1 /*-
2 * Copyright (c) 2006,2008,2010 Joseph Koshy
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS `AS IS' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 */
26
27 #include <assert.h>
28 #include <ctype.h>
29 #include <libelf.h>
30 #include <stdlib.h>
31 #include <string.h>
32
33 #include "_libelf.h"
34 #include "_libelf_ar.h"
35
36 ELFTC_VCSID("$Id: libelf_ar.c 3174 2015-03-27 17:13:41Z emaste $");
37
38 #define LIBELF_NALLOC_SIZE 16
39
40 /*
41 * `ar' archive handling.
42 *
43 * `ar' archives start with signature `ARMAG'. Each archive member is
44 * preceded by a header containing meta-data for the member. This
45 * header is described in <ar.h> (struct ar_hdr). The header always
46 * starts on an even address. File data is padded with "\n"
47 * characters to keep this invariant.
48 *
49 * Special considerations for `ar' archives:
50 *
51 * There are two variants of the `ar' archive format: traditional BSD
52 * and SVR4. These differ in the way long file names are treated, and
53 * in the layout of the archive symbol table.
54 *
55 * The `ar' header only has space for a 16 character file name.
56 *
57 * In the SVR4 format, file names are terminated with a '/', so this
58 * effectively leaves 15 characters for the actual file name. Longer
59 * file names stored in a separate 'string table' and referenced
60 * indirectly from the name field. The string table itself appears as
61 * an archive member with name "// ". An `indirect' file name in an
62 * `ar' header matches the pattern "/[0-9]*". The digits form a
63 * decimal number that corresponds to a byte offset into the string
64 * table where the actual file name of the object starts. Strings in
65 * the string table are padded to start on even addresses.
66 *
67 * In the BSD format, file names can be upto 16 characters. File
68 * names shorter than 16 characters are padded to 16 characters using
69 * (ASCII) space characters. File names with embedded spaces and file
70 * names longer than 16 characters are stored immediately after the
71 * archive header and the name field set to a special indirect name
72 * matching the pattern "#1/[0-9]+". The digits form a decimal number
73 * that corresponds to the actual length of the file name following
74 * the archive header. The content of the archive member immediately
75 * follows the file name, and the size field of the archive member
76 * holds the sum of the sizes of the member and of the appended file
77 * name.
78 *
79 * Archives may also have a symbol table (see ranlib(1)), mapping
80 * program symbols to object files inside the archive.
81 *
82 * In the SVR4 format, a symbol table uses a file name of "/ " in its
83 * archive header. The symbol table is structured as:
84 * - a 4-byte count of entries stored as a binary value, MSB first
85 * - 'n' 4-byte offsets, stored as binary values, MSB first
86 * - 'n' NUL-terminated strings, for ELF symbol names, stored unpadded.
87 *
88 * In the BSD format, the symbol table uses a file name of "__.SYMDEF".
89 * It is structured as two parts:
90 * - The first part is an array of "ranlib" structures preceded by
91 * the size of the array in bytes. Each "ranlib" structure
92 * describes one symbol. Each structure contains an offset into
93 * the string table for the symbol name, and a file offset into the
94 * archive for the member defining the symbol.
95 * - The second part is a string table containing NUL-terminated
96 * strings, preceded by the size of the string table in bytes.
97 *
98 * If the symbol table and string table are is present in an archive
99 * they must be the very first objects and in that order.
100 */
101
102
103 /*
104 * Retrieve an archive header descriptor.
105 */
106
107 Elf_Arhdr *
_libelf_ar_gethdr(Elf * e)108 _libelf_ar_gethdr(Elf *e)
109 {
110 Elf *parent;
111 Elf_Arhdr *eh;
112 char *namelen;
113 size_t n, nlen;
114 struct ar_hdr *arh;
115
116 if ((parent = e->e_parent) == NULL) {
117 LIBELF_SET_ERROR(ARGUMENT, 0);
118 return (NULL);
119 }
120
121 assert((e->e_flags & LIBELF_F_AR_HEADER) == 0);
122
123 arh = (struct ar_hdr *) (uintptr_t) e->e_hdr.e_rawhdr;
124
125 assert((uintptr_t) arh >= (uintptr_t) parent->e_rawfile + SARMAG);
126 assert((uintptr_t) arh <= (uintptr_t) parent->e_rawfile +
127 parent->e_rawsize - sizeof(struct ar_hdr));
128
129 if ((eh = malloc(sizeof(Elf_Arhdr))) == NULL) {
130 LIBELF_SET_ERROR(RESOURCE, 0);
131 return (NULL);
132 }
133
134 e->e_hdr.e_arhdr = eh;
135 e->e_flags |= LIBELF_F_AR_HEADER;
136
137 eh->ar_name = eh->ar_rawname = NULL;
138
139 if ((eh->ar_name = _libelf_ar_get_translated_name(arh, parent)) ==
140 NULL)
141 goto error;
142
143 if (_libelf_ar_get_number(arh->ar_uid, sizeof(arh->ar_uid), 10,
144 &n) == 0)
145 goto error;
146 eh->ar_uid = (uid_t) n;
147
148 if (_libelf_ar_get_number(arh->ar_gid, sizeof(arh->ar_gid), 10,
149 &n) == 0)
150 goto error;
151 eh->ar_gid = (gid_t) n;
152
153 if (_libelf_ar_get_number(arh->ar_mode, sizeof(arh->ar_mode), 8,
154 &n) == 0)
155 goto error;
156 eh->ar_mode = (mode_t) n;
157
158 if (_libelf_ar_get_number(arh->ar_size, sizeof(arh->ar_size), 10,
159 &n) == 0)
160 goto error;
161
162 /*
163 * Get the true size of the member if extended naming is being used.
164 */
165 if (IS_EXTENDED_BSD_NAME(arh->ar_name)) {
166 namelen = arh->ar_name +
167 LIBELF_AR_BSD_EXTENDED_NAME_PREFIX_SIZE;
168 if (_libelf_ar_get_number(namelen, sizeof(arh->ar_name) -
169 LIBELF_AR_BSD_EXTENDED_NAME_PREFIX_SIZE, 10, &nlen) == 0)
170 goto error;
171 n -= nlen;
172 }
173
174 eh->ar_size = n;
175
176 if ((eh->ar_rawname = _libelf_ar_get_raw_name(arh)) == NULL)
177 goto error;
178
179 eh->ar_flags = 0;
180
181 return (eh);
182
183 error:
184 if (eh) {
185 if (eh->ar_name)
186 free(eh->ar_name);
187 if (eh->ar_rawname)
188 free(eh->ar_rawname);
189 free(eh);
190 }
191
192 e->e_flags &= ~LIBELF_F_AR_HEADER;
193 e->e_hdr.e_rawhdr = (unsigned char *) arh;
194
195 return (NULL);
196 }
197
198 Elf *
_libelf_ar_open_member(int fd,Elf_Cmd c,Elf * elf)199 _libelf_ar_open_member(int fd, Elf_Cmd c, Elf *elf)
200 {
201 Elf *e;
202 off_t next;
203 size_t nsz, sz;
204 struct ar_hdr *arh;
205 char *member, *namelen;
206
207 assert(elf->e_kind == ELF_K_AR);
208
209 next = elf->e_u.e_ar.e_next;
210
211 /*
212 * `next' is only set to zero by elf_next() when the last
213 * member of an archive is processed.
214 */
215 if (next == (off_t) 0)
216 return (NULL);
217
218 assert((next & 1) == 0);
219
220 arh = (struct ar_hdr *) (elf->e_rawfile + next);
221
222 /*
223 * Retrieve the size of the member.
224 */
225 if (_libelf_ar_get_number(arh->ar_size, sizeof(arh->ar_size), 10,
226 &sz) == 0) {
227 LIBELF_SET_ERROR(ARCHIVE, 0);
228 return (NULL);
229 }
230
231 /*
232 * Adjust the size field for members in BSD archives using
233 * extended naming.
234 */
235 if (IS_EXTENDED_BSD_NAME(arh->ar_name)) {
236 namelen = arh->ar_name +
237 LIBELF_AR_BSD_EXTENDED_NAME_PREFIX_SIZE;
238 if (_libelf_ar_get_number(namelen, sizeof(arh->ar_name) -
239 LIBELF_AR_BSD_EXTENDED_NAME_PREFIX_SIZE, 10, &nsz) == 0) {
240 LIBELF_SET_ERROR(ARCHIVE, 0);
241 return (NULL);
242 }
243
244 member = (char *) (arh + 1) + nsz;
245 sz -= nsz;
246 } else
247 member = (char *) (arh + 1);
248
249
250 if ((e = elf_memory(member, sz)) == NULL)
251 return (NULL);
252
253 e->e_fd = fd;
254 e->e_cmd = c;
255 e->e_hdr.e_rawhdr = (unsigned char *) arh;
256
257 elf->e_u.e_ar.e_nchildren++;
258 e->e_parent = elf;
259
260 return (e);
261 }
262
263 /*
264 * A BSD-style ar(1) symbol table has the following layout:
265 *
266 * - A count of bytes used by the following array of 'ranlib'
267 * structures, stored as a 'long'.
268 * - An array of 'ranlib' structures. Each array element is
269 * two 'long's in size.
270 * - A count of bytes used for the following symbol table.
271 * - The symbol table itself.
272 */
273
274 /*
275 * A helper macro to read in a 'long' value from the archive.
276 *
277 * We use memcpy() since the source pointer may be misaligned with
278 * respect to the natural alignment for a C 'long'.
279 */
280 #define GET_LONG(P, V)do { \
281 memcpy(&(V), (P), sizeof(long)); \
282 (P) += sizeof(long); \
283 } while (0)
284
285 Elf_Arsym *
_libelf_ar_process_bsd_symtab(Elf * e,size_t * count)286 _libelf_ar_process_bsd_symtab(Elf *e, size_t *count)
287 {
288 Elf_Arsym *symtab, *sym;
289 unsigned int n, nentries;
290 unsigned char *end, *p, *p0, *s, *s0;
291 const size_t entrysize = 2 * sizeof(long);
292 long arraysize, fileoffset, stroffset, strtabsize;
293
294 assert(e != NULL);
295 assert(count != NULL);
296 assert(e->e_u.e_ar.e_symtab == NULL);
297
298 symtab = NULL;
299
300 /*
301 * The BSD symbol table always contains the count fields even
302 * if there are no entries in it.
303 */
304 if (e->e_u.e_ar.e_rawsymtabsz < 2 * sizeof(long))
305 goto symtaberror;
306
307 p = p0 = (unsigned char *) e->e_u.e_ar.e_rawsymtab;
308 end = p0 + e->e_u.e_ar.e_rawsymtabsz;
309
310 /*
311 * Retrieve the size of the array of ranlib descriptors and
312 * check it for validity.
313 */
314 GET_LONG(p, arraysize);
315
316 if (arraysize < 0 || p0 + arraysize >= end ||
317 ((size_t) arraysize % entrysize != 0))
318 goto symtaberror;
319
320 /*
321 * Check the value of the string table size.
322 */
323 s = p + arraysize;
324 GET_LONG(s, strtabsize);
325
326 s0 = s; /* Start of string table. */
327 if (strtabsize < 0 || s0 + strtabsize > end)
328 goto symtaberror;
329
330 nentries = (size_t) arraysize / entrysize;
331
332 /*
333 * Allocate space for the returned Elf_Arsym array.
334 */
335 if ((symtab = malloc(sizeof(Elf_Arsym) * (nentries + 1))) == NULL) {
336 LIBELF_SET_ERROR(RESOURCE, 0);
337 return (NULL);
338 }
339
340 /* Read in symbol table entries. */
341 for (n = 0, sym = symtab; n < nentries; n++, sym++) {
342 GET_LONG(p, stroffset);
343 GET_LONG(p, fileoffset);
344
345 if (stroffset < 0 || fileoffset < 0 ||
346 (size_t) fileoffset >= e->e_rawsize)
347 goto symtaberror;
348
349 s = s0 + stroffset;
350
351 if (s >= end)
352 goto symtaberror;
353
354 sym->as_off = (off_t) fileoffset;
355 sym->as_hash = elf_hash((char *) s);
356 sym->as_name = (char *) s;
357 }
358
359 /* Fill up the sentinel entry. */
360 sym->as_name = NULL;
361 sym->as_hash = ~0UL;
362 sym->as_off = (off_t) 0;
363
364 /* Remember the processed symbol table. */
365 e->e_u.e_ar.e_symtab = symtab;
366
367 *count = e->e_u.e_ar.e_symtabsz = nentries + 1;
368
369 return (symtab);
370
371 symtaberror:
372 if (symtab)
373 free(symtab);
374 LIBELF_SET_ERROR(ARCHIVE, 0);
375 return (NULL);
376 }
377
378 /*
379 * An SVR4-style ar(1) symbol table has the following layout:
380 *
381 * - The first 4 bytes are a binary count of the number of entries in the
382 * symbol table, stored MSB-first.
383 * - Then there are 'n' 4-byte binary offsets, also stored MSB first.
384 * - Following this, there are 'n' null-terminated strings.
385 */
386
387 #define GET_WORD(P, V) do { \
388 (V) = 0; \
389 (V) = (P)[0]; (V) <<= 8; \
390 (V) += (P)[1]; (V) <<= 8; \
391 (V) += (P)[2]; (V) <<= 8; \
392 (V) += (P)[3]; \
393 } while (0)
394
395 #define INTSZ 4
396
397
398 Elf_Arsym *
_libelf_ar_process_svr4_symtab(Elf * e,size_t * count)399 _libelf_ar_process_svr4_symtab(Elf *e, size_t *count)
400 {
401 uint32_t off;
402 size_t n, nentries;
403 Elf_Arsym *symtab, *sym;
404 unsigned char *p, *s, *end;
405
406 assert(e != NULL);
407 assert(count != NULL);
408 assert(e->e_u.e_ar.e_symtab == NULL);
409
410 symtab = NULL;
411
412 if (e->e_u.e_ar.e_rawsymtabsz < INTSZ)
413 goto symtaberror;
414
415 p = (unsigned char *) e->e_u.e_ar.e_rawsymtab;
416 end = p + e->e_u.e_ar.e_rawsymtabsz;
417
418 GET_WORD(p, nentries);
419 p += INTSZ;
420
421 if (nentries == 0 || p + nentries * INTSZ >= end)
422 goto symtaberror;
423
424 /* Allocate space for a nentries + a sentinel. */
425 if ((symtab = malloc(sizeof(Elf_Arsym) * (nentries+1))) == NULL) {
426 LIBELF_SET_ERROR(RESOURCE, 0);
427 return (NULL);
428 }
429
430 s = p + (nentries * INTSZ); /* start of the string table. */
431
432 for (n = nentries, sym = symtab; n > 0; n--) {
433 if (s >= end)
434 goto symtaberror;
435
436 GET_WORD(p, off);
437 if (off >= e->e_rawsize)
438 goto symtaberror;
439
440 sym->as_off = (off_t) off;
441 sym->as_hash = elf_hash((char *) s);
442 sym->as_name = (char *) s;
443
444 p += INTSZ;
445 sym++;
446
447 for (; s < end && *s++ != '\0';) /* skip to next string */
448 ;
449 }
450
451 /* Fill up the sentinel entry. */
452 sym->as_name = NULL;
453 sym->as_hash = ~0UL;
454 sym->as_off = (off_t) 0;
455
456 *count = e->e_u.e_ar.e_symtabsz = nentries + 1;
457 e->e_u.e_ar.e_symtab = symtab;
458
459 return (symtab);
460
461 symtaberror:
462 if (symtab)
463 free(symtab);
464 LIBELF_SET_ERROR(ARCHIVE, 0);
465 return (NULL);
466 }
467