Age Owner TLA Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * xlogreader.c
4 : * Generic XLog reading facility
5 : *
6 : * Portions Copyright (c) 2013-2023, 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 "miscadmin.h"
38 : #include "pgstat.h"
39 : #include "utils/memutils.h"
40 : #else
41 : #include "common/logging.h"
42 : #endif
43 :
44 : static void report_invalid_record(XLogReaderState *state, const char *fmt,...)
45 : pg_attribute_printf(2, 3);
46 : static bool allocate_recordbuf(XLogReaderState *state, uint32 reclength);
47 : static int ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr,
48 : int reqLen);
49 : static void XLogReaderInvalReadState(XLogReaderState *state);
50 : static XLogPageReadResult XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking);
51 : static bool ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr,
52 : XLogRecPtr PrevRecPtr, XLogRecord *record, bool randAccess);
53 : static bool ValidXLogRecord(XLogReaderState *state, XLogRecord *record,
54 : XLogRecPtr recptr);
55 : static void ResetDecoder(XLogReaderState *state);
56 : static void WALOpenSegmentInit(WALOpenSegment *seg, WALSegmentContext *segcxt,
57 : int segsize, const char *waldir);
58 :
59 : /* size of the buffer allocated for error message. */
60 : #define MAX_ERRORMSG_LEN 1000
61 :
62 : /*
63 : * Default size; large enough that typical users of XLogReader won't often need
64 : * to use the 'oversized' memory allocation code path.
65 : */
66 : #define DEFAULT_DECODE_BUFFER_SIZE (64 * 1024)
67 :
68 : /*
69 : * Construct a string in state->errormsg_buf explaining what's wrong with
70 : * the current record being read.
71 : */
72 : static void
3735 alvherre 73 CBC 188 : report_invalid_record(XLogReaderState *state, const char *fmt,...)
74 : {
75 : va_list args;
76 :
77 188 : fmt = _(fmt);
78 :
79 188 : va_start(args, fmt);
80 188 : vsnprintf(state->errormsg_buf, MAX_ERRORMSG_LEN, fmt, args);
81 188 : va_end(args);
82 :
387 tmunro 83 188 : state->errormsg_deferred = true;
84 188 : }
85 :
86 : /*
87 : * Set the size of the decoding buffer. A pointer to a caller supplied memory
88 : * region may also be passed in, in which case non-oversized records will be
89 : * decoded there.
90 : */
91 : void
92 1176 : XLogReaderSetDecodeBuffer(XLogReaderState *state, void *buffer, size_t size)
93 : {
94 1176 : Assert(state->decode_buffer == NULL);
95 :
96 1176 : state->decode_buffer = buffer;
97 1176 : state->decode_buffer_size = size;
98 1176 : state->decode_buffer_tail = buffer;
99 1176 : state->decode_buffer_head = buffer;
3735 alvherre 100 1176 : }
101 :
102 : /*
103 : * Allocate and initialize a new XLogReader.
104 : *
105 : * Returns NULL if the xlogreader couldn't be allocated.
106 : */
107 : XLogReaderState *
1293 108 2624 : XLogReaderAllocate(int wal_segment_size, const char *waldir,
109 : XLogReaderRoutine *routine, void *private_data)
110 : {
111 : XLogReaderState *state;
112 :
113 : state = (XLogReaderState *)
2928 fujii 114 2624 : palloc_extended(sizeof(XLogReaderState),
115 : MCXT_ALLOC_NO_OOM | MCXT_ALLOC_ZERO);
116 2624 : if (!state)
2928 fujii 117 UBC 0 : return NULL;
118 :
119 : /* initialize caller-provided support functions */
699 tmunro 120 CBC 2624 : state->routine = *routine;
121 :
122 : /*
123 : * Permanently allocate readBuf. We do it this way, rather than just
124 : * making a static array, for two reasons: (1) no need to waste the
125 : * storage in most instantiations of the backend; (2) a static char array
126 : * isn't guaranteed to have any particular alignment, whereas
127 : * palloc_extended() will provide MAXALIGN'd storage.
128 : */
2928 fujii 129 2624 : state->readBuf = (char *) palloc_extended(XLOG_BLCKSZ,
130 : MCXT_ALLOC_NO_OOM);
131 2624 : if (!state->readBuf)
132 : {
2928 fujii 133 UBC 0 : pfree(state);
134 0 : return NULL;
135 : }
136 :
137 : /* Initialize segment info. */
1293 alvherre 138 CBC 2624 : WALOpenSegmentInit(&state->seg, &state->segcxt, wal_segment_size,
139 : waldir);
140 :
141 : /* system_identifier initialized to zeroes above */
699 tmunro 142 2624 : state->private_data = private_data;
143 : /* ReadRecPtr, EndRecPtr and readLen initialized to zeroes above */
2928 fujii 144 2624 : state->errormsg_buf = palloc_extended(MAX_ERRORMSG_LEN + 1,
145 : MCXT_ALLOC_NO_OOM);
146 2624 : if (!state->errormsg_buf)
147 : {
2928 fujii 148 UBC 0 : pfree(state->readBuf);
149 0 : pfree(state);
150 0 : return NULL;
151 : }
3735 alvherre 152 CBC 2624 : state->errormsg_buf[0] = '\0';
153 :
154 : /*
155 : * Allocate an initial readRecordBuf of minimal size, which can later be
156 : * enlarged if necessary.
157 : */
158 2624 : if (!allocate_recordbuf(state, 0))
159 : {
3062 heikki.linnakangas 160 UBC 0 : pfree(state->errormsg_buf);
161 0 : pfree(state->readBuf);
162 0 : pfree(state);
3735 alvherre 163 0 : return NULL;
164 : }
165 :
3735 alvherre 166 CBC 2624 : return state;
167 : }
168 :
169 : void
170 2257 : XLogReaderFree(XLogReaderState *state)
171 : {
699 tmunro 172 2257 : if (state->seg.ws_file != -1)
173 1055 : state->routine.segment_close(state);
174 :
387 175 2257 : if (state->decode_buffer && state->free_decode_buffer)
176 2197 : pfree(state->decode_buffer);
177 :
3062 heikki.linnakangas 178 2257 : pfree(state->errormsg_buf);
3735 alvherre 179 2257 : if (state->readRecordBuf)
3062 heikki.linnakangas 180 2257 : pfree(state->readRecordBuf);
181 2257 : pfree(state->readBuf);
182 2257 : pfree(state);
3735 alvherre 183 2257 : }
184 :
185 : /*
186 : * Allocate readRecordBuf to fit a record of at least the given length.
187 : * Returns true if successful, false if out of memory.
188 : *
189 : * readRecordBufSize is set to the new buffer size.
190 : *
191 : * To avoid useless small increases, round its size to a multiple of
192 : * XLOG_BLCKSZ, and make sure it's at least 5*Max(BLCKSZ, XLOG_BLCKSZ) to start
193 : * with. (That is enough for all "normal" records, but very large commit or
194 : * abort records might need more space.)
195 : */
196 : static bool
197 2642 : allocate_recordbuf(XLogReaderState *state, uint32 reclength)
198 : {
199 2642 : uint32 newSize = reclength;
200 :
201 2642 : newSize += XLOG_BLCKSZ - (newSize % XLOG_BLCKSZ);
202 2642 : newSize = Max(newSize, 5 * Max(BLCKSZ, XLOG_BLCKSZ));
203 :
204 : #ifndef FRONTEND
205 :
206 : /*
207 : * Note that in much unlucky circumstances, the random data read from a
208 : * recycled segment can cause this routine to be called with a size
209 : * causing a hard failure at allocation. For a standby, this would cause
210 : * the instance to stop suddenly with a hard failure, preventing it to
211 : * retry fetching WAL from one of its sources which could allow it to move
212 : * on with replay without a manual restart. If the data comes from a past
213 : * recycled segment and is still valid, then the allocation may succeed
214 : * but record checks are going to fail so this would be short-lived. If
215 : * the allocation fails because of a memory shortage, then this is not a
216 : * hard failure either per the guarantee given by MCXT_ALLOC_NO_OOM.
217 : */
1756 michael 218 2642 : if (!AllocSizeIsValid(newSize))
1756 michael 219 UBC 0 : return false;
220 :
221 : #endif
222 :
3735 alvherre 223 CBC 2642 : if (state->readRecordBuf)
3062 heikki.linnakangas 224 18 : pfree(state->readRecordBuf);
2928 fujii 225 2642 : state->readRecordBuf =
226 2642 : (char *) palloc_extended(newSize, MCXT_ALLOC_NO_OOM);
227 2642 : if (state->readRecordBuf == NULL)
228 : {
2928 fujii 229 UBC 0 : state->readRecordBufSize = 0;
230 0 : return false;
231 : }
3735 alvherre 232 CBC 2642 : state->readRecordBufSize = newSize;
233 2642 : return true;
234 : }
235 :
236 : /*
237 : * Initialize the passed segment structs.
238 : */
239 : static void
1293 240 2624 : WALOpenSegmentInit(WALOpenSegment *seg, WALSegmentContext *segcxt,
241 : int segsize, const char *waldir)
242 : {
243 2624 : seg->ws_file = -1;
244 2624 : seg->ws_segno = 0;
245 2624 : seg->ws_tli = 0;
246 :
247 2624 : segcxt->ws_segsize = segsize;
248 2624 : if (waldir)
1293 alvherre 249 UBC 0 : snprintf(segcxt->ws_dir, MAXPGPATH, "%s", waldir);
1293 alvherre 250 CBC 2624 : }
251 :
252 : /*
253 : * Begin reading WAL at 'RecPtr'.
254 : *
255 : * 'RecPtr' should point to the beginning of a valid WAL record. Pointing at
256 : * the beginning of a page is also OK, if there is a new record right after
257 : * the page header, i.e. not a continuation.
258 : *
259 : * This does not make any attempt to read the WAL yet, and hence cannot fail.
260 : * If the starting address is not correct, the first call to XLogReadRecord()
261 : * will error out.
262 : */
263 : void
1169 heikki.linnakangas 264 3709 : XLogBeginRead(XLogReaderState *state, XLogRecPtr RecPtr)
265 : {
266 3709 : Assert(!XLogRecPtrIsInvalid(RecPtr));
267 :
268 3709 : ResetDecoder(state);
269 :
270 : /* Begin at the passed-in record pointer. */
271 3709 : state->EndRecPtr = RecPtr;
387 tmunro 272 3709 : state->NextRecPtr = RecPtr;
1169 heikki.linnakangas 273 3709 : state->ReadRecPtr = InvalidXLogRecPtr;
387 tmunro 274 3709 : state->DecodeRecPtr = InvalidXLogRecPtr;
275 3709 : }
276 :
277 : /*
278 : * Release the last record that was returned by XLogNextRecord(), if any, to
279 : * free up space. Returns the LSN past the end of the record.
280 : */
281 : XLogRecPtr
282 10026716 : XLogReleasePreviousRecord(XLogReaderState *state)
283 : {
284 : DecodedXLogRecord *record;
285 : XLogRecPtr next_lsn;
286 :
287 10026716 : if (!state->record)
213 288 5017120 : return InvalidXLogRecPtr;
289 :
290 : /*
291 : * Remove it from the decoded record queue. It must be the oldest item
292 : * decoded, decode_queue_head.
293 : */
387 294 5009596 : record = state->record;
213 295 5009596 : next_lsn = record->next_lsn;
387 296 5009596 : Assert(record == state->decode_queue_head);
297 5009596 : state->record = NULL;
298 5009596 : state->decode_queue_head = record->next;
299 :
300 : /* It might also be the newest item decoded, decode_queue_tail. */
301 5009596 : if (state->decode_queue_tail == record)
302 2532526 : state->decode_queue_tail = NULL;
303 :
304 : /* Release the space. */
305 5009596 : if (unlikely(record->oversized))
306 : {
307 : /* It's not in the decode buffer, so free it to release space. */
308 10 : pfree(record);
309 : }
310 : else
311 : {
312 : /* It must be the head (oldest) record in the decode buffer. */
313 5009586 : Assert(state->decode_buffer_head == (char *) record);
314 :
315 : /*
316 : * We need to update head to point to the next record that is in the
317 : * decode buffer, if any, being careful to skip oversized ones
318 : * (they're not in the decode buffer).
319 : */
320 5009586 : record = record->next;
321 5009586 : while (unlikely(record && record->oversized))
387 tmunro 322 UBC 0 : record = record->next;
323 :
387 tmunro 324 CBC 5009586 : if (record)
325 : {
326 : /* Adjust head to release space up to the next record. */
327 2477070 : state->decode_buffer_head = (char *) record;
328 : }
329 : else
330 : {
331 : /*
332 : * Otherwise we might as well just reset head and tail to the
333 : * start of the buffer space, because we're empty. This means
334 : * we'll keep overwriting the same piece of memory if we're not
335 : * doing any prefetching.
336 : */
337 2532516 : state->decode_buffer_head = state->decode_buffer;
338 2532516 : state->decode_buffer_tail = state->decode_buffer;
339 : }
340 : }
341 :
213 342 5009596 : return next_lsn;
343 : }
344 :
345 : /*
346 : * Attempt to read an XLOG record.
347 : *
348 : * XLogBeginRead() or XLogFindNextRecord() and then XLogReadAhead() must be
349 : * called before the first call to XLogNextRecord(). This functions returns
350 : * records and errors that were put into an internal queue by XLogReadAhead().
351 : *
352 : * On success, a record is returned.
353 : *
354 : * The returned record (or *errormsg) points to an internal buffer that's
355 : * valid until the next call to XLogNextRecord.
356 : */
357 : DecodedXLogRecord *
387 358 5013282 : XLogNextRecord(XLogReaderState *state, char **errormsg)
359 : {
360 : /* Release the last record returned by XLogNextRecord(). */
361 5013282 : XLogReleasePreviousRecord(state);
362 :
363 5013282 : if (state->decode_queue_head == NULL)
364 : {
365 406 : *errormsg = NULL;
366 406 : if (state->errormsg_deferred)
367 : {
368 182 : if (state->errormsg_buf[0] != '\0')
369 182 : *errormsg = state->errormsg_buf;
370 182 : state->errormsg_deferred = false;
371 : }
372 :
373 : /*
374 : * state->EndRecPtr is expected to have been set by the last call to
375 : * XLogBeginRead() or XLogNextRecord(), and is the location of the
376 : * error.
377 : */
378 406 : Assert(!XLogRecPtrIsInvalid(state->EndRecPtr));
379 :
380 406 : return NULL;
381 : }
382 :
383 : /*
384 : * Record this as the most recent record returned, so that we'll release
385 : * it next time. This also exposes it to the traditional
386 : * XLogRecXXX(xlogreader) macros, which work with the decoder rather than
387 : * the record for historical reasons.
388 : */
389 5012876 : state->record = state->decode_queue_head;
390 :
391 : /*
392 : * Update the pointers to the beginning and one-past-the-end of this
393 : * record, again for the benefit of historical code that expected the
394 : * decoder to track this rather than accessing these fields of the record
395 : * itself.
396 : */
397 5012876 : state->ReadRecPtr = state->record->lsn;
398 5012876 : state->EndRecPtr = state->record->next_lsn;
399 :
400 5012876 : *errormsg = NULL;
401 :
402 5012876 : return state->record;
403 : }
404 :
405 : /*
406 : * Attempt to read an XLOG record.
407 : *
408 : * XLogBeginRead() or XLogFindNextRecord() must be called before the first call
409 : * to XLogReadRecord().
410 : *
411 : * If the page_read callback fails to read the requested data, NULL is
412 : * returned. The callback is expected to have reported the error; errormsg
413 : * is set to NULL.
414 : *
415 : * If the reading fails for some other reason, NULL is also returned, and
416 : * *errormsg is set to a string with details of the failure.
417 : *
418 : * The returned pointer (or *errormsg) points to an internal buffer that's
419 : * valid until the next call to XLogReadRecord.
420 : */
421 : XLogRecord *
699 422 2506751 : XLogReadRecord(XLogReaderState *state, char **errormsg)
423 : {
424 : DecodedXLogRecord *decoded;
425 :
426 : /*
427 : * Release last returned record, if there is one. We need to do this so
428 : * that we can check for empty decode queue accurately.
429 : */
387 430 2506751 : XLogReleasePreviousRecord(state);
431 :
432 : /*
433 : * Call XLogReadAhead() in blocking mode to make sure there is something
434 : * in the queue, though we don't use the result.
435 : */
436 2506751 : if (!XLogReaderHasQueuedRecordOrError(state))
437 2506751 : XLogReadAhead(state, false /* nonblocking */ );
438 :
439 : /* Consume the head record or error. */
440 2506630 : decoded = XLogNextRecord(state, errormsg);
441 2506630 : if (decoded)
442 : {
443 : /*
444 : * This function returns a pointer to the record's header, not the
445 : * actual decoded record. The caller will access the decoded record
446 : * through the XLogRecGetXXX() macros, which reach the decoded
447 : * recorded as xlogreader->record.
448 : */
449 2506419 : Assert(state->record == decoded);
450 2506419 : return &decoded->header;
451 : }
452 :
453 211 : return NULL;
454 : }
455 :
456 : /*
457 : * Allocate space for a decoded record. The only member of the returned
458 : * object that is initialized is the 'oversized' flag, indicating that the
459 : * decoded record wouldn't fit in the decode buffer and must eventually be
460 : * freed explicitly.
461 : *
462 : * The caller is responsible for adjusting decode_buffer_tail with the real
463 : * size after successfully decoding a record into this space. This way, if
464 : * decoding fails, then there is nothing to undo unless the 'oversized' flag
465 : * was set and pfree() must be called.
466 : *
467 : * Return NULL if there is no space in the decode buffer and allow_oversized
468 : * is false, or if memory allocation fails for an oversized buffer.
469 : */
470 : static DecodedXLogRecord *
471 5013073 : XLogReadRecordAlloc(XLogReaderState *state, size_t xl_tot_len, bool allow_oversized)
472 : {
473 5013073 : size_t required_space = DecodeXLogRecordRequiredSpace(xl_tot_len);
474 5013073 : DecodedXLogRecord *decoded = NULL;
475 :
476 : /* Allocate a circular decode buffer if we don't have one already. */
477 5013073 : if (unlikely(state->decode_buffer == NULL))
478 : {
479 2368 : if (state->decode_buffer_size == 0)
480 1192 : state->decode_buffer_size = DEFAULT_DECODE_BUFFER_SIZE;
481 2368 : state->decode_buffer = palloc(state->decode_buffer_size);
482 2368 : state->decode_buffer_head = state->decode_buffer;
483 2368 : state->decode_buffer_tail = state->decode_buffer;
484 2368 : state->free_decode_buffer = true;
485 : }
486 :
487 : /* Try to allocate space in the circular decode buffer. */
488 5013073 : if (state->decode_buffer_tail >= state->decode_buffer_head)
489 : {
490 : /* Empty, or tail is to the right of head. */
491 4991318 : if (state->decode_buffer_tail + required_space <=
492 4991318 : state->decode_buffer + state->decode_buffer_size)
493 : {
494 : /* There is space between tail and end. */
495 4990708 : decoded = (DecodedXLogRecord *) state->decode_buffer_tail;
496 4990708 : decoded->oversized = false;
497 4990708 : return decoded;
498 : }
499 610 : else if (state->decode_buffer + required_space <
500 610 : state->decode_buffer_head)
501 : {
502 : /* There is space between start and head. */
503 593 : decoded = (DecodedXLogRecord *) state->decode_buffer;
504 593 : decoded->oversized = false;
505 593 : return decoded;
506 : }
507 : }
508 : else
509 : {
510 : /* Tail is to the left of head. */
511 21755 : if (state->decode_buffer_tail + required_space <
512 21755 : state->decode_buffer_head)
513 : {
514 : /* There is space between tail and head. */
515 21755 : decoded = (DecodedXLogRecord *) state->decode_buffer_tail;
516 21755 : decoded->oversized = false;
517 21755 : return decoded;
518 : }
519 : }
520 :
521 : /* Not enough space in the decode buffer. Are we allowed to allocate? */
522 17 : if (allow_oversized)
523 : {
524 13 : decoded = palloc_extended(required_space, MCXT_ALLOC_NO_OOM);
525 13 : if (decoded == NULL)
387 tmunro 526 UBC 0 : return NULL;
387 tmunro 527 CBC 13 : decoded->oversized = true;
528 13 : return decoded;
529 : }
530 :
531 4 : return NULL;
532 : }
533 :
534 : static XLogPageReadResult
535 5026864 : XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking)
536 : {
537 : XLogRecPtr RecPtr;
538 : XLogRecord *record;
539 : XLogRecPtr targetPagePtr;
540 : bool randAccess;
541 : uint32 len,
542 : total_len;
543 : uint32 targetRecOff;
544 : uint32 pageHeaderSize;
545 : bool assembled;
546 : bool gotheader;
547 : int readOff;
548 : DecodedXLogRecord *decoded;
549 : char *errormsg; /* not used */
550 :
551 : /*
552 : * randAccess indicates whether to verify the previous-record pointer of
553 : * the record we're reading. We only do this if we're reading
554 : * sequentially, which is what we initially assume.
555 : */
699 556 5026864 : randAccess = false;
557 :
558 : /* reset error state */
559 5026864 : state->errormsg_buf[0] = '\0';
387 560 5026864 : decoded = NULL;
561 :
557 alvherre 562 5026864 : state->abortedRecPtr = InvalidXLogRecPtr;
563 5026864 : state->missingContrecPtr = InvalidXLogRecPtr;
564 :
387 tmunro 565 5026864 : RecPtr = state->NextRecPtr;
566 :
567 5026864 : if (state->DecodeRecPtr != InvalidXLogRecPtr)
568 : {
569 : /* read the record after the one we just read */
570 :
571 : /*
572 : * NextRecPtr is pointing to end+1 of the previous WAL record. If
573 : * we're at a page boundary, no more records can fit on the current
574 : * page. We must skip over the page header, but we can't do that until
575 : * we've read in the page, since the header size is variable.
576 : */
577 : }
578 : else
579 : {
580 : /*
581 : * Caller supplied a position to start at.
582 : *
583 : * In this case, NextRecPtr should already be pointing either to a
584 : * valid record starting position or alternatively to the beginning of
585 : * a page. See the header comments for XLogBeginRead.
586 : */
234 rhaas 587 GNC 3655 : Assert(RecPtr % XLOG_BLCKSZ == 0 || XRecOffIsValid(RecPtr));
699 tmunro 588 CBC 3655 : randAccess = true;
731 tmunro 589 ECB : }
590 :
557 alvherre 591 GIC 5026864 : restart:
387 tmunro 592 CBC 5026865 : state->nonblocking = nonblocking;
699 593 5026865 : state->currRecPtr = RecPtr;
557 alvherre 594 5026865 : assembled = false;
731 tmunro 595 ECB :
699 tmunro 596 GIC 5026865 : targetPagePtr = RecPtr - (RecPtr % XLOG_BLCKSZ);
699 tmunro 597 CBC 5026865 : targetRecOff = RecPtr % XLOG_BLCKSZ;
731 tmunro 598 ECB :
599 : /*
600 : * Read the page containing the record into state->readBuf. Request enough
601 : * byte to cover the whole record header, or at least the part of it that
602 : * fits on the same page.
603 : */
699 tmunro 604 GIC 5026865 : readOff = ReadPageInternal(state, targetPagePtr,
699 tmunro 605 CBC 5026865 : Min(targetRecOff + SizeOfXLogRecord, XLOG_BLCKSZ));
387 606 5026714 : if (readOff == XLREAD_WOULDBLOCK)
607 13224 : return XLREAD_WOULDBLOCK;
608 5013490 : else if (readOff < 0)
699 609 238 : goto err;
731 tmunro 610 ECB :
611 : /*
612 : * ReadPageInternal always returns at least the page header, so we can
613 : * examine it now.
614 : */
699 tmunro 615 GIC 5013252 : pageHeaderSize = XLogPageHeaderSize((XLogPageHeader) state->readBuf);
699 tmunro 616 CBC 5013252 : if (targetRecOff == 0)
699 tmunro 617 ECB : {
618 : /*
619 : * At page start, so skip over page header.
620 : */
699 tmunro 621 GIC 5117 : RecPtr += pageHeaderSize;
699 tmunro 622 CBC 5117 : targetRecOff = pageHeaderSize;
731 tmunro 623 ECB : }
699 tmunro 624 GIC 5008135 : else if (targetRecOff < pageHeaderSize)
731 tmunro 625 ECB : {
38 peter 626 UNC 0 : report_invalid_record(state, "invalid record offset at %X/%X: expected at least %u, got %u",
627 0 : LSN_FORMAT_ARGS(RecPtr),
628 : pageHeaderSize, targetRecOff);
699 tmunro 629 UBC 0 : goto err;
630 : }
731 tmunro 631 EUB :
699 tmunro 632 GIC 5013252 : if ((((XLogPageHeader) state->readBuf)->xlp_info & XLP_FIRST_IS_CONTRECORD) &&
633 : targetRecOff == pageHeaderSize)
731 tmunro 634 ECB : {
699 tmunro 635 UIC 0 : report_invalid_record(state, "contrecord is requested by %X/%X",
636 0 : LSN_FORMAT_ARGS(RecPtr));
699 tmunro 637 UBC 0 : goto err;
731 tmunro 638 EUB : }
639 :
640 : /* ReadPageInternal has verified the page header */
699 tmunro 641 GIC 5013252 : Assert(pageHeaderSize <= readOff);
642 :
699 tmunro 643 ECB : /*
644 : * Read the record length.
645 : *
646 : * NB: Even though we use an XLogRecord pointer here, the whole record
647 : * header might not fit on this page. xl_tot_len is the first field of the
648 : * struct, so it must be on this page (the records are MAXALIGNed), but we
649 : * cannot access any other fields until we've verified that we got the
650 : * whole header.
651 : */
699 tmunro 652 GIC 5013252 : record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ);
653 5013252 : total_len = record->xl_tot_len;
731 tmunro 654 ECB :
699 655 : /*
656 : * If the whole record header is on this page, validate it immediately.
657 : * Otherwise do just a basic sanity check on xl_tot_len, and validate the
658 : * rest of the header after reading it from the next page. The xl_tot_len
659 : * check is necessary here to ensure that we enter the "Need to reassemble
660 : * record" code path below; otherwise we might fail to apply
661 : * ValidXLogRecordHeader at all.
662 : */
699 tmunro 663 GIC 5013252 : if (targetRecOff <= XLOG_BLCKSZ - SizeOfXLogRecord)
664 : {
387 tmunro 665 CBC 5003337 : if (!ValidXLogRecordHeader(state, RecPtr, state->DecodeRecPtr, record,
666 : randAccess))
699 667 179 : goto err;
699 tmunro 668 GIC 5003158 : gotheader = true;
731 tmunro 669 ECB : }
670 : else
671 : {
672 : /* XXX: more validation should be done here */
699 tmunro 673 GIC 9915 : if (total_len < SizeOfXLogRecord)
674 : {
699 tmunro 675 LBC 0 : report_invalid_record(state,
676 : "invalid record length at %X/%X: expected at least %u, got %u",
699 tmunro 677 UBC 0 : LSN_FORMAT_ARGS(RecPtr),
678 : (uint32) SizeOfXLogRecord, total_len);
679 0 : goto err;
680 : }
699 tmunro 681 GBC 9915 : gotheader = false;
682 : }
731 tmunro 683 ECB :
684 : /*
685 : * Find space to decode this record. Don't allow oversized allocation if
686 : * the caller requested nonblocking. Otherwise, we *have* to try to
687 : * decode the record now because the caller has nothing else to do, so
688 : * allow an oversized record to be palloc'd if that turns out to be
689 : * necessary.
690 : */
387 tmunro 691 GIC 5013073 : decoded = XLogReadRecordAlloc(state,
692 : total_len,
387 tmunro 693 CBC 5013073 : !nonblocking /* allow_oversized */ );
387 tmunro 694 GIC 5013073 : if (decoded == NULL)
387 tmunro 695 ECB : {
696 : /*
697 : * There is no space in the decode buffer. The caller should help
698 : * with that problem by consuming some records.
699 : */
387 tmunro 700 GIC 4 : if (nonblocking)
701 4 : return XLREAD_WOULDBLOCK;
387 tmunro 702 ECB :
703 : /* We failed to allocate memory for an oversized record. */
387 tmunro 704 UIC 0 : report_invalid_record(state,
705 : "out of memory while trying to decode a record of length %u", total_len);
387 tmunro 706 UBC 0 : goto err;
707 : }
387 tmunro 708 EUB :
699 tmunro 709 GIC 5013069 : len = XLOG_BLCKSZ - RecPtr % XLOG_BLCKSZ;
710 5013069 : if (total_len > len)
731 tmunro 711 ECB : {
699 712 : /* Need to reassemble record */
713 : char *contdata;
714 : XLogPageHeader pageHeader;
715 : char *buffer;
716 : uint32 gotlen;
717 :
557 alvherre 718 GIC 75444 : assembled = true;
719 :
699 tmunro 720 ECB : /*
721 : * Enlarge readRecordBuf as needed.
722 : */
699 tmunro 723 GIC 75444 : if (total_len > state->readRecordBufSize &&
724 18 : !allocate_recordbuf(state, total_len))
699 tmunro 725 ECB : {
726 : /* We treat this as a "bogus data" condition */
699 tmunro 727 UIC 0 : report_invalid_record(state, "record length %u at %X/%X too long",
728 0 : total_len, LSN_FORMAT_ARGS(RecPtr));
699 tmunro 729 UBC 0 : goto err;
699 tmunro 730 EUB : }
3735 alvherre 731 :
732 : /* Copy the first fragment of the record from the first page. */
699 tmunro 733 GIC 75444 : memcpy(state->readRecordBuf,
734 75444 : state->readBuf + RecPtr % XLOG_BLCKSZ, len);
699 tmunro 735 CBC 75444 : buffer = state->readRecordBuf + len;
736 75444 : gotlen = len;
3062 heikki.linnakangas 737 ECB :
699 tmunro 738 : do
739 : {
740 : /* Calculate pointer to beginning of next page */
699 tmunro 741 GIC 78233 : targetPagePtr += XLOG_BLCKSZ;
742 :
699 tmunro 743 ECB : /* Wait for the next page to become available */
699 tmunro 744 GIC 78233 : readOff = ReadPageInternal(state, targetPagePtr,
745 78233 : Min(total_len - gotlen + SizeOfXLogShortPHD,
699 tmunro 746 ECB : XLOG_BLCKSZ));
3735 alvherre 747 :
387 tmunro 748 GIC 78232 : if (readOff == XLREAD_WOULDBLOCK)
749 182 : return XLREAD_WOULDBLOCK;
387 tmunro 750 CBC 78050 : else if (readOff < 0)
699 751 3 : goto err;
3735 alvherre 752 ECB :
699 tmunro 753 CBC 78047 : Assert(SizeOfXLogShortPHD <= readOff);
754 :
755 78047 : pageHeader = (XLogPageHeader) state->readBuf;
756 :
557 alvherre 757 ECB : /*
758 : * If we were expecting a continuation record and got an
759 : * "overwrite contrecord" flag, that means the continuation record
760 : * was overwritten with a different record. Restart the read by
761 : * assuming the address to read is the location where we found
762 : * this flag; but keep track of the LSN of the record we were
763 : * reading, for later verification.
764 : */
557 alvherre 765 GIC 78047 : if (pageHeader->xlp_info & XLP_FIRST_IS_OVERWRITE_CONTRECORD)
766 : {
499 alvherre 767 CBC 1 : state->overwrittenRecPtr = RecPtr;
557 alvherre 768 GIC 1 : RecPtr = targetPagePtr;
557 alvherre 769 CBC 1 : goto restart;
557 alvherre 770 ECB : }
771 :
772 : /* Check that the continuation on next page looks valid */
699 tmunro 773 GIC 78046 : if (!(pageHeader->xlp_info & XLP_FIRST_IS_CONTRECORD))
774 : {
699 tmunro 775 LBC 0 : report_invalid_record(state,
776 : "there is no contrecord flag at %X/%X",
699 tmunro 777 UBC 0 : LSN_FORMAT_ARGS(RecPtr));
699 tmunro 778 UIC 0 : goto err;
731 tmunro 779 EUB : }
1602 michael 780 :
781 : /*
782 : * Cross-check that xlp_rem_len agrees with how much of the record
783 : * we expect there to be left.
784 : */
699 tmunro 785 GIC 78046 : if (pageHeader->xlp_rem_len == 0 ||
786 78046 : total_len != (pageHeader->xlp_rem_len + gotlen))
731 tmunro 787 ECB : {
699 tmunro 788 LBC 0 : report_invalid_record(state,
789 : "invalid contrecord length %u (expected %lld) at %X/%X",
699 tmunro 790 EUB : pageHeader->xlp_rem_len,
699 tmunro 791 UIC 0 : ((long long) total_len) - gotlen,
792 0 : LSN_FORMAT_ARGS(RecPtr));
699 tmunro 793 UBC 0 : goto err;
699 tmunro 794 EUB : }
3735 alvherre 795 :
796 : /* Append the continuation from this page to the buffer */
699 tmunro 797 GIC 78046 : pageHeaderSize = XLogPageHeaderSize(pageHeader);
798 :
699 tmunro 799 CBC 78046 : if (readOff < pageHeaderSize)
699 tmunro 800 UIC 0 : readOff = ReadPageInternal(state, targetPagePtr,
699 tmunro 801 ECB : pageHeaderSize);
3735 alvherre 802 EUB :
699 tmunro 803 GIC 78046 : Assert(pageHeaderSize <= readOff);
804 :
699 tmunro 805 CBC 78046 : contdata = (char *) state->readBuf + pageHeaderSize;
699 tmunro 806 GIC 78046 : len = XLOG_BLCKSZ - pageHeaderSize;
699 tmunro 807 CBC 78046 : if (pageHeader->xlp_rem_len < len)
808 75256 : len = pageHeader->xlp_rem_len;
3735 alvherre 809 ECB :
699 tmunro 810 CBC 78046 : if (readOff < pageHeaderSize + len)
699 tmunro 811 UIC 0 : readOff = ReadPageInternal(state, targetPagePtr,
699 tmunro 812 LBC 0 : pageHeaderSize + len);
731 tmunro 813 EUB :
699 tmunro 814 GBC 78046 : memcpy(buffer, (char *) contdata, len);
699 tmunro 815 GIC 78046 : buffer += len;
699 tmunro 816 CBC 78046 : gotlen += len;
731 tmunro 817 ECB :
699 818 : /* If we just reassembled the record header, validate it. */
699 tmunro 819 GIC 78046 : if (!gotheader)
820 : {
699 tmunro 821 CBC 9881 : record = (XLogRecord *) state->readRecordBuf;
387 tmunro 822 GIC 9881 : if (!ValidXLogRecordHeader(state, RecPtr, state->DecodeRecPtr,
699 tmunro 823 ECB : record, randAccess))
3735 alvherre 824 LBC 0 : goto err;
699 tmunro 825 GIC 9881 : gotheader = true;
731 tmunro 826 EUB : }
699 tmunro 827 CBC 78046 : } while (gotlen < total_len);
828 :
829 75257 : Assert(gotheader);
830 :
831 75257 : record = (XLogRecord *) state->readRecordBuf;
699 tmunro 832 GIC 75257 : if (!ValidXLogRecord(state, record, RecPtr))
699 tmunro 833 LBC 0 : goto err;
731 tmunro 834 ECB :
699 tmunro 835 GBC 75257 : pageHeaderSize = XLogPageHeaderSize((XLogPageHeader) state->readBuf);
387 tmunro 836 GIC 75257 : state->DecodeRecPtr = RecPtr;
387 tmunro 837 CBC 75257 : state->NextRecPtr = targetPagePtr + pageHeaderSize
699 838 75257 : + MAXALIGN(pageHeader->xlp_rem_len);
699 tmunro 839 ECB : }
840 : else
841 : {
842 : /* Wait for the record data to become available */
699 tmunro 843 GIC 4937625 : readOff = ReadPageInternal(state, targetPagePtr,
844 4937625 : Min(targetRecOff + total_len, XLOG_BLCKSZ));
387 tmunro 845 CBC 4937625 : if (readOff == XLREAD_WOULDBLOCK)
387 tmunro 846 LBC 0 : return XLREAD_WOULDBLOCK;
387 tmunro 847 CBC 4937625 : else if (readOff < 0)
699 tmunro 848 UBC 0 : goto err;
3735 alvherre 849 ECB :
699 tmunro 850 EUB : /* Record does not cross a page boundary */
699 tmunro 851 GIC 4937625 : if (!ValidXLogRecord(state, record, RecPtr))
699 tmunro 852 UIC 0 : goto err;
3735 alvherre 853 ECB :
387 tmunro 854 GBC 4937625 : state->NextRecPtr = RecPtr + MAXALIGN(total_len);
855 :
387 tmunro 856 CBC 4937625 : state->DecodeRecPtr = RecPtr;
857 : }
3735 alvherre 858 ECB :
859 : /*
860 : * Special processing if it's an XLOG SWITCH record
861 : */
731 tmunro 862 GIC 5012882 : if (record->xl_rmid == RM_XLOG_ID &&
863 32993 : (record->xl_info & ~XLR_INFO_MASK) == XLOG_SWITCH)
3735 alvherre 864 ECB : {
865 : /* Pretend it extends to end of segment */
387 tmunro 866 GIC 112 : state->NextRecPtr += state->segcxt.ws_segsize - 1;
867 112 : state->NextRecPtr -= XLogSegmentOffset(state->NextRecPtr, state->segcxt.ws_segsize);
3735 alvherre 868 ECB : }
869 :
387 tmunro 870 GIC 5012882 : if (DecodeXLogRecord(state, decoded, record, RecPtr, &errormsg))
871 : {
387 tmunro 872 ECB : /* Record the location of the next record. */
387 tmunro 873 GIC 5012882 : decoded->next_lsn = state->NextRecPtr;
874 :
387 tmunro 875 ECB : /*
876 : * If it's in the decode buffer, mark the decode buffer space as
877 : * occupied.
878 : */
387 tmunro 879 GIC 5012882 : if (!decoded->oversized)
880 : {
387 tmunro 881 ECB : /* The new decode buffer head must be MAXALIGNed. */
387 tmunro 882 GIC 5012872 : Assert(decoded->size == MAXALIGN(decoded->size));
883 5012872 : if ((char *) decoded == state->decode_buffer)
387 tmunro 884 CBC 2536389 : state->decode_buffer_tail = state->decode_buffer + decoded->size;
387 tmunro 885 ECB : else
387 tmunro 886 CBC 2476483 : state->decode_buffer_tail += decoded->size;
887 : }
387 tmunro 888 ECB :
889 : /* Insert it into the queue of decoded records. */
387 tmunro 890 GIC 5012882 : Assert(state->decode_queue_tail != decoded);
891 5012882 : if (state->decode_queue_tail)
387 tmunro 892 CBC 2477076 : state->decode_queue_tail->next = decoded;
893 5012882 : state->decode_queue_tail = decoded;
894 5012882 : if (!state->decode_queue_head)
895 2535806 : state->decode_queue_head = decoded;
896 5012882 : return XLREAD_SUCCESS;
387 tmunro 897 ECB : }
699 898 : else
387 tmunro 899 UIC 0 : return XLREAD_FAIL;
900 :
3735 alvherre 901 GBC 420 : err:
557 alvherre 902 GIC 420 : if (assembled)
557 alvherre 903 ECB : {
904 : /*
905 : * We get here when a record that spans multiple pages needs to be
906 : * assembled, but something went wrong -- perhaps a contrecord piece
907 : * was lost. If caller is WAL replay, it will know where the aborted
908 : * record was and where to direct followup WAL to be written, marking
909 : * the next piece with XLP_FIRST_IS_OVERWRITE_CONTRECORD, which will
910 : * in turn signal downstream WAL consumers that the broken WAL record
911 : * is to be ignored.
912 : */
557 alvherre 913 GIC 3 : state->abortedRecPtr = RecPtr;
914 3 : state->missingContrecPtr = targetPagePtr;
213 tmunro 915 ECB :
916 : /*
917 : * If we got here without reporting an error, report one now so that
918 : * XLogPrefetcherReadRecord() doesn't bring us back a second time and
919 : * clobber the above state. Otherwise, the existing error takes
920 : * precedence.
921 : */
213 tmunro 922 GIC 3 : if (!state->errormsg_buf[0])
923 3 : report_invalid_record(state,
213 tmunro 924 ECB : "missing contrecord at %X/%X",
213 tmunro 925 CBC 3 : LSN_FORMAT_ARGS(RecPtr));
926 : }
3735 alvherre 927 ECB :
387 tmunro 928 GIC 420 : if (decoded && decoded->oversized)
929 2 : pfree(decoded);
387 tmunro 930 ECB :
3735 alvherre 931 : /*
932 : * Invalidate the read state. We might read from a different source after
933 : * failure.
934 : */
2566 alvherre 935 GIC 420 : XLogReaderInvalReadState(state);
936 :
387 tmunro 937 ECB : /*
938 : * If an error was written to errmsg_buf, it'll be returned to the caller
939 : * of XLogReadRecord() after all successfully decoded records from the
940 : * read queue.
941 : */
942 :
387 tmunro 943 GIC 420 : return XLREAD_FAIL;
944 : }
387 tmunro 945 ECB :
946 : /*
947 : * Try to decode the next available record, and return it. The record will
948 : * also be returned to XLogNextRecord(), which must be called to 'consume'
949 : * each record.
950 : *
951 : * If nonblocking is true, may return NULL due to lack of data or WAL decoding
952 : * space.
953 : */
954 : DecodedXLogRecord *
387 tmunro 955 GIC 5026864 : XLogReadAhead(XLogReaderState *state, bool nonblocking)
956 : {
387 tmunro 957 ECB : XLogPageReadResult result;
958 :
387 tmunro 959 GIC 5026864 : if (state->errormsg_deferred)
387 tmunro 960 UIC 0 : return NULL;
387 tmunro 961 ECB :
387 tmunro 962 GBC 5026864 : result = XLogDecodeNextRecord(state, nonblocking);
387 tmunro 963 GIC 5026712 : if (result == XLREAD_SUCCESS)
387 tmunro 964 ECB : {
387 tmunro 965 CBC 5012882 : Assert(state->decode_queue_tail != NULL);
387 tmunro 966 GIC 5012882 : return state->decode_queue_tail;
387 tmunro 967 ECB : }
3735 alvherre 968 :
699 tmunro 969 GIC 13830 : return NULL;
970 : }
3735 alvherre 971 ECB :
972 : /*
973 : * Read a single xlog page including at least [pageptr, reqLen] of valid data
974 : * via the page_read() callback.
975 : *
976 : * Returns XLREAD_FAIL if the required page cannot be read for some
977 : * reason; errormsg_buf is set in that case (unless the error occurs in the
978 : * page_read callback).
979 : *
980 : * Returns XLREAD_WOULDBLOCK if the requested data can't be read without
981 : * waiting. This can be returned only if the installed page_read callback
982 : * respects the state->nonblocking flag, and cannot read the requested data
983 : * immediately.
984 : *
985 : * We fetch the page from a reader-local cache if we know we have the required
986 : * data and if there hasn't been any error since caching the data.
987 : */
988 : static int
699 tmunro 989 GIC 10042753 : ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr, int reqLen)
990 : {
699 tmunro 991 ECB : int readLen;
992 : uint32 targetPageOff;
993 : XLogSegNo targetSegNo;
994 : XLogPageHeader hdr;
995 :
699 tmunro 996 GIC 10042753 : Assert((pageptr % XLOG_BLCKSZ) == 0);
997 :
699 tmunro 998 CBC 10042753 : XLByteToSeg(pageptr, targetSegNo, state->segcxt.ws_segsize);
699 tmunro 999 GIC 10042753 : targetPageOff = XLogSegmentOffset(pageptr, state->segcxt.ws_segsize);
3735 alvherre 1000 ECB :
699 tmunro 1001 : /* check whether we have all the requested data already */
699 tmunro 1002 GIC 10042753 : if (targetSegNo == state->seg.ws_segno &&
1003 10039861 : targetPageOff == state->segoff && reqLen <= state->readLen)
699 tmunro 1004 CBC 9928596 : return state->readLen;
3735 alvherre 1005 ECB :
218 tmunro 1006 : /*
1007 : * Invalidate contents of internal buffer before read attempt. Just set
1008 : * the length to 0, rather than a full XLogReaderInvalReadState(), so we
1009 : * don't forget the segment we last successfully read.
1010 : */
218 tmunro 1011 GIC 114157 : state->readLen = 0;
1012 :
699 tmunro 1013 ECB : /*
1014 : * Data is not in our buffer.
1015 : *
1016 : * Every time we actually read the segment, even if we looked at parts of
1017 : * it before, we need to do verification as the page_read callback might
1018 : * now be rereading data from a different source.
1019 : *
1020 : * Whenever switching to a new WAL segment, we read the first page of the
1021 : * file and validate its header, even if that's not where the target
1022 : * record is. This is so that we can check the additional identification
1023 : * info that is present in the first page's "long" header.
1024 : */
699 tmunro 1025 GIC 114157 : if (targetSegNo != state->seg.ws_segno && targetPageOff != 0)
1026 : {
699 tmunro 1027 CBC 2224 : XLogRecPtr targetSegmentPtr = pageptr - targetPageOff;
1028 :
1029 2224 : readLen = state->routine.page_read(state, targetSegmentPtr, XLOG_BLCKSZ,
1030 : state->currRecPtr,
699 tmunro 1031 ECB : state->readBuf);
387 tmunro 1032 GIC 2224 : if (readLen == XLREAD_WOULDBLOCK)
387 tmunro 1033 UIC 0 : return XLREAD_WOULDBLOCK;
387 tmunro 1034 CBC 2224 : else if (readLen < 0)
699 tmunro 1035 UBC 0 : goto err;
731 tmunro 1036 ECB :
699 tmunro 1037 EUB : /* we can be sure to have enough WAL available, we scrolled back */
699 tmunro 1038 GIC 2224 : Assert(readLen == XLOG_BLCKSZ);
1039 :
699 tmunro 1040 CBC 2224 : if (!XLogReaderValidatePageHeader(state, targetSegmentPtr,
1041 : state->readBuf))
699 tmunro 1042 LBC 0 : goto err;
1043 : }
3735 alvherre 1044 EUB :
1045 : /*
1046 : * First, read the requested data length, but at least a short page header
1047 : * so that we can validate it.
1048 : */
699 tmunro 1049 GIC 114157 : readLen = state->routine.page_read(state, pageptr, Max(reqLen, SizeOfXLogShortPHD),
1050 : state->currRecPtr,
699 tmunro 1051 ECB : state->readBuf);
387 tmunro 1052 GIC 114005 : if (readLen == XLREAD_WOULDBLOCK)
1053 13406 : return XLREAD_WOULDBLOCK;
387 tmunro 1054 CBC 100599 : else if (readLen < 0)
699 1055 241 : goto err;
3735 alvherre 1056 ECB :
699 tmunro 1057 CBC 100358 : Assert(readLen <= XLOG_BLCKSZ);
1058 :
699 tmunro 1059 ECB : /* Do we have enough data to check the header length? */
699 tmunro 1060 GIC 100358 : if (readLen <= SizeOfXLogShortPHD)
699 tmunro 1061 UIC 0 : goto err;
3735 alvherre 1062 ECB :
699 tmunro 1063 GBC 100358 : Assert(readLen >= reqLen);
1064 :
699 tmunro 1065 CBC 100358 : hdr = (XLogPageHeader) state->readBuf;
1066 :
699 tmunro 1067 ECB : /* still not enough */
699 tmunro 1068 GIC 100358 : if (readLen < XLogPageHeaderSize(hdr))
1069 : {
699 tmunro 1070 LBC 0 : readLen = state->routine.page_read(state, pageptr, XLogPageHeaderSize(hdr),
1071 : state->currRecPtr,
699 tmunro 1072 EUB : state->readBuf);
387 tmunro 1073 UIC 0 : if (readLen == XLREAD_WOULDBLOCK)
1074 0 : return XLREAD_WOULDBLOCK;
387 tmunro 1075 UBC 0 : else if (readLen < 0)
699 1076 0 : goto err;
3735 alvherre 1077 EUB : }
1078 :
1079 : /*
1080 : * Now that we know we have the full header, validate it.
1081 : */
699 tmunro 1082 GIC 100358 : if (!XLogReaderValidatePageHeader(state, pageptr, (char *) hdr))
699 tmunro 1083 UIC 0 : goto err;
699 tmunro 1084 ECB :
699 tmunro 1085 EUB : /* update read state information */
699 tmunro 1086 GIC 100358 : state->seg.ws_segno = targetSegNo;
1087 100358 : state->segoff = targetPageOff;
699 tmunro 1088 CBC 100358 : state->readLen = readLen;
699 tmunro 1089 ECB :
699 tmunro 1090 CBC 100358 : return readLen;
1091 :
1092 241 : err:
218 tmunro 1093 GIC 241 : XLogReaderInvalReadState(state);
218 tmunro 1094 ECB :
387 tmunro 1095 CBC 241 : return XLREAD_FAIL;
1096 : }
2566 alvherre 1097 ECB :
1098 : /*
1099 : * Invalidate the xlogreader's read state to force a re-read.
1100 : */
1101 : static void
2566 alvherre 1102 GIC 661 : XLogReaderInvalReadState(XLogReaderState *state)
1103 : {
699 tmunro 1104 CBC 661 : state->seg.ws_segno = 0;
699 tmunro 1105 GIC 661 : state->segoff = 0;
699 tmunro 1106 CBC 661 : state->readLen = 0;
3735 alvherre 1107 661 : }
3735 alvherre 1108 ECB :
1109 : /*
1110 : * Validate an XLOG record header.
1111 : *
1112 : * This is just a convenience subroutine to avoid duplicated code in
1113 : * XLogReadRecord. It's not intended for use from anywhere else.
1114 : */
1115 : static bool
3735 alvherre 1116 GIC 5013218 : ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr,
1117 : XLogRecPtr PrevRecPtr, XLogRecord *record,
699 tmunro 1118 ECB : bool randAccess)
1119 : {
3062 heikki.linnakangas 1120 GIC 5013218 : if (record->xl_tot_len < SizeOfXLogRecord)
1121 : {
3735 alvherre 1122 CBC 178 : report_invalid_record(state,
1123 : "invalid record length at %X/%X: expected at least %u, got %u",
775 peter 1124 178 : LSN_FORMAT_ARGS(RecPtr),
1125 : (uint32) SizeOfXLogRecord, record->xl_tot_len);
3735 alvherre 1126 178 : return false;
1127 : }
367 jdavis 1128 5013040 : if (!RmgrIdIsValid(record->xl_rmid))
1129 : {
3735 alvherre 1130 LBC 0 : report_invalid_record(state,
1131 : "invalid resource manager ID %u at %X/%X",
775 peter 1132 UBC 0 : record->xl_rmid, LSN_FORMAT_ARGS(RecPtr));
3735 alvherre 1133 UIC 0 : return false;
3735 alvherre 1134 EUB : }
699 tmunro 1135 GBC 5013040 : if (randAccess)
1136 : {
3735 alvherre 1137 ECB : /*
1138 : * We can't exactly verify the prev-link, but surely it should be less
1139 : * than the record's own address.
1140 : */
3735 alvherre 1141 GIC 3655 : if (!(record->xl_prev < RecPtr))
1142 : {
3735 alvherre 1143 LBC 0 : report_invalid_record(state,
1144 : "record with incorrect prev-link %X/%X at %X/%X",
775 peter 1145 UBC 0 : LSN_FORMAT_ARGS(record->xl_prev),
775 peter 1146 UIC 0 : LSN_FORMAT_ARGS(RecPtr));
3735 alvherre 1147 UBC 0 : return false;
3735 alvherre 1148 EUB : }
1149 : }
1150 : else
1151 : {
1152 : /*
1153 : * Record's prev-link should exactly match our previous location. This
1154 : * check guards against torn WAL pages where a stale but valid-looking
1155 : * WAL record starts on a sector boundary.
1156 : */
3735 alvherre 1157 GIC 5009385 : if (record->xl_prev != PrevRecPtr)
1158 : {
3735 alvherre 1159 CBC 1 : report_invalid_record(state,
1160 : "record with incorrect prev-link %X/%X at %X/%X",
775 peter 1161 1 : LSN_FORMAT_ARGS(record->xl_prev),
775 peter 1162 GIC 1 : LSN_FORMAT_ARGS(RecPtr));
3735 alvherre 1163 CBC 1 : return false;
3735 alvherre 1164 ECB : }
1165 : }
1166 :
3735 alvherre 1167 GIC 5013039 : return true;
1168 : }
3735 alvherre 1169 ECB :
1170 :
1171 : /*
1172 : * CRC-check an XLOG record. We do not believe the contents of an XLOG
1173 : * record (other than to the minimal extent of computing the amount of
1174 : * data to read in) until we've checked the CRCs.
1175 : *
1176 : * We assume all of the record (that is, xl_tot_len bytes) has been read
1177 : * into memory at *record. Also, ValidXLogRecordHeader() has accepted the
1178 : * record's header, which means in particular that xl_tot_len is at least
1179 : * SizeOfXLogRecord.
1180 : */
1181 : static bool
3735 alvherre 1182 GIC 5012882 : ValidXLogRecord(XLogReaderState *state, XLogRecord *record, XLogRecPtr recptr)
1183 : {
2917 heikki.linnakangas 1184 ECB : pg_crc32c crc;
1185 :
1186 : /* Calculate the CRC */
3078 heikki.linnakangas 1187 GIC 5012882 : INIT_CRC32C(crc);
3062 1188 5012882 : COMP_CRC32C(crc, ((char *) record) + SizeOfXLogRecord, record->xl_tot_len - SizeOfXLogRecord);
3062 heikki.linnakangas 1189 ECB : /* include the record header last */
3078 heikki.linnakangas 1190 CBC 5012882 : COMP_CRC32C(crc, (char *) record, offsetof(XLogRecord, xl_crc));
3078 heikki.linnakangas 1191 GIC 5012882 : FIN_CRC32C(crc);
3735 alvherre 1192 ECB :
3078 heikki.linnakangas 1193 CBC 5012882 : if (!EQ_CRC32C(record->xl_crc, crc))
1194 : {
3735 alvherre 1195 LBC 0 : report_invalid_record(state,
1196 : "incorrect resource manager data checksum in record at %X/%X",
775 peter 1197 UBC 0 : LSN_FORMAT_ARGS(recptr));
3735 alvherre 1198 UIC 0 : return false;
3735 alvherre 1199 EUB : }
1200 :
3735 alvherre 1201 GIC 5012882 : return true;
1202 : }
3735 alvherre 1203 ECB :
1204 : /*
1205 : * Validate a page header.
1206 : *
1207 : * Check if 'phdr' is valid as the header of the XLog page at position
1208 : * 'recptr'.
1209 : */
1210 : bool
1800 heikki.linnakangas 1211 GIC 161029 : XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr,
1212 : char *phdr)
3735 alvherre 1213 ECB : {
1214 : XLogSegNo segno;
1215 : int32 offset;
1800 heikki.linnakangas 1216 GIC 161029 : XLogPageHeader hdr = (XLogPageHeader) phdr;
3735 alvherre 1217 ECB :
3735 alvherre 1218 GIC 161029 : Assert((recptr % XLOG_BLCKSZ) == 0);
3735 alvherre 1219 ECB :
1293 alvherre 1220 GIC 161029 : XLByteToSeg(recptr, segno, state->segcxt.ws_segsize);
1293 alvherre 1221 CBC 161029 : offset = XLogSegmentOffset(recptr, state->segcxt.ws_segsize);
3735 alvherre 1222 ECB :
3735 alvherre 1223 GIC 161029 : if (hdr->xlp_magic != XLOG_PAGE_MAGIC)
1224 : {
1225 : char fname[MAXFNAMELEN];
3735 alvherre 1226 ECB :
1293 alvherre 1227 GIC 3 : XLogFileName(fname, state->seg.ws_tli, segno, state->segcxt.ws_segsize);
3735 alvherre 1228 ECB :
3735 alvherre 1229 GIC 3 : report_invalid_record(state,
1230 : "invalid magic number %04X in WAL segment %s, LSN %X/%X, offset %u",
1231 3 : hdr->xlp_magic,
3735 alvherre 1232 ECB : fname,
125 michael 1233 GNC 3 : LSN_FORMAT_ARGS(recptr),
1234 : offset);
3735 alvherre 1235 CBC 3 : return false;
1236 : }
1237 :
1238 161026 : if ((hdr->xlp_info & ~XLP_ALL_FLAGS) != 0)
1239 : {
1240 : char fname[MAXFNAMELEN];
1241 :
1293 alvherre 1242 UBC 0 : XLogFileName(fname, state->seg.ws_tli, segno, state->segcxt.ws_segsize);
1243 :
3735 1244 0 : report_invalid_record(state,
1245 : "invalid info bits %04X in WAL segment %s, LSN %X/%X, offset %u",
1246 0 : hdr->xlp_info,
1247 : fname,
125 michael 1248 UNC 0 : LSN_FORMAT_ARGS(recptr),
3735 alvherre 1249 EUB : offset);
3735 alvherre 1250 UIC 0 : return false;
3735 alvherre 1251 EUB : }
1252 :
3735 alvherre 1253 GIC 161026 : if (hdr->xlp_info & XLP_LONG_HEADER)
3735 alvherre 1254 ECB : {
3735 alvherre 1255 GIC 3056 : XLogLongPageHeader longhdr = (XLogLongPageHeader) hdr;
3735 alvherre 1256 ECB :
3735 alvherre 1257 GIC 3056 : if (state->system_identifier &&
3735 alvherre 1258 CBC 1748 : longhdr->xlp_sysid != state->system_identifier)
3735 alvherre 1259 ECB : {
3735 alvherre 1260 UIC 0 : report_invalid_record(state,
1403 peter 1261 EUB : "WAL file is from different database system: WAL file database system identifier is %llu, pg_control database system identifier is %llu",
1403 peter 1262 UIC 0 : (unsigned long long) longhdr->xlp_sysid,
1403 peter 1263 UBC 0 : (unsigned long long) state->system_identifier);
3735 alvherre 1264 0 : return false;
3735 alvherre 1265 EUB : }
1293 alvherre 1266 GIC 3056 : else if (longhdr->xlp_seg_size != state->segcxt.ws_segsize)
3735 alvherre 1267 ECB : {
3735 alvherre 1268 UIC 0 : report_invalid_record(state,
2028 andres 1269 EUB : "WAL file is from different database system: incorrect segment size in page header");
3735 alvherre 1270 UIC 0 : return false;
3735 alvherre 1271 EUB : }
3735 alvherre 1272 GIC 3056 : else if (longhdr->xlp_xlog_blcksz != XLOG_BLCKSZ)
3735 alvherre 1273 ECB : {
3735 alvherre 1274 UIC 0 : report_invalid_record(state,
2747 peter_e 1275 EUB : "WAL file is from different database system: incorrect XLOG_BLCKSZ in page header");
3735 alvherre 1276 UIC 0 : return false;
3735 alvherre 1277 EUB : }
1278 : }
3735 alvherre 1279 GIC 157970 : else if (offset == 0)
3735 alvherre 1280 ECB : {
1281 : char fname[MAXFNAMELEN];
1282 :
1293 alvherre 1283 UIC 0 : XLogFileName(fname, state->seg.ws_tli, segno, state->segcxt.ws_segsize);
3735 alvherre 1284 EUB :
1285 : /* hmm, first page of file doesn't have a long header? */
3735 alvherre 1286 UIC 0 : report_invalid_record(state,
1287 : "invalid info bits %04X in WAL segment %s, LSN %X/%X, offset %u",
1288 0 : hdr->xlp_info,
3735 alvherre 1289 EUB : fname,
125 michael 1290 UNC 0 : LSN_FORMAT_ARGS(recptr),
1291 : offset);
3735 alvherre 1292 UBC 0 : return false;
1293 : }
3735 alvherre 1294 EUB :
1295 : /*
1296 : * Check that the address on the page agrees with what we expected. This
1297 : * check typically fails when an old WAL segment is recycled, and hasn't
1298 : * yet been overwritten with new data yet.
1299 : */
179 michael 1300 GNC 161026 : if (hdr->xlp_pageaddr != recptr)
1301 : {
3735 alvherre 1302 ECB : char fname[MAXFNAMELEN];
1303 :
1293 alvherre 1304 GIC 3 : XLogFileName(fname, state->seg.ws_tli, segno, state->segcxt.ws_segsize);
1305 :
3735 alvherre 1306 CBC 3 : report_invalid_record(state,
1307 : "unexpected pageaddr %X/%X in WAL segment %s, LSN %X/%X, offset %u",
775 peter 1308 3 : LSN_FORMAT_ARGS(hdr->xlp_pageaddr),
1309 : fname,
125 michael 1310 GNC 3 : LSN_FORMAT_ARGS(recptr),
3735 alvherre 1311 ECB : offset);
3735 alvherre 1312 GIC 3 : return false;
3735 alvherre 1313 ECB : }
1314 :
1315 : /*
1316 : * Since child timelines are always assigned a TLI greater than their
1317 : * immediate parent's TLI, we should never see TLI go backwards across
1318 : * successive pages of a consistent WAL sequence.
1319 : *
1320 : * Sometimes we re-read a segment that's already been (partially) read. So
1321 : * we only verify TLIs for pages that are later than the last remembered
1322 : * LSN.
1323 : */
3735 alvherre 1324 GIC 161023 : if (recptr > state->latestPagePtr)
1325 : {
1326 87646 : if (hdr->xlp_tli < state->latestPageTLI)
3735 alvherre 1327 ECB : {
1328 : char fname[MAXFNAMELEN];
1329 :
1293 alvherre 1330 UIC 0 : XLogFileName(fname, state->seg.ws_tli, segno, state->segcxt.ws_segsize);
1331 :
3735 1332 0 : report_invalid_record(state,
1333 : "out-of-sequence timeline ID %u (after %u) in WAL segment %s, LSN %X/%X, offset %u",
1334 : hdr->xlp_tli,
3735 alvherre 1335 EUB : state->latestPageTLI,
1336 : fname,
125 michael 1337 UNC 0 : LSN_FORMAT_ARGS(recptr),
1338 : offset);
3735 alvherre 1339 UIC 0 : return false;
1340 : }
3735 alvherre 1341 EUB : }
3735 alvherre 1342 GIC 161023 : state->latestPagePtr = recptr;
3735 alvherre 1343 GBC 161023 : state->latestPageTLI = hdr->xlp_tli;
1344 :
3735 alvherre 1345 GIC 161023 : return true;
3735 alvherre 1346 ECB : }
1347 :
1348 : /*
218 tmunro 1349 : * Forget about an error produced by XLogReaderValidatePageHeader().
1350 : */
1351 : void
218 tmunro 1352 GIC 6 : XLogReaderResetError(XLogReaderState *state)
1353 : {
1354 6 : state->errormsg_buf[0] = '\0';
1355 6 : state->errormsg_deferred = false;
218 tmunro 1356 CBC 6 : }
1357 :
3735 alvherre 1358 ECB : /*
3454 heikki.linnakangas 1359 : * Find the first record with an lsn >= RecPtr.
3735 alvherre 1360 : *
1361 : * This is different from XLogBeginRead() in that RecPtr doesn't need to point
1362 : * to a valid record boundary. Useful for checking whether RecPtr is a valid
1363 : * xlog address for reading, and to find the first valid address after some
1364 : * address when dumping records for debugging purposes.
1365 : *
1366 : * This positions the reader, like XLogBeginRead(), so that the next call to
1367 : * XLogReadRecord() will read the next valid record.
1368 : */
1369 : XLogRecPtr
699 tmunro 1370 GIC 15 : XLogFindNextRecord(XLogReaderState *state, XLogRecPtr RecPtr)
1371 : {
1372 : XLogRecPtr tmpRecPtr;
1373 15 : XLogRecPtr found = InvalidXLogRecPtr;
3735 alvherre 1374 ECB : XLogPageHeader header;
1375 : char *errormsg;
1376 :
699 tmunro 1377 CBC 15 : Assert(!XLogRecPtrIsInvalid(RecPtr));
1378 :
1379 : /* Make sure ReadPageInternal() can't return XLREAD_WOULDBLOCK. */
387 tmunro 1380 GIC 15 : state->nonblocking = false;
387 tmunro 1381 ECB :
1382 : /*
1383 : * skip over potential continuation data, keeping in mind that it may span
2414 fujii 1384 : * multiple pages
1385 : */
699 tmunro 1386 GIC 15 : tmpRecPtr = RecPtr;
1387 : while (true)
2414 fujii 1388 UIC 0 : {
1389 : XLogRecPtr targetPagePtr;
2414 fujii 1390 ECB : int targetRecOff;
1391 : uint32 pageHeaderSize;
699 tmunro 1392 EUB : int readLen;
1393 :
1394 : /*
1395 : * Compute targetRecOff. It should typically be equal or greater than
1396 : * short page-header since a valid record can't start anywhere before
1397 : * that, except when caller has explicitly specified the offset that
1398 : * falls somewhere there or when we are skipping multi-page
1399 : * continuation record. It doesn't matter though because
1400 : * ReadPageInternal() is prepared to handle that and will read at
1401 : * least short page-header worth of data
1402 : */
699 tmunro 1403 GIC 15 : targetRecOff = tmpRecPtr % XLOG_BLCKSZ;
1404 :
1405 : /* scroll back to page boundary */
1406 15 : targetPagePtr = tmpRecPtr - targetRecOff;
3735 alvherre 1407 ECB :
1408 : /* Read the page containing the record */
699 tmunro 1409 GIC 15 : readLen = ReadPageInternal(state, targetPagePtr, targetRecOff);
699 tmunro 1410 CBC 15 : if (readLen < 0)
2414 fujii 1411 UIC 0 : goto err;
1412 :
699 tmunro 1413 CBC 15 : header = (XLogPageHeader) state->readBuf;
3735 alvherre 1414 ECB :
2414 fujii 1415 GBC 15 : pageHeaderSize = XLogPageHeaderSize(header);
1416 :
699 tmunro 1417 ECB : /* make sure we have enough data for the page header */
699 tmunro 1418 GIC 15 : readLen = ReadPageInternal(state, targetPagePtr, pageHeaderSize);
699 tmunro 1419 CBC 15 : if (readLen < 0)
699 tmunro 1420 UIC 0 : goto err;
1421 :
2414 fujii 1422 ECB : /* skip over potential continuation data */
2414 fujii 1423 CBC 15 : if (header->xlp_info & XLP_FIRST_IS_CONTRECORD)
2414 fujii 1424 EUB : {
1425 : /*
1426 : * If the length of the remaining continuation data is more than
2414 fujii 1427 ECB : * what can fit in this page, the continuation record crosses over
1428 : * this page. Read the next page and try again. xlp_rem_len in the
1429 : * next page header will contain the remaining length of the
1430 : * continuation data
1431 : *
1432 : * Note that record headers are MAXALIGN'ed
1433 : */
1249 fujii 1434 GIC 13 : if (MAXALIGN(header->xlp_rem_len) >= (XLOG_BLCKSZ - pageHeaderSize))
699 tmunro 1435 UIC 0 : tmpRecPtr = targetPagePtr + XLOG_BLCKSZ;
1436 : else
1437 : {
2414 fujii 1438 ECB : /*
2414 fujii 1439 EUB : * The previous continuation record ends in this page. Set
1440 : * tmpRecPtr to point to the first valid record
1441 : */
699 tmunro 1442 GIC 13 : tmpRecPtr = targetPagePtr + pageHeaderSize
2414 fujii 1443 13 : + MAXALIGN(header->xlp_rem_len);
1444 13 : break;
1445 : }
2414 fujii 1446 ECB : }
1447 : else
1448 : {
699 tmunro 1449 GIC 2 : tmpRecPtr = targetPagePtr + pageHeaderSize;
2414 fujii 1450 2 : break;
1451 : }
1452 : }
3735 alvherre 1453 ECB :
1454 : /*
1455 : * we know now that tmpRecPtr is an address pointing to a valid XLogRecord
1456 : * because either we're at the first record after the beginning of a page
1457 : * or we just jumped over the remaining data of a continuation.
1458 : */
699 tmunro 1459 GIC 15 : XLogBeginRead(state, tmpRecPtr);
1460 612 : while (XLogReadRecord(state, &errormsg) != NULL)
1461 : {
1462 : /* past the record we've found, break out */
699 tmunro 1463 CBC 612 : if (RecPtr <= state->ReadRecPtr)
3735 alvherre 1464 ECB : {
1465 : /* Rewind the reader to the beginning of the last record. */
699 tmunro 1466 GIC 15 : found = state->ReadRecPtr;
699 tmunro 1467 CBC 15 : XLogBeginRead(state, found);
699 tmunro 1468 GIC 15 : return found;
1469 : }
3735 alvherre 1470 ECB : }
1471 :
3735 alvherre 1472 LBC 0 : err:
699 tmunro 1473 UIC 0 : XLogReaderInvalReadState(state);
1474 :
1475 0 : return InvalidXLogRecPtr;
3735 alvherre 1476 EUB : }
1477 :
1478 : /*
699 tmunro 1479 : * Helper function to ease writing of XLogRoutine->page_read callbacks.
1480 : * If this function is used, caller must supply a segment_open callback in
1481 : * 'state', as that is used here.
1482 : *
1483 : * Read 'count' bytes into 'buf', starting at location 'startptr', from WAL
1484 : * fetched from timeline 'tli'.
1485 : *
1486 : * Returns true if succeeded, false if an error occurs, in which case
1487 : * 'errinfo' receives error details.
1488 : *
1489 : * XXX probably this should be improved to suck data directly from the
1490 : * WAL buffers when possible.
1491 : */
1492 : bool
1066 alvherre 1493 GIC 61072 : WALRead(XLogReaderState *state,
1494 : char *buf, XLogRecPtr startptr, Size count, TimeLineID tli,
1495 : WALReadError *errinfo)
1496 : {
1231 alvherre 1497 ECB : char *p;
1498 : XLogRecPtr recptr;
1499 : Size nbytes;
1500 :
1231 alvherre 1501 GIC 61072 : p = buf;
1502 61072 : recptr = startptr;
1503 61072 : nbytes = count;
1504 :
1231 alvherre 1505 CBC 122159 : while (nbytes > 0)
1231 alvherre 1506 ECB : {
1507 : uint32 startoff;
1508 : int segbytes;
1509 : int readbytes;
1510 :
1061 alvherre 1511 GIC 61088 : startoff = XLogSegmentOffset(recptr, state->segcxt.ws_segsize);
1512 :
1513 : /*
1514 : * If the data we want is not in a segment we have open, close what we
1231 alvherre 1515 ECB : * have (if anything) and open the next one, using the caller's
1516 : * provided openSegment callback.
1517 : */
1061 alvherre 1518 GIC 61088 : if (state->seg.ws_file < 0 ||
1519 59716 : !XLByteInSeg(recptr, state->seg.ws_segno, state->segcxt.ws_segsize) ||
1520 59556 : tli != state->seg.ws_tli)
1521 : {
1231 alvherre 1522 ECB : XLogSegNo nextSegNo;
1523 :
1061 alvherre 1524 CBC 1532 : if (state->seg.ws_file >= 0)
699 tmunro 1525 GIC 160 : state->routine.segment_close(state);
1526 :
1061 alvherre 1527 1532 : XLByteToSeg(recptr, nextSegNo, state->segcxt.ws_segsize);
699 tmunro 1528 CBC 1532 : state->routine.segment_open(state, nextSegNo, &tli);
1061 alvherre 1529 ECB :
1530 : /* This shouldn't happen -- indicates a bug in segment_open */
1061 alvherre 1531 CBC 1531 : Assert(state->seg.ws_file >= 0);
1231 alvherre 1532 ECB :
1533 : /* Update the current segment info. */
1061 alvherre 1534 GIC 1531 : state->seg.ws_tli = tli;
1061 alvherre 1535 CBC 1531 : state->seg.ws_segno = nextSegNo;
1536 : }
1537 :
1231 alvherre 1538 ECB : /* How many bytes are within this segment? */
1061 alvherre 1539 CBC 61087 : if (nbytes > (state->segcxt.ws_segsize - startoff))
1061 alvherre 1540 GIC 16 : segbytes = state->segcxt.ws_segsize - startoff;
1541 : else
1231 1542 61071 : segbytes = nbytes;
1231 alvherre 1543 ECB :
1544 : #ifndef FRONTEND
1231 alvherre 1545 GIC 61087 : pgstat_report_wait_start(WAIT_EVENT_WAL_READ);
1231 alvherre 1546 ECB : #endif
1547 :
1548 : /* Reset errno first; eases reporting non-errno-affecting errors */
1231 alvherre 1549 CBC 61087 : errno = 0;
192 tmunro 1550 GIC 61087 : readbytes = pg_pread(state->seg.ws_file, p, segbytes, (off_t) startoff);
1551 :
1552 : #ifndef FRONTEND
1231 alvherre 1553 CBC 61087 : pgstat_report_wait_end();
1231 alvherre 1554 ECB : #endif
1555 :
1231 alvherre 1556 GIC 61087 : if (readbytes <= 0)
1231 alvherre 1557 ECB : {
1231 alvherre 1558 UIC 0 : errinfo->wre_errno = errno;
1559 0 : errinfo->wre_req = segbytes;
1231 alvherre 1560 LBC 0 : errinfo->wre_read = readbytes;
1231 alvherre 1561 UIC 0 : errinfo->wre_off = startoff;
1061 alvherre 1562 UBC 0 : errinfo->wre_seg = state->seg;
1231 1563 0 : return false;
1231 alvherre 1564 EUB : }
1565 :
1566 : /* Update state for read */
1231 alvherre 1567 GBC 61087 : recptr += readbytes;
1231 alvherre 1568 GIC 61087 : nbytes -= readbytes;
1569 61087 : p += readbytes;
1570 : }
1231 alvherre 1571 ECB :
1231 alvherre 1572 CBC 61071 : return true;
1231 alvherre 1573 ECB : }
1574 :
1575 : /* ----------------------------------------
3062 heikki.linnakangas 1576 : * Functions for decoding the data and block references in a record.
1577 : * ----------------------------------------
1578 : */
1579 :
1580 : /*
1581 : * Private function to reset the state, forgetting all decoded records, if we
1582 : * are asked to move to a new read position.
1583 : */
1584 : static void
3062 heikki.linnakangas 1585 GIC 3709 : ResetDecoder(XLogReaderState *state)
1586 : {
1587 : DecodedXLogRecord *r;
1588 :
387 tmunro 1589 ECB : /* Reset the decoded record queue, freeing any oversized records. */
387 tmunro 1590 GIC 8596 : while ((r = state->decode_queue_head) != NULL)
1591 : {
1592 1178 : state->decode_queue_head = r->next;
1593 1178 : if (r->oversized)
387 tmunro 1594 LBC 0 : pfree(r);
1595 : }
387 tmunro 1596 CBC 3709 : state->decode_queue_tail = NULL;
1597 3709 : state->decode_queue_head = NULL;
387 tmunro 1598 GBC 3709 : state->record = NULL;
1599 :
387 tmunro 1600 ECB : /* Reset the decode buffer to empty. */
387 tmunro 1601 CBC 3709 : state->decode_buffer_tail = state->decode_buffer;
1602 3709 : state->decode_buffer_head = state->decode_buffer;
1603 :
1604 : /* Clear error state. */
1605 3709 : state->errormsg_buf[0] = '\0';
1606 3709 : state->errormsg_deferred = false;
3062 heikki.linnakangas 1607 GIC 3709 : }
1608 :
3062 heikki.linnakangas 1609 ECB : /*
387 tmunro 1610 : * Compute the maximum possible amount of padding that could be required to
1611 : * decode a record, given xl_tot_len from the record's header. This is the
1612 : * amount of output buffer space that we need to decode a record, though we
1613 : * might not finish up using it all.
1614 : *
1615 : * This computation is pessimistic and assumes the maximum possible number of
1616 : * blocks, due to lack of better information.
1617 : */
1618 : size_t
387 tmunro 1619 GIC 10039173 : DecodeXLogRecordRequiredSpace(size_t xl_tot_len)
1620 : {
1621 10039173 : size_t size = 0;
1622 :
387 tmunro 1623 ECB : /* Account for the fixed size part of the decoded record struct. */
387 tmunro 1624 GIC 10039173 : size += offsetof(DecodedXLogRecord, blocks[0]);
387 tmunro 1625 ECB : /* Account for the flexible blocks array of maximum possible size. */
387 tmunro 1626 GIC 10039173 : size += sizeof(DecodedBkpBlock) * (XLR_MAX_BLOCK_ID + 1);
1627 : /* Account for all the raw main and block data. */
387 tmunro 1628 CBC 10039173 : size += xl_tot_len;
1629 : /* We might insert padding before main_data. */
1630 10039173 : size += (MAXIMUM_ALIGNOF - 1);
1631 : /* We might insert padding before each block's data. */
1632 10039173 : size += (MAXIMUM_ALIGNOF - 1) * (XLR_MAX_BLOCK_ID + 1);
1633 : /* We might insert padding at the end. */
1634 10039173 : size += (MAXIMUM_ALIGNOF - 1);
1635 :
1636 10039173 : return size;
1637 : }
387 tmunro 1638 ECB :
1639 : /*
1640 : * Decode a record. "decoded" must point to a MAXALIGNed memory area that has
1641 : * space for at least DecodeXLogRecordRequiredSpace(record) bytes. On
1642 : * success, decoded->size contains the actual space occupied by the decoded
1643 : * record, which may turn out to be less.
1644 : *
1645 : * Only decoded->oversized member must be initialized already, and will not be
1646 : * modified. Other members will be initialized as required.
1647 : *
1648 : * On error, a human-readable error message is returned in *errormsg, and
1649 : * the return value is false.
1650 : */
1651 : bool
387 tmunro 1652 GIC 5012882 : DecodeXLogRecord(XLogReaderState *state,
1653 : DecodedXLogRecord *decoded,
1654 : XLogRecord *record,
1655 : XLogRecPtr lsn,
387 tmunro 1656 ECB : char **errormsg)
1657 : {
1658 : /*
1659 : * read next _size bytes from record buffer, but check for overrun first.
1660 : */
1661 : #define COPY_HEADER_FIELD(_dst, _size) \
1662 : do { \
1663 : if (remaining < _size) \
1664 : goto shortdata_err; \
1665 : memcpy(_dst, ptr, _size); \
1666 : ptr += _size; \
1667 : remaining -= _size; \
1668 : } while(0)
1669 :
1670 : char *ptr;
1671 : char *out;
1672 : uint32 remaining;
1673 : uint32 datatotal;
277 rhaas 1674 GNC 5012882 : RelFileLocator *rlocator = NULL;
1675 : uint8 block_id;
1676 :
387 tmunro 1677 GIC 5012882 : decoded->header = *record;
387 tmunro 1678 CBC 5012882 : decoded->lsn = lsn;
387 tmunro 1679 GIC 5012882 : decoded->next = NULL;
1680 5012882 : decoded->record_origin = InvalidRepOriginId;
387 tmunro 1681 CBC 5012882 : decoded->toplevel_xid = InvalidTransactionId;
1682 5012882 : decoded->main_data = NULL;
1683 5012882 : decoded->main_data_len = 0;
1684 5012882 : decoded->max_block_id = -1;
3062 heikki.linnakangas 1685 5012882 : ptr = (char *) record;
1686 5012882 : ptr += SizeOfXLogRecord;
1687 5012882 : remaining = record->xl_tot_len - SizeOfXLogRecord;
3062 heikki.linnakangas 1688 ECB :
1689 : /* Decode the headers */
3062 heikki.linnakangas 1690 CBC 5012882 : datatotal = 0;
1691 10315759 : while (remaining > datatotal)
1692 : {
3062 heikki.linnakangas 1693 GIC 10284929 : COPY_HEADER_FIELD(&block_id, sizeof(uint8));
3062 heikki.linnakangas 1694 ECB :
3062 heikki.linnakangas 1695 CBC 10284929 : if (block_id == XLR_BLOCK_ID_DATA_SHORT)
1696 : {
3062 heikki.linnakangas 1697 ECB : /* XLogRecordDataHeaderShort */
1698 : uint8 main_data_len;
1699 :
3062 heikki.linnakangas 1700 GIC 4973683 : COPY_HEADER_FIELD(&main_data_len, sizeof(uint8));
1701 :
387 tmunro 1702 4973683 : decoded->main_data_len = main_data_len;
3062 heikki.linnakangas 1703 4973683 : datatotal += main_data_len;
3062 heikki.linnakangas 1704 CBC 4973683 : break; /* by convention, the main data fragment is
1705 : * always last */
3062 heikki.linnakangas 1706 ECB : }
3062 heikki.linnakangas 1707 CBC 5311246 : else if (block_id == XLR_BLOCK_ID_DATA_LONG)
3062 heikki.linnakangas 1708 ECB : {
1709 : /* XLogRecordDataHeaderLong */
1710 : uint32 main_data_len;
1711 :
3062 heikki.linnakangas 1712 GIC 8369 : COPY_HEADER_FIELD(&main_data_len, sizeof(uint32));
387 tmunro 1713 8369 : decoded->main_data_len = main_data_len;
3062 heikki.linnakangas 1714 8369 : datatotal += main_data_len;
1715 8369 : break; /* by convention, the main data fragment is
3062 heikki.linnakangas 1716 ECB : * always last */
1717 : }
2902 andres 1718 CBC 5302877 : else if (block_id == XLR_BLOCK_ID_ORIGIN)
2902 andres 1719 ECB : {
387 tmunro 1720 GIC 12899 : COPY_HEADER_FIELD(&decoded->record_origin, sizeof(RepOriginId));
1721 : }
993 akapila 1722 CBC 5289978 : else if (block_id == XLR_BLOCK_ID_TOPLEVEL_XID)
1723 : {
387 tmunro 1724 677 : COPY_HEADER_FIELD(&decoded->toplevel_xid, sizeof(TransactionId));
1725 : }
3062 heikki.linnakangas 1726 5289301 : else if (block_id <= XLR_MAX_BLOCK_ID)
1727 : {
3062 heikki.linnakangas 1728 ECB : /* XLogRecordBlockHeader */
1729 : DecodedBkpBlock *blk;
1730 : uint8 fork_flags;
1731 :
1732 : /* mark any intervening block IDs as not in use */
387 tmunro 1733 GIC 5291658 : for (int i = decoded->max_block_id + 1; i < block_id; ++i)
1734 2357 : decoded->blocks[i].in_use = false;
1735 :
1736 5289301 : if (block_id <= decoded->max_block_id)
3062 heikki.linnakangas 1737 ECB : {
3062 heikki.linnakangas 1738 LBC 0 : report_invalid_record(state,
1739 : "out-of-order block_id %u at %X/%X",
3062 heikki.linnakangas 1740 ECB : block_id,
775 peter 1741 UIC 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
3062 heikki.linnakangas 1742 UBC 0 : goto err;
1743 : }
387 tmunro 1744 GIC 5289301 : decoded->max_block_id = block_id;
3062 heikki.linnakangas 1745 EUB :
387 tmunro 1746 GBC 5289301 : blk = &decoded->blocks[block_id];
3062 heikki.linnakangas 1747 GIC 5289301 : blk->in_use = true;
2251 rhaas 1748 CBC 5289301 : blk->apply_image = false;
1749 :
3062 heikki.linnakangas 1750 5289301 : COPY_HEADER_FIELD(&fork_flags, sizeof(uint8));
1751 5289301 : blk->forknum = fork_flags & BKPBLOCK_FORK_MASK;
1752 5289301 : blk->flags = fork_flags;
3062 heikki.linnakangas 1753 GIC 5289301 : blk->has_image = ((fork_flags & BKPBLOCK_HAS_IMAGE) != 0);
3062 heikki.linnakangas 1754 CBC 5289301 : blk->has_data = ((fork_flags & BKPBLOCK_HAS_DATA) != 0);
3062 heikki.linnakangas 1755 ECB :
367 tmunro 1756 CBC 5289301 : blk->prefetch_buffer = InvalidBuffer;
367 tmunro 1757 ECB :
3062 heikki.linnakangas 1758 CBC 5289301 : COPY_HEADER_FIELD(&blk->data_len, sizeof(uint16));
1759 : /* cross-check that the HAS_DATA flag is set iff data_length > 0 */
1760 5289301 : if (blk->has_data && blk->data_len == 0)
1761 : {
3062 heikki.linnakangas 1762 LBC 0 : report_invalid_record(state,
1763 : "BKPBLOCK_HAS_DATA set, but no data included at %X/%X",
775 peter 1764 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
2953 fujii 1765 UIC 0 : goto err;
2953 fujii 1766 EUB : }
3062 heikki.linnakangas 1767 GIC 5289301 : if (!blk->has_data && blk->data_len != 0)
2953 fujii 1768 EUB : {
3062 heikki.linnakangas 1769 UBC 0 : report_invalid_record(state,
1770 : "BKPBLOCK_HAS_DATA not set, but data length is %u at %X/%X",
3062 heikki.linnakangas 1771 LBC 0 : (unsigned int) blk->data_len,
775 peter 1772 UIC 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
2953 fujii 1773 UBC 0 : goto err;
1774 : }
3062 heikki.linnakangas 1775 GBC 5289301 : datatotal += blk->data_len;
3062 heikki.linnakangas 1776 EUB :
3062 heikki.linnakangas 1777 GBC 5289301 : if (blk->has_image)
1778 : {
2951 fujii 1779 CBC 44761 : COPY_HEADER_FIELD(&blk->bimg_len, sizeof(uint16));
3062 heikki.linnakangas 1780 GIC 44761 : COPY_HEADER_FIELD(&blk->hole_offset, sizeof(uint16));
2951 fujii 1781 CBC 44761 : COPY_HEADER_FIELD(&blk->bimg_info, sizeof(uint8));
1782 :
2251 rhaas 1783 44761 : blk->apply_image = ((blk->bimg_info & BKPIMAGE_APPLY) != 0);
2251 rhaas 1784 ECB :
649 michael 1785 CBC 44761 : if (BKPIMAGE_COMPRESSED(blk->bimg_info))
1786 : {
2951 fujii 1787 LBC 0 : if (blk->bimg_info & BKPIMAGE_HAS_HOLE)
2951 fujii 1788 UIC 0 : COPY_HEADER_FIELD(&blk->hole_length, sizeof(uint16));
2951 fujii 1789 ECB : else
2951 fujii 1790 UIC 0 : blk->hole_length = 0;
2951 fujii 1791 EUB : }
1792 : else
2951 fujii 1793 GIC 44761 : blk->hole_length = BLCKSZ - blk->bimg_len;
2951 fujii 1794 GBC 44761 : datatotal += blk->bimg_len;
1795 :
1796 : /*
2951 fujii 1797 ECB : * cross-check that hole_offset > 0, hole_length > 0 and
1798 : * bimg_len < BLCKSZ if the HAS_HOLE flag is set.
1799 : */
2951 fujii 1800 GIC 44761 : if ((blk->bimg_info & BKPIMAGE_HAS_HOLE) &&
1801 41467 : (blk->hole_offset == 0 ||
1802 41467 : blk->hole_length == 0 ||
1803 41467 : blk->bimg_len == BLCKSZ))
2951 fujii 1804 ECB : {
2951 fujii 1805 LBC 0 : report_invalid_record(state,
2878 bruce 1806 ECB : "BKPIMAGE_HAS_HOLE set, but hole offset %u length %u block image length %u at %X/%X",
2951 fujii 1807 LBC 0 : (unsigned int) blk->hole_offset,
2951 fujii 1808 UIC 0 : (unsigned int) blk->hole_length,
2951 fujii 1809 UBC 0 : (unsigned int) blk->bimg_len,
775 peter 1810 UIC 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
2951 fujii 1811 UBC 0 : goto err;
2951 fujii 1812 EUB : }
2878 bruce 1813 :
2951 fujii 1814 : /*
2878 bruce 1815 : * cross-check that hole_offset == 0 and hole_length == 0 if
1816 : * the HAS_HOLE flag is not set.
1817 : */
2951 fujii 1818 GIC 44761 : if (!(blk->bimg_info & BKPIMAGE_HAS_HOLE) &&
1819 3294 : (blk->hole_offset != 0 || blk->hole_length != 0))
1820 : {
2951 fujii 1821 UIC 0 : report_invalid_record(state,
2878 bruce 1822 ECB : "BKPIMAGE_HAS_HOLE not set, but hole offset %u length %u at %X/%X",
2951 fujii 1823 LBC 0 : (unsigned int) blk->hole_offset,
2951 fujii 1824 UIC 0 : (unsigned int) blk->hole_length,
775 peter 1825 UBC 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
2951 fujii 1826 UIC 0 : goto err;
2951 fujii 1827 EUB : }
2878 bruce 1828 :
2951 fujii 1829 : /*
649 michael 1830 : * Cross-check that bimg_len < BLCKSZ if it is compressed.
1831 : */
649 michael 1832 GIC 44761 : if (BKPIMAGE_COMPRESSED(blk->bimg_info) &&
2951 fujii 1833 UIC 0 : blk->bimg_len == BLCKSZ)
1834 : {
1835 0 : report_invalid_record(state,
649 michael 1836 ECB : "BKPIMAGE_COMPRESSED set, but block image length %u at %X/%X",
2951 fujii 1837 UBC 0 : (unsigned int) blk->bimg_len,
775 peter 1838 UIC 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
2951 fujii 1839 UBC 0 : goto err;
1840 : }
2878 bruce 1841 EUB :
2951 fujii 1842 : /*
649 michael 1843 : * cross-check that bimg_len = BLCKSZ if neither HAS_HOLE is
1844 : * set nor COMPRESSED().
1845 : */
2951 fujii 1846 GIC 44761 : if (!(blk->bimg_info & BKPIMAGE_HAS_HOLE) &&
649 michael 1847 3294 : !BKPIMAGE_COMPRESSED(blk->bimg_info) &&
2951 fujii 1848 3294 : blk->bimg_len != BLCKSZ)
1849 : {
2951 fujii 1850 LBC 0 : report_invalid_record(state,
649 michael 1851 ECB : "neither BKPIMAGE_HAS_HOLE nor BKPIMAGE_COMPRESSED set, but block image length is %u at %X/%X",
2951 fujii 1852 LBC 0 : (unsigned int) blk->data_len,
775 peter 1853 UIC 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
2951 fujii 1854 UBC 0 : goto err;
1855 : }
3062 heikki.linnakangas 1856 EUB : }
3062 heikki.linnakangas 1857 GBC 5289301 : if (!(fork_flags & BKPBLOCK_SAME_REL))
3062 heikki.linnakangas 1858 EUB : {
277 rhaas 1859 GNC 4919523 : COPY_HEADER_FIELD(&blk->rlocator, sizeof(RelFileLocator));
1860 4919523 : rlocator = &blk->rlocator;
3062 heikki.linnakangas 1861 ECB : }
1862 : else
1863 : {
277 rhaas 1864 GNC 369778 : if (rlocator == NULL)
1865 : {
3062 heikki.linnakangas 1866 UIC 0 : report_invalid_record(state,
1867 : "BKPBLOCK_SAME_REL set but no previous rel at %X/%X",
775 peter 1868 LBC 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
3062 heikki.linnakangas 1869 UIC 0 : goto err;
3062 heikki.linnakangas 1870 EUB : }
1871 :
277 rhaas 1872 GNC 369778 : blk->rlocator = *rlocator;
3062 heikki.linnakangas 1873 EUB : }
3062 heikki.linnakangas 1874 GIC 5289301 : COPY_HEADER_FIELD(&blk->blkno, sizeof(BlockNumber));
1875 : }
3062 heikki.linnakangas 1876 ECB : else
1877 : {
3062 heikki.linnakangas 1878 LBC 0 : report_invalid_record(state,
1879 : "invalid block_id %u at %X/%X",
775 peter 1880 UIC 0 : block_id, LSN_FORMAT_ARGS(state->ReadRecPtr));
3062 heikki.linnakangas 1881 0 : goto err;
3062 heikki.linnakangas 1882 EUB : }
1883 : }
1884 :
3062 heikki.linnakangas 1885 GBC 5012882 : if (remaining != datatotal)
3062 heikki.linnakangas 1886 UIC 0 : goto shortdata_err;
1887 :
1888 : /*
3062 heikki.linnakangas 1889 ECB : * Ok, we've parsed the fragment headers, and verified that the total
3062 heikki.linnakangas 1890 EUB : * length of the payload in the fragments is equal to the amount of data
1891 : * left. Copy the data of each fragment to contiguous space after the
1892 : * blocks array, inserting alignment padding before the data fragments so
1893 : * they can be cast to struct pointers by REDO routines.
1894 : */
387 tmunro 1895 GIC 5012882 : out = ((char *) decoded) +
1896 5012882 : offsetof(DecodedXLogRecord, blocks) +
1897 5012882 : sizeof(decoded->blocks[0]) * (decoded->max_block_id + 1);
1898 :
3062 heikki.linnakangas 1899 ECB : /* block data first */
387 tmunro 1900 CBC 10304540 : for (block_id = 0; block_id <= decoded->max_block_id; block_id++)
3062 heikki.linnakangas 1901 ECB : {
387 tmunro 1902 GIC 5291658 : DecodedBkpBlock *blk = &decoded->blocks[block_id];
1903 :
3062 heikki.linnakangas 1904 CBC 5291658 : if (!blk->in_use)
3062 heikki.linnakangas 1905 GIC 2357 : continue;
2251 rhaas 1906 ECB :
2251 rhaas 1907 GIC 5289301 : Assert(blk->has_image || !blk->apply_image);
2251 rhaas 1908 ECB :
3062 heikki.linnakangas 1909 CBC 5289301 : if (blk->has_image)
1910 : {
387 tmunro 1911 ECB : /* no need to align image */
387 tmunro 1912 GIC 44761 : blk->bkp_image = out;
387 tmunro 1913 CBC 44761 : memcpy(out, ptr, blk->bimg_len);
2951 fujii 1914 GIC 44761 : ptr += blk->bimg_len;
387 tmunro 1915 44761 : out += blk->bimg_len;
3062 heikki.linnakangas 1916 ECB : }
3062 heikki.linnakangas 1917 CBC 5289301 : if (blk->has_data)
3062 heikki.linnakangas 1918 ECB : {
387 tmunro 1919 CBC 4027016 : out = (char *) MAXALIGN(out);
387 tmunro 1920 GIC 4027016 : blk->data = out;
3062 heikki.linnakangas 1921 CBC 4027016 : memcpy(blk->data, ptr, blk->data_len);
3062 heikki.linnakangas 1922 GIC 4027016 : ptr += blk->data_len;
387 tmunro 1923 CBC 4027016 : out += blk->data_len;
3062 heikki.linnakangas 1924 ECB : }
1925 : }
1926 :
1927 : /* and finally, the main data */
387 tmunro 1928 GIC 5012882 : if (decoded->main_data_len > 0)
1929 : {
1930 4982052 : out = (char *) MAXALIGN(out);
1931 4982052 : decoded->main_data = out;
387 tmunro 1932 CBC 4982052 : memcpy(decoded->main_data, ptr, decoded->main_data_len);
387 tmunro 1933 GIC 4982052 : ptr += decoded->main_data_len;
387 tmunro 1934 CBC 4982052 : out += decoded->main_data_len;
699 tmunro 1935 ECB : }
731 1936 :
387 1937 : /* Report the actual size we used. */
387 tmunro 1938 CBC 5012882 : decoded->size = MAXALIGN(out - (char *) decoded);
387 tmunro 1939 GIC 5012882 : Assert(DecodeXLogRecordRequiredSpace(record->xl_tot_len) >=
1940 : decoded->size);
1941 :
3062 heikki.linnakangas 1942 CBC 5012882 : return true;
3062 heikki.linnakangas 1943 ECB :
3062 heikki.linnakangas 1944 UIC 0 : shortdata_err:
1945 0 : report_invalid_record(state,
3062 heikki.linnakangas 1946 ECB : "record with invalid length at %X/%X",
775 peter 1947 UIC 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
3062 heikki.linnakangas 1948 UBC 0 : err:
1949 0 : *errormsg = state->errormsg_buf;
1950 :
1951 0 : return false;
3062 heikki.linnakangas 1952 EUB : }
1953 :
1954 : /*
1955 : * Returns information about the block that a block reference refers to.
1956 : *
1957 : * This is like XLogRecGetBlockTagExtended, except that the block reference
1958 : * must exist and there's no access to prefetch_buffer.
1959 : */
1960 : void
3062 heikki.linnakangas 1961 GIC 3310125 : XLogRecGetBlockTag(XLogReaderState *record, uint8 block_id,
1962 : RelFileLocator *rlocator, ForkNumber *forknum,
1963 : BlockNumber *blknum)
1964 : {
277 rhaas 1965 GNC 3310125 : if (!XLogRecGetBlockTagExtended(record, block_id, rlocator, forknum,
1966 : blknum, NULL))
1967 : {
1968 : #ifndef FRONTEND
197 peter 1969 UIC 0 : elog(ERROR, "could not locate backup block with ID %d in WAL record",
363 tgl 1970 ECB : block_id);
1971 : #else
1972 : pg_fatal("could not locate backup block with ID %d in WAL record",
1973 : block_id);
363 tgl 1974 EUB : #endif
1975 : }
367 tmunro 1976 GIC 3310125 : }
1977 :
1978 : /*
1979 : * Returns information about the block that a block reference refers to,
1980 : * optionally including the buffer that the block may already be in.
367 tmunro 1981 ECB : *
1982 : * If the WAL record contains a block reference with the given ID, *rlocator,
1983 : * *forknum, *blknum and *prefetch_buffer are filled in (if not NULL), and
1984 : * returns true. Otherwise returns false.
1985 : */
1986 : bool
367 tmunro 1987 GIC 6089138 : XLogRecGetBlockTagExtended(XLogReaderState *record, uint8 block_id,
1988 : RelFileLocator *rlocator, ForkNumber *forknum,
1989 : BlockNumber *blknum,
1990 : Buffer *prefetch_buffer)
1991 : {
3062 heikki.linnakangas 1992 ECB : DecodedBkpBlock *bkpb;
1993 :
363 tgl 1994 GIC 6089138 : if (!XLogRecHasBlockRef(record, block_id))
3062 heikki.linnakangas 1995 33965 : return false;
1996 :
387 tmunro 1997 6055173 : bkpb = &record->record->blocks[block_id];
277 rhaas 1998 GNC 6055173 : if (rlocator)
1999 5996547 : *rlocator = bkpb->rlocator;
3062 heikki.linnakangas 2000 CBC 6055173 : if (forknum)
3062 heikki.linnakangas 2001 GIC 2691196 : *forknum = bkpb->forknum;
3062 heikki.linnakangas 2002 CBC 6055173 : if (blknum)
2003 4369662 : *blknum = bkpb->blkno;
367 tmunro 2004 6055173 : if (prefetch_buffer)
2005 2654768 : *prefetch_buffer = bkpb->prefetch_buffer;
3062 heikki.linnakangas 2006 6055173 : return true;
3062 heikki.linnakangas 2007 ECB : }
2008 :
2009 : /*
2010 : * Returns the data associated with a block reference, or NULL if there is
2011 : * no data (e.g. because a full-page image was taken instead). The returned
2012 : * pointer points to a MAXALIGNed buffer.
2013 : */
2014 : char *
3062 heikki.linnakangas 2015 GIC 3436025 : XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len)
2016 : {
2017 : DecodedBkpBlock *bkpb;
2018 :
387 tmunro 2019 3436025 : if (block_id > record->record->max_block_id ||
387 tmunro 2020 CBC 3436025 : !record->record->blocks[block_id].in_use)
3062 heikki.linnakangas 2021 UIC 0 : return NULL;
2022 :
387 tmunro 2023 GIC 3436025 : bkpb = &record->record->blocks[block_id];
3062 heikki.linnakangas 2024 ECB :
3062 heikki.linnakangas 2025 CBC 3436025 : if (!bkpb->has_data)
3062 heikki.linnakangas 2026 EUB : {
3062 heikki.linnakangas 2027 GIC 224 : if (len)
3062 heikki.linnakangas 2028 CBC 224 : *len = 0;
3062 heikki.linnakangas 2029 GIC 224 : return NULL;
3062 heikki.linnakangas 2030 ECB : }
2031 : else
2032 : {
3062 heikki.linnakangas 2033 CBC 3435801 : if (len)
2034 3432270 : *len = bkpb->data_len;
3062 heikki.linnakangas 2035 GIC 3435801 : return bkpb->data;
2036 : }
2037 : }
3062 heikki.linnakangas 2038 ECB :
2039 : /*
2040 : * Restore a full-page image from a backup block attached to an XLOG record.
2041 : *
2042 : * Returns true if a full-page image is restored, and false on failure with
2043 : * an error to be consumed by the caller.
2044 : */
2045 : bool
3062 heikki.linnakangas 2046 GIC 36508 : RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page)
2047 : {
2048 : DecodedBkpBlock *bkpb;
2049 : char *ptr;
2050 : PGAlignedBlock tmp;
3062 heikki.linnakangas 2051 ECB :
387 tmunro 2052 GIC 36508 : if (block_id > record->record->max_block_id ||
2053 36508 : !record->record->blocks[block_id].in_use)
2054 : {
212 michael 2055 UIC 0 : report_invalid_record(record,
2056 : "could not restore image at %X/%X with invalid block %d specified",
212 michael 2057 LBC 0 : LSN_FORMAT_ARGS(record->ReadRecPtr),
212 michael 2058 ECB : block_id);
3062 heikki.linnakangas 2059 UIC 0 : return false;
212 michael 2060 EUB : }
387 tmunro 2061 GIC 36508 : if (!record->record->blocks[block_id].has_image)
212 michael 2062 EUB : {
212 michael 2063 UIC 0 : report_invalid_record(record, "could not restore image at %X/%X with invalid state, block %d",
212 michael 2064 UBC 0 : LSN_FORMAT_ARGS(record->ReadRecPtr),
2065 : block_id);
3062 heikki.linnakangas 2066 LBC 0 : return false;
2067 : }
3062 heikki.linnakangas 2068 EUB :
387 tmunro 2069 GBC 36508 : bkpb = &record->record->blocks[block_id];
2951 fujii 2070 GIC 36508 : ptr = bkpb->bkp_image;
2951 fujii 2071 EUB :
649 michael 2072 GIC 36508 : if (BKPIMAGE_COMPRESSED(bkpb->bimg_info))
2073 : {
2951 fujii 2074 ECB : /* If a backup block image is compressed, decompress it */
649 michael 2075 LBC 0 : bool decomp_success = true;
2076 :
2077 0 : if ((bkpb->bimg_info & BKPIMAGE_COMPRESS_PGLZ) != 0)
2078 : {
649 michael 2079 UIC 0 : if (pglz_decompress(ptr, bkpb->bimg_len, tmp.data,
649 michael 2080 UBC 0 : BLCKSZ - bkpb->hole_length, true) < 0)
649 michael 2081 UIC 0 : decomp_success = false;
649 michael 2082 EUB : }
649 michael 2083 UIC 0 : else if ((bkpb->bimg_info & BKPIMAGE_COMPRESS_LZ4) != 0)
649 michael 2084 EUB : {
2085 : #ifdef USE_LZ4
649 michael 2086 UBC 0 : if (LZ4_decompress_safe(ptr, tmp.data,
649 michael 2087 UIC 0 : bkpb->bimg_len, BLCKSZ - bkpb->hole_length) <= 0)
649 michael 2088 UBC 0 : decomp_success = false;
2089 : #else
2090 : report_invalid_record(record, "could not restore image at %X/%X compressed with %s not supported by build, block %d",
639 michael 2091 EUB : LSN_FORMAT_ARGS(record->ReadRecPtr),
649 2092 : "LZ4",
2093 : block_id);
2094 : return false;
2095 : #endif
2096 : }
394 michael 2097 UIC 0 : else if ((bkpb->bimg_info & BKPIMAGE_COMPRESS_ZSTD) != 0)
2098 : {
2099 : #ifdef USE_ZSTD
2100 0 : size_t decomp_result = ZSTD_decompress(tmp.data,
2101 0 : BLCKSZ - bkpb->hole_length,
394 michael 2102 UBC 0 : ptr, bkpb->bimg_len);
2103 :
394 michael 2104 UIC 0 : if (ZSTD_isError(decomp_result))
394 michael 2105 UBC 0 : decomp_success = false;
394 michael 2106 EUB : #else
212 2107 : report_invalid_record(record, "could not restore image at %X/%X compressed with %s not supported by build, block %d",
2108 : LSN_FORMAT_ARGS(record->ReadRecPtr),
394 2109 : "zstd",
2110 : block_id);
2111 : return false;
2112 : #endif
2113 : }
2114 : else
2115 : {
212 michael 2116 UIC 0 : report_invalid_record(record, "could not restore image at %X/%X compressed with unknown method, block %d",
639 2117 0 : LSN_FORMAT_ARGS(record->ReadRecPtr),
2118 : block_id);
649 2119 0 : return false;
2120 : }
649 michael 2121 EUB :
649 michael 2122 UBC 0 : if (!decomp_success)
2123 : {
212 2124 0 : report_invalid_record(record, "could not decompress image at %X/%X, block %d",
775 peter 2125 UIC 0 : LSN_FORMAT_ARGS(record->ReadRecPtr),
2126 : block_id);
2951 fujii 2127 UBC 0 : return false;
2128 : }
649 michael 2129 EUB :
1681 tgl 2130 UBC 0 : ptr = tmp.data;
2131 : }
3062 heikki.linnakangas 2132 EUB :
2133 : /* generate page, taking into account hole if necessary */
3062 heikki.linnakangas 2134 GIC 36508 : if (bkpb->hole_length == 0)
3062 heikki.linnakangas 2135 EUB : {
2951 fujii 2136 GIC 1510 : memcpy(page, ptr, BLCKSZ);
2137 : }
2138 : else
3062 heikki.linnakangas 2139 ECB : {
2951 fujii 2140 GIC 34998 : memcpy(page, ptr, bkpb->hole_offset);
3062 heikki.linnakangas 2141 ECB : /* must zero-fill the hole */
3062 heikki.linnakangas 2142 GIC 258577 : MemSet(page + bkpb->hole_offset, 0, bkpb->hole_length);
2143 34998 : memcpy(page + (bkpb->hole_offset + bkpb->hole_length),
2951 fujii 2144 34998 : ptr + bkpb->hole_offset,
3062 heikki.linnakangas 2145 CBC 34998 : BLCKSZ - (bkpb->hole_offset + bkpb->hole_length));
2146 : }
3062 heikki.linnakangas 2147 ECB :
3062 heikki.linnakangas 2148 CBC 36508 : return true;
3062 heikki.linnakangas 2149 ECB : }
1364 tmunro 2150 :
2151 : #ifndef FRONTEND
2152 :
2153 : /*
2154 : * Extract the FullTransactionId from a WAL record.
2155 : */
2156 : FullTransactionId
1364 tmunro 2157 UIC 0 : XLogRecGetFullXid(XLogReaderState *record)
2158 : {
2159 : TransactionId xid,
2160 : next_xid;
2161 : uint32 epoch;
1364 tmunro 2162 EUB :
2163 : /*
2164 : * This function is only safe during replay, because it depends on the
2165 : * replay state. See AdvanceNextFullTransactionIdPastXid() for more.
2166 : */
1364 tmunro 2167 UIC 0 : Assert(AmStartupProcess() || !IsUnderPostmaster);
2168 :
2169 0 : xid = XLogRecGetXid(record);
971 andres 2170 0 : next_xid = XidFromFullTransactionId(ShmemVariableCache->nextXid);
2171 0 : epoch = EpochFromFullTransactionId(ShmemVariableCache->nextXid);
1364 tmunro 2172 EUB :
2173 : /*
1060 tgl 2174 : * If xid is numerically greater than next_xid, it has to be from the last
2175 : * epoch.
1364 tmunro 2176 : */
1364 tmunro 2177 UIC 0 : if (unlikely(xid > next_xid))
2178 0 : --epoch;
2179 :
2180 0 : return FullTransactionIdFromEpochAndXid(epoch, xid);
2181 : }
1364 tmunro 2182 EUB :
2183 : #endif
|
