ZSTD) != 0)
2108 : : {
2109 : : #ifdef USE_ZSTD
2110 : 0 : size_t decomp_result = ZSTD_decompress(tmp.data,
2111 : 0 : BLCKSZ - bkpb->hole_length,
2112 : 0 : ptr, bkpb->bimg_len);
2113 : :
2114 [ # # ]: 0 : if (ZSTD_isError(decomp_result))
2115 : 0 : decomp_success = false;
2116 : : #else
2117 : : report_invalid_record(record, "could not restore image at %X/%X compressed with %s not supported by build, block %d",
2118 : : LSN_FORMAT_ARGS(record->ReadRecPtr),
2119 : : "zstd",
2120 : : block_id);
2121 : : return false;
2122 : : #endif
2123 : : }
2124 : : else
2125 : : {
583 2126 : 0 : report_invalid_record(record, "could not restore image at %X/%X compressed with unknown method, block %d",
1010 2127 : 0 : LSN_FORMAT_ARGS(record->ReadRecPtr),
2128 : : block_id);
1020 2129 : 0 : return false;
2130 : : }
2131 : :
2132 [ # # ]: 0 : if (!decomp_success)
2133 : : {
583 2134 : 0 : report_invalid_record(record, "could not decompress image at %X/%X, block %d",
1146 peter@eisentraut.org 2135 : 0 : LSN_FORMAT_ARGS(record->ReadRecPtr),
2136 : : block_id);
3322 fujii@postgresql.org 2137 : 0 : return false;
2138 : : }
2139 : :
2052 tgl@sss.pgh.pa.us 2140 : 0 : ptr = tmp.data;
2141 : : }
2142 : :
2143 : : /* generate page, taking into account hole if necessary */
3433 heikki.linnakangas@i 2144 [ + + ]:CBC 61431 : if (bkpb->hole_length == 0)
2145 : : {
3322 fujii@postgresql.org 2146 : 3770 : memcpy(page, ptr, BLCKSZ);
2147 : : }
2148 : : else
2149 : : {
2150 : 57661 : memcpy(page, ptr, bkpb->hole_offset);
2151 : : /* must zero-fill the hole */
3433 heikki.linnakangas@i 2152 [ + + + - : 430741 : MemSet(page + bkpb->hole_offset, 0, bkpb->hole_length);
+ - + + +
+ ]
2153 : 57661 : memcpy(page + (bkpb->hole_offset + bkpb->hole_length),
3322 fujii@postgresql.org 2154 : 57661 : ptr + bkpb->hole_offset,
3433 heikki.linnakangas@i 2155 : 57661 : BLCKSZ - (bkpb->hole_offset + bkpb->hole_length));
2156 : : }
2157 : :
2158 : 61431 : return true;
2159 : : }
2160 : :
2161 : : #ifndef FRONTEND
2162 : :
2163 : : /*
2164 : : * Extract the FullTransactionId from a WAL record.
2165 : : */
2166 : : FullTransactionId
1735 tmunro@postgresql.or 2167 :UBC 0 : XLogRecGetFullXid(XLogReaderState *record)
2168 : : {
2169 : : TransactionId xid,
2170 : : next_xid;
2171 : : uint32 epoch;
2172 : :
2173 : : /*
2174 : : * This function is only safe during replay, because it depends on the
2175 : : * replay state. See AdvanceNextFullTransactionIdPastXid() for more.
2176 : : */
2177 [ # # # # ]: 0 : Assert(AmStartupProcess() || !IsUnderPostmaster);
2178 : :
2179 : 0 : xid = XLogRecGetXid(record);
128 heikki.linnakangas@i 2180 :UNC 0 : next_xid = XidFromFullTransactionId(TransamVariables->nextXid);
2181 : 0 : epoch = EpochFromFullTransactionId(TransamVariables->nextXid);
2182 : :
2183 : : /*
2184 : : * If xid is numerically greater than next_xid, it has to be from the last
2185 : : * epoch.
2186 : : */
1735 tmunro@postgresql.or 2187 [ # # ]:UBC 0 : if (unlikely(xid > next_xid))
2188 : 0 : --epoch;
2189 : :
2190 : 0 : return FullTransactionIdFromEpochAndXid(epoch, xid);
2191 : : }
2192 : :
2193 : : #endif
|