PostgreSQL Source Code  git master
blkreftable.c
Go to the documentation of this file.
1 /*-------------------------------------------------------------------------
2  *
3  * blkreftable.c
4  * Block reference tables.
5  *
6  * A block reference table is used to keep track of which blocks have
7  * been modified by WAL records within a certain LSN range.
8  *
9  * For each relation fork, we keep track of all blocks that have appeared
10  * in block reference in the WAL. We also keep track of the "limit block",
11  * which is the smallest relation length in blocks known to have occurred
12  * during that range of WAL records. This should be set to 0 if the relation
13  * fork is created or destroyed, and to the post-truncation length if
14  * truncated.
15  *
16  * Whenever we set the limit block, we also forget about any modified blocks
17  * beyond that point. Those blocks don't exist any more. Such blocks can
18  * later be marked as modified again; if that happens, it means the relation
19  * was re-extended.
20  *
21  * Portions Copyright (c) 2010-2024, PostgreSQL Global Development Group
22  *
23  * src/common/blkreftable.c
24  *
25  *-------------------------------------------------------------------------
26  */
27 
28 
29 #ifndef FRONTEND
30 #include "postgres.h"
31 #else
32 #include "postgres_fe.h"
33 #endif
34 
35 #ifdef FRONTEND
36 #include "common/logging.h"
37 #endif
38 
39 #include "common/blkreftable.h"
40 #include "common/hashfn.h"
41 #include "port/pg_crc32c.h"
42 
43 /*
44  * A block reference table keeps track of the status of each relation
45  * fork individually.
46  */
47 typedef struct BlockRefTableKey
48 {
52 
53 /*
54  * We could need to store data either for a relation in which only a
55  * tiny fraction of the blocks have been modified or for a relation in
56  * which nearly every block has been modified, and we want a
57  * space-efficient representation in both cases. To accomplish this,
58  * we divide the relation into chunks of 2^16 blocks and choose between
59  * an array representation and a bitmap representation for each chunk.
60  *
61  * When the number of modified blocks in a given chunk is small, we
62  * essentially store an array of block numbers, but we need not store the
63  * entire block number: instead, we store each block number as a 2-byte
64  * offset from the start of the chunk.
65  *
66  * When the number of modified blocks in a given chunk is large, we switch
67  * to a bitmap representation.
68  *
69  * These same basic representational choices are used both when a block
70  * reference table is stored in memory and when it is serialized to disk.
71  *
72  * In the in-memory representation, we initially allocate each chunk with
73  * space for a number of entries given by INITIAL_ENTRIES_PER_CHUNK and
74  * increase that as necessary until we reach MAX_ENTRIES_PER_CHUNK.
75  * Any chunk whose allocated size reaches MAX_ENTRIES_PER_CHUNK is converted
76  * to a bitmap, and thus never needs to grow further.
77  */
78 #define BLOCKS_PER_CHUNK (1 << 16)
79 #define BLOCKS_PER_ENTRY (BITS_PER_BYTE * sizeof(uint16))
80 #define MAX_ENTRIES_PER_CHUNK (BLOCKS_PER_CHUNK / BLOCKS_PER_ENTRY)
81 #define INITIAL_ENTRIES_PER_CHUNK 16
83 
84 /*
85  * State for one relation fork.
86  *
87  * 'rlocator' and 'forknum' identify the relation fork to which this entry
88  * pertains.
89  *
90  * 'limit_block' is the shortest known length of the relation in blocks
91  * within the LSN range covered by a particular block reference table.
92  * It should be set to 0 if the relation fork is created or dropped. If the
93  * relation fork is truncated, it should be set to the number of blocks that
94  * remain after truncation.
95  *
96  * 'nchunks' is the allocated length of each of the three arrays that follow.
97  * We can only represent the status of block numbers less than nchunks *
98  * BLOCKS_PER_CHUNK.
99  *
100  * 'chunk_size' is an array storing the allocated size of each chunk.
101  *
102  * 'chunk_usage' is an array storing the number of elements used in each
103  * chunk. If that value is less than MAX_ENTRIES_PER_CHUNK, the corresponding
104  * chunk is used as an array; else the corresponding chunk is used as a bitmap.
105  * When used as a bitmap, the least significant bit of the first array element
106  * is the status of the lowest-numbered block covered by this chunk.
107  *
108  * 'chunk_data' is the array of chunks.
109  */
111 {
114  char status;
119 };
120 
121 /* Declare and define a hash table over type BlockRefTableEntry. */
122 #define SH_PREFIX blockreftable
123 #define SH_ELEMENT_TYPE BlockRefTableEntry
124 #define SH_KEY_TYPE BlockRefTableKey
125 #define SH_KEY key
126 #define SH_HASH_KEY(tb, key) \
127  hash_bytes((const unsigned char *) &key, sizeof(BlockRefTableKey))
128 #define SH_EQUAL(tb, a, b) (memcmp(&a, &b, sizeof(BlockRefTableKey)) == 0)
129 #define SH_SCOPE static inline
130 #ifdef FRONTEND
131 #define SH_RAW_ALLOCATOR pg_malloc0
132 #endif
133 #define SH_DEFINE
134 #define SH_DECLARE
135 #include "lib/simplehash.h"
136 
137 /*
138  * A block reference table is basically just the hash table, but we don't
139  * want to expose that to outside callers.
140  *
141  * We keep track of the memory context in use explicitly too, so that it's
142  * easy to place all of our allocations in the same context.
143  */
145 {
146  blockreftable_hash *hash;
147 #ifndef FRONTEND
149 #endif
150 };
151 
152 /*
153  * On-disk serialization format for block reference table entries.
154  */
156 {
162 
163 /*
164  * Buffer size, so that we avoid doing many small I/Os.
165  */
166 #define BUFSIZE 65536
167 
168 /*
169  * Ad-hoc buffer for file I/O.
170  */
171 typedef struct BlockRefTableBuffer
172 {
175  char data[BUFSIZE];
176  int used;
177  int cursor;
180 
181 /*
182  * State for keeping track of progress while incrementally reading a block
183  * table reference file from disk.
184  *
185  * total_chunks means the number of chunks for the RelFileLocator/ForkNumber
186  * combination that is currently being read, and consumed_chunks is the number
187  * of those that have been read. (We always read all the information for
188  * a single chunk at one time, so we don't need to be able to represent the
189  * state where a chunk has been partially read.)
190  *
191  * chunk_size is the array of chunk sizes. The length is given by total_chunks.
192  *
193  * chunk_data holds the current chunk.
194  *
195  * chunk_position helps us figure out how much progress we've made in returning
196  * the block numbers for the current chunk to the caller. If the chunk is a
197  * bitmap, it's the number of bits we've scanned; otherwise, it's the number
198  * of chunk entries we've scanned.
199  */
201 {
211 };
212 
213 /*
214  * State for keeping track of progress while incrementally writing a block
215  * reference table file to disk.
216  */
218 {
220 };
221 
222 /* Function prototypes. */
223 static int BlockRefTableComparator(const void *a, const void *b);
224 static void BlockRefTableFlush(BlockRefTableBuffer *buffer);
225 static void BlockRefTableRead(BlockRefTableReader *reader, void *data,
226  int length);
227 static void BlockRefTableWrite(BlockRefTableBuffer *buffer, void *data,
228  int length);
230 
231 /*
232  * Create an empty block reference table.
233  */
236 {
237  BlockRefTable *brtab = palloc(sizeof(BlockRefTable));
238 
239  /*
240  * Even completely empty database has a few hundred relation forks, so it
241  * seems best to size the hash on the assumption that we're going to have
242  * at least a few thousand entries.
243  */
244 #ifdef FRONTEND
245  brtab->hash = blockreftable_create(4096, NULL);
246 #else
247  brtab->mcxt = CurrentMemoryContext;
248  brtab->hash = blockreftable_create(brtab->mcxt, 4096, NULL);
249 #endif
250 
251  return brtab;
252 }
253 
254 /*
255  * Set the "limit block" for a relation fork and forget any modified blocks
256  * with equal or higher block numbers.
257  *
258  * The "limit block" is the shortest known length of the relation within the
259  * range of WAL records covered by this block reference table.
260  */
261 void
263  const RelFileLocator *rlocator,
264  ForkNumber forknum,
265  BlockNumber limit_block)
266 {
267  BlockRefTableEntry *brtentry;
268  BlockRefTableKey key = {{0}}; /* make sure any padding is zero */
269  bool found;
270 
271  memcpy(&key.rlocator, rlocator, sizeof(RelFileLocator));
272  key.forknum = forknum;
273  brtentry = blockreftable_insert(brtab->hash, key, &found);
274 
275  if (!found)
276  {
277  /*
278  * We have no existing data about this relation fork, so just record
279  * the limit_block value supplied by the caller, and make sure other
280  * parts of the entry are properly initialized.
281  */
282  brtentry->limit_block = limit_block;
283  brtentry->nchunks = 0;
284  brtentry->chunk_size = NULL;
285  brtentry->chunk_usage = NULL;
286  brtentry->chunk_data = NULL;
287  return;
288  }
289 
290  BlockRefTableEntrySetLimitBlock(brtentry, limit_block);
291 }
292 
293 /*
294  * Mark a block in a given relation fork as known to have been modified.
295  */
296 void
298  const RelFileLocator *rlocator,
299  ForkNumber forknum,
300  BlockNumber blknum)
301 {
302  BlockRefTableEntry *brtentry;
303  BlockRefTableKey key = {{0}}; /* make sure any padding is zero */
304  bool found;
305 #ifndef FRONTEND
306  MemoryContext oldcontext = MemoryContextSwitchTo(brtab->mcxt);
307 #endif
308 
309  memcpy(&key.rlocator, rlocator, sizeof(RelFileLocator));
310  key.forknum = forknum;
311  brtentry = blockreftable_insert(brtab->hash, key, &found);
312 
313  if (!found)
314  {
315  /*
316  * We want to set the initial limit block value to something higher
317  * than any legal block number. InvalidBlockNumber fits the bill.
318  */
319  brtentry->limit_block = InvalidBlockNumber;
320  brtentry->nchunks = 0;
321  brtentry->chunk_size = NULL;
322  brtentry->chunk_usage = NULL;
323  brtentry->chunk_data = NULL;
324  }
325 
326  BlockRefTableEntryMarkBlockModified(brtentry, forknum, blknum);
327 
328 #ifndef FRONTEND
329  MemoryContextSwitchTo(oldcontext);
330 #endif
331 }
332 
333 /*
334  * Get an entry from a block reference table.
335  *
336  * If the entry does not exist, this function returns NULL. Otherwise, it
337  * returns the entry and sets *limit_block to the value from the entry.
338  */
341  ForkNumber forknum, BlockNumber *limit_block)
342 {
343  BlockRefTableKey key = {{0}}; /* make sure any padding is zero */
344  BlockRefTableEntry *entry;
345 
346  Assert(limit_block != NULL);
347 
348  memcpy(&key.rlocator, rlocator, sizeof(RelFileLocator));
349  key.forknum = forknum;
350  entry = blockreftable_lookup(brtab->hash, key);
351 
352  if (entry != NULL)
353  *limit_block = entry->limit_block;
354 
355  return entry;
356 }
357 
358 /*
359  * Get block numbers from a table entry.
360  *
361  * 'blocks' must point to enough space to hold at least 'nblocks' block
362  * numbers, and any block numbers we manage to get will be written there.
363  * The return value is the number of block numbers actually written.
364  *
365  * We do not return block numbers unless they are greater than or equal to
366  * start_blkno and strictly less than stop_blkno.
367  */
368 int
370  BlockNumber start_blkno,
371  BlockNumber stop_blkno,
372  BlockNumber *blocks,
373  int nblocks)
374 {
375  uint32 start_chunkno;
376  uint32 stop_chunkno;
377  uint32 chunkno;
378  int nresults = 0;
379 
380  Assert(entry != NULL);
381 
382  /*
383  * Figure out which chunks could potentially contain blocks of interest.
384  *
385  * We need to be careful about overflow here, because stop_blkno could be
386  * InvalidBlockNumber or something very close to it.
387  */
388  start_chunkno = start_blkno / BLOCKS_PER_CHUNK;
389  stop_chunkno = stop_blkno / BLOCKS_PER_CHUNK;
390  if ((stop_blkno % BLOCKS_PER_CHUNK) != 0)
391  ++stop_chunkno;
392  if (stop_chunkno > entry->nchunks)
393  stop_chunkno = entry->nchunks;
394 
395  /*
396  * Loop over chunks.
397  */
398  for (chunkno = start_chunkno; chunkno < stop_chunkno; ++chunkno)
399  {
400  uint16 chunk_usage = entry->chunk_usage[chunkno];
401  BlockRefTableChunk chunk_data = entry->chunk_data[chunkno];
402  unsigned start_offset = 0;
403  unsigned stop_offset = BLOCKS_PER_CHUNK;
404 
405  /*
406  * If the start and/or stop block number falls within this chunk, the
407  * whole chunk may not be of interest. Figure out which portion we
408  * care about, if it's not the whole thing.
409  */
410  if (chunkno == start_chunkno)
411  start_offset = start_blkno % BLOCKS_PER_CHUNK;
412  if (chunkno == stop_chunkno - 1)
413  {
414  Assert(stop_blkno > chunkno * BLOCKS_PER_CHUNK);
415  stop_offset = stop_blkno - (chunkno * BLOCKS_PER_CHUNK);
416  Assert(stop_offset <= BLOCKS_PER_CHUNK);
417  }
418 
419  /*
420  * Handling differs depending on whether this is an array of offsets
421  * or a bitmap.
422  */
423  if (chunk_usage == MAX_ENTRIES_PER_CHUNK)
424  {
425  unsigned i;
426 
427  /* It's a bitmap, so test every relevant bit. */
428  for (i = start_offset; i < stop_offset; ++i)
429  {
430  uint16 w = chunk_data[i / BLOCKS_PER_ENTRY];
431 
432  if ((w & (1 << (i % BLOCKS_PER_ENTRY))) != 0)
433  {
434  BlockNumber blkno = chunkno * BLOCKS_PER_CHUNK + i;
435 
436  blocks[nresults++] = blkno;
437 
438  /* Early exit if we run out of output space. */
439  if (nresults == nblocks)
440  return nresults;
441  }
442  }
443  }
444  else
445  {
446  unsigned i;
447 
448  /* It's an array of offsets, so check each one. */
449  for (i = 0; i < chunk_usage; ++i)
450  {
451  uint16 offset = chunk_data[i];
452 
453  if (offset >= start_offset && offset < stop_offset)
454  {
455  BlockNumber blkno = chunkno * BLOCKS_PER_CHUNK + offset;
456 
457  blocks[nresults++] = blkno;
458 
459  /* Early exit if we run out of output space. */
460  if (nresults == nblocks)
461  return nresults;
462  }
463  }
464  }
465  }
466 
467  return nresults;
468 }
469 
470 /*
471  * Serialize a block reference table to a file.
472  */
473 void
475  io_callback_fn write_callback,
476  void *write_callback_arg)
477 {
478  BlockRefTableSerializedEntry *sdata = NULL;
479  BlockRefTableBuffer buffer;
480  uint32 magic = BLOCKREFTABLE_MAGIC;
481 
482  /* Prepare buffer. */
483  memset(&buffer, 0, sizeof(BlockRefTableBuffer));
484  buffer.io_callback = write_callback;
485  buffer.io_callback_arg = write_callback_arg;
486  INIT_CRC32C(buffer.crc);
487 
488  /* Write magic number. */
489  BlockRefTableWrite(&buffer, &magic, sizeof(uint32));
490 
491  /* Write the entries, assuming there are some. */
492  if (brtab->hash->members > 0)
493  {
494  unsigned i = 0;
495  blockreftable_iterator it;
496  BlockRefTableEntry *brtentry;
497 
498  /* Extract entries into serializable format and sort them. */
499  sdata =
500  palloc(brtab->hash->members * sizeof(BlockRefTableSerializedEntry));
501  blockreftable_start_iterate(brtab->hash, &it);
502  while ((brtentry = blockreftable_iterate(brtab->hash, &it)) != NULL)
503  {
504  BlockRefTableSerializedEntry *sentry = &sdata[i++];
505 
506  sentry->rlocator = brtentry->key.rlocator;
507  sentry->forknum = brtentry->key.forknum;
508  sentry->limit_block = brtentry->limit_block;
509  sentry->nchunks = brtentry->nchunks;
510 
511  /* trim trailing zero entries */
512  while (sentry->nchunks > 0 &&
513  brtentry->chunk_usage[sentry->nchunks - 1] == 0)
514  sentry->nchunks--;
515  }
516  Assert(i == brtab->hash->members);
517  qsort(sdata, i, sizeof(BlockRefTableSerializedEntry),
519 
520  /* Loop over entries in sorted order and serialize each one. */
521  for (i = 0; i < brtab->hash->members; ++i)
522  {
523  BlockRefTableSerializedEntry *sentry = &sdata[i];
524  BlockRefTableKey key = {{0}}; /* make sure any padding is zero */
525  unsigned j;
526 
527  /* Write the serialized entry itself. */
528  BlockRefTableWrite(&buffer, sentry,
530 
531  /* Look up the original entry so we can access the chunks. */
532  memcpy(&key.rlocator, &sentry->rlocator, sizeof(RelFileLocator));
533  key.forknum = sentry->forknum;
534  brtentry = blockreftable_lookup(brtab->hash, key);
535  Assert(brtentry != NULL);
536 
537  /* Write the untruncated portion of the chunk length array. */
538  if (sentry->nchunks != 0)
539  BlockRefTableWrite(&buffer, brtentry->chunk_usage,
540  sentry->nchunks * sizeof(uint16));
541 
542  /* Write the contents of each chunk. */
543  for (j = 0; j < brtentry->nchunks; ++j)
544  {
545  if (brtentry->chunk_usage[j] == 0)
546  continue;
547  BlockRefTableWrite(&buffer, brtentry->chunk_data[j],
548  brtentry->chunk_usage[j] * sizeof(uint16));
549  }
550  }
551  }
552 
553  /* Write out appropriate terminator and CRC and flush buffer. */
555 }
556 
557 /*
558  * Prepare to incrementally read a block reference table file.
559  *
560  * 'read_callback' is a function that can be called to read data from the
561  * underlying file (or other data source) into our internal buffer.
562  *
563  * 'read_callback_arg' is an opaque argument to be passed to read_callback.
564  *
565  * 'error_filename' is the filename that should be included in error messages
566  * if the file is found to be malformed. The value is not copied, so the
567  * caller should ensure that it remains valid until done with this
568  * BlockRefTableReader.
569  *
570  * 'error_callback' is a function to be called if the file is found to be
571  * malformed. This is not used for I/O errors, which must be handled internally
572  * by read_callback.
573  *
574  * 'error_callback_arg' is an opaque argument to be passed to error_callback.
575  */
578  void *read_callback_arg,
579  char *error_filename,
580  report_error_fn error_callback,
581  void *error_callback_arg)
582 {
583  BlockRefTableReader *reader;
584  uint32 magic;
585 
586  /* Initialize data structure. */
587  reader = palloc0(sizeof(BlockRefTableReader));
588  reader->buffer.io_callback = read_callback;
589  reader->buffer.io_callback_arg = read_callback_arg;
590  reader->error_filename = error_filename;
591  reader->error_callback = error_callback;
592  reader->error_callback_arg = error_callback_arg;
593  INIT_CRC32C(reader->buffer.crc);
594 
595  /* Verify magic number. */
596  BlockRefTableRead(reader, &magic, sizeof(uint32));
597  if (magic != BLOCKREFTABLE_MAGIC)
598  error_callback(error_callback_arg,
599  "file \"%s\" has wrong magic number: expected %u, found %u",
600  error_filename,
601  BLOCKREFTABLE_MAGIC, magic);
602 
603  return reader;
604 }
605 
606 /*
607  * Read next relation fork covered by this block reference table file.
608  *
609  * After calling this function, you must call BlockRefTableReaderGetBlocks
610  * until it returns 0 before calling it again.
611  */
612 bool
614  RelFileLocator *rlocator,
615  ForkNumber *forknum,
616  BlockNumber *limit_block)
617 {
619  BlockRefTableSerializedEntry zentry = {{0}};
620 
621  /*
622  * Sanity check: caller must read all blocks from all chunks before moving
623  * on to the next relation.
624  */
625  Assert(reader->total_chunks == reader->consumed_chunks);
626 
627  /* Read serialized entry. */
628  BlockRefTableRead(reader, &sentry,
630 
631  /*
632  * If we just read the sentinel entry indicating that we've reached the
633  * end, read and check the CRC.
634  */
635  if (memcmp(&sentry, &zentry, sizeof(BlockRefTableSerializedEntry)) == 0)
636  {
637  pg_crc32c expected_crc;
638  pg_crc32c actual_crc;
639 
640  /*
641  * We want to know the CRC of the file excluding the 4-byte CRC
642  * itself, so copy the current value of the CRC accumulator before
643  * reading those bytes, and use the copy to finalize the calculation.
644  */
645  expected_crc = reader->buffer.crc;
646  FIN_CRC32C(expected_crc);
647 
648  /* Now we can read the actual value. */
649  BlockRefTableRead(reader, &actual_crc, sizeof(pg_crc32c));
650 
651  /* Throw an error if there is a mismatch. */
652  if (!EQ_CRC32C(expected_crc, actual_crc))
653  reader->error_callback(reader->error_callback_arg,
654  "file \"%s\" has wrong checksum: expected %08X, found %08X",
655  reader->error_filename, expected_crc, actual_crc);
656 
657  return false;
658  }
659 
660  /* Read chunk size array. */
661  if (reader->chunk_size != NULL)
662  pfree(reader->chunk_size);
663  reader->chunk_size = palloc(sentry.nchunks * sizeof(uint16));
664  BlockRefTableRead(reader, reader->chunk_size,
665  sentry.nchunks * sizeof(uint16));
666 
667  /* Set up for chunk scan. */
668  reader->total_chunks = sentry.nchunks;
669  reader->consumed_chunks = 0;
670 
671  /* Return data to caller. */
672  memcpy(rlocator, &sentry.rlocator, sizeof(RelFileLocator));
673  *forknum = sentry.forknum;
674  *limit_block = sentry.limit_block;
675  return true;
676 }
677 
678 /*
679  * Get modified blocks associated with the relation fork returned by
680  * the most recent call to BlockRefTableReaderNextRelation.
681  *
682  * On return, block numbers will be written into the 'blocks' array, whose
683  * length should be passed via 'nblocks'. The return value is the number of
684  * entries actually written into the 'blocks' array, which may be less than
685  * 'nblocks' if we run out of modified blocks in the relation fork before
686  * we run out of room in the array.
687  */
688 unsigned
690  BlockNumber *blocks,
691  int nblocks)
692 {
693  unsigned blocks_found = 0;
694 
695  /* Must provide space for at least one block number to be returned. */
696  Assert(nblocks > 0);
697 
698  /* Loop collecting blocks to return to caller. */
699  for (;;)
700  {
701  uint16 next_chunk_size;
702 
703  /*
704  * If we've read at least one chunk, maybe it contains some block
705  * numbers that could satisfy caller's request.
706  */
707  if (reader->consumed_chunks > 0)
708  {
709  uint32 chunkno = reader->consumed_chunks - 1;
710  uint16 chunk_size = reader->chunk_size[chunkno];
711 
712  if (chunk_size == MAX_ENTRIES_PER_CHUNK)
713  {
714  /* Bitmap format, so search for bits that are set. */
715  while (reader->chunk_position < BLOCKS_PER_CHUNK &&
716  blocks_found < nblocks)
717  {
718  uint16 chunkoffset = reader->chunk_position;
719  uint16 w;
720 
721  w = reader->chunk_data[chunkoffset / BLOCKS_PER_ENTRY];
722  if ((w & (1u << (chunkoffset % BLOCKS_PER_ENTRY))) != 0)
723  blocks[blocks_found++] =
724  chunkno * BLOCKS_PER_CHUNK + chunkoffset;
725  ++reader->chunk_position;
726  }
727  }
728  else
729  {
730  /* Not in bitmap format, so each entry is a 2-byte offset. */
731  while (reader->chunk_position < chunk_size &&
732  blocks_found < nblocks)
733  {
734  blocks[blocks_found++] = chunkno * BLOCKS_PER_CHUNK
735  + reader->chunk_data[reader->chunk_position];
736  ++reader->chunk_position;
737  }
738  }
739  }
740 
741  /* We found enough blocks, so we're done. */
742  if (blocks_found >= nblocks)
743  break;
744 
745  /*
746  * We didn't find enough blocks, so we must need the next chunk. If
747  * there are none left, though, then we're done anyway.
748  */
749  if (reader->consumed_chunks == reader->total_chunks)
750  break;
751 
752  /*
753  * Read data for next chunk and reset scan position to beginning of
754  * chunk. Note that the next chunk might be empty, in which case we
755  * consume the chunk without actually consuming any bytes from the
756  * underlying file.
757  */
758  next_chunk_size = reader->chunk_size[reader->consumed_chunks];
759  if (next_chunk_size > 0)
760  BlockRefTableRead(reader, reader->chunk_data,
761  next_chunk_size * sizeof(uint16));
762  ++reader->consumed_chunks;
763  reader->chunk_position = 0;
764  }
765 
766  return blocks_found;
767 }
768 
769 /*
770  * Release memory used while reading a block reference table from a file.
771  */
772 void
774 {
775  if (reader->chunk_size != NULL)
776  {
777  pfree(reader->chunk_size);
778  reader->chunk_size = NULL;
779  }
780  pfree(reader);
781 }
782 
783 /*
784  * Prepare to write a block reference table file incrementally.
785  *
786  * Caller must be able to supply BlockRefTableEntry objects sorted in the
787  * appropriate order.
788  */
791  void *write_callback_arg)
792 {
793  BlockRefTableWriter *writer;
794  uint32 magic = BLOCKREFTABLE_MAGIC;
795 
796  /* Prepare buffer and CRC check and save callbacks. */
797  writer = palloc0(sizeof(BlockRefTableWriter));
798  writer->buffer.io_callback = write_callback;
799  writer->buffer.io_callback_arg = write_callback_arg;
800  INIT_CRC32C(writer->buffer.crc);
801 
802  /* Write magic number. */
803  BlockRefTableWrite(&writer->buffer, &magic, sizeof(uint32));
804 
805  return writer;
806 }
807 
808 /*
809  * Append one entry to a block reference table file.
810  *
811  * Note that entries must be written in the proper order, that is, sorted by
812  * tablespace, then database, then relfilenumber, then fork number. Caller
813  * is responsible for supplying data in the correct order. If that seems hard,
814  * use an in-memory BlockRefTable instead.
815  */
816 void
818 {
820  unsigned j;
821 
822  /* Convert to serialized entry format. */
823  sentry.rlocator = entry->key.rlocator;
824  sentry.forknum = entry->key.forknum;
825  sentry.limit_block = entry->limit_block;
826  sentry.nchunks = entry->nchunks;
827 
828  /* Trim trailing zero entries. */
829  while (sentry.nchunks > 0 && entry->chunk_usage[sentry.nchunks - 1] == 0)
830  sentry.nchunks--;
831 
832  /* Write the serialized entry itself. */
833  BlockRefTableWrite(&writer->buffer, &sentry,
835 
836  /* Write the untruncated portion of the chunk length array. */
837  if (sentry.nchunks != 0)
838  BlockRefTableWrite(&writer->buffer, entry->chunk_usage,
839  sentry.nchunks * sizeof(uint16));
840 
841  /* Write the contents of each chunk. */
842  for (j = 0; j < entry->nchunks; ++j)
843  {
844  if (entry->chunk_usage[j] == 0)
845  continue;
846  BlockRefTableWrite(&writer->buffer, entry->chunk_data[j],
847  entry->chunk_usage[j] * sizeof(uint16));
848  }
849 }
850 
851 /*
852  * Finalize an incremental write of a block reference table file.
853  */
854 void
856 {
858  pfree(writer);
859 }
860 
861 /*
862  * Allocate a standalone BlockRefTableEntry.
863  *
864  * When we're manipulating a full in-memory BlockRefTable, the entries are
865  * part of the hash table and are allocated by simplehash. This routine is
866  * used by callers that want to write out a BlockRefTable to a file without
867  * needing to store the whole thing in memory at once.
868  *
869  * Entries allocated by this function can be manipulated using the functions
870  * BlockRefTableEntrySetLimitBlock and BlockRefTableEntryMarkBlockModified
871  * and then written using BlockRefTableWriteEntry and freed using
872  * BlockRefTableFreeEntry.
873  */
876 {
878 
879  memcpy(&entry->key.rlocator, &rlocator, sizeof(RelFileLocator));
880  entry->key.forknum = forknum;
882 
883  return entry;
884 }
885 
886 /*
887  * Update a BlockRefTableEntry with a new value for the "limit block" and
888  * forget any equal-or-higher-numbered modified blocks.
889  *
890  * The "limit block" is the shortest known length of the relation within the
891  * range of WAL records covered by this block reference table.
892  */
893 void
895  BlockNumber limit_block)
896 {
897  unsigned chunkno;
898  unsigned limit_chunkno;
899  unsigned limit_chunkoffset;
900  BlockRefTableChunk limit_chunk;
901 
902  /* If we already have an equal or lower limit block, do nothing. */
903  if (limit_block >= entry->limit_block)
904  return;
905 
906  /* Record the new limit block value. */
907  entry->limit_block = limit_block;
908 
909  /*
910  * Figure out which chunk would store the state of the new limit block,
911  * and which offset within that chunk.
912  */
913  limit_chunkno = limit_block / BLOCKS_PER_CHUNK;
914  limit_chunkoffset = limit_block % BLOCKS_PER_CHUNK;
915 
916  /*
917  * If the number of chunks is not large enough for any blocks with equal
918  * or higher block numbers to exist, then there is nothing further to do.
919  */
920  if (limit_chunkno >= entry->nchunks)
921  return;
922 
923  /* Discard entire contents of any higher-numbered chunks. */
924  for (chunkno = limit_chunkno + 1; chunkno < entry->nchunks; ++chunkno)
925  entry->chunk_usage[chunkno] = 0;
926 
927  /*
928  * Next, we need to discard any offsets within the chunk that would
929  * contain the limit_block. We must handle this differently depending on
930  * whether the chunk that would contain limit_block is a bitmap or an
931  * array of offsets.
932  */
933  limit_chunk = entry->chunk_data[limit_chunkno];
934  if (entry->chunk_usage[limit_chunkno] == MAX_ENTRIES_PER_CHUNK)
935  {
936  unsigned chunkoffset;
937 
938  /* It's a bitmap. Unset bits. */
939  for (chunkoffset = limit_chunkoffset; chunkoffset < BLOCKS_PER_CHUNK;
940  ++chunkoffset)
941  limit_chunk[chunkoffset / BLOCKS_PER_ENTRY] &=
942  ~(1 << (chunkoffset % BLOCKS_PER_ENTRY));
943  }
944  else
945  {
946  unsigned i,
947  j = 0;
948 
949  /* It's an offset array. Filter out large offsets. */
950  for (i = 0; i < entry->chunk_usage[limit_chunkno]; ++i)
951  {
952  Assert(j <= i);
953  if (limit_chunk[i] < limit_chunkoffset)
954  limit_chunk[j++] = limit_chunk[i];
955  }
956  Assert(j <= entry->chunk_usage[limit_chunkno]);
957  entry->chunk_usage[limit_chunkno] = j;
958  }
959 }
960 
961 /*
962  * Mark a block in a given BlockRefTableEntry as known to have been modified.
963  */
964 void
966  ForkNumber forknum,
967  BlockNumber blknum)
968 {
969  unsigned chunkno;
970  unsigned chunkoffset;
971  unsigned i;
972 
973  /*
974  * Which chunk should store the state of this block? And what is the
975  * offset of this block relative to the start of that chunk?
976  */
977  chunkno = blknum / BLOCKS_PER_CHUNK;
978  chunkoffset = blknum % BLOCKS_PER_CHUNK;
979 
980  /*
981  * If 'nchunks' isn't big enough for us to be able to represent the state
982  * of this block, we need to enlarge our arrays.
983  */
984  if (chunkno >= entry->nchunks)
985  {
986  unsigned max_chunks;
987  unsigned extra_chunks;
988 
989  /*
990  * New array size is a power of 2, at least 16, big enough so that
991  * chunkno will be a valid array index.
992  */
993  max_chunks = Max(16, entry->nchunks);
994  while (max_chunks < chunkno + 1)
995  max_chunks *= 2;
996  extra_chunks = max_chunks - entry->nchunks;
997 
998  if (entry->nchunks == 0)
999  {
1000  entry->chunk_size = palloc0(sizeof(uint16) * max_chunks);
1001  entry->chunk_usage = palloc0(sizeof(uint16) * max_chunks);
1002  entry->chunk_data =
1003  palloc0(sizeof(BlockRefTableChunk) * max_chunks);
1004  }
1005  else
1006  {
1007  entry->chunk_size = repalloc(entry->chunk_size,
1008  sizeof(uint16) * max_chunks);
1009  memset(&entry->chunk_size[entry->nchunks], 0,
1010  extra_chunks * sizeof(uint16));
1011  entry->chunk_usage = repalloc(entry->chunk_usage,
1012  sizeof(uint16) * max_chunks);
1013  memset(&entry->chunk_usage[entry->nchunks], 0,
1014  extra_chunks * sizeof(uint16));
1015  entry->chunk_data = repalloc(entry->chunk_data,
1016  sizeof(BlockRefTableChunk) * max_chunks);
1017  memset(&entry->chunk_data[entry->nchunks], 0,
1018  extra_chunks * sizeof(BlockRefTableChunk));
1019  }
1020  entry->nchunks = max_chunks;
1021  }
1022 
1023  /*
1024  * If the chunk that covers this block number doesn't exist yet, create it
1025  * as an array and add the appropriate offset to it. We make it pretty
1026  * small initially, because there might only be 1 or a few block
1027  * references in this chunk and we don't want to use up too much memory.
1028  */
1029  if (entry->chunk_size[chunkno] == 0)
1030  {
1031  entry->chunk_data[chunkno] =
1033  entry->chunk_size[chunkno] = INITIAL_ENTRIES_PER_CHUNK;
1034  entry->chunk_data[chunkno][0] = chunkoffset;
1035  entry->chunk_usage[chunkno] = 1;
1036  return;
1037  }
1038 
1039  /*
1040  * If the number of entries in this chunk is already maximum, it must be a
1041  * bitmap. Just set the appropriate bit.
1042  */
1043  if (entry->chunk_usage[chunkno] == MAX_ENTRIES_PER_CHUNK)
1044  {
1045  BlockRefTableChunk chunk = entry->chunk_data[chunkno];
1046 
1047  chunk[chunkoffset / BLOCKS_PER_ENTRY] |=
1048  1 << (chunkoffset % BLOCKS_PER_ENTRY);
1049  return;
1050  }
1051 
1052  /*
1053  * There is an existing chunk and it's in array format. Let's find out
1054  * whether it already has an entry for this block. If so, we do not need
1055  * to do anything.
1056  */
1057  for (i = 0; i < entry->chunk_usage[chunkno]; ++i)
1058  {
1059  if (entry->chunk_data[chunkno][i] == chunkoffset)
1060  return;
1061  }
1062 
1063  /*
1064  * If the number of entries currently used is one less than the maximum,
1065  * it's time to convert to bitmap format.
1066  */
1067  if (entry->chunk_usage[chunkno] == MAX_ENTRIES_PER_CHUNK - 1)
1068  {
1069  BlockRefTableChunk newchunk;
1070  unsigned j;
1071 
1072  /* Allocate a new chunk. */
1073  newchunk = palloc0(MAX_ENTRIES_PER_CHUNK * sizeof(uint16));
1074 
1075  /* Set the bit for each existing entry. */
1076  for (j = 0; j < entry->chunk_usage[chunkno]; ++j)
1077  {
1078  unsigned coff = entry->chunk_data[chunkno][j];
1079 
1080  newchunk[coff / BLOCKS_PER_ENTRY] |=
1081  1 << (coff % BLOCKS_PER_ENTRY);
1082  }
1083 
1084  /* Set the bit for the new entry. */
1085  newchunk[chunkoffset / BLOCKS_PER_ENTRY] |=
1086  1 << (chunkoffset % BLOCKS_PER_ENTRY);
1087 
1088  /* Swap the new chunk into place and update metadata. */
1089  pfree(entry->chunk_data[chunkno]);
1090  entry->chunk_data[chunkno] = newchunk;
1091  entry->chunk_size[chunkno] = MAX_ENTRIES_PER_CHUNK;
1092  entry->chunk_usage[chunkno] = MAX_ENTRIES_PER_CHUNK;
1093  return;
1094  }
1095 
1096  /*
1097  * OK, we currently have an array, and we don't need to convert to a
1098  * bitmap, but we do need to add a new element. If there's not enough
1099  * room, we'll have to expand the array.
1100  */
1101  if (entry->chunk_usage[chunkno] == entry->chunk_size[chunkno])
1102  {
1103  unsigned newsize = entry->chunk_size[chunkno] * 2;
1104 
1105  Assert(newsize <= MAX_ENTRIES_PER_CHUNK);
1106  entry->chunk_data[chunkno] = repalloc(entry->chunk_data[chunkno],
1107  newsize * sizeof(uint16));
1108  entry->chunk_size[chunkno] = newsize;
1109  }
1110 
1111  /* Now we can add the new entry. */
1112  entry->chunk_data[chunkno][entry->chunk_usage[chunkno]] =
1113  chunkoffset;
1114  entry->chunk_usage[chunkno]++;
1115 }
1116 
1117 /*
1118  * Release memory for a BlockRefTableEntry that was created by
1119  * CreateBlockRefTableEntry.
1120  */
1121 void
1123 {
1124  if (entry->chunk_size != NULL)
1125  {
1126  pfree(entry->chunk_size);
1127  entry->chunk_size = NULL;
1128  }
1129 
1130  if (entry->chunk_usage != NULL)
1131  {
1132  pfree(entry->chunk_usage);
1133  entry->chunk_usage = NULL;
1134  }
1135 
1136  if (entry->chunk_data != NULL)
1137  {
1138  pfree(entry->chunk_data);
1139  entry->chunk_data = NULL;
1140  }
1141 
1142  pfree(entry);
1143 }
1144 
1145 /*
1146  * Comparator for BlockRefTableSerializedEntry objects.
1147  *
1148  * We make the tablespace OID the first column of the sort key to match
1149  * the on-disk tree structure.
1150  */
1151 static int
1152 BlockRefTableComparator(const void *a, const void *b)
1153 {
1155  const BlockRefTableSerializedEntry *sb = b;
1156 
1157  if (sa->rlocator.spcOid > sb->rlocator.spcOid)
1158  return 1;
1159  if (sa->rlocator.spcOid < sb->rlocator.spcOid)
1160  return -1;
1161 
1162  if (sa->rlocator.dbOid > sb->rlocator.dbOid)
1163  return 1;
1164  if (sa->rlocator.dbOid < sb->rlocator.dbOid)
1165  return -1;
1166 
1167  if (sa->rlocator.relNumber > sb->rlocator.relNumber)
1168  return 1;
1169  if (sa->rlocator.relNumber < sb->rlocator.relNumber)
1170  return -1;
1171 
1172  if (sa->forknum > sb->forknum)
1173  return 1;
1174  if (sa->forknum < sb->forknum)
1175  return -1;
1176 
1177  return 0;
1178 }
1179 
1180 /*
1181  * Flush any buffered data out of a BlockRefTableBuffer.
1182  */
1183 static void
1185 {
1186  buffer->io_callback(buffer->io_callback_arg, buffer->data, buffer->used);
1187  buffer->used = 0;
1188 }
1189 
1190 /*
1191  * Read data from a BlockRefTableBuffer, and update the running CRC
1192  * calculation for the returned data (but not any data that we may have
1193  * buffered but not yet actually returned).
1194  */
1195 static void
1196 BlockRefTableRead(BlockRefTableReader *reader, void *data, int length)
1197 {
1198  BlockRefTableBuffer *buffer = &reader->buffer;
1199 
1200  /* Loop until read is fully satisfied. */
1201  while (length > 0)
1202  {
1203  if (buffer->cursor < buffer->used)
1204  {
1205  /*
1206  * If any buffered data is available, use that to satisfy as much
1207  * of the request as possible.
1208  */
1209  int bytes_to_copy = Min(length, buffer->used - buffer->cursor);
1210 
1211  memcpy(data, &buffer->data[buffer->cursor], bytes_to_copy);
1212  COMP_CRC32C(buffer->crc, &buffer->data[buffer->cursor],
1213  bytes_to_copy);
1214  buffer->cursor += bytes_to_copy;
1215  data = ((char *) data) + bytes_to_copy;
1216  length -= bytes_to_copy;
1217  }
1218  else if (length >= BUFSIZE)
1219  {
1220  /*
1221  * If the request length is long, read directly into caller's
1222  * buffer.
1223  */
1224  int bytes_read;
1225 
1226  bytes_read = buffer->io_callback(buffer->io_callback_arg,
1227  data, length);
1228  COMP_CRC32C(buffer->crc, data, bytes_read);
1229  data = ((char *) data) + bytes_read;
1230  length -= bytes_read;
1231 
1232  /* If we didn't get anything, that's bad. */
1233  if (bytes_read == 0)
1234  reader->error_callback(reader->error_callback_arg,
1235  "file \"%s\" ends unexpectedly",
1236  reader->error_filename);
1237  }
1238  else
1239  {
1240  /*
1241  * Refill our buffer.
1242  */
1243  buffer->used = buffer->io_callback(buffer->io_callback_arg,
1244  buffer->data, BUFSIZE);
1245  buffer->cursor = 0;
1246 
1247  /* If we didn't get anything, that's bad. */
1248  if (buffer->used == 0)
1249  reader->error_callback(reader->error_callback_arg,
1250  "file \"%s\" ends unexpectedly",
1251  reader->error_filename);
1252  }
1253  }
1254 }
1255 
1256 /*
1257  * Supply data to a BlockRefTableBuffer for write to the underlying File,
1258  * and update the running CRC calculation for that data.
1259  */
1260 static void
1261 BlockRefTableWrite(BlockRefTableBuffer *buffer, void *data, int length)
1262 {
1263  /* Update running CRC calculation. */
1264  COMP_CRC32C(buffer->crc, data, length);
1265 
1266  /* If the new data can't fit into the buffer, flush the buffer. */
1267  if (buffer->used + length > BUFSIZE)
1268  {
1269  buffer->io_callback(buffer->io_callback_arg, buffer->data,
1270  buffer->used);
1271  buffer->used = 0;
1272  }
1273 
1274  /* If the new data would fill the buffer, or more, write it directly. */
1275  if (length >= BUFSIZE)
1276  {
1277  buffer->io_callback(buffer->io_callback_arg, data, length);
1278  return;
1279  }
1280 
1281  /* Otherwise, copy the new data into the buffer. */
1282  memcpy(&buffer->data[buffer->used], data, length);
1283  buffer->used += length;
1284  Assert(buffer->used <= BUFSIZE);
1285 }
1286 
1287 /*
1288  * Generate the sentinel and CRC required at the end of a block reference
1289  * table file and flush them out of our internal buffer.
1290  */
1291 static void
1293 {
1294  BlockRefTableSerializedEntry zentry = {{0}};
1295  pg_crc32c crc;
1296 
1297  /* Write a sentinel indicating that there are no more entries. */
1298  BlockRefTableWrite(buffer, &zentry,
1300 
1301  /*
1302  * Writing the checksum will perturb the ongoing checksum calculation, so
1303  * copy the state first and finalize the computation using the copy.
1304  */
1305  crc = buffer->crc;
1306  FIN_CRC32C(crc);
1307  BlockRefTableWrite(buffer, &crc, sizeof(pg_crc32c));
1308 
1309  /* Flush any leftover data out of our buffer. */
1310  BlockRefTableFlush(buffer);
1311 }
void BlockRefTableFreeEntry(BlockRefTableEntry *entry)
Definition: blkreftable.c:1122
struct BlockRefTableSerializedEntry BlockRefTableSerializedEntry
BlockRefTableWriter * CreateBlockRefTableWriter(io_callback_fn write_callback, void *write_callback_arg)
Definition: blkreftable.c:790
BlockRefTable * CreateEmptyBlockRefTable(void)
Definition: blkreftable.c:235
BlockRefTableReader * CreateBlockRefTableReader(io_callback_fn read_callback, void *read_callback_arg, char *error_filename, report_error_fn error_callback, void *error_callback_arg)
Definition: blkreftable.c:577
bool BlockRefTableReaderNextRelation(BlockRefTableReader *reader, RelFileLocator *rlocator, ForkNumber *forknum, BlockNumber *limit_block)
Definition: blkreftable.c:613
int BlockRefTableEntryGetBlocks(BlockRefTableEntry *entry, BlockNumber start_blkno, BlockNumber stop_blkno, BlockNumber *blocks, int nblocks)
Definition: blkreftable.c:369
void BlockRefTableMarkBlockModified(BlockRefTable *brtab, const RelFileLocator *rlocator, ForkNumber forknum, BlockNumber blknum)
Definition: blkreftable.c:297
#define BLOCKS_PER_CHUNK
Definition: blkreftable.c:78
void BlockRefTableWriteEntry(BlockRefTableWriter *writer, BlockRefTableEntry *entry)
Definition: blkreftable.c:817
static void BlockRefTableRead(BlockRefTableReader *reader, void *data, int length)
Definition: blkreftable.c:1196
BlockRefTableEntry * BlockRefTableGetEntry(BlockRefTable *brtab, const RelFileLocator *rlocator, ForkNumber forknum, BlockNumber *limit_block)
Definition: blkreftable.c:340
void BlockRefTableEntryMarkBlockModified(BlockRefTableEntry *entry, ForkNumber forknum, BlockNumber blknum)
Definition: blkreftable.c:965
BlockRefTableEntry * CreateBlockRefTableEntry(RelFileLocator rlocator, ForkNumber forknum)
Definition: blkreftable.c:875
unsigned BlockRefTableReaderGetBlocks(BlockRefTableReader *reader, BlockNumber *blocks, int nblocks)
Definition: blkreftable.c:689
#define BLOCKS_PER_ENTRY
Definition: blkreftable.c:79
void BlockRefTableSetLimitBlock(BlockRefTable *brtab, const RelFileLocator *rlocator, ForkNumber forknum, BlockNumber limit_block)
Definition: blkreftable.c:262
void BlockRefTableEntrySetLimitBlock(BlockRefTableEntry *entry, BlockNumber limit_block)
Definition: blkreftable.c:894
struct BlockRefTableBuffer BlockRefTableBuffer
void WriteBlockRefTable(BlockRefTable *brtab, io_callback_fn write_callback, void *write_callback_arg)
Definition: blkreftable.c:474
static void BlockRefTableFileTerminate(BlockRefTableBuffer *buffer)
Definition: blkreftable.c:1292
void DestroyBlockRefTableReader(BlockRefTableReader *reader)
Definition: blkreftable.c:773
#define MAX_ENTRIES_PER_CHUNK
Definition: blkreftable.c:80
void DestroyBlockRefTableWriter(BlockRefTableWriter *writer)
Definition: blkreftable.c:855
#define INITIAL_ENTRIES_PER_CHUNK
Definition: blkreftable.c:81
#define BUFSIZE
Definition: blkreftable.c:166
uint16 * BlockRefTableChunk
Definition: blkreftable.c:82
static void BlockRefTableWrite(BlockRefTableBuffer *buffer, void *data, int length)
Definition: blkreftable.c:1261
static void BlockRefTableFlush(BlockRefTableBuffer *buffer)
Definition: blkreftable.c:1184
struct BlockRefTableKey BlockRefTableKey
static int BlockRefTableComparator(const void *a, const void *b)
Definition: blkreftable.c:1152
void(* report_error_fn)(void *calblack_arg, char *msg,...) pg_attribute_printf(2
Definition: blkreftable.h:47
int(* io_callback_fn)(void *callback_arg, void *data, int length)
Definition: blkreftable.h:46
#define BLOCKREFTABLE_MAGIC
Definition: blkreftable.h:30
uint32 BlockNumber
Definition: block.h:31
#define InvalidBlockNumber
Definition: block.h:33
unsigned short uint16
Definition: c.h:505
unsigned int uint32
Definition: c.h:506
#define Min(x, y)
Definition: c.h:1004
#define Max(x, y)
Definition: c.h:998
#define Assert(condition)
Definition: c.h:858
uint64 chunk
int b
Definition: isn.c:70
int a
Definition: isn.c:69
int j
Definition: isn.c:74
int i
Definition: isn.c:73
if(TABLE==NULL||TABLE_index==NULL)
Definition: isn.c:77
void pfree(void *pointer)
Definition: mcxt.c:1520
void * palloc0(Size size)
Definition: mcxt.c:1346
MemoryContext CurrentMemoryContext
Definition: mcxt.c:143
void * repalloc(void *pointer, Size size)
Definition: mcxt.c:1540
void * palloc(Size size)
Definition: mcxt.c:1316
uint32 pg_crc32c
Definition: pg_crc32c.h:38
#define COMP_CRC32C(crc, data, len)
Definition: pg_crc32c.h:98
#define EQ_CRC32C(c1, c2)
Definition: pg_crc32c.h:42
#define INIT_CRC32C(crc)
Definition: pg_crc32c.h:41
#define FIN_CRC32C(crc)
Definition: pg_crc32c.h:103
const void * data
return crc
#define qsort(a, b, c, d)
Definition: port.h:449
MemoryContextSwitchTo(old_ctx)
ForkNumber
Definition: relpath.h:48
io_callback_fn io_callback
Definition: blkreftable.c:173
char data[BUFSIZE]
Definition: blkreftable.c:175
BlockRefTableKey key
Definition: blkreftable.c:112
uint16 * chunk_usage
Definition: blkreftable.c:117
BlockNumber limit_block
Definition: blkreftable.c:113
BlockRefTableChunk * chunk_data
Definition: blkreftable.c:118
RelFileLocator rlocator
Definition: blkreftable.c:49
ForkNumber forknum
Definition: blkreftable.c:50
BlockRefTableBuffer buffer
Definition: blkreftable.c:202
uint16 chunk_data[MAX_ENTRIES_PER_CHUNK]
Definition: blkreftable.c:209
report_error_fn error_callback
Definition: blkreftable.c:204
BlockRefTableBuffer buffer
Definition: blkreftable.c:219
blockreftable_hash * hash
Definition: blkreftable.c:146
MemoryContext mcxt
Definition: blkreftable.c:148
RelFileNumber relNumber