2 This file is part of GNUnet.
3 Copyright (C) 2003, 2004, 2006, 2009 GNUnet e.V.
5 GNUnet is free software: you can redistribute it and/or modify it
6 under the terms of the GNU Affero General Public License as published
7 by the Free Software Foundation, either version 3 of the License,
8 or (at your option) any later version.
10 GNUnet is distributed in the hope that it will be useful, but
11 WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 Affero General Public License for more details.
15 You should have received a copy of the GNU Affero General Public License
16 along with this program. If not, see <http://www.gnu.org/licenses/>.
18 SPDX-License-Identifier: AGPL3.0-or-later
22 * @file fs/fs_directory.c
23 * @brief Helper functions for building directories.
24 * @author Christian Grothoff
27 * - modify directory builder API to support incremental
28 * generation of directories (to allow directories that
29 * would not fit into memory to be created)
30 * - modify directory processor API to support incremental
31 * iteration over FULL directories (without missing entries)
32 * to allow access to directories that do not fit entirely
36 #include "gnunet_fs_service.h"
40 * String that is used to indicate that a file
41 * is a GNUnet directory.
43 #define GNUNET_DIRECTORY_MAGIC "\211GND\r\n\032\n"
47 * Does the meta-data claim that this is a directory?
48 * Checks if the mime-type is that of a GNUnet directory.
50 * @return #GNUNET_YES if it is, #GNUNET_NO if it is not, #GNUNET_SYSERR if
51 * we have no mime-type information (treat as #GNUNET_NO)
54 GNUNET_FS_meta_data_test_for_directory (const struct
55 GNUNET_CONTAINER_MetaData *md)
62 mime = GNUNET_CONTAINER_meta_data_get_by_type (md,
63 EXTRACTOR_METATYPE_MIMETYPE);
66 ret = (0 == strcasecmp (mime, GNUNET_FS_DIRECTORY_MIME)) ? GNUNET_YES :
74 * Set the MIMETYPE information for the given
75 * metadata to "application/gnunet-directory".
77 * @param md metadata to add mimetype to
80 GNUNET_FS_meta_data_make_directory (struct GNUNET_CONTAINER_MetaData *md)
85 GNUNET_CONTAINER_meta_data_get_by_type (md, EXTRACTOR_METATYPE_MIMETYPE);
88 GNUNET_break (0 == strcmp (mime, GNUNET_FS_DIRECTORY_MIME));
92 GNUNET_CONTAINER_meta_data_insert (md, "<gnunet>",
93 EXTRACTOR_METATYPE_MIMETYPE,
94 EXTRACTOR_METAFORMAT_UTF8, "text/plain",
95 GNUNET_FS_DIRECTORY_MIME,
96 strlen (GNUNET_FS_DIRECTORY_MIME) + 1);
101 * Closure for 'find_full_data'.
103 struct GetFullDataClosure
106 * Extracted binary meta data.
111 * Number of bytes stored in data.
118 * Type of a function that libextractor calls for each
119 * meta data item found.
121 * @param cls closure (user-defined)
122 * @param plugin_name name of the plugin that produced this value;
123 * special values can be used (i.e. '<zlib>' for zlib being
124 * used in the main libextractor library and yielding
126 * @param type libextractor-type describing the meta data
127 * @param format basic format information about data
128 * @param data_mime_type mime-type of data (not of the original file);
129 * can be NULL (if mime-type is not known)
130 * @param data actual meta-data found
131 * @param data_len number of bytes in data
132 * @return 0 to continue extracting, 1 to abort
135 find_full_data (void *cls, const char *plugin_name,
136 enum EXTRACTOR_MetaType type, enum EXTRACTOR_MetaFormat format,
137 const char *data_mime_type, const char *data, size_t data_len)
139 struct GetFullDataClosure *gfdc = cls;
141 if (type == EXTRACTOR_METATYPE_GNUNET_FULL_DATA)
143 gfdc->size = data_len;
146 gfdc->data = GNUNET_malloc (data_len);
147 GNUNET_memcpy (gfdc->data, data, data_len);
156 * Iterate over all entries in a directory. Note that directories
157 * are structured such that it is possible to iterate over the
158 * individual blocks as well as over the entire directory. Thus
159 * a client can call this function on the buffer in the
160 * GNUNET_FS_ProgressCallback. Also, directories can optionally
161 * include the contents of (small) files embedded in the directory
162 * itself; for those files, the processor may be given the
163 * contents of the file directly by this function.
166 * Note that this function maybe called on parts of directories. Thus
167 * parser errors should not be reported _at all_ (with GNUNET_break).
168 * Still, if some entries can be recovered despite these parsing
169 * errors, the function should try to do this.
171 * @param size number of bytes in data
172 * @param data pointer to the beginning of the directory
173 * @param offset offset of data in the directory
174 * @param dep function to call on each entry
175 * @param dep_cls closure for @a dep
176 * @return #GNUNET_OK if this could be a block in a directory,
177 * #GNUNET_NO if this could be part of a directory (but not 100% OK)
178 * #GNUNET_SYSERR if @a data does not represent a directory
181 GNUNET_FS_directory_list_contents (size_t size,
184 GNUNET_FS_DirectoryEntryProcessor dep,
187 struct GetFullDataClosure full_data;
188 const char *cdata = data;
194 struct GNUNET_FS_Uri *uri;
195 struct GNUNET_CONTAINER_MetaData *md;
199 ((size < 8 + sizeof(uint32_t)) ||
201 GNUNET_FS_DIRECTORY_MAGIC,
203 return GNUNET_SYSERR;
207 GNUNET_memcpy (&mdSize,
210 mdSize = ntohl (mdSize);
211 if (mdSize > size - 8 - sizeof(uint32_t))
214 GNUNET_log (GNUNET_ERROR_TYPE_WARNING,
215 _ ("MAGIC mismatch. This is not a GNUnet directory.\n"));
216 return GNUNET_SYSERR;
218 md = GNUNET_CONTAINER_meta_data_deserialize (&cdata[8 + sizeof(uint32_t)],
223 return GNUNET_SYSERR; /* malformed ! */
231 GNUNET_CONTAINER_meta_data_destroy (md);
232 pos = 8 + sizeof(uint32_t) + mdSize;
236 /* find end of URI */
237 if (cdata[pos] == '\0')
239 /* URI is never empty, must be end of block,
240 * skip to next alignment */
241 align = ((pos / DBLOCK_SIZE) + 1) * DBLOCK_SIZE;
244 /* if we were already aligned, still skip a block! */
245 align += DBLOCK_SIZE;
250 /* malformed - or partial download... */
255 while ((epos < size) && (cdata[epos] != '\0'))
258 return GNUNET_NO; /* malformed - or partial download */
260 uri = GNUNET_FS_uri_parse (&cdata[pos], &emsg);
265 pos--; /* go back to '\0' to force going to next alignment */
268 if (GNUNET_FS_uri_test_ksk (uri))
270 GNUNET_FS_uri_destroy (uri);
272 return GNUNET_NO; /* illegal in directory! */
275 GNUNET_memcpy (&mdSize,
278 mdSize = ntohl (mdSize);
279 pos += sizeof(uint32_t);
280 if (pos + mdSize > size)
282 GNUNET_FS_uri_destroy (uri);
283 return GNUNET_NO; /* malformed - or partial download */
286 md = GNUNET_CONTAINER_meta_data_deserialize (&cdata[pos],
290 GNUNET_FS_uri_destroy (uri);
292 return GNUNET_NO; /* malformed ! */
296 GNUNET_CONTAINER_meta_data_get_by_type (md,
297 EXTRACTOR_METATYPE_GNUNET_ORIGINAL_FILENAME);
299 full_data.data = NULL;
300 GNUNET_CONTAINER_meta_data_iterate (md,
312 GNUNET_free_non_null (full_data.data);
313 GNUNET_free_non_null (filename);
314 GNUNET_CONTAINER_meta_data_destroy (md);
315 GNUNET_FS_uri_destroy (uri);
321 * Entries in the directory (builder).
326 * This is a linked list.
328 struct BuilderEntry *next;
331 * Length of this entry.
337 * Internal state of a directory builder.
339 struct GNUNET_FS_DirectoryBuilder
342 * Meta-data for the directory itself.
344 struct GNUNET_CONTAINER_MetaData *meta;
347 * Head of linked list of entries.
349 struct BuilderEntry *head;
352 * Number of entires in the directory.
359 * Create a directory builder.
361 * @param mdir metadata for the directory
363 struct GNUNET_FS_DirectoryBuilder *
364 GNUNET_FS_directory_builder_create (const struct GNUNET_CONTAINER_MetaData
367 struct GNUNET_FS_DirectoryBuilder *ret;
369 ret = GNUNET_new (struct GNUNET_FS_DirectoryBuilder);
371 ret->meta = GNUNET_CONTAINER_meta_data_duplicate (mdir);
373 ret->meta = GNUNET_CONTAINER_meta_data_create ();
374 GNUNET_FS_meta_data_make_directory (ret->meta);
380 * Add an entry to a directory.
382 * @param bld directory to extend
383 * @param uri uri of the entry (must not be a KSK)
384 * @param md metadata of the entry
385 * @param data raw data of the entry, can be NULL, otherwise
386 * data must point to exactly the number of bytes specified
387 * by the uri which must be of type LOC or CHK
390 GNUNET_FS_directory_builder_add (struct GNUNET_FS_DirectoryBuilder *bld,
391 const struct GNUNET_FS_Uri *uri,
392 const struct GNUNET_CONTAINER_MetaData *md,
395 struct GNUNET_FS_Uri *curi;
396 struct BuilderEntry *e;
406 struct GNUNET_CONTAINER_MetaData *meta;
407 const struct GNUNET_CONTAINER_MetaData *meta_use;
409 GNUNET_assert (! GNUNET_FS_uri_test_ksk (uri));
412 GNUNET_assert (! GNUNET_FS_uri_test_sks (uri));
413 if (GNUNET_FS_uri_test_chk (uri))
415 fsize = GNUNET_FS_uri_chk_get_file_size (uri);
419 curi = GNUNET_FS_uri_loc_get_uri (uri);
420 GNUNET_assert (NULL != curi);
421 fsize = GNUNET_FS_uri_chk_get_file_size (curi);
422 GNUNET_FS_uri_destroy (curi);
427 fsize = 0; /* not given */
429 if (fsize > MAX_INLINE_SIZE)
430 fsize = 0; /* too large */
431 uris = GNUNET_FS_uri_to_string (uri);
432 slen = strlen (uris) + 1;
433 mds = GNUNET_CONTAINER_meta_data_get_serialized_size (md);
438 meta = GNUNET_CONTAINER_meta_data_duplicate (md);
439 GNUNET_CONTAINER_meta_data_insert (meta, "<gnunet>",
440 EXTRACTOR_METATYPE_GNUNET_FULL_DATA,
441 EXTRACTOR_METAFORMAT_BINARY, NULL, data,
443 mdxs = GNUNET_CONTAINER_meta_data_get_serialized_size (meta);
444 if ((slen + sizeof(uint32_t) + mdxs - 1) / DBLOCK_SIZE ==
445 (slen + sizeof(uint32_t) + mds - 1) / DBLOCK_SIZE)
447 /* adding full data would not cause us to cross
448 * additional blocks, so add it! */
454 if (mds > GNUNET_MAX_MALLOC_CHECKED / 2)
455 mds = GNUNET_MAX_MALLOC_CHECKED / 2;
456 e = GNUNET_malloc (sizeof(struct BuilderEntry) + slen + mds
458 ser = (char *) &e[1];
459 GNUNET_memcpy (ser, uris, slen);
461 sptr = &ser[slen + sizeof(uint32_t)];
463 GNUNET_CONTAINER_meta_data_serialize (meta_use, &sptr, mds,
464 GNUNET_CONTAINER_META_DATA_SERIALIZE_PART);
466 GNUNET_CONTAINER_meta_data_destroy (meta);
472 GNUNET_memcpy (&ser[slen], &big, sizeof(uint32_t));
473 e->len = slen + sizeof(uint32_t) + mds;
481 * Given the start and end position of a block of
482 * data, return the end position of that data
483 * after alignment to the DBLOCK_SIZE.
486 do_align (size_t start_position, size_t end_position)
490 align = (end_position / DBLOCK_SIZE) * DBLOCK_SIZE;
491 if ((start_position < align) && (end_position > align))
492 return align + end_position - start_position;
498 * Compute a permuation of the blocks to
499 * minimize the cost of alignment. Greedy packer.
501 * @param start starting position for the first block
502 * @param count size of the two arrays
503 * @param sizes the sizes of the individual blocks
504 * @param perm the permutation of the blocks (updated)
507 block_align (size_t start, unsigned int count, const size_t *sizes,
521 for (i = 0; i < count; i++)
524 badness = 0x7FFFFFFF;
526 for (j = i; j < count; j++)
529 cend = cpos + sizes[cval];
530 if (cpos % DBLOCK_SIZE == 0)
532 /* prefer placing the largest blocks first */
533 cbad = -(cend % DBLOCK_SIZE);
537 if (cpos / DBLOCK_SIZE == cend / DBLOCK_SIZE)
539 /* Data fits into the same block! Prefer small left-overs! */
540 cbad = DBLOCK_SIZE - cend % DBLOCK_SIZE;
544 /* Would have to waste space to re-align, add big factor, this
545 * case is a real loss (proportional to space wasted)! */
546 cbad = DBLOCK_SIZE * (DBLOCK_SIZE - cpos % DBLOCK_SIZE);
555 GNUNET_assert (best != -1);
557 perm[i] = perm[best];
559 cpos += sizes[perm[i]];
560 cpos = do_align (start, cpos);
566 * Finish building the directory. Frees the
567 * builder context and returns the directory
570 * @param bld directory to finish
571 * @param rsize set to the number of bytes needed
572 * @param rdata set to the encoded directory
573 * @return #GNUNET_OK on success
576 GNUNET_FS_directory_builder_finish (struct GNUNET_FS_DirectoryBuilder *bld,
586 struct BuilderEntry *pos;
587 struct BuilderEntry **bes;
594 size = strlen (GNUNET_DIRECTORY_MAGIC) + sizeof(uint32_t);
595 size += GNUNET_CONTAINER_meta_data_get_serialized_size (bld->meta);
601 sizes = GNUNET_new_array (bld->count,
603 perm = GNUNET_new_array (bld->count,
605 bes = GNUNET_new_array (bld->count,
606 struct BuilderEntry *);
608 for (i = 0; i < bld->count; i++)
615 block_align (size, bld->count, sizes, perm);
616 /* compute final size with alignment */
617 for (i = 0; i < bld->count; i++)
620 size += sizes[perm[i]];
621 size = do_align (psize, size);
625 data = GNUNET_malloc_large (size);
628 GNUNET_log_strerror (GNUNET_ERROR_TYPE_ERROR,
632 GNUNET_free_non_null (sizes);
633 GNUNET_free_non_null (perm);
634 GNUNET_free_non_null (bes);
635 return GNUNET_SYSERR;
639 GNUNET_DIRECTORY_MAGIC,
640 strlen (GNUNET_DIRECTORY_MAGIC));
641 off = strlen (GNUNET_DIRECTORY_MAGIC);
643 sptr = &data[off + sizeof(uint32_t)];
645 GNUNET_CONTAINER_meta_data_serialize (bld->meta,
647 size - off - sizeof(uint32_t),
648 GNUNET_CONTAINER_META_DATA_SERIALIZE_FULL);
649 GNUNET_assert (ret != -1);
651 GNUNET_memcpy (&data[off],
654 off += sizeof(uint32_t) + ret;
655 for (j = 0; j < bld->count; j++)
660 off = do_align (psize, off);
661 GNUNET_memcpy (&data[off - sizes[i]], &(bes[i])[1], sizes[i]);
662 GNUNET_free (bes[i]);
664 GNUNET_free_non_null (sizes);
665 GNUNET_free_non_null (perm);
666 GNUNET_free_non_null (bes);
667 GNUNET_assert (off == size);
668 GNUNET_CONTAINER_meta_data_destroy (bld->meta);
674 /* end of fs_directory.c */