Line data Source code
1 : /*
2 : * contrib/pgstattuple/pgstattuple.c
3 : *
4 : * Copyright (c) 2001,2002 Tatsuo Ishii
5 : *
6 : * Permission to use, copy, modify, and distribute this software and
7 : * its documentation for any purpose, without fee, and without a
8 : * written agreement is hereby granted, provided that the above
9 : * copyright notice and this paragraph and the following two
10 : * paragraphs appear in all copies.
11 : *
12 : * IN NO EVENT SHALL THE AUTHOR BE LIABLE TO ANY PARTY FOR DIRECT,
13 : * INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, INCLUDING
14 : * LOST PROFITS, ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS
15 : * DOCUMENTATION, EVEN IF THE UNIVERSITY OF CALIFORNIA HAS BEEN ADVISED
16 : * OF THE POSSIBILITY OF SUCH DAMAGE.
17 : *
18 : * THE AUTHOR SPECIFICALLY DISCLAIMS ANY WARRANTIES, INCLUDING, BUT NOT
19 : * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
20 : * A PARTICULAR PURPOSE. THE SOFTWARE PROVIDED HEREUNDER IS ON AN "AS
21 : * IS" BASIS, AND THE AUTHOR HAS NO OBLIGATIONS TO PROVIDE MAINTENANCE,
22 : * SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
23 : */
24 :
25 : #include "postgres.h"
26 :
27 : #include "access/gist_private.h"
28 : #include "access/hash.h"
29 : #include "access/heapam.h"
30 : #include "access/nbtree.h"
31 : #include "access/relscan.h"
32 : #include "access/tableam.h"
33 : #include "catalog/namespace.h"
34 : #include "catalog/pg_am_d.h"
35 : #include "funcapi.h"
36 : #include "miscadmin.h"
37 : #include "storage/bufmgr.h"
38 : #include "storage/lmgr.h"
39 : #include "utils/builtins.h"
40 : #include "utils/varlena.h"
41 :
42 2 : PG_MODULE_MAGIC;
43 :
44 2 : PG_FUNCTION_INFO_V1(pgstattuple);
45 4 : PG_FUNCTION_INFO_V1(pgstattuple_v1_5);
46 2 : PG_FUNCTION_INFO_V1(pgstattuplebyid);
47 4 : PG_FUNCTION_INFO_V1(pgstattuplebyid_v1_5);
48 :
49 : /*
50 : * struct pgstattuple_type
51 : *
52 : * tuple_percent, dead_tuple_percent and free_percent are computable,
53 : * so not defined here.
54 : */
55 : typedef struct pgstattuple_type
56 : {
57 : uint64 table_len;
58 : uint64 tuple_count;
59 : uint64 tuple_len;
60 : uint64 dead_tuple_count;
61 : uint64 dead_tuple_len;
62 : uint64 free_space; /* free/reusable space in bytes */
63 : } pgstattuple_type;
64 :
65 : typedef void (*pgstat_page) (pgstattuple_type *, Relation, BlockNumber,
66 : BufferAccessStrategy);
67 :
68 : static Datum build_pgstattuple_type(pgstattuple_type *stat,
69 : FunctionCallInfo fcinfo);
70 : static Datum pgstat_relation(Relation rel, FunctionCallInfo fcinfo);
71 : static Datum pgstat_heap(Relation rel, FunctionCallInfo fcinfo);
72 : static void pgstat_btree_page(pgstattuple_type *stat,
73 : Relation rel, BlockNumber blkno,
74 : BufferAccessStrategy bstrategy);
75 : static void pgstat_hash_page(pgstattuple_type *stat,
76 : Relation rel, BlockNumber blkno,
77 : BufferAccessStrategy bstrategy);
78 : static void pgstat_gist_page(pgstattuple_type *stat,
79 : Relation rel, BlockNumber blkno,
80 : BufferAccessStrategy bstrategy);
81 : static Datum pgstat_index(Relation rel, BlockNumber start,
82 : pgstat_page pagefn, FunctionCallInfo fcinfo);
83 : static void pgstat_index_page(pgstattuple_type *stat, Page page,
84 : OffsetNumber minoff, OffsetNumber maxoff);
85 :
86 : /*
87 : * build_pgstattuple_type -- build a pgstattuple_type tuple
88 : */
89 : static Datum
90 16 : build_pgstattuple_type(pgstattuple_type *stat, FunctionCallInfo fcinfo)
91 : {
92 : #define NCOLUMNS 9
93 : #define NCHARS 314
94 :
95 : HeapTuple tuple;
96 : char *values[NCOLUMNS];
97 : char values_buf[NCOLUMNS][NCHARS];
98 : int i;
99 : double tuple_percent;
100 : double dead_tuple_percent;
101 : double free_percent; /* free/reusable space in % */
102 : TupleDesc tupdesc;
103 : AttInMetadata *attinmeta;
104 :
105 : /* Build a tuple descriptor for our result type */
106 16 : if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
107 0 : elog(ERROR, "return type must be a row type");
108 :
109 : /*
110 : * Generate attribute metadata needed later to produce tuples from raw C
111 : * strings
112 : */
113 16 : attinmeta = TupleDescGetAttInMetadata(tupdesc);
114 :
115 16 : if (stat->table_len == 0)
116 : {
117 16 : tuple_percent = 0.0;
118 16 : dead_tuple_percent = 0.0;
119 16 : free_percent = 0.0;
120 : }
121 : else
122 : {
123 0 : tuple_percent = 100.0 * stat->tuple_len / stat->table_len;
124 0 : dead_tuple_percent = 100.0 * stat->dead_tuple_len / stat->table_len;
125 0 : free_percent = 100.0 * stat->free_space / stat->table_len;
126 : }
127 :
128 : /*
129 : * Prepare a values array for constructing the tuple. This should be an
130 : * array of C strings which will be processed later by the appropriate
131 : * "in" functions.
132 : */
133 160 : for (i = 0; i < NCOLUMNS; i++)
134 144 : values[i] = values_buf[i];
135 16 : i = 0;
136 16 : snprintf(values[i++], NCHARS, INT64_FORMAT, stat->table_len);
137 16 : snprintf(values[i++], NCHARS, INT64_FORMAT, stat->tuple_count);
138 16 : snprintf(values[i++], NCHARS, INT64_FORMAT, stat->tuple_len);
139 16 : snprintf(values[i++], NCHARS, "%.2f", tuple_percent);
140 16 : snprintf(values[i++], NCHARS, INT64_FORMAT, stat->dead_tuple_count);
141 16 : snprintf(values[i++], NCHARS, INT64_FORMAT, stat->dead_tuple_len);
142 16 : snprintf(values[i++], NCHARS, "%.2f", dead_tuple_percent);
143 16 : snprintf(values[i++], NCHARS, INT64_FORMAT, stat->free_space);
144 16 : snprintf(values[i++], NCHARS, "%.2f", free_percent);
145 :
146 : /* build a tuple */
147 16 : tuple = BuildTupleFromCStrings(attinmeta, values);
148 :
149 : /* make the tuple into a datum */
150 16 : return HeapTupleGetDatum(tuple);
151 : }
152 :
153 : /* ----------
154 : * pgstattuple:
155 : * returns live/dead tuples info
156 : *
157 : * C FUNCTION definition
158 : * pgstattuple(text) returns pgstattuple_type
159 : *
160 : * The superuser() check here must be kept as the library might be upgraded
161 : * without the extension being upgraded, meaning that in pre-1.5 installations
162 : * these functions could be called by any user.
163 : * ----------
164 : */
165 :
166 : Datum
167 0 : pgstattuple(PG_FUNCTION_ARGS)
168 : {
169 0 : text *relname = PG_GETARG_TEXT_PP(0);
170 : RangeVar *relrv;
171 : Relation rel;
172 :
173 0 : if (!superuser())
174 0 : ereport(ERROR,
175 : (errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
176 : errmsg("must be superuser to use pgstattuple functions")));
177 :
178 : /* open relation */
179 0 : relrv = makeRangeVarFromNameList(textToQualifiedNameList(relname));
180 0 : rel = relation_openrv(relrv, AccessShareLock);
181 :
182 0 : PG_RETURN_DATUM(pgstat_relation(rel, fcinfo));
183 : }
184 :
185 : /*
186 : * As of pgstattuple version 1.5, we no longer need to check if the user
187 : * is a superuser because we REVOKE EXECUTE on the function from PUBLIC.
188 : * Users can then grant access to it based on their policies.
189 : *
190 : * Otherwise identical to pgstattuple (above).
191 : */
192 : Datum
193 18 : pgstattuple_v1_5(PG_FUNCTION_ARGS)
194 : {
195 18 : text *relname = PG_GETARG_TEXT_PP(0);
196 : RangeVar *relrv;
197 : Relation rel;
198 :
199 : /* open relation */
200 18 : relrv = makeRangeVarFromNameList(textToQualifiedNameList(relname));
201 18 : rel = relation_openrv(relrv, AccessShareLock);
202 :
203 18 : PG_RETURN_DATUM(pgstat_relation(rel, fcinfo));
204 : }
205 :
206 : /* Must keep superuser() check, see above. */
207 : Datum
208 0 : pgstattuplebyid(PG_FUNCTION_ARGS)
209 : {
210 0 : Oid relid = PG_GETARG_OID(0);
211 : Relation rel;
212 :
213 0 : if (!superuser())
214 0 : ereport(ERROR,
215 : (errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
216 : errmsg("must be superuser to use pgstattuple functions")));
217 :
218 : /* open relation */
219 0 : rel = relation_open(relid, AccessShareLock);
220 :
221 0 : PG_RETURN_DATUM(pgstat_relation(rel, fcinfo));
222 : }
223 :
224 : /* Remove superuser() check for 1.5 version, see above */
225 : Datum
226 6 : pgstattuplebyid_v1_5(PG_FUNCTION_ARGS)
227 : {
228 6 : Oid relid = PG_GETARG_OID(0);
229 : Relation rel;
230 :
231 : /* open relation */
232 6 : rel = relation_open(relid, AccessShareLock);
233 :
234 6 : PG_RETURN_DATUM(pgstat_relation(rel, fcinfo));
235 : }
236 :
237 : /*
238 : * pgstat_relation
239 : */
240 : static Datum
241 24 : pgstat_relation(Relation rel, FunctionCallInfo fcinfo)
242 : {
243 : const char *err;
244 :
245 : /*
246 : * Reject attempts to read non-local temporary relations; we would be
247 : * likely to get wrong data since we have no visibility into the owning
248 : * session's local buffers.
249 : */
250 24 : if (RELATION_IS_OTHER_TEMP(rel))
251 0 : ereport(ERROR,
252 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
253 : errmsg("cannot access temporary tables of other sessions")));
254 :
255 24 : if (RELKIND_HAS_TABLE_AM(rel->rd_rel->relkind) ||
256 8 : rel->rd_rel->relkind == RELKIND_SEQUENCE)
257 : {
258 16 : return pgstat_heap(rel, fcinfo);
259 : }
260 8 : else if (rel->rd_rel->relkind == RELKIND_INDEX)
261 : {
262 0 : switch (rel->rd_rel->relam)
263 : {
264 0 : case BTREE_AM_OID:
265 0 : return pgstat_index(rel, BTREE_METAPAGE + 1,
266 : pgstat_btree_page, fcinfo);
267 0 : case HASH_AM_OID:
268 0 : return pgstat_index(rel, HASH_METAPAGE + 1,
269 : pgstat_hash_page, fcinfo);
270 0 : case GIST_AM_OID:
271 0 : return pgstat_index(rel, GIST_ROOT_BLKNO + 1,
272 : pgstat_gist_page, fcinfo);
273 0 : case GIN_AM_OID:
274 0 : err = "gin index";
275 0 : break;
276 0 : case SPGIST_AM_OID:
277 0 : err = "spgist index";
278 0 : break;
279 0 : case BRIN_AM_OID:
280 0 : err = "brin index";
281 0 : break;
282 0 : default:
283 0 : err = "unknown index";
284 0 : break;
285 : }
286 0 : ereport(ERROR,
287 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
288 : errmsg("index \"%s\" (%s) is not supported",
289 : RelationGetRelationName(rel), err)));
290 : }
291 : else
292 : {
293 8 : ereport(ERROR,
294 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
295 : errmsg("cannot get tuple-level statistics for relation \"%s\"",
296 : RelationGetRelationName(rel)),
297 : errdetail_relkind_not_supported(rel->rd_rel->relkind)));
298 : }
299 :
300 : return 0; /* should not happen */
301 : }
302 :
303 : /*
304 : * pgstat_heap -- returns live/dead tuples info in a heap
305 : */
306 : static Datum
307 16 : pgstat_heap(Relation rel, FunctionCallInfo fcinfo)
308 : {
309 : TableScanDesc scan;
310 : HeapScanDesc hscan;
311 : HeapTuple tuple;
312 : BlockNumber nblocks;
313 16 : BlockNumber block = 0; /* next block to count free space in */
314 : BlockNumber tupblock;
315 : Buffer buffer;
316 16 : pgstattuple_type stat = {0};
317 : SnapshotData SnapshotDirty;
318 :
319 16 : if (rel->rd_rel->relam != HEAP_TABLE_AM_OID)
320 0 : ereport(ERROR,
321 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
322 : errmsg("only heap AM is supported")));
323 :
324 : /* Disable syncscan because we assume we scan from block zero upwards */
325 16 : scan = table_beginscan_strat(rel, SnapshotAny, 0, NULL, true, false);
326 16 : hscan = (HeapScanDesc) scan;
327 :
328 16 : InitDirtySnapshot(SnapshotDirty);
329 :
330 16 : nblocks = hscan->rs_nblocks; /* # blocks to be scanned */
331 :
332 : /* scan the relation */
333 16 : while ((tuple = heap_getnext(scan, ForwardScanDirection)) != NULL)
334 : {
335 0 : CHECK_FOR_INTERRUPTS();
336 :
337 : /* must hold a buffer lock to call HeapTupleSatisfiesVisibility */
338 0 : LockBuffer(hscan->rs_cbuf, BUFFER_LOCK_SHARE);
339 :
340 0 : if (HeapTupleSatisfiesVisibility(tuple, &SnapshotDirty, hscan->rs_cbuf))
341 : {
342 0 : stat.tuple_len += tuple->t_len;
343 0 : stat.tuple_count++;
344 : }
345 : else
346 : {
347 0 : stat.dead_tuple_len += tuple->t_len;
348 0 : stat.dead_tuple_count++;
349 : }
350 :
351 0 : LockBuffer(hscan->rs_cbuf, BUFFER_LOCK_UNLOCK);
352 :
353 : /*
354 : * To avoid physically reading the table twice, try to do the
355 : * free-space scan in parallel with the heap scan. However,
356 : * heap_getnext may find no tuples on a given page, so we cannot
357 : * simply examine the pages returned by the heap scan.
358 : */
359 0 : tupblock = ItemPointerGetBlockNumber(&tuple->t_self);
360 :
361 0 : while (block <= tupblock)
362 : {
363 0 : CHECK_FOR_INTERRUPTS();
364 :
365 0 : buffer = ReadBufferExtended(rel, MAIN_FORKNUM, block,
366 : RBM_NORMAL, hscan->rs_strategy);
367 0 : LockBuffer(buffer, BUFFER_LOCK_SHARE);
368 0 : stat.free_space += PageGetHeapFreeSpace((Page) BufferGetPage(buffer));
369 0 : UnlockReleaseBuffer(buffer);
370 0 : block++;
371 : }
372 : }
373 :
374 16 : while (block < nblocks)
375 : {
376 0 : CHECK_FOR_INTERRUPTS();
377 :
378 0 : buffer = ReadBufferExtended(rel, MAIN_FORKNUM, block,
379 : RBM_NORMAL, hscan->rs_strategy);
380 0 : LockBuffer(buffer, BUFFER_LOCK_SHARE);
381 0 : stat.free_space += PageGetHeapFreeSpace((Page) BufferGetPage(buffer));
382 0 : UnlockReleaseBuffer(buffer);
383 0 : block++;
384 : }
385 :
386 16 : table_endscan(scan);
387 16 : relation_close(rel, AccessShareLock);
388 :
389 16 : stat.table_len = (uint64) nblocks * BLCKSZ;
390 :
391 16 : return build_pgstattuple_type(&stat, fcinfo);
392 : }
393 :
394 : /*
395 : * pgstat_btree_page -- check tuples in a btree page
396 : */
397 : static void
398 0 : pgstat_btree_page(pgstattuple_type *stat, Relation rel, BlockNumber blkno,
399 : BufferAccessStrategy bstrategy)
400 : {
401 : Buffer buf;
402 : Page page;
403 :
404 0 : buf = ReadBufferExtended(rel, MAIN_FORKNUM, blkno, RBM_NORMAL, bstrategy);
405 0 : LockBuffer(buf, BT_READ);
406 0 : page = BufferGetPage(buf);
407 :
408 : /* Page is valid, see what to do with it */
409 0 : if (PageIsNew(page))
410 : {
411 : /* fully empty page */
412 0 : stat->free_space += BLCKSZ;
413 : }
414 : else
415 : {
416 : BTPageOpaque opaque;
417 :
418 0 : opaque = BTPageGetOpaque(page);
419 0 : if (P_IGNORE(opaque))
420 : {
421 : /* deleted or half-dead page */
422 0 : stat->free_space += BLCKSZ;
423 : }
424 0 : else if (P_ISLEAF(opaque))
425 : {
426 0 : pgstat_index_page(stat, page, P_FIRSTDATAKEY(opaque),
427 0 : PageGetMaxOffsetNumber(page));
428 : }
429 : else
430 : {
431 : /* internal page */
432 : }
433 : }
434 :
435 0 : _bt_relbuf(rel, buf);
436 0 : }
437 :
438 : /*
439 : * pgstat_hash_page -- check tuples in a hash page
440 : */
441 : static void
442 0 : pgstat_hash_page(pgstattuple_type *stat, Relation rel, BlockNumber blkno,
443 : BufferAccessStrategy bstrategy)
444 : {
445 : Buffer buf;
446 : Page page;
447 :
448 0 : buf = _hash_getbuf_with_strategy(rel, blkno, HASH_READ, 0, bstrategy);
449 0 : page = BufferGetPage(buf);
450 :
451 0 : if (PageGetSpecialSize(page) == MAXALIGN(sizeof(HashPageOpaqueData)))
452 : {
453 : HashPageOpaque opaque;
454 :
455 0 : opaque = HashPageGetOpaque(page);
456 0 : switch (opaque->hasho_flag & LH_PAGE_TYPE)
457 : {
458 0 : case LH_UNUSED_PAGE:
459 0 : stat->free_space += BLCKSZ;
460 0 : break;
461 0 : case LH_BUCKET_PAGE:
462 : case LH_OVERFLOW_PAGE:
463 0 : pgstat_index_page(stat, page, FirstOffsetNumber,
464 0 : PageGetMaxOffsetNumber(page));
465 0 : break;
466 0 : case LH_BITMAP_PAGE:
467 : case LH_META_PAGE:
468 : default:
469 0 : break;
470 : }
471 : }
472 : else
473 : {
474 : /* maybe corrupted */
475 : }
476 :
477 0 : _hash_relbuf(rel, buf);
478 0 : }
479 :
480 : /*
481 : * pgstat_gist_page -- check tuples in a gist page
482 : */
483 : static void
484 0 : pgstat_gist_page(pgstattuple_type *stat, Relation rel, BlockNumber blkno,
485 : BufferAccessStrategy bstrategy)
486 : {
487 : Buffer buf;
488 : Page page;
489 :
490 0 : buf = ReadBufferExtended(rel, MAIN_FORKNUM, blkno, RBM_NORMAL, bstrategy);
491 0 : LockBuffer(buf, GIST_SHARE);
492 0 : gistcheckpage(rel, buf);
493 0 : page = BufferGetPage(buf);
494 :
495 0 : if (GistPageIsLeaf(page))
496 : {
497 0 : pgstat_index_page(stat, page, FirstOffsetNumber,
498 0 : PageGetMaxOffsetNumber(page));
499 : }
500 : else
501 : {
502 : /* root or node */
503 : }
504 :
505 0 : UnlockReleaseBuffer(buf);
506 0 : }
507 :
508 : /*
509 : * pgstat_index -- returns live/dead tuples info in a generic index
510 : */
511 : static Datum
512 0 : pgstat_index(Relation rel, BlockNumber start, pgstat_page pagefn,
513 : FunctionCallInfo fcinfo)
514 : {
515 : BlockNumber nblocks;
516 : BlockNumber blkno;
517 : BufferAccessStrategy bstrategy;
518 0 : pgstattuple_type stat = {0};
519 :
520 : /* prepare access strategy for this index */
521 0 : bstrategy = GetAccessStrategy(BAS_BULKREAD);
522 :
523 0 : blkno = start;
524 : for (;;)
525 : {
526 : /* Get the current relation length */
527 0 : LockRelationForExtension(rel, ExclusiveLock);
528 0 : nblocks = RelationGetNumberOfBlocks(rel);
529 0 : UnlockRelationForExtension(rel, ExclusiveLock);
530 :
531 : /* Quit if we've scanned the whole relation */
532 0 : if (blkno >= nblocks)
533 : {
534 0 : stat.table_len = (uint64) nblocks * BLCKSZ;
535 :
536 0 : break;
537 : }
538 :
539 0 : for (; blkno < nblocks; blkno++)
540 : {
541 0 : CHECK_FOR_INTERRUPTS();
542 :
543 0 : pagefn(&stat, rel, blkno, bstrategy);
544 : }
545 : }
546 :
547 0 : relation_close(rel, AccessShareLock);
548 :
549 0 : return build_pgstattuple_type(&stat, fcinfo);
550 : }
551 :
552 : /*
553 : * pgstat_index_page -- for generic index page
554 : */
555 : static void
556 0 : pgstat_index_page(pgstattuple_type *stat, Page page,
557 : OffsetNumber minoff, OffsetNumber maxoff)
558 : {
559 : OffsetNumber i;
560 :
561 0 : stat->free_space += PageGetFreeSpace(page);
562 :
563 0 : for (i = minoff; i <= maxoff; i = OffsetNumberNext(i))
564 : {
565 0 : ItemId itemid = PageGetItemId(page, i);
566 :
567 0 : if (ItemIdIsDead(itemid))
568 : {
569 0 : stat->dead_tuple_count++;
570 0 : stat->dead_tuple_len += ItemIdGetLength(itemid);
571 : }
572 : else
573 : {
574 0 : stat->tuple_count++;
575 0 : stat->tuple_len += ItemIdGetLength(itemid);
576 : }
577 : }
578 0 : }
|