1/*-
2 * Copyright (c) 2007-2009 Kai Wang
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 <sys/param.h>
28#include <sys/stat.h>
29#include <err.h>
30#include <stdlib.h>
31#include <string.h>
32#include <unistd.h>
33
34#ifndef LIBELF_AR
35#include <archive.h>
36#include <archive_entry.h>
37#endif	/* ! LIBELF_AR */
38
39#include "elfcopy.h"
40
41ELFTC_VCSID("$Id: archive.c 3490 2016-08-31 00:12:22Z emaste $");
42
43#define _ARMAG_LEN 8		/* length of ar magic string */
44#define _ARHDR_LEN 60		/* length of ar header */
45#define _INIT_AS_CAP 128	/* initial archive string table size */
46#define _INIT_SYMOFF_CAP (256*(sizeof(uint32_t))) /* initial so table size */
47#define _INIT_SYMNAME_CAP 1024			  /* initial sn table size */
48#define _MAXNAMELEN_SVR4 15	/* max member name length in svr4 variant */
49
50#ifndef LIBELF_AR
51static void ac_read_objs(struct elfcopy *ecp, int ifd);
52static void ac_write_cleanup(struct elfcopy *ecp);
53static void ac_write_data(struct archive *a, const void *buf, size_t s);
54static void ac_write_objs(struct elfcopy *ecp, int ofd);
55#endif	/* ! LIBELF_AR */
56static void add_to_ar_str_table(struct elfcopy *elfcopy, const char *name);
57static void add_to_ar_sym_table(struct elfcopy *ecp, const char *name);
58static void extract_arsym(struct elfcopy *ecp);
59static void process_ar_obj(struct elfcopy *ecp, struct ar_obj *obj);
60static void sync_ar(struct elfcopy *ecp);
61
62
63static void
64process_ar_obj(struct elfcopy *ecp, struct ar_obj *obj)
65{
66	struct stat	 sb;
67	char		*tempfile;
68	int		 fd;
69
70	/* Output to a temporary file. */
71	create_tempfile(NULL, &tempfile, &fd);
72	if ((ecp->eout = elf_begin(fd, ELF_C_WRITE, NULL)) == NULL)
73		errx(EXIT_FAILURE, "elf_begin() failed: %s",
74		    elf_errmsg(-1));
75	elf_flagelf(ecp->eout, ELF_C_SET, ELF_F_LAYOUT);
76	create_elf(ecp);
77	elf_end(ecp->ein);
78	elf_end(ecp->eout);
79	free(obj->buf);
80	obj->buf = NULL;
81
82	/* Extract archive symbols. */
83	if (lseek(fd, 0, SEEK_SET) < 0)
84		err(EXIT_FAILURE, "lseek failed for '%s'", tempfile);
85	if ((ecp->eout = elf_begin(fd, ELF_C_READ, NULL)) == NULL)
86		errx(EXIT_FAILURE, "elf_begin() failed: %s",
87		    elf_errmsg(-1));
88	extract_arsym(ecp);
89	elf_end(ecp->eout);
90
91	if (fstat(fd, &sb) == -1)
92		err(EXIT_FAILURE, "fstat %s failed", tempfile);
93	if (lseek(fd, 0, SEEK_SET) < 0)
94		err(EXIT_FAILURE, "lseek %s failed", tempfile);
95	obj->size = sb.st_size;
96	if ((obj->maddr = malloc(obj->size)) == NULL)
97		err(EXIT_FAILURE, "memory allocation failed for '%s'",
98		    tempfile);
99	if ((size_t) read(fd, obj->maddr, obj->size) != obj->size)
100		err(EXIT_FAILURE, "read failed for '%s'", tempfile);
101	if (unlink(tempfile))
102		err(EXIT_FAILURE, "unlink %s failed", tempfile);
103	free(tempfile);
104	close(fd);
105	if (strlen(obj->name) > _MAXNAMELEN_SVR4)
106		add_to_ar_str_table(ecp, obj->name);
107	ecp->rela_off += _ARHDR_LEN + obj->size + obj->size % 2;
108	STAILQ_INSERT_TAIL(&ecp->v_arobj, obj, objs);
109}
110
111/*
112 * Append to the archive string table buffer.
113 */
114static void
115add_to_ar_str_table(struct elfcopy *ecp, const char *name)
116{
117
118	if (ecp->as == NULL) {
119		ecp->as_cap = _INIT_AS_CAP;
120		ecp->as_sz = 0;
121		if ((ecp->as = malloc(ecp->as_cap)) == NULL)
122			err(EXIT_FAILURE, "malloc failed");
123	}
124
125	/*
126	 * The space required for holding one member name in as table includes:
127	 * strlen(name) + (1 for '/') + (1 for '\n') + (possibly 1 for padding).
128	 */
129	while (ecp->as_sz + strlen(name) + 3 > ecp->as_cap) {
130		ecp->as_cap *= 2;
131		ecp->as = realloc(ecp->as, ecp->as_cap);
132		if (ecp->as == NULL)
133			err(EXIT_FAILURE, "realloc failed");
134	}
135	strncpy(&ecp->as[ecp->as_sz], name, strlen(name));
136	ecp->as_sz += strlen(name);
137	ecp->as[ecp->as_sz++] = '/';
138	ecp->as[ecp->as_sz++] = '\n';
139}
140
141/*
142 * Append to the archive symbol table buffer.
143 */
144static void
145add_to_ar_sym_table(struct elfcopy *ecp, const char *name)
146{
147
148	if (ecp->s_so == NULL) {
149		if ((ecp->s_so = malloc(_INIT_SYMOFF_CAP)) == NULL)
150			err(EXIT_FAILURE, "malloc failed");
151		ecp->s_so_cap = _INIT_SYMOFF_CAP;
152		ecp->s_cnt = 0;
153	}
154
155	if (ecp->s_sn == NULL) {
156		if ((ecp->s_sn = malloc(_INIT_SYMNAME_CAP)) == NULL)
157			err(EXIT_FAILURE, "malloc failed");
158		ecp->s_sn_cap = _INIT_SYMNAME_CAP;
159		ecp->s_sn_sz = 0;
160	}
161
162	if (ecp->s_cnt * sizeof(uint32_t) >= ecp->s_so_cap) {
163		ecp->s_so_cap *= 2;
164		ecp->s_so = realloc(ecp->s_so, ecp->s_so_cap);
165		if (ecp->s_so == NULL)
166			err(EXIT_FAILURE, "realloc failed");
167	}
168	ecp->s_so[ecp->s_cnt] = ecp->rela_off;
169	ecp->s_cnt++;
170
171	/*
172	 * The space required for holding one symbol name in sn table includes:
173	 * strlen(name) + (1 for '\n') + (possibly 1 for padding).
174	 */
175	while (ecp->s_sn_sz + strlen(name) + 2 > ecp->s_sn_cap) {
176		ecp->s_sn_cap *= 2;
177		ecp->s_sn = realloc(ecp->s_sn, ecp->s_sn_cap);
178		if (ecp->s_sn == NULL)
179			err(EXIT_FAILURE, "realloc failed");
180	}
181	strncpy(&ecp->s_sn[ecp->s_sn_sz], name, strlen(name));
182	ecp->s_sn_sz += strlen(name);
183	ecp->s_sn[ecp->s_sn_sz++] = '\0';
184}
185
186static void
187sync_ar(struct elfcopy *ecp)
188{
189	size_t s_sz;		/* size of archive symbol table. */
190	size_t pm_sz;		/* size of pseudo members */
191	int i;
192
193	/*
194	 * Pad the symbol name string table. It is treated specially because
195	 * symbol name table should be padded by a '\0', not the common '\n'
196	 * for other members. The size of sn table includes the pad bit.
197	 */
198	if (ecp->s_cnt != 0 && ecp->s_sn_sz % 2 != 0)
199		ecp->s_sn[ecp->s_sn_sz++] = '\0';
200
201	/*
202	 * Archive string table is padded by a "\n" as the normal members.
203	 * The difference is that the size of archive string table counts
204	 * in the pad bit, while normal members' size fileds do not.
205	 */
206	if (ecp->as != NULL && ecp->as_sz % 2 != 0)
207		ecp->as[ecp->as_sz++] = '\n';
208
209	/*
210	 * If there is a symbol table, calculate the size of pseudo members,
211	 * convert previously stored relative offsets to absolute ones, and
212	 * then make them Big Endian.
213	 *
214	 * absolute_offset = htobe32(relative_offset + size_of_pseudo_members)
215	 */
216
217	if (ecp->s_cnt != 0) {
218		s_sz = (ecp->s_cnt + 1) * sizeof(uint32_t) + ecp->s_sn_sz;
219		pm_sz = _ARMAG_LEN + (_ARHDR_LEN + s_sz);
220		if (ecp->as != NULL)
221			pm_sz += _ARHDR_LEN + ecp->as_sz;
222		for (i = 0; (size_t)i < ecp->s_cnt; i++)
223			*(ecp->s_so + i) = htobe32(*(ecp->s_so + i) +
224			    pm_sz);
225	}
226}
227
228/*
229 * Extract global symbols from archive members.
230 */
231static void
232extract_arsym(struct elfcopy *ecp)
233{
234	Elf_Scn		*scn;
235	GElf_Shdr	 shdr;
236	GElf_Sym	 sym;
237	Elf_Data	*data;
238	char		*name;
239	size_t		 n, shstrndx;
240	int		 elferr, tabndx, len, i;
241
242	if (elf_kind(ecp->eout) != ELF_K_ELF) {
243		warnx("internal: cannot extract symbols from non-elf object");
244		return;
245	}
246	if (elf_getshstrndx(ecp->eout, &shstrndx) == 0) {
247		warnx("elf_getshstrndx failed: %s", elf_errmsg(-1));
248		return;
249	}
250
251	tabndx = -1;
252	scn = NULL;
253	while ((scn = elf_nextscn(ecp->eout, scn)) != NULL) {
254		if (gelf_getshdr(scn, &shdr) != &shdr) {
255			warnx("elf_getshdr failed: %s", elf_errmsg(-1));
256			continue;
257		}
258		if ((name = elf_strptr(ecp->eout, shstrndx, shdr.sh_name)) ==
259		    NULL) {
260			warnx("elf_strptr failed: %s", elf_errmsg(-1));
261			continue;
262		}
263		if (strcmp(name, ".strtab") == 0) {
264			tabndx = elf_ndxscn(scn);
265			break;
266		}
267	}
268	elferr = elf_errno();
269	if (elferr != 0)
270		warnx("elf_nextscn failed: %s", elf_errmsg(elferr));
271
272	/* Ignore members without symbol table. */
273	if (tabndx == -1)
274		return;
275
276	scn = NULL;
277	while ((scn = elf_nextscn(ecp->eout, scn)) != NULL) {
278		if (gelf_getshdr(scn, &shdr) != &shdr) {
279			warnx("elf_getshdr failed: %s", elf_errmsg(-1));
280			continue;
281		}
282		if (shdr.sh_type != SHT_SYMTAB)
283			continue;
284
285		data = NULL;
286		n = 0;
287		while (n < shdr.sh_size &&
288		    (data = elf_getdata(scn, data)) != NULL) {
289			len = data->d_size / shdr.sh_entsize;
290			for (i = 0; i < len; i++) {
291				if (gelf_getsym(data, i, &sym) != &sym) {
292					warnx("gelf_getsym failed: %s",
293					     elf_errmsg(-1));
294					continue;
295				}
296
297				/* keep only global or weak symbols */
298				if (GELF_ST_BIND(sym.st_info) != STB_GLOBAL &&
299				    GELF_ST_BIND(sym.st_info) != STB_WEAK)
300					continue;
301
302				/* keep only defined symbols */
303				if (sym.st_shndx == SHN_UNDEF)
304					continue;
305
306				if ((name = elf_strptr(ecp->eout, tabndx,
307				    sym.st_name)) == NULL) {
308					warnx("elf_strptr failed: %s",
309					     elf_errmsg(-1));
310					continue;
311				}
312
313				add_to_ar_sym_table(ecp, name);
314			}
315		}
316	}
317	elferr = elf_errno();
318	if (elferr != 0)
319		warnx("elf_nextscn failed: %s", elf_errmsg(elferr));
320}
321
322#ifndef LIBELF_AR
323
324/*
325 * Convenient wrapper for general libarchive error handling.
326 */
327#define	AC(CALL) do {							\
328	if ((CALL))							\
329		errx(EXIT_FAILURE, "%s", archive_error_string(a));	\
330} while (0)
331
332/* Earlier versions of libarchive had some functions that returned 'void'. */
333#if	ARCHIVE_VERSION_NUMBER >= 2000000
334#define	ACV(CALL) 	AC(CALL)
335#else
336#define	ACV(CALL)	do {						\
337		(CALL);							\
338	} while (0)
339#endif
340
341int
342ac_detect_ar(int ifd)
343{
344	struct archive		*a;
345	struct archive_entry	*entry;
346	int			 r;
347
348	r = -1;
349	if ((a = archive_read_new()) == NULL)
350		return (0);
351	archive_read_support_format_ar(a);
352	if (archive_read_open_fd(a, ifd, 10240) == ARCHIVE_OK)
353		r = archive_read_next_header(a, &entry);
354	archive_read_close(a);
355	archive_read_free(a);
356
357	return (r == ARCHIVE_OK);
358}
359
360void
361ac_create_ar(struct elfcopy *ecp, int ifd, int ofd)
362{
363
364	ac_read_objs(ecp, ifd);
365	sync_ar(ecp);
366	ac_write_objs(ecp, ofd);
367	ac_write_cleanup(ecp);
368}
369
370static void
371ac_read_objs(struct elfcopy *ecp, int ifd)
372{
373	struct archive		*a;
374	struct archive_entry	*entry;
375	struct ar_obj		*obj;
376	const char		*name;
377	char			*buff;
378	size_t			 size;
379	int			 r;
380
381	ecp->rela_off = 0;
382	if (lseek(ifd, 0, SEEK_SET) == -1)
383		err(EXIT_FAILURE, "lseek failed");
384	if ((a = archive_read_new()) == NULL)
385		errx(EXIT_FAILURE, "archive_read_new failed");
386	archive_read_support_format_ar(a);
387	AC(archive_read_open_fd(a, ifd, 10240));
388	for(;;) {
389		r = archive_read_next_header(a, &entry);
390		if (r == ARCHIVE_FATAL)
391			errx(EXIT_FAILURE, "%s", archive_error_string(a));
392		if (r == ARCHIVE_EOF)
393			break;
394		if (r == ARCHIVE_WARN || r == ARCHIVE_RETRY)
395			warnx("%s", archive_error_string(a));
396		if (r == ARCHIVE_RETRY)
397			continue;
398
399		name = archive_entry_pathname(entry);
400
401		/* skip pseudo members. */
402		if (strcmp(name, "/") == 0 || strcmp(name, "//") == 0)
403			continue;
404
405		size = archive_entry_size(entry);
406
407		if (size > 0) {
408			if ((buff = malloc(size)) == NULL)
409				err(EXIT_FAILURE, "malloc failed");
410			if (archive_read_data(a, buff, size) != (ssize_t)size) {
411				warnx("%s", archive_error_string(a));
412				free(buff);
413				continue;
414			}
415			if ((obj = malloc(sizeof(*obj))) == NULL)
416				err(EXIT_FAILURE, "malloc failed");
417			if ((obj->name = strdup(name)) == NULL)
418				err(EXIT_FAILURE, "strdup failed");
419			obj->buf = buff;
420			obj->uid = archive_entry_uid(entry);
421			obj->gid = archive_entry_gid(entry);
422			obj->md = archive_entry_mode(entry);
423			obj->mtime = archive_entry_mtime(entry);
424			if ((ecp->ein = elf_memory(buff, size)) == NULL)
425				errx(EXIT_FAILURE, "elf_memory() failed: %s",
426				    elf_errmsg(-1));
427			if (elf_kind(ecp->ein) != ELF_K_ELF)
428				errx(EXIT_FAILURE,
429				    "file format not recognized");
430			process_ar_obj(ecp, obj);
431		}
432	}
433	AC(archive_read_close(a));
434	ACV(archive_read_free(a));
435}
436
437static void
438ac_write_objs(struct elfcopy *ecp, int ofd)
439{
440	struct archive		*a;
441	struct archive_entry	*entry;
442	struct ar_obj		*obj;
443	time_t			 timestamp;
444	int			 nr;
445
446	if ((a = archive_write_new()) == NULL)
447		errx(EXIT_FAILURE, "archive_write_new failed");
448	archive_write_set_format_ar_svr4(a);
449	AC(archive_write_open_fd(a, ofd));
450
451	/* Write the archive symbol table, even if it's empty. */
452	entry = archive_entry_new();
453	archive_entry_copy_pathname(entry, "/");
454	if (elftc_timestamp(&timestamp) != 0)
455		err(EXIT_FAILURE, "elftc_timestamp");
456	archive_entry_set_mtime(entry, timestamp, 0);
457	archive_entry_set_size(entry, (ecp->s_cnt + 1) * sizeof(uint32_t) +
458	    ecp->s_sn_sz);
459	AC(archive_write_header(a, entry));
460	nr = htobe32(ecp->s_cnt);
461	ac_write_data(a, &nr, sizeof(uint32_t));
462	ac_write_data(a, ecp->s_so, sizeof(uint32_t) * ecp->s_cnt);
463	ac_write_data(a, ecp->s_sn, ecp->s_sn_sz);
464	archive_entry_free(entry);
465
466	/* Write the archive string table, if exist. */
467	if (ecp->as != NULL) {
468		entry = archive_entry_new();
469		archive_entry_copy_pathname(entry, "//");
470		archive_entry_set_size(entry, ecp->as_sz);
471		AC(archive_write_header(a, entry));
472		ac_write_data(a, ecp->as, ecp->as_sz);
473		archive_entry_free(entry);
474	}
475
476	/* Write normal members. */
477	STAILQ_FOREACH(obj, &ecp->v_arobj, objs) {
478		entry = archive_entry_new();
479		archive_entry_copy_pathname(entry, obj->name);
480		archive_entry_set_uid(entry, obj->uid);
481		archive_entry_set_gid(entry, obj->gid);
482		archive_entry_set_mode(entry, obj->md);
483		archive_entry_set_size(entry, obj->size);
484		archive_entry_set_mtime(entry, obj->mtime, 0);
485		archive_entry_set_filetype(entry, AE_IFREG);
486		AC(archive_write_header(a, entry));
487		ac_write_data(a, obj->maddr, obj->size);
488		archive_entry_free(entry);
489	}
490
491	AC(archive_write_close(a));
492	ACV(archive_write_free(a));
493}
494
495static void
496ac_write_cleanup(struct elfcopy *ecp)
497{
498	struct ar_obj		*obj, *obj_temp;
499
500	STAILQ_FOREACH_SAFE(obj, &ecp->v_arobj, objs, obj_temp) {
501		STAILQ_REMOVE(&ecp->v_arobj, obj, ar_obj, objs);
502		if (obj->maddr != NULL)
503			free(obj->maddr);
504		free(obj->name);
505		free(obj);
506	}
507
508	free(ecp->as);
509	free(ecp->s_so);
510	free(ecp->s_sn);
511	ecp->as = NULL;
512	ecp->s_so = NULL;
513	ecp->s_sn = NULL;
514}
515
516/*
517 * Wrapper for archive_write_data().
518 */
519static void
520ac_write_data(struct archive *a, const void *buf, size_t s)
521{
522	if (archive_write_data(a, buf, s) != (ssize_t)s)
523		errx(EXIT_FAILURE, "%s", archive_error_string(a));
524}
525
526#endif	/* ! LIBELF_AR */
527