Age Owner Branch data TLA Line data Source code
1 : : /*-------------------------------------------------------------------------
2 : : *
3 : : * astreamer_file.c
4 : : *
5 : : * Archive streamers that write to files. astreamer_plain_writer writes
6 : : * the whole archive to a single file, and astreamer_extractor writes
7 : : * each archive member to a separate file in a given directory.
8 : : *
9 : : * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group
10 : : *
11 : : * IDENTIFICATION
12 : : * src/fe_utils/astreamer_file.c
13 : : *-------------------------------------------------------------------------
14 : : */
15 : :
16 : : #include "postgres_fe.h"
17 : :
18 : : #include <unistd.h>
19 : :
20 : : #include "common/file_perm.h"
21 : : #include "common/logging.h"
22 : : #include "fe_utils/astreamer.h"
23 : :
24 : : typedef struct astreamer_plain_writer
25 : : {
26 : : astreamer base;
27 : : char *pathname;
28 : : FILE *file;
29 : : bool should_close_file;
30 : : } astreamer_plain_writer;
31 : :
32 : : typedef struct astreamer_extractor
33 : : {
34 : : astreamer base;
35 : : char *basepath;
36 : : const char *(*link_map) (const char *);
37 : : void (*report_output_file) (const char *);
38 : : char filename[MAXPGPATH];
39 : : FILE *file;
40 : : } astreamer_extractor;
41 : :
42 : : static void astreamer_plain_writer_content(astreamer *streamer,
43 : : astreamer_member *member,
44 : : const char *data, int len,
45 : : astreamer_archive_context context);
46 : : static void astreamer_plain_writer_finalize(astreamer *streamer);
47 : : static void astreamer_plain_writer_free(astreamer *streamer);
48 : :
49 : : static const astreamer_ops astreamer_plain_writer_ops = {
50 : : .content = astreamer_plain_writer_content,
51 : : .finalize = astreamer_plain_writer_finalize,
52 : : .free = astreamer_plain_writer_free
53 : : };
54 : :
55 : : static void astreamer_extractor_content(astreamer *streamer,
56 : : astreamer_member *member,
57 : : const char *data, int len,
58 : : astreamer_archive_context context);
59 : : static void astreamer_extractor_finalize(astreamer *streamer);
60 : : static void astreamer_extractor_free(astreamer *streamer);
61 : : static void extract_directory(const char *filename, mode_t mode);
62 : : static void extract_link(const char *filename, const char *linktarget);
63 : : static FILE *create_file_for_extract(const char *filename, mode_t mode);
64 : :
65 : : static const astreamer_ops astreamer_extractor_ops = {
66 : : .content = astreamer_extractor_content,
67 : : .finalize = astreamer_extractor_finalize,
68 : : .free = astreamer_extractor_free
69 : : };
70 : :
71 : : /*
72 : : * Create a astreamer that just writes data to a file.
73 : : *
74 : : * The caller must specify a pathname and may specify a file. The pathname is
75 : : * used for error-reporting purposes either way. If file is NULL, the pathname
76 : : * also identifies the file to which the data should be written: it is opened
77 : : * for writing and closed when done. If file is not NULL, the data is written
78 : : * there.
79 : : */
80 : : astreamer *
638 rhaas@postgresql.org 81 :CBC 22 : astreamer_plain_writer_new(char *pathname, FILE *file)
82 : : {
83 : : astreamer_plain_writer *streamer;
84 : :
147 michael@paquier.xyz 85 :GNC 22 : streamer = palloc0_object(astreamer_plain_writer);
638 rhaas@postgresql.org 86 :CBC 22 : *((const astreamer_ops **) &streamer->base.bbs_ops) =
87 : : &astreamer_plain_writer_ops;
88 : :
1642 89 : 22 : streamer->pathname = pstrdup(pathname);
90 : 22 : streamer->file = file;
91 : :
92 [ + - ]: 22 : if (file == NULL)
93 : : {
94 : 22 : streamer->file = fopen(pathname, "wb");
95 [ - + ]: 22 : if (streamer->file == NULL)
1488 tgl@sss.pgh.pa.us 96 :UBC 0 : pg_fatal("could not create file \"%s\": %m", pathname);
1642 rhaas@postgresql.org 97 :CBC 22 : streamer->should_close_file = true;
98 : : }
99 : :
100 : 22 : return &streamer->base;
101 : : }
102 : :
103 : : /*
104 : : * Write archive content to file.
105 : : */
106 : : static void
638 107 : 30756 : astreamer_plain_writer_content(astreamer *streamer,
108 : : astreamer_member *member, const char *data,
109 : : int len, astreamer_archive_context context)
110 : : {
111 : : astreamer_plain_writer *mystreamer;
112 : :
113 : 30756 : mystreamer = (astreamer_plain_writer *) streamer;
114 : :
1642 115 [ - + ]: 30756 : if (len == 0)
1642 rhaas@postgresql.org 116 :UBC 0 : return;
117 : :
1642 rhaas@postgresql.org 118 :CBC 30756 : errno = 0;
119 [ - + ]: 30756 : if (fwrite(data, len, 1, mystreamer->file) != 1)
120 : : {
121 : : /* if write didn't set errno, assume problem is no disk space */
1642 rhaas@postgresql.org 122 [ # # ]:UBC 0 : if (errno == 0)
123 : 0 : errno = ENOSPC;
1488 tgl@sss.pgh.pa.us 124 : 0 : pg_fatal("could not write to file \"%s\": %m",
125 : : mystreamer->pathname);
126 : : }
127 : : }
128 : :
129 : : /*
130 : : * End-of-archive processing when writing to a plain file consists of closing
131 : : * the file if we opened it, but not if the caller provided it.
132 : : */
133 : : static void
638 rhaas@postgresql.org 134 :CBC 22 : astreamer_plain_writer_finalize(astreamer *streamer)
135 : : {
136 : : astreamer_plain_writer *mystreamer;
137 : :
138 : 22 : mystreamer = (astreamer_plain_writer *) streamer;
139 : :
1642 140 [ + - - + ]: 22 : if (mystreamer->should_close_file && fclose(mystreamer->file) != 0)
1488 tgl@sss.pgh.pa.us 141 :UBC 0 : pg_fatal("could not close file \"%s\": %m",
142 : : mystreamer->pathname);
143 : :
1642 rhaas@postgresql.org 144 :CBC 22 : mystreamer->file = NULL;
145 : 22 : mystreamer->should_close_file = false;
146 : 22 : }
147 : :
148 : : /*
149 : : * Free memory associated with this astreamer.
150 : : */
151 : : static void
638 152 : 22 : astreamer_plain_writer_free(astreamer *streamer)
153 : : {
154 : : astreamer_plain_writer *mystreamer;
155 : :
156 : 22 : mystreamer = (astreamer_plain_writer *) streamer;
157 : :
1642 158 [ - + ]: 22 : Assert(!mystreamer->should_close_file);
159 [ - + ]: 22 : Assert(mystreamer->base.bbs_next == NULL);
160 : :
161 : 22 : pfree(mystreamer->pathname);
162 : 22 : pfree(mystreamer);
163 : 22 : }
164 : :
165 : : /*
166 : : * Create a astreamer that extracts an archive.
167 : : *
168 : : * All pathnames in the archive are interpreted relative to basepath.
169 : : *
170 : : * Unlike e.g. astreamer_plain_writer_new() we can't do anything useful here
171 : : * with untyped chunks; we need typed chunks which follow the rules described
172 : : * in astreamer.h. Assuming we have that, we don't need to worry about the
173 : : * original archive format; it's enough to just look at the member information
174 : : * provided and write to the corresponding file.
175 : : *
176 : : * 'link_map' is a function that will be applied to the target of any
177 : : * symbolic link, and which should return a replacement pathname to be used
178 : : * in its place. If NULL, the symbolic link target is used without
179 : : * modification.
180 : : *
181 : : * 'report_output_file' is a function that will be called each time we open a
182 : : * new output file. The pathname to that file is passed as an argument. If
183 : : * NULL, the call is skipped.
184 : : */
185 : : astreamer *
638 186 : 179 : astreamer_extractor_new(const char *basepath,
187 : : const char *(*link_map) (const char *),
188 : : void (*report_output_file) (const char *))
189 : : {
190 : : astreamer_extractor *streamer;
191 : :
147 michael@paquier.xyz 192 :GNC 179 : streamer = palloc0_object(astreamer_extractor);
638 rhaas@postgresql.org 193 :CBC 179 : *((const astreamer_ops **) &streamer->base.bbs_ops) =
194 : : &astreamer_extractor_ops;
1642 195 : 179 : streamer->basepath = pstrdup(basepath);
196 : 179 : streamer->link_map = link_map;
197 : 179 : streamer->report_output_file = report_output_file;
198 : :
199 : 179 : return &streamer->base;
200 : : }
201 : :
202 : : /*
203 : : * Extract archive contents to the filesystem.
204 : : */
205 : : static void
638 206 : 509743 : astreamer_extractor_content(astreamer *streamer, astreamer_member *member,
207 : : const char *data, int len,
208 : : astreamer_archive_context context)
209 : : {
210 : 509743 : astreamer_extractor *mystreamer = (astreamer_extractor *) streamer;
211 : : int fnamelen;
212 : :
213 [ + + - + ]: 509743 : Assert(member != NULL || context == ASTREAMER_ARCHIVE_TRAILER);
214 [ - + ]: 509743 : Assert(context != ASTREAMER_UNKNOWN);
215 : :
1642 216 [ + + + + : 509743 : switch (context)
- ]
217 : : {
638 218 : 161407 : case ASTREAMER_MEMBER_HEADER:
1642 219 [ - + ]: 161407 : Assert(mystreamer->file == NULL);
220 : :
221 : : /* Prepend basepath. */
222 : 161407 : snprintf(mystreamer->filename, sizeof(mystreamer->filename),
223 : 161407 : "%s/%s", mystreamer->basepath, member->pathname);
224 : :
225 : : /* Remove any trailing slash. */
226 : 161407 : fnamelen = strlen(mystreamer->filename);
227 [ + + ]: 161407 : if (mystreamer->filename[fnamelen - 1] == '/')
228 : 3947 : mystreamer->filename[fnamelen - 1] = '\0';
229 : :
230 : : /* Dispatch based on file type. */
33 tgl@sss.pgh.pa.us 231 [ + + ]: 161407 : if (member->is_regular)
232 : 157460 : mystreamer->file =
233 : 157460 : create_file_for_extract(mystreamer->filename,
234 : : member->mode);
235 [ + + ]: 3947 : else if (member->is_directory)
1642 rhaas@postgresql.org 236 : 3931 : extract_directory(mystreamer->filename, member->mode);
33 tgl@sss.pgh.pa.us 237 [ + - ]: 16 : else if (member->is_symlink)
238 : : {
1642 rhaas@postgresql.org 239 : 16 : const char *linktarget = member->linktarget;
240 : :
241 [ + - ]: 16 : if (mystreamer->link_map)
242 : 16 : linktarget = mystreamer->link_map(linktarget);
243 : 16 : extract_link(mystreamer->filename, linktarget);
244 : : }
245 : :
246 : : /* Report output file change. */
247 [ + - ]: 161407 : if (mystreamer->report_output_file)
248 : 161407 : mystreamer->report_output_file(mystreamer->filename);
249 : 161407 : break;
250 : :
638 251 : 186752 : case ASTREAMER_MEMBER_CONTENTS:
1642 252 [ - + ]: 186752 : if (mystreamer->file == NULL)
1642 rhaas@postgresql.org 253 :UBC 0 : break;
254 : :
1642 rhaas@postgresql.org 255 :CBC 186752 : errno = 0;
256 [ + + - + ]: 186752 : if (len > 0 && fwrite(data, len, 1, mystreamer->file) != 1)
257 : : {
258 : : /* if write didn't set errno, assume problem is no disk space */
1642 rhaas@postgresql.org 259 [ # # ]:UBC 0 : if (errno == 0)
260 : 0 : errno = ENOSPC;
1488 tgl@sss.pgh.pa.us 261 : 0 : pg_fatal("could not write to file \"%s\": %m",
262 : : mystreamer->filename);
263 : : }
1642 rhaas@postgresql.org 264 :CBC 186752 : break;
265 : :
638 266 : 161407 : case ASTREAMER_MEMBER_TRAILER:
1642 267 [ + + ]: 161407 : if (mystreamer->file == NULL)
268 : 3947 : break;
43 andrew@dunslane.net 269 [ - + ]: 157460 : if (fclose(mystreamer->file) != 0)
43 andrew@dunslane.net 270 :UBC 0 : pg_fatal("could not close file \"%s\": %m",
271 : : mystreamer->filename);
1642 rhaas@postgresql.org 272 :CBC 157460 : mystreamer->file = NULL;
273 : 157460 : break;
274 : :
638 275 : 177 : case ASTREAMER_ARCHIVE_TRAILER:
1642 276 : 177 : break;
277 : :
1642 rhaas@postgresql.org 278 :UBC 0 : default:
279 : : /* Shouldn't happen. */
1488 tgl@sss.pgh.pa.us 280 : 0 : pg_fatal("unexpected state while extracting archive");
281 : : }
1642 rhaas@postgresql.org 282 :CBC 509743 : }
283 : :
284 : : /*
285 : : * Should we tolerate an already-existing directory?
286 : : *
287 : : * When streaming WAL, pg_wal (or pg_xlog for pre-9.6 clusters) will have been
288 : : * created by the wal receiver process. Also, when the WAL directory location
289 : : * was specified, pg_wal (or pg_xlog) has already been created as a symbolic
290 : : * link before starting the actual backup. So just ignore creation failures
291 : : * on related directories.
292 : : *
293 : : * If in-place tablespaces are used, pg_tblspc and subdirectories may already
294 : : * exist when we get here. So tolerate that case, too.
295 : : */
296 : : static bool
1113 297 : 433 : should_allow_existing_directory(const char *pathname)
298 : : {
299 : 433 : const char *filename = last_dir_separator(pathname) + 1;
300 : :
301 [ + + ]: 433 : if (strcmp(filename, "pg_wal") == 0 ||
302 [ + - ]: 296 : strcmp(filename, "pg_xlog") == 0 ||
303 [ + + ]: 296 : strcmp(filename, "archive_status") == 0 ||
867 304 [ + + ]: 159 : strcmp(filename, "summaries") == 0 ||
1113 305 [ + + ]: 22 : strcmp(filename, "pg_tblspc") == 0)
306 : 419 : return true;
307 : :
308 [ + - ]: 14 : if (strspn(filename, "0123456789") == strlen(filename))
309 : : {
310 : 14 : const char *pg_tblspc = strstr(pathname, "/pg_tblspc/");
311 : :
312 [ + - + - ]: 14 : return pg_tblspc != NULL && pg_tblspc + 11 == filename;
313 : : }
314 : :
1113 rhaas@postgresql.org 315 :UBC 0 : return false;
316 : : }
317 : :
318 : : /*
319 : : * Create a directory.
320 : : */
321 : : static void
1642 rhaas@postgresql.org 322 :CBC 3931 : extract_directory(const char *filename, mode_t mode)
323 : : {
1113 324 [ + + ]: 3931 : if (mkdir(filename, pg_dir_create_mode) != 0 &&
325 [ + - - + ]: 433 : (errno != EEXIST || !should_allow_existing_directory(filename)))
1113 rhaas@postgresql.org 326 :UBC 0 : pg_fatal("could not create directory \"%s\": %m",
327 : : filename);
328 : :
329 : : #ifndef WIN32
1642 rhaas@postgresql.org 330 [ - + ]:CBC 3931 : if (chmod(filename, mode))
1488 tgl@sss.pgh.pa.us 331 :UBC 0 : pg_fatal("could not set permissions on directory \"%s\": %m",
332 : : filename);
333 : : #endif
1642 rhaas@postgresql.org 334 :CBC 3931 : }
335 : :
336 : : /*
337 : : * Create a symbolic link.
338 : : *
339 : : * It's most likely a link in pg_tblspc directory, to the location of a
340 : : * tablespace. Apply any tablespace mapping given on the command line
341 : : * (--tablespace-mapping). (We blindly apply the mapping without checking that
342 : : * the link really is inside pg_tblspc. We don't expect there to be other
343 : : * symlinks in a data directory, but if there are, you can call it an
344 : : * undocumented feature that you can map them too.)
345 : : */
346 : : static void
347 : 16 : extract_link(const char *filename, const char *linktarget)
348 : : {
349 [ - + ]: 16 : if (symlink(linktarget, filename) != 0)
1488 tgl@sss.pgh.pa.us 350 :UBC 0 : pg_fatal("could not create symbolic link from \"%s\" to \"%s\": %m",
351 : : filename, linktarget);
1642 rhaas@postgresql.org 352 :CBC 16 : }
353 : :
354 : : /*
355 : : * Create a regular file.
356 : : *
357 : : * Return the resulting handle so we can write the content to the file.
358 : : */
359 : : static FILE *
360 : 157460 : create_file_for_extract(const char *filename, mode_t mode)
361 : : {
362 : : FILE *file;
363 : :
364 : 157460 : file = fopen(filename, "wb");
365 [ - + ]: 157460 : if (file == NULL)
1488 tgl@sss.pgh.pa.us 366 :UBC 0 : pg_fatal("could not create file \"%s\": %m", filename);
367 : :
368 : : #ifndef WIN32
1642 rhaas@postgresql.org 369 [ - + ]:CBC 157460 : if (chmod(filename, mode))
1488 tgl@sss.pgh.pa.us 370 :UBC 0 : pg_fatal("could not set permissions on file \"%s\": %m",
371 : : filename);
372 : : #endif
373 : :
1642 rhaas@postgresql.org 374 :CBC 157460 : return file;
375 : : }
376 : :
377 : : /*
378 : : * End-of-stream processing for extracting an archive.
379 : : *
380 : : * There's nothing to do here but sanity checking.
381 : : */
382 : : static void
638 383 : 177 : astreamer_extractor_finalize(astreamer *streamer)
384 : : {
385 : 177 : astreamer_extractor *mystreamer PG_USED_FOR_ASSERTS_ONLY
386 : : = (astreamer_extractor *) streamer;
387 : :
1642 388 [ - + ]: 177 : Assert(mystreamer->file == NULL);
389 : 177 : }
390 : :
391 : : /*
392 : : * Free memory.
393 : : */
394 : : static void
638 395 : 177 : astreamer_extractor_free(astreamer *streamer)
396 : : {
397 : 177 : astreamer_extractor *mystreamer = (astreamer_extractor *) streamer;
398 : :
1642 399 : 177 : pfree(mystreamer->basepath);
400 : 177 : pfree(mystreamer);
401 : 177 : }
|
