2 This file is part of GNUnet.
3 (C) 2012 Christian Grothoff (and other contributing authors)
5 GNUnet is free software; you can redistribute it and/or modify
6 it under the terms of the GNU General Public License as published
7 by the Free Software Foundation; either version 3, or (at your
8 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 General Public License for more details.
15 You should have received a copy of the GNU General Public License
16 along with GNUnet; see the file COPYING. If not, write to the
17 Free Software Foundation, Inc., 59 Temple Place - Suite 330,
18 Boston, MA 02111-1307, USA.
22 * @file src/fs/gnunet-helper-fs-publish.c
23 * @brief Tool to help extract meta data asynchronously
24 * @author Christian Grothoff
26 * This program will scan a directory for files with meta data
27 * and report the results to stdout.
30 #include "gnunet_fs_service.h"
34 * A node of a directory tree.
40 * This is a doubly-linked list
42 struct ScanTreeNode *next;
45 * This is a doubly-linked list
47 struct ScanTreeNode *prev;
50 * Parent of this node, NULL for top-level entries.
52 struct ScanTreeNode *parent;
55 * This is a doubly-linked tree
56 * NULL for files and empty directories
58 struct ScanTreeNode *children_head;
61 * This is a doubly-linked tree
62 * NULL for files and empty directories
64 struct ScanTreeNode *children_tail;
67 * Name of the file/directory
72 * Size of the file (if it is a file), in bytes.
73 * At the moment it is set to 0 for directories.
78 * GNUNET_YES if this is a directory
86 * List of libextractor plugins to use for extracting.
88 static struct EXTRACTOR_PluginList *plugins;
91 * File descriptor we use for IPC with the parent.
93 static int output_stream;
97 * Add meta data that libextractor finds to our meta data
100 * @param cls closure, our meta data container
101 * @param plugin_name name of the plugin that produced this value;
102 * special values can be used (i.e. '<zlib>' for zlib being
103 * used in the main libextractor library and yielding
105 * @param type libextractor-type describing the meta data
106 * @param format basic format information about data
107 * @param data_mime_type mime-type of data (not of the original file);
108 * can be NULL (if mime-type is not known)
109 * @param data actual meta-data found
110 * @param data_len number of bytes in data
111 * @return always 0 to continue extracting
114 add_to_md (void *cls, const char *plugin_name, enum EXTRACTOR_MetaType type,
115 enum EXTRACTOR_MetaFormat format, const char *data_mime_type,
116 const char *data, size_t data_len)
118 struct GNUNET_CONTAINER_MetaData *md = cls;
120 (void) GNUNET_CONTAINER_meta_data_insert (md, plugin_name, type, format,
121 data_mime_type, data, data_len);
127 * Free memory of the 'tree' structure
129 * @param tree tree to free
132 free_tree (struct ScanTreeNode *tree)
134 struct ScanTreeNode *pos;
136 while (NULL != (pos = tree->children_head))
138 if (NULL != tree->parent)
139 GNUNET_CONTAINER_DLL_remove (tree->parent->children_head,
140 tree->parent->children_tail,
142 GNUNET_free (tree->filename);
148 * Write @a size bytes from @a buf into the #output_stream.
150 * @param buf buffer with data to write
151 * @param size number of bytes to write
152 * @return #GNUNET_OK on success, #GNUNET_SYSERR on error
155 write_all (const void *buf,
158 const char *cbuf = buf;
165 wr = write (output_stream,
170 } while ( (wr > 0) && (total < size) );
172 GNUNET_log (GNUNET_ERROR_TYPE_DEBUG,
173 "Failed to write to stdout: %s\n",
175 return (total == size) ? GNUNET_OK : GNUNET_SYSERR;
180 * Write message to the master process.
182 * @param message_type message type to use
183 * @param data data to append, NULL for none
184 * @param data_length number of bytes in @a data
185 * @return #GNUNET_SYSERR to stop scanning (the pipe was broken somehow)
188 write_message (uint16_t message_type,
192 struct GNUNET_MessageHeader hdr;
196 "Helper sends %u-byte message of type %u\n",
197 (unsigned int) (sizeof (struct GNUNET_MessageHeader) + data_length),
198 (unsigned int) message_type);
200 hdr.type = htons (message_type);
201 hdr.size = htons (sizeof (struct GNUNET_MessageHeader) + data_length);
208 return GNUNET_SYSERR;
214 * Function called to (recursively) add all of the files in the
215 * directory to the tree. Called by the directory scanner to initiate
216 * the scan. Does NOT yet add any metadata.
218 * @param filename file or directory to scan
219 * @param dst where to store the resulting share tree item;
220 * NULL is stored in @a dst upon recoverable errors (#GNUNET_OK is returned)
221 * @return #GNUNET_OK on success, #GNUNET_SYSERR on error
224 preprocess_file (const char *filename,
225 struct ScanTreeNode **dst);
229 * Closure for the 'scan_callback'
231 struct RecursionContext
234 * Parent to add the files to.
236 struct ScanTreeNode *parent;
239 * Flag to set to GNUNET_YES on serious errors.
246 * Function called by the directory iterator to (recursively) add all
247 * of the files in the directory to the tree. Called by the directory
248 * scanner to initiate the scan. Does NOT yet add any metadata.
250 * @param cls the `struct RecursionContext`
251 * @param filename file or directory to scan
252 * @return #GNUNET_OK on success, #GNUNET_SYSERR on error
255 scan_callback (void *cls,
256 const char *filename)
258 struct RecursionContext *rc = cls;
259 struct ScanTreeNode *chld;
262 preprocess_file (filename,
265 rc->stop = GNUNET_YES;
266 return GNUNET_SYSERR;
270 chld->parent = rc->parent;
271 GNUNET_CONTAINER_DLL_insert (rc->parent->children_head,
272 rc->parent->children_tail,
279 * Function called to (recursively) add all of the files in the
280 * directory to the tree. Called by the directory scanner to initiate
281 * the scan. Does NOT yet add any metadata.
283 * @param filename file or directory to scan
284 * @param dst where to store the resulting share tree item;
285 * NULL is stored in @a dst upon recoverable errors (#GNUNET_OK is returned)
286 * @return #GNUNET_OK on success, #GNUNET_SYSERR on error
289 preprocess_file (const char *filename,
290 struct ScanTreeNode **dst)
292 struct ScanTreeNode *item;
296 if ((0 != STAT (filename, &sbuf)) ||
297 ((!S_ISDIR (sbuf.st_mode)) && (GNUNET_OK != GNUNET_DISK_file_size (
298 filename, &fsize, GNUNET_NO, GNUNET_YES))))
300 /* If the file doesn't exist (or is not stat-able for any other reason)
301 skip it (but report it), but do continue. */
303 write_message (GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_SKIP_FILE,
304 filename, strlen (filename) + 1))
305 return GNUNET_SYSERR;
306 /* recoverable error, store 'NULL' in *dst */
311 /* Report the progress */
313 write_message (S_ISDIR (sbuf.st_mode)
314 ? GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_PROGRESS_DIRECTORY
315 : GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_PROGRESS_FILE,
316 filename, strlen (filename) + 1))
317 return GNUNET_SYSERR;
318 item = GNUNET_malloc (sizeof (struct ScanTreeNode));
319 item->filename = GNUNET_strdup (filename);
320 item->is_directory = (S_ISDIR (sbuf.st_mode)) ? GNUNET_YES : GNUNET_NO;
321 item->file_size = fsize;
322 if (GNUNET_YES == item->is_directory)
324 struct RecursionContext rc;
328 GNUNET_DISK_directory_scan (filename,
331 if ( (GNUNET_YES == rc.stop) ||
333 write_message (GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_PROGRESS_DIRECTORY,
337 return GNUNET_SYSERR;
346 * Extract metadata from files.
348 * @param item entry we are processing
349 * @return #GNUNET_OK on success, #GNUNET_SYSERR on fatal errors
352 extract_files (struct ScanTreeNode *item)
354 struct GNUNET_CONTAINER_MetaData *meta;
358 if (GNUNET_YES == item->is_directory)
360 /* for directories, we simply only descent, no extraction, no
361 progress reporting */
362 struct ScanTreeNode *pos;
364 for (pos = item->children_head; NULL != pos; pos = pos->next)
367 return GNUNET_SYSERR;
371 /* this is the expensive operation, *afterwards* we'll check for aborts */
372 meta = GNUNET_CONTAINER_meta_data_create ();
373 EXTRACTOR_extract (plugins, item->filename, NULL, 0, &add_to_md, meta);
374 slen = strlen (item->filename) + 1;
375 size = GNUNET_CONTAINER_meta_data_get_serialized_size (meta);
379 GNUNET_CONTAINER_meta_data_destroy (meta);
381 write_message (GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_META_DATA,
382 item->filename, slen))
383 return GNUNET_SYSERR;
386 else if (size > (UINT16_MAX - sizeof (struct GNUNET_MessageHeader) - slen))
388 /* We can't transfer more than 64k bytes in one message. */
389 size = UINT16_MAX - sizeof (struct GNUNET_MessageHeader) - slen;
392 char buf[size + slen];
393 char *dst = &buf[slen];
395 memcpy (buf, item->filename, slen);
396 size = GNUNET_CONTAINER_meta_data_serialize (meta,
398 GNUNET_CONTAINER_META_DATA_SERIALIZE_PART);
404 GNUNET_CONTAINER_meta_data_destroy (meta);
406 write_message (GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_META_DATA,
409 return GNUNET_SYSERR;
417 * Install a signal handler to ignore SIGPIPE.
422 struct sigaction oldsig;
423 struct sigaction sig;
425 memset (&sig, 0, sizeof (struct sigaction));
426 sig.sa_handler = SIG_IGN;
427 sigemptyset (&sig.sa_mask);
429 sig.sa_flags = SA_INTERRUPT; /* SunOS */
431 sig.sa_flags = SA_RESTART;
433 if (0 != sigaction (SIGPIPE, &sig, &oldsig))
435 "Failed to install SIGPIPE handler: %s\n", strerror (errno));
440 * Turn the given file descriptor in to '/dev/null'.
442 * @param fd fd to bind to /dev/null
443 * @param flags flags to use (O_RDONLY or O_WRONLY)
446 make_dev_zero (int fd,
451 GNUNET_assert (0 == close (fd));
452 z = open ("/dev/null", flags);
453 GNUNET_assert (-1 != z);
457 GNUNET_assert (0 == close (z));
464 * Main function of the helper process to extract meta data.
466 * @param argc should be 3
467 * @param argv [0] our binary name
468 * [1] name of the file or directory to process
469 * [2] "-" to disable extraction, NULL for defaults,
470 * otherwise custom plugins to load from LE
471 * @return 0 on success
477 const char *filename_expanded;
479 struct ScanTreeNode *root;
482 /* We're using stdout to communicate binary data back to the parent; use
485 _setmode (1, _O_BINARY);
486 /* Get utf-8-encoded arguments */
487 if (GNUNET_OK != GNUNET_STRINGS_get_utf8_args (argc, argv, &argc, &argv))
489 output_stream = 1; /* stdout */
492 /* move stdout to some other FD for IPC, bind
493 stdout/stderr to /dev/null */
494 output_stream = dup (1);
495 make_dev_zero (1, O_WRONLY);
496 make_dev_zero (2, O_WRONLY);
499 /* parse command line */
500 if ( (3 != argc) && (2 != argc) )
504 "gnunet-helper-fs-publish needs exactly one or two arguments\n");
506 GNUNET_free ((void*) argv);
510 filename_expanded = argv[1];
513 (0 != strcmp (ex, "-")) )
515 plugins = EXTRACTOR_plugin_add_defaults (EXTRACTOR_OPTION_DEFAULT_POLICY);
517 plugins = EXTRACTOR_plugin_add_config (plugins, ex,
518 EXTRACTOR_OPTION_DEFAULT_POLICY);
521 /* scan tree to find out how much work there is to be done */
522 if (GNUNET_OK != preprocess_file (filename_expanded,
525 (void) write_message (GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_ERROR, NULL, 0);
526 EXTRACTOR_plugin_remove_all (plugins);
528 GNUNET_free ((void*) argv);
532 /* signal that we're done counting files, so that a percentage of
533 progress can now be calculated */
535 write_message (GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_COUNTING_DONE, NULL, 0))
537 EXTRACTOR_plugin_remove_all (plugins);
539 GNUNET_free ((void*) argv);
546 extract_files (root))
548 (void) write_message (GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_ERROR, NULL, 0);
550 EXTRACTOR_plugin_remove_all (plugins);
552 GNUNET_free ((void*) argv);
558 /* enable "clean" shutdown by telling parent that we are done */
559 (void) write_message (GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_FINISHED, NULL, 0);
560 EXTRACTOR_plugin_remove_all (plugins);
562 GNUNET_free ((void*) argv);
567 /* end of gnunet-helper-fs-publish.c */