1/*	$NetBSD: crunchide.c,v 1.8 1997/11/01 06:51:45 lukem Exp $	*/
2/*
3 * Copyright (c) 1997 Christopher G. Demetriou.  All rights reserved.
4 * Copyright (c) 1994 University of Maryland
5 * All Rights Reserved.
6 *
7 * Permission to use, copy, modify, distribute, and sell this software and its
8 * documentation for any purpose is hereby granted without fee, provided that
9 * the above copyright notice appear in all copies and that both that
10 * copyright notice and this permission notice appear in supporting
11 * documentation, and that the name of U.M. not be used in advertising or
12 * publicity pertaining to distribution of the software without specific,
13 * written prior permission.  U.M. makes no representations about the
14 * suitability of this software for any purpose.  It is provided "as is"
15 * without express or implied warranty.
16 *
17 * U.M. DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL U.M.
19 * BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
20 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
21 * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
22 * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
23 *
24 * Author: James da Silva, Systems Design and Analysis Group
25 *			   Computer Science Department
26 *			   University of Maryland at College Park
27 */
28/*
29 * crunchide.c - tiptoes through a symbol table, hiding all defined
30 *	global symbols.  Allows the user to supply a "keep list" of symbols
31 *	that are not to be hidden.  This program relies on the use of the
32 * 	linker's -dc flag to actually put global bss data into the file's
33 * 	bss segment (rather than leaving it as undefined "common" data).
34 *
35 * 	The point of all this is to allow multiple programs to be linked
36 *	together without getting multiple-defined errors.
37 *
38 *	For example, consider a program "foo.c".  It can be linked with a
39 *	small stub routine, called "foostub.c", eg:
40 *	    int foo_main(int argc, char **argv){ return main(argc, argv); }
41 *      like so:
42 *	    cc -c foo.c foostub.c
43 *	    ld -dc -r foo.o foostub.o -o foo.combined.o
44 *	    crunchide -k _foo_main foo.combined.o
45 *	at this point, foo.combined.o can be linked with another program
46 * 	and invoked with "foo_main(argc, argv)".  foo's main() and any
47 * 	other globals are hidden and will not conflict with other symbols.
48 *
49 * TODO:
50 *	- resolve the theoretical hanging reloc problem (see check_reloc()
51 *	  below). I have yet to see this problem actually occur in any real
52 *	  program. In what cases will gcc/gas generate code that needs a
53 *	  relative reloc from a global symbol, other than PIC?  The
54 *	  solution is to not hide the symbol from the linker in this case,
55 *	  but to generate some random name for it so that it doesn't link
56 *	  with anything but holds the place for the reloc.
57 *      - arrange that all the BSS segments start at the same address, so
58 *	  that the final crunched binary BSS size is the max of all the
59 *	  component programs' BSS sizes, rather than their sum.
60 */
61
62#include <sys/cdefs.h>
63#ifndef lint
64__RCSID("$NetBSD: crunchide.c,v 1.8 1997/11/01 06:51:45 lukem Exp $");
65#endif
66__FBSDID("$FreeBSD$");
67
68#include <sys/types.h>
69#include <sys/stat.h>
70#include <sys/errno.h>
71#include <unistd.h>
72#include <stdio.h>
73#include <stdlib.h>
74#include <string.h>
75#include <fcntl.h>
76
77#include "extern.h"
78
79char *pname = "crunchide";
80
81void usage(void);
82
83void add_to_keep_list(char *symbol);
84void add_file_to_keep_list(char *filename);
85
86int hide_syms(const char *filename);
87
88int verbose;
89
90int main(int, char *[]);
91
92int
93main(int argc, char **argv)
94{
95    int ch, errors;
96
97    if(argc > 0) pname = argv[0];
98
99    while ((ch = getopt(argc, argv, "k:f:v")) != -1)
100	switch(ch) {
101	case 'k':
102	    add_to_keep_list(optarg);
103	    break;
104	case 'f':
105	    add_file_to_keep_list(optarg);
106	    break;
107	case 'v':
108	    verbose = 1;
109	    break;
110	default:
111	    usage();
112	}
113
114    argc -= optind;
115    argv += optind;
116
117    if(argc == 0) usage();
118
119    errors = 0;
120    while(argc) {
121	if (hide_syms(*argv))
122		errors = 1;
123	argc--, argv++;
124    }
125
126    return errors;
127}
128
129void
130usage(void)
131{
132    fprintf(stderr,
133	    "usage: %s [-k <symbol-name>] [-f <keep-list-file>] <files> ...\n",
134	    pname);
135    exit(1);
136}
137
138/* ---------------------------- */
139
140struct keep {
141    struct keep *next;
142    char *sym;
143} *keep_list;
144
145void
146add_to_keep_list(char *symbol)
147{
148    struct keep *newp, *prevp, *curp;
149    int cmp;
150
151    cmp = 0;
152
153    for(curp = keep_list, prevp = NULL; curp; prevp = curp, curp = curp->next)
154	if((cmp = strcmp(symbol, curp->sym)) <= 0) break;
155
156    if(curp && cmp == 0)
157	return;	/* already in table */
158
159    newp = (struct keep *) malloc(sizeof(struct keep));
160    if(newp) newp->sym = strdup(symbol);
161    if(newp == NULL || newp->sym == NULL) {
162	fprintf(stderr, "%s: out of memory for keep list\n", pname);
163	exit(1);
164    }
165
166    newp->next = curp;
167    if(prevp) prevp->next = newp;
168    else keep_list = newp;
169}
170
171int
172in_keep_list(const char *symbol)
173{
174    struct keep *curp;
175    int cmp;
176
177    cmp = 0;
178
179    for(curp = keep_list; curp; curp = curp->next)
180	if((cmp = strcmp(symbol, curp->sym)) <= 0) break;
181
182    return curp && cmp == 0;
183}
184
185void
186add_file_to_keep_list(char *filename)
187{
188    FILE *keepf;
189    char symbol[1024];
190    int len;
191
192    if((keepf = fopen(filename, "r")) == NULL) {
193	perror(filename);
194	usage();
195    }
196
197    while(fgets(symbol, sizeof(symbol), keepf)) {
198	len = strlen(symbol);
199	if(len && symbol[len-1] == '\n')
200	    symbol[len-1] = '\0';
201
202	add_to_keep_list(symbol);
203    }
204    fclose(keepf);
205}
206
207/* ---------------------------- */
208
209struct {
210	const char *name;
211	int	(*check)(int, const char *);	/* 1 if match, zero if not */
212	int	(*hide)(int, const char *);	/* non-zero if error */
213} exec_formats[] = {
214#ifdef NLIST_ELF32
215	{	"ELF32",	check_elf32,	hide_elf32,	},
216#endif
217#ifdef NLIST_ELF64
218	{	"ELF64",	check_elf64,	hide_elf64,	},
219#endif
220};
221
222int
223hide_syms(const char *filename)
224{
225	int fd, i, n, rv;
226
227	fd = open(filename, O_RDWR, 0);
228	if (fd == -1) {
229		perror(filename);
230		return 1;
231	}
232
233	rv = 0;
234
235        n = sizeof exec_formats / sizeof exec_formats[0];
236        for (i = 0; i < n; i++) {
237		if (lseek(fd, 0, SEEK_SET) != 0) {
238			perror(filename);
239			goto err;
240		}
241                if ((*exec_formats[i].check)(fd, filename) != 0)
242                        break;
243	}
244	if (i == n) {
245		fprintf(stderr, "%s: unknown executable format\n", filename);
246		goto err;
247	}
248
249	if (verbose)
250		fprintf(stderr, "%s is an %s binary\n", filename,
251		    exec_formats[i].name);
252
253	if (lseek(fd, 0, SEEK_SET) != 0) {
254		perror(filename);
255		goto err;
256	}
257	rv = (*exec_formats[i].hide)(fd, filename);
258
259out:
260	close (fd);
261	return (rv);
262
263err:
264	rv = 1;
265	goto out;
266}
267