Age Owner Branch data TLA Line data Source code
1 : : /*-------------------------------------------------------------------------
2 : : *
3 : : * xlogreader.c
4 : : * Generic XLog reading facility
5 : : *
6 : : * Portions Copyright (c) 2013-2026, 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 : : #include "storage/bufmgr.h"
39 : : #include "utils/wait_event.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 void 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
4806 alvherre@alvh.no-ip. 73 :CBC 274 : report_invalid_record(XLogReaderState *state, const char *fmt,...)
74 : : {
75 : : va_list args;
76 : :
77 : 274 : fmt = _(fmt);
78 : :
79 : 274 : va_start(args, fmt);
80 : 274 : vsnprintf(state->errormsg_buf, MAX_ERRORMSG_LEN, fmt, args);
81 : 274 : va_end(args);
82 : :
1458 tmunro@postgresql.or 83 : 274 : state->errormsg_deferred = true;
84 : 274 : }
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 : 1004 : XLogReaderSetDecodeBuffer(XLogReaderState *state, void *buffer, size_t size)
93 : : {
94 [ - + ]: 1004 : Assert(state->decode_buffer == NULL);
95 : :
96 : 1004 : state->decode_buffer = buffer;
97 : 1004 : state->decode_buffer_size = size;
98 : 1004 : state->decode_buffer_tail = buffer;
99 : 1004 : state->decode_buffer_head = buffer;
4806 alvherre@alvh.no-ip. 100 : 1004 : }
101 : :
102 : : /*
103 : : * Allocate and initialize a new XLogReader.
104 : : *
105 : : * Returns NULL if the xlogreader couldn't be allocated.
106 : : */
107 : : XLogReaderState *
2364 108 : 2997 : XLogReaderAllocate(int wal_segment_size, const char *waldir,
109 : : XLogReaderRoutine *routine, void *private_data)
110 : : {
111 : : XLogReaderState *state;
112 : :
113 : : state = (XLogReaderState *)
3999 fujii@postgresql.org 114 : 2997 : palloc_extended(sizeof(XLogReaderState),
115 : : MCXT_ALLOC_NO_OOM | MCXT_ALLOC_ZERO);
116 [ - + ]: 2997 : if (!state)
3999 fujii@postgresql.org 117 :UBC 0 : return NULL;
118 : :
119 : : /* initialize caller-provided support functions */
1770 tmunro@postgresql.or 120 :CBC 2997 : 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 : : */
3999 fujii@postgresql.org 129 : 2997 : state->readBuf = (char *) palloc_extended(XLOG_BLCKSZ,
130 : : MCXT_ALLOC_NO_OOM);
131 [ - + ]: 2997 : if (!state->readBuf)
132 : : {
3999 fujii@postgresql.org 133 :UBC 0 : pfree(state);
134 : 0 : return NULL;
135 : : }
136 : :
137 : : /* Initialize segment info. */
2364 alvherre@alvh.no-ip. 138 :CBC 2997 : WALOpenSegmentInit(&state->seg, &state->segcxt, wal_segment_size,
139 : : waldir);
140 : :
141 : : /* system_identifier initialized to zeroes above */
1770 tmunro@postgresql.or 142 : 2997 : state->private_data = private_data;
143 : : /* ReadRecPtr, EndRecPtr and readLen initialized to zeroes above */
3999 fujii@postgresql.org 144 : 2997 : state->errormsg_buf = palloc_extended(MAX_ERRORMSG_LEN + 1,
145 : : MCXT_ALLOC_NO_OOM);
146 [ - + ]: 2997 : if (!state->errormsg_buf)
147 : : {
3999 fujii@postgresql.org 148 :UBC 0 : pfree(state->readBuf);
149 : 0 : pfree(state);
150 : 0 : return NULL;
151 : : }
4806 alvherre@alvh.no-ip. 152 :CBC 2997 : state->errormsg_buf[0] = '\0';
153 : :
154 : : /*
155 : : * Allocate an initial readRecordBuf of minimal size, which can later be
156 : : * enlarged if necessary.
157 : : */
894 michael@paquier.xyz 158 : 2997 : allocate_recordbuf(state, 0);
4806 alvherre@alvh.no-ip. 159 : 2997 : return state;
160 : : }
161 : :
162 : : void
163 : 2405 : XLogReaderFree(XLogReaderState *state)
164 : : {
1770 tmunro@postgresql.or 165 [ + + ]: 2405 : if (state->seg.ws_file != -1)
166 : 1380 : state->routine.segment_close(state);
167 : :
1458 168 [ + + + - ]: 2405 : if (state->decode_buffer && state->free_decode_buffer)
169 : 2364 : pfree(state->decode_buffer);
170 : :
4133 heikki.linnakangas@i 171 : 2405 : pfree(state->errormsg_buf);
4806 alvherre@alvh.no-ip. 172 [ + - ]: 2405 : if (state->readRecordBuf)
4133 heikki.linnakangas@i 173 : 2405 : pfree(state->readRecordBuf);
174 : 2405 : pfree(state->readBuf);
175 : 2405 : pfree(state);
4806 alvherre@alvh.no-ip. 176 : 2405 : }
177 : :
178 : : /*
179 : : * Allocate readRecordBuf to fit a record of at least the given length.
180 : : *
181 : : * readRecordBufSize is set to the new buffer size.
182 : : *
183 : : * To avoid useless small increases, round its size to a multiple of
184 : : * XLOG_BLCKSZ, and make sure it's at least 5*Max(BLCKSZ, XLOG_BLCKSZ) to start
185 : : * with. (That is enough for all "normal" records, but very large commit or
186 : : * abort records might need more space.)
187 : : *
188 : : * Note: This routine should *never* be called for xl_tot_len until the header
189 : : * of the record has been fully validated.
190 : : */
191 : : static void
192 : 3034 : allocate_recordbuf(XLogReaderState *state, uint32 reclength)
193 : : {
194 : 3034 : uint32 newSize = reclength;
195 : :
196 : 3034 : newSize += XLOG_BLCKSZ - (newSize % XLOG_BLCKSZ);
197 : 3034 : newSize = Max(newSize, 5 * Max(BLCKSZ, XLOG_BLCKSZ));
198 : :
199 [ + + ]: 3034 : if (state->readRecordBuf)
4133 heikki.linnakangas@i 200 : 37 : pfree(state->readRecordBuf);
894 michael@paquier.xyz 201 : 3034 : state->readRecordBuf = (char *) palloc(newSize);
4806 alvherre@alvh.no-ip. 202 : 3034 : state->readRecordBufSize = newSize;
203 : 3034 : }
204 : :
205 : : /*
206 : : * Initialize the passed segment structs.
207 : : */
208 : : static void
2364 209 : 2997 : WALOpenSegmentInit(WALOpenSegment *seg, WALSegmentContext *segcxt,
210 : : int segsize, const char *waldir)
211 : : {
212 : 2997 : seg->ws_file = -1;
213 : 2997 : seg->ws_segno = 0;
214 : 2997 : seg->ws_tli = 0;
215 : :
216 : 2997 : segcxt->ws_segsize = segsize;
217 [ + + ]: 2997 : if (waldir)
218 : 112 : snprintf(segcxt->ws_dir, MAXPGPATH, "%s", waldir);
219 : 2997 : }
220 : :
221 : : /*
222 : : * Begin reading WAL at 'RecPtr'.
223 : : *
224 : : * 'RecPtr' should point to the beginning of a valid WAL record. Pointing at
225 : : * the beginning of a page is also OK, if there is a new record right after
226 : : * the page header, i.e. not a continuation.
227 : : *
228 : : * This does not make any attempt to read the WAL yet, and hence cannot fail.
229 : : * If the starting address is not correct, the first call to XLogReadRecord()
230 : : * will error out.
231 : : */
232 : : void
2240 heikki.linnakangas@i 233 : 6628 : XLogBeginRead(XLogReaderState *state, XLogRecPtr RecPtr)
234 : : {
129 alvherre@kurilemu.de 235 [ - + ]:GNC 6628 : Assert(XLogRecPtrIsValid(RecPtr));
236 : :
2240 heikki.linnakangas@i 237 :CBC 6628 : ResetDecoder(state);
238 : :
239 : : /* Begin at the passed-in record pointer. */
240 : 6628 : state->EndRecPtr = RecPtr;
1458 tmunro@postgresql.or 241 : 6628 : state->NextRecPtr = RecPtr;
2240 heikki.linnakangas@i 242 : 6628 : state->ReadRecPtr = InvalidXLogRecPtr;
1458 tmunro@postgresql.or 243 : 6628 : state->DecodeRecPtr = InvalidXLogRecPtr;
244 : 6628 : }
245 : :
246 : : /*
247 : : * Release the last record that was returned by XLogNextRecord(), if any, to
248 : : * free up space. Returns the LSN past the end of the record.
249 : : */
250 : : XLogRecPtr
251 : 11902195 : XLogReleasePreviousRecord(XLogReaderState *state)
252 : : {
253 : : DecodedXLogRecord *record;
254 : : XLogRecPtr next_lsn;
255 : :
256 [ + + ]: 11902195 : if (!state->record)
1284 257 : 5963902 : return InvalidXLogRecPtr;
258 : :
259 : : /*
260 : : * Remove it from the decoded record queue. It must be the oldest item
261 : : * decoded, decode_queue_head.
262 : : */
1458 263 : 5938293 : record = state->record;
1284 264 : 5938293 : next_lsn = record->next_lsn;
1458 265 [ - + ]: 5938293 : Assert(record == state->decode_queue_head);
266 : 5938293 : state->record = NULL;
267 : 5938293 : state->decode_queue_head = record->next;
268 : :
269 : : /* It might also be the newest item decoded, decode_queue_tail. */
270 [ + + ]: 5938293 : if (state->decode_queue_tail == record)
271 : 3105114 : state->decode_queue_tail = NULL;
272 : :
273 : : /* Release the space. */
274 [ + + ]: 5938293 : if (unlikely(record->oversized))
275 : : {
276 : : /* It's not in the decode buffer, so free it to release space. */
277 : 57 : pfree(record);
278 : : }
279 : : else
280 : : {
281 : : /* It must be the head (oldest) record in the decode buffer. */
282 [ - + ]: 5938236 : Assert(state->decode_buffer_head == (char *) record);
283 : :
284 : : /*
285 : : * We need to update head to point to the next record that is in the
286 : : * decode buffer, if any, being careful to skip oversized ones
287 : : * (they're not in the decode buffer).
288 : : */
289 : 5938236 : record = record->next;
290 [ + + - + : 5938236 : while (unlikely(record && record->oversized))
- + ]
1458 tmunro@postgresql.or 291 :UBC 0 : record = record->next;
292 : :
1458 tmunro@postgresql.or 293 [ + + ]:CBC 5938236 : if (record)
294 : : {
295 : : /* Adjust head to release space up to the next record. */
296 : 2833179 : state->decode_buffer_head = (char *) record;
297 : : }
298 : : else
299 : : {
300 : : /*
301 : : * Otherwise we might as well just reset head and tail to the
302 : : * start of the buffer space, because we're empty. This means
303 : : * we'll keep overwriting the same piece of memory if we're not
304 : : * doing any prefetching.
305 : : */
306 : 3105057 : state->decode_buffer_head = state->decode_buffer;
307 : 3105057 : state->decode_buffer_tail = state->decode_buffer;
308 : : }
309 : : }
310 : :
1284 311 : 5938293 : return next_lsn;
312 : : }
313 : :
314 : : /*
315 : : * Attempt to read an XLOG record.
316 : : *
317 : : * XLogBeginRead() or XLogFindNextRecord() and then XLogReadAhead() must be
318 : : * called before the first call to XLogNextRecord(). This functions returns
319 : : * records and errors that were put into an internal queue by XLogReadAhead().
320 : : *
321 : : * On success, a record is returned.
322 : : *
323 : : * The returned record (or *errormsg) points to an internal buffer that's
324 : : * valid until the next call to XLogNextRecord.
325 : : */
326 : : DecodedXLogRecord *
1458 327 : 5950965 : XLogNextRecord(XLogReaderState *state, char **errormsg)
328 : : {
329 : : /* Release the last record returned by XLogNextRecord(). */
330 : 5950965 : XLogReleasePreviousRecord(state);
331 : :
332 [ + + ]: 5950965 : if (state->decode_queue_head == NULL)
333 : : {
334 : 6754 : *errormsg = NULL;
335 [ + + ]: 6754 : if (state->errormsg_deferred)
336 : : {
337 [ + + ]: 275 : if (state->errormsg_buf[0] != '\0')
338 : 271 : *errormsg = state->errormsg_buf;
339 : 275 : state->errormsg_deferred = false;
340 : : }
341 : :
342 : : /*
343 : : * state->EndRecPtr is expected to have been set by the last call to
344 : : * XLogBeginRead() or XLogNextRecord(), and is the location of the
345 : : * error.
346 : : */
129 alvherre@kurilemu.de 347 [ - + ]:GNC 6754 : Assert(XLogRecPtrIsValid(state->EndRecPtr));
348 : :
1458 tmunro@postgresql.or 349 :CBC 6754 : return NULL;
350 : : }
351 : :
352 : : /*
353 : : * Record this as the most recent record returned, so that we'll release
354 : : * it next time. This also exposes it to the traditional
355 : : * XLogRecXXX(xlogreader) macros, which work with the decoder rather than
356 : : * the record for historical reasons.
357 : : */
358 : 5944211 : state->record = state->decode_queue_head;
359 : :
360 : : /*
361 : : * Update the pointers to the beginning and one-past-the-end of this
362 : : * record, again for the benefit of historical code that expected the
363 : : * decoder to track this rather than accessing these fields of the record
364 : : * itself.
365 : : */
366 : 5944211 : state->ReadRecPtr = state->record->lsn;
367 : 5944211 : state->EndRecPtr = state->record->next_lsn;
368 : :
369 : 5944211 : *errormsg = NULL;
370 : :
371 : 5944211 : return state->record;
372 : : }
373 : :
374 : : /*
375 : : * Attempt to read an XLOG record.
376 : : *
377 : : * XLogBeginRead() or XLogFindNextRecord() must be called before the first call
378 : : * to XLogReadRecord().
379 : : *
380 : : * If the page_read callback fails to read the requested data, NULL is
381 : : * returned. The callback is expected to have reported the error; errormsg
382 : : * is set to NULL.
383 : : *
384 : : * If the reading fails for some other reason, NULL is also returned, and
385 : : * *errormsg is set to a string with details of the failure.
386 : : *
387 : : * The returned pointer (or *errormsg) points to an internal buffer that's
388 : : * valid until the next call to XLogReadRecord.
389 : : */
390 : : XLogRecord *
1770 391 : 3086569 : XLogReadRecord(XLogReaderState *state, char **errormsg)
392 : : {
393 : : DecodedXLogRecord *decoded;
394 : :
395 : : /*
396 : : * Release last returned record, if there is one. We need to do this so
397 : : * that we can check for empty decode queue accurately.
398 : : */
1458 399 : 3086569 : XLogReleasePreviousRecord(state);
400 : :
401 : : /*
402 : : * Call XLogReadAhead() in blocking mode to make sure there is something
403 : : * in the queue, though we don't use the result.
404 : : */
405 [ + - ]: 3086569 : if (!XLogReaderHasQueuedRecordOrError(state))
406 : 3086569 : XLogReadAhead(state, false /* nonblocking */ );
407 : :
408 : : /* Consume the head record or error. */
409 : 3086362 : decoded = XLogNextRecord(state, errormsg);
410 [ + + ]: 3086362 : if (decoded)
411 : : {
412 : : /*
413 : : * This function returns a pointer to the record's header, not the
414 : : * actual decoded record. The caller will access the decoded record
415 : : * through the XLogRecGetXXX() macros, which reach the decoded
416 : : * recorded as xlogreader->record.
417 : : */
418 [ - + ]: 3079904 : Assert(state->record == decoded);
419 : 3079904 : return &decoded->header;
420 : : }
421 : :
422 : 6458 : return NULL;
423 : : }
424 : :
425 : : /*
426 : : * Allocate space for a decoded record. The only member of the returned
427 : : * object that is initialized is the 'oversized' flag, indicating that the
428 : : * decoded record wouldn't fit in the decode buffer and must eventually be
429 : : * freed explicitly.
430 : : *
431 : : * The caller is responsible for adjusting decode_buffer_tail with the real
432 : : * size after successfully decoding a record into this space. This way, if
433 : : * decoding fails, then there is nothing to undo unless the 'oversized' flag
434 : : * was set and pfree() must be called.
435 : : *
436 : : * Return NULL if there is no space in the decode buffer and allow_oversized
437 : : * is false, or if memory allocation fails for an oversized buffer.
438 : : */
439 : : static DecodedXLogRecord *
440 : 5949590 : XLogReadRecordAlloc(XLogReaderState *state, size_t xl_tot_len, bool allow_oversized)
441 : : {
442 : 5949590 : size_t required_space = DecodeXLogRecordRequiredSpace(xl_tot_len);
443 : 5949590 : DecodedXLogRecord *decoded = NULL;
444 : :
445 : : /* Allocate a circular decode buffer if we don't have one already. */
446 [ + + ]: 5949590 : if (unlikely(state->decode_buffer == NULL))
447 : : {
448 [ + + ]: 2665 : if (state->decode_buffer_size == 0)
449 : 1662 : state->decode_buffer_size = DEFAULT_DECODE_BUFFER_SIZE;
450 : 2665 : state->decode_buffer = palloc(state->decode_buffer_size);
451 : 2665 : state->decode_buffer_head = state->decode_buffer;
452 : 2665 : state->decode_buffer_tail = state->decode_buffer;
453 : 2665 : state->free_decode_buffer = true;
454 : : }
455 : :
456 : : /* Try to allocate space in the circular decode buffer. */
457 [ + + ]: 5949590 : if (state->decode_buffer_tail >= state->decode_buffer_head)
458 : : {
459 : : /* Empty, or tail is to the right of head. */
828 460 : 5009097 : if (required_space <=
461 : 5009097 : state->decode_buffer_size -
462 [ + + ]: 5009097 : (state->decode_buffer_tail - state->decode_buffer))
463 : : {
464 : : /*-
465 : : * There is space between tail and end.
466 : : *
467 : : * +-----+--------------------+-----+
468 : : * | |////////////////////|here!|
469 : : * +-----+--------------------+-----+
470 : : * ^ ^
471 : : * | |
472 : : * h t
473 : : */
1458 474 : 4989506 : decoded = (DecodedXLogRecord *) state->decode_buffer_tail;
475 : 4989506 : decoded->oversized = false;
476 : 4989506 : return decoded;
477 : : }
828 478 : 19591 : else if (required_space <
479 [ + + ]: 19591 : state->decode_buffer_head - state->decode_buffer)
480 : : {
481 : : /*-
482 : : * There is space between start and head.
483 : : *
484 : : * +-----+--------------------+-----+
485 : : * |here!|////////////////////| |
486 : : * +-----+--------------------+-----+
487 : : * ^ ^
488 : : * | |
489 : : * h t
490 : : */
1458 491 : 17042 : decoded = (DecodedXLogRecord *) state->decode_buffer;
492 : 17042 : decoded->oversized = false;
493 : 17042 : return decoded;
494 : : }
495 : : }
496 : : else
497 : : {
498 : : /* Tail is to the left of head. */
828 499 : 940493 : if (required_space <
500 [ + + ]: 940493 : state->decode_buffer_head - state->decode_buffer_tail)
501 : : {
502 : : /*-
503 : : * There is space between tail and head.
504 : : *
505 : : * +-----+--------------------+-----+
506 : : * |/////|here! |/////|
507 : : * +-----+--------------------+-----+
508 : : * ^ ^
509 : : * | |
510 : : * t h
511 : : */
1458 512 : 940343 : decoded = (DecodedXLogRecord *) state->decode_buffer_tail;
513 : 940343 : decoded->oversized = false;
514 : 940343 : return decoded;
515 : : }
516 : : }
517 : :
518 : : /* Not enough space in the decode buffer. Are we allowed to allocate? */
519 [ + + ]: 2699 : if (allow_oversized)
520 : : {
894 michael@paquier.xyz 521 : 57 : decoded = palloc(required_space);
1458 tmunro@postgresql.or 522 : 57 : decoded->oversized = true;
523 : 57 : return decoded;
524 : : }
525 : :
526 : 2642 : return NULL;
527 : : }
528 : :
529 : : static XLogPageReadResult
530 : 5965707 : XLogDecodeNextRecord(XLogReaderState *state, bool nonblocking)
531 : : {
532 : : XLogRecPtr RecPtr;
533 : : XLogRecord *record;
534 : : XLogRecPtr targetPagePtr;
535 : : bool randAccess;
536 : : uint32 len,
537 : : total_len;
538 : : uint32 targetRecOff;
539 : : uint32 pageHeaderSize;
540 : : bool assembled;
541 : : bool gotheader;
542 : : int readOff;
543 : : DecodedXLogRecord *decoded;
544 : : char *errormsg; /* not used */
545 : :
546 : : /*
547 : : * randAccess indicates whether to verify the previous-record pointer of
548 : : * the record we're reading. We only do this if we're reading
549 : : * sequentially, which is what we initially assume.
550 : : */
1770 551 : 5965707 : randAccess = false;
552 : :
553 : : /* reset error state */
554 : 5965707 : state->errormsg_buf[0] = '\0';
1458 555 : 5965707 : decoded = NULL;
556 : :
1628 alvherre@alvh.no-ip. 557 : 5965707 : state->abortedRecPtr = InvalidXLogRecPtr;
558 : 5965707 : state->missingContrecPtr = InvalidXLogRecPtr;
559 : :
1458 tmunro@postgresql.or 560 : 5965707 : RecPtr = state->NextRecPtr;
561 : :
129 alvherre@kurilemu.de 562 [ + + ]:GNC 5965707 : if (XLogRecPtrIsValid(state->DecodeRecPtr))
563 : : {
564 : : /* read the record after the one we just read */
565 : :
566 : : /*
567 : : * NextRecPtr is pointing to end+1 of the previous WAL record. If
568 : : * we're at a page boundary, no more records can fit on the current
569 : : * page. We must skip over the page header, but we can't do that until
570 : : * we've read in the page, since the header size is variable.
571 : : */
572 : : }
573 : : else
574 : : {
575 : : /*
576 : : * Caller supplied a position to start at.
577 : : *
578 : : * In this case, NextRecPtr should already be pointing either to a
579 : : * valid record starting position or alternatively to the beginning of
580 : : * a page. See the header comments for XLogBeginRead.
581 : : */
1305 rhaas@postgresql.org 582 [ + + - + ]:CBC 6595 : Assert(RecPtr % XLOG_BLCKSZ == 0 || XRecOffIsValid(RecPtr));
1770 tmunro@postgresql.or 583 : 6595 : randAccess = true;
584 : : }
585 : :
1628 alvherre@alvh.no-ip. 586 : 5965707 : restart:
1458 tmunro@postgresql.or 587 : 5965709 : state->nonblocking = nonblocking;
1770 588 : 5965709 : state->currRecPtr = RecPtr;
1628 alvherre@alvh.no-ip. 589 : 5965709 : assembled = false;
590 : :
1770 tmunro@postgresql.or 591 : 5965709 : targetPagePtr = RecPtr - (RecPtr % XLOG_BLCKSZ);
592 : 5965709 : targetRecOff = RecPtr % XLOG_BLCKSZ;
593 : :
594 : : /*
595 : : * Read the page containing the record into state->readBuf. Request enough
596 : : * byte to cover the whole record header, or at least the part of it that
597 : : * fits on the same page.
598 : : */
599 : 5965709 : readOff = ReadPageInternal(state, targetPagePtr,
600 : 5965709 : Min(targetRecOff + SizeOfXLogRecord, XLOG_BLCKSZ));
1458 601 [ + + ]: 5965447 : if (readOff == XLREAD_WOULDBLOCK)
602 : 9154 : return XLREAD_WOULDBLOCK;
603 [ + + ]: 5956293 : else if (readOff < 0)
1770 604 : 6504 : goto err;
605 : :
606 : : /*
607 : : * ReadPageInternal always returns at least the page header, so we can
608 : : * examine it now.
609 : : */
610 [ + + ]: 5949789 : pageHeaderSize = XLogPageHeaderSize((XLogPageHeader) state->readBuf);
611 [ + + ]: 5949789 : if (targetRecOff == 0)
612 : : {
613 : : /*
614 : : * At page start, so skip over page header.
615 : : */
616 : 6164 : RecPtr += pageHeaderSize;
617 : 6164 : targetRecOff = pageHeaderSize;
618 : : }
619 [ - + ]: 5943625 : else if (targetRecOff < pageHeaderSize)
620 : : {
251 alvherre@kurilemu.de 621 :UNC 0 : report_invalid_record(state, "invalid record offset at %X/%08X: expected at least %u, got %u",
1109 peter@eisentraut.org 622 :UBC 0 : LSN_FORMAT_ARGS(RecPtr),
623 : : pageHeaderSize, targetRecOff);
1770 tmunro@postgresql.or 624 : 0 : goto err;
625 : : }
626 : :
1770 tmunro@postgresql.or 627 [ + + - + ]:CBC 5949789 : if ((((XLogPageHeader) state->readBuf)->xlp_info & XLP_FIRST_IS_CONTRECORD) &&
628 : : targetRecOff == pageHeaderSize)
629 : : {
251 alvherre@kurilemu.de 630 :UNC 0 : report_invalid_record(state, "contrecord is requested by %X/%08X",
1770 tmunro@postgresql.or 631 :UBC 0 : LSN_FORMAT_ARGS(RecPtr));
632 : 0 : goto err;
633 : : }
634 : :
635 : : /* ReadPageInternal has verified the page header */
1770 tmunro@postgresql.or 636 [ - + ]:CBC 5949789 : Assert(pageHeaderSize <= readOff);
637 : :
638 : : /*
639 : : * Read the record length.
640 : : *
641 : : * NB: Even though we use an XLogRecord pointer here, the whole record
642 : : * header might not fit on this page. xl_tot_len is the first field of the
643 : : * struct, so it must be on this page (the records are MAXALIGNed), but we
644 : : * cannot access any other fields until we've verified that we got the
645 : : * whole header.
646 : : */
647 : 5949789 : record = (XLogRecord *) (state->readBuf + RecPtr % XLOG_BLCKSZ);
648 : 5949789 : total_len = record->xl_tot_len;
649 : :
650 : : /*
651 : : * If the whole record header is on this page, validate it immediately.
652 : : * Otherwise do just a basic sanity check on xl_tot_len, and validate the
653 : : * rest of the header after reading it from the next page. The xl_tot_len
654 : : * check is necessary here to ensure that we enter the "Need to reassemble
655 : : * record" code path below; otherwise we might fail to apply
656 : : * ValidXLogRecordHeader at all.
657 : : */
658 [ + + ]: 5949789 : if (targetRecOff <= XLOG_BLCKSZ - SizeOfXLogRecord)
659 : : {
1458 660 [ + + ]: 5938131 : if (!ValidXLogRecordHeader(state, RecPtr, state->DecodeRecPtr, record,
661 : : randAccess))
1770 662 : 255 : goto err;
663 : 5937876 : gotheader = true;
664 : : }
665 : : else
666 : : {
667 : : /* There may be no next page if it's too small. */
901 668 [ + + ]: 11658 : if (total_len < SizeOfXLogRecord)
669 : : {
670 : 1 : report_invalid_record(state,
671 : : "invalid record length at %X/%08X: expected at least %u, got %u",
672 : 1 : LSN_FORMAT_ARGS(RecPtr),
673 : : (uint32) SizeOfXLogRecord, total_len);
674 : 1 : goto err;
675 : : }
676 : : /* We'll validate the header once we have the next page. */
1770 677 : 11657 : gotheader = false;
678 : : }
679 : :
680 : : /*
681 : : * Try to find space to decode this record, if we can do so without
682 : : * calling palloc. If we can't, we'll try again below after we've
683 : : * validated that total_len isn't garbage bytes from a recycled WAL page.
684 : : */
1458 685 : 5949533 : decoded = XLogReadRecordAlloc(state,
686 : : total_len,
687 : : false /* allow_oversized */ );
904 688 [ + + + + ]: 5949533 : if (decoded == NULL && nonblocking)
689 : : {
690 : : /*
691 : : * There is no space in the circular decode buffer, and the caller is
692 : : * only reading ahead. The caller should consume existing records to
693 : : * make space.
694 : : */
695 : 2573 : return XLREAD_WOULDBLOCK;
696 : : }
697 : :
1770 698 : 5946960 : len = XLOG_BLCKSZ - RecPtr % XLOG_BLCKSZ;
699 [ + + ]: 5946960 : if (total_len > len)
700 : : {
701 : : /* Need to reassemble record */
702 : : char *contdata;
703 : : XLogPageHeader pageHeader;
704 : : char *buffer;
705 : : uint32 gotlen;
706 : :
1628 alvherre@alvh.no-ip. 707 : 1461331 : assembled = true;
708 : :
709 : : /*
710 : : * We always have space for a couple of pages, enough to validate a
711 : : * boundary-spanning record header.
712 : : */
904 tmunro@postgresql.or 713 [ - + ]: 1461331 : Assert(state->readRecordBufSize >= XLOG_BLCKSZ * 2);
714 [ - + ]: 1461331 : Assert(state->readRecordBufSize >= len);
715 : :
716 : : /* Copy the first fragment of the record from the first page. */
1770 717 : 1461331 : memcpy(state->readRecordBuf,
718 : 1461331 : state->readBuf + RecPtr % XLOG_BLCKSZ, len);
719 : 1461331 : buffer = state->readRecordBuf + len;
720 : 1461331 : gotlen = len;
721 : :
722 : : do
723 : : {
724 : : /* Calculate pointer to beginning of next page */
725 : 1491654 : targetPagePtr += XLOG_BLCKSZ;
726 : :
727 : : /*
728 : : * Read the page header before processing the record data, so we
729 : : * can handle the case where the previous record ended as being a
730 : : * partial one.
731 : : */
239 akorotkov@postgresql 732 : 1491654 : readOff = ReadPageInternal(state, targetPagePtr, SizeOfXLogShortPHD);
1458 tmunro@postgresql.or 733 [ + + ]: 1491651 : if (readOff == XLREAD_WOULDBLOCK)
734 : 2654 : return XLREAD_WOULDBLOCK;
735 [ + + ]: 1488997 : else if (readOff < 0)
1770 736 : 15 : goto err;
737 : :
738 [ - + ]: 1488982 : Assert(SizeOfXLogShortPHD <= readOff);
739 : :
740 : 1488982 : pageHeader = (XLogPageHeader) state->readBuf;
741 : :
742 : : /*
743 : : * If we were expecting a continuation record and got an
744 : : * "overwrite contrecord" flag, that means the continuation record
745 : : * was overwritten with a different record. Restart the read by
746 : : * assuming the address to read is the location where we found
747 : : * this flag; but keep track of the LSN of the record we were
748 : : * reading, for later verification.
749 : : */
1628 alvherre@alvh.no-ip. 750 [ + + ]: 1488982 : if (pageHeader->xlp_info & XLP_FIRST_IS_OVERWRITE_CONTRECORD)
751 : : {
1570 752 : 2 : state->overwrittenRecPtr = RecPtr;
1628 753 : 2 : RecPtr = targetPagePtr;
754 : 2 : goto restart;
755 : : }
756 : :
757 : : /* Check that the continuation on next page looks valid */
1770 tmunro@postgresql.or 758 [ + + ]: 1488980 : if (!(pageHeader->xlp_info & XLP_FIRST_IS_CONTRECORD))
759 : : {
760 : 1 : report_invalid_record(state,
761 : : "there is no contrecord flag at %X/%08X",
762 : 1 : LSN_FORMAT_ARGS(RecPtr));
763 : 1 : goto err;
764 : : }
765 : :
766 : : /*
767 : : * Cross-check that xlp_rem_len agrees with how much of the record
768 : : * we expect there to be left.
769 : : */
770 [ + - ]: 1488979 : if (pageHeader->xlp_rem_len == 0 ||
771 [ + + ]: 1488979 : total_len != (pageHeader->xlp_rem_len + gotlen))
772 : : {
773 : 2 : report_invalid_record(state,
774 : : "invalid contrecord length %u (expected %lld) at %X/%08X",
775 : : pageHeader->xlp_rem_len,
776 : 2 : ((long long) total_len) - gotlen,
777 : 2 : LSN_FORMAT_ARGS(RecPtr));
778 : 2 : goto err;
779 : : }
780 : :
781 : : /* Wait for the next page to become available */
239 akorotkov@postgresql 782 : 1488977 : readOff = ReadPageInternal(state, targetPagePtr,
783 : 1488977 : Min(total_len - gotlen + SizeOfXLogShortPHD,
784 : : XLOG_BLCKSZ));
785 [ - + ]: 1488977 : if (readOff == XLREAD_WOULDBLOCK)
239 akorotkov@postgresql 786 :UBC 0 : return XLREAD_WOULDBLOCK;
239 akorotkov@postgresql 787 [ - + ]:CBC 1488977 : else if (readOff < 0)
239 akorotkov@postgresql 788 :UBC 0 : goto err;
789 : :
790 : : /* Append the continuation from this page to the buffer */
1770 tmunro@postgresql.or 791 [ + + ]:CBC 1488977 : pageHeaderSize = XLogPageHeaderSize(pageHeader);
792 : :
793 [ - + ]: 1488977 : if (readOff < pageHeaderSize)
1770 tmunro@postgresql.or 794 :UBC 0 : readOff = ReadPageInternal(state, targetPagePtr,
795 : : pageHeaderSize);
796 : :
1770 tmunro@postgresql.or 797 [ - + ]:CBC 1488977 : Assert(pageHeaderSize <= readOff);
798 : :
799 : 1488977 : contdata = (char *) state->readBuf + pageHeaderSize;
800 : 1488977 : len = XLOG_BLCKSZ - pageHeaderSize;
801 [ + + ]: 1488977 : if (pageHeader->xlp_rem_len < len)
802 : 1458639 : len = pageHeader->xlp_rem_len;
803 : :
804 [ - + ]: 1488977 : if (readOff < pageHeaderSize + len)
1770 tmunro@postgresql.or 805 :UBC 0 : readOff = ReadPageInternal(state, targetPagePtr,
806 : 0 : pageHeaderSize + len);
807 : :
396 peter@eisentraut.org 808 :CBC 1488977 : memcpy(buffer, contdata, len);
1770 tmunro@postgresql.or 809 : 1488977 : buffer += len;
810 : 1488977 : gotlen += len;
811 : :
812 : : /* If we just reassembled the record header, validate it. */
813 [ + + ]: 1488977 : if (!gotheader)
814 : : {
815 : 11634 : record = (XLogRecord *) state->readRecordBuf;
1458 816 [ - + ]: 11634 : if (!ValidXLogRecordHeader(state, RecPtr, state->DecodeRecPtr,
817 : : record, randAccess))
4806 alvherre@alvh.no-ip. 818 :UBC 0 : goto err;
1770 tmunro@postgresql.or 819 :CBC 11634 : gotheader = true;
820 : : }
821 : :
822 : : /*
823 : : * We might need a bigger buffer. We have validated the record
824 : : * header, in the case that it split over a page boundary. We've
825 : : * also cross-checked total_len against xlp_rem_len on the second
826 : : * page, and verified xlp_pageaddr on both.
827 : : */
904 828 [ + + ]: 1488977 : if (total_len > state->readRecordBufSize)
829 : : {
830 : : char save_copy[XLOG_BLCKSZ * 2];
831 : :
832 : : /*
833 : : * Save and restore the data we already had. It can't be more
834 : : * than two pages.
835 : : */
836 [ - + ]: 37 : Assert(gotlen <= lengthof(save_copy));
837 [ - + ]: 37 : Assert(gotlen <= state->readRecordBufSize);
838 : 37 : memcpy(save_copy, state->readRecordBuf, gotlen);
894 michael@paquier.xyz 839 : 37 : allocate_recordbuf(state, total_len);
904 tmunro@postgresql.or 840 : 37 : memcpy(state->readRecordBuf, save_copy, gotlen);
841 : 37 : buffer = state->readRecordBuf + gotlen;
842 : : }
843 [ + + ]: 1488977 : } while (gotlen < total_len);
1770 844 [ - + ]: 1458654 : Assert(gotheader);
845 : :
846 : 1458654 : record = (XLogRecord *) state->readRecordBuf;
847 [ - + ]: 1458654 : if (!ValidXLogRecord(state, record, RecPtr))
1770 tmunro@postgresql.or 848 :UBC 0 : goto err;
849 : :
1770 tmunro@postgresql.or 850 [ + + ]:CBC 1458654 : pageHeaderSize = XLogPageHeaderSize((XLogPageHeader) state->readBuf);
1458 851 : 1458654 : state->DecodeRecPtr = RecPtr;
852 : 1458654 : state->NextRecPtr = targetPagePtr + pageHeaderSize
1770 853 : 1458654 : + MAXALIGN(pageHeader->xlp_rem_len);
854 : : }
855 : : else
856 : : {
857 : : /* Wait for the record data to become available */
858 : 4485629 : readOff = ReadPageInternal(state, targetPagePtr,
859 : 4485629 : Min(targetRecOff + total_len, XLOG_BLCKSZ));
1458 860 [ - + ]: 4485629 : if (readOff == XLREAD_WOULDBLOCK)
1458 tmunro@postgresql.or 861 :UBC 0 : return XLREAD_WOULDBLOCK;
1458 tmunro@postgresql.or 862 [ - + ]:CBC 4485629 : else if (readOff < 0)
1770 tmunro@postgresql.or 863 :UBC 0 : goto err;
864 : :
865 : : /* Record does not cross a page boundary */
1770 tmunro@postgresql.or 866 [ + + ]:CBC 4485629 : if (!ValidXLogRecord(state, record, RecPtr))
867 : 1 : goto err;
868 : :
1458 869 : 4485628 : state->NextRecPtr = RecPtr + MAXALIGN(total_len);
870 : :
871 : 4485628 : state->DecodeRecPtr = RecPtr;
872 : : }
873 : :
874 : : /*
875 : : * Special processing if it's an XLOG SWITCH record
876 : : */
1802 877 [ + + ]: 5944282 : if (record->xl_rmid == RM_XLOG_ID &&
878 [ + + ]: 138929 : (record->xl_info & ~XLR_INFO_MASK) == XLOG_SWITCH)
879 : : {
880 : : /* Pretend it extends to end of segment */
1458 881 : 559 : state->NextRecPtr += state->segcxt.ws_segsize - 1;
882 : 559 : state->NextRecPtr -= XLogSegmentOffset(state->NextRecPtr, state->segcxt.ws_segsize);
883 : : }
884 : :
885 : : /*
886 : : * If we got here without a DecodedXLogRecord, it means we needed to
887 : : * validate total_len before trusting it, but by now we've done that.
888 : : */
904 889 [ + + ]: 5944282 : if (decoded == NULL)
890 : : {
891 [ - + ]: 57 : Assert(!nonblocking);
892 : 57 : decoded = XLogReadRecordAlloc(state,
893 : : total_len,
894 : : true /* allow_oversized */ );
895 : : /* allocation should always happen under allow_oversized */
894 michael@paquier.xyz 896 [ - + ]: 57 : Assert(decoded != NULL);
897 : : }
898 : :
1458 tmunro@postgresql.or 899 [ + - ]: 5944282 : if (DecodeXLogRecord(state, decoded, record, RecPtr, &errormsg))
900 : : {
901 : : /* Record the location of the next record. */
902 : 5944282 : decoded->next_lsn = state->NextRecPtr;
903 : :
904 : : /*
905 : : * If it's in the decode buffer, mark the decode buffer space as
906 : : * occupied.
907 : : */
908 [ + + ]: 5944282 : if (!decoded->oversized)
909 : : {
910 : : /* The new decode buffer head must be MAXALIGNed. */
911 [ - + ]: 5944225 : Assert(decoded->size == MAXALIGN(decoded->size));
912 [ + + ]: 5944225 : if ((char *) decoded == state->decode_buffer)
913 : 3128003 : state->decode_buffer_tail = state->decode_buffer + decoded->size;
914 : : else
915 : 2816222 : state->decode_buffer_tail += decoded->size;
916 : : }
917 : :
918 : : /* Insert it into the queue of decoded records. */
919 [ - + ]: 5944282 : Assert(state->decode_queue_tail != decoded);
920 [ + + ]: 5944282 : if (state->decode_queue_tail)
921 : 2833250 : state->decode_queue_tail->next = decoded;
922 : 5944282 : state->decode_queue_tail = decoded;
923 [ + + ]: 5944282 : if (!state->decode_queue_head)
924 : 3111032 : state->decode_queue_head = decoded;
925 : 5944282 : return XLREAD_SUCCESS;
926 : : }
927 : :
4806 alvherre@alvh.no-ip. 928 :UBC 0 : err:
1628 alvherre@alvh.no-ip. 929 [ + + ]:CBC 6779 : if (assembled)
930 : : {
931 : : /*
932 : : * We get here when a record that spans multiple pages needs to be
933 : : * assembled, but something went wrong -- perhaps a contrecord piece
934 : : * was lost. If caller is WAL replay, it will know where the aborted
935 : : * record was and where to direct followup WAL to be written, marking
936 : : * the next piece with XLP_FIRST_IS_OVERWRITE_CONTRECORD, which will
937 : : * in turn signal downstream WAL consumers that the broken WAL record
938 : : * is to be ignored.
939 : : */
940 : 18 : state->abortedRecPtr = RecPtr;
941 : 18 : state->missingContrecPtr = targetPagePtr;
942 : :
943 : : /*
944 : : * If we got here without reporting an error, make sure an error is
945 : : * queued so that XLogPrefetcherReadRecord() doesn't bring us back a
946 : : * second time and clobber the above state.
947 : : */
986 tmunro@postgresql.or 948 : 18 : state->errormsg_deferred = true;
949 : : }
950 : :
1458 951 [ + + - + ]: 6779 : if (decoded && decoded->oversized)
1458 tmunro@postgresql.or 952 :UBC 0 : pfree(decoded);
953 : :
954 : : /*
955 : : * Invalidate the read state. We might read from a different source after
956 : : * failure.
957 : : */
3637 alvherre@alvh.no-ip. 958 :CBC 6779 : XLogReaderInvalReadState(state);
959 : :
960 : : /*
961 : : * If an error was written to errormsg_buf, it'll be returned to the
962 : : * caller of XLogReadRecord() after all successfully decoded records from
963 : : * the read queue.
964 : : */
965 : :
1458 tmunro@postgresql.or 966 : 6779 : return XLREAD_FAIL;
967 : : }
968 : :
969 : : /*
970 : : * Try to decode the next available record, and return it. The record will
971 : : * also be returned to XLogNextRecord(), which must be called to 'consume'
972 : : * each record.
973 : : *
974 : : * If nonblocking is true, may return NULL due to lack of data or WAL decoding
975 : : * space.
976 : : */
977 : : DecodedXLogRecord *
978 : 5965707 : XLogReadAhead(XLogReaderState *state, bool nonblocking)
979 : : {
980 : : XLogPageReadResult result;
981 : :
982 [ - + ]: 5965707 : if (state->errormsg_deferred)
1458 tmunro@postgresql.or 983 :UBC 0 : return NULL;
984 : :
1458 tmunro@postgresql.or 985 :CBC 5965707 : result = XLogDecodeNextRecord(state, nonblocking);
986 [ + + ]: 5965442 : if (result == XLREAD_SUCCESS)
987 : : {
988 [ - + ]: 5944282 : Assert(state->decode_queue_tail != NULL);
989 : 5944282 : return state->decode_queue_tail;
990 : : }
991 : :
1770 992 : 21160 : return NULL;
993 : : }
994 : :
995 : : /*
996 : : * Read a single xlog page including at least [pageptr, reqLen] of valid data
997 : : * via the page_read() callback.
998 : : *
999 : : * Returns XLREAD_FAIL if the required page cannot be read for some
1000 : : * reason; errormsg_buf is set in that case (unless the error occurs in the
1001 : : * page_read callback).
1002 : : *
1003 : : * Returns XLREAD_WOULDBLOCK if the requested data can't be read without
1004 : : * waiting. This can be returned only if the installed page_read callback
1005 : : * respects the state->nonblocking flag, and cannot read the requested data
1006 : : * immediately.
1007 : : *
1008 : : * We fetch the page from a reader-local cache if we know we have the required
1009 : : * data and if there hasn't been any error since caching the data.
1010 : : */
1011 : : static int
1012 : 13432143 : ReadPageInternal(XLogReaderState *state, XLogRecPtr pageptr, int reqLen)
1013 : : {
1014 : : int readLen;
1015 : : uint32 targetPageOff;
1016 : : XLogSegNo targetSegNo;
1017 : : XLogPageHeader hdr;
1018 : :
1019 [ - + ]: 13432143 : Assert((pageptr % XLOG_BLCKSZ) == 0);
1020 : :
1021 : 13432143 : XLByteToSeg(pageptr, targetSegNo, state->segcxt.ws_segsize);
1022 : 13432143 : targetPageOff = XLogSegmentOffset(pageptr, state->segcxt.ws_segsize);
1023 : :
1024 : : /* check whether we have all the requested data already */
1025 [ + + ]: 13432143 : if (targetSegNo == state->seg.ws_segno &&
1026 [ + + + + ]: 13421367 : targetPageOff == state->segoff && reqLen <= state->readLen)
1027 : 11901069 : return state->readLen;
1028 : :
1029 : : /*
1030 : : * Invalidate contents of internal buffer before read attempt. Just set
1031 : : * the length to 0, rather than a full XLogReaderInvalReadState(), so we
1032 : : * don't forget the segment we last successfully read.
1033 : : */
1289 1034 : 1531074 : state->readLen = 0;
1035 : :
1036 : : /*
1037 : : * Data is not in our buffer.
1038 : : *
1039 : : * Every time we actually read the segment, even if we looked at parts of
1040 : : * it before, we need to do verification as the page_read callback might
1041 : : * now be rereading data from a different source.
1042 : : *
1043 : : * Whenever switching to a new WAL segment, we read the first page of the
1044 : : * file and validate its header, even if that's not where the target
1045 : : * record is. This is so that we can check the additional identification
1046 : : * info that is present in the first page's "long" header.
1047 : : */
1770 1048 [ + + + + ]: 1531074 : if (targetSegNo != state->seg.ws_segno && targetPageOff != 0)
1049 : : {
1050 : 8741 : XLogRecPtr targetSegmentPtr = pageptr - targetPageOff;
1051 : :
1052 : 8741 : readLen = state->routine.page_read(state, targetSegmentPtr, XLOG_BLCKSZ,
1053 : : state->currRecPtr,
1054 : : state->readBuf);
1458 1055 [ - + ]: 8734 : if (readLen == XLREAD_WOULDBLOCK)
1458 tmunro@postgresql.or 1056 :UBC 0 : return XLREAD_WOULDBLOCK;
1458 tmunro@postgresql.or 1057 [ + + ]:CBC 8734 : else if (readLen < 0)
1770 tmunro@postgresql.or 1058 :GBC 4 : goto err;
1059 : :
1060 : : /* we can be sure to have enough WAL available, we scrolled back */
1770 tmunro@postgresql.or 1061 [ - + ]:CBC 8730 : Assert(readLen == XLOG_BLCKSZ);
1062 : :
1063 [ - + ]: 8730 : if (!XLogReaderValidatePageHeader(state, targetSegmentPtr,
1064 : : state->readBuf))
1770 tmunro@postgresql.or 1065 :UBC 0 : goto err;
1066 : : }
1067 : :
1068 : : /*
1069 : : * First, read the requested data length, but at least a short page header
1070 : : * so that we can validate it.
1071 : : */
1770 tmunro@postgresql.or 1072 :CBC 1531063 : readLen = state->routine.page_read(state, pageptr, Max(reqLen, SizeOfXLogShortPHD),
1073 : : state->currRecPtr,
1074 : : state->readBuf);
1458 1075 [ + + ]: 1530805 : if (readLen == XLREAD_WOULDBLOCK)
1076 : 11808 : return XLREAD_WOULDBLOCK;
1077 [ + + ]: 1518997 : else if (readLen < 0)
1770 1078 : 6504 : goto err;
1079 : :
1080 [ - + ]: 1512493 : Assert(readLen <= XLOG_BLCKSZ);
1081 : :
1082 : : /* Do we have enough data to check the header length? */
1083 [ - + ]: 1512493 : if (readLen <= SizeOfXLogShortPHD)
1770 tmunro@postgresql.or 1084 :UBC 0 : goto err;
1085 : :
1770 tmunro@postgresql.or 1086 [ - + ]:CBC 1512493 : Assert(readLen >= reqLen);
1087 : :
1088 : 1512493 : hdr = (XLogPageHeader) state->readBuf;
1089 : :
1090 : : /* still not enough */
1091 [ + + - + ]: 1512493 : if (readLen < XLogPageHeaderSize(hdr))
1092 : : {
1770 tmunro@postgresql.or 1093 [ # # ]:UBC 0 : readLen = state->routine.page_read(state, pageptr, XLogPageHeaderSize(hdr),
1094 : : state->currRecPtr,
1095 : : state->readBuf);
1458 1096 [ # # ]: 0 : if (readLen == XLREAD_WOULDBLOCK)
1097 : 0 : return XLREAD_WOULDBLOCK;
1098 [ # # ]: 0 : else if (readLen < 0)
1770 1099 : 0 : goto err;
1100 : : }
1101 : :
1102 : : /*
1103 : : * Now that we know we have the full header, validate it.
1104 : : */
1770 tmunro@postgresql.or 1105 [ + + ]:CBC 1512493 : if (!XLogReaderValidatePageHeader(state, pageptr, (char *) hdr))
1106 : 11 : goto err;
1107 : :
1108 : : /* update read state information */
1109 : 1512482 : state->seg.ws_segno = targetSegNo;
1110 : 1512482 : state->segoff = targetPageOff;
1111 : 1512482 : state->readLen = readLen;
1112 : :
1113 : 1512482 : return readLen;
1114 : :
1115 : 6519 : err:
1289 1116 : 6519 : XLogReaderInvalReadState(state);
1117 : :
1458 1118 : 6519 : return XLREAD_FAIL;
1119 : : }
1120 : :
1121 : : /*
1122 : : * Invalidate the xlogreader's read state to force a re-read.
1123 : : */
1124 : : static void
3637 alvherre@alvh.no-ip. 1125 : 13298 : XLogReaderInvalReadState(XLogReaderState *state)
1126 : : {
1770 tmunro@postgresql.or 1127 : 13298 : state->seg.ws_segno = 0;
1128 : 13298 : state->segoff = 0;
1129 : 13298 : state->readLen = 0;
4806 alvherre@alvh.no-ip. 1130 : 13298 : }
1131 : :
1132 : : /*
1133 : : * Validate an XLOG record header.
1134 : : *
1135 : : * This is just a convenience subroutine to avoid duplicated code in
1136 : : * XLogReadRecord. It's not intended for use from anywhere else.
1137 : : */
1138 : : static bool
1139 : 5949765 : ValidXLogRecordHeader(XLogReaderState *state, XLogRecPtr RecPtr,
1140 : : XLogRecPtr PrevRecPtr, XLogRecord *record,
1141 : : bool randAccess)
1142 : : {
4133 heikki.linnakangas@i 1143 [ + + ]: 5949765 : if (record->xl_tot_len < SizeOfXLogRecord)
1144 : : {
4806 alvherre@alvh.no-ip. 1145 : 246 : report_invalid_record(state,
1146 : : "invalid record length at %X/%08X: expected at least %u, got %u",
1846 peter@eisentraut.org 1147 : 246 : LSN_FORMAT_ARGS(RecPtr),
1148 : : (uint32) SizeOfXLogRecord, record->xl_tot_len);
4806 alvherre@alvh.no-ip. 1149 : 246 : return false;
1150 : : }
1438 jdavis@postgresql.or 1151 [ + + + + ]: 5949519 : if (!RmgrIdIsValid(record->xl_rmid))
1152 : : {
4806 alvherre@alvh.no-ip. 1153 : 1 : report_invalid_record(state,
1154 : : "invalid resource manager ID %u at %X/%08X",
1846 peter@eisentraut.org 1155 : 1 : record->xl_rmid, LSN_FORMAT_ARGS(RecPtr));
4806 alvherre@alvh.no-ip. 1156 : 1 : return false;
1157 : : }
1770 tmunro@postgresql.or 1158 [ + + ]: 5949518 : if (randAccess)
1159 : : {
1160 : : /*
1161 : : * We can't exactly verify the prev-link, but surely it should be less
1162 : : * than the record's own address.
1163 : : */
4806 alvherre@alvh.no-ip. 1164 [ - + ]: 6591 : if (!(record->xl_prev < RecPtr))
1165 : : {
4806 alvherre@alvh.no-ip. 1166 :UBC 0 : report_invalid_record(state,
1167 : : "record with incorrect prev-link %X/%08X at %X/%08X",
1846 peter@eisentraut.org 1168 : 0 : LSN_FORMAT_ARGS(record->xl_prev),
1169 : 0 : LSN_FORMAT_ARGS(RecPtr));
4806 alvherre@alvh.no-ip. 1170 : 0 : return false;
1171 : : }
1172 : : }
1173 : : else
1174 : : {
1175 : : /*
1176 : : * Record's prev-link should exactly match our previous location. This
1177 : : * check guards against torn WAL pages where a stale but valid-looking
1178 : : * WAL record starts on a sector boundary.
1179 : : */
4806 alvherre@alvh.no-ip. 1180 [ + + ]:CBC 5942927 : if (record->xl_prev != PrevRecPtr)
1181 : : {
1182 : 8 : report_invalid_record(state,
1183 : : "record with incorrect prev-link %X/%08X at %X/%08X",
1846 peter@eisentraut.org 1184 : 8 : LSN_FORMAT_ARGS(record->xl_prev),
1185 : 8 : LSN_FORMAT_ARGS(RecPtr));
4806 alvherre@alvh.no-ip. 1186 : 8 : return false;
1187 : : }
1188 : : }
1189 : :
1190 : 5949510 : return true;
1191 : : }
1192 : :
1193 : :
1194 : : /*
1195 : : * CRC-check an XLOG record. We do not believe the contents of an XLOG
1196 : : * record (other than to the minimal extent of computing the amount of
1197 : : * data to read in) until we've checked the CRCs.
1198 : : *
1199 : : * We assume all of the record (that is, xl_tot_len bytes) has been read
1200 : : * into memory at *record. Also, ValidXLogRecordHeader() has accepted the
1201 : : * record's header, which means in particular that xl_tot_len is at least
1202 : : * SizeOfXLogRecord.
1203 : : */
1204 : : static bool
1205 : 5944283 : ValidXLogRecord(XLogReaderState *state, XLogRecord *record, XLogRecPtr recptr)
1206 : : {
1207 : : pg_crc32c crc;
1208 : :
901 tmunro@postgresql.or 1209 [ - + ]: 5944283 : Assert(record->xl_tot_len >= SizeOfXLogRecord);
1210 : :
1211 : : /* Calculate the CRC */
4149 heikki.linnakangas@i 1212 : 5944283 : INIT_CRC32C(crc);
4133 1213 : 5944283 : COMP_CRC32C(crc, ((char *) record) + SizeOfXLogRecord, record->xl_tot_len - SizeOfXLogRecord);
1214 : : /* include the record header last */
4149 1215 : 5944283 : COMP_CRC32C(crc, (char *) record, offsetof(XLogRecord, xl_crc));
1216 : 5944283 : FIN_CRC32C(crc);
1217 : :
1218 [ + + ]: 5944283 : if (!EQ_CRC32C(record->xl_crc, crc))
1219 : : {
4806 alvherre@alvh.no-ip. 1220 : 1 : report_invalid_record(state,
1221 : : "incorrect resource manager data checksum in record at %X/%08X",
1846 peter@eisentraut.org 1222 : 1 : LSN_FORMAT_ARGS(recptr));
4806 alvherre@alvh.no-ip. 1223 : 1 : return false;
1224 : : }
1225 : :
1226 : 5944282 : return true;
1227 : : }
1228 : :
1229 : : /*
1230 : : * Validate a page header.
1231 : : *
1232 : : * Check if 'phdr' is valid as the header of the XLog page at position
1233 : : * 'recptr'.
1234 : : */
1235 : : bool
2871 heikki.linnakangas@i 1236 : 1522637 : XLogReaderValidatePageHeader(XLogReaderState *state, XLogRecPtr recptr,
1237 : : char *phdr)
1238 : : {
1239 : : XLogSegNo segno;
1240 : : int32 offset;
1241 : 1522637 : XLogPageHeader hdr = (XLogPageHeader) phdr;
1242 : :
4806 alvherre@alvh.no-ip. 1243 [ - + ]: 1522637 : Assert((recptr % XLOG_BLCKSZ) == 0);
1244 : :
2364 1245 : 1522637 : XLByteToSeg(recptr, segno, state->segcxt.ws_segsize);
1246 : 1522637 : offset = XLogSegmentOffset(recptr, state->segcxt.ws_segsize);
1247 : :
4806 1248 [ + + ]: 1522637 : if (hdr->xlp_magic != XLOG_PAGE_MAGIC)
1249 : : {
1250 : : char fname[MAXFNAMELEN];
1251 : :
2364 1252 : 8 : XLogFileName(fname, state->seg.ws_tli, segno, state->segcxt.ws_segsize);
1253 : :
4806 1254 : 8 : report_invalid_record(state,
1255 : : "invalid magic number %04X in WAL segment %s, LSN %X/%08X, offset %u",
1256 : 8 : hdr->xlp_magic,
1257 : : fname,
1196 michael@paquier.xyz 1258 : 8 : LSN_FORMAT_ARGS(recptr),
1259 : : offset);
4806 alvherre@alvh.no-ip. 1260 : 8 : return false;
1261 : : }
1262 : :
1263 [ + + ]: 1522629 : if ((hdr->xlp_info & ~XLP_ALL_FLAGS) != 0)
1264 : : {
1265 : : char fname[MAXFNAMELEN];
1266 : :
2364 1267 : 1 : XLogFileName(fname, state->seg.ws_tli, segno, state->segcxt.ws_segsize);
1268 : :
4806 1269 : 1 : report_invalid_record(state,
1270 : : "invalid info bits %04X in WAL segment %s, LSN %X/%08X, offset %u",
1271 : 1 : hdr->xlp_info,
1272 : : fname,
1196 michael@paquier.xyz 1273 : 1 : LSN_FORMAT_ARGS(recptr),
1274 : : offset);
4806 alvherre@alvh.no-ip. 1275 : 1 : return false;
1276 : : }
1277 : :
1278 [ + + ]: 1522628 : if (hdr->xlp_info & XLP_LONG_HEADER)
1279 : : {
1280 : 11868 : XLogLongPageHeader longhdr = (XLogLongPageHeader) hdr;
1281 : :
1282 [ + + ]: 11868 : if (state->system_identifier &&
1283 [ - + ]: 3886 : longhdr->xlp_sysid != state->system_identifier)
1284 : : {
4806 alvherre@alvh.no-ip. 1285 :UBC 0 : report_invalid_record(state,
1286 : : "WAL file is from different database system: WAL file database system identifier is %" PRIu64 ", pg_control database system identifier is %" PRIu64,
1287 : : longhdr->xlp_sysid,
1288 : : state->system_identifier);
1289 : 0 : return false;
1290 : : }
2364 alvherre@alvh.no-ip. 1291 [ - + ]:CBC 11868 : else if (longhdr->xlp_seg_size != state->segcxt.ws_segsize)
1292 : : {
4806 alvherre@alvh.no-ip. 1293 :UBC 0 : report_invalid_record(state,
1294 : : "WAL file is from different database system: incorrect segment size in page header");
1295 : 0 : return false;
1296 : : }
4806 alvherre@alvh.no-ip. 1297 [ - + ]:CBC 11868 : else if (longhdr->xlp_xlog_blcksz != XLOG_BLCKSZ)
1298 : : {
4806 alvherre@alvh.no-ip. 1299 :UBC 0 : report_invalid_record(state,
1300 : : "WAL file is from different database system: incorrect XLOG_BLCKSZ in page header");
1301 : 0 : return false;
1302 : : }
1303 : : }
4806 alvherre@alvh.no-ip. 1304 [ - + ]:CBC 1510760 : else if (offset == 0)
1305 : : {
1306 : : char fname[MAXFNAMELEN];
1307 : :
2364 alvherre@alvh.no-ip. 1308 :UBC 0 : XLogFileName(fname, state->seg.ws_tli, segno, state->segcxt.ws_segsize);
1309 : :
1310 : : /* hmm, first page of file doesn't have a long header? */
4806 1311 : 0 : report_invalid_record(state,
1312 : : "invalid info bits %04X in WAL segment %s, LSN %X/%08X, offset %u",
1313 : 0 : hdr->xlp_info,
1314 : : fname,
1196 michael@paquier.xyz 1315 : 0 : LSN_FORMAT_ARGS(recptr),
1316 : : offset);
4806 alvherre@alvh.no-ip. 1317 : 0 : return false;
1318 : : }
1319 : :
1320 : : /*
1321 : : * Check that the address on the page agrees with what we expected. This
1322 : : * check typically fails when an old WAL segment is recycled, and hasn't
1323 : : * yet been overwritten with new data yet.
1324 : : */
1250 michael@paquier.xyz 1325 [ + + ]:CBC 1522628 : if (hdr->xlp_pageaddr != recptr)
1326 : : {
1327 : : char fname[MAXFNAMELEN];
1328 : :
2364 alvherre@alvh.no-ip. 1329 : 5 : XLogFileName(fname, state->seg.ws_tli, segno, state->segcxt.ws_segsize);
1330 : :
4806 1331 : 5 : report_invalid_record(state,
1332 : : "unexpected pageaddr %X/%08X in WAL segment %s, LSN %X/%08X, offset %u",
1846 peter@eisentraut.org 1333 : 5 : LSN_FORMAT_ARGS(hdr->xlp_pageaddr),
1334 : : fname,
1196 michael@paquier.xyz 1335 : 5 : LSN_FORMAT_ARGS(recptr),
1336 : : offset);
4806 alvherre@alvh.no-ip. 1337 : 5 : return false;
1338 : : }
1339 : :
1340 : : /*
1341 : : * Since child timelines are always assigned a TLI greater than their
1342 : : * immediate parent's TLI, we should never see TLI go backwards across
1343 : : * successive pages of a consistent WAL sequence.
1344 : : *
1345 : : * Sometimes we re-read a segment that's already been (partially) read. So
1346 : : * we only verify TLIs for pages that are later than the last remembered
1347 : : * LSN.
1348 : : */
1349 [ + + ]: 1522623 : if (recptr > state->latestPagePtr)
1350 : : {
1351 [ - + ]: 1500323 : if (hdr->xlp_tli < state->latestPageTLI)
1352 : : {
1353 : : char fname[MAXFNAMELEN];
1354 : :
2364 alvherre@alvh.no-ip. 1355 :UBC 0 : XLogFileName(fname, state->seg.ws_tli, segno, state->segcxt.ws_segsize);
1356 : :
4806 1357 : 0 : report_invalid_record(state,
1358 : : "out-of-sequence timeline ID %u (after %u) in WAL segment %s, LSN %X/%08X, offset %u",
1359 : : hdr->xlp_tli,
1360 : : state->latestPageTLI,
1361 : : fname,
1196 michael@paquier.xyz 1362 : 0 : LSN_FORMAT_ARGS(recptr),
1363 : : offset);
4806 alvherre@alvh.no-ip. 1364 : 0 : return false;
1365 : : }
1366 : : }
4806 alvherre@alvh.no-ip. 1367 :CBC 1522623 : state->latestPagePtr = recptr;
1368 : 1522623 : state->latestPageTLI = hdr->xlp_tli;
1369 : :
1370 : 1522623 : return true;
1371 : : }
1372 : :
1373 : : /*
1374 : : * Forget about an error produced by XLogReaderValidatePageHeader().
1375 : : */
1376 : : void
1289 tmunro@postgresql.or 1377 : 3 : XLogReaderResetError(XLogReaderState *state)
1378 : : {
1379 : 3 : state->errormsg_buf[0] = '\0';
1380 : 3 : state->errormsg_deferred = false;
1381 : 3 : }
1382 : :
1383 : : /*
1384 : : * Find the first record with an lsn >= RecPtr.
1385 : : *
1386 : : * This is different from XLogBeginRead() in that RecPtr doesn't need to point
1387 : : * to a valid record boundary. Useful for checking whether RecPtr is a valid
1388 : : * xlog address for reading, and to find the first valid address after some
1389 : : * address when dumping records for debugging purposes.
1390 : : *
1391 : : * This positions the reader, like XLogBeginRead(), so that the next call to
1392 : : * XLogReadRecord() will read the next valid record.
1393 : : */
1394 : : XLogRecPtr
1770 1395 : 87 : XLogFindNextRecord(XLogReaderState *state, XLogRecPtr RecPtr)
1396 : : {
1397 : : XLogRecPtr tmpRecPtr;
1398 : 87 : XLogRecPtr found = InvalidXLogRecPtr;
1399 : : XLogPageHeader header;
1400 : : char *errormsg;
1401 : :
129 alvherre@kurilemu.de 1402 [ - + ]:GNC 87 : Assert(XLogRecPtrIsValid(RecPtr));
1403 : :
1404 : : /* Make sure ReadPageInternal() can't return XLREAD_WOULDBLOCK. */
1458 tmunro@postgresql.or 1405 :CBC 87 : state->nonblocking = false;
1406 : :
1407 : : /*
1408 : : * skip over potential continuation data, keeping in mind that it may span
1409 : : * multiple pages
1410 : : */
1770 1411 : 87 : tmpRecPtr = RecPtr;
1412 : : while (true)
3485 fujii@postgresql.org 1413 :UBC 0 : {
1414 : : XLogRecPtr targetPagePtr;
1415 : : int targetRecOff;
1416 : : uint32 pageHeaderSize;
1417 : : int readLen;
1418 : :
1419 : : /*
1420 : : * Compute targetRecOff. It should typically be equal or greater than
1421 : : * short page-header since a valid record can't start anywhere before
1422 : : * that, except when caller has explicitly specified the offset that
1423 : : * falls somewhere there or when we are skipping multi-page
1424 : : * continuation record. It doesn't matter though because
1425 : : * ReadPageInternal() is prepared to handle that and will read at
1426 : : * least short page-header worth of data
1427 : : */
1770 tmunro@postgresql.or 1428 :CBC 87 : targetRecOff = tmpRecPtr % XLOG_BLCKSZ;
1429 : :
1430 : : /* scroll back to page boundary */
1431 : 87 : targetPagePtr = tmpRecPtr - targetRecOff;
1432 : :
1433 : : /* Read the page containing the record */
1434 : 87 : readLen = ReadPageInternal(state, targetPagePtr, targetRecOff);
1435 [ - + ]: 87 : if (readLen < 0)
3485 fujii@postgresql.org 1436 :UBC 0 : goto err;
1437 : :
1770 tmunro@postgresql.or 1438 :CBC 87 : header = (XLogPageHeader) state->readBuf;
1439 : :
3485 fujii@postgresql.org 1440 [ + + ]: 87 : pageHeaderSize = XLogPageHeaderSize(header);
1441 : :
1442 : : /* make sure we have enough data for the page header */
1770 tmunro@postgresql.or 1443 : 87 : readLen = ReadPageInternal(state, targetPagePtr, pageHeaderSize);
1444 [ - + ]: 87 : if (readLen < 0)
1770 tmunro@postgresql.or 1445 :UBC 0 : goto err;
1446 : :
1447 : : /* skip over potential continuation data */
3485 fujii@postgresql.org 1448 [ + + ]:CBC 87 : if (header->xlp_info & XLP_FIRST_IS_CONTRECORD)
1449 : : {
1450 : : /*
1451 : : * If the length of the remaining continuation data is more than
1452 : : * what can fit in this page, the continuation record crosses over
1453 : : * this page. Read the next page and try again. xlp_rem_len in the
1454 : : * next page header will contain the remaining length of the
1455 : : * continuation data
1456 : : *
1457 : : * Note that record headers are MAXALIGN'ed
1458 : : */
2320 1459 [ - + ]: 30 : if (MAXALIGN(header->xlp_rem_len) >= (XLOG_BLCKSZ - pageHeaderSize))
1770 tmunro@postgresql.or 1460 :UBC 0 : tmpRecPtr = targetPagePtr + XLOG_BLCKSZ;
1461 : : else
1462 : : {
1463 : : /*
1464 : : * The previous continuation record ends in this page. Set
1465 : : * tmpRecPtr to point to the first valid record
1466 : : */
1770 tmunro@postgresql.or 1467 :CBC 30 : tmpRecPtr = targetPagePtr + pageHeaderSize
3485 fujii@postgresql.org 1468 : 30 : + MAXALIGN(header->xlp_rem_len);
1469 : 30 : break;
1470 : : }
1471 : : }
1472 : : else
1473 : : {
1770 tmunro@postgresql.or 1474 : 57 : tmpRecPtr = targetPagePtr + pageHeaderSize;
3485 fujii@postgresql.org 1475 : 57 : break;
1476 : : }
1477 : : }
1478 : :
1479 : : /*
1480 : : * we know now that tmpRecPtr is an address pointing to a valid XLogRecord
1481 : : * because either we're at the first record after the beginning of a page
1482 : : * or we just jumped over the remaining data of a continuation.
1483 : : */
1770 tmunro@postgresql.or 1484 : 87 : XLogBeginRead(state, tmpRecPtr);
1485 [ + - ]: 764 : while (XLogReadRecord(state, &errormsg) != NULL)
1486 : : {
1487 : : /* past the record we've found, break out */
1488 [ + + ]: 764 : if (RecPtr <= state->ReadRecPtr)
1489 : : {
1490 : : /* Rewind the reader to the beginning of the last record. */
1491 : 87 : found = state->ReadRecPtr;
1492 : 87 : XLogBeginRead(state, found);
1493 : 87 : return found;
1494 : : }
1495 : : }
1496 : :
4806 alvherre@alvh.no-ip. 1497 :UBC 0 : err:
1770 tmunro@postgresql.or 1498 : 0 : XLogReaderInvalReadState(state);
1499 : :
1500 : 0 : return InvalidXLogRecPtr;
1501 : : }
1502 : :
1503 : : /*
1504 : : * Helper function to ease writing of XLogReaderRoutine->page_read callbacks.
1505 : : * If this function is used, caller must supply a segment_open callback in
1506 : : * 'state', as that is used here.
1507 : : *
1508 : : * Read 'count' bytes into 'buf', starting at location 'startptr', from WAL
1509 : : * fetched from timeline 'tli'.
1510 : : *
1511 : : * Returns true if succeeded, false if an error occurs, in which case
1512 : : * 'errinfo' receives error details.
1513 : : */
1514 : : bool
2137 alvherre@alvh.no-ip. 1515 :CBC 158634 : WALRead(XLogReaderState *state,
1516 : : char *buf, XLogRecPtr startptr, Size count, TimeLineID tli,
1517 : : WALReadError *errinfo)
1518 : : {
1519 : : char *p;
1520 : : XLogRecPtr recptr;
1521 : : Size nbytes;
1522 : : #ifndef FRONTEND
1523 : : instr_time io_start;
1524 : : #endif
1525 : :
2302 1526 : 158634 : p = buf;
1527 : 158634 : recptr = startptr;
1528 : 158634 : nbytes = count;
1529 : :
1530 [ + + ]: 317656 : while (nbytes > 0)
1531 : : {
1532 : : uint32 startoff;
1533 : : int segbytes;
1534 : : int readbytes;
1535 : :
2132 1536 : 159023 : startoff = XLogSegmentOffset(recptr, state->segcxt.ws_segsize);
1537 : :
1538 : : /*
1539 : : * If the data we want is not in a segment we have open, close what we
1540 : : * have (if anything) and open the next one, using the caller's
1541 : : * provided segment_open callback.
1542 : : */
1543 [ + + ]: 159023 : if (state->seg.ws_file < 0 ||
1544 [ + + ]: 157137 : !XLByteInSeg(recptr, state->seg.ws_segno, state->segcxt.ws_segsize) ||
1545 [ - + ]: 150137 : tli != state->seg.ws_tli)
1546 : : {
1547 : : XLogSegNo nextSegNo;
1548 : :
1549 [ + + ]: 8886 : if (state->seg.ws_file >= 0)
1770 tmunro@postgresql.or 1550 : 7000 : state->routine.segment_close(state);
1551 : :
2132 alvherre@alvh.no-ip. 1552 : 8886 : XLByteToSeg(recptr, nextSegNo, state->segcxt.ws_segsize);
1770 tmunro@postgresql.or 1553 : 8886 : state->routine.segment_open(state, nextSegNo, &tli);
1554 : :
1555 : : /* This shouldn't happen -- indicates a bug in segment_open */
2132 alvherre@alvh.no-ip. 1556 [ - + ]: 8885 : Assert(state->seg.ws_file >= 0);
1557 : :
1558 : : /* Update the current segment info. */
1559 : 8885 : state->seg.ws_tli = tli;
1560 : 8885 : state->seg.ws_segno = nextSegNo;
1561 : : }
1562 : :
1563 : : /* How many bytes are within this segment? */
1564 [ + + ]: 159022 : if (nbytes > (state->segcxt.ws_segsize - startoff))
1565 : 389 : segbytes = state->segcxt.ws_segsize - startoff;
1566 : : else
2302 1567 : 158633 : segbytes = nbytes;
1568 : :
1569 : : #ifndef FRONTEND
1570 : : /* Measure I/O timing when reading segment */
382 michael@paquier.xyz 1571 : 140224 : io_start = pgstat_prepare_io_time(track_wal_io_timing);
1572 : :
2302 alvherre@alvh.no-ip. 1573 : 140224 : pgstat_report_wait_start(WAIT_EVENT_WAL_READ);
1574 : : #endif
1575 : :
1576 : : /* Reset errno first; eases reporting non-errno-affecting errors */
1577 : 159022 : errno = 0;
122 michael@paquier.xyz 1578 :GNC 159022 : readbytes = pg_pread(state->seg.ws_file, p, segbytes, (pgoff_t) startoff);
1579 : :
1580 : : #ifndef FRONTEND
2302 alvherre@alvh.no-ip. 1581 :CBC 140224 : pgstat_report_wait_end();
1582 : :
404 michael@paquier.xyz 1583 : 140224 : pgstat_count_io_op_time(IOOBJECT_WAL, IOCONTEXT_NORMAL, IOOP_READ,
1584 : : io_start, 1, readbytes);
1585 : : #endif
1586 : :
2302 alvherre@alvh.no-ip. 1587 [ - + ]: 159022 : if (readbytes <= 0)
1588 : : {
2302 alvherre@alvh.no-ip. 1589 :UBC 0 : errinfo->wre_errno = errno;
1590 : 0 : errinfo->wre_req = segbytes;
1591 : 0 : errinfo->wre_read = readbytes;
1592 : 0 : errinfo->wre_off = startoff;
2132 1593 : 0 : errinfo->wre_seg = state->seg;
2302 1594 : 0 : return false;
1595 : : }
1596 : :
1597 : : /* Update state for read */
2302 alvherre@alvh.no-ip. 1598 :CBC 159022 : recptr += readbytes;
1599 : 159022 : nbytes -= readbytes;
1600 : 159022 : p += readbytes;
1601 : : }
1602 : :
1603 : 158633 : return true;
1604 : : }
1605 : :
1606 : : /* ----------------------------------------
1607 : : * Functions for decoding the data and block references in a record.
1608 : : * ----------------------------------------
1609 : : */
1610 : :
1611 : : /*
1612 : : * Private function to reset the state, forgetting all decoded records, if we
1613 : : * are asked to move to a new read position.
1614 : : */
1615 : : static void
4133 heikki.linnakangas@i 1616 : 6628 : ResetDecoder(XLogReaderState *state)
1617 : : {
1618 : : DecodedXLogRecord *r;
1619 : :
1620 : : /* Reset the decoded record queue, freeing any oversized records. */
1458 tmunro@postgresql.or 1621 [ + + ]: 17098 : while ((r = state->decode_queue_head) != NULL)
1622 : : {
1623 : 3842 : state->decode_queue_head = r->next;
1624 [ + - ]: 3842 : if (r->oversized)
1458 tmunro@postgresql.or 1625 :UBC 0 : pfree(r);
1626 : : }
1458 tmunro@postgresql.or 1627 :CBC 6628 : state->decode_queue_tail = NULL;
1628 : 6628 : state->decode_queue_head = NULL;
1629 : 6628 : state->record = NULL;
1630 : :
1631 : : /* Reset the decode buffer to empty. */
1632 : 6628 : state->decode_buffer_tail = state->decode_buffer;
1633 : 6628 : state->decode_buffer_head = state->decode_buffer;
1634 : :
1635 : : /* Clear error state. */
1636 : 6628 : state->errormsg_buf[0] = '\0';
1637 : 6628 : state->errormsg_deferred = false;
4133 heikki.linnakangas@i 1638 : 6628 : }
1639 : :
1640 : : /*
1641 : : * Compute the maximum possible amount of padding that could be required to
1642 : : * decode a record, given xl_tot_len from the record's header. This is the
1643 : : * amount of output buffer space that we need to decode a record, though we
1644 : : * might not finish up using it all.
1645 : : *
1646 : : * This computation is pessimistic and assumes the maximum possible number of
1647 : : * blocks, due to lack of better information.
1648 : : */
1649 : : size_t
1458 tmunro@postgresql.or 1650 : 11915302 : DecodeXLogRecordRequiredSpace(size_t xl_tot_len)
1651 : : {
1652 : 11915302 : size_t size = 0;
1653 : :
1654 : : /* Account for the fixed size part of the decoded record struct. */
1655 : 11915302 : size += offsetof(DecodedXLogRecord, blocks[0]);
1656 : : /* Account for the flexible blocks array of maximum possible size. */
1657 : 11915302 : size += sizeof(DecodedBkpBlock) * (XLR_MAX_BLOCK_ID + 1);
1658 : : /* Account for all the raw main and block data. */
1659 : 11915302 : size += xl_tot_len;
1660 : : /* We might insert padding before main_data. */
1661 : 11915302 : size += (MAXIMUM_ALIGNOF - 1);
1662 : : /* We might insert padding before each block's data. */
1663 : 11915302 : size += (MAXIMUM_ALIGNOF - 1) * (XLR_MAX_BLOCK_ID + 1);
1664 : : /* We might insert padding at the end. */
1665 : 11915302 : size += (MAXIMUM_ALIGNOF - 1);
1666 : :
1667 : 11915302 : return size;
1668 : : }
1669 : :
1670 : : /*
1671 : : * Decode a record. "decoded" must point to a MAXALIGNed memory area that has
1672 : : * space for at least DecodeXLogRecordRequiredSpace(record) bytes. On
1673 : : * success, decoded->size contains the actual space occupied by the decoded
1674 : : * record, which may turn out to be less.
1675 : : *
1676 : : * Only decoded->oversized member must be initialized already, and will not be
1677 : : * modified. Other members will be initialized as required.
1678 : : *
1679 : : * On error, a human-readable error message is returned in *errormsg, and
1680 : : * the return value is false.
1681 : : */
1682 : : bool
1683 : 5944282 : DecodeXLogRecord(XLogReaderState *state,
1684 : : DecodedXLogRecord *decoded,
1685 : : XLogRecord *record,
1686 : : XLogRecPtr lsn,
1687 : : char **errormsg)
1688 : : {
1689 : : /*
1690 : : * read next _size bytes from record buffer, but check for overrun first.
1691 : : */
1692 : : #define COPY_HEADER_FIELD(_dst, _size) \
1693 : : do { \
1694 : : if (remaining < _size) \
1695 : : goto shortdata_err; \
1696 : : memcpy(_dst, ptr, _size); \
1697 : : ptr += _size; \
1698 : : remaining -= _size; \
1699 : : } while(0)
1700 : :
1701 : : char *ptr;
1702 : : char *out;
1703 : : uint32 remaining;
1704 : : uint32 datatotal;
1348 rhaas@postgresql.org 1705 : 5944282 : RelFileLocator *rlocator = NULL;
1706 : : uint8 block_id;
1707 : :
1458 tmunro@postgresql.or 1708 : 5944282 : decoded->header = *record;
1709 : 5944282 : decoded->lsn = lsn;
1710 : 5944282 : decoded->next = NULL;
46 msawada@postgresql.o 1711 :GNC 5944282 : decoded->record_origin = InvalidReplOriginId;
1458 tmunro@postgresql.or 1712 :CBC 5944282 : decoded->toplevel_xid = InvalidTransactionId;
1713 : 5944282 : decoded->main_data = NULL;
1714 : 5944282 : decoded->main_data_len = 0;
1715 : 5944282 : decoded->max_block_id = -1;
4133 heikki.linnakangas@i 1716 : 5944282 : ptr = (char *) record;
1717 : 5944282 : ptr += SizeOfXLogRecord;
1718 : 5944282 : remaining = record->xl_tot_len - SizeOfXLogRecord;
1719 : :
1720 : : /* Decode the headers */
1721 : 5944282 : datatotal = 0;
1722 [ + + ]: 12150729 : while (remaining > datatotal)
1723 : : {
1724 [ - + ]: 12078111 : COPY_HEADER_FIELD(&block_id, sizeof(uint8));
1725 : :
1726 [ + + ]: 12078111 : if (block_id == XLR_BLOCK_ID_DATA_SHORT)
1727 : : {
1728 : : /* XLogRecordDataHeaderShort */
1729 : : uint8 main_data_len;
1730 : :
1731 [ - + ]: 5859828 : COPY_HEADER_FIELD(&main_data_len, sizeof(uint8));
1732 : :
1458 tmunro@postgresql.or 1733 : 5859828 : decoded->main_data_len = main_data_len;
4133 heikki.linnakangas@i 1734 : 5859828 : datatotal += main_data_len;
1735 : 5859828 : break; /* by convention, the main data fragment is
1736 : : * always last */
1737 : : }
1738 [ + + ]: 6218283 : else if (block_id == XLR_BLOCK_ID_DATA_LONG)
1739 : : {
1740 : : /* XLogRecordDataHeaderLong */
1741 : : uint32 main_data_len;
1742 : :
1743 [ - + ]: 11836 : COPY_HEADER_FIELD(&main_data_len, sizeof(uint32));
1458 tmunro@postgresql.or 1744 : 11836 : decoded->main_data_len = main_data_len;
4133 heikki.linnakangas@i 1745 : 11836 : datatotal += main_data_len;
1746 : 11836 : break; /* by convention, the main data fragment is
1747 : : * always last */
1748 : : }
3973 andres@anarazel.de 1749 [ + + ]: 6206447 : else if (block_id == XLR_BLOCK_ID_ORIGIN)
1750 : : {
46 msawada@postgresql.o 1751 [ - + ]:GNC 13024 : COPY_HEADER_FIELD(&decoded->record_origin, sizeof(ReplOriginId));
1752 : : }
2064 akapila@postgresql.o 1753 [ + + ]:CBC 6193423 : else if (block_id == XLR_BLOCK_ID_TOPLEVEL_XID)
1754 : : {
1458 tmunro@postgresql.or 1755 [ - + ]: 683 : COPY_HEADER_FIELD(&decoded->toplevel_xid, sizeof(TransactionId));
1756 : : }
4133 heikki.linnakangas@i 1757 [ + - ]: 6192740 : else if (block_id <= XLR_MAX_BLOCK_ID)
1758 : : {
1759 : : /* XLogRecordBlockHeader */
1760 : : DecodedBkpBlock *blk;
1761 : : uint8 fork_flags;
1762 : :
1763 : : /* mark any intervening block IDs as not in use */
1458 tmunro@postgresql.or 1764 [ + + ]: 6195875 : for (int i = decoded->max_block_id + 1; i < block_id; ++i)
1765 : 3135 : decoded->blocks[i].in_use = false;
1766 : :
1767 [ - + ]: 6192740 : if (block_id <= decoded->max_block_id)
1768 : : {
4133 heikki.linnakangas@i 1769 :UBC 0 : report_invalid_record(state,
1770 : : "out-of-order block_id %u at %X/%08X",
1771 : : block_id,
1846 peter@eisentraut.org 1772 : 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
4133 heikki.linnakangas@i 1773 : 0 : goto err;
1774 : : }
1458 tmunro@postgresql.or 1775 :CBC 6192740 : decoded->max_block_id = block_id;
1776 : :
1777 : 6192740 : blk = &decoded->blocks[block_id];
4133 heikki.linnakangas@i 1778 : 6192740 : blk->in_use = true;
3322 rhaas@postgresql.org 1779 : 6192740 : blk->apply_image = false;
1780 : :
4133 heikki.linnakangas@i 1781 [ - + ]: 6192740 : COPY_HEADER_FIELD(&fork_flags, sizeof(uint8));
1782 : 6192740 : blk->forknum = fork_flags & BKPBLOCK_FORK_MASK;
1783 : 6192740 : blk->flags = fork_flags;
1784 : 6192740 : blk->has_image = ((fork_flags & BKPBLOCK_HAS_IMAGE) != 0);
1785 : 6192740 : blk->has_data = ((fork_flags & BKPBLOCK_HAS_DATA) != 0);
1786 : :
1438 tmunro@postgresql.or 1787 : 6192740 : blk->prefetch_buffer = InvalidBuffer;
1788 : :
4133 heikki.linnakangas@i 1789 [ - + ]: 6192740 : COPY_HEADER_FIELD(&blk->data_len, sizeof(uint16));
1790 : : /* cross-check that the HAS_DATA flag is set iff data_length > 0 */
1791 [ + + - + ]: 6192740 : if (blk->has_data && blk->data_len == 0)
1792 : : {
4133 heikki.linnakangas@i 1793 :UBC 0 : report_invalid_record(state,
1794 : : "BKPBLOCK_HAS_DATA set, but no data included at %X/%08X",
1846 peter@eisentraut.org 1795 : 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
4024 fujii@postgresql.org 1796 : 0 : goto err;
1797 : : }
4133 heikki.linnakangas@i 1798 [ + + - + ]:CBC 6192740 : if (!blk->has_data && blk->data_len != 0)
1799 : : {
4133 heikki.linnakangas@i 1800 :UBC 0 : report_invalid_record(state,
1801 : : "BKPBLOCK_HAS_DATA not set, but data length is %d at %X/%08X",
96 peter@eisentraut.org 1802 :UNC 0 : blk->data_len,
1846 peter@eisentraut.org 1803 :UBC 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
4024 fujii@postgresql.org 1804 : 0 : goto err;
1805 : : }
4133 heikki.linnakangas@i 1806 :CBC 6192740 : datatotal += blk->data_len;
1807 : :
1808 [ + + ]: 6192740 : if (blk->has_image)
1809 : : {
4022 fujii@postgresql.org 1810 [ - + ]: 2470791 : COPY_HEADER_FIELD(&blk->bimg_len, sizeof(uint16));
4133 heikki.linnakangas@i 1811 [ - + ]: 2470791 : COPY_HEADER_FIELD(&blk->hole_offset, sizeof(uint16));
4022 fujii@postgresql.org 1812 [ - + ]: 2470791 : COPY_HEADER_FIELD(&blk->bimg_info, sizeof(uint8));
1813 : :
3322 rhaas@postgresql.org 1814 : 2470791 : blk->apply_image = ((blk->bimg_info & BKPIMAGE_APPLY) != 0);
1815 : :
1720 michael@paquier.xyz 1816 [ - + ]: 2470791 : if (BKPIMAGE_COMPRESSED(blk->bimg_info))
1817 : : {
4022 fujii@postgresql.org 1818 [ # # ]:UBC 0 : if (blk->bimg_info & BKPIMAGE_HAS_HOLE)
1819 [ # # ]: 0 : COPY_HEADER_FIELD(&blk->hole_length, sizeof(uint16));
1820 : : else
1821 : 0 : blk->hole_length = 0;
1822 : : }
1823 : : else
4022 fujii@postgresql.org 1824 :CBC 2470791 : blk->hole_length = BLCKSZ - blk->bimg_len;
1825 : 2470791 : datatotal += blk->bimg_len;
1826 : :
1827 : : /*
1828 : : * cross-check that hole_offset > 0, hole_length > 0 and
1829 : : * bimg_len < BLCKSZ if the HAS_HOLE flag is set.
1830 : : */
1831 [ + + ]: 2470791 : if ((blk->bimg_info & BKPIMAGE_HAS_HOLE) &&
1832 [ + - ]: 2449625 : (blk->hole_offset == 0 ||
1833 [ + - ]: 2449625 : blk->hole_length == 0 ||
1834 [ - + ]: 2449625 : blk->bimg_len == BLCKSZ))
1835 : : {
4022 fujii@postgresql.org 1836 :UBC 0 : report_invalid_record(state,
1837 : : "BKPIMAGE_HAS_HOLE set, but hole offset %d length %d block image length %d at %X/%08X",
96 peter@eisentraut.org 1838 :UNC 0 : blk->hole_offset,
1839 : 0 : blk->hole_length,
1840 : 0 : blk->bimg_len,
1846 peter@eisentraut.org 1841 :UBC 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
4022 fujii@postgresql.org 1842 : 0 : goto err;
1843 : : }
1844 : :
1845 : : /*
1846 : : * cross-check that hole_offset == 0 and hole_length == 0 if
1847 : : * the HAS_HOLE flag is not set.
1848 : : */
4022 fujii@postgresql.org 1849 [ + + ]:CBC 2470791 : if (!(blk->bimg_info & BKPIMAGE_HAS_HOLE) &&
1850 [ + - - + ]: 21166 : (blk->hole_offset != 0 || blk->hole_length != 0))
1851 : : {
4022 fujii@postgresql.org 1852 :UBC 0 : report_invalid_record(state,
1853 : : "BKPIMAGE_HAS_HOLE not set, but hole offset %d length %d at %X/%08X",
96 peter@eisentraut.org 1854 :UNC 0 : blk->hole_offset,
1855 : 0 : blk->hole_length,
1846 peter@eisentraut.org 1856 :UBC 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
4022 fujii@postgresql.org 1857 : 0 : goto err;
1858 : : }
1859 : :
1860 : : /*
1861 : : * Cross-check that bimg_len < BLCKSZ if it is compressed.
1862 : : */
1720 michael@paquier.xyz 1863 [ - + ]:CBC 2470791 : if (BKPIMAGE_COMPRESSED(blk->bimg_info) &&
4022 fujii@postgresql.org 1864 [ # # ]:UBC 0 : blk->bimg_len == BLCKSZ)
1865 : : {
1866 : 0 : report_invalid_record(state,
1867 : : "BKPIMAGE_COMPRESSED set, but block image length %d at %X/%08X",
96 peter@eisentraut.org 1868 :UNC 0 : blk->bimg_len,
1846 peter@eisentraut.org 1869 :UBC 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
4022 fujii@postgresql.org 1870 : 0 : goto err;
1871 : : }
1872 : :
1873 : : /*
1874 : : * cross-check that bimg_len = BLCKSZ if neither HAS_HOLE is
1875 : : * set nor COMPRESSED().
1876 : : */
4022 fujii@postgresql.org 1877 [ + + ]:CBC 2470791 : if (!(blk->bimg_info & BKPIMAGE_HAS_HOLE) &&
1720 michael@paquier.xyz 1878 [ + - ]: 21166 : !BKPIMAGE_COMPRESSED(blk->bimg_info) &&
4022 fujii@postgresql.org 1879 [ - + ]: 21166 : blk->bimg_len != BLCKSZ)
1880 : : {
4022 fujii@postgresql.org 1881 :UBC 0 : report_invalid_record(state,
1882 : : "neither BKPIMAGE_HAS_HOLE nor BKPIMAGE_COMPRESSED set, but block image length is %d at %X/%08X",
96 peter@eisentraut.org 1883 :UNC 0 : blk->data_len,
1846 peter@eisentraut.org 1884 :UBC 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
4022 fujii@postgresql.org 1885 : 0 : goto err;
1886 : : }
1887 : : }
4133 heikki.linnakangas@i 1888 [ + + ]:CBC 6192740 : if (!(fork_flags & BKPBLOCK_SAME_REL))
1889 : : {
1348 rhaas@postgresql.org 1890 [ - + ]: 5739919 : COPY_HEADER_FIELD(&blk->rlocator, sizeof(RelFileLocator));
1891 : 5739919 : rlocator = &blk->rlocator;
1892 : : }
1893 : : else
1894 : : {
1895 [ - + ]: 452821 : if (rlocator == NULL)
1896 : : {
4133 heikki.linnakangas@i 1897 :UBC 0 : report_invalid_record(state,
1898 : : "BKPBLOCK_SAME_REL set but no previous rel at %X/%08X",
1846 peter@eisentraut.org 1899 : 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
4133 heikki.linnakangas@i 1900 : 0 : goto err;
1901 : : }
1902 : :
1348 rhaas@postgresql.org 1903 :CBC 452821 : blk->rlocator = *rlocator;
1904 : : }
4133 heikki.linnakangas@i 1905 [ - + ]: 6192740 : COPY_HEADER_FIELD(&blk->blkno, sizeof(BlockNumber));
1906 : : }
1907 : : else
1908 : : {
4133 heikki.linnakangas@i 1909 :UBC 0 : report_invalid_record(state,
1910 : : "invalid block_id %u at %X/%08X",
1846 peter@eisentraut.org 1911 : 0 : block_id, LSN_FORMAT_ARGS(state->ReadRecPtr));
4133 heikki.linnakangas@i 1912 : 0 : goto err;
1913 : : }
1914 : : }
1915 : :
4133 heikki.linnakangas@i 1916 [ - + ]:CBC 5944282 : if (remaining != datatotal)
4133 heikki.linnakangas@i 1917 :UBC 0 : goto shortdata_err;
1918 : :
1919 : : /*
1920 : : * Ok, we've parsed the fragment headers, and verified that the total
1921 : : * length of the payload in the fragments is equal to the amount of data
1922 : : * left. Copy the data of each fragment to contiguous space after the
1923 : : * blocks array, inserting alignment padding before the data fragments so
1924 : : * they can be cast to struct pointers by REDO routines.
1925 : : */
1458 tmunro@postgresql.or 1926 :CBC 5944282 : out = ((char *) decoded) +
1927 : 5944282 : offsetof(DecodedXLogRecord, blocks) +
1928 : 5944282 : sizeof(decoded->blocks[0]) * (decoded->max_block_id + 1);
1929 : :
1930 : : /* block data first */
1931 [ + + ]: 12140157 : for (block_id = 0; block_id <= decoded->max_block_id; block_id++)
1932 : : {
1933 : 6195875 : DecodedBkpBlock *blk = &decoded->blocks[block_id];
1934 : :
4133 heikki.linnakangas@i 1935 [ + + ]: 6195875 : if (!blk->in_use)
1936 : 3135 : continue;
1937 : :
3322 rhaas@postgresql.org 1938 [ + + - + ]: 6192740 : Assert(blk->has_image || !blk->apply_image);
1939 : :
4133 heikki.linnakangas@i 1940 [ + + ]: 6192740 : if (blk->has_image)
1941 : : {
1942 : : /* no need to align image */
1458 tmunro@postgresql.or 1943 : 2470791 : blk->bkp_image = out;
1944 : 2470791 : memcpy(out, ptr, blk->bimg_len);
4022 fujii@postgresql.org 1945 : 2470791 : ptr += blk->bimg_len;
1458 tmunro@postgresql.or 1946 : 2470791 : out += blk->bimg_len;
1947 : : }
4133 heikki.linnakangas@i 1948 [ + + ]: 6192740 : if (blk->has_data)
1949 : : {
1458 tmunro@postgresql.or 1950 : 4696364 : out = (char *) MAXALIGN(out);
1951 : 4696364 : blk->data = out;
4133 heikki.linnakangas@i 1952 : 4696364 : memcpy(blk->data, ptr, blk->data_len);
1953 : 4696364 : ptr += blk->data_len;
1458 tmunro@postgresql.or 1954 : 4696364 : out += blk->data_len;
1955 : : }
1956 : : }
1957 : :
1958 : : /* and finally, the main data */
1959 [ + + ]: 5944282 : if (decoded->main_data_len > 0)
1960 : : {
1961 : 5871664 : out = (char *) MAXALIGN(out);
1962 : 5871664 : decoded->main_data = out;
1963 : 5871664 : memcpy(decoded->main_data, ptr, decoded->main_data_len);
1964 : 5871664 : ptr += decoded->main_data_len;
1965 : 5871664 : out += decoded->main_data_len;
1966 : : }
1967 : :
1968 : : /* Report the actual size we used. */
1969 : 5944282 : decoded->size = MAXALIGN(out - (char *) decoded);
1970 [ - + ]: 5944282 : Assert(DecodeXLogRecordRequiredSpace(record->xl_tot_len) >=
1971 : : decoded->size);
1972 : :
4133 heikki.linnakangas@i 1973 : 5944282 : return true;
1974 : :
4133 heikki.linnakangas@i 1975 :UBC 0 : shortdata_err:
1976 : 0 : report_invalid_record(state,
1977 : : "record with invalid length at %X/%08X",
1846 peter@eisentraut.org 1978 : 0 : LSN_FORMAT_ARGS(state->ReadRecPtr));
4133 heikki.linnakangas@i 1979 : 0 : err:
1980 : 0 : *errormsg = state->errormsg_buf;
1981 : :
1982 : 0 : return false;
1983 : : }
1984 : :
1985 : : /*
1986 : : * Returns information about the block that a block reference refers to.
1987 : : *
1988 : : * This is like XLogRecGetBlockTagExtended, except that the block reference
1989 : : * must exist and there's no access to prefetch_buffer.
1990 : : */
1991 : : void
4133 heikki.linnakangas@i 1992 :CBC 3353032 : XLogRecGetBlockTag(XLogReaderState *record, uint8 block_id,
1993 : : RelFileLocator *rlocator, ForkNumber *forknum,
1994 : : BlockNumber *blknum)
1995 : : {
1348 rhaas@postgresql.org 1996 [ - + ]: 3353032 : if (!XLogRecGetBlockTagExtended(record, block_id, rlocator, forknum,
1997 : : blknum, NULL))
1998 : : {
1999 : : #ifndef FRONTEND
1268 peter@eisentraut.org 2000 [ # # ]:UBC 0 : elog(ERROR, "could not locate backup block with ID %d in WAL record",
2001 : : block_id);
2002 : : #else
2003 : 0 : pg_fatal("could not locate backup block with ID %d in WAL record",
2004 : : block_id);
2005 : : #endif
2006 : : }
1438 tmunro@postgresql.or 2007 :CBC 3353032 : }
2008 : :
2009 : : /*
2010 : : * Returns information about the block that a block reference refers to,
2011 : : * optionally including the buffer that the block may already be in.
2012 : : *
2013 : : * If the WAL record contains a block reference with the given ID, *rlocator,
2014 : : * *forknum, *blknum and *prefetch_buffer are filled in (if not NULL), and
2015 : : * returns true. Otherwise returns false.
2016 : : */
2017 : : bool
2018 : 9387138 : XLogRecGetBlockTagExtended(XLogReaderState *record, uint8 block_id,
2019 : : RelFileLocator *rlocator, ForkNumber *forknum,
2020 : : BlockNumber *blknum,
2021 : : Buffer *prefetch_buffer)
2022 : : {
2023 : : DecodedBkpBlock *bkpb;
2024 : :
1434 tgl@sss.pgh.pa.us 2025 [ + + + + ]: 9387138 : if (!XLogRecHasBlockRef(record, block_id))
4133 heikki.linnakangas@i 2026 : 43937 : return false;
2027 : :
1458 tmunro@postgresql.or 2028 : 9343201 : bkpb = &record->record->blocks[block_id];
1348 rhaas@postgresql.org 2029 [ + + ]: 9343201 : if (rlocator)
2030 : 9282743 : *rlocator = bkpb->rlocator;
4133 heikki.linnakangas@i 2031 [ + + ]: 9343201 : if (forknum)
2032 : 5934981 : *forknum = bkpb->forknum;
2033 [ + + ]: 9343201 : if (blknum)
2034 : 7776423 : *blknum = bkpb->blkno;
1438 tmunro@postgresql.or 2035 [ + + ]: 9343201 : if (prefetch_buffer)
2036 : 2946651 : *prefetch_buffer = bkpb->prefetch_buffer;
4133 heikki.linnakangas@i 2037 : 9343201 : return true;
2038 : : }
2039 : :
2040 : : /*
2041 : : * Returns the data associated with a block reference, or NULL if there is
2042 : : * no data (e.g. because a full-page image was taken instead). The returned
2043 : : * pointer points to a MAXALIGNed buffer.
2044 : : */
2045 : : char *
2046 : 3571663 : XLogRecGetBlockData(XLogReaderState *record, uint8 block_id, Size *len)
2047 : : {
2048 : : DecodedBkpBlock *bkpb;
2049 : :
1458 tmunro@postgresql.or 2050 [ + - ]: 3571663 : if (block_id > record->record->max_block_id ||
2051 [ - + ]: 3571663 : !record->record->blocks[block_id].in_use)
4133 heikki.linnakangas@i 2052 :UBC 0 : return NULL;
2053 : :
1458 tmunro@postgresql.or 2054 :CBC 3571663 : bkpb = &record->record->blocks[block_id];
2055 : :
4133 heikki.linnakangas@i 2056 [ + + ]: 3571663 : if (!bkpb->has_data)
2057 : : {
2058 [ + - ]: 226 : if (len)
2059 : 226 : *len = 0;
2060 : 226 : return NULL;
2061 : : }
2062 : : else
2063 : : {
2064 [ + + ]: 3571437 : if (len)
2065 : 3567169 : *len = bkpb->data_len;
2066 : 3571437 : return bkpb->data;
2067 : : }
2068 : : }
2069 : :
2070 : : /*
2071 : : * Restore a full-page image from a backup block attached to an XLOG record.
2072 : : *
2073 : : * Returns true if a full-page image is restored, and false on failure with
2074 : : * an error to be consumed by the caller.
2075 : : */
2076 : : bool
2077 : 2430376 : RestoreBlockImage(XLogReaderState *record, uint8 block_id, char *page)
2078 : : {
2079 : : DecodedBkpBlock *bkpb;
2080 : : char *ptr;
2081 : : PGAlignedBlock tmp;
2082 : :
1458 tmunro@postgresql.or 2083 [ + - ]: 2430376 : if (block_id > record->record->max_block_id ||
2084 [ - + ]: 2430376 : !record->record->blocks[block_id].in_use)
2085 : : {
1283 michael@paquier.xyz 2086 :UBC 0 : report_invalid_record(record,
2087 : : "could not restore image at %X/%08X with invalid block %d specified",
2088 : 0 : LSN_FORMAT_ARGS(record->ReadRecPtr),
2089 : : block_id);
4133 heikki.linnakangas@i 2090 : 0 : return false;
2091 : : }
1458 tmunro@postgresql.or 2092 [ - + ]:CBC 2430376 : if (!record->record->blocks[block_id].has_image)
2093 : : {
251 alvherre@kurilemu.de 2094 :UNC 0 : report_invalid_record(record, "could not restore image at %X/%08X with invalid state, block %d",
1283 michael@paquier.xyz 2095 :UBC 0 : LSN_FORMAT_ARGS(record->ReadRecPtr),
2096 : : block_id);
4133 heikki.linnakangas@i 2097 : 0 : return false;
2098 : : }
2099 : :
1458 tmunro@postgresql.or 2100 :CBC 2430376 : bkpb = &record->record->blocks[block_id];
4022 fujii@postgresql.org 2101 : 2430376 : ptr = bkpb->bkp_image;
2102 : :
1720 michael@paquier.xyz 2103 [ - + ]: 2430376 : if (BKPIMAGE_COMPRESSED(bkpb->bimg_info))
2104 : : {
2105 : : /* If a backup block image is compressed, decompress it */
1720 michael@paquier.xyz 2106 :UBC 0 : bool decomp_success = true;
2107 : :
2108 [ # # ]: 0 : if ((bkpb->bimg_info & BKPIMAGE_COMPRESS_PGLZ) != 0)
2109 : : {
2110 [ # # ]: 0 : if (pglz_decompress(ptr, bkpb->bimg_len, tmp.data,
2111 : 0 : BLCKSZ - bkpb->hole_length, true) < 0)
2112 : 0 : decomp_success = false;
2113 : : }
2114 [ # # ]: 0 : else if ((bkpb->bimg_info & BKPIMAGE_COMPRESS_LZ4) != 0)
2115 : : {
2116 : : #ifdef USE_LZ4
2117 [ # # ]: 0 : if (LZ4_decompress_safe(ptr, tmp.data,
2118 : 0 : bkpb->bimg_len, BLCKSZ - bkpb->hole_length) <= 0)
2119 : 0 : decomp_success = false;
2120 : : #else
2121 : : report_invalid_record(record, "could not restore image at %X/%08X compressed with %s not supported by build, block %d",
2122 : : LSN_FORMAT_ARGS(record->ReadRecPtr),
2123 : : "LZ4",
2124 : : block_id);
2125 : : return false;
2126 : : #endif
2127 : : }
1465 2128 [ # # ]: 0 : else if ((bkpb->bimg_info & BKPIMAGE_COMPRESS_ZSTD) != 0)
2129 : : {
2130 : : #ifdef USE_ZSTD
2131 : 0 : size_t decomp_result = ZSTD_decompress(tmp.data,
2132 : 0 : BLCKSZ - bkpb->hole_length,
2133 : 0 : ptr, bkpb->bimg_len);
2134 : :
2135 [ # # ]: 0 : if (ZSTD_isError(decomp_result))
2136 : 0 : decomp_success = false;
2137 : : #else
2138 : : report_invalid_record(record, "could not restore image at %X/%08X compressed with %s not supported by build, block %d",
2139 : : LSN_FORMAT_ARGS(record->ReadRecPtr),
2140 : : "zstd",
2141 : : block_id);
2142 : : return false;
2143 : : #endif
2144 : : }
2145 : : else
2146 : : {
251 alvherre@kurilemu.de 2147 :UNC 0 : report_invalid_record(record, "could not restore image at %X/%08X compressed with unknown method, block %d",
1710 michael@paquier.xyz 2148 :UBC 0 : LSN_FORMAT_ARGS(record->ReadRecPtr),
2149 : : block_id);
1720 2150 : 0 : return false;
2151 : : }
2152 : :
2153 [ # # ]: 0 : if (!decomp_success)
2154 : : {
251 alvherre@kurilemu.de 2155 :UNC 0 : report_invalid_record(record, "could not decompress image at %X/%08X, block %d",
1846 peter@eisentraut.org 2156 :UBC 0 : LSN_FORMAT_ARGS(record->ReadRecPtr),
2157 : : block_id);
4022 fujii@postgresql.org 2158 : 0 : return false;
2159 : : }
2160 : :
2752 tgl@sss.pgh.pa.us 2161 : 0 : ptr = tmp.data;
2162 : : }
2163 : :
2164 : : /* generate page, taking into account hole if necessary */
4133 heikki.linnakangas@i 2165 [ + + ]:CBC 2430376 : if (bkpb->hole_length == 0)
2166 : : {
4022 fujii@postgresql.org 2167 : 16877 : memcpy(page, ptr, BLCKSZ);
2168 : : }
2169 : : else
2170 : : {
2171 : 2413499 : memcpy(page, ptr, bkpb->hole_offset);
2172 : : /* must zero-fill the hole */
4133 heikki.linnakangas@i 2173 [ + + + - : 12682368 : MemSet(page + bkpb->hole_offset, 0, bkpb->hole_length);
+ - + + +
+ ]
2174 : 2413499 : memcpy(page + (bkpb->hole_offset + bkpb->hole_length),
4022 fujii@postgresql.org 2175 : 2413499 : ptr + bkpb->hole_offset,
4133 heikki.linnakangas@i 2176 : 2413499 : BLCKSZ - (bkpb->hole_offset + bkpb->hole_length));
2177 : : }
2178 : :
2179 : 2430376 : return true;
2180 : : }
2181 : :
2182 : : #ifndef FRONTEND
2183 : :
2184 : : /*
2185 : : * Extract the FullTransactionId from a WAL record.
2186 : : */
2187 : : FullTransactionId
2435 tmunro@postgresql.or 2188 :UBC 0 : XLogRecGetFullXid(XLogReaderState *record)
2189 : : {
2190 : : /*
2191 : : * This function is only safe during replay, because it depends on the
2192 : : * replay state. See AdvanceNextFullTransactionIdPastXid() for more.
2193 : : */
2194 [ # # # # ]: 0 : Assert(AmStartupProcess() || !IsUnderPostmaster);
2195 : :
414 noah@leadboat.com 2196 : 0 : return FullTransactionIdFromAllowableAt(TransamVariables->nextXid,
2197 : 0 : XLogRecGetXid(record));
2198 : : }
2199 : :
2200 : : #endif
|
