Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * inv_api.c
4 : * routines for manipulating inversion fs large objects. This file
5 : * contains the user-level large object application interface routines.
6 : *
7 : *
8 : * Note: we access pg_largeobject.data using its C struct declaration.
9 : * This is safe because it immediately follows pageno which is an int4 field,
10 : * and therefore the data field will always be 4-byte aligned, even if it
11 : * is in the short 1-byte-header format. We have to detoast it since it's
12 : * quite likely to be in compressed or short format. We also need to check
13 : * for NULLs, since initdb will mark loid and pageno but not data as NOT NULL.
14 : *
15 : * Note: many of these routines leak memory in CurrentMemoryContext, as indeed
16 : * does most of the backend code. We expect that CurrentMemoryContext will
17 : * be a short-lived context. Data that must persist across function calls
18 : * is kept either in CacheMemoryContext (the Relation structs) or in the
19 : * memory context given to inv_open (for LargeObjectDesc structs).
20 : *
21 : *
22 : * Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group
23 : * Portions Copyright (c) 1994, Regents of the University of California
24 : *
25 : *
26 : * IDENTIFICATION
27 : * src/backend/storage/large_object/inv_api.c
28 : *
29 : *-------------------------------------------------------------------------
30 : */
31 : #include "postgres.h"
32 :
33 : #include <limits.h>
34 :
35 : #include "access/detoast.h"
36 : #include "access/genam.h"
37 : #include "access/htup_details.h"
38 : #include "access/table.h"
39 : #include "access/xact.h"
40 : #include "catalog/dependency.h"
41 : #include "catalog/indexing.h"
42 : #include "catalog/objectaccess.h"
43 : #include "catalog/pg_largeobject.h"
44 : #include "libpq/libpq-fs.h"
45 : #include "miscadmin.h"
46 : #include "storage/large_object.h"
47 : #include "utils/acl.h"
48 : #include "utils/fmgroids.h"
49 : #include "utils/rel.h"
50 : #include "utils/snapmgr.h"
51 :
52 :
53 : /*
54 : * GUC: backwards-compatibility flag to suppress LO permission checks
55 : */
56 : bool lo_compat_privileges;
57 :
58 : /*
59 : * All accesses to pg_largeobject and its index make use of a single
60 : * Relation reference. To guarantee that the relcache entry remains
61 : * in the cache, on the first reference inside a subtransaction, we
62 : * execute a slightly klugy maneuver to assign ownership of the
63 : * Relation reference to TopTransactionResourceOwner.
64 : */
65 : static Relation lo_heap_r = NULL;
66 : static Relation lo_index_r = NULL;
67 :
68 :
69 : /*
70 : * Open pg_largeobject and its index, if not already done in current xact
71 : */
72 : static void
73 3176 : open_lo_relation(void)
74 : {
75 : ResourceOwner currentOwner;
76 :
77 3176 : if (lo_heap_r && lo_index_r)
78 2808 : return; /* already open in current xact */
79 :
80 : /* Arrange for the top xact to own these relation references */
81 368 : currentOwner = CurrentResourceOwner;
82 368 : CurrentResourceOwner = TopTransactionResourceOwner;
83 :
84 : /* Use RowExclusiveLock since we might either read or write */
85 368 : if (lo_heap_r == NULL)
86 368 : lo_heap_r = table_open(LargeObjectRelationId, RowExclusiveLock);
87 368 : if (lo_index_r == NULL)
88 368 : lo_index_r = index_open(LargeObjectLOidPNIndexId, RowExclusiveLock);
89 :
90 368 : CurrentResourceOwner = currentOwner;
91 : }
92 :
93 : /*
94 : * Clean up at main transaction end
95 : */
96 : void
97 534 : close_lo_relation(bool isCommit)
98 : {
99 534 : if (lo_heap_r || lo_index_r)
100 : {
101 : /*
102 : * Only bother to close if committing; else abort cleanup will handle
103 : * it
104 : */
105 368 : if (isCommit)
106 : {
107 : ResourceOwner currentOwner;
108 :
109 276 : currentOwner = CurrentResourceOwner;
110 276 : CurrentResourceOwner = TopTransactionResourceOwner;
111 :
112 276 : if (lo_index_r)
113 276 : index_close(lo_index_r, NoLock);
114 276 : if (lo_heap_r)
115 276 : table_close(lo_heap_r, NoLock);
116 :
117 276 : CurrentResourceOwner = currentOwner;
118 : }
119 368 : lo_heap_r = NULL;
120 368 : lo_index_r = NULL;
121 : }
122 534 : }
123 :
124 :
125 : /*
126 : * Extract data field from a pg_largeobject tuple, detoasting if needed
127 : * and verifying that the length is sane. Returns data pointer (a bytea *),
128 : * data length, and an indication of whether to pfree the data pointer.
129 : */
130 : static void
131 10312 : getdatafield(Form_pg_largeobject tuple,
132 : bytea **pdatafield,
133 : int *plen,
134 : bool *pfreeit)
135 : {
136 : bytea *datafield;
137 : int len;
138 : bool freeit;
139 :
140 10312 : datafield = &(tuple->data); /* see note at top of file */
141 10312 : freeit = false;
142 10312 : if (VARATT_IS_EXTENDED(datafield))
143 : {
144 : datafield = (bytea *)
145 10146 : detoast_attr((struct varlena *) datafield);
146 10146 : freeit = true;
147 : }
148 10312 : len = VARSIZE(datafield) - VARHDRSZ;
149 10312 : if (len < 0 || len > LOBLKSIZE)
150 0 : ereport(ERROR,
151 : (errcode(ERRCODE_DATA_CORRUPTED),
152 : errmsg("pg_largeobject entry for OID %u, page %d has invalid data field size %d",
153 : tuple->loid, tuple->pageno, len)));
154 10312 : *pdatafield = datafield;
155 10312 : *plen = len;
156 10312 : *pfreeit = freeit;
157 10312 : }
158 :
159 :
160 : /*
161 : * inv_create -- create a new large object
162 : *
163 : * Arguments:
164 : * lobjId - OID to use for new large object, or InvalidOid to pick one
165 : *
166 : * Returns:
167 : * OID of new object
168 : *
169 : * If lobjId is not InvalidOid, then an error occurs if the OID is already
170 : * in use.
171 : */
172 : Oid
173 168 : inv_create(Oid lobjId)
174 : {
175 : Oid lobjId_new;
176 :
177 : /*
178 : * Create a new largeobject with empty data pages
179 : */
180 168 : lobjId_new = LargeObjectCreate(lobjId);
181 :
182 : /*
183 : * dependency on the owner of largeobject
184 : *
185 : * Note that LO dependencies are recorded using classId
186 : * LargeObjectRelationId for backwards-compatibility reasons. Using
187 : * LargeObjectMetadataRelationId instead would simplify matters for the
188 : * backend, but it'd complicate pg_dump and possibly break other clients.
189 : */
190 168 : recordDependencyOnOwner(LargeObjectRelationId,
191 : lobjId_new, GetUserId());
192 :
193 : /* Post creation hook for new large object */
194 168 : InvokeObjectPostCreateHook(LargeObjectRelationId, lobjId_new, 0);
195 :
196 : /*
197 : * Advance command counter to make new tuple visible to later operations.
198 : */
199 168 : CommandCounterIncrement();
200 :
201 168 : return lobjId_new;
202 : }
203 :
204 : /*
205 : * inv_open -- access an existing large object.
206 : *
207 : * Returns a large object descriptor, appropriately filled in.
208 : * The descriptor and subsidiary data are allocated in the specified
209 : * memory context, which must be suitably long-lived for the caller's
210 : * purposes. If the returned descriptor has a snapshot associated
211 : * with it, the caller must ensure that it also lives long enough,
212 : * e.g. by calling RegisterSnapshotOnOwner
213 : */
214 : LargeObjectDesc *
215 550 : inv_open(Oid lobjId, int flags, MemoryContext mcxt)
216 : {
217 : LargeObjectDesc *retval;
218 550 : Snapshot snapshot = NULL;
219 550 : int descflags = 0;
220 :
221 : /*
222 : * Historically, no difference is made between (INV_WRITE) and (INV_WRITE
223 : * | INV_READ), the caller being allowed to read the large object
224 : * descriptor in either case.
225 : */
226 550 : if (flags & INV_WRITE)
227 186 : descflags |= IFS_WRLOCK | IFS_RDLOCK;
228 550 : if (flags & INV_READ)
229 394 : descflags |= IFS_RDLOCK;
230 :
231 550 : if (descflags == 0)
232 0 : ereport(ERROR,
233 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
234 : errmsg("invalid flags for opening a large object: %d",
235 : flags)));
236 :
237 : /* Get snapshot. If write is requested, use an instantaneous snapshot. */
238 550 : if (descflags & IFS_WRLOCK)
239 186 : snapshot = NULL;
240 : else
241 364 : snapshot = GetActiveSnapshot();
242 :
243 : /* Can't use LargeObjectExists here because we need to specify snapshot */
244 550 : if (!LargeObjectExistsWithSnapshot(lobjId, snapshot))
245 10 : ereport(ERROR,
246 : (errcode(ERRCODE_UNDEFINED_OBJECT),
247 : errmsg("large object %u does not exist", lobjId)));
248 :
249 : /* Apply permission checks, again specifying snapshot */
250 540 : if ((descflags & IFS_RDLOCK) != 0)
251 : {
252 1062 : if (!lo_compat_privileges &&
253 522 : pg_largeobject_aclcheck_snapshot(lobjId,
254 : GetUserId(),
255 : ACL_SELECT,
256 : snapshot) != ACLCHECK_OK)
257 42 : ereport(ERROR,
258 : (errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
259 : errmsg("permission denied for large object %u",
260 : lobjId)));
261 : }
262 498 : if ((descflags & IFS_WRLOCK) != 0)
263 : {
264 312 : if (!lo_compat_privileges &&
265 150 : pg_largeobject_aclcheck_snapshot(lobjId,
266 : GetUserId(),
267 : ACL_UPDATE,
268 : snapshot) != ACLCHECK_OK)
269 12 : ereport(ERROR,
270 : (errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
271 : errmsg("permission denied for large object %u",
272 : lobjId)));
273 : }
274 :
275 : /* OK to create a descriptor */
276 486 : retval = (LargeObjectDesc *) MemoryContextAlloc(mcxt,
277 : sizeof(LargeObjectDesc));
278 486 : retval->id = lobjId;
279 486 : retval->offset = 0;
280 486 : retval->flags = descflags;
281 :
282 : /* caller sets if needed, not used by the functions in this file */
283 486 : retval->subid = InvalidSubTransactionId;
284 :
285 : /*
286 : * The snapshot (if any) is just the currently active snapshot. The
287 : * caller will replace it with a longer-lived copy if needed.
288 : */
289 486 : retval->snapshot = snapshot;
290 :
291 486 : return retval;
292 : }
293 :
294 : /*
295 : * Closes a large object descriptor previously made by inv_open(), and
296 : * releases the long-term memory used by it.
297 : */
298 : void
299 456 : inv_close(LargeObjectDesc *obj_desc)
300 : {
301 : Assert(obj_desc);
302 456 : pfree(obj_desc);
303 456 : }
304 :
305 : /*
306 : * Destroys an existing large object (not to be confused with a descriptor!)
307 : *
308 : * Note we expect caller to have done any required permissions check.
309 : */
310 : int
311 88 : inv_drop(Oid lobjId)
312 : {
313 : ObjectAddress object;
314 :
315 : /*
316 : * Delete any comments and dependencies on the large object
317 : */
318 88 : object.classId = LargeObjectRelationId;
319 88 : object.objectId = lobjId;
320 88 : object.objectSubId = 0;
321 88 : performDeletion(&object, DROP_CASCADE, 0);
322 :
323 : /*
324 : * Advance command counter so that tuple removal will be seen by later
325 : * large-object operations in this transaction.
326 : */
327 88 : CommandCounterIncrement();
328 :
329 : /* For historical reasons, we always return 1 on success. */
330 88 : return 1;
331 : }
332 :
333 : /*
334 : * Determine size of a large object
335 : *
336 : * NOTE: LOs can contain gaps, just like Unix files. We actually return
337 : * the offset of the last byte + 1.
338 : */
339 : static uint64
340 138 : inv_getsize(LargeObjectDesc *obj_desc)
341 : {
342 138 : uint64 lastbyte = 0;
343 : ScanKeyData skey[1];
344 : SysScanDesc sd;
345 : HeapTuple tuple;
346 :
347 : Assert(obj_desc);
348 :
349 138 : open_lo_relation();
350 :
351 138 : ScanKeyInit(&skey[0],
352 : Anum_pg_largeobject_loid,
353 : BTEqualStrategyNumber, F_OIDEQ,
354 : ObjectIdGetDatum(obj_desc->id));
355 :
356 138 : sd = systable_beginscan_ordered(lo_heap_r, lo_index_r,
357 : obj_desc->snapshot, 1, skey);
358 :
359 : /*
360 : * Because the pg_largeobject index is on both loid and pageno, but we
361 : * constrain only loid, a backwards scan should visit all pages of the
362 : * large object in reverse pageno order. So, it's sufficient to examine
363 : * the first valid tuple (== last valid page).
364 : */
365 138 : tuple = systable_getnext_ordered(sd, BackwardScanDirection);
366 138 : if (HeapTupleIsValid(tuple))
367 : {
368 : Form_pg_largeobject data;
369 : bytea *datafield;
370 : int len;
371 : bool pfreeit;
372 :
373 120 : if (HeapTupleHasNulls(tuple)) /* paranoia */
374 0 : elog(ERROR, "null field found in pg_largeobject");
375 120 : data = (Form_pg_largeobject) GETSTRUCT(tuple);
376 120 : getdatafield(data, &datafield, &len, &pfreeit);
377 120 : lastbyte = (uint64) data->pageno * LOBLKSIZE + len;
378 120 : if (pfreeit)
379 42 : pfree(datafield);
380 : }
381 :
382 138 : systable_endscan_ordered(sd);
383 :
384 138 : return lastbyte;
385 : }
386 :
387 : int64
388 288 : inv_seek(LargeObjectDesc *obj_desc, int64 offset, int whence)
389 : {
390 : int64 newoffset;
391 :
392 : Assert(obj_desc);
393 :
394 : /*
395 : * We allow seek/tell if you have either read or write permission, so no
396 : * need for a permission check here.
397 : */
398 :
399 : /*
400 : * Note: overflow in the additions is possible, but since we will reject
401 : * negative results, we don't need any extra test for that.
402 : */
403 288 : switch (whence)
404 : {
405 132 : case SEEK_SET:
406 132 : newoffset = offset;
407 132 : break;
408 18 : case SEEK_CUR:
409 18 : newoffset = obj_desc->offset + offset;
410 18 : break;
411 138 : case SEEK_END:
412 138 : newoffset = inv_getsize(obj_desc) + offset;
413 138 : break;
414 0 : default:
415 0 : ereport(ERROR,
416 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
417 : errmsg("invalid whence setting: %d", whence)));
418 : newoffset = 0; /* keep compiler quiet */
419 : break;
420 : }
421 :
422 : /*
423 : * use errmsg_internal here because we don't want to expose INT64_FORMAT
424 : * in translatable strings; doing better is not worth the trouble
425 : */
426 288 : if (newoffset < 0 || newoffset > MAX_LARGE_OBJECT_SIZE)
427 0 : ereport(ERROR,
428 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
429 : errmsg_internal("invalid large object seek target: " INT64_FORMAT,
430 : newoffset)));
431 :
432 288 : obj_desc->offset = newoffset;
433 288 : return newoffset;
434 : }
435 :
436 : int64
437 48 : inv_tell(LargeObjectDesc *obj_desc)
438 : {
439 : Assert(obj_desc);
440 :
441 : /*
442 : * We allow seek/tell if you have either read or write permission, so no
443 : * need for a permission check here.
444 : */
445 :
446 48 : return obj_desc->offset;
447 : }
448 :
449 : int
450 1430 : inv_read(LargeObjectDesc *obj_desc, char *buf, int nbytes)
451 : {
452 1430 : int nread = 0;
453 : int64 n;
454 : int64 off;
455 : int len;
456 1430 : int32 pageno = (int32) (obj_desc->offset / LOBLKSIZE);
457 : uint64 pageoff;
458 : ScanKeyData skey[2];
459 : SysScanDesc sd;
460 : HeapTuple tuple;
461 :
462 : Assert(obj_desc);
463 : Assert(buf != NULL);
464 :
465 1430 : if ((obj_desc->flags & IFS_RDLOCK) == 0)
466 0 : ereport(ERROR,
467 : (errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
468 : errmsg("permission denied for large object %u",
469 : obj_desc->id)));
470 :
471 1430 : if (nbytes <= 0)
472 18 : return 0;
473 :
474 1412 : open_lo_relation();
475 :
476 1412 : ScanKeyInit(&skey[0],
477 : Anum_pg_largeobject_loid,
478 : BTEqualStrategyNumber, F_OIDEQ,
479 : ObjectIdGetDatum(obj_desc->id));
480 :
481 1412 : ScanKeyInit(&skey[1],
482 : Anum_pg_largeobject_pageno,
483 : BTGreaterEqualStrategyNumber, F_INT4GE,
484 : Int32GetDatum(pageno));
485 :
486 1412 : sd = systable_beginscan_ordered(lo_heap_r, lo_index_r,
487 : obj_desc->snapshot, 2, skey);
488 :
489 10488 : while ((tuple = systable_getnext_ordered(sd, ForwardScanDirection)) != NULL)
490 : {
491 : Form_pg_largeobject data;
492 : bytea *datafield;
493 : bool pfreeit;
494 :
495 10162 : if (HeapTupleHasNulls(tuple)) /* paranoia */
496 0 : elog(ERROR, "null field found in pg_largeobject");
497 10162 : data = (Form_pg_largeobject) GETSTRUCT(tuple);
498 :
499 : /*
500 : * We expect the indexscan will deliver pages in order. However,
501 : * there may be missing pages if the LO contains unwritten "holes". We
502 : * want missing sections to read out as zeroes.
503 : */
504 10162 : pageoff = ((uint64) data->pageno) * LOBLKSIZE;
505 10162 : if (pageoff > obj_desc->offset)
506 : {
507 12 : n = pageoff - obj_desc->offset;
508 12 : n = (n <= (nbytes - nread)) ? n : (nbytes - nread);
509 12 : MemSet(buf + nread, 0, n);
510 12 : nread += n;
511 12 : obj_desc->offset += n;
512 : }
513 :
514 10162 : if (nread < nbytes)
515 : {
516 : Assert(obj_desc->offset >= pageoff);
517 10156 : off = (int) (obj_desc->offset - pageoff);
518 : Assert(off >= 0 && off < LOBLKSIZE);
519 :
520 10156 : getdatafield(data, &datafield, &len, &pfreeit);
521 10156 : if (len > off)
522 : {
523 10046 : n = len - off;
524 10046 : n = (n <= (nbytes - nread)) ? n : (nbytes - nread);
525 10046 : memcpy(buf + nread, VARDATA(datafield) + off, n);
526 10046 : nread += n;
527 10046 : obj_desc->offset += n;
528 : }
529 10156 : if (pfreeit)
530 10080 : pfree(datafield);
531 : }
532 :
533 10162 : if (nread >= nbytes)
534 1086 : break;
535 : }
536 :
537 1412 : systable_endscan_ordered(sd);
538 :
539 1412 : return nread;
540 : }
541 :
542 : int
543 1584 : inv_write(LargeObjectDesc *obj_desc, const char *buf, int nbytes)
544 : {
545 1584 : int nwritten = 0;
546 : int n;
547 : int off;
548 : int len;
549 1584 : int32 pageno = (int32) (obj_desc->offset / LOBLKSIZE);
550 : ScanKeyData skey[2];
551 : SysScanDesc sd;
552 : HeapTuple oldtuple;
553 : Form_pg_largeobject olddata;
554 : bool neednextpage;
555 : bytea *datafield;
556 : bool pfreeit;
557 : union
558 : {
559 : alignas(int32) bytea hdr;
560 : /* this is to make the union big enough for a LO data chunk: */
561 : char data[LOBLKSIZE + VARHDRSZ];
562 1584 : } workbuf = {0};
563 1584 : char *workb = VARDATA(&workbuf.hdr);
564 : HeapTuple newtup;
565 : Datum values[Natts_pg_largeobject];
566 : bool nulls[Natts_pg_largeobject];
567 : bool replace[Natts_pg_largeobject];
568 : CatalogIndexState indstate;
569 :
570 : Assert(obj_desc);
571 : Assert(buf != NULL);
572 :
573 : /* enforce writability because snapshot is probably wrong otherwise */
574 1584 : if ((obj_desc->flags & IFS_WRLOCK) == 0)
575 0 : ereport(ERROR,
576 : (errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
577 : errmsg("permission denied for large object %u",
578 : obj_desc->id)));
579 :
580 1584 : if (nbytes <= 0)
581 0 : return 0;
582 :
583 : /* this addition can't overflow because nbytes is only int32 */
584 1584 : if ((nbytes + obj_desc->offset) > MAX_LARGE_OBJECT_SIZE)
585 0 : ereport(ERROR,
586 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
587 : errmsg("invalid large object write request size: %d",
588 : nbytes)));
589 :
590 1584 : open_lo_relation();
591 :
592 1584 : indstate = CatalogOpenIndexes(lo_heap_r);
593 :
594 1584 : ScanKeyInit(&skey[0],
595 : Anum_pg_largeobject_loid,
596 : BTEqualStrategyNumber, F_OIDEQ,
597 : ObjectIdGetDatum(obj_desc->id));
598 :
599 1584 : ScanKeyInit(&skey[1],
600 : Anum_pg_largeobject_pageno,
601 : BTGreaterEqualStrategyNumber, F_INT4GE,
602 : Int32GetDatum(pageno));
603 :
604 1584 : sd = systable_beginscan_ordered(lo_heap_r, lo_index_r,
605 : obj_desc->snapshot, 2, skey);
606 :
607 1584 : oldtuple = NULL;
608 1584 : olddata = NULL;
609 1584 : neednextpage = true;
610 :
611 9564 : while (nwritten < nbytes)
612 : {
613 : /*
614 : * If possible, get next pre-existing page of the LO. We expect the
615 : * indexscan will deliver these in order --- but there may be holes.
616 : */
617 7980 : if (neednextpage)
618 : {
619 1590 : if ((oldtuple = systable_getnext_ordered(sd, ForwardScanDirection)) != NULL)
620 : {
621 24 : if (HeapTupleHasNulls(oldtuple)) /* paranoia */
622 0 : elog(ERROR, "null field found in pg_largeobject");
623 24 : olddata = (Form_pg_largeobject) GETSTRUCT(oldtuple);
624 : Assert(olddata->pageno >= pageno);
625 : }
626 1590 : neednextpage = false;
627 : }
628 :
629 : /*
630 : * If we have a pre-existing page, see if it is the page we want to
631 : * write, or a later one.
632 : */
633 7980 : if (olddata != NULL && olddata->pageno == pageno)
634 : {
635 : /*
636 : * Update an existing page with fresh data.
637 : *
638 : * First, load old data into workbuf
639 : */
640 24 : getdatafield(olddata, &datafield, &len, &pfreeit);
641 24 : memcpy(workb, VARDATA(datafield), len);
642 24 : if (pfreeit)
643 18 : pfree(datafield);
644 :
645 : /*
646 : * Fill any hole
647 : */
648 24 : off = (int) (obj_desc->offset % LOBLKSIZE);
649 24 : if (off > len)
650 0 : MemSet(workb + len, 0, off - len);
651 :
652 : /*
653 : * Insert appropriate portion of new data
654 : */
655 24 : n = LOBLKSIZE - off;
656 24 : n = (n <= (nbytes - nwritten)) ? n : (nbytes - nwritten);
657 24 : memcpy(workb + off, buf + nwritten, n);
658 24 : nwritten += n;
659 24 : obj_desc->offset += n;
660 24 : off += n;
661 : /* compute valid length of new page */
662 24 : len = (len >= off) ? len : off;
663 24 : SET_VARSIZE(&workbuf.hdr, len + VARHDRSZ);
664 :
665 : /*
666 : * Form and insert updated tuple
667 : */
668 24 : memset(values, 0, sizeof(values));
669 24 : memset(nulls, false, sizeof(nulls));
670 24 : memset(replace, false, sizeof(replace));
671 24 : values[Anum_pg_largeobject_data - 1] = PointerGetDatum(&workbuf);
672 24 : replace[Anum_pg_largeobject_data - 1] = true;
673 24 : newtup = heap_modify_tuple(oldtuple, RelationGetDescr(lo_heap_r),
674 : values, nulls, replace);
675 24 : CatalogTupleUpdateWithInfo(lo_heap_r, &newtup->t_self, newtup,
676 : indstate);
677 24 : heap_freetuple(newtup);
678 :
679 : /*
680 : * We're done with this old page.
681 : */
682 24 : oldtuple = NULL;
683 24 : olddata = NULL;
684 24 : neednextpage = true;
685 : }
686 : else
687 : {
688 : /*
689 : * Write a brand new page.
690 : *
691 : * First, fill any hole
692 : */
693 7956 : off = (int) (obj_desc->offset % LOBLKSIZE);
694 7956 : if (off > 0)
695 6 : MemSet(workb, 0, off);
696 :
697 : /*
698 : * Insert appropriate portion of new data
699 : */
700 7956 : n = LOBLKSIZE - off;
701 7956 : n = (n <= (nbytes - nwritten)) ? n : (nbytes - nwritten);
702 7956 : memcpy(workb + off, buf + nwritten, n);
703 7956 : nwritten += n;
704 7956 : obj_desc->offset += n;
705 : /* compute valid length of new page */
706 7956 : len = off + n;
707 7956 : SET_VARSIZE(&workbuf.hdr, len + VARHDRSZ);
708 :
709 : /*
710 : * Form and insert updated tuple
711 : */
712 7956 : memset(values, 0, sizeof(values));
713 7956 : memset(nulls, false, sizeof(nulls));
714 7956 : values[Anum_pg_largeobject_loid - 1] = ObjectIdGetDatum(obj_desc->id);
715 7956 : values[Anum_pg_largeobject_pageno - 1] = Int32GetDatum(pageno);
716 7956 : values[Anum_pg_largeobject_data - 1] = PointerGetDatum(&workbuf);
717 7956 : newtup = heap_form_tuple(lo_heap_r->rd_att, values, nulls);
718 7956 : CatalogTupleInsertWithInfo(lo_heap_r, newtup, indstate);
719 7956 : heap_freetuple(newtup);
720 : }
721 7980 : pageno++;
722 : }
723 :
724 1584 : systable_endscan_ordered(sd);
725 :
726 1584 : CatalogCloseIndexes(indstate);
727 :
728 : /*
729 : * Advance command counter so that my tuple updates will be seen by later
730 : * large-object operations in this transaction.
731 : */
732 1584 : CommandCounterIncrement();
733 :
734 1584 : return nwritten;
735 : }
736 :
737 : void
738 42 : inv_truncate(LargeObjectDesc *obj_desc, int64 len)
739 : {
740 42 : int32 pageno = (int32) (len / LOBLKSIZE);
741 : int32 off;
742 : ScanKeyData skey[2];
743 : SysScanDesc sd;
744 : HeapTuple oldtuple;
745 : Form_pg_largeobject olddata;
746 : union
747 : {
748 : alignas(int32) bytea hdr;
749 : /* this is to make the union big enough for a LO data chunk: */
750 : char data[LOBLKSIZE + VARHDRSZ];
751 42 : } workbuf = {0};
752 42 : char *workb = VARDATA(&workbuf.hdr);
753 : HeapTuple newtup;
754 : Datum values[Natts_pg_largeobject];
755 : bool nulls[Natts_pg_largeobject];
756 : bool replace[Natts_pg_largeobject];
757 : CatalogIndexState indstate;
758 :
759 : Assert(obj_desc);
760 :
761 : /* enforce writability because snapshot is probably wrong otherwise */
762 42 : if ((obj_desc->flags & IFS_WRLOCK) == 0)
763 0 : ereport(ERROR,
764 : (errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
765 : errmsg("permission denied for large object %u",
766 : obj_desc->id)));
767 :
768 : /*
769 : * use errmsg_internal here because we don't want to expose INT64_FORMAT
770 : * in translatable strings; doing better is not worth the trouble
771 : */
772 42 : if (len < 0 || len > MAX_LARGE_OBJECT_SIZE)
773 0 : ereport(ERROR,
774 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
775 : errmsg_internal("invalid large object truncation target: " INT64_FORMAT,
776 : len)));
777 :
778 42 : open_lo_relation();
779 :
780 42 : indstate = CatalogOpenIndexes(lo_heap_r);
781 :
782 : /*
783 : * Set up to find all pages with desired loid and pageno >= target
784 : */
785 42 : ScanKeyInit(&skey[0],
786 : Anum_pg_largeobject_loid,
787 : BTEqualStrategyNumber, F_OIDEQ,
788 : ObjectIdGetDatum(obj_desc->id));
789 :
790 42 : ScanKeyInit(&skey[1],
791 : Anum_pg_largeobject_pageno,
792 : BTGreaterEqualStrategyNumber, F_INT4GE,
793 : Int32GetDatum(pageno));
794 :
795 42 : sd = systable_beginscan_ordered(lo_heap_r, lo_index_r,
796 : obj_desc->snapshot, 2, skey);
797 :
798 : /*
799 : * If possible, get the page the truncation point is in. The truncation
800 : * point may be beyond the end of the LO or in a hole.
801 : */
802 42 : olddata = NULL;
803 42 : if ((oldtuple = systable_getnext_ordered(sd, ForwardScanDirection)) != NULL)
804 : {
805 24 : if (HeapTupleHasNulls(oldtuple)) /* paranoia */
806 0 : elog(ERROR, "null field found in pg_largeobject");
807 24 : olddata = (Form_pg_largeobject) GETSTRUCT(oldtuple);
808 : Assert(olddata->pageno >= pageno);
809 : }
810 :
811 : /*
812 : * If we found the page of the truncation point we need to truncate the
813 : * data in it. Otherwise if we're in a hole, we need to create a page to
814 : * mark the end of data.
815 : */
816 42 : if (olddata != NULL && olddata->pageno == pageno)
817 12 : {
818 : /* First, load old data into workbuf */
819 : bytea *datafield;
820 : int pagelen;
821 : bool pfreeit;
822 :
823 12 : getdatafield(olddata, &datafield, &pagelen, &pfreeit);
824 12 : memcpy(workb, VARDATA(datafield), pagelen);
825 12 : if (pfreeit)
826 6 : pfree(datafield);
827 :
828 : /*
829 : * Fill any hole
830 : */
831 12 : off = len % LOBLKSIZE;
832 12 : if (off > pagelen)
833 6 : MemSet(workb + pagelen, 0, off - pagelen);
834 :
835 : /* compute length of new page */
836 12 : SET_VARSIZE(&workbuf.hdr, off + VARHDRSZ);
837 :
838 : /*
839 : * Form and insert updated tuple
840 : */
841 12 : memset(values, 0, sizeof(values));
842 12 : memset(nulls, false, sizeof(nulls));
843 12 : memset(replace, false, sizeof(replace));
844 12 : values[Anum_pg_largeobject_data - 1] = PointerGetDatum(&workbuf);
845 12 : replace[Anum_pg_largeobject_data - 1] = true;
846 12 : newtup = heap_modify_tuple(oldtuple, RelationGetDescr(lo_heap_r),
847 : values, nulls, replace);
848 12 : CatalogTupleUpdateWithInfo(lo_heap_r, &newtup->t_self, newtup,
849 : indstate);
850 12 : heap_freetuple(newtup);
851 : }
852 : else
853 : {
854 : /*
855 : * If the first page we found was after the truncation point, we're in
856 : * a hole that we'll fill, but we need to delete the later page
857 : * because the loop below won't visit it again.
858 : */
859 30 : if (olddata != NULL)
860 : {
861 : Assert(olddata->pageno > pageno);
862 12 : CatalogTupleDelete(lo_heap_r, &oldtuple->t_self);
863 : }
864 :
865 : /*
866 : * Write a brand new page.
867 : *
868 : * Fill the hole up to the truncation point
869 : */
870 30 : off = len % LOBLKSIZE;
871 30 : if (off > 0)
872 30 : MemSet(workb, 0, off);
873 :
874 : /* compute length of new page */
875 30 : SET_VARSIZE(&workbuf.hdr, off + VARHDRSZ);
876 :
877 : /*
878 : * Form and insert new tuple
879 : */
880 30 : memset(values, 0, sizeof(values));
881 30 : memset(nulls, false, sizeof(nulls));
882 30 : values[Anum_pg_largeobject_loid - 1] = ObjectIdGetDatum(obj_desc->id);
883 30 : values[Anum_pg_largeobject_pageno - 1] = Int32GetDatum(pageno);
884 30 : values[Anum_pg_largeobject_data - 1] = PointerGetDatum(&workbuf);
885 30 : newtup = heap_form_tuple(lo_heap_r->rd_att, values, nulls);
886 30 : CatalogTupleInsertWithInfo(lo_heap_r, newtup, indstate);
887 30 : heap_freetuple(newtup);
888 : }
889 :
890 : /*
891 : * Delete any pages after the truncation point. If the initial search
892 : * didn't find a page, then of course there's nothing more to do.
893 : */
894 42 : if (olddata != NULL)
895 : {
896 30 : while ((oldtuple = systable_getnext_ordered(sd, ForwardScanDirection)) != NULL)
897 : {
898 6 : CatalogTupleDelete(lo_heap_r, &oldtuple->t_self);
899 : }
900 : }
901 :
902 42 : systable_endscan_ordered(sd);
903 :
904 42 : CatalogCloseIndexes(indstate);
905 :
906 : /*
907 : * Advance command counter so that tuple updates will be seen by later
908 : * large-object operations in this transaction.
909 : */
910 42 : CommandCounterIncrement();
911 42 : }
|
