PolarDB-for-PostgreSQL

Форк
0
477 строк · 12.1 Кб
1
/*-------------------------------------------------------------------------
2
 *
3
 * heapfuncs.c
4
 *	  Functions to investigate heap pages
5
 *
6
 * We check the input to these functions for corrupt pointers etc. that
7
 * might cause crashes, but at the same time we try to print out as much
8
 * information as possible, even if it's nonsense. That's because if a
9
 * page is corrupt, we don't know why and how exactly it is corrupt, so we
10
 * let the user judge it.
11
 *
12
 * These functions are restricted to superusers for the fear of introducing
13
 * security holes if the input checking isn't as water-tight as it should be.
14
 * You'd need to be superuser to obtain a raw page image anyway, so
15
 * there's hardly any use case for using these without superuser-rights
16
 * anyway.
17
 *
18
 * Copyright (c) 2007-2018, PostgreSQL Global Development Group
19
 *
20
 * IDENTIFICATION
21
 *	  contrib/pageinspect/heapfuncs.c
22
 *
23
 *-------------------------------------------------------------------------
24
 */
25

26
#include "postgres.h"
27

28
#include "pageinspect.h"
29

30
#include "access/htup_details.h"
31
#include "funcapi.h"
32
#include "catalog/pg_type.h"
33
#include "miscadmin.h"
34
#include "utils/array.h"
35
#include "utils/builtins.h"
36
#include "utils/rel.h"
37

38

39
/*
40
 * bits_to_text
41
 *
42
 * Converts a bits8-array of 'len' bits to a human-readable
43
 * c-string representation.
44
 */
45
static char *
46
bits_to_text(bits8 *bits, int len)
47
{
48
	int			i;
49
	char	   *str;
50

51
	str = palloc(len + 1);
52

53
	for (i = 0; i < len; i++)
54
		str[i] = (bits[(i / 8)] & (1 << (i % 8))) ? '1' : '0';
55

56
	str[i] = '\0';
57

58
	return str;
59
}
60

61

62
/*
63
 * text_to_bits
64
 *
65
 * Converts a c-string representation of bits into a bits8-array. This is
66
 * the reverse operation of previous routine.
67
 */
68
static bits8 *
69
text_to_bits(char *str, int len)
70
{
71
	bits8	   *bits;
72
	int			off = 0;
73
	char		byte = 0;
74

75
	bits = palloc(len + 1);
76

77
	while (off < len)
78
	{
79
		if (off % 8 == 0)
80
			byte = 0;
81

82
		if ((str[off] == '0') || (str[off] == '1'))
83
			byte = byte | ((str[off] - '0') << off % 8);
84
		else
85
			ereport(ERROR,
86
					(errcode(ERRCODE_DATA_CORRUPTED),
87
					 errmsg("illegal character '%c' in t_bits string", str[off])));
88

89
		if (off % 8 == 7)
90
			bits[off / 8] = byte;
91

92
		off++;
93
	}
94

95
	return bits;
96
}
97

98
/*
99
 * heap_page_items
100
 *
101
 * Allows inspection of line pointers and tuple headers of a heap page.
102
 */
103
PG_FUNCTION_INFO_V1(heap_page_items);
104

105
typedef struct heap_page_items_state
106
{
107
	TupleDesc	tupd;
108
	Page		page;
109
	uint16		offset;
110
} heap_page_items_state;
111

112
Datum
113
heap_page_items(PG_FUNCTION_ARGS)
114
{
115
	bytea	   *raw_page = PG_GETARG_BYTEA_P(0);
116
	heap_page_items_state *inter_call_data = NULL;
117
	FuncCallContext *fctx;
118
	int			raw_page_size;
119

120
	if (!superuser())
121
		ereport(ERROR,
122
				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
123
				 (errmsg("must be superuser to use raw page functions"))));
124

125
	raw_page_size = VARSIZE(raw_page) - VARHDRSZ;
126

127
	if (SRF_IS_FIRSTCALL())
128
	{
129
		TupleDesc	tupdesc;
130
		MemoryContext mctx;
131

132
		if (raw_page_size < SizeOfPageHeaderData)
133
			ereport(ERROR,
134
					(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
135
					 errmsg("input page too small (%d bytes)", raw_page_size)));
136

137
		fctx = SRF_FIRSTCALL_INIT();
138
		mctx = MemoryContextSwitchTo(fctx->multi_call_memory_ctx);
139

140
		inter_call_data = palloc(sizeof(heap_page_items_state));
141

142
		/* Build a tuple descriptor for our result type */
143
		if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
144
			elog(ERROR, "return type must be a row type");
145

146
		inter_call_data->tupd = tupdesc;
147

148
		inter_call_data->offset = FirstOffsetNumber;
149
		inter_call_data->page = VARDATA(raw_page);
150

151
		fctx->max_calls = PageGetMaxOffsetNumber(inter_call_data->page);
152
		fctx->user_fctx = inter_call_data;
153

154
		MemoryContextSwitchTo(mctx);
155
	}
156

157
	fctx = SRF_PERCALL_SETUP();
158
	inter_call_data = fctx->user_fctx;
159

160
	if (fctx->call_cntr < fctx->max_calls)
161
	{
162
		Page		page = inter_call_data->page;
163
		HeapTuple	resultTuple;
164
		Datum		result;
165
		ItemId		id;
166
		Datum		values[14];
167
		bool		nulls[14];
168
		uint16		lp_offset;
169
		uint16		lp_flags;
170
		uint16		lp_len;
171

172
		memset(nulls, 0, sizeof(nulls));
173

174
		/* Extract information from the line pointer */
175

176
		id = PageGetItemId(page, inter_call_data->offset);
177

178
		lp_offset = ItemIdGetOffset(id);
179
		lp_flags = ItemIdGetFlags(id);
180
		lp_len = ItemIdGetLength(id);
181

182
		values[0] = UInt16GetDatum(inter_call_data->offset);
183
		values[1] = UInt16GetDatum(lp_offset);
184
		values[2] = UInt16GetDatum(lp_flags);
185
		values[3] = UInt16GetDatum(lp_len);
186

187
		/*
188
		 * We do just enough validity checking to make sure we don't reference
189
		 * data outside the page passed to us. The page could be corrupt in
190
		 * many other ways, but at least we won't crash.
191
		 */
192
		if (ItemIdHasStorage(id) &&
193
			lp_len >= MinHeapTupleSize &&
194
			lp_offset == MAXALIGN(lp_offset) &&
195
			lp_offset + lp_len <= raw_page_size)
196
		{
197
			HeapTupleHeader tuphdr;
198
			bytea	   *tuple_data_bytea;
199
			int			tuple_data_len;
200

201
			/* Extract information from the tuple header */
202

203
			tuphdr = (HeapTupleHeader) PageGetItem(page, id);
204

205
			values[4] = UInt32GetDatum(HeapTupleHeaderGetRawXmin(tuphdr));
206
			values[5] = UInt32GetDatum(HeapTupleHeaderGetRawXmax(tuphdr));
207
			/* shared with xvac */
208
			values[6] = UInt32GetDatum(HeapTupleHeaderGetRawCommandId(tuphdr));
209
			values[7] = PointerGetDatum(&tuphdr->t_ctid);
210
			values[8] = UInt32GetDatum(tuphdr->t_infomask2);
211
			values[9] = UInt32GetDatum(tuphdr->t_infomask);
212
			values[10] = UInt8GetDatum(tuphdr->t_hoff);
213

214
			/* Copy raw tuple data into bytea attribute */
215
			tuple_data_len = lp_len - tuphdr->t_hoff;
216
			tuple_data_bytea = (bytea *) palloc(tuple_data_len + VARHDRSZ);
217
			SET_VARSIZE(tuple_data_bytea, tuple_data_len + VARHDRSZ);
218
			memcpy(VARDATA(tuple_data_bytea), (char *) tuphdr + tuphdr->t_hoff,
219
				   tuple_data_len);
220
			values[13] = PointerGetDatum(tuple_data_bytea);
221

222
			/*
223
			 * We already checked that the item is completely within the raw
224
			 * page passed to us, with the length given in the line pointer.
225
			 * Let's check that t_hoff doesn't point over lp_len, before using
226
			 * it to access t_bits and oid.
227
			 */
228
			if (tuphdr->t_hoff >= SizeofHeapTupleHeader &&
229
				tuphdr->t_hoff <= lp_len &&
230
				tuphdr->t_hoff == MAXALIGN(tuphdr->t_hoff))
231
			{
232
				if (tuphdr->t_infomask & HEAP_HASNULL)
233
				{
234
					int			bits_len;
235

236
					bits_len =
237
						BITMAPLEN(HeapTupleHeaderGetNatts(tuphdr)) * BITS_PER_BYTE;
238
					values[11] = CStringGetTextDatum(
239
													 bits_to_text(tuphdr->t_bits, bits_len));
240
				}
241
				else
242
					nulls[11] = true;
243

244
				if (tuphdr->t_infomask & HEAP_HASOID)
245
					values[12] = HeapTupleHeaderGetOid(tuphdr);
246
				else
247
					nulls[12] = true;
248
			}
249
			else
250
			{
251
				nulls[11] = true;
252
				nulls[12] = true;
253
			}
254
		}
255
		else
256
		{
257
			/*
258
			 * The line pointer is not used, or it's invalid. Set the rest of
259
			 * the fields to NULL
260
			 */
261
			int			i;
262

263
			for (i = 4; i <= 13; i++)
264
				nulls[i] = true;
265
		}
266

267
		/* Build and return the result tuple. */
268
		resultTuple = heap_form_tuple(inter_call_data->tupd, values, nulls);
269
		result = HeapTupleGetDatum(resultTuple);
270

271
		inter_call_data->offset++;
272

273
		SRF_RETURN_NEXT(fctx, result);
274
	}
275
	else
276
		SRF_RETURN_DONE(fctx);
277
}
278

279
/*
280
 * tuple_data_split_internal
281
 *
282
 * Split raw tuple data taken directly from a page into an array of bytea
283
 * elements. This routine does a lookup on NULL values and creates array
284
 * elements accordingly. This is a reimplementation of nocachegetattr()
285
 * in heaptuple.c simplified for educational purposes.
286
 */
287
static Datum
288
tuple_data_split_internal(Oid relid, char *tupdata,
289
						  uint16 tupdata_len, uint16 t_infomask,
290
						  uint16 t_infomask2, bits8 *t_bits,
291
						  bool do_detoast)
292
{
293
	ArrayBuildState *raw_attrs;
294
	int			nattrs;
295
	int			i;
296
	int			off = 0;
297
	Relation	rel;
298
	TupleDesc	tupdesc;
299

300
	/* Get tuple descriptor from relation OID */
301
	rel = relation_open(relid, AccessShareLock);
302
	tupdesc = RelationGetDescr(rel);
303

304
	raw_attrs = initArrayResult(BYTEAOID, CurrentMemoryContext, false);
305
	nattrs = tupdesc->natts;
306

307
	if (nattrs < (t_infomask2 & HEAP_NATTS_MASK))
308
		ereport(ERROR,
309
				(errcode(ERRCODE_DATA_CORRUPTED),
310
				 errmsg("number of attributes in tuple header is greater than number of attributes in tuple descriptor")));
311

312
	for (i = 0; i < nattrs; i++)
313
	{
314
		Form_pg_attribute attr;
315
		bool		is_null;
316
		bytea	   *attr_data = NULL;
317

318
		attr = TupleDescAttr(tupdesc, i);
319

320
		/*
321
		 * Tuple header can specify less attributes than tuple descriptor as
322
		 * ALTER TABLE ADD COLUMN without DEFAULT keyword does not actually
323
		 * change tuples in pages, so attributes with numbers greater than
324
		 * (t_infomask2 & HEAP_NATTS_MASK) should be treated as NULL.
325
		 */
326
		if (i >= (t_infomask2 & HEAP_NATTS_MASK))
327
			is_null = true;
328
		else
329
			is_null = (t_infomask & HEAP_HASNULL) && att_isnull(i, t_bits);
330

331
		if (!is_null)
332
		{
333
			int			len;
334

335
			if (attr->attlen == -1)
336
			{
337
				off = att_align_pointer(off, attr->attalign, -1,
338
										tupdata + off);
339

340
				/*
341
				 * As VARSIZE_ANY throws an exception if it can't properly
342
				 * detect the type of external storage in macros VARTAG_SIZE,
343
				 * this check is repeated to have a nicer error handling.
344
				 */
345
				if (VARATT_IS_EXTERNAL(tupdata + off) &&
346
					!VARATT_IS_EXTERNAL_ONDISK(tupdata + off) &&
347
					!VARATT_IS_EXTERNAL_INDIRECT(tupdata + off))
348
					ereport(ERROR,
349
							(errcode(ERRCODE_DATA_CORRUPTED),
350
							 errmsg("first byte of varlena attribute is incorrect for attribute %d", i)));
351

352
				len = VARSIZE_ANY(tupdata + off);
353
			}
354
			else
355
			{
356
				off = att_align_nominal(off, attr->attalign);
357
				len = attr->attlen;
358
			}
359

360
			if (tupdata_len < off + len)
361
				ereport(ERROR,
362
						(errcode(ERRCODE_DATA_CORRUPTED),
363
						 errmsg("unexpected end of tuple data")));
364

365
			if (attr->attlen == -1 && do_detoast)
366
				attr_data = DatumGetByteaPCopy(tupdata + off);
367
			else
368
			{
369
				attr_data = (bytea *) palloc(len + VARHDRSZ);
370
				SET_VARSIZE(attr_data, len + VARHDRSZ);
371
				memcpy(VARDATA(attr_data), tupdata + off, len);
372
			}
373

374
			off = att_addlength_pointer(off, attr->attlen,
375
										tupdata + off);
376
		}
377

378
		raw_attrs = accumArrayResult(raw_attrs, PointerGetDatum(attr_data),
379
									 is_null, BYTEAOID, CurrentMemoryContext);
380
		if (attr_data)
381
			pfree(attr_data);
382
	}
383

384
	if (tupdata_len != off)
385
		ereport(ERROR,
386
				(errcode(ERRCODE_DATA_CORRUPTED),
387
				 errmsg("end of tuple reached without looking at all its data")));
388

389
	relation_close(rel, AccessShareLock);
390

391
	return makeArrayResult(raw_attrs, CurrentMemoryContext);
392
}
393

394
/*
395
 * tuple_data_split
396
 *
397
 * Split raw tuple data taken directly from page into distinct elements
398
 * taking into account null values.
399
 */
400
PG_FUNCTION_INFO_V1(tuple_data_split);
401

402
Datum
403
tuple_data_split(PG_FUNCTION_ARGS)
404
{
405
	Oid			relid;
406
	bytea	   *raw_data;
407
	uint16		t_infomask;
408
	uint16		t_infomask2;
409
	char	   *t_bits_str;
410
	bool		do_detoast = false;
411
	bits8	   *t_bits = NULL;
412
	Datum		res;
413

414
	relid = PG_GETARG_OID(0);
415
	raw_data = PG_ARGISNULL(1) ? NULL : PG_GETARG_BYTEA_P(1);
416
	t_infomask = PG_GETARG_INT16(2);
417
	t_infomask2 = PG_GETARG_INT16(3);
418
	t_bits_str = PG_ARGISNULL(4) ? NULL :
419
		text_to_cstring(PG_GETARG_TEXT_PP(4));
420

421
	if (PG_NARGS() >= 6)
422
		do_detoast = PG_GETARG_BOOL(5);
423

424
	if (!superuser())
425
		ereport(ERROR,
426
				(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
427
				 errmsg("must be superuser to use raw page functions")));
428

429
	if (!raw_data)
430
		PG_RETURN_NULL();
431

432
	/*
433
	 * Convert t_bits string back to the bits8 array as represented in the
434
	 * tuple header.
435
	 */
436
	if (t_infomask & HEAP_HASNULL)
437
	{
438
		int			bits_str_len;
439
		int			bits_len;
440

441
		bits_len = BITMAPLEN(t_infomask2 & HEAP_NATTS_MASK) * BITS_PER_BYTE;
442
		if (!t_bits_str)
443
			ereport(ERROR,
444
					(errcode(ERRCODE_DATA_CORRUPTED),
445
					 errmsg("argument of t_bits is null, but it is expected to be null and %d character long",
446
							bits_len)));
447

448
		bits_str_len = strlen(t_bits_str);
449
		if (bits_len != bits_str_len)
450
			ereport(ERROR,
451
					(errcode(ERRCODE_DATA_CORRUPTED),
452
					 errmsg("unexpected length of t_bits %u, expected %d",
453
							bits_str_len, bits_len)));
454

455
		/* do the conversion */
456
		t_bits = text_to_bits(t_bits_str, bits_str_len);
457
	}
458
	else
459
	{
460
		if (t_bits_str)
461
			ereport(ERROR,
462
					(errcode(ERRCODE_DATA_CORRUPTED),
463
					 errmsg("t_bits string is expected to be NULL, but instead it is %zu bytes length",
464
							strlen(t_bits_str))));
465
	}
466

467
	/* Split tuple data */
468
	res = tuple_data_split_internal(relid, (char *) raw_data + VARHDRSZ,
469
									VARSIZE(raw_data) - VARHDRSZ,
470
									t_infomask, t_infomask2, t_bits,
471
									do_detoast);
472

473
	if (t_bits)
474
		pfree(t_bits);
475

476
	PG_RETURN_ARRAYTYPE_P(res);
477
}
478

Использование cookies

Мы используем файлы cookie в соответствии с Политикой конфиденциальности и Политикой использования cookies.

Нажимая кнопку «Принимаю», Вы даете АО «СберТех» согласие на обработку Ваших персональных данных в целях совершенствования нашего веб-сайта и Сервиса GitVerse, а также повышения удобства их использования.

Запретить использование cookies Вы можете самостоятельно в настройках Вашего браузера.