GNUnet  0.10.x
Data Structures | Functions
fs_dirmetascan.c File Reference

code to asynchronously build a 'struct GNUNET_FS_ShareTreeItem' from an on-disk directory for publishing; use the 'gnunet-helper-fs-publish'. More...

#include "platform.h"
#include "gnunet_fs_service.h"
#include "gnunet_scheduler_lib.h"
#include <pthread.h>
Include dependency graph for fs_dirmetascan.c:

Go to the source code of this file.

Data Structures

struct  GNUNET_FS_DirScanner
 An opaque structure a pointer to which is returned to the caller to be used to control the scanner. More...
 

Functions

void GNUNET_FS_directory_scan_abort (struct GNUNET_FS_DirScanner *ds)
 Abort the scan. More...
 
struct GNUNET_FS_ShareTreeItemGNUNET_FS_directory_scan_get_result (struct GNUNET_FS_DirScanner *ds)
 Obtain the result of the scan after the scan has signalled completion. More...
 
static struct GNUNET_FS_ShareTreeItemadvance (struct GNUNET_FS_ShareTreeItem *pos)
 Move in the directory from the given position to the next file in DFS traversal. More...
 
static struct GNUNET_FS_ShareTreeItemexpand_tree (struct GNUNET_FS_ShareTreeItem *parent, const char *filename, int is_directory)
 Add another child node to the tree. More...
 
static void finish_scan (void *cls)
 Task run last to shut everything down. More...
 
static int process_helper_msgs (void *cls, const struct GNUNET_MessageHeader *msg)
 Called every time there is data to read from the scanner. More...
 
static void helper_died_cb (void *cls)
 Function called if our helper process died. More...
 
struct GNUNET_FS_DirScannerGNUNET_FS_directory_scan_start (const char *filename, int disable_extractor, const char *ex, GNUNET_FS_DirScannerProgressCallback cb, void *cb_cls)
 Start a directory scanner thread. More...
 

Detailed Description

code to asynchronously build a 'struct GNUNET_FS_ShareTreeItem' from an on-disk directory for publishing; use the 'gnunet-helper-fs-publish'.

Author
LRN
Christian Grothoff

Definition in file fs_dirmetascan.c.

Function Documentation

◆ advance()

static struct GNUNET_FS_ShareTreeItem* advance ( struct GNUNET_FS_ShareTreeItem pos)
static

Move in the directory from the given position to the next file in DFS traversal.

Parameters
poscurrent position
Returns
next file, NULL for none

Definition at line 145 of file fs_dirmetascan.c.

References GNUNET_FS_ShareTreeItem::children_head, GNUNET_assert, GNUNET_YES, GNUNET_FS_ShareTreeItem::is_directory, GNUNET_FS_ShareTreeItem::next, GNUNET_FS_ShareTreeItem::parent, and GNUNET_FS_DirScanner::pos.

Referenced by process_helper_msgs().

146 {
147  int moved;
148 
149  GNUNET_assert(NULL != pos);
150  moved = 0; /* must not terminate, even on file, otherwise "normal" */
151  while ((pos->is_directory == GNUNET_YES) || (0 == moved))
152  {
153  if ((moved != -1) && (NULL != pos->children_head))
154  {
155  pos = pos->children_head;
156  moved = 1; /* can terminate if file */
157  continue;
158  }
159  if (NULL != pos->next)
160  {
161  pos = pos->next;
162  moved = 1; /* can terminate if file */
163  continue;
164  }
165  if (NULL != pos->parent)
166  {
167  pos = pos->parent;
168  moved = -1; /* force move to 'next' or 'parent' */
169  continue;
170  }
171  /* no more options, end of traversal */
172  return NULL;
173  }
174  return pos;
175 }
#define GNUNET_assert(cond)
Use this for fatal errors that cannot be handled.
struct GNUNET_FS_ShareTreeItem * parent
This is a doubly-linked tree NULL for top-level entries.
struct GNUNET_FS_ShareTreeItem * next
This is a doubly-linked list.
struct GNUNET_FS_ShareTreeItem * children_head
This is a doubly-linked tree NULL for files and empty directories.
int is_directory
GNUNET_YES if this is a directory
#define GNUNET_YES
Definition: gnunet_common.h:77
Here is the caller graph for this function:

◆ expand_tree()

static struct GNUNET_FS_ShareTreeItem* expand_tree ( struct GNUNET_FS_ShareTreeItem parent,
const char *  filename,
int  is_directory 
)
static

Add another child node to the tree.

Parameters
parentparent of the child, NULL for top level
filenamename of the file or directory
is_directoryGNUNET_YES for directories
Returns
new entry that was just created

Definition at line 187 of file fs_dirmetascan.c.

References GNUNET_FS_ShareTreeItem::children_head, GNUNET_FS_ShareTreeItem::children_tail, GNUNET_FS_ShareTreeItem::filename, GNUNET_asprintf(), GNUNET_CONTAINER_DLL_insert, GNUNET_new, GNUNET_strdup, GNUNET_STRINGS_get_short_name(), GNUNET_YES, GNUNET_FS_ShareTreeItem::is_directory, GNUNET_FS_ShareTreeItem::parent, and GNUNET_FS_ShareTreeItem::short_filename.

Referenced by process_helper_msgs().

190 {
191  struct GNUNET_FS_ShareTreeItem *chld;
192  size_t slen;
193 
194  chld = GNUNET_new(struct GNUNET_FS_ShareTreeItem);
195  chld->parent = parent;
198  "%s%s",
200  is_directory == GNUNET_YES ? "/" : "");
201  /* make sure we do not end with '//' */
202  slen = strlen(chld->short_filename);
203  if ((slen >= 2) && (chld->short_filename[slen - 1] == '/') &&
204  (chld->short_filename[slen - 2] == '/'))
205  chld->short_filename[slen - 1] = '\0';
206  chld->is_directory = is_directory;
207  if (NULL != parent)
209  parent->children_tail,
210  chld);
211  return chld;
212 }
#define GNUNET_CONTAINER_DLL_insert(head, tail, element)
Insert an element at the head of a DLL.
struct GNUNET_FS_ShareTreeItem * parent
This is a doubly-linked tree NULL for top-level entries.
struct GNUNET_FS_ShareTreeItem * children_tail
This is a doubly-linked tree NULL for files and empty directories.
#define GNUNET_new(type)
Allocate a struct or union of the given type.
#define GNUNET_strdup(a)
Wrapper around GNUNET_xstrdup_.
int GNUNET_asprintf(char **buf, const char *format,...)
Like asprintf, just portable.
A node of a directory tree (produced by dirscanner)
struct GNUNET_FS_ShareTreeItem * children_head
This is a doubly-linked tree NULL for files and empty directories.
static char * filename
char * short_filename
Base name of the file/directory.
const char * GNUNET_STRINGS_get_short_name(const char *filename)
"man basename" Returns a pointer to a part of filename (allocates nothing)!
Definition: strings.c:776
int is_directory
GNUNET_YES if this is a directory
char * filename
Name of the file/directory.
#define GNUNET_YES
Definition: gnunet_common.h:77
Here is the call graph for this function:
Here is the caller graph for this function:

◆ finish_scan()

static void finish_scan ( void *  cls)
static

Task run last to shut everything down.

Parameters
clsthe 'struct GNUNET_FS_DirScanner'

Definition at line 221 of file fs_dirmetascan.c.

References ds, GNUNET_FS_DIRSCANNER_FINISHED, GNUNET_HELPER_stop(), GNUNET_NO, GNUNET_SYSERR, GNUNET_FS_DirScanner::helper, GNUNET_FS_DirScanner::progress_callback, GNUNET_FS_DirScanner::progress_callback_cls, and GNUNET_FS_DirScanner::stop_task.

Referenced by process_helper_msgs().

222 {
223  struct GNUNET_FS_DirScanner *ds = cls;
224 
225  ds->stop_task = NULL;
226  if (NULL != ds->helper)
227  {
229  ds->helper = NULL;
230  }
232  NULL,
235 }
struct GNUNET_HELPER_Handle * helper
Helper process.
void * progress_callback_cls
A closure for progress_callback.
void GNUNET_HELPER_stop(struct GNUNET_HELPER_Handle *h, int soft_kill)
Kills the helper, closes the pipe, frees the handle and calls wait() on the helper process...
Definition: helper.c:563
#define GNUNET_NO
Definition: gnunet_common.h:78
An opaque structure a pointer to which is returned to the caller to be used to control the scanner...
struct GNUNET_SCHEDULER_Task * stop_task
Task scheduled when we are done.
GNUNET_FS_DirScannerProgressCallback progress_callback
The function that will be called every time there&#39;s a progress message.
#define GNUNET_SYSERR
Definition: gnunet_common.h:76
Last call to the progress function: we have finished scanning the directory.
static struct GNUNET_FS_DirScanner * ds
Handle to the directory scanner (for recursive insertions).
Here is the call graph for this function:
Here is the caller graph for this function:

◆ process_helper_msgs()

static int process_helper_msgs ( void *  cls,
const struct GNUNET_MessageHeader msg 
)
static

Called every time there is data to read from the scanner.

Calls the scanner progress handler.

Parameters
clsthe closure (directory scanner object)
msgmessage from the helper process
Returns
GNUNET_OK on success, GNUNET_NO to stop further processing (no error) GNUNET_SYSERR to stop further processing with error

Definition at line 249 of file fs_dirmetascan.c.

References advance(), ds, end, expand_tree(), EXTRACTOR_METAFORMAT_UTF8, EXTRACTOR_METATYPE_FILENAME, EXTRACTOR_METATYPE_GNUNET_ORIGINAL_FILENAME, filename, GNUNET_FS_ShareTreeItem::filename, finish_scan(), GNUNET_assert, GNUNET_break, GNUNET_CONTAINER_meta_data_delete(), GNUNET_CONTAINER_meta_data_deserialize(), GNUNET_CONTAINER_meta_data_insert(), GNUNET_FS_DIRSCANNER_ALL_COUNTED, GNUNET_FS_DIRSCANNER_EXTRACT_FINISHED, GNUNET_FS_DIRSCANNER_FILE_IGNORED, GNUNET_FS_DIRSCANNER_FILE_START, GNUNET_FS_DIRSCANNER_INTERNAL_ERROR, GNUNET_FS_uri_ksk_create_from_meta_data(), GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_COUNTING_DONE, GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_ERROR, GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_FINISHED, GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_META_DATA, GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_PROGRESS_DIRECTORY, GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_PROGRESS_FILE, GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_SKIP_FILE, GNUNET_NO, GNUNET_OK, GNUNET_SCHEDULER_add_now(), GNUNET_SYSERR, GNUNET_YES, GNUNET_FS_ShareTreeItem::is_directory, GNUNET_FS_ShareTreeItem::ksk_uri, GNUNET_FS_ShareTreeItem::meta, GNUNET_FS_ShareTreeItem::parent, GNUNET_FS_DirScanner::pos, GNUNET_FS_DirScanner::progress_callback, GNUNET_FS_DirScanner::progress_callback_cls, GNUNET_FS_ShareTreeItem::short_filename, GNUNET_MessageHeader::size, GNUNET_FS_DirScanner::stop_task, GNUNET_FS_DirScanner::toplevel, and GNUNET_MessageHeader::type.

Referenced by GNUNET_FS_directory_scan_start().

250 {
251  struct GNUNET_FS_DirScanner *ds = cls;
252  const char *filename;
253  size_t left;
254 
255 #if 0
256  fprintf(stderr,
257  "DMS parses %u-byte message of type %u\n",
258  (unsigned int)ntohs(msg->size),
259  (unsigned int)ntohs(msg->type));
260 #endif
261  left = ntohs(msg->size) - sizeof(struct GNUNET_MessageHeader);
262  filename = (const char *)&msg[1];
263  switch (ntohs(msg->type))
264  {
266  if (filename[left - 1] != '\0')
267  {
268  GNUNET_break(0);
269  break;
270  }
272  filename,
273  GNUNET_NO,
275  if (NULL == ds->toplevel)
276  {
277  ds->toplevel = expand_tree(ds->pos, filename, GNUNET_NO);
278  }
279  else
280  {
281  GNUNET_assert(NULL != ds->pos);
282  (void)expand_tree(ds->pos, filename, GNUNET_NO);
283  }
284  return GNUNET_OK;
285 
287  if (filename[left - 1] != '\0')
288  {
289  GNUNET_break(0);
290  break;
291  }
292  if (0 == strcmp("..", filename))
293  {
294  if (NULL == ds->pos)
295  {
296  GNUNET_break(0);
297  break;
298  }
299  ds->pos = ds->pos->parent;
300  return GNUNET_OK;
301  }
303  filename,
304  GNUNET_YES,
306  ds->pos = expand_tree(ds->pos, filename, GNUNET_YES);
307  if (NULL == ds->toplevel)
308  ds->toplevel = ds->pos;
309  return GNUNET_OK;
310 
312  break;
313 
315  if ('\0' != filename[left - 1])
316  break;
318  filename,
321  return GNUNET_OK;
322 
324  if (0 != left)
325  {
326  GNUNET_break(0);
327  break;
328  }
329  if (NULL == ds->toplevel)
330  break;
332  NULL,
335  ds->pos = ds->toplevel;
336  if (GNUNET_YES == ds->pos->is_directory)
337  ds->pos = advance(ds->pos);
338  return GNUNET_OK;
339 
341  size_t nlen;
342  const char *end;
343 
344  if (NULL == ds->pos)
345  {
346  GNUNET_break(0);
347  break;
348  }
349  end = memchr(filename, 0, left);
350  if (NULL == end)
351  {
352  GNUNET_break(0);
353  break;
354  }
355  end++;
356  nlen = end - filename;
357  left -= nlen;
358  if (0 != strcmp(filename, ds->pos->filename))
359  {
360  GNUNET_break(0);
361  break;
362  }
364  filename,
365  GNUNET_YES,
367  if (0 < left)
368  {
370  if (NULL == ds->pos->meta)
371  {
372  GNUNET_break(0);
373  break;
374  }
375  /* having full filenames is too dangerous; always make sure we clean them up */
378  NULL,
379  0);
380  /* instead, put in our 'safer' original filename */
382  "<libgnunetfs>",
385  "text/plain",
386  ds->pos->short_filename,
387  strlen(ds->pos->short_filename) + 1);
388  }
390  ds->pos = advance(ds->pos);
391  return GNUNET_OK;
392  }
393 
395  if (NULL != ds->pos)
396  {
397  GNUNET_break(0);
398  break;
399  }
400  if (0 != left)
401  {
402  GNUNET_break(0);
403  break;
404  }
405  if (NULL == ds->toplevel)
406  break;
408  return GNUNET_OK;
409 
410  default:
411  GNUNET_break(0);
412  break;
413  }
415  NULL,
418  return GNUNET_OK;
419 }
static void finish_scan(void *cls)
Task run last to shut everything down.
static int end
Set if we are to shutdown all services (including ARM).
Definition: gnunet-arm.c:34
#define GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_ERROR
Error signal from the helper.
static struct GNUNET_FS_ShareTreeItem * expand_tree(struct GNUNET_FS_ShareTreeItem *parent, const char *filename, int is_directory)
Add another child node to the tree.
void * progress_callback_cls
A closure for progress_callback.
#define GNUNET_assert(cond)
Use this for fatal errors that cannot be handled.
struct GNUNET_FS_ShareTreeItem * parent
This is a doubly-linked tree NULL for top-level entries.
struct GNUNET_FS_Uri * ksk_uri
Keywords for this file or directory (derived from metadata).
struct GNUNET_CONTAINER_MetaData * GNUNET_CONTAINER_meta_data_deserialize(const char *input, size_t size)
Deserialize meta-data.
0-terminated, UTF-8 encoded string.
#define GNUNET_NO
Definition: gnunet_common.h:78
#define GNUNET_OK
Named constants for return values.
Definition: gnunet_common.h:75
An opaque structure a pointer to which is returned to the caller to be used to control the scanner...
uint16_t size
The length of the struct (in bytes, including the length field itself), in big-endian format...
struct GNUNET_SCHEDULER_Task * stop_task
Task scheduled when we are done.
We&#39;ve found all files (in the pre-pass).
#define GNUNET_break(cond)
Use this for internal assertion violations that are not fatal (can be handled) but should not occur...
uint16_t type
The type of the message (GNUNET_MESSAGE_TYPE_XXXX), in big-endian format.
GNUNET_FS_DirScannerProgressCallback progress_callback
The function that will be called every time there&#39;s a progress message.
struct GNUNET_FS_ShareTreeItem * pos
Current position during processing.
struct GNUNET_SCHEDULER_Task * GNUNET_SCHEDULER_add_now(GNUNET_SCHEDULER_TaskCallback task, void *task_cls)
Schedule a new task to be run as soon as possible.
Definition: scheduler.c:1264
static char * filename
There was an internal error.
int GNUNET_CONTAINER_meta_data_insert(struct GNUNET_CONTAINER_MetaData *md, const char *plugin_name, enum EXTRACTOR_MetaType type, enum EXTRACTOR_MetaFormat format, const char *data_mime_type, const char *data, size_t data_size)
Extend metadata.
int GNUNET_CONTAINER_meta_data_delete(struct GNUNET_CONTAINER_MetaData *md, enum EXTRACTOR_MetaType type, const char *data, size_t data_size)
Remove an item.
char * short_filename
Base name of the file/directory.
#define GNUNET_SYSERR
Definition: gnunet_common.h:76
int is_directory
GNUNET_YES if this is a directory
char * filename
Name of the file/directory.
static struct GNUNET_FS_ShareTreeItem * advance(struct GNUNET_FS_ShareTreeItem *pos)
Move in the directory from the given position to the next file in DFS traversal.
#define GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_META_DATA
Extracted meta data from the helper.
struct GNUNET_CONTAINER_MetaData * meta
Metadata for this file or directory.
#define GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_PROGRESS_DIRECTORY
Progress information from the helper: found a directory.
We&#39;ve started processing a file or directory.
struct GNUNET_FS_Uri * GNUNET_FS_uri_ksk_create_from_meta_data(const struct GNUNET_CONTAINER_MetaData *md)
Construct a keyword-URI from meta-data (take all entries in the meta-data and construct one large key...
Definition: fs_uri.c:1778
#define GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_COUNTING_DONE
Signal that helper is done scanning the directory tree.
#define GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_PROGRESS_FILE
Progress information from the helper: found a file.
#define GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_SKIP_FILE
Signal that helper skipped a file.
We&#39;ve finished extracting meta data from a file.
#define GNUNET_MESSAGE_TYPE_FS_PUBLISH_HELPER_FINISHED
Signal that helper is done.
struct GNUNET_FS_ShareTreeItem * toplevel
After the scan is finished, it will contain a pointer to the top-level directory entry in the directo...
Header for all communications.
#define GNUNET_YES
Definition: gnunet_common.h:77
static struct GNUNET_FS_DirScanner * ds
Handle to the directory scanner (for recursive insertions).
We&#39;re having trouble accessing a file (soft-error); it will be ignored.
Here is the call graph for this function:
Here is the caller graph for this function:

◆ helper_died_cb()

static void helper_died_cb ( void *  cls)
static

Function called if our helper process died.

Parameters
clsthe 'struct GNUNET_FS_DirScanner' callback.

Definition at line 428 of file fs_dirmetascan.c.

References ds, GNUNET_FS_DIRSCANNER_INTERNAL_ERROR, GNUNET_SYSERR, GNUNET_FS_DirScanner::helper, GNUNET_FS_DirScanner::progress_callback, GNUNET_FS_DirScanner::progress_callback_cls, and GNUNET_FS_DirScanner::stop_task.

Referenced by GNUNET_FS_directory_scan_start().

429 {
430  struct GNUNET_FS_DirScanner *ds = cls;
431 
432  ds->helper = NULL;
433  if (NULL != ds->stop_task)
434  return; /* normal death, was finished */
436  NULL,
439 }
struct GNUNET_HELPER_Handle * helper
Helper process.
void * progress_callback_cls
A closure for progress_callback.
An opaque structure a pointer to which is returned to the caller to be used to control the scanner...
struct GNUNET_SCHEDULER_Task * stop_task
Task scheduled when we are done.
GNUNET_FS_DirScannerProgressCallback progress_callback
The function that will be called every time there&#39;s a progress message.
There was an internal error.
#define GNUNET_SYSERR
Definition: gnunet_common.h:76
static struct GNUNET_FS_DirScanner * ds
Handle to the directory scanner (for recursive insertions).
Here is the caller graph for this function: