xref: /freebsd/usr.sbin/crunch/crunchide/exec_elf32.c (revision 02e9120893770924227138ba49df1edb3896112a)
1 /*-
2  * SPDX-License-Identifier: BSD-4-Clause
3  *
4  * Copyright (c) 1997 Christopher G. Demetriou.  All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  * 3. All advertising materials mentioning features or use of this software
15  *    must display the following acknowledgement:
16  *      This product includes software developed by Christopher G. Demetriou
17  *	for the NetBSD Project.
18  * 4. The name of the author may not be used to endorse or promote products
19  *    derived from this software without specific prior written permission
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
22  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
23  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
24  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
26  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
27  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
28  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
29  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
30  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
31  */
32 
33 #include <sys/cdefs.h>
34 #ifndef lint
35 #if 0
36 __RCSID("$NetBSD: exec_elf32.c,v 1.6 1999/09/20 04:12:16 christos Exp $");
37 #endif
38 #endif
39 #ifndef ELFSIZE
40 #define ELFSIZE         32
41 #endif
42 
43 #include <sys/types.h>
44 #include <sys/endian.h>
45 #include <sys/stat.h>
46 
47 #include <errno.h>
48 #include <limits.h>
49 #include <stddef.h>
50 #include <stdio.h>
51 #include <stdlib.h>
52 #include <string.h>
53 #include <unistd.h>
54 
55 #include "extern.h"
56 
57 #if (defined(NLIST_ELF32) && (ELFSIZE == 32)) || \
58     (defined(NLIST_ELF64) && (ELFSIZE == 64))
59 
60 #define	__ELF_WORD_SIZE ELFSIZE
61 #if (ELFSIZE == 32)
62 #include <sys/elf32.h>
63 #define	xewtoh(x)	((data == ELFDATA2MSB) ? be32toh(x) : le32toh(x))
64 #define	htoxew(x)	((data == ELFDATA2MSB) ? htobe32(x) : htole32(x))
65 #define	wewtoh(x)	((data == ELFDATA2MSB) ? be32toh(x) : le32toh(x))
66 #define	htowew(x)	((data == ELFDATA2MSB) ? htobe32(x) : htole32(x))
67 #elif (ELFSIZE == 64)
68 #include <sys/elf64.h>
69 #define	xewtoh(x)	((data == ELFDATA2MSB) ? be64toh(x) : le64toh(x))
70 #define	htoxew(x)	((data == ELFDATA2MSB) ? htobe64(x) : htole64(x))
71 /* elf64 Elf64_Word are 32 bits */
72 #define	wewtoh(x)	((data == ELFDATA2MSB) ? be32toh(x) : le32toh(x))
73 #define	htowew(x)	((data == ELFDATA2MSB) ? htobe32(x) : htole32(x))
74 #endif
75 #include <sys/elf_generic.h>
76 
77 #define CONCAT(x,y)     __CONCAT(x,y)
78 #define ELFNAME(x)      CONCAT(elf,CONCAT(ELFSIZE,CONCAT(_,x)))
79 #define ELFNAME2(x,y)   CONCAT(x,CONCAT(_elf,CONCAT(ELFSIZE,CONCAT(_,y))))
80 #define ELFNAMEEND(x)   CONCAT(x,CONCAT(_elf,ELFSIZE))
81 #define ELFDEFNNAME(x)  CONCAT(ELF,CONCAT(ELFSIZE,CONCAT(_,x)))
82 #ifndef ELFCLASS
83 #define ELFCLASS	CONCAT(ELFCLASS,ELFSIZE)
84 #endif
85 
86 #define	xe16toh(x)	((data == ELFDATA2MSB) ? be16toh(x) : le16toh(x))
87 #define	xe32toh(x)	((data == ELFDATA2MSB) ? be32toh(x) : le32toh(x))
88 #define	htoxe32(x)	((data == ELFDATA2MSB) ? htobe32(x) : htole32(x))
89 
90 struct shlayout {
91 	Elf_Shdr *shdr;
92 	void *bufp;
93 };
94 
95 static ssize_t
96 xreadatoff(int fd, void *buf, off_t off, size_t size, const char *fn)
97 {
98 	ssize_t rv;
99 
100 	if (lseek(fd, off, SEEK_SET) != off) {
101 		perror(fn);
102 		return -1;
103 	}
104 	if ((size_t)(rv = read(fd, buf, size)) != size) {
105 		fprintf(stderr, "%s: read error: %s\n", fn,
106 		    rv == -1 ? strerror(errno) : "short read");
107 		return -1;
108 	}
109 	return size;
110 }
111 
112 static ssize_t
113 xwriteatoff(int fd, void *buf, off_t off, size_t size, const char *fn)
114 {
115 	ssize_t rv;
116 
117 	if (lseek(fd, off, SEEK_SET) != off) {
118 		perror(fn);
119 		return -1;
120 	}
121 	if ((size_t)(rv = write(fd, buf, size)) != size) {
122 		fprintf(stderr, "%s: write error: %s\n", fn,
123 		    rv == -1 ? strerror(errno) : "short write");
124 		return -1;
125 	}
126 	return size;
127 }
128 
129 static void *
130 xmalloc(size_t size, const char *fn, const char *use)
131 {
132 	void *rv;
133 
134 	rv = malloc(size);
135 	if (rv == NULL)
136 		fprintf(stderr, "%s: out of memory (allocating for %s)\n",
137 		    fn, use);
138 	return (rv);
139 }
140 
141 static void *
142 xrealloc(void *ptr, size_t size, const char *fn, const char *use)
143 {
144 	void *rv;
145 
146 	rv = realloc(ptr, size);
147 	if (rv == NULL) {
148 		free(ptr);
149 		fprintf(stderr, "%s: out of memory (reallocating for %s)\n",
150 		    fn, use);
151 	}
152 	return (rv);
153 }
154 
155 int
156 ELFNAMEEND(check)(int fd, const char *fn __unused)
157 {
158 	Elf_Ehdr eh;
159 	struct stat sb;
160 	unsigned char data;
161 
162 	/*
163 	 * Check the header to make sure it's an ELF file (of the
164 	 * appropriate size).
165 	 */
166 	if (fstat(fd, &sb) == -1)
167 		return 0;
168 	if (sb.st_size < (off_t)(sizeof eh))
169 		return 0;
170 	if (read(fd, &eh, sizeof eh) != sizeof eh)
171 		return 0;
172 
173 	if (IS_ELF(eh) == 0 || eh.e_ident[EI_CLASS] != ELFCLASS)
174                 return 0;
175 
176 	data = eh.e_ident[EI_DATA];
177 
178 	switch (xe16toh(eh.e_machine)) {
179 	case EM_386: break;
180 	case EM_ALPHA: break;
181 #ifndef EM_AARCH64
182 #define	EM_AARCH64	183
183 #endif
184 	case EM_AARCH64: break;
185 	case EM_ARM: break;
186 	case EM_MIPS: break;
187 	case /* EM_MIPS_RS3_LE */ EM_MIPS_RS4_BE: break;
188 	case EM_PPC: break;
189 	case EM_PPC64: break;
190 #ifndef EM_RISCV
191 #define	EM_RISCV	243
192 #endif
193 	case EM_RISCV: break;
194 	case EM_S390: break;
195 	case EM_SPARCV9: break;
196 	case EM_X86_64: break;
197 /*        ELFDEFNNAME(MACHDEP_ID_CASES) */
198 
199         default:
200                 return 0;
201         }
202 
203 	return 1;
204 }
205 
206 /*
207  * This function 'hides' (some of) ELF executable file's symbols.
208  * It hides them by renaming them to "_$$hide$$ <filename> <symbolname>".
209  * Symbols in the global keep list, or which are marked as being undefined,
210  * are left alone.
211  *
212  * An old version of this code shuffled various tables around, turning
213  * global symbols to be hidden into local symbols.  That lost on the
214  * mips, because CALL16 relocs must reference global symbols, and, if
215  * those symbols were being hidden, they were no longer global.
216  *
217  * The new renaming behaviour doesn't take global symbols out of the
218  * namespace.  However, it's ... unlikely that there will ever be
219  * any collisions in practice because of the new method.
220  */
221 int
222 ELFNAMEEND(hide)(int fd, const char *fn)
223 {
224 	Elf_Ehdr ehdr;
225 	struct shlayout *layoutp = NULL;
226 	Elf_Shdr *shdrp = NULL, *symtabshdr, *strtabshdr, *shstrtabshdr;
227 	Elf_Shdr shdrshdr;
228 	Elf_Sym *symtabp = NULL;
229 	char *shstrtabp = NULL, *strtabp = NULL;
230 	Elf_Size nsyms, ewi;
231 	Elf_Off off;
232 	ssize_t shdrsize;
233 	int rv, i, weird, l, m, r, strtabidx;
234 	size_t nstrtab_size, nstrtab_nextoff, fn_size, size;
235 	char *nstrtabp = NULL;
236 	unsigned char data;
237 	const char *weirdreason = NULL;
238 	void *buf;
239 	Elf_Half shnum;
240 
241 	rv = 0;
242 	if (xreadatoff(fd, &ehdr, 0, sizeof ehdr, fn) != sizeof ehdr)
243 		goto bad;
244 
245 	data = ehdr.e_ident[EI_DATA];
246 	shnum = xe16toh(ehdr.e_shnum);
247 
248 	shdrsize = shnum * xe16toh(ehdr.e_shentsize);
249 	if ((shdrp = xmalloc(shdrsize, fn, "section header table")) == NULL)
250 		goto bad;
251 	if (xreadatoff(fd, shdrp, xewtoh(ehdr.e_shoff), shdrsize, fn) !=
252 	    shdrsize)
253 		goto bad;
254 
255 	symtabshdr = strtabshdr = shstrtabshdr = NULL;
256 	weird = 0;
257 	for (i = 0; i < shnum; i++) {
258 		switch (xe32toh(shdrp[i].sh_type)) {
259 		case SHT_SYMTAB:
260 			if (symtabshdr != NULL) {
261 				weird = 1;
262 				weirdreason = "multiple symbol tables";
263 			}
264 			symtabshdr = &shdrp[i];
265 			strtabshdr = &shdrp[xe32toh(shdrp[i].sh_link)];
266 			break;
267 		case SHT_STRTAB:
268 			if (i == xe16toh(ehdr.e_shstrndx))
269 				shstrtabshdr = &shdrp[i];
270 			break;
271 		}
272 	}
273 	if (symtabshdr == NULL)
274 		goto out;
275 	if (strtabshdr == NULL) {
276 		weird = 1;
277 		weirdreason = "string table does not exist";
278 	}
279 	if (shstrtabshdr == NULL) {
280 		weird = 1;
281 		weirdreason = "section header string table does not exist";
282 	}
283 	if (strtabshdr == shstrtabshdr) {
284 		weird = 1;
285 		weirdreason = "combined strtab and shstrtab not supported";
286 	}
287 	if (weirdreason == NULL)
288 		weirdreason = "unsupported";
289 	if (weird) {
290 		fprintf(stderr, "%s: weird executable (%s)\n", fn, weirdreason);
291 		goto bad;
292 	}
293 
294 	/*
295 	 * sort section layout table by offset
296 	 */
297 	layoutp = xmalloc((shnum + 1) * sizeof(struct shlayout),
298 	    fn, "layout table");
299 	if (layoutp == NULL)
300 		goto bad;
301 
302 	/* add a pseudo entry to represent the section header table */
303 	shdrshdr.sh_offset = ehdr.e_shoff;
304 	shdrshdr.sh_size = htoxew(shdrsize);
305 	shdrshdr.sh_addralign = htoxew(ELFSIZE / 8);
306 	layoutp[shnum].shdr = &shdrshdr;
307 
308 	/* insert and sort normal section headers */
309 	for (i = shnum; i-- != 0;) {
310 		l = i + 1;
311 		r = shnum;
312 		while (l <= r) {
313 			m = ( l + r) / 2;
314 			if (xewtoh(shdrp[i].sh_offset) >
315 			    xewtoh(layoutp[m].shdr->sh_offset))
316 				l = m + 1;
317 			else
318 				r = m - 1;
319 		}
320 
321 		if (r != i) {
322 			memmove(&layoutp[i], &layoutp[i + 1],
323 			    sizeof(struct shlayout) * (r - i));
324 		}
325 
326 		layoutp[r].shdr = &shdrp[i];
327 		layoutp[r].bufp = NULL;
328 	}
329 	++shnum;
330 
331 	/*
332 	 * load up everything we need
333 	 */
334 
335 	/* load section string table for debug use */
336 	if ((size = xewtoh(shstrtabshdr->sh_size)) == 0)
337 		goto bad;
338 	if ((shstrtabp = xmalloc(size, fn, "section string table")) == NULL)
339 		goto bad;
340 	if ((size_t)xreadatoff(fd, shstrtabp, xewtoh(shstrtabshdr->sh_offset),
341 	    size, fn) != size)
342 		goto bad;
343 	if (shstrtabp[size - 1] != '\0')
344 		goto bad;
345 
346 	/* we need symtab, strtab, and everything behind strtab */
347 	strtabidx = INT_MAX;
348 	for (i = 0; i < shnum; i++) {
349 		if (layoutp[i].shdr == &shdrshdr) {
350 			/* not load section header again */
351 			layoutp[i].bufp = shdrp;
352 			continue;
353 		}
354 		if (layoutp[i].shdr == shstrtabshdr) {
355 			/* not load section string table again */
356 			layoutp[i].bufp = shstrtabp;
357 			continue;
358 		}
359 
360 		if (layoutp[i].shdr == strtabshdr)
361 			strtabidx = i;
362 		if (layoutp[i].shdr == symtabshdr || i >= strtabidx) {
363 			off = xewtoh(layoutp[i].shdr->sh_offset);
364 			if ((size = xewtoh(layoutp[i].shdr->sh_size)) == 0)
365 				goto bad;
366 			layoutp[i].bufp = xmalloc(size, fn,
367 			    shstrtabp + xewtoh(layoutp[i].shdr->sh_name));
368 			if (layoutp[i].bufp == NULL)
369 				goto bad;
370 			if ((size_t)xreadatoff(fd, layoutp[i].bufp, off, size, fn) !=
371 			    size)
372 				goto bad;
373 
374 			/* set symbol table and string table */
375 			if (layoutp[i].shdr == symtabshdr) {
376 				symtabp = layoutp[i].bufp;
377 			} else if (layoutp[i].shdr == strtabshdr) {
378 				strtabp = layoutp[i].bufp;
379 				if (strtabp[size - 1] != '\0')
380 					goto bad;
381 			}
382 		}
383 	}
384 
385 	nstrtab_size = 256;
386 	nstrtabp = xmalloc(nstrtab_size, fn, "new string table");
387 	if (nstrtabp == NULL)
388 		goto bad;
389 	nstrtab_nextoff = 0;
390 
391 	fn_size = strlen(fn);
392 
393 	/* Prepare data structures for symbol movement. */
394 	nsyms = xewtoh(symtabshdr->sh_size) / xewtoh(symtabshdr->sh_entsize);
395 
396 	/* move symbols, making them local */
397 	for (ewi = 0; ewi < nsyms; ewi++) {
398 		Elf_Sym *sp = &symtabp[ewi];
399 		const char *symname = strtabp + xe32toh(sp->st_name);
400 		size_t newent_len;
401 		/*
402 		 * make sure there's size for the next entry, even if it's
403 		 * as large as it can be.
404 		 *
405 		 * "_$$hide$$ <filename> <symname><NUL>" ->
406 		 *    9 + 3 + sizes of fn and sym name
407 		 */
408 		while ((nstrtab_size - nstrtab_nextoff) <
409 		    strlen(symname) + fn_size + 12) {
410 			nstrtab_size *= 2;
411 			nstrtabp = xrealloc(nstrtabp, nstrtab_size, fn,
412 			    "new string table");
413 			if (nstrtabp == NULL)
414 				goto bad;
415 		}
416 
417 		sp->st_name = htowew(nstrtab_nextoff);
418 
419 		/* if it's a keeper or is undefined, don't rename it. */
420 		if (in_keep_list(symname) ||
421 		    (xe16toh(sp->st_shndx) == SHN_UNDEF)) {
422 			newent_len = sprintf(nstrtabp + nstrtab_nextoff,
423 			    "%s", symname) + 1;
424 		} else {
425 			newent_len = sprintf(nstrtabp + nstrtab_nextoff,
426 			    "_$$hide$$ %s %s", fn, symname) + 1;
427 		}
428 		nstrtab_nextoff += newent_len;
429 	}
430 	strtabshdr->sh_size = htoxew(nstrtab_nextoff);
431 
432 	/*
433 	 * update section header table in ascending order of offset
434 	 */
435 	for (i = strtabidx + 1; i < shnum; i++) {
436 		Elf_Off soff, align;
437 		soff = xewtoh(layoutp[i - 1].shdr->sh_offset) +
438 		    xewtoh(layoutp[i - 1].shdr->sh_size);
439 		align = xewtoh(layoutp[i].shdr->sh_addralign);
440 		soff = (soff + (align - 1)) & ~(align - 1);
441 		layoutp[i].shdr->sh_offset = htoxew(soff);
442 	}
443 
444 	/*
445 	 * write data to the file in descending order of offset
446 	 */
447 	for (i = shnum; i-- != 0;) {
448 		if (layoutp[i].shdr == strtabshdr) {
449 			/* new string table */
450 			buf = nstrtabp;
451 		} else
452 			buf = layoutp[i].bufp;
453 
454 		if (layoutp[i].shdr == &shdrshdr ||
455 		    layoutp[i].shdr == symtabshdr || i >= strtabidx) {
456 			if (buf == NULL)
457 				goto bad;
458 
459 			/*
460 			 * update the offset of section header table in elf
461 			 * header if needed.
462 			 */
463 			if (layoutp[i].shdr == &shdrshdr &&
464 			    ehdr.e_shoff != shdrshdr.sh_offset) {
465 				ehdr.e_shoff = shdrshdr.sh_offset;
466 				off = offsetof(Elf_Ehdr, e_shoff);
467 				size = sizeof(Elf_Off);
468 				if ((size_t)xwriteatoff(fd, &ehdr.e_shoff, off, size,
469 				    fn) != size)
470 					goto bad;
471 			}
472 
473 			off = xewtoh(layoutp[i].shdr->sh_offset);
474 			size = xewtoh(layoutp[i].shdr->sh_size);
475 			if ((size_t)xwriteatoff(fd, buf, off, size, fn) != size)
476 				goto bad;
477 		}
478 	}
479 
480 out:
481 	if (layoutp != NULL) {
482 		for (i = 0; i < shnum; i++) {
483 			if (layoutp[i].bufp != NULL)
484 				free(layoutp[i].bufp);
485 		}
486 		free(layoutp);
487 	}
488 	free(nstrtabp);
489 	return (rv);
490 
491 bad:
492 	rv = 1;
493 	goto out;
494 }
495 
496 #endif /* include this size of ELF */
497