2 * NOTE: this is part of libzzipfseeko (i.e. it is not libzzip).
4 * These routines are fully independent from the traditional zzip
5 * implementation. They assume a readonly seekable stdio handle
6 * representing a complete zip file. The functions show how to
7 * parse the structure, find files and return a decoded bytestream.
9 * These routines are a bit simple and really here for documenting
10 * the way to access a zip file. The complexity of zip access comes
11 * from staggered reading of bytes and reposition of a filepointer in
12 * a big archive with lots of files and long compressed datastreams.
13 * Plus varaints of drop-in stdio replacements, obfuscation routines,
14 * auto fileextensions, drop-in dirent replacements, and so on...
16 * btw, we can _not_ use fgetpos/fsetpos since an fpos_t has no asserted
17 * relation to a linear seek value as specified in zip info headers. In
18 * general it is not a problem if your system has no fseeko/ftello pair
19 * since we can fallback to fseek/ftell which limits the zip disk size
20 * to 2MiBs but the zip-storable seek values are 32bit limited anyway.
23 * Guido Draheim <guidod@gmx.de>
25 * Copyright (c) 2003,2004 Guido Draheim
26 * All rights reserved,
27 * use under the restrictions of the
28 * Lesser GNU General Public License
29 * or alternatively the restrictions
30 * of the Mozilla Public License 1.1
33 #define _LARGEFILE_SOURCE 1
34 #define _ZZIP_ENTRY_STRUCT 1
36 #include <zzip/types.h>
42 #if defined ZZIP_HAVE_STRING_H
44 #elif defined ZZIP_HAVE_STRINGS_H
49 #include <zzip/format.h>
50 #include <zzip/fseeko.h>
51 #include <zzip/fetch.h>
52 #include <zzip/__mmap.h>
53 #include <zzip/__fnmatch.h>
55 #if __STDC_VERSION__+0 > 199900L
63 #ifndef ZZIP_HAVE_FSEEKO
68 /* note that the struct zzip_entry inherits the zzip_disk_entry values
69 * and usually carries a copy of its values (in disk format!). To make the
70 * following code more readable, we use a shorthand notation for the
71 * upcast needed in C (not needed in C++) as "disk_(entry)".
73 #ifdef __zzip_entry_extends_zzip_disk_entry
74 #define disk_(_entry_) _entry_
76 #define disk_(_entry_) (& (_entry_)->head)
79 /* we try to round all seeks to the pagesize - since we do not use
80 * the sys/mmap interface we have to guess a good value here: */
83 /* ====================================================================== */
84 /* helper functions */
86 /** => zzip_entry_data_offset
87 * This functions read the correspoding struct zzip_file_header from
88 * the zip disk of the given "entry". The returned off_t points to the
89 * end of the file_header where the current fseek pointer has stopped.
90 * This is used to immediatly parse out any filename/extras block following
91 * the file_header. The return value is null on error.
94 zzip_entry_fread_file_header (ZZIP_ENTRY* entry,
95 struct zzip_file_header* file_header)
97 if (! entry || ! file_header) return 0;
98 ___ zzip_off_t offset = zzip_disk_entry_fileoffset (disk_(entry));
99 if (0 > offset || offset >= entry->disksize) return 0;
101 fseeko (entry->diskfile, offset, SEEK_SET);
102 return (fread (file_header, sizeof(*file_header), 1, entry->diskfile)
103 ? offset+sizeof(*file_header) : 0 ); ____;
106 /** helper functions for (fseeko) zip access api
108 * This functions returns the seekval offset of the data portion of the
109 * file referenced by the given zzip_entry. It requires an intermediate
110 * check of the file_header structure (i.e. it reads it from disk). After
111 * this call, the contained diskfile readposition is already set to the
112 * data_offset returned here. On error -1 is returned.
115 zzip_entry_data_offset(ZZIP_ENTRY* entry)
117 struct zzip_file_header file_header;
118 if (! entry) return -1;
119 ___ zzip_off_t offset =
120 zzip_entry_fread_file_header (entry, & file_header);
121 if (! offset) return -1;
122 offset += zzip_file_header_sizeof_tails (& file_header);
123 fseeko (entry->diskfile, offset, SEEK_SET);
127 /** => zzip_entry_data_offset
128 * This function is a big helper despite its little name: in a zip file the
129 * encoded filenames are usually NOT zero-terminated but for common usage
130 * with libc we need it that way. Secondly, the filename SHOULD be present
131 * in the zip central directory but if not then we fallback to the filename
132 * given in the file_header of each compressed data portion.
135 zzip_entry_strdup_name(ZZIP_ENTRY* entry)
137 if (! entry) return 0;
140 if ((len = zzip_disk_entry_namlen (disk_(entry)))) {
141 char* name = malloc (len+1);
142 if (! name) return 0;
143 memcpy (name, entry->tail, len);
147 ___ auto struct zzip_file_header header;
148 if (zzip_entry_fread_file_header (entry, &header)
149 && ( len = zzip_file_header_namlen(&header) )) {
150 char* name = malloc (len+1);
151 if (! name) return 0;
152 fread (name, 1, len, entry->diskfile);
161 prescan_entry(ZZIP_ENTRY* entry)
164 ___ zzip_off_t tailsize = zzip_disk_entry_sizeof_tails (disk_(entry));
165 if (tailsize+1 > entry->tailalloc) {
166 char* newtail = realloc (entry->tail, tailsize+1);
167 if (! newtail) return ENOMEM;
168 entry->tail = newtail;
169 entry->tailalloc = tailsize+1;
171 fread (entry->tail, 1, tailsize, entry->diskfile);
172 /* name + comment + extras */
177 prescan_clear(ZZIP_ENTRY* entry)
180 if (entry->tail) free (entry->tail);
181 entry->tail = 0; entry->tailalloc = 0;
184 /* ====================================================================== */
186 /** => zzip_entry_findfile
188 * This function is the first call of all the zip access functions here.
189 * It contains the code to find the first entry of the zip central directory.
190 * Here we require the stdio handle to represent a real zip file where the
191 * disk_trailer is _last_ in the file area, so that its position would be at
192 * a fixed offset from the end of the file area if not for the comment field
193 * allowed to be of variable length (which needs us to do a little search
194 * for the disk_tailer). However, in this simple implementation we disregard
195 * any disk_trailer info telling about multidisk archives, so we just return
196 * a pointer to the first entry in the zip central directory of that file.
198 * For an actual means, we are going to search backwards from the end
199 * of the mmaped block looking for the PK-magic signature of a
200 * disk_trailer. If we see one then we check the rootseek value to
201 * find the first disk_entry of the root central directory. If we find
202 * the correct PK-magic signature of a disk_entry over there then we
203 * assume we are done and we are going to return a pointer to that label.
205 * The return value is a pointer to the first zzip_disk_entry being checked
206 * to be within the bounds of the file area specified by the arguments. If
207 * no disk_trailer was found then null is returned, and likewise we only
208 * accept a disk_trailer with a seekvalue that points to a disk_entry and
209 * both parts have valid PK-magic parts. Beyond some sanity check we try to
210 * catch a common brokeness with zip archives that still allows us to find
211 * the start of the zip central directory.
213 ZZIP_ENTRY* _zzip_new
214 zzip_entry_findfirst(FILE* disk)
216 if (! disk) return 0;
217 fseeko (disk, 0, SEEK_END);
218 ___ zzip_off_t disksize = ftello (disk);
219 if (disksize < (zzip_off_t) sizeof(struct zzip_disk_trailer)) return 0;
220 /* we read out chunks of 8 KiB in the hope to match disk granularity */
221 ___ zzip_off_t pagesize = PAGESIZE; /* getpagesize() */
222 ___ ZZIP_ENTRY* entry = malloc (sizeof(*entry)); if (! entry) return 0;
223 ___ char* buffer = malloc (pagesize); if (! buffer) goto nomem;
225 assert (pagesize/2 > (zzip_off_t) sizeof (struct zzip_disk_trailer));
226 /* at each step, we will fread a pagesize block which overlaps with the
227 * previous read by means of pagesize/2 step at the end of the while(1) */
228 ___ zzip_off_t mapoffs = disksize &~ (pagesize-1);
229 ___ zzip_off_t mapsize = disksize - mapoffs;
230 if (mapoffs && mapsize < pagesize/2) {
231 mapoffs -= pagesize/2; mapsize += pagesize/2; }
233 fseeko (disk, mapoffs, SEEK_SET);
234 fread (buffer, 1, mapsize, disk);
235 ___ char* p = buffer + mapsize - sizeof(struct zzip_disk_trailer);
236 for (; p >= buffer ; p--)
238 zzip_off_t root; /* (struct zzip_disk_entry*) */
239 if (zzip_disk_trailer_check_magic(p)) {
240 root = zzip_disk_trailer_rootseek (
241 (struct zzip_disk_trailer*)p);
242 if (root > disksize - (long)sizeof(struct zzip_disk_trailer)) {
243 /* first disk_entry is after the disk_trailer? can't be! */
244 zzip_off_t rootsize = zzip_disk_trailer_rootsize (
245 (struct zzip_disk_trailer*)p);
246 if (rootsize > mapoffs) continue;
247 /* a common brokeness that can be fixed: we just assume the
248 * central directory was written directly before : */
249 root = mapoffs - rootsize;
251 } else if (zzip_disk64_trailer_check_magic(p)) {
252 if (sizeof(zzip_off_t) < 8) return 0;
253 root = zzip_disk64_trailer_rootseek (
254 (struct zzip_disk64_trailer*)p);
257 assert (0 <= root && root < mapsize);
258 fseeko (disk, root, SEEK_SET);
259 fread (disk_(entry), 1, sizeof(*disk_(entry)), disk);
260 if (zzip_disk_entry_check_magic(entry)) {
262 entry->headseek = root;
263 entry->diskfile = disk;
264 entry->disksize = disksize;
265 if (prescan_entry(entry)) goto nomem;
269 if (! mapoffs) break; assert (mapsize >= pagesize/2);
270 mapoffs -= pagesize/2; /* mapsize += pagesize/2; */
271 mapsize = pagesize; /* if (mapsize > pagesize) ... */
272 if (disksize - mapoffs > 64*1024) break;
276 free (entry); ____;____;____;____;____;____;
280 /** => zzip_entry_findfile
282 * This function takes an existing "entry" in the central root directory
283 * (e.g. from zzip_entry_findfirst) and moves it to point to the next entry.
284 * On error it returns 0, otherwise the old entry. If no further match is
285 * found then null is returned and the entry already free()d. If you want
286 * to stop searching for matches before that case then please call
287 * => zzip_entry_free on the cursor struct ZZIP_ENTRY.
289 ZZIP_ENTRY* _zzip_new
290 zzip_entry_findnext(ZZIP_ENTRY* _zzip_restrict entry)
292 if (! entry) return entry;
293 if (! zzip_disk_entry_check_magic (entry)) goto err;
294 ___ zzip_off_t seek =
295 entry->headseek + zzip_disk_entry_sizeto_end (disk_(entry));
296 if (seek + (zzip_off_t) sizeof(*disk_(entry)) > entry->disksize) goto err;
298 fseeko (entry->diskfile, seek, SEEK_SET);
299 fread (disk_(entry), 1, sizeof(*disk_(entry)), entry->diskfile);
300 entry->headseek = seek;
301 if (! zzip_disk_entry_check_magic (entry)) goto err;
302 if (prescan_entry(entry)) goto err;
305 zzip_entry_free (entry);
309 /** => zzip_entry_findfile
310 * this function releases the malloc()ed areas needed for zzip_entry, the
311 * pointer is invalid afterwards. This function has #define synonyms of
312 * zzip_entry_findlast(), zzip_entry_findlastfile(), zzip_entry_findlastmatch()
315 zzip_entry_free(ZZIP_ENTRY* entry)
317 if (! entry) return 0;
318 prescan_clear (entry);
323 /** search for files in the (fseeko) zip central directory
325 * This function is given a filename as an additional argument, to find the
326 * disk_entry matching a given filename. The compare-function is usually
327 * strcmp or strcasecmp or perhaps strcoll, if null then strcmp is used.
328 * - use null as argument for "old"-entry when searching the first
329 * matching entry, otherwise the last returned value if you look for other
330 * entries with a special "compare" function (if null then a doubled search
331 * is rather useless with this variant of _findfile). If no further entry is
332 * found then null is returned and any "old"-entry gets already free()d.
334 ZZIP_ENTRY* _zzip_new
335 zzip_entry_findfile(FILE* disk, char* filename,
336 ZZIP_ENTRY* _zzip_restrict entry,
337 zzip_strcmp_fn_t compare)
339 if (! filename || ! disk) return 0;
340 entry = ( ! entry ) ? zzip_entry_findfirst (disk)
341 : zzip_entry_findnext (entry);
342 if (! compare) compare = (zzip_strcmp_fn_t)(strcmp);
344 for (; entry ; entry = zzip_entry_findnext (entry))
345 { /* filenames within zip files are often not null-terminated! */
346 char* realname = zzip_entry_strdup_name (entry);
347 if (! realname) continue;
348 if (! compare (filename, realname)) {
349 free (realname); return entry;
351 free (realname); continue;
357 #ifdef ZZIP_HAVE_FNMATCH_H
358 #define _zzip_fnmatch fnmatch
360 # define _zzip_fnmatch_CASEFOLD FNM_CASEFOLD
362 # define _zzip_fnmatch_CASEFOLD 0
365 # define _zzip_fnmatch_CASEFOLD 0
366 /* if your system does not have fnmatch, we fall back to strcmp: */
367 static int _zzip_fnmatch(char* pattern, char* string, int flags)
369 puts ("<zzip:strcmp>");
370 return strcmp (pattern, string);
374 /** => zzip_entry_findfile
376 * This function uses a compare-function with an additional argument
377 * and it is called just like fnmatch(3) from POSIX.2 AD:1993), i.e.
378 * the argument filespec first and the ziplocal filename second with
379 * the integer-flags put in as third to the indirect call. If the
380 * platform has fnmatch available then null-compare will use that one
381 * and otherwise we fall back to mere strcmp, so if you need fnmatch
382 * searching then please provide an implementation somewhere else.
383 * - use null as argument for "after"-entry when searching the first
384 * matching entry, or the last disk_entry return-value to find the
385 * next entry matching the given filespec. If no further entry is
386 * found then null is returned and any "old"-entry gets already free()d.
388 ZZIP_ENTRY* _zzip_new
389 zzip_entry_findmatch(FILE* disk, char* filespec,
390 ZZIP_ENTRY* _zzip_restrict entry,
391 zzip_fnmatch_fn_t compare, int flags)
393 if (! filespec || ! disk) return 0;
394 entry = ( ! entry ) ? zzip_entry_findfirst (disk)
395 : zzip_entry_findnext (entry);
396 if (! compare) compare = (zzip_fnmatch_fn_t) _zzip_fnmatch;
398 for (; entry ; entry = zzip_entry_findnext (entry))
399 { /* filenames within zip files are often not null-terminated! */
400 char* realname = zzip_entry_strdup_name (entry);
401 if (! realname) continue;
402 if (! compare (filespec, realname, flags)) {
403 free (realname); return entry;
405 free (realname); continue;
411 /* ====================================================================== */
414 * typedef struct zzip_disk_file ZZIP_ENTRY_FILE;
416 struct zzip_entry_file /* : zzip_file_header */
418 struct zzip_file_header header; /* fopen detected header */
419 ZZIP_ENTRY* entry; /* fopen entry */
420 zzip_off_t data; /* for stored blocks */
421 zzip_size_t avail; /* memorized for checks on EOF */
422 zzip_size_t compressed; /* compressed flag and datasize */
423 zzip_size_t dataoff; /* offset from data start */
424 z_stream zlib; /* for inflated blocks */
425 char buffer[PAGESIZE]; /* work buffer for inflate algorithm */
428 /** open a file within a zip disk for reading
430 * This function does take an "entry" argument and copies it (or just takes
431 * it over as owner) to a new ZZIP_ENTRY_FILE handle structure. That
432 * structure contains also a zlib buffer for decoding. This function does
433 * seek to the file_header of the given "entry" and validates it for the
434 * data buffer following it. We do also prefetch some data from the data
435 * buffer thereby trying to match the disk pagesize for faster access later.
436 * The => zzip_entry_fread will then read in chunks of pagesizes which is
437 * the size of the internal readahead buffer. If an error occurs then null
440 ZZIP_ENTRY_FILE* _zzip_new
441 zzip_entry_fopen (ZZIP_ENTRY* entry, int takeover)
443 if (! entry) return 0;
445 ZZIP_ENTRY* found = malloc (sizeof(*entry));
446 if (! found) return 0;
447 memcpy (found, entry, sizeof(*entry)); /* prescan_copy */
448 found->tail = malloc (found->tailalloc);
449 if (! found->tail) { free (found); return 0; }
450 memcpy (found->tail, entry->tail, entry->tailalloc);
453 ___ ZZIP_ENTRY_FILE* file = malloc(sizeof(*file));
454 if (! file) goto fail1;
456 if (! zzip_entry_fread_file_header (entry, &file->header))
458 file->avail = zzip_file_header_usize (&file->header);
459 file->data = zzip_entry_data_offset (entry);
462 if (! file->avail || zzip_file_header_data_stored (&file->header))
463 { file->compressed = 0; return file; }
465 file->compressed = zzip_file_header_csize (&file->header);
466 file->zlib.opaque = 0;
467 file->zlib.zalloc = Z_NULL;
468 file->zlib.zfree = Z_NULL;
470 ___ zzip_off_t seek = file->data;
471 seek += sizeof(file->buffer); seek -= seek & (sizeof(file->buffer)-1);
472 assert (file->data < seek); /* pre-read to next PAGESIZE boundary... */
473 fseeko (file->entry->diskfile, file->data + file->dataoff, SEEK_SET);
474 file->zlib.next_in = file->buffer;
475 file->zlib.avail_in = fread (file->buffer, 1, seek - file->data,
476 file->entry->diskfile);
477 file->dataoff += file->zlib.avail_in; ____;
479 if (! zzip_file_header_data_deflated (&file->header)
480 || inflateInit2 (& file->zlib, -MAX_WBITS) != Z_OK) goto fail2;
486 zzip_entry_free (entry);
490 /** => zzip_entry_fopen
492 * This function opens a file found by name, so it does a search into
493 * the zip central directory with => zzip_entry_findfile and whatever
494 * is found first is given to => zzip_entry_fopen
496 ZZIP_ENTRY_FILE* _zzip_new
497 zzip_entry_ffile (FILE* disk, char* filename)
499 ZZIP_ENTRY* entry = zzip_entry_findfile (disk, filename, 0, 0);
500 if (! entry) return 0;
501 return zzip_entry_fopen (entry, 1);
505 /** => zzip_entry_fopen
507 * This function reads more bytes into the output buffer specified as
508 * arguments. The return value is null on eof or error, the stdio-like
509 * interface can not distinguish between these so you need to check
510 * with => zzip_entry_feof for the difference.
513 zzip_entry_fread (void* ptr, zzip_size_t sized, zzip_size_t nmemb,
514 ZZIP_ENTRY_FILE* file)
516 if (! file) return 0;
517 ___ zzip_size_t size = sized*nmemb;
518 if (! file->compressed) {
519 if (size > file->avail) size = file->avail;
520 fread (ptr, 1, size, file->entry->diskfile);
521 file->dataoff += size;
526 file->zlib.avail_out = size;
527 file->zlib.next_out = ptr;
528 ___ zzip_size_t total_old = file->zlib.total_out;
530 if (! file->zlib.avail_in) {
531 size = file->compressed - file->dataoff;
532 if (size > sizeof(file->buffer)) size = sizeof(file->buffer);
533 /* fseek (file->data + file->dataoff, file->entry->diskfile); */
534 file->zlib.avail_in = fread (file->buffer, 1, size,
535 file->entry->diskfile);
536 file->zlib.next_in = file->buffer;
537 file->dataoff += file->zlib.avail_in;
539 if (! file->zlib.avail_in) return 0;
541 ___ int err = inflate (& file->zlib, Z_NO_FLUSH);
542 if (err == Z_STREAM_END)
544 else if (err == Z_OK)
545 file->avail -= file->zlib.total_out - total_old;
549 if (file->zlib.avail_out && ! file->zlib.avail_in) continue;
550 return file->zlib.total_out - total_old;
554 /** => zzip_entry_fopen
555 * This function releases any zlib decoder info needed for decompression
556 * and dumps the ZZIP_ENTRY_FILE struct then.
559 zzip_entry_fclose (ZZIP_ENTRY_FILE* file)
561 if (! file) return 0;
562 if (file->compressed)
563 inflateEnd (& file->zlib);
564 zzip_entry_free (file->entry);
569 /** => zzip_entry_fopen
571 * This function allows to distinguish an error from an eof condition.
572 * Actually, if we found an error but we did already reach eof then we
573 * just keep on saying that it was an eof, so the app can just continue.
576 zzip_entry_feof (ZZIP_ENTRY_FILE* file)
578 return ! file || ! file->avail;