Age Owner Branch data TLA Line data Source code
1 : : /*-------------------------------------------------------------------------
2 : : *
3 : : * basebackup.c
4 : : * code for taking a base backup and streaming it to a standby
5 : : *
6 : : * Portions Copyright (c) 2010-2024, PostgreSQL Global Development Group
7 : : *
8 : : * IDENTIFICATION
9 : : * src/backend/backup/basebackup.c
10 : : *
11 : : *-------------------------------------------------------------------------
12 : : */
13 : : #include "postgres.h"
14 : :
15 : : #include <sys/stat.h>
16 : : #include <unistd.h>
17 : : #include <time.h>
18 : :
19 : : #include "access/xlog_internal.h"
20 : : #include "access/xlogbackup.h"
21 : : #include "backup/backup_manifest.h"
22 : : #include "backup/basebackup.h"
23 : : #include "backup/basebackup_incremental.h"
24 : : #include "backup/basebackup_sink.h"
25 : : #include "backup/basebackup_target.h"
26 : : #include "catalog/pg_tablespace_d.h"
27 : : #include "commands/defrem.h"
28 : : #include "common/compression.h"
29 : : #include "common/file_perm.h"
30 : : #include "common/file_utils.h"
31 : : #include "lib/stringinfo.h"
32 : : #include "miscadmin.h"
33 : : #include "nodes/pg_list.h"
34 : : #include "pgstat.h"
35 : : #include "pgtar.h"
36 : : #include "port.h"
37 : : #include "postmaster/syslogger.h"
38 : : #include "postmaster/walsummarizer.h"
39 : : #include "replication/walsender.h"
40 : : #include "replication/walsender_private.h"
41 : : #include "storage/bufpage.h"
42 : : #include "storage/checksum.h"
43 : : #include "storage/dsm_impl.h"
44 : : #include "storage/ipc.h"
45 : : #include "storage/reinit.h"
46 : : #include "utils/builtins.h"
47 : : #include "utils/guc.h"
48 : : #include "utils/ps_status.h"
49 : : #include "utils/relcache.h"
50 : : #include "utils/resowner.h"
51 : :
52 : : /*
53 : : * How much data do we want to send in one CopyData message? Note that
54 : : * this may also result in reading the underlying files in chunks of this
55 : : * size.
56 : : *
57 : : * NB: The buffer size is required to be a multiple of the system block
58 : : * size, so use that value instead if it's bigger than our preference.
59 : : */
60 : : #define SINK_BUFFER_LENGTH Max(32768, BLCKSZ)
61 : :
62 : : typedef struct
63 : : {
64 : : const char *label;
65 : : bool progress;
66 : : bool fastcheckpoint;
67 : : bool nowait;
68 : : bool includewal;
69 : : bool incremental;
70 : : uint32 maxrate;
71 : : bool sendtblspcmapfile;
72 : : bool send_to_client;
73 : : bool use_copytblspc;
74 : : BaseBackupTargetHandle *target_handle;
75 : : backup_manifest_option manifest;
76 : : pg_compress_algorithm compression;
77 : : pg_compress_specification compression_specification;
78 : : pg_checksum_type manifest_checksum_type;
79 : : } basebackup_options;
80 : :
81 : : static int64 sendTablespace(bbsink *sink, char *path, Oid spcoid, bool sizeonly,
82 : : struct backup_manifest_info *manifest,
83 : : IncrementalBackupInfo *ib);
84 : : static int64 sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly,
85 : : List *tablespaces, bool sendtblspclinks,
86 : : backup_manifest_info *manifest, Oid spcoid,
87 : : IncrementalBackupInfo *ib);
88 : : static bool sendFile(bbsink *sink, const char *readfilename, const char *tarfilename,
89 : : struct stat *statbuf, bool missing_ok,
90 : : Oid dboid, Oid spcoid, RelFileNumber relfilenumber,
91 : : unsigned segno,
92 : : backup_manifest_info *manifest,
93 : : unsigned num_incremental_blocks,
94 : : BlockNumber *incremental_blocks,
95 : : unsigned truncation_block_length);
96 : : static off_t read_file_data_into_buffer(bbsink *sink,
97 : : const char *readfilename, int fd,
98 : : off_t offset, size_t length,
99 : : BlockNumber blkno,
100 : : bool verify_checksum,
101 : : int *checksum_failures);
102 : : static void push_to_sink(bbsink *sink, pg_checksum_context *checksum_ctx,
103 : : size_t *bytes_done, void *data, size_t length);
104 : : static bool verify_page_checksum(Page page, XLogRecPtr start_lsn,
105 : : BlockNumber blkno,
106 : : uint16 *expected_checksum);
107 : : static void sendFileWithContent(bbsink *sink, const char *filename,
108 : : const char *content, int len,
109 : : backup_manifest_info *manifest);
110 : : static int64 _tarWriteHeader(bbsink *sink, const char *filename,
111 : : const char *linktarget, struct stat *statbuf,
112 : : bool sizeonly);
113 : : static void _tarWritePadding(bbsink *sink, int len);
114 : : static void convert_link_to_directory(const char *pathbuf, struct stat *statbuf);
115 : : static void perform_base_backup(basebackup_options *opt, bbsink *sink,
116 : : IncrementalBackupInfo *ib);
117 : : static void parse_basebackup_options(List *options, basebackup_options *opt);
118 : : static int compareWalFileNames(const ListCell *a, const ListCell *b);
119 : : static ssize_t basebackup_read_file(int fd, char *buf, size_t nbytes, off_t offset,
120 : : const char *filename, bool partial_read_ok);
121 : :
122 : : /* Was the backup currently in-progress initiated in recovery mode? */
123 : : static bool backup_started_in_recovery = false;
124 : :
125 : : /* Total number of checksum failures during base backup. */
126 : : static long long int total_checksum_failures;
127 : :
128 : : /* Do not verify checksums. */
129 : : static bool noverify_checksums = false;
130 : :
131 : : /*
132 : : * Definition of one element part of an exclusion list, used for paths part
133 : : * of checksum validation or base backups. "name" is the name of the file
134 : : * or path to check for exclusion. If "match_prefix" is true, any items
135 : : * matching the name as prefix are excluded.
136 : : */
137 : : struct exclude_list_item
138 : : {
139 : : const char *name;
140 : : bool match_prefix;
141 : : };
142 : :
143 : : /*
144 : : * The contents of these directories are removed or recreated during server
145 : : * start so they are not included in backups. The directories themselves are
146 : : * kept and included as empty to preserve access permissions.
147 : : *
148 : : * Note: this list should be kept in sync with the filter lists in pg_rewind's
149 : : * filemap.c.
150 : : */
151 : : static const char *const excludeDirContents[] =
152 : : {
153 : : /*
154 : : * Skip temporary statistics files. PG_STAT_TMP_DIR must be skipped
155 : : * because extensions like pg_stat_statements store data there.
156 : : */
157 : : PG_STAT_TMP_DIR,
158 : :
159 : : /*
160 : : * It is generally not useful to backup the contents of this directory
161 : : * even if the intention is to restore to another primary. See backup.sgml
162 : : * for a more detailed description.
163 : : */
164 : : "pg_replslot",
165 : :
166 : : /* Contents removed on startup, see dsm_cleanup_for_mmap(). */
167 : : PG_DYNSHMEM_DIR,
168 : :
169 : : /* Contents removed on startup, see AsyncShmemInit(). */
170 : : "pg_notify",
171 : :
172 : : /*
173 : : * Old contents are loaded for possible debugging but are not required for
174 : : * normal operation, see SerialInit().
175 : : */
176 : : "pg_serial",
177 : :
178 : : /* Contents removed on startup, see DeleteAllExportedSnapshotFiles(). */
179 : : "pg_snapshots",
180 : :
181 : : /* Contents zeroed on startup, see StartupSUBTRANS(). */
182 : : "pg_subtrans",
183 : :
184 : : /* end of list */
185 : : NULL
186 : : };
187 : :
188 : : /*
189 : : * List of files excluded from backups.
190 : : */
191 : : static const struct exclude_list_item excludeFiles[] =
192 : : {
193 : : /* Skip auto conf temporary file. */
194 : : {PG_AUTOCONF_FILENAME ".tmp", false},
195 : :
196 : : /* Skip current log file temporary file */
197 : : {LOG_METAINFO_DATAFILE_TMP, false},
198 : :
199 : : /*
200 : : * Skip relation cache because it is rebuilt on startup. This includes
201 : : * temporary files.
202 : : */
203 : : {RELCACHE_INIT_FILENAME, true},
204 : :
205 : : /*
206 : : * backup_label and tablespace_map should not exist in a running cluster
207 : : * capable of doing an online backup, but exclude them just in case.
208 : : */
209 : : {BACKUP_LABEL_FILE, false},
210 : : {TABLESPACE_MAP, false},
211 : :
212 : : /*
213 : : * If there's a backup_manifest, it belongs to a backup that was used to
214 : : * start this server. It is *not* correct for this backup. Our
215 : : * backup_manifest is injected into the backup separately if users want
216 : : * it.
217 : : */
218 : : {"backup_manifest", false},
219 : :
220 : : {"postmaster.pid", false},
221 : : {"postmaster.opts", false},
222 : :
223 : : /* end of list */
224 : : {NULL, false}
225 : : };
226 : :
227 : : /*
228 : : * Actually do a base backup for the specified tablespaces.
229 : : *
230 : : * This is split out mainly to avoid complaints about "variable might be
231 : : * clobbered by longjmp" from stupider versions of gcc.
232 : : */
233 : : static void
116 rhaas@postgresql.org 234 :GNC 149 : perform_base_backup(basebackup_options *opt, bbsink *sink,
235 : : IncrementalBackupInfo *ib)
236 : : {
237 : : bbsink_state state;
238 : : XLogRecPtr endptr;
239 : : TimeLineID endtli;
240 : : backup_manifest_info manifest;
241 : : BackupState *backup_state;
242 : : StringInfo tablespace_map;
243 : :
244 : : /* Initial backup state, insofar as we know it now. */
891 rhaas@postgresql.org 245 :CBC 149 : state.tablespaces = NIL;
246 : 149 : state.tablespace_num = 0;
247 : 149 : state.bytes_done = 0;
248 : 149 : state.bytes_total = 0;
249 : 149 : state.bytes_total_is_valid = false;
250 : :
251 : : /* we're going to use a BufFile, so we need a ResourceOwner */
1472 252 [ - + ]: 149 : Assert(CurrentResourceOwner == NULL);
253 : 149 : CurrentResourceOwner = ResourceOwnerCreate(NULL, "base backup");
254 : :
4140 heikki.linnakangas@i 255 : 149 : backup_started_in_recovery = RecoveryInProgress();
256 : :
1452 rhaas@postgresql.org 257 : 149 : InitializeBackupManifest(&manifest, opt->manifest,
258 : : opt->manifest_checksum_type);
259 : :
2203 magnus@hagander.net 260 : 149 : total_checksum_failures = 0;
261 : :
262 : : /* Allocate backup related variables. */
566 michael@paquier.xyz 263 : 149 : backup_state = (BackupState *) palloc0(sizeof(BackupState));
264 : 149 : tablespace_map = makeStringInfo();
265 : :
891 rhaas@postgresql.org 266 : 149 : basebackup_progress_wait_checkpoint();
566 michael@paquier.xyz 267 : 149 : do_pg_backup_start(opt->label, opt->fastcheckpoint, &state.tablespaces,
268 : : backup_state, tablespace_map);
269 : :
270 : 149 : state.startptr = backup_state->startpoint;
271 : 149 : state.starttli = backup_state->starttli;
272 : :
273 : : /*
274 : : * Once do_pg_backup_start has been called, ensure that any failure causes
275 : : * us to abort the backup so we don't "leak" a backup counter. For this
276 : : * reason, *all* functionality between do_pg_backup_start() and the end of
277 : : * do_pg_backup_stop() should be inside the error cleanup block!
278 : : */
279 : :
1578 rhaas@postgresql.org 280 [ + + ]: 149 : PG_ENSURE_ERROR_CLEANUP(do_pg_abort_backup, BoolGetDatum(false));
281 : : {
282 : : ListCell *lc;
283 : : tablespaceinfo *newti;
284 : :
285 : : /* If this is an incremental backup, execute preparatory steps. */
116 rhaas@postgresql.org 286 [ + + ]:GNC 149 : if (ib != NULL)
287 : 7 : PrepareForIncrementalBackup(ib, backup_state);
288 : :
289 : : /* Add a node for the base directory at the end */
557 drowley@postgresql.o 290 :CBC 149 : newti = palloc0(sizeof(tablespaceinfo));
291 : 149 : newti->size = -1;
292 : 149 : state.tablespaces = lappend(state.tablespaces, newti);
293 : :
294 : : /*
295 : : * Calculate the total backup size by summing up the size of each
296 : : * tablespace
297 : : */
1503 fujii@postgresql.org 298 [ + - ]: 149 : if (opt->progress)
299 : : {
891 rhaas@postgresql.org 300 : 149 : basebackup_progress_estimate_backup_size();
301 : :
302 [ + - + + : 329 : foreach(lc, state.tablespaces)
+ + ]
303 : : {
1503 fujii@postgresql.org 304 : 180 : tablespaceinfo *tmp = (tablespaceinfo *) lfirst(lc);
305 : :
1397 rhaas@postgresql.org 306 [ + + ]: 180 : if (tmp->path == NULL)
891 307 : 149 : tmp->size = sendDir(sink, ".", 1, true, state.tablespaces,
308 : : true, NULL, InvalidOid, NULL);
309 : : else
310 : 31 : tmp->size = sendTablespace(sink, tmp->path, tmp->oid, true,
311 : : NULL, NULL);
312 : 180 : state.bytes_total += tmp->size;
313 : : }
314 : 149 : state.bytes_total_is_valid = true;
315 : : }
316 : :
317 : : /* notify basebackup sink about start of backup */
318 : 149 : bbsink_begin_backup(sink, &state, SINK_BUFFER_LENGTH);
319 : :
320 : : /* Send off our tablespaces one by one */
321 [ + - + + : 323 : foreach(lc, state.tablespaces)
+ + ]
322 : : {
4842 tgl@sss.pgh.pa.us 323 : 180 : tablespaceinfo *ti = (tablespaceinfo *) lfirst(lc);
324 : :
4463 simon@2ndQuadrant.co 325 [ + + ]: 180 : if (ti->path == NULL)
326 : : {
327 : : struct stat statbuf;
1068 tgl@sss.pgh.pa.us 328 : 149 : bool sendtblspclinks = true;
329 : : char *backup_label;
330 : :
891 rhaas@postgresql.org 331 : 149 : bbsink_begin_archive(sink, "base.tar");
332 : :
333 : : /* In the main tar, include the backup_label first... */
566 michael@paquier.xyz 334 : 149 : backup_label = build_backup_content(backup_state, false);
335 : 149 : sendFileWithContent(sink, BACKUP_LABEL_FILE,
336 : : backup_label, -1, &manifest);
337 : 149 : pfree(backup_label);
338 : :
339 : : /* Then the tablespace_map file, if required... */
1397 rhaas@postgresql.org 340 [ + + ]: 149 : if (opt->sendtblspcmapfile)
341 : : {
566 michael@paquier.xyz 342 : 24 : sendFileWithContent(sink, TABLESPACE_MAP,
153 michael@paquier.xyz 343 :GNC 24 : tablespace_map->data, -1, &manifest);
1397 rhaas@postgresql.org 344 :CBC 24 : sendtblspclinks = false;
345 : : }
346 : :
347 : : /* Then the bulk of the files... */
891 348 : 149 : sendDir(sink, ".", 1, false, state.tablespaces,
349 : : sendtblspclinks, &manifest, InvalidOid, ib);
350 : :
351 : : /* ... and pg_control after everything else. */
4463 simon@2ndQuadrant.co 352 [ - + ]: 143 : if (lstat(XLOG_CONTROL_FILE, &statbuf) != 0)
4463 simon@2ndQuadrant.co 353 [ # # ]:UBC 0 : ereport(ERROR,
354 : : (errcode_for_file_access(),
355 : : errmsg("could not stat file \"%s\": %m",
356 : : XLOG_CONTROL_FILE)));
891 rhaas@postgresql.org 357 :CBC 143 : sendFile(sink, XLOG_CONTROL_FILE, XLOG_CONTROL_FILE, &statbuf,
358 : : false, InvalidOid, InvalidOid,
359 : : InvalidRelFileNumber, 0, &manifest, 0, NULL, 0);
360 : : }
361 : : else
362 : : {
174 rhaas@postgresql.org 363 :GNC 31 : char *archive_name = psprintf("%u.tar", ti->oid);
364 : :
891 rhaas@postgresql.org 365 :CBC 31 : bbsink_begin_archive(sink, archive_name);
366 : :
116 rhaas@postgresql.org 367 :GNC 31 : sendTablespace(sink, ti->path, ti->oid, false, &manifest, ib);
368 : : }
369 : :
370 : : /*
371 : : * If we're including WAL, and this is the main data directory we
372 : : * don't treat this as the end of the tablespace. Instead, we will
373 : : * include the xlog files below and stop afterwards. This is safe
374 : : * since the main data directory is always sent *last*.
375 : : */
4823 magnus@hagander.net 376 [ + + + - ]:CBC 174 : if (opt->includewal && ti->path == NULL)
377 : : {
891 rhaas@postgresql.org 378 [ - + ]: 20 : Assert(lnext(state.tablespaces, lc) == NULL);
379 : : }
380 : : else
381 : : {
382 : : /* Properly terminate the tarfile. */
383 : : StaticAssertDecl(2 * TAR_BLOCK_SIZE <= BLCKSZ,
384 : : "BLCKSZ too small for 2 tar blocks");
887 385 : 154 : memset(sink->bbs_buffer, 0, 2 * TAR_BLOCK_SIZE);
386 : 154 : bbsink_archive_contents(sink, 2 * TAR_BLOCK_SIZE);
387 : :
388 : : /* OK, that's the end of the archive. */
891 389 : 154 : bbsink_end_archive(sink);
390 : : }
391 : : }
392 : :
393 : 143 : basebackup_progress_wait_wal_archive(&state);
566 michael@paquier.xyz 394 : 143 : do_pg_backup_stop(backup_state, !opt->nowait);
395 : :
396 : 143 : endptr = backup_state->stoppoint;
397 : 143 : endtli = backup_state->stoptli;
398 : :
399 : : /* Deallocate backup-related variables. */
29 dgustafsson@postgres 400 :GNC 143 : destroyStringInfo(tablespace_map);
566 michael@paquier.xyz 401 :CBC 143 : pfree(backup_state);
402 : : }
1578 rhaas@postgresql.org 403 [ - + ]: 144 : PG_END_ENSURE_ERROR_CLEANUP(do_pg_abort_backup, BoolGetDatum(false));
404 : :
405 : :
4823 magnus@hagander.net 406 [ + + ]: 143 : if (opt->includewal)
407 : : {
408 : : /*
409 : : * We've left the last tar file "open", so we can now append the
410 : : * required WAL files to it.
411 : : */
412 : : char pathbuf[MAXPGPATH];
413 : : XLogSegNo segno;
414 : : XLogSegNo startsegno;
415 : : XLogSegNo endsegno;
416 : : struct stat statbuf;
4119 heikki.linnakangas@i 417 : 20 : List *historyFileList = NIL;
418 : 20 : List *walFileList = NIL;
419 : : char firstoff[MAXFNAMELEN];
420 : : char lastoff[MAXFNAMELEN];
421 : : DIR *dir;
422 : : struct dirent *de;
423 : : ListCell *lc;
424 : : TimeLineID tli;
425 : :
891 rhaas@postgresql.org 426 : 20 : basebackup_progress_transfer_wal();
427 : :
428 : : /*
429 : : * I'd rather not worry about timelines here, so scan pg_wal and
430 : : * include all WAL files in the range between 'startptr' and 'endptr',
431 : : * regardless of the timeline the file is stamped with. If there are
432 : : * some spurious WAL files belonging to timelines that don't belong in
433 : : * this server's history, they will be included too. Normally there
434 : : * shouldn't be such files, but if there are, there's little harm in
435 : : * including them.
436 : : */
437 : 20 : XLByteToSeg(state.startptr, startsegno, wal_segment_size);
438 : 20 : XLogFileName(firstoff, state.starttli, startsegno, wal_segment_size);
2399 andres@anarazel.de 439 : 20 : XLByteToPrevSeg(endptr, endsegno, wal_segment_size);
898 rhaas@postgresql.org 440 : 20 : XLogFileName(lastoff, endtli, endsegno, wal_segment_size);
441 : :
2733 442 : 20 : dir = AllocateDir("pg_wal");
443 [ + + ]: 143 : while ((de = ReadDir(dir, "pg_wal")) != NULL)
444 : : {
445 : : /* Does it look like a WAL segment, and is it in the range? */
3264 heikki.linnakangas@i 446 [ + + ]: 123 : if (IsXLogFileName(de->d_name) &&
4119 447 [ + - ]: 43 : strcmp(de->d_name + 8, firstoff + 8) >= 0 &&
448 [ + + ]: 43 : strcmp(de->d_name + 8, lastoff + 8) <= 0)
449 : : {
450 : 20 : walFileList = lappend(walFileList, pstrdup(de->d_name));
451 : : }
452 : : /* Does it look like a timeline history file? */
3264 453 [ - + ]: 103 : else if (IsTLHistoryFileName(de->d_name))
454 : : {
4119 heikki.linnakangas@i 455 :UBC 0 : historyFileList = lappend(historyFileList, pstrdup(de->d_name));
456 : : }
457 : : }
4119 heikki.linnakangas@i 458 :CBC 20 : FreeDir(dir);
459 : :
460 : : /*
461 : : * Before we go any further, check that none of the WAL segments we
462 : : * need were removed.
463 : : */
891 rhaas@postgresql.org 464 : 20 : CheckXLogRemoved(startsegno, state.starttli);
465 : :
466 : : /*
467 : : * Sort the WAL filenames. We want to send the files in order from
468 : : * oldest to newest, to reduce the chance that a file is recycled
469 : : * before we get a chance to send it over.
470 : : */
1734 tgl@sss.pgh.pa.us 471 : 20 : list_sort(walFileList, compareWalFileNames);
472 : :
473 : : /*
474 : : * There must be at least one xlog file in the pg_wal directory, since
475 : : * we are doing backup-including-xlog.
476 : : */
477 [ - + ]: 20 : if (walFileList == NIL)
3886 magnus@hagander.net 478 [ # # ]:UBC 0 : ereport(ERROR,
479 : : (errmsg("could not find any WAL files")));
480 : :
481 : : /*
482 : : * Sanity check: the first and last segment should cover startptr and
483 : : * endptr, with no gaps in between.
484 : : */
1734 tgl@sss.pgh.pa.us 485 :CBC 20 : XLogFromFileName((char *) linitial(walFileList),
486 : : &tli, &segno, wal_segment_size);
4119 heikki.linnakangas@i 487 [ - + ]: 20 : if (segno != startsegno)
488 : : {
489 : : char startfname[MAXFNAMELEN];
490 : :
891 rhaas@postgresql.org 491 :UBC 0 : XLogFileName(startfname, state.starttli, startsegno,
492 : : wal_segment_size);
4119 heikki.linnakangas@i 493 [ # # ]: 0 : ereport(ERROR,
494 : : (errmsg("could not find WAL file \"%s\"", startfname)));
495 : : }
1734 tgl@sss.pgh.pa.us 496 [ + - + + :CBC 40 : foreach(lc, walFileList)
+ + ]
497 : : {
498 : 20 : char *walFileName = (char *) lfirst(lc);
3973 bruce@momjian.us 499 : 20 : XLogSegNo currsegno = segno;
500 : 20 : XLogSegNo nextsegno = segno + 1;
501 : :
1734 tgl@sss.pgh.pa.us 502 : 20 : XLogFromFileName(walFileName, &tli, &segno, wal_segment_size);
4119 heikki.linnakangas@i 503 [ + - - + ]: 20 : if (!(nextsegno == segno || currsegno == segno))
504 : : {
505 : : char nextfname[MAXFNAMELEN];
506 : :
898 rhaas@postgresql.org 507 :UBC 0 : XLogFileName(nextfname, tli, nextsegno, wal_segment_size);
4119 heikki.linnakangas@i 508 [ # # ]: 0 : ereport(ERROR,
509 : : (errmsg("could not find WAL file \"%s\"", nextfname)));
510 : : }
511 : : }
4119 heikki.linnakangas@i 512 [ - + ]:CBC 20 : if (segno != endsegno)
513 : : {
514 : : char endfname[MAXFNAMELEN];
515 : :
898 rhaas@postgresql.org 516 :UBC 0 : XLogFileName(endfname, endtli, endsegno, wal_segment_size);
4119 heikki.linnakangas@i 517 [ # # ]: 0 : ereport(ERROR,
518 : : (errmsg("could not find WAL file \"%s\"", endfname)));
519 : : }
520 : :
521 : : /* Ok, we have everything we need. Send the WAL files. */
1734 tgl@sss.pgh.pa.us 522 [ + - + + :CBC 40 : foreach(lc, walFileList)
+ + ]
523 : : {
524 : 20 : char *walFileName = (char *) lfirst(lc);
525 : : int fd;
526 : : ssize_t cnt;
4119 heikki.linnakangas@i 527 : 20 : pgoff_t len = 0;
528 : :
1734 tgl@sss.pgh.pa.us 529 : 20 : snprintf(pathbuf, MAXPGPATH, XLOGDIR "/%s", walFileName);
530 : 20 : XLogFromFileName(walFileName, &tli, &segno, wal_segment_size);
531 : :
1397 rhaas@postgresql.org 532 : 20 : fd = OpenTransientFile(pathbuf, O_RDONLY | PG_BINARY);
533 [ - + ]: 20 : if (fd < 0)
534 : : {
2120 michael@paquier.xyz 535 :UBC 0 : int save_errno = errno;
536 : :
537 : : /*
538 : : * Most likely reason for this is that the file was already
539 : : * removed by a checkpoint, so check for that to get a better
540 : : * error message.
541 : : */
4119 heikki.linnakangas@i 542 : 0 : CheckXLogRemoved(segno, tli);
543 : :
2120 michael@paquier.xyz 544 : 0 : errno = save_errno;
4119 heikki.linnakangas@i 545 [ # # ]: 0 : ereport(ERROR,
546 : : (errcode_for_file_access(),
547 : : errmsg("could not open file \"%s\": %m", pathbuf)));
548 : : }
549 : :
1397 rhaas@postgresql.org 550 [ - + ]:CBC 20 : if (fstat(fd, &statbuf) != 0)
4119 heikki.linnakangas@i 551 [ # # ]:UBC 0 : ereport(ERROR,
552 : : (errcode_for_file_access(),
553 : : errmsg("could not stat file \"%s\": %m",
554 : : pathbuf)));
2399 andres@anarazel.de 555 [ - + ]:CBC 20 : if (statbuf.st_size != wal_segment_size)
556 : : {
4119 heikki.linnakangas@i 557 :UBC 0 : CheckXLogRemoved(segno, tli);
558 [ # # ]: 0 : ereport(ERROR,
559 : : (errcode_for_file_access(),
560 : : errmsg("unexpected WAL file size \"%s\"", walFileName)));
561 : : }
562 : :
563 : : /* send the WAL file itself */
891 rhaas@postgresql.org 564 :CBC 20 : _tarWriteHeader(sink, pathbuf, NULL, &statbuf, false);
565 : :
566 : 20 : while ((cnt = basebackup_read_file(fd, sink->bbs_buffer,
567 : 10240 : Min(sink->bbs_buffer_length,
568 : : wal_segment_size - len),
1397 569 [ + - ]: 10240 : len, pathbuf, true)) > 0)
570 : : {
4119 heikki.linnakangas@i 571 : 10240 : CheckXLogRemoved(segno, tli);
891 rhaas@postgresql.org 572 : 10240 : bbsink_archive_contents(sink, cnt);
573 : :
4119 heikki.linnakangas@i 574 : 10240 : len += cnt;
575 : :
2399 andres@anarazel.de 576 [ + + ]: 10240 : if (len == wal_segment_size)
4119 heikki.linnakangas@i 577 : 20 : break;
578 : : }
579 : :
2399 andres@anarazel.de 580 [ - + ]: 20 : if (len != wal_segment_size)
581 : : {
4119 heikki.linnakangas@i 582 :UBC 0 : CheckXLogRemoved(segno, tli);
583 [ # # ]: 0 : ereport(ERROR,
584 : : (errcode_for_file_access(),
585 : : errmsg("unexpected WAL file size \"%s\"", walFileName)));
586 : : }
587 : :
588 : : /*
589 : : * wal_segment_size is a multiple of TAR_BLOCK_SIZE, so no need
590 : : * for padding.
591 : : */
1451 rhaas@postgresql.org 592 [ - + ]:CBC 20 : Assert(wal_segment_size % TAR_BLOCK_SIZE == 0);
593 : :
1397 594 : 20 : CloseTransientFile(fd);
595 : :
596 : : /*
597 : : * Mark file as archived, otherwise files can get archived again
598 : : * after promotion of a new node. This is in line with
599 : : * walreceiver.c always doing an XLogArchiveForceDone() after a
600 : : * complete segment.
601 : : */
1734 tgl@sss.pgh.pa.us 602 : 20 : StatusFilePath(pathbuf, walFileName, ".done");
153 michael@paquier.xyz 603 :GNC 20 : sendFileWithContent(sink, pathbuf, "", -1, &manifest);
604 : : }
605 : :
606 : : /*
607 : : * Send timeline history files too. Only the latest timeline history
608 : : * file is required for recovery, and even that only if there happens
609 : : * to be a timeline switch in the first WAL segment that contains the
610 : : * checkpoint record, or if we're taking a base backup from a standby
611 : : * server and the target timeline changes while the backup is taken.
612 : : * But they are small and highly useful for debugging purposes, so
613 : : * better include them all, always.
614 : : */
4119 heikki.linnakangas@i 615 [ - + - - :CBC 20 : foreach(lc, historyFileList)
- + ]
616 : : {
3973 bruce@momjian.us 617 :UBC 0 : char *fname = lfirst(lc);
618 : :
4119 heikki.linnakangas@i 619 : 0 : snprintf(pathbuf, MAXPGPATH, XLOGDIR "/%s", fname);
620 : :
621 [ # # ]: 0 : if (lstat(pathbuf, &statbuf) != 0)
622 [ # # ]: 0 : ereport(ERROR,
623 : : (errcode_for_file_access(),
624 : : errmsg("could not stat file \"%s\": %m", pathbuf)));
625 : :
174 rhaas@postgresql.org 626 :UNC 0 : sendFile(sink, pathbuf, pathbuf, &statbuf, false,
627 : : InvalidOid, InvalidOid, InvalidRelFileNumber, 0,
628 : : &manifest, 0, NULL, 0);
629 : :
630 : : /* unconditionally mark file as archived */
3389 andres@anarazel.de 631 :UBC 0 : StatusFilePath(pathbuf, fname, ".done");
153 michael@paquier.xyz 632 :UNC 0 : sendFileWithContent(sink, pathbuf, "", -1, &manifest);
633 : : }
634 : :
635 : : /* Properly terminate the tar file. */
636 : : StaticAssertStmt(2 * TAR_BLOCK_SIZE <= BLCKSZ,
637 : : "BLCKSZ too small for 2 tar blocks");
887 rhaas@postgresql.org 638 :CBC 20 : memset(sink->bbs_buffer, 0, 2 * TAR_BLOCK_SIZE);
639 : 20 : bbsink_archive_contents(sink, 2 * TAR_BLOCK_SIZE);
640 : :
641 : : /* OK, that's the end of the archive. */
891 642 : 20 : bbsink_end_archive(sink);
643 : : }
644 : :
645 : 143 : AddWALInfoToBackupManifest(&manifest, state.startptr, state.starttli,
646 : : endptr, endtli);
647 : :
648 : 143 : SendBackupManifest(&manifest, sink);
649 : :
650 : 143 : bbsink_end_backup(sink, endptr, endtli);
651 : :
2203 magnus@hagander.net 652 [ + + ]: 143 : if (total_checksum_failures)
653 : : {
654 [ + + ]: 3 : if (total_checksum_failures > 1)
655 [ + - ]: 2 : ereport(WARNING,
656 : : (errmsg_plural("%lld total checksum verification failure",
657 : : "%lld total checksum verification failures",
658 : : total_checksum_failures,
659 : : total_checksum_failures)));
660 : :
661 [ + - ]: 3 : ereport(ERROR,
662 : : (errcode(ERRCODE_DATA_CORRUPTED),
663 : : errmsg("checksum verification failure during base backup")));
664 : : }
665 : :
666 : : /*
667 : : * Make sure to free the manifest before the resource owners as manifests
668 : : * use cryptohash contexts that may depend on resource owners (like
669 : : * OpenSSL).
670 : : */
1227 michael@paquier.xyz 671 : 140 : FreeBackupManifest(&manifest);
672 : :
673 : : /* clean up the resource owner we created */
1472 rhaas@postgresql.org 674 : 140 : WalSndResourceCleanup(true);
675 : :
891 676 : 140 : basebackup_progress_done();
4842 tgl@sss.pgh.pa.us 677 : 140 : }
678 : :
679 : : /*
680 : : * list_sort comparison function, to compare log/seg portion of WAL segment
681 : : * filenames, ignoring the timeline portion.
682 : : */
683 : : static int
1734 tgl@sss.pgh.pa.us 684 :UBC 0 : compareWalFileNames(const ListCell *a, const ListCell *b)
685 : : {
686 : 0 : char *fna = (char *) lfirst(a);
687 : 0 : char *fnb = (char *) lfirst(b);
688 : :
4119 heikki.linnakangas@i 689 : 0 : return strcmp(fna + 8, fnb + 8);
690 : : }
691 : :
692 : : /*
693 : : * Parse the base backup options passed down by the parser
694 : : */
695 : : static void
4830 magnus@hagander.net 696 :CBC 165 : parse_basebackup_options(List *options, basebackup_options *opt)
697 : : {
698 : : ListCell *lopt;
699 : 165 : bool o_label = false;
700 : 165 : bool o_progress = false;
922 rhaas@postgresql.org 701 : 165 : bool o_checkpoint = false;
4813 magnus@hagander.net 702 : 165 : bool o_nowait = false;
4823 703 : 165 : bool o_wal = false;
116 rhaas@postgresql.org 704 :GNC 165 : bool o_incremental = false;
3699 alvherre@alvh.no-ip. 705 :CBC 165 : bool o_maxrate = false;
3260 andrew@dunslane.net 706 : 165 : bool o_tablespace_map = false;
2203 magnus@hagander.net 707 : 165 : bool o_noverify_checksums = false;
1472 rhaas@postgresql.org 708 : 165 : bool o_manifest = false;
709 : 165 : bool o_manifest_checksums = false;
817 710 : 165 : bool o_target = false;
880 711 : 165 : bool o_target_detail = false;
761 712 : 165 : char *target_str = NULL;
713 : 165 : char *target_detail_str = NULL;
811 714 : 165 : bool o_compression = false;
753 715 : 165 : bool o_compression_detail = false;
716 : 165 : char *compression_detail_str = NULL;
717 : :
4828 magnus@hagander.net 718 [ + - + - : 1815 : MemSet(opt, 0, sizeof(*opt));
+ - + - +
+ ]
1472 rhaas@postgresql.org 719 : 165 : opt->manifest = MANIFEST_OPTION_NO;
720 : 165 : opt->manifest_checksum_type = CHECKSUM_TYPE_CRC32C;
733 michael@paquier.xyz 721 : 165 : opt->compression = PG_COMPRESSION_NONE;
722 : 165 : opt->compression_specification.algorithm = PG_COMPRESSION_NONE;
723 : :
4830 magnus@hagander.net 724 [ + - + + : 1255 : foreach(lopt, options)
+ + ]
725 : : {
726 : 1092 : DefElem *defel = (DefElem *) lfirst(lopt);
727 : :
728 [ + + ]: 1092 : if (strcmp(defel->defname, "label") == 0)
729 : : {
730 [ - + ]: 165 : if (o_label)
4830 magnus@hagander.net 731 [ # # ]:UBC 0 : ereport(ERROR,
732 : : (errcode(ERRCODE_SYNTAX_ERROR),
733 : : errmsg("duplicate option \"%s\"", defel->defname)));
922 rhaas@postgresql.org 734 :CBC 165 : opt->label = defGetString(defel);
4830 magnus@hagander.net 735 : 165 : o_label = true;
736 : : }
737 [ + + ]: 927 : else if (strcmp(defel->defname, "progress") == 0)
738 : : {
739 [ - + ]: 165 : if (o_progress)
4830 magnus@hagander.net 740 [ # # ]:UBC 0 : ereport(ERROR,
741 : : (errcode(ERRCODE_SYNTAX_ERROR),
742 : : errmsg("duplicate option \"%s\"", defel->defname)));
922 rhaas@postgresql.org 743 :CBC 165 : opt->progress = defGetBoolean(defel);
4830 magnus@hagander.net 744 : 165 : o_progress = true;
745 : : }
922 rhaas@postgresql.org 746 [ + + ]: 762 : else if (strcmp(defel->defname, "checkpoint") == 0)
747 : : {
748 : 155 : char *optval = defGetString(defel);
749 : :
750 [ - + ]: 155 : if (o_checkpoint)
4830 magnus@hagander.net 751 [ # # ]:UBC 0 : ereport(ERROR,
752 : : (errcode(ERRCODE_SYNTAX_ERROR),
753 : : errmsg("duplicate option \"%s\"", defel->defname)));
922 rhaas@postgresql.org 754 [ + - ]:CBC 155 : if (pg_strcasecmp(optval, "fast") == 0)
755 : 155 : opt->fastcheckpoint = true;
922 rhaas@postgresql.org 756 [ # # ]:UBC 0 : else if (pg_strcasecmp(optval, "spread") == 0)
757 : 0 : opt->fastcheckpoint = false;
758 : : else
759 [ # # ]: 0 : ereport(ERROR,
760 : : (errcode(ERRCODE_SYNTAX_ERROR),
761 : : errmsg("unrecognized checkpoint type: \"%s\"",
762 : : optval)));
922 rhaas@postgresql.org 763 :CBC 155 : o_checkpoint = true;
764 : : }
765 [ + + ]: 607 : else if (strcmp(defel->defname, "wait") == 0)
766 : : {
4813 magnus@hagander.net 767 [ - + ]: 158 : if (o_nowait)
4813 magnus@hagander.net 768 [ # # ]:UBC 0 : ereport(ERROR,
769 : : (errcode(ERRCODE_SYNTAX_ERROR),
770 : : errmsg("duplicate option \"%s\"", defel->defname)));
922 rhaas@postgresql.org 771 :CBC 158 : opt->nowait = !defGetBoolean(defel);
4813 magnus@hagander.net 772 : 158 : o_nowait = true;
773 : : }
4823 774 [ + + ]: 449 : else if (strcmp(defel->defname, "wal") == 0)
775 : : {
776 [ - + ]: 24 : if (o_wal)
4823 magnus@hagander.net 777 [ # # ]:UBC 0 : ereport(ERROR,
778 : : (errcode(ERRCODE_SYNTAX_ERROR),
779 : : errmsg("duplicate option \"%s\"", defel->defname)));
922 rhaas@postgresql.org 780 :CBC 24 : opt->includewal = defGetBoolean(defel);
4823 magnus@hagander.net 781 : 24 : o_wal = true;
782 : : }
116 rhaas@postgresql.org 783 [ + + ]:GNC 425 : else if (strcmp(defel->defname, "incremental") == 0)
784 : : {
785 [ - + ]: 7 : if (o_incremental)
116 rhaas@postgresql.org 786 [ # # ]:UNC 0 : ereport(ERROR,
787 : : (errcode(ERRCODE_SYNTAX_ERROR),
788 : : errmsg("duplicate option \"%s\"", defel->defname)));
116 rhaas@postgresql.org 789 :GNC 7 : opt->incremental = defGetBoolean(defel);
790 [ + - - + ]: 7 : if (opt->incremental && !summarize_wal)
116 rhaas@postgresql.org 791 [ # # ]:UNC 0 : ereport(ERROR,
792 : : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
793 : : errmsg("incremental backups cannot be taken unless WAL summarization is enabled")));
116 rhaas@postgresql.org 794 :GNC 7 : o_incremental = true;
795 : : }
3699 alvherre@alvh.no-ip. 796 [ + + ]:CBC 418 : else if (strcmp(defel->defname, "max_rate") == 0)
797 : : {
798 : : int64 maxrate;
799 : :
800 [ - + ]: 1 : if (o_maxrate)
3699 alvherre@alvh.no-ip. 801 [ # # ]:UBC 0 : ereport(ERROR,
802 : : (errcode(ERRCODE_SYNTAX_ERROR),
803 : : errmsg("duplicate option \"%s\"", defel->defname)));
804 : :
922 rhaas@postgresql.org 805 :CBC 1 : maxrate = defGetInt64(defel);
3699 alvherre@alvh.no-ip. 806 [ + - - + ]: 1 : if (maxrate < MAX_RATE_LOWER || maxrate > MAX_RATE_UPPER)
3699 alvherre@alvh.no-ip. 807 [ # # ]:UBC 0 : ereport(ERROR,
808 : : (errcode(ERRCODE_NUMERIC_VALUE_OUT_OF_RANGE),
809 : : errmsg("%d is outside the valid range for parameter \"%s\" (%d .. %d)",
810 : : (int) maxrate, "MAX_RATE", MAX_RATE_LOWER, MAX_RATE_UPPER)));
811 : :
3699 alvherre@alvh.no-ip. 812 :CBC 1 : opt->maxrate = (uint32) maxrate;
813 : 1 : o_maxrate = true;
814 : : }
3260 andrew@dunslane.net 815 [ + + ]: 417 : else if (strcmp(defel->defname, "tablespace_map") == 0)
816 : : {
817 [ - + ]: 29 : if (o_tablespace_map)
3260 andrew@dunslane.net 818 [ # # ]:UBC 0 : ereport(ERROR,
819 : : (errcode(ERRCODE_SYNTAX_ERROR),
820 : : errmsg("duplicate option \"%s\"", defel->defname)));
922 rhaas@postgresql.org 821 :CBC 29 : opt->sendtblspcmapfile = defGetBoolean(defel);
3260 andrew@dunslane.net 822 : 29 : o_tablespace_map = true;
823 : : }
922 rhaas@postgresql.org 824 [ + + ]: 388 : else if (strcmp(defel->defname, "verify_checksums") == 0)
825 : : {
2203 magnus@hagander.net 826 [ - + ]: 1 : if (o_noverify_checksums)
2203 magnus@hagander.net 827 [ # # ]:UBC 0 : ereport(ERROR,
828 : : (errcode(ERRCODE_SYNTAX_ERROR),
829 : : errmsg("duplicate option \"%s\"", defel->defname)));
922 rhaas@postgresql.org 830 :CBC 1 : noverify_checksums = !defGetBoolean(defel);
2203 magnus@hagander.net 831 : 1 : o_noverify_checksums = true;
832 : : }
1472 rhaas@postgresql.org 833 [ + + ]: 387 : else if (strcmp(defel->defname, "manifest") == 0)
834 : : {
922 835 : 164 : char *optval = defGetString(defel);
836 : : bool manifest_bool;
837 : :
1472 838 [ - + ]: 164 : if (o_manifest)
1472 rhaas@postgresql.org 839 [ # # ]:UBC 0 : ereport(ERROR,
840 : : (errcode(ERRCODE_SYNTAX_ERROR),
841 : : errmsg("duplicate option \"%s\"", defel->defname)));
1472 rhaas@postgresql.org 842 [ + + ]:CBC 164 : if (parse_bool(optval, &manifest_bool))
843 : : {
844 [ + - ]: 163 : if (manifest_bool)
845 : 163 : opt->manifest = MANIFEST_OPTION_YES;
846 : : else
1472 rhaas@postgresql.org 847 :UBC 0 : opt->manifest = MANIFEST_OPTION_NO;
848 : : }
1472 rhaas@postgresql.org 849 [ + - ]:CBC 1 : else if (pg_strcasecmp(optval, "force-encode") == 0)
850 : 1 : opt->manifest = MANIFEST_OPTION_FORCE_ENCODE;
851 : : else
1472 rhaas@postgresql.org 852 [ # # ]:UBC 0 : ereport(ERROR,
853 : : (errcode(ERRCODE_SYNTAX_ERROR),
854 : : errmsg("unrecognized manifest option: \"%s\"",
855 : : optval)));
1472 rhaas@postgresql.org 856 :CBC 164 : o_manifest = true;
857 : : }
858 [ + + ]: 223 : else if (strcmp(defel->defname, "manifest_checksums") == 0)
859 : : {
922 860 : 7 : char *optval = defGetString(defel);
861 : :
1472 862 [ - + ]: 7 : if (o_manifest_checksums)
1472 rhaas@postgresql.org 863 [ # # ]:UBC 0 : ereport(ERROR,
864 : : (errcode(ERRCODE_SYNTAX_ERROR),
865 : : errmsg("duplicate option \"%s\"", defel->defname)));
1472 rhaas@postgresql.org 866 [ + + ]:CBC 7 : if (!pg_checksum_parse_type(optval,
867 : : &opt->manifest_checksum_type))
868 [ + - ]: 1 : ereport(ERROR,
869 : : (errcode(ERRCODE_SYNTAX_ERROR),
870 : : errmsg("unrecognized checksum algorithm: \"%s\"",
871 : : optval)));
872 : 6 : o_manifest_checksums = true;
873 : : }
817 874 [ + + ]: 216 : else if (strcmp(defel->defname, "target") == 0)
875 : : {
876 [ - + ]: 164 : if (o_target)
817 rhaas@postgresql.org 877 [ # # ]:UBC 0 : ereport(ERROR,
878 : : (errcode(ERRCODE_SYNTAX_ERROR),
879 : : errmsg("duplicate option \"%s\"", defel->defname)));
761 rhaas@postgresql.org 880 :CBC 164 : target_str = defGetString(defel);
817 881 : 164 : o_target = true;
882 : : }
880 883 [ + + ]: 52 : else if (strcmp(defel->defname, "target_detail") == 0)
884 : : {
885 : 9 : char *optval = defGetString(defel);
886 : :
887 [ - + ]: 9 : if (o_target_detail)
880 rhaas@postgresql.org 888 [ # # ]:UBC 0 : ereport(ERROR,
889 : : (errcode(ERRCODE_SYNTAX_ERROR),
890 : : errmsg("duplicate option \"%s\"", defel->defname)));
761 rhaas@postgresql.org 891 :CBC 9 : target_detail_str = optval;
880 892 : 9 : o_target_detail = true;
893 : : }
811 894 [ + + ]: 43 : else if (strcmp(defel->defname, "compression") == 0)
895 : : {
896 : 29 : char *optval = defGetString(defel);
897 : :
898 [ - + ]: 29 : if (o_compression)
811 rhaas@postgresql.org 899 [ # # ]:UBC 0 : ereport(ERROR,
900 : : (errcode(ERRCODE_SYNTAX_ERROR),
901 : : errmsg("duplicate option \"%s\"", defel->defname)));
733 michael@paquier.xyz 902 [ + + ]:CBC 29 : if (!parse_compress_algorithm(optval, &opt->compression))
811 rhaas@postgresql.org 903 [ + - ]: 1 : ereport(ERROR,
904 : : (errcode(ERRCODE_SYNTAX_ERROR),
905 : : errmsg("unrecognized compression algorithm: \"%s\"",
906 : : optval)));
907 : 28 : o_compression = true;
908 : : }
753 909 [ + - ]: 14 : else if (strcmp(defel->defname, "compression_detail") == 0)
910 : : {
911 [ - + ]: 14 : if (o_compression_detail)
811 rhaas@postgresql.org 912 [ # # ]:UBC 0 : ereport(ERROR,
913 : : (errcode(ERRCODE_SYNTAX_ERROR),
914 : : errmsg("duplicate option \"%s\"", defel->defname)));
753 rhaas@postgresql.org 915 :CBC 14 : compression_detail_str = defGetString(defel);
916 : 14 : o_compression_detail = true;
917 : : }
918 : : else
811 rhaas@postgresql.org 919 [ # # ]:UBC 0 : ereport(ERROR,
920 : : (errcode(ERRCODE_SYNTAX_ERROR),
921 : : errmsg("unrecognized base backup option: \"%s\"",
922 : : defel->defname)));
923 : : }
924 : :
4830 magnus@hagander.net 925 [ - + ]:CBC 163 : if (opt->label == NULL)
4830 magnus@hagander.net 926 :UBC 0 : opt->label = "base backup";
1472 rhaas@postgresql.org 927 [ + + ]:CBC 163 : if (opt->manifest == MANIFEST_OPTION_NO)
928 : : {
929 [ - + ]: 1 : if (o_manifest_checksums)
1472 rhaas@postgresql.org 930 [ # # ]:UBC 0 : ereport(ERROR,
931 : : (errcode(ERRCODE_SYNTAX_ERROR),
932 : : errmsg("manifest checksums require a backup manifest")));
1472 rhaas@postgresql.org 933 :CBC 1 : opt->manifest_checksum_type = CHECKSUM_TYPE_NONE;
934 : : }
935 : :
761 936 [ - + ]: 163 : if (target_str == NULL)
937 : : {
761 rhaas@postgresql.org 938 [ # # ]:UBC 0 : if (target_detail_str != NULL)
880 939 [ # # ]: 0 : ereport(ERROR,
940 : : (errcode(ERRCODE_SYNTAX_ERROR),
941 : : errmsg("target detail cannot be used without target")));
761 942 : 0 : opt->use_copytblspc = true;
943 : 0 : opt->send_to_client = true;
944 : : }
761 rhaas@postgresql.org 945 [ + + ]:CBC 163 : else if (strcmp(target_str, "client") == 0)
946 : : {
947 [ - + ]: 148 : if (target_detail_str != NULL)
880 rhaas@postgresql.org 948 [ # # ]:UBC 0 : ereport(ERROR,
949 : : (errcode(ERRCODE_SYNTAX_ERROR),
950 : : errmsg("target \"%s\" does not accept a target detail",
951 : : target_str)));
761 rhaas@postgresql.org 952 :CBC 148 : opt->send_to_client = true;
953 : : }
954 : : else
955 : 13 : opt->target_handle =
956 : 15 : BaseBackupGetTargetHandle(target_str, target_detail_str);
957 : :
753 958 [ + + - + ]: 161 : if (o_compression_detail && !o_compression)
811 rhaas@postgresql.org 959 [ # # ]:UBC 0 : ereport(ERROR,
960 : : (errcode(ERRCODE_SYNTAX_ERROR),
961 : : errmsg("compression detail cannot be specified unless compression is enabled")));
962 : :
753 rhaas@postgresql.org 963 [ + + ]:CBC 161 : if (o_compression)
964 : : {
965 : : char *error_detail;
966 : :
733 michael@paquier.xyz 967 : 26 : parse_compress_specification(opt->compression, compression_detail_str,
968 : : &opt->compression_specification);
969 : : error_detail =
970 : 26 : validate_compress_specification(&opt->compression_specification);
753 rhaas@postgresql.org 971 [ + + ]: 26 : if (error_detail != NULL)
972 [ + - ]: 9 : ereport(ERROR,
973 : : errcode(ERRCODE_SYNTAX_ERROR),
974 : : errmsg("invalid compression specification: %s",
975 : : error_detail));
976 : : }
4830 magnus@hagander.net 977 : 152 : }
978 : :
979 : :
980 : : /*
981 : : * SendBaseBackup() - send a complete base backup.
982 : : *
983 : : * The function will put the system into backup mode like pg_backup_start()
984 : : * does, so that the backup is consistent even though we read directly from
985 : : * the filesystem, bypassing the buffer cache.
986 : : */
987 : : void
116 rhaas@postgresql.org 988 :GNC 166 : SendBaseBackup(BaseBackupCmd *cmd, IncrementalBackupInfo *ib)
989 : : {
990 : : basebackup_options opt;
991 : : bbsink *sink;
642 fujii@postgresql.org 992 :CBC 166 : SessionBackupState status = get_backup_status();
993 : :
994 [ + + ]: 166 : if (status == SESSION_BACKUP_RUNNING)
995 [ + - ]: 1 : ereport(ERROR,
996 : : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
997 : : errmsg("a backup is already in progress in this session")));
998 : :
4830 magnus@hagander.net 999 : 165 : parse_basebackup_options(cmd->options, &opt);
1000 : :
4842 1001 : 152 : WalSndSetState(WALSNDSTATE_BACKUP);
1002 : :
4843 1003 [ + - ]: 152 : if (update_process_title)
1004 : : {
1005 : : char activitymsg[50];
1006 : :
1007 : 152 : snprintf(activitymsg, sizeof(activitymsg), "sending backup \"%s\"",
1008 : : opt.label);
1495 peter@eisentraut.org 1009 : 152 : set_ps_display(activitymsg);
1010 : : }
1011 : :
1012 : : /*
1013 : : * If we're asked to perform an incremental backup and the user has not
1014 : : * supplied a manifest, that's an ERROR.
1015 : : *
1016 : : * If we're asked to perform a full backup and the user did supply a
1017 : : * manifest, just ignore it.
1018 : : */
116 rhaas@postgresql.org 1019 [ + + ]:GNC 152 : if (!opt.incremental)
1020 : 145 : ib = NULL;
1021 [ - + ]: 7 : else if (ib == NULL)
116 rhaas@postgresql.org 1022 [ # # ]:UNC 0 : ereport(ERROR,
1023 : : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1024 : : errmsg("must UPLOAD_MANIFEST before performing an incremental BASE_BACKUP")));
1025 : :
1026 : : /*
1027 : : * If the target is specifically 'client' then set up to stream the backup
1028 : : * to the client; otherwise, it's being sent someplace else and should not
1029 : : * be sent to the client. BaseBackupGetSink has the job of setting up a
1030 : : * sink to send the backup data wherever it needs to go.
1031 : : */
761 rhaas@postgresql.org 1032 :CBC 152 : sink = bbsink_copystream_new(opt.send_to_client);
1033 [ + + ]: 152 : if (opt.target_handle != NULL)
1034 : 13 : sink = BaseBackupGetSink(opt.target_handle, sink);
1035 : :
1036 : : /* Set up network throttling, if client requested it */
891 1037 [ + + ]: 149 : if (opt.maxrate > 0)
1038 : 1 : sink = bbsink_throttle_new(sink, opt.maxrate);
1039 : :
1040 : : /* Set up server-side compression, if client requested it */
733 michael@paquier.xyz 1041 [ + + ]: 149 : if (opt.compression == PG_COMPRESSION_GZIP)
753 rhaas@postgresql.org 1042 : 2 : sink = bbsink_gzip_new(sink, &opt.compression_specification);
733 michael@paquier.xyz 1043 [ + + ]: 147 : else if (opt.compression == PG_COMPRESSION_LZ4)
753 rhaas@postgresql.org 1044 : 2 : sink = bbsink_lz4_new(sink, &opt.compression_specification);
733 michael@paquier.xyz 1045 [ + + ]: 145 : else if (opt.compression == PG_COMPRESSION_ZSTD)
753 rhaas@postgresql.org 1046 : 4 : sink = bbsink_zstd_new(sink, &opt.compression_specification);
1047 : :
1048 : : /* Set up progress reporting. */
891 1049 : 149 : sink = bbsink_progress_new(sink, opt.progress);
1050 : :
1051 : : /*
1052 : : * Perform the base backup, but make sure we clean up the bbsink even if
1053 : : * an error occurs.
1054 : : */
1055 [ + + ]: 149 : PG_TRY();
1056 : : {
116 rhaas@postgresql.org 1057 :GNC 149 : perform_base_backup(&opt, sink, ib);
1058 : : }
891 rhaas@postgresql.org 1059 :CBC 4 : PG_FINALLY();
1060 : : {
1061 : 144 : bbsink_cleanup(sink);
1062 : : }
1063 [ + + ]: 144 : PG_END_TRY();
4819 magnus@hagander.net 1064 : 140 : }
1065 : :
1066 : : /*
1067 : : * Inject a file with given name and content in the output tar stream.
1068 : : *
1069 : : * "len" can optionally be set to an arbitrary length of data sent. If set
1070 : : * to -1, the content sent is treated as a string with strlen() as length.
1071 : : */
1072 : : static void
891 rhaas@postgresql.org 1073 : 193 : sendFileWithContent(bbsink *sink, const char *filename, const char *content,
1074 : : int len, backup_manifest_info *manifest)
1075 : : {
1076 : : struct stat statbuf;
153 michael@paquier.xyz 1077 :GNC 193 : int bytes_done = 0;
1078 : : pg_checksum_context checksum_ctx;
1079 : :
1229 michael@paquier.xyz 1080 [ - + ]:CBC 193 : if (pg_checksum_init(&checksum_ctx, manifest->checksum_type) < 0)
1229 michael@paquier.xyz 1081 [ # # ]:UBC 0 : elog(ERROR, "could not initialize checksum of file \"%s\"",
1082 : : filename);
1083 : :
153 michael@paquier.xyz 1084 [ + - ]:GNC 193 : if (len < 0)
1085 : 193 : len = strlen(content);
1086 : :
1087 : : /*
1088 : : * Construct a stat struct for the file we're injecting in the tar.
1089 : : */
1090 : :
1091 : : /* Windows doesn't have the concept of uid and gid */
1092 : : #ifdef WIN32
1093 : : statbuf.st_uid = 0;
1094 : : statbuf.st_gid = 0;
1095 : : #else
4822 heikki.linnakangas@i 1096 :CBC 193 : statbuf.st_uid = geteuid();
1097 : 193 : statbuf.st_gid = getegid();
1098 : : #endif
1099 : 193 : statbuf.st_mtime = time(NULL);
2199 sfrost@snowman.net 1100 : 193 : statbuf.st_mode = pg_file_create_mode;
4822 heikki.linnakangas@i 1101 : 193 : statbuf.st_size = len;
1102 : :
891 rhaas@postgresql.org 1103 : 193 : _tarWriteHeader(sink, filename, NULL, &statbuf, false);
1104 : :
1105 [ - + ]: 193 : if (pg_checksum_update(&checksum_ctx, (uint8 *) content, len) < 0)
891 rhaas@postgresql.org 1106 [ # # ]:UBC 0 : elog(ERROR, "could not update checksum of file \"%s\"",
1107 : : filename);
1108 : :
891 rhaas@postgresql.org 1109 [ + + ]:CBC 344 : while (bytes_done < len)
1110 : : {
1111 : 151 : size_t remaining = len - bytes_done;
1112 : 151 : size_t nbytes = Min(sink->bbs_buffer_length, remaining);
1113 : :
1114 : 151 : memcpy(sink->bbs_buffer, content, nbytes);
1115 : 151 : bbsink_archive_contents(sink, nbytes);
1116 : 151 : bytes_done += nbytes;
489 1117 : 151 : content += nbytes;
1118 : : }
1119 : :
891 1120 : 193 : _tarWritePadding(sink, len);
1121 : :
174 rhaas@postgresql.org 1122 :GNC 193 : AddFileToBackupManifest(manifest, InvalidOid, filename, len,
1452 rhaas@postgresql.org 1123 :CBC 193 : (pg_time_t) statbuf.st_mtime, &checksum_ctx);
4822 heikki.linnakangas@i 1124 : 193 : }
1125 : :
1126 : : /*
1127 : : * Include the tablespace directory pointed to by 'path' in the output tar
1128 : : * stream. If 'sizeonly' is true, we just calculate a total length and return
1129 : : * it, without actually sending anything.
1130 : : *
1131 : : * Only used to send auxiliary tablespaces, not PGDATA.
1132 : : */
1133 : : static int64
174 rhaas@postgresql.org 1134 :GNC 62 : sendTablespace(bbsink *sink, char *path, Oid spcoid, bool sizeonly,
1135 : : backup_manifest_info *manifest, IncrementalBackupInfo *ib)
1136 : : {
1137 : : int64 size;
1138 : : char pathbuf[MAXPGPATH];
1139 : : struct stat statbuf;
1140 : :
1141 : : /*
1142 : : * 'path' points to the tablespace location, but we only want to include
1143 : : * the version directory in it that belongs to us.
1144 : : */
4038 heikki.linnakangas@i 1145 :CBC 62 : snprintf(pathbuf, sizeof(pathbuf), "%s/%s", path,
1146 : : TABLESPACE_VERSION_DIRECTORY);
1147 : :
1148 : : /*
1149 : : * Store a directory entry in the tar file so we get the permissions
1150 : : * right.
1151 : : */
1152 [ - + ]: 62 : if (lstat(pathbuf, &statbuf) != 0)
1153 : : {
4038 heikki.linnakangas@i 1154 [ # # ]:UBC 0 : if (errno != ENOENT)
1155 [ # # ]: 0 : ereport(ERROR,
1156 : : (errcode_for_file_access(),
1157 : : errmsg("could not stat file or directory \"%s\": %m",
1158 : : pathbuf)));
1159 : :
1160 : : /* If the tablespace went away while scanning, it's no error. */
1161 : 0 : return 0;
1162 : : }
1163 : :
891 rhaas@postgresql.org 1164 :CBC 62 : size = _tarWriteHeader(sink, TABLESPACE_VERSION_DIRECTORY, NULL, &statbuf,
1165 : : sizeonly);
1166 : :
1167 : : /* Send all the files in the tablespace version directory */
1168 : 62 : size += sendDir(sink, pathbuf, strlen(path), sizeonly, NIL, true, manifest,
1169 : : spcoid, ib);
1170 : :
4038 heikki.linnakangas@i 1171 : 62 : return size;
1172 : : }
1173 : :
1174 : : /*
1175 : : * Include all files from the given directory in the output tar stream. If
1176 : : * 'sizeonly' is true, we just calculate a total length and return it, without
1177 : : * actually sending anything.
1178 : : *
1179 : : * Omit any directory in the tablespaces list, to avoid backing up
1180 : : * tablespaces twice when they were created inside PGDATA.
1181 : : *
1182 : : * If sendtblspclinks is true, we need to include symlink
1183 : : * information in the tar file. If not, we can skip that
1184 : : * as it will be sent separately in the tablespace_map file.
1185 : : */
1186 : : static int64
891 rhaas@postgresql.org 1187 : 5107 : sendDir(bbsink *sink, const char *path, int basepathlen, bool sizeonly,
1188 : : List *tablespaces, bool sendtblspclinks, backup_manifest_info *manifest,
1189 : : Oid spcoid, IncrementalBackupInfo *ib)
1190 : : {
1191 : : DIR *dir;
1192 : : struct dirent *de;
1193 : : char pathbuf[MAXPGPATH * 2];
1194 : : struct stat statbuf;
4843 magnus@hagander.net 1195 : 5107 : int64 size = 0;
1196 : : const char *lastDir; /* Split last dir from parent path. */
152 rhaas@postgresql.org 1197 :GNC 5107 : bool isRelationDir = false; /* Does directory contain relations? */
116 1198 : 5107 : bool isGlobalDir = false;
152 1199 : 5107 : Oid dboid = InvalidOid;
116 1200 : 5107 : BlockNumber *relative_block_numbers = NULL;
1201 : :
1202 : : /*
1203 : : * Since this array is relatively large, avoid putting it on the stack.
1204 : : * But we don't need it at all if this is not an incremental backup.
1205 : : */
1206 [ + + ]: 5107 : if (ib != NULL)
1207 : 121 : relative_block_numbers = palloc(sizeof(BlockNumber) * RELSEG_SIZE);
1208 : :
1209 : : /*
1210 : : * Determine if the current path is a database directory that can contain
1211 : : * relations.
1212 : : *
1213 : : * Start by finding the location of the delimiter between the parent path
1214 : : * and the current path.
1215 : : */
2214 teodor@sigaev.ru 1216 :CBC 5107 : lastDir = last_dir_separator(path);
1217 : :
1218 : : /* Does this path look like a database path (i.e. all digits)? */
1219 [ + + ]: 5107 : if (lastDir != NULL &&
1220 [ + + ]: 4809 : strspn(lastDir + 1, "0123456789") == strlen(lastDir + 1))
2214 teodor@sigaev.ru 1221 :GIC 949 : {
1222 : : /* Part of path that contains the parent directory. */
2180 tgl@sss.pgh.pa.us 1223 :CBC 949 : int parentPathLen = lastDir - path;
1224 : :
1225 : : /*
1226 : : * Mark path as a database directory if the parent path is either
1227 : : * $PGDATA/base or a tablespace version path.
1228 : : */
2214 teodor@sigaev.ru 1229 [ + + ]: 949 : if (strncmp(path, "./base", parentPathLen) == 0 ||
1230 [ + - ]: 40 : (parentPathLen >= (sizeof(TABLESPACE_VERSION_DIRECTORY) - 1) &&
1231 [ + - ]: 40 : strncmp(lastDir - (sizeof(TABLESPACE_VERSION_DIRECTORY) - 1),
1232 : : TABLESPACE_VERSION_DIRECTORY,
1233 : : sizeof(TABLESPACE_VERSION_DIRECTORY) - 1) == 0))
1234 : : {
152 rhaas@postgresql.org 1235 :GNC 949 : isRelationDir = true;
1236 : 949 : dboid = atooid(lastDir + 1);
1237 : : }
1238 : : }
1239 [ + + ]: 4158 : else if (strcmp(path, "./global") == 0)
1240 : : {
1241 : 298 : isRelationDir = true;
116 1242 : 298 : isGlobalDir = true;
1243 : : }
1244 : :
4843 magnus@hagander.net 1245 :CBC 5107 : dir = AllocateDir(path);
1246 [ + + ]: 317088 : while ((de = ReadDir(dir, path)) != NULL)
1247 : : {
1248 : : int excludeIdx;
1249 : : bool excludeFound;
152 rhaas@postgresql.org 1250 :GNC 311992 : RelFileNumber relfilenumber = InvalidRelFileNumber;
1251 : 311992 : ForkNumber relForkNum = InvalidForkNumber;
1252 : 311992 : unsigned segno = 0;
1253 : 311992 : bool isRelationFile = false;
1254 : :
1255 : : /* Skip special stuff */
4843 magnus@hagander.net 1256 [ + + + + ]:CBC 311992 : if (strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0)
1257 : 14493 : continue;
1258 : :
1259 : : /* Skip temporary files */
heikki.linnakangas@i 1260 [ + + ]: 301778 : if (strncmp(de->d_name,
1261 : : PG_TEMP_FILE_PREFIX,
1262 : : strlen(PG_TEMP_FILE_PREFIX)) == 0)
1263 : 293 : continue;
1264 : :
1265 : : /* Skip macOS system files */
61 dgustafsson@postgres 1266 [ + + ]: 301485 : if (strcmp(de->d_name, ".DS_Store") == 0)
1267 : 67 : continue;
1268 : :
1269 : : /*
1270 : : * Check if the postmaster has signaled us to exit, and abort with an
1271 : : * error in that case. The error handler further up will call
1272 : : * do_pg_abort_backup() for us. Also check that if the backup was
1273 : : * started while still in recovery, the server wasn't promoted.
1274 : : * do_pg_backup_stop() will check that too, but it's better to stop
1275 : : * the backup early than continue to the end and fail there.
1276 : : */
4140 heikki.linnakangas@i 1277 [ + + ]: 301418 : CHECK_FOR_INTERRUPTS();
1278 [ - + ]: 301413 : if (RecoveryInProgress() != backup_started_in_recovery)
4839 magnus@hagander.net 1279 [ # # ]:UBC 0 : ereport(ERROR,
1280 : : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1281 : : errmsg("the standby was promoted during online backup"),
1282 : : errhint("This means that the backup being taken is corrupt "
1283 : : "and should not be used. "
1284 : : "Try taking another online backup.")));
1285 : :
1286 : : /* Scan for files that should be excluded */
2755 peter_e@gmx.net 1287 :CBC 301413 : excludeFound = false;
1511 michael@paquier.xyz 1288 [ + + ]: 2708253 : for (excludeIdx = 0; excludeFiles[excludeIdx].name != NULL; excludeIdx++)
1289 : : {
1290 : 2408001 : int cmplen = strlen(excludeFiles[excludeIdx].name);
1291 : :
1292 [ + + ]: 2408001 : if (!excludeFiles[excludeIdx].match_prefix)
1293 : 2106722 : cmplen++;
1294 [ + + ]: 2408001 : if (strncmp(de->d_name, excludeFiles[excludeIdx].name, cmplen) == 0)
1295 : : {
2755 peter_e@gmx.net 1296 [ + + ]: 1161 : elog(DEBUG1, "file \"%s\" excluded from backup", de->d_name);
1297 : 1161 : excludeFound = true;
1298 : 1161 : break;
1299 : : }
1300 : : }
1301 : :
1302 [ + + ]: 301413 : if (excludeFound)
4843 magnus@hagander.net 1303 : 1161 : continue;
1304 : :
1305 : : /*
1306 : : * If there could be non-temporary relation files in this directory,
1307 : : * try to parse the filename.
1308 : : */
152 rhaas@postgresql.org 1309 [ + + ]:GNC 300252 : if (isRelationDir)
1310 : : isRelationFile =
1311 : 290319 : parse_filename_for_nontemp_relation(de->d_name,
1312 : : &relfilenumber,
1313 : : &relForkNum, &segno);
1314 : :
1315 : : /* Exclude all forks for unlogged tables except the init fork */
1316 [ + + + + ]: 300252 : if (isRelationFile && relForkNum != INIT_FORKNUM)
1317 : : {
1318 : : char initForkFile[MAXPGPATH];
1319 : :
1320 : : /*
1321 : : * If any other type of fork, check if there is an init fork with
1322 : : * the same RelFileNumber. If so, the file can be excluded.
1323 : : */
1324 : 287794 : snprintf(initForkFile, sizeof(initForkFile), "%s/%u_init",
1325 : : path, relfilenumber);
1326 : :
1327 [ + + ]: 287794 : if (lstat(initForkFile, &statbuf) == 0)
1328 : : {
1329 [ - + ]: 71 : elog(DEBUG2,
1330 : : "unlogged relation file \"%s\" excluded from backup",
1331 : : de->d_name);
1332 : :
1333 : 71 : continue;
1334 : : }
1335 : : }
1336 : :
1337 : : /* Exclude temporary relations */
1338 [ + + + + ]: 300181 : if (OidIsValid(dboid) && looks_like_temp_rel_name(de->d_name))
1339 : : {
2210 teodor@sigaev.ru 1340 [ - + ]:CBC 40 : elog(DEBUG2,
1341 : : "temporary relation file \"%s\" excluded from backup",
1342 : : de->d_name);
1343 : :
1344 : 40 : continue;
1345 : : }
1346 : :
2560 peter_e@gmx.net 1347 : 300141 : snprintf(pathbuf, sizeof(pathbuf), "%s/%s", path, de->d_name);
1348 : :
1349 : : /* Skip pg_control here to back up it last */
4463 simon@2ndQuadrant.co 1350 [ + + ]: 300141 : if (strcmp(pathbuf, "./global/pg_control") == 0)
1351 : 298 : continue;
1352 : :
4843 magnus@hagander.net 1353 [ - + ]: 299843 : if (lstat(pathbuf, &statbuf) != 0)
1354 : : {
4843 magnus@hagander.net 1355 [ # # ]:UBC 0 : if (errno != ENOENT)
1356 [ # # ]: 0 : ereport(ERROR,
1357 : : (errcode_for_file_access(),
1358 : : errmsg("could not stat file or directory \"%s\": %m",
1359 : : pathbuf)));
1360 : :
1361 : : /* If the file went away while scanning, it's not an error. */
1362 : 0 : continue;
1363 : : }
1364 : :
1365 : : /* Scan for directories whose contents should be excluded */
2755 peter_e@gmx.net 1366 :CBC 299843 : excludeFound = false;
1367 [ + + ]: 2390540 : for (excludeIdx = 0; excludeDirContents[excludeIdx] != NULL; excludeIdx++)
1368 : : {
1369 [ + + ]: 2092748 : if (strcmp(de->d_name, excludeDirContents[excludeIdx]) == 0)
1370 : : {
1371 [ + + ]: 2051 : elog(DEBUG1, "contents of directory \"%s\" excluded from backup", de->d_name);
915 rhaas@postgresql.org 1372 : 2051 : convert_link_to_directory(pathbuf, &statbuf);
891 1373 : 2051 : size += _tarWriteHeader(sink, pathbuf + basepathlen + 1, NULL,
1374 : : &statbuf, sizeonly);
2755 peter_e@gmx.net 1375 : 2051 : excludeFound = true;
1376 : 2051 : break;
1377 : : }
1378 : : }
1379 : :
1380 [ + + ]: 299843 : if (excludeFound)
1381 : 2051 : continue;
1382 : :
1383 : : /*
1384 : : * We can skip pg_wal, the WAL segments need to be fetched from the
1385 : : * WAL archive anyway. But include it as an empty directory anyway, so
1386 : : * we get permissions right.
1387 : : */
2733 rhaas@postgresql.org 1388 [ + + ]: 297792 : if (strcmp(pathbuf, "./pg_wal") == 0)
1389 : : {
1390 : : /* If pg_wal is a symlink, write it as a directory anyway */
915 1391 : 298 : convert_link_to_directory(pathbuf, &statbuf);
891 1392 : 298 : size += _tarWriteHeader(sink, pathbuf + basepathlen + 1, NULL,
1393 : : &statbuf, sizeonly);
1394 : :
1395 : : /*
1396 : : * Also send archive_status and summaries directories (by
1397 : : * hackishly reusing statbuf from above ...).
1398 : : */
1399 : 298 : size += _tarWriteHeader(sink, "./pg_wal/archive_status", NULL,
1400 : : &statbuf, sizeonly);
116 rhaas@postgresql.org 1401 :GNC 298 : size += _tarWriteHeader(sink, "./pg_wal/summaries", NULL,
1402 : : &statbuf, sizeonly);
1403 : :
2733 rhaas@postgresql.org 1404 :CBC 298 : continue; /* don't recurse into pg_wal */
1405 : : }
1406 : :
1407 : : /* Allow symbolic links in pg_tblspc only */
617 tmunro@postgresql.or 1408 [ + + + + ]: 297494 : if (strcmp(path, "./pg_tblspc") == 0 && S_ISLNK(statbuf.st_mode))
4843 magnus@hagander.net 1409 : 31 : {
1410 : : char linkpath[MAXPGPATH];
1411 : : int rllen;
1412 : :
4512 tgl@sss.pgh.pa.us 1413 : 31 : rllen = readlink(pathbuf, linkpath, sizeof(linkpath));
1414 [ - + ]: 31 : if (rllen < 0)
4843 magnus@hagander.net 1415 [ # # ]:UBC 0 : ereport(ERROR,
1416 : : (errcode_for_file_access(),
1417 : : errmsg("could not read symbolic link \"%s\": %m",
1418 : : pathbuf)));
4512 tgl@sss.pgh.pa.us 1419 [ - + ]:CBC 31 : if (rllen >= sizeof(linkpath))
4512 tgl@sss.pgh.pa.us 1420 [ # # ]:UBC 0 : ereport(ERROR,
1421 : : (errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
1422 : : errmsg("symbolic link \"%s\" target is too long",
1423 : : pathbuf)));
4512 tgl@sss.pgh.pa.us 1424 :CBC 31 : linkpath[rllen] = '\0';
1425 : :
891 rhaas@postgresql.org 1426 : 31 : size += _tarWriteHeader(sink, pathbuf + basepathlen + 1, linkpath,
1427 : : &statbuf, sizeonly);
1428 : : }
4843 magnus@hagander.net 1429 [ + + ]: 297463 : else if (S_ISDIR(statbuf.st_mode))
1430 : : {
3750 1431 : 4799 : bool skip_this_dir = false;
1432 : : ListCell *lc;
1433 : :
1434 : : /*
1435 : : * Store a directory entry in the tar file so we can get the
1436 : : * permissions right.
1437 : : */
891 rhaas@postgresql.org 1438 : 4799 : size += _tarWriteHeader(sink, pathbuf + basepathlen + 1, NULL, &statbuf,
1439 : : sizeonly);
1440 : :
1441 : : /*
1442 : : * Call ourselves recursively for a directory, unless it happens
1443 : : * to be a separate tablespace located within PGDATA.
1444 : : */
3750 magnus@hagander.net 1445 [ + + + + : 10547 : foreach(lc, tablespaces)
+ + ]
1446 : : {
1447 : 5776 : tablespaceinfo *ti = (tablespaceinfo *) lfirst(lc);
1448 : :
1449 : : /*
1450 : : * ti->rpath is the tablespace relative path within PGDATA, or
1451 : : * NULL if the tablespace has been properly located somewhere
1452 : : * else.
1453 : : *
1454 : : * Skip past the leading "./" in pathbuf when comparing.
1455 : : */
1456 [ + + + + ]: 5776 : if (ti->rpath && strcmp(ti->rpath, pathbuf + 2) == 0)
1457 : : {
1458 : 28 : skip_this_dir = true;
1459 : 28 : break;
1460 : : }
1461 : : }
1462 : :
1463 : : /*
1464 : : * skip sending directories inside pg_tblspc, if not required.
1465 : : */
3260 andrew@dunslane.net 1466 [ + + + + ]: 4799 : if (strcmp(pathbuf, "./pg_tblspc") == 0 && !sendtblspclinks)
1467 : 24 : skip_this_dir = true;
1468 : :
3750 magnus@hagander.net 1469 [ + + ]: 4799 : if (!skip_this_dir)
891 rhaas@postgresql.org 1470 : 4747 : size += sendDir(sink, pathbuf, basepathlen, sizeonly, tablespaces,
1471 : : sendtblspclinks, manifest, spcoid, ib);
1472 : : }
4843 magnus@hagander.net 1473 [ + - ]: 292664 : else if (S_ISREG(statbuf.st_mode))
1474 : : {
3973 bruce@momjian.us 1475 : 292664 : bool sent = false;
116 rhaas@postgresql.org 1476 :GNC 292664 : unsigned num_blocks_required = 0;
1477 : 292664 : unsigned truncation_block_length = 0;
1478 : : char tarfilenamebuf[MAXPGPATH * 2];
1479 : 292664 : char *tarfilename = pathbuf + basepathlen + 1;
1480 : 292664 : FileBackupMethod method = BACK_UP_FILE_FULLY;
1481 : :
1482 [ + + + + ]: 292664 : if (ib != NULL && isRelationFile)
1483 : : {
1484 : : Oid relspcoid;
1485 : : char *lookup_path;
1486 : :
1487 [ - + ]: 7255 : if (OidIsValid(spcoid))
1488 : : {
116 rhaas@postgresql.org 1489 :UNC 0 : relspcoid = spcoid;
1490 : 0 : lookup_path = psprintf("pg_tblspc/%u/%s", spcoid,
1491 : : tarfilename);
1492 : : }
1493 : : else
1494 : : {
116 rhaas@postgresql.org 1495 [ + + ]:GNC 7255 : if (isGlobalDir)
1496 : 420 : relspcoid = GLOBALTABLESPACE_OID;
1497 : : else
1498 : 6835 : relspcoid = DEFAULTTABLESPACE_OID;
1499 : 7255 : lookup_path = pstrdup(tarfilename);
1500 : : }
1501 : :
1502 : 7255 : method = GetFileBackupMethod(ib, lookup_path, dboid, relspcoid,
1503 : : relfilenumber, relForkNum,
1504 : 7255 : segno, statbuf.st_size,
1505 : : &num_blocks_required,
1506 : : relative_block_numbers,
1507 : : &truncation_block_length);
1508 [ + + ]: 7255 : if (method == BACK_UP_FILE_INCREMENTALLY)
1509 : : {
1510 : 4651 : statbuf.st_size =
1511 : 4651 : GetIncrementalFileSize(num_blocks_required);
1512 : 4651 : snprintf(tarfilenamebuf, sizeof(tarfilenamebuf),
1513 : : "%s/INCREMENTAL.%s",
1514 : 4651 : path + basepathlen + 1,
1515 : 4651 : de->d_name);
1516 : 4651 : tarfilename = tarfilenamebuf;
1517 : : }
1518 : :
1519 : 7255 : pfree(lookup_path);
1520 : : }
1521 : :
4843 magnus@hagander.net 1522 [ + + ]:CBC 292664 : if (!sizeonly)
116 rhaas@postgresql.org 1523 [ + + ]:GNC 143980 : sent = sendFile(sink, pathbuf, tarfilename, &statbuf,
1524 : : true, dboid, spcoid,
1525 : : relfilenumber, segno, manifest,
1526 : : num_blocks_required,
1527 : : method == BACK_UP_FILE_INCREMENTALLY ? relative_block_numbers : NULL,
1528 : : truncation_block_length);
1529 : :
4132 heikki.linnakangas@i 1530 [ + + + - ]:CBC 292663 : if (sent || sizeonly)
1531 : : {
1532 : : /* Add size. */
1451 rhaas@postgresql.org 1533 : 292663 : size += statbuf.st_size;
1534 : :
1535 : : /* Pad to a multiple of the tar block size. */
1536 : 292663 : size += tarPaddingBytesRequired(statbuf.st_size);
1537 : :
1538 : : /* Size of the header for the file. */
1539 : 292663 : size += TAR_BLOCK_SIZE;
1540 : : }
1541 : : }
1542 : : else
4843 magnus@hagander.net 1543 [ # # ]:UBC 0 : ereport(WARNING,
1544 : : (errmsg("skipping special file \"%s\"", pathbuf)));
1545 : : }
1546 : :
116 rhaas@postgresql.org 1547 [ + + ]:GNC 5096 : if (relative_block_numbers != NULL)
1548 : 121 : pfree(relative_block_numbers);
1549 : :
4843 magnus@hagander.net 1550 :CBC 5096 : FreeDir(dir);
1551 : 5096 : return size;
1552 : : }
1553 : :
1554 : : /*
1555 : : * Given the member, write the TAR header & send the file.
1556 : : *
1557 : : * If 'missing_ok' is true, will not throw an error if the file is not found.
1558 : : *
1559 : : * If dboid is anything other than InvalidOid then any checksum failures
1560 : : * detected will get reported to the cumulative stats system.
1561 : : *
1562 : : * If the file is to be sent incrementally, then num_incremental_blocks
1563 : : * should be the number of blocks to be sent, and incremental_blocks
1564 : : * an array of block numbers relative to the start of the current segment.
1565 : : * If the whole file is to be sent, then incremental_blocks should be NULL,
1566 : : * and num_incremental_blocks can have any value, as it will be ignored.
1567 : : *
1568 : : * Returns true if the file was successfully sent, false if 'missing_ok',
1569 : : * and the file did not exist.
1570 : : */
1571 : : static bool
891 rhaas@postgresql.org 1572 : 144123 : sendFile(bbsink *sink, const char *readfilename, const char *tarfilename,
1573 : : struct stat *statbuf, bool missing_ok, Oid dboid, Oid spcoid,
1574 : : RelFileNumber relfilenumber, unsigned segno,
1575 : : backup_manifest_info *manifest, unsigned num_incremental_blocks,
1576 : : BlockNumber *incremental_blocks, unsigned truncation_block_length)
1577 : : {
1578 : : int fd;
2203 magnus@hagander.net 1579 : 144123 : BlockNumber blkno = 0;
1580 : 144123 : int checksum_failures = 0;
1581 : : off_t cnt;
194 rhaas@postgresql.org 1582 :GNC 144123 : pgoff_t bytes_done = 0;
2203 magnus@hagander.net 1583 :CBC 144123 : bool verify_checksum = false;
1584 : : pg_checksum_context checksum_ctx;
116 rhaas@postgresql.org 1585 :GNC 144123 : int ibindex = 0;
1586 : :
1229 michael@paquier.xyz 1587 [ - + ]:CBC 144123 : if (pg_checksum_init(&checksum_ctx, manifest->checksum_type) < 0)
1229 michael@paquier.xyz 1588 [ # # ]:UBC 0 : elog(ERROR, "could not initialize checksum of file \"%s\"",
1589 : : readfilename);
1590 : :
1397 rhaas@postgresql.org 1591 :CBC 144123 : fd = OpenTransientFile(readfilename, O_RDONLY | PG_BINARY);
1592 [ - + ]: 144123 : if (fd < 0)
1593 : : {
4132 heikki.linnakangas@i 1594 [ # # # # ]:UBC 0 : if (errno == ENOENT && missing_ok)
1595 : 0 : return false;
4843 magnus@hagander.net 1596 [ # # ]: 0 : ereport(ERROR,
1597 : : (errcode_for_file_access(),
1598 : : errmsg("could not open file \"%s\": %m", readfilename)));
1599 : : }
1600 : :
891 rhaas@postgresql.org 1601 :CBC 144123 : _tarWriteHeader(sink, tarfilename, NULL, statbuf, false);
1602 : :
1603 : : /*
1604 : : * Checksums are verified in multiples of BLCKSZ, so the buffer length
1605 : : * should be a multiple of the block size as well.
1606 : : */
194 rhaas@postgresql.org 1607 [ - + ]:GNC 144122 : Assert((sink->bbs_buffer_length % BLCKSZ) == 0);
1608 : :
1609 : : /*
1610 : : * If we weren't told not to verify checksums, and if checksums are
1611 : : * enabled for this cluster, and if this is a relation file, then verify
1612 : : * the checksum.
1613 : : */
152 1614 [ + + + + : 144122 : if (!noverify_checksums && DataChecksumsEnabled() &&
+ + ]
1615 : : RelFileNumberIsValid(relfilenumber))
1616 : 28609 : verify_checksum = true;
1617 : :
1618 : : /*
1619 : : * If we're sending an incremental file, write the file header.
1620 : : */
116 1621 [ + + ]: 144122 : if (incremental_blocks != NULL)
1622 : : {
1623 : 4651 : unsigned magic = INCREMENTAL_MAGIC;
1624 : 4651 : size_t header_bytes_done = 0;
1625 : : char padding[BLCKSZ];
1626 : : size_t paddinglen;
1627 : :
1628 : : /* Emit header data. */
1629 : 4651 : push_to_sink(sink, &checksum_ctx, &header_bytes_done,
1630 : : &magic, sizeof(magic));
1631 : 4651 : push_to_sink(sink, &checksum_ctx, &header_bytes_done,
1632 : : &num_incremental_blocks, sizeof(num_incremental_blocks));
1633 : 4651 : push_to_sink(sink, &checksum_ctx, &header_bytes_done,
1634 : : &truncation_block_length, sizeof(truncation_block_length));
1635 : 4651 : push_to_sink(sink, &checksum_ctx, &header_bytes_done,
1636 : : incremental_blocks,
1637 : : sizeof(BlockNumber) * num_incremental_blocks);
1638 : :
1639 : : /*
1640 : : * Add padding to align header to a multiple of BLCKSZ, but only if
1641 : : * the incremental file has some blocks, and the alignment is actually
1642 : : * needed (i.e. header is not already a multiple of BLCKSZ). If there
1643 : : * are no blocks we don't want to make the file unnecessarily large,
1644 : : * as that might make some filesystem optimizations impossible.
1645 : : */
0 tomas.vondra@postgre 1646 [ + + + - ]: 4651 : if ((num_incremental_blocks > 0) && (header_bytes_done % BLCKSZ != 0))
1647 : : {
9 1648 : 23 : paddinglen = (BLCKSZ - (header_bytes_done % BLCKSZ));
1649 : :
1650 : 23 : memset(padding, 0, paddinglen);
1651 : 23 : bytes_done += paddinglen;
1652 : :
1653 : 23 : push_to_sink(sink, &checksum_ctx, &header_bytes_done,
1654 : : padding, paddinglen);
1655 : : }
1656 : :
1657 : : /* Flush out any data still in the buffer so it's again empty. */
116 rhaas@postgresql.org 1658 [ + - ]: 4651 : if (header_bytes_done > 0)
1659 : : {
1660 : 4651 : bbsink_archive_contents(sink, header_bytes_done);
1661 [ - + ]: 4651 : if (pg_checksum_update(&checksum_ctx,
1662 : 4651 : (uint8 *) sink->bbs_buffer,
1663 : : header_bytes_done) < 0)
116 rhaas@postgresql.org 1664 [ # # ]:UNC 0 : elog(ERROR, "could not update checksum of base backup");
1665 : : }
1666 : :
1667 : : /* Update our notion of file position. */
116 rhaas@postgresql.org 1668 :GNC 4651 : bytes_done += sizeof(magic);
1669 : 4651 : bytes_done += sizeof(num_incremental_blocks);
1670 : 4651 : bytes_done += sizeof(truncation_block_length);
1671 : 4651 : bytes_done += sizeof(BlockNumber) * num_incremental_blocks;
1672 : : }
1673 : :
1674 : : /*
1675 : : * Loop until we read the amount of data the caller told us to expect. The
1676 : : * file could be longer, if it was extended while we were sending it, but
1677 : : * for a base backup we can ignore such extended data. It will be restored
1678 : : * from WAL.
1679 : : */
1680 : : while (1)
1681 : : {
1682 : : /*
1683 : : * Determine whether we've read all the data that we need, and if not,
1684 : : * read some more.
1685 : : */
1686 [ + + ]: 312382 : if (incremental_blocks == NULL)
1687 : : {
1688 : 307696 : size_t remaining = statbuf->st_size - bytes_done;
1689 : :
1690 : : /*
1691 : : * If we've read the required number of bytes, then it's time to
1692 : : * stop.
1693 : : */
1694 [ + + ]: 307696 : if (bytes_done >= statbuf->st_size)
1695 : 139471 : break;
1696 : :
1697 : : /*
1698 : : * Read as many bytes as will fit in the buffer, or however many
1699 : : * are left to read, whichever is less.
1700 : : */
1701 : 168225 : cnt = read_file_data_into_buffer(sink, readfilename, fd,
1702 : : bytes_done, remaining,
1703 : 168225 : blkno + segno * RELSEG_SIZE,
1704 : : verify_checksum,
1705 : : &checksum_failures);
1706 : : }
1707 : : else
1708 : : {
1709 : : BlockNumber relative_blkno;
1710 : :
1711 : : /*
1712 : : * If we've read all the blocks, then it's time to stop.
1713 : : */
1714 [ + + ]: 4686 : if (ibindex >= num_incremental_blocks)
1715 : 4651 : break;
1716 : :
1717 : : /*
1718 : : * Read just one block, whichever one is the next that we're
1719 : : * supposed to include.
1720 : : */
1721 : 35 : relative_blkno = incremental_blocks[ibindex++];
1722 : 35 : cnt = read_file_data_into_buffer(sink, readfilename, fd,
1723 : 35 : relative_blkno * BLCKSZ,
1724 : : BLCKSZ,
1725 : 35 : relative_blkno + segno * RELSEG_SIZE,
1726 : : verify_checksum,
1727 : : &checksum_failures);
1728 : :
1729 : : /*
1730 : : * If we get a partial read, that must mean that the relation is
1731 : : * being truncated. Ultimately, it should be truncated to a
1732 : : * multiple of BLCKSZ, since this path should only be reached for
1733 : : * relation files, but we might transiently observe an
1734 : : * intermediate value.
1735 : : *
1736 : : * It should be fine to treat this just as if the entire block had
1737 : : * been truncated away - i.e. fill this and all later blocks with
1738 : : * zeroes. WAL replay will fix things up.
1739 : : */
1740 [ - + ]: 35 : if (cnt < BLCKSZ)
116 rhaas@postgresql.org 1741 :UNC 0 : break;
1742 : : }
1743 : :
1744 : : /*
1745 : : * If the amount of data we were able to read was not a multiple of
1746 : : * BLCKSZ, we cannot verify checksums, which are block-level.
1747 : : */
2203 magnus@hagander.net 1748 [ + + - + ]:CBC 168260 : if (verify_checksum && (cnt % BLCKSZ != 0))
1749 : : {
2203 magnus@hagander.net 1750 [ # # ]:UBC 0 : ereport(WARNING,
1751 : : (errmsg("could not verify checksum in file \"%s\", block "
1752 : : "%u: read buffer size %d and page size %d "
1753 : : "differ",
1754 : : readfilename, blkno, (int) cnt, BLCKSZ)));
1755 : 0 : verify_checksum = false;
1756 : : }
1757 : :
1758 : : /*
1759 : : * If we hit end-of-file, a concurrent truncation must have occurred.
1760 : : * That's not an error condition, because WAL replay will fix things
1761 : : * up.
1762 : : */
437 rhaas@postgresql.org 1763 [ - + ]:CBC 168260 : if (cnt == 0)
437 rhaas@postgresql.org 1764 :UBC 0 : break;
1765 : :
1766 : : /* Update block number and # of bytes done for next loop iteration. */
194 rhaas@postgresql.org 1767 :GNC 168260 : blkno += cnt / BLCKSZ;
1768 : 168260 : bytes_done += cnt;
1769 : :
1770 : : /*
1771 : : * Make sure incremental files with block data are properly aligned
1772 : : * (header is a multiple of BLCKSZ, blocks are BLCKSZ too).
1773 : : */
9 tomas.vondra@postgre 1774 [ + + + - : 168260 : Assert(!((incremental_blocks != NULL && num_incremental_blocks > 0) &&
- + ]
1775 : : (bytes_done % BLCKSZ != 0)));
1776 : :
1777 : : /* Archive the data we just read. */
891 rhaas@postgresql.org 1778 :CBC 168260 : bbsink_archive_contents(sink, cnt);
1779 : :
1780 : : /* Also feed it to the checksum machinery. */
1781 [ - + ]: 168260 : if (pg_checksum_update(&checksum_ctx,
1782 : 168260 : (uint8 *) sink->bbs_buffer, cnt) < 0)
1229 michael@paquier.xyz 1783 [ # # ]:UBC 0 : elog(ERROR, "could not update checksum of base backup");
1784 : : }
1785 : :
1786 : : /* If the file was truncated while we were sending it, pad it with zeros */
194 rhaas@postgresql.org 1787 [ - + ]:GNC 144122 : while (bytes_done < statbuf->st_size)
1788 : : {
194 rhaas@postgresql.org 1789 :UNC 0 : size_t remaining = statbuf->st_size - bytes_done;
891 rhaas@postgresql.org 1790 :UBC 0 : size_t nbytes = Min(sink->bbs_buffer_length, remaining);
1791 : :
1792 [ # # # # : 0 : MemSet(sink->bbs_buffer, 0, nbytes);
# # # # #
# ]
1793 [ # # ]: 0 : if (pg_checksum_update(&checksum_ctx,
1794 : 0 : (uint8 *) sink->bbs_buffer,
1795 : : nbytes) < 0)
1796 [ # # ]: 0 : elog(ERROR, "could not update checksum of base backup");
1797 : 0 : bbsink_archive_contents(sink, nbytes);
194 rhaas@postgresql.org 1798 :UNC 0 : bytes_done += nbytes;
1799 : : }
1800 : :
1801 : : /*
1802 : : * Pad to a block boundary, per tar format requirements. (This small piece
1803 : : * of data is probably not worth throttling, and is not checksummed
1804 : : * because it's not actually part of the file.)
1805 : : */
194 rhaas@postgresql.org 1806 :GNC 144122 : _tarWritePadding(sink, bytes_done);
1807 : :
1397 rhaas@postgresql.org 1808 :CBC 144122 : CloseTransientFile(fd);
1809 : :
2203 magnus@hagander.net 1810 [ + + ]: 144122 : if (checksum_failures > 1)
1811 : : {
1812 [ + - ]: 2 : ereport(WARNING,
1813 : : (errmsg_plural("file \"%s\" has a total of %d checksum verification failure",
1814 : : "file \"%s\" has a total of %d checksum verification failures",
1815 : : checksum_failures,
1816 : : readfilename, checksum_failures)));
1817 : :
1829 1818 : 2 : pgstat_report_checksum_failures_in_db(dboid, checksum_failures);
1819 : : }
1820 : :
2203 1821 : 144122 : total_checksum_failures += checksum_failures;
1822 : :
1452 rhaas@postgresql.org 1823 : 144122 : AddFileToBackupManifest(manifest, spcoid, tarfilename, statbuf->st_size,
1824 : 144122 : (pg_time_t) statbuf->st_mtime, &checksum_ctx);
1825 : :
4132 heikki.linnakangas@i 1826 : 144122 : return true;
1827 : : }
1828 : :
1829 : : /*
1830 : : * Read some more data from the file into the bbsink's buffer, verifying
1831 : : * checksums as required.
1832 : : *
1833 : : * 'offset' is the file offset from which we should begin to read, and
1834 : : * 'length' is the amount of data that should be read. The actual amount
1835 : : * of data read will be less than the requested amount if the bbsink's
1836 : : * buffer isn't big enough to hold it all, or if the underlying file has
1837 : : * been truncated. The return value is the number of bytes actually read.
1838 : : *
1839 : : * 'blkno' is the block number of the first page in the bbsink's buffer
1840 : : * relative to the start of the relation.
1841 : : *
1842 : : * 'verify_checksum' indicates whether we should try to verify checksums
1843 : : * for the blocks we read. If we do this, we'll update *checksum_failures
1844 : : * and issue warnings as appropriate.
1845 : : */
1846 : : static off_t
194 rhaas@postgresql.org 1847 :GNC 168260 : read_file_data_into_buffer(bbsink *sink, const char *readfilename, int fd,
1848 : : off_t offset, size_t length, BlockNumber blkno,
1849 : : bool verify_checksum, int *checksum_failures)
1850 : : {
1851 : : off_t cnt;
1852 : : int i;
1853 : : char *page;
1854 : :
1855 : : /* Try to read some more data. */
1856 : 168260 : cnt = basebackup_read_file(fd, sink->bbs_buffer,
1857 : 168260 : Min(sink->bbs_buffer_length, length),
1858 : : offset, readfilename, true);
1859 : :
1860 : : /* Can't verify checksums if read length is not a multiple of BLCKSZ. */
1861 [ + + - + ]: 168260 : if (!verify_checksum || (cnt % BLCKSZ) != 0)
1862 : 133371 : return cnt;
1863 : :
1864 : : /* Verify checksum for each block. */
1865 [ + + ]: 121438 : for (i = 0; i < cnt / BLCKSZ; i++)
1866 : : {
1867 : : int reread_cnt;
1868 : : uint16 expected_checksum;
1869 : :
1870 : 86549 : page = sink->bbs_buffer + BLCKSZ * i;
1871 : :
1872 : : /* If the page is OK, go on to the next one. */
1873 [ + + ]: 86549 : if (verify_page_checksum(page, sink->bbs_state->startptr, blkno + i,
1874 : : &expected_checksum))
1875 : 86535 : continue;
1876 : :
1877 : : /*
1878 : : * Retry the block on the first failure. It's possible that we read
1879 : : * the first 4K page of the block just before postgres updated the
1880 : : * entire block so it ends up looking torn to us. If, before we retry
1881 : : * the read, the concurrent write of the block finishes, the page LSN
1882 : : * will be updated and we'll realize that we should ignore this block.
1883 : : *
1884 : : * There's no guarantee that this will actually happen, though: the
1885 : : * torn write could take an arbitrarily long time to complete.
1886 : : * Retrying multiple times wouldn't fix this problem, either, though
1887 : : * it would reduce the chances of it happening in practice. The only
1888 : : * real fix here seems to be to have some kind of interlock that
1889 : : * allows us to wait until we can be certain that no write to the
1890 : : * block is in progress. Since we don't have any such thing right now,
1891 : : * we just do this and hope for the best.
1892 : : */
1893 : 14 : reread_cnt =
1894 : 14 : basebackup_read_file(fd, sink->bbs_buffer + BLCKSZ * i,
1895 : 14 : BLCKSZ, offset + BLCKSZ * i,
1896 : : readfilename, false);
1897 [ - + ]: 14 : if (reread_cnt == 0)
1898 : : {
1899 : : /*
1900 : : * If we hit end-of-file, a concurrent truncation must have
1901 : : * occurred, so reduce cnt to reflect only the blocks already
1902 : : * processed and break out of this loop.
1903 : : */
194 rhaas@postgresql.org 1904 :UNC 0 : cnt = BLCKSZ * i;
1905 : 0 : break;
1906 : : }
1907 : :
1908 : : /* If the page now looks OK, go on to the next one. */
194 rhaas@postgresql.org 1909 [ - + ]:GNC 14 : if (verify_page_checksum(page, sink->bbs_state->startptr, blkno + i,
1910 : : &expected_checksum))
194 rhaas@postgresql.org 1911 :UNC 0 : continue;
1912 : :
1913 : : /* Handle checksum failure. */
194 rhaas@postgresql.org 1914 :GNC 14 : (*checksum_failures)++;
1915 [ + + ]: 14 : if (*checksum_failures <= 5)
1916 [ + - ]: 12 : ereport(WARNING,
1917 : : (errmsg("checksum verification failed in "
1918 : : "file \"%s\", block %u: calculated "
1919 : : "%X but expected %X",
1920 : : readfilename, blkno + i, expected_checksum,
1921 : : ((PageHeader) page)->pd_checksum)));
1922 [ + + ]: 14 : if (*checksum_failures == 5)
1923 [ + - ]: 2 : ereport(WARNING,
1924 : : (errmsg("further checksum verification "
1925 : : "failures in file \"%s\" will not "
1926 : : "be reported", readfilename)));
1927 : : }
1928 : :
1929 : 34889 : return cnt;
1930 : : }
1931 : :
1932 : : /*
1933 : : * Push data into a bbsink.
1934 : : *
1935 : : * It's better, when possible, to read data directly into the bbsink's buffer,
1936 : : * rather than using this function to copy it into the buffer; this function is
1937 : : * for cases where that approach is not practical.
1938 : : *
1939 : : * bytes_done should point to a count of the number of bytes that are
1940 : : * currently used in the bbsink's buffer. Upon return, the bytes identified by
1941 : : * data and length will have been copied into the bbsink's buffer, flushing
1942 : : * as required, and *bytes_done will have been updated accordingly. If the
1943 : : * buffer was flushed, the previous contents will also have been fed to
1944 : : * checksum_ctx.
1945 : : *
1946 : : * Note that after one or more calls to this function it is the caller's
1947 : : * responsibility to perform any required final flush.
1948 : : */
1949 : : static void
116 1950 : 18627 : push_to_sink(bbsink *sink, pg_checksum_context *checksum_ctx,
1951 : : size_t *bytes_done, void *data, size_t length)
1952 : : {
1953 [ + + ]: 18627 : while (length > 0)
1954 : : {
1955 : : size_t bytes_to_copy;
1956 : :
1957 : : /*
1958 : : * We use < here rather than <= so that if the data exactly fills the
1959 : : * remaining buffer space, we trigger a flush now.
1960 : : */
1961 [ + - ]: 13999 : if (length < sink->bbs_buffer_length - *bytes_done)
1962 : : {
1963 : : /* Append remaining data to buffer. */
1964 : 13999 : memcpy(sink->bbs_buffer + *bytes_done, data, length);
1965 : 13999 : *bytes_done += length;
1966 : 13999 : return;
1967 : : }
1968 : :
1969 : : /* Copy until buffer is full and flush it. */
116 rhaas@postgresql.org 1970 :UNC 0 : bytes_to_copy = sink->bbs_buffer_length - *bytes_done;
1971 : 0 : memcpy(sink->bbs_buffer + *bytes_done, data, bytes_to_copy);
1972 : 0 : data = ((char *) data) + bytes_to_copy;
1973 : 0 : length -= bytes_to_copy;
1974 : 0 : bbsink_archive_contents(sink, sink->bbs_buffer_length);
1975 [ # # ]: 0 : if (pg_checksum_update(checksum_ctx, (uint8 *) sink->bbs_buffer,
1976 : : sink->bbs_buffer_length) < 0)
1977 [ # # ]: 0 : elog(ERROR, "could not update checksum");
1978 : 0 : *bytes_done = 0;
1979 : : }
1980 : : }
1981 : :
1982 : : /*
1983 : : * Try to verify the checksum for the provided page, if it seems appropriate
1984 : : * to do so.
1985 : : *
1986 : : * Returns true if verification succeeds or if we decide not to check it,
1987 : : * and false if verification fails. When return false, it also sets
1988 : : * *expected_checksum to the computed value.
1989 : : */
1990 : : static bool
194 rhaas@postgresql.org 1991 :GNC 86563 : verify_page_checksum(Page page, XLogRecPtr start_lsn, BlockNumber blkno,
1992 : : uint16 *expected_checksum)
1993 : : {
1994 : : PageHeader phdr;
1995 : : uint16 checksum;
1996 : :
1997 : : /*
1998 : : * Only check pages which have not been modified since the start of the
1999 : : * base backup. Otherwise, they might have been written only halfway and
2000 : : * the checksum would not be valid. However, replaying WAL would
2001 : : * reinstate the correct page in this case. We also skip completely new
2002 : : * pages, since they don't have a checksum yet.
2003 : : */
2004 [ + + - + ]: 86563 : if (PageIsNew(page) || PageGetLSN(page) >= start_lsn)
2005 : 76 : return true;
2006 : :
2007 : : /* Perform the actual checksum calculation. */
2008 : 86487 : checksum = pg_checksum_page(page, blkno);
2009 : :
2010 : : /* See whether it matches the value from the page. */
2011 : 86487 : phdr = (PageHeader) page;
2012 [ + + ]: 86487 : if (phdr->pd_checksum == checksum)
2013 : 86459 : return true;
2014 : 28 : *expected_checksum = checksum;
2015 : 28 : return false;
2016 : : }
2017 : :
2018 : : static int64
891 rhaas@postgresql.org 2019 :CBC 152173 : _tarWriteHeader(bbsink *sink, const char *filename, const char *linktarget,
2020 : : struct stat *statbuf, bool sizeonly)
2021 : : {
2022 : : enum tarError rc;
2023 : :
2755 peter_e@gmx.net 2024 [ + + ]: 152173 : if (!sizeonly)
2025 : : {
2026 : : /*
2027 : : * As of this writing, the smallest supported block size is 1kB, which
2028 : : * is twice TAR_BLOCK_SIZE. Since the buffer size is required to be a
2029 : : * multiple of BLCKSZ, it should be safe to assume that the buffer is
2030 : : * large enough to fit an entire tar block. We double-check by means
2031 : : * of these assertions.
2032 : : */
2033 : : StaticAssertDecl(TAR_BLOCK_SIZE <= BLCKSZ,
2034 : : "BLCKSZ too small for tar block");
891 rhaas@postgresql.org 2035 [ - + ]: 148198 : Assert(sink->bbs_buffer_length >= TAR_BLOCK_SIZE);
2036 : :
2037 : 148198 : rc = tarCreateHeader(sink->bbs_buffer, filename, linktarget,
2038 : : statbuf->st_size, statbuf->st_mode,
2039 : : statbuf->st_uid, statbuf->st_gid,
2040 : : statbuf->st_mtime);
2041 : :
2755 peter_e@gmx.net 2042 [ + + - - ]: 148198 : switch (rc)
2043 : : {
2044 : 148197 : case TAR_OK:
2045 : 148197 : break;
2046 : 1 : case TAR_NAME_TOO_LONG:
2047 [ + - ]: 1 : ereport(ERROR,
2048 : : (errmsg("file name too long for tar format: \"%s\"",
2049 : : filename)));
2050 : : break;
2755 peter_e@gmx.net 2051 :UBC 0 : case TAR_SYMLINK_TOO_LONG:
2052 [ # # ]: 0 : ereport(ERROR,
2053 : : (errmsg("symbolic link target too long for tar format: "
2054 : : "file name \"%s\", target \"%s\"",
2055 : : filename, linktarget)));
2056 : : break;
2057 : 0 : default:
2058 [ # # ]: 0 : elog(ERROR, "unrecognized tar error: %d", rc);
2059 : : }
2060 : :
891 rhaas@postgresql.org 2061 :CBC 148197 : bbsink_archive_contents(sink, TAR_BLOCK_SIZE);
2062 : : }
2063 : :
2064 : 152172 : return TAR_BLOCK_SIZE;
2065 : : }
2066 : :
2067 : : /*
2068 : : * Pad with zero bytes out to a multiple of TAR_BLOCK_SIZE.
2069 : : */
2070 : : static void
2071 : 144315 : _tarWritePadding(bbsink *sink, int len)
2072 : : {
2073 : 144315 : int pad = tarPaddingBytesRequired(len);
2074 : :
2075 : : /*
2076 : : * As in _tarWriteHeader, it should be safe to assume that the buffer is
2077 : : * large enough that we don't need to do this in multiple chunks.
2078 : : */
2079 [ - + ]: 144315 : Assert(sink->bbs_buffer_length >= TAR_BLOCK_SIZE);
2080 [ - + ]: 144315 : Assert(pad <= TAR_BLOCK_SIZE);
2081 : :
2082 [ + + ]: 144315 : if (pad > 0)
2083 : : {
2084 [ + - + + : 32574 : MemSet(sink->bbs_buffer, 0, pad);
+ - + - +
+ ]
2085 : 6727 : bbsink_archive_contents(sink, pad);
2086 : : }
3699 alvherre@alvh.no-ip. 2087 : 144315 : }
2088 : :
2089 : : /*
2090 : : * If the entry in statbuf is a link, then adjust statbuf to make it look like a
2091 : : * directory, so that it will be written that way.
2092 : : */
2093 : : static void
891 rhaas@postgresql.org 2094 : 2349 : convert_link_to_directory(const char *pathbuf, struct stat *statbuf)
2095 : : {
2096 : : /* If symlink, write it as a directory anyway */
2097 [ + + ]: 2349 : if (S_ISLNK(statbuf->st_mode))
2098 : 63 : statbuf->st_mode = S_IFDIR | pg_dir_create_mode;
1503 fujii@postgresql.org 2099 : 2349 : }
2100 : :
2101 : : /*
2102 : : * Read some data from a file, setting a wait event and reporting any error
2103 : : * encountered.
2104 : : *
2105 : : * If partial_read_ok is false, also report an error if the number of bytes
2106 : : * read is not equal to the number of bytes requested.
2107 : : *
2108 : : * Returns the number of bytes read.
2109 : : */
2110 : : static ssize_t
1397 rhaas@postgresql.org 2111 : 178514 : basebackup_read_file(int fd, char *buf, size_t nbytes, off_t offset,
2112 : : const char *filename, bool partial_read_ok)
2113 : : {
2114 : : ssize_t rc;
2115 : :
2116 : 178514 : pgstat_report_wait_start(WAIT_EVENT_BASEBACKUP_READ);
563 tmunro@postgresql.or 2117 : 178514 : rc = pg_pread(fd, buf, nbytes, offset);
1397 rhaas@postgresql.org 2118 : 178514 : pgstat_report_wait_end();
2119 : :
2120 [ - + ]: 178514 : if (rc < 0)
1397 rhaas@postgresql.org 2121 [ # # ]:UBC 0 : ereport(ERROR,
2122 : : (errcode_for_file_access(),
2123 : : errmsg("could not read file \"%s\": %m", filename)));
1397 rhaas@postgresql.org 2124 [ + + + - :CBC 178514 : if (!partial_read_ok && rc > 0 && rc != nbytes)
- + ]
1397 rhaas@postgresql.org 2125 [ # # ]:UBC 0 : ereport(ERROR,
2126 : : (errcode_for_file_access(),
2127 : : errmsg("could not read file \"%s\": read %zd of %zu",
2128 : : filename, rc, nbytes)));
2129 : :
1397 rhaas@postgresql.org 2130 :CBC 178514 : return rc;
2131 : : }
|
