1
/*-------------------------------------------------------------------------
5
* Implements the custom output format.
7
* The comments with the routined in this code are a good place to
8
* understand how to write a new format.
10
* See the headers to pg_restore for more details.
12
* Copyright (c) 2000, Philip Warner
13
* Rights are granted to use this software in any way so long
14
* as this notice is not removed.
16
* The author is not responsible for loss or damages that may
17
* and any liability will be limited to the time taken to fix any
22
* $PostgreSQL: pgsql/src/bin/pg_dump/pg_backup_custom.c,v 1.29.4.1 2005-01-25 22:44:47 tgl Exp $
24
*-------------------------------------------------------------------------
27
#include "pg_backup.h"
28
#include "pg_backup_archiver.h"
33
* Routines in the format interface
37
static void _ArchiveEntry(ArchiveHandle *AH, TocEntry *te);
38
static void _StartData(ArchiveHandle *AH, TocEntry *te);
39
static size_t _WriteData(ArchiveHandle *AH, const void *data, size_t dLen);
40
static void _EndData(ArchiveHandle *AH, TocEntry *te);
41
static int _WriteByte(ArchiveHandle *AH, const int i);
42
static int _ReadByte(ArchiveHandle *);
43
static size_t _WriteBuf(ArchiveHandle *AH, const void *buf, size_t len);
44
static size_t _ReadBuf(ArchiveHandle *AH, void *buf, size_t len);
45
static void _CloseArchive(ArchiveHandle *AH);
46
static void _PrintTocData(ArchiveHandle *AH, TocEntry *te, RestoreOptions *ropt);
47
static void _WriteExtraToc(ArchiveHandle *AH, TocEntry *te);
48
static void _ReadExtraToc(ArchiveHandle *AH, TocEntry *te);
49
static void _PrintExtraToc(ArchiveHandle *AH, TocEntry *te);
51
static void _PrintData(ArchiveHandle *AH);
52
static void _skipData(ArchiveHandle *AH);
53
static void _skipBlobs(ArchiveHandle *AH);
55
static void _StartBlobs(ArchiveHandle *AH, TocEntry *te);
56
static void _StartBlob(ArchiveHandle *AH, TocEntry *te, Oid oid);
57
static void _EndBlob(ArchiveHandle *AH, TocEntry *te, Oid oid);
58
static void _EndBlobs(ArchiveHandle *AH, TocEntry *te);
59
static void _LoadBlobs(ArchiveHandle *AH);
62
* Buffers used in zlib compression and extra data stored in archive and
66
#define zlibOutSize 4096
67
#define zlibInSize 4096
91
static void _readBlockHeader(ArchiveHandle *AH, int *type, int *id);
92
static void _StartDataCompressor(ArchiveHandle *AH, TocEntry *te);
93
static void _EndDataCompressor(ArchiveHandle *AH, TocEntry *te);
94
static off_t _getFilePos(ArchiveHandle *AH, lclContext *ctx);
95
static int _DoDeflate(ArchiveHandle *AH, lclContext *ctx, int flush);
97
static char *modulename = gettext_noop("custom archiver");
102
* Init routine required by ALL formats. This is a global routine
103
* and should be declared in pg_backup_archiver.h
105
* It's task is to create any extra archive context (using AH->formatData),
106
* and to initialize the supported function pointers.
108
* It should also prepare whatever it's input source is for reading/writing,
109
* and in the case of a read mode connection, it should load the Header & TOC.
112
InitArchiveFmt_Custom(ArchiveHandle *AH)
116
/* Assuming static functions, this can be copied for each format. */
117
AH->ArchiveEntryPtr = _ArchiveEntry;
118
AH->StartDataPtr = _StartData;
119
AH->WriteDataPtr = _WriteData;
120
AH->EndDataPtr = _EndData;
121
AH->WriteBytePtr = _WriteByte;
122
AH->ReadBytePtr = _ReadByte;
123
AH->WriteBufPtr = _WriteBuf;
124
AH->ReadBufPtr = _ReadBuf;
125
AH->ClosePtr = _CloseArchive;
126
AH->PrintTocDataPtr = _PrintTocData;
127
AH->ReadExtraTocPtr = _ReadExtraToc;
128
AH->WriteExtraTocPtr = _WriteExtraToc;
129
AH->PrintExtraTocPtr = _PrintExtraToc;
131
AH->StartBlobsPtr = _StartBlobs;
132
AH->StartBlobPtr = _StartBlob;
133
AH->EndBlobPtr = _EndBlob;
134
AH->EndBlobsPtr = _EndBlobs;
137
* Set up some special context used in compressing data.
139
ctx = (lclContext *) calloc(1, sizeof(lclContext));
141
die_horribly(AH, modulename, "out of memory\n");
142
AH->formatData = (void *) ctx;
144
ctx->zp = (z_streamp) malloc(sizeof(z_stream));
146
die_horribly(AH, modulename, "out of memory\n");
148
/* Initialize LO buffering */
149
AH->lo_buf_size = LOBBUFSIZE;
150
AH->lo_buf = (void *) malloc(LOBBUFSIZE);
151
if (AH->lo_buf == NULL)
152
die_horribly(AH, modulename, "out of memory\n");
155
* zlibOutSize is the buffer size we tell zlib it can output to. We
156
* actually allocate one extra byte because some routines want to
157
* append a trailing zero byte to the zlib output. The input buffer
158
* is expansible and is always of size ctx->inSize; zlibInSize is just
159
* the initial default size for it.
161
ctx->zlibOut = (char *) malloc(zlibOutSize + 1);
162
ctx->zlibIn = (char *) malloc(zlibInSize);
163
ctx->inSize = zlibInSize;
166
if (ctx->zlibOut == NULL || ctx->zlibIn == NULL)
167
die_horribly(AH, modulename, "out of memory\n");
172
if (AH->mode == archModeWrite)
174
if (AH->fSpec && strcmp(AH->fSpec, "") != 0)
175
AH->FH = fopen(AH->fSpec, PG_BINARY_W);
180
die_horribly(AH, modulename, "could not open archive file \"%s\": %s\n", AH->fSpec, strerror(errno));
182
ctx->hasSeek = checkSeek(AH->FH);
186
if (AH->fSpec && strcmp(AH->fSpec, "") != 0)
187
AH->FH = fopen(AH->fSpec, PG_BINARY_R);
191
die_horribly(AH, modulename, "could not open archive file \"%s\": %s\n", AH->fSpec, strerror(errno));
193
ctx->hasSeek = checkSeek(AH->FH);
197
ctx->dataStart = _getFilePos(AH, ctx);
203
* Called by the Archiver when the dumper creates a new TOC entry.
207
* Set up extrac format-related TOC data.
210
_ArchiveEntry(ArchiveHandle *AH, TocEntry *te)
214
ctx = (lclTocEntry *) calloc(1, sizeof(lclTocEntry));
216
ctx->dataState = K_OFFSET_POS_NOT_SET;
218
ctx->dataState = K_OFFSET_NO_DATA;
220
te->formatData = (void *) ctx;
224
* Called by the Archiver to save any extra format-related TOC entry
229
* Use the Archiver routines to write data - they are non-endian, and
230
* maintain other important file information.
233
_WriteExtraToc(ArchiveHandle *AH, TocEntry *te)
235
lclTocEntry *ctx = (lclTocEntry *) te->formatData;
237
WriteOffset(AH, ctx->dataPos, ctx->dataState);
241
* Called by the Archiver to read any extra format-related TOC data.
245
* Needs to match the order defined in _WriteExtraToc, and sould also
246
* use the Archiver input routines.
249
_ReadExtraToc(ArchiveHandle *AH, TocEntry *te)
252
lclTocEntry *ctx = (lclTocEntry *) te->formatData;
256
ctx = (lclTocEntry *) calloc(1, sizeof(lclTocEntry));
257
te->formatData = (void *) ctx;
260
ctx->dataState = ReadOffset(AH, &(ctx->dataPos));
263
* Prior to V1.7 (pg7.3), we dumped the data size as an int now we
264
* don't dump it at all.
266
if (AH->version < K_VERS_1_7)
271
* Called by the Archiver when restoring an archive to output a comment
272
* that includes useful information about the TOC entry.
278
_PrintExtraToc(ArchiveHandle *AH, TocEntry *te)
280
lclTocEntry *ctx = (lclTocEntry *) te->formatData;
282
if (AH->public.verbose)
283
ahprintf(AH, "-- Data Pos: " INT64_FORMAT "\n",
284
(int64) ctx->dataPos);
288
* Called by the archiver when saving TABLE DATA (not schema). This routine
289
* should save whatever format-specific information is needed to read
292
* It is called just prior to the dumper's 'DataDumper' routine being called.
294
* Optional, but strongly recommended.
298
_StartData(ArchiveHandle *AH, TocEntry *te)
300
lclContext *ctx = (lclContext *) AH->formatData;
301
lclTocEntry *tctx = (lclTocEntry *) te->formatData;
303
tctx->dataPos = _getFilePos(AH, ctx);
304
tctx->dataState = K_OFFSET_POS_SET;
306
_WriteByte(AH, BLK_DATA); /* Block type */
307
WriteInt(AH, te->dumpId); /* For sanity check */
309
_StartDataCompressor(AH, te);
313
* Called by archiver when dumper calls WriteData. This routine is
314
* called for both BLOB and TABLE data; it is the responsibility of
315
* the format to manage each kind of data using StartBlob/StartData.
317
* It should only be called from withing a DataDumper routine.
323
_WriteData(ArchiveHandle *AH, const void *data, size_t dLen)
325
lclContext *ctx = (lclContext *) AH->formatData;
326
z_streamp zp = ctx->zp;
328
zp->next_in = (void *) data;
331
while (zp->avail_in != 0)
333
/* printf("Deflating %lu bytes\n", (unsigned long) dLen); */
334
_DoDeflate(AH, ctx, 0);
340
* Called by the archiver when a dumper's 'DataDumper' routine has
347
_EndData(ArchiveHandle *AH, TocEntry *te)
349
/* lclContext *ctx = (lclContext *) AH->formatData; */
350
/* lclTocEntry *tctx = (lclTocEntry *) te->formatData; */
352
_EndDataCompressor(AH, te);
356
* Called by the archiver when starting to save all BLOB DATA (not schema).
357
* This routine should save whatever format-specific information is needed
358
* to read the BLOBs back into memory.
360
* It is called just prior to the dumper's DataDumper routine.
362
* Optional, but strongly recommended.
366
_StartBlobs(ArchiveHandle *AH, TocEntry *te)
368
lclContext *ctx = (lclContext *) AH->formatData;
369
lclTocEntry *tctx = (lclTocEntry *) te->formatData;
371
tctx->dataPos = _getFilePos(AH, ctx);
372
tctx->dataState = K_OFFSET_POS_SET;
374
_WriteByte(AH, BLK_BLOBS); /* Block type */
375
WriteInt(AH, te->dumpId); /* For sanity check */
379
* Called by the archiver when the dumper calls StartBlob.
383
* Must save the passed OID for retrieval at restore-time.
386
_StartBlob(ArchiveHandle *AH, TocEntry *te, Oid oid)
389
die_horribly(AH, modulename, "invalid OID for large object\n");
392
_StartDataCompressor(AH, te);
396
* Called by the archiver when the dumper calls EndBlob.
402
_EndBlob(ArchiveHandle *AH, TocEntry *te, Oid oid)
404
_EndDataCompressor(AH, te);
408
* Called by the archiver when finishing saving all BLOB DATA.
414
_EndBlobs(ArchiveHandle *AH, TocEntry *te)
416
/* Write out a fake zero OID to mark end-of-blobs. */
421
* Print data for a given TOC entry
424
_PrintTocData(ArchiveHandle *AH, TocEntry *te, RestoreOptions *ropt)
426
lclContext *ctx = (lclContext *) AH->formatData;
428
lclTocEntry *tctx = (lclTocEntry *) te->formatData;
432
if (tctx->dataState == K_OFFSET_NO_DATA)
435
if (!ctx->hasSeek || tctx->dataState == K_OFFSET_POS_NOT_SET)
437
/* Skip over unnecessary blocks until we get the one we want. */
441
_readBlockHeader(AH, &blkType, &id);
443
while (id != te->dumpId)
445
if ((TocIDRequired(AH, id, ropt) & REQ_DATA) != 0)
446
die_horribly(AH, modulename,
447
"Dumping a specific TOC data block out of order is not supported"
448
" without ID on this input stream (fseek required)\n");
460
default: /* Always have a default */
461
die_horribly(AH, modulename,
462
"unrecognized data block type (%d) while searching archive\n",
466
_readBlockHeader(AH, &blkType, &id);
472
if (fseeko(AH->FH, tctx->dataPos, SEEK_SET) != 0)
473
die_horribly(AH, modulename, "error during file seek: %s\n", strerror(errno));
475
_readBlockHeader(AH, &blkType, &id);
479
if (id != te->dumpId)
480
die_horribly(AH, modulename, "found unexpected block ID (%d) when reading data -- expected %d\n",
491
die_horribly(AH, modulename, "large objects cannot be loaded without a database connection\n");
496
default: /* Always have a default */
497
die_horribly(AH, modulename, "unrecognized data block type %d while restoring archive\n",
504
* Print data from current file position.
507
_PrintData(ArchiveHandle *AH)
509
lclContext *ctx = (lclContext *) AH->formatData;
510
z_streamp zp = ctx->zp;
512
char *in = ctx->zlibIn;
517
char *out = ctx->zlibOut;
524
if (AH->compression != 0)
530
if (inflateInit(zp) != Z_OK)
531
die_horribly(AH, modulename, "could not initialize compression library: %s\n", zp->msg);
535
blkLen = ReadInt(AH);
538
if (blkLen + 1 > ctx->inSize)
542
ctx->zlibIn = (char *) malloc(blkLen + 1);
544
die_horribly(AH, modulename, "out of memory\n");
546
ctx->inSize = blkLen + 1;
550
cnt = fread(in, 1, blkLen, AH->FH);
552
die_horribly(AH, modulename,
553
"could not read data block -- expected %lu, got %lu\n",
554
(unsigned long) blkLen, (unsigned long) cnt);
556
ctx->filePos += blkLen;
559
zp->avail_in = blkLen;
563
if (AH->compression != 0)
565
while (zp->avail_in != 0)
568
zp->avail_out = zlibOutSize;
569
res = inflate(zp, 0);
570
if (res != Z_OK && res != Z_STREAM_END)
571
die_horribly(AH, modulename, "could not uncompress data: %s\n", zp->msg);
573
out[zlibOutSize - zp->avail_out] = '\0';
574
ahwrite(out, 1, zlibOutSize - zp->avail_out, AH);
580
in[zp->avail_in] = '\0';
581
ahwrite(in, 1, zp->avail_in, AH);
587
blkLen = ReadInt(AH);
591
if (AH->compression != 0)
595
while (res != Z_STREAM_END)
598
zp->avail_out = zlibOutSize;
599
res = inflate(zp, 0);
600
if (res != Z_OK && res != Z_STREAM_END)
601
die_horribly(AH, modulename, "could not uncompress data: %s\n", zp->msg);
603
out[zlibOutSize - zp->avail_out] = '\0';
604
ahwrite(out, 1, zlibOutSize - zp->avail_out, AH);
606
if (inflateEnd(zp) != Z_OK)
607
die_horribly(AH, modulename, "could not close compression library: %s\n", zp->msg);
613
_LoadBlobs(ArchiveHandle *AH)
617
StartRestoreBlobs(AH);
622
StartRestoreBlob(AH, oid);
624
EndRestoreBlob(AH, oid);
632
* Skip the BLOBs from the current file position.
633
* BLOBS are written sequentially as data blocks (see below).
634
* Each BLOB is preceded by it's original OID.
635
* A zero OID indicated the end of the BLOBS
638
_skipBlobs(ArchiveHandle *AH)
651
* Skip data from current file position.
652
* Data blocks are formatted as an integer length, followed by data.
653
* A zero length denoted the end of the block.
656
_skipData(ArchiveHandle *AH)
658
lclContext *ctx = (lclContext *) AH->formatData;
660
char *in = ctx->zlibIn;
663
blkLen = ReadInt(AH);
666
if (blkLen > ctx->inSize)
669
ctx->zlibIn = (char *) malloc(blkLen);
670
ctx->inSize = blkLen;
673
cnt = fread(in, 1, blkLen, AH->FH);
675
die_horribly(AH, modulename,
676
"could not read data block -- expected %lu, got %lu\n",
677
(unsigned long) blkLen, (unsigned long) cnt);
679
ctx->filePos += blkLen;
681
blkLen = ReadInt(AH);
686
* Write a byte of data to the archive.
690
* Called by the archiver to do integer & byte output to the archive.
691
* These routines are only used to read & write headers & TOC.
695
_WriteByte(ArchiveHandle *AH, const int i)
697
lclContext *ctx = (lclContext *) AH->formatData;
700
res = fputc(i, AH->FH);
704
die_horribly(AH, modulename, "could not write byte: %s\n", strerror(errno));
709
* Read a byte of data from the archive.
713
* Called by the archiver to read bytes & integers from the archive.
714
* These routines are only used to read & write headers & TOC.
718
_ReadByte(ArchiveHandle *AH)
720
lclContext *ctx = (lclContext *) AH->formatData;
730
* Write a buffer of data to the archive.
734
* Called by the archiver to write a block of bytes to the archive.
735
* These routines are only used to read & write headers & TOC.
739
_WriteBuf(ArchiveHandle *AH, const void *buf, size_t len)
741
lclContext *ctx = (lclContext *) AH->formatData;
744
res = fwrite(buf, 1, len, AH->FH);
747
die_horribly(AH, modulename,
748
"write error in _WriteBuf (%lu != %lu)\n",
749
(unsigned long) res, (unsigned long) len);
756
* Read a block of bytes from the archive.
760
* Called by the archiver to read a block of bytes from the archive
761
* These routines are only used to read & write headers & TOC.
765
_ReadBuf(ArchiveHandle *AH, void *buf, size_t len)
767
lclContext *ctx = (lclContext *) AH->formatData;
770
res = fread(buf, 1, len, AH->FH);
781
* When writing the archive, this is the routine that actually starts
782
* the process of saving it to files. No data should be written prior
783
* to this point, since the user could sort the TOC after creating it.
785
* If an archive is to be written, this toutine must call:
786
* WriteHead to save the archive header
787
* WriteToc to save the TOC entries
788
* WriteDataChunks to save all DATA & BLOBs.
792
_CloseArchive(ArchiveHandle *AH)
794
lclContext *ctx = (lclContext *) AH->formatData;
797
if (AH->mode == archModeWrite)
800
tpos = ftello(AH->FH);
802
ctx->dataStart = _getFilePos(AH, ctx);
806
* This is not an essential operation - it is really only needed
807
* if we expect to be doing seeks to read the data back - it may
808
* be ok to just use the existing self-consistent block
813
fseeko(AH->FH, tpos, SEEK_SET);
818
if (fclose(AH->FH) != 0)
819
die_horribly(AH, modulename, "could not close archive file: %s\n", strerror(errno));
824
/*--------------------------------------------------
825
* END OF FORMAT CALLBACKS
826
*--------------------------------------------------
830
* Get the current position in the archive file.
833
_getFilePos(ArchiveHandle *AH, lclContext *ctx)
839
pos = ftello(AH->FH);
840
if (pos != ctx->filePos)
842
write_msg(modulename, "WARNING: ftell mismatch with expected position -- ftell used\n");
845
* Prior to 1.7 (pg7.3) we relied on the internally maintained
846
* pointer. Now we rely on off_t always. pos = ctx->filePos;
856
* Read a data block header. The format changed in V1.3, so we
857
* put the code here for simplicity.
860
_readBlockHeader(ArchiveHandle *AH, int *type, int *id)
862
if (AH->version < K_VERS_1_3)
865
*type = _ReadByte(AH);
871
* If zlib is available, then startit up. This is called from
872
* StartData & StartBlob. The buffers are setup in the Init routine.
876
_StartDataCompressor(ArchiveHandle *AH, TocEntry *te)
878
lclContext *ctx = (lclContext *) AH->formatData;
879
z_streamp zp = ctx->zp;
883
if (AH->compression < 0 || AH->compression > 9)
884
AH->compression = Z_DEFAULT_COMPRESSION;
886
if (AH->compression != 0)
892
if (deflateInit(zp, AH->compression) != Z_OK)
893
die_horribly(AH, modulename, "could not initialize compression library: %s\n", zp->msg);
901
/* Just be paranoid - maybe End is called after Start, with no Write */
902
zp->next_out = ctx->zlibOut;
903
zp->avail_out = zlibOutSize;
907
* Send compressed data to the output stream (via ahwrite).
908
* Each data chunk is preceded by it's length.
909
* In the case of Z0, or no zlib, just write the raw data.
913
_DoDeflate(ArchiveHandle *AH, lclContext *ctx, int flush)
915
z_streamp zp = ctx->zp;
918
char *out = ctx->zlibOut;
921
if (AH->compression != 0)
923
res = deflate(zp, flush);
924
if (res == Z_STREAM_ERROR)
925
die_horribly(AH, modulename, "could not compress data: %s\n", zp->msg);
927
if (((flush == Z_FINISH) && (zp->avail_out < zlibOutSize))
928
|| (zp->avail_out == 0)
929
|| (zp->avail_in != 0)
933
* Extra paranoia: avoid zero-length chunks since a zero
934
* length chunk is the EOF marker. This should never happen
937
if (zp->avail_out < zlibOutSize)
940
* printf("Wrote %lu byte deflated chunk\n", (unsigned
941
* long) (zlibOutSize - zp->avail_out));
943
WriteInt(AH, zlibOutSize - zp->avail_out);
944
if (fwrite(out, 1, zlibOutSize - zp->avail_out, AH->FH) != (zlibOutSize - zp->avail_out))
945
die_horribly(AH, modulename, "could not write compressed chunk\n");
946
ctx->filePos += zlibOutSize - zp->avail_out;
949
zp->avail_out = zlibOutSize;
955
if (zp->avail_in > 0)
957
WriteInt(AH, zp->avail_in);
958
if (fwrite(zp->next_in, 1, zp->avail_in, AH->FH) != zp->avail_in)
959
die_horribly(AH, modulename, "could not write uncompressed chunk\n");
960
ctx->filePos += zp->avail_in;
966
if (flush == Z_FINISH)
980
* Terminate zlib context and flush it's buffers. If no zlib
985
_EndDataCompressor(ArchiveHandle *AH, TocEntry *te)
989
lclContext *ctx = (lclContext *) AH->formatData;
990
z_streamp zp = ctx->zp;
993
if (AH->compression != 0)
1000
/* printf("Ending data output\n"); */
1001
res = _DoDeflate(AH, ctx, Z_FINISH);
1002
} while (res != Z_STREAM_END);
1004
if (deflateEnd(zp) != Z_OK)
1005
die_horribly(AH, modulename, "could not close compression stream: %s\n", zp->msg);
1009
/* Send the end marker */