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 an a.out 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/types.h>
63 #include <sys/stat.h>
64 #include <sys/errno.h>
65 #include <unistd.h>
66 #include <stdio.h>
67 #include <stdlib.h>
68 #include <string.h>
69 #include <fcntl.h>
70 #include <a.out.h>
71 
72 #include "extern.h"
73 
74 char *pname = "crunchide";
75 
76 void usage(void);
77 
78 void add_to_keep_list(char *symbol);
79 void add_file_to_keep_list(char *filename);
80 
81 int hide_syms(const char *filename);
82 
83 int verbose;
84 
85 int main(int, char *[]);
86 
87 int
88 main(int argc, char **argv)
89 {
90     int ch, errors;
91 
92     if(argc > 0) pname = argv[0];
93 
94     while ((ch = getopt(argc, argv, "k:f:v")) != -1)
95 	switch(ch) {
96 	case 'k':
97 	    add_to_keep_list(optarg);
98 	    break;
99 	case 'f':
100 	    add_file_to_keep_list(optarg);
101 	    break;
102 	case 'v':
103 	    verbose = 1;
104 	    break;
105 	default:
106 	    usage();
107 	}
108 
109     argc -= optind;
110     argv += optind;
111 
112     if(argc == 0) usage();
113 
114     errors = 0;
115     while(argc) {
116 	if (hide_syms(*argv))
117 		errors = 1;
118 	argc--, argv++;
119     }
120 
121     return errors;
122 }
123 
124 void
125 usage(void)
126 {
127     fprintf(stderr,
128 	    "usage: %s [-k <symbol-name>] [-f <keep-list-file>] <files> ...\n",
129 	    pname);
130     exit(1);
131 }
132 
133 /* ---------------------------- */
134 
135 struct keep {
136     struct keep *next;
137     char *sym;
138 } *keep_list;
139 
140 void
141 add_to_keep_list(char *symbol)
142 {
143     struct keep *newp, *prevp, *curp;
144     int cmp;
145 
146     cmp = 0;
147 
148     for(curp = keep_list, prevp = NULL; curp; prevp = curp, curp = curp->next)
149 	if((cmp = strcmp(symbol, curp->sym)) <= 0) break;
150 
151     if(curp && cmp == 0)
152 	return;	/* already in table */
153 
154     newp = (struct keep *) malloc(sizeof(struct keep));
155     if(newp) newp->sym = strdup(symbol);
156     if(newp == NULL || newp->sym == NULL) {
157 	fprintf(stderr, "%s: out of memory for keep list\n", pname);
158 	exit(1);
159     }
160 
161     newp->next = curp;
162     if(prevp) prevp->next = newp;
163     else keep_list = newp;
164 }
165 
166 int
167 in_keep_list(const char *symbol)
168 {
169     struct keep *curp;
170     int cmp;
171 
172     cmp = 0;
173 
174     for(curp = keep_list; curp; curp = curp->next)
175 	if((cmp = strcmp(symbol, curp->sym)) <= 0) break;
176 
177     return curp && cmp == 0;
178 }
179 
180 void
181 add_file_to_keep_list(char *filename)
182 {
183     FILE *keepf;
184     char symbol[1024];
185     int len;
186 
187     if((keepf = fopen(filename, "r")) == NULL) {
188 	perror(filename);
189 	usage();
190     }
191 
192     while(fgets(symbol, sizeof(symbol), keepf)) {
193 	len = strlen(symbol);
194 	if(len && symbol[len-1] == '\n')
195 	    symbol[len-1] = '\0';
196 
197 	add_to_keep_list(symbol);
198     }
199     fclose(keepf);
200 }
201 
202 /* ---------------------------- */
203 
204 struct {
205 	const char *name;
206 	int	(*check)(int, const char *);	/* 1 if match, zero if not */
207 	int	(*hide)(int, const char *);	/* non-zero if error */
208 } exec_formats[] = {
209 #ifdef NLIST_AOUT
210 	{	"a.out",	check_aout,	hide_aout,	},
211 #endif
212 #ifdef NLIST_ECOFF
213 	{	"ECOFF",	check_elf64,	hide_elf64,	},
214 #endif
215 #ifdef NLIST_ELF32
216 	{	"ELF32",	check_elf32,	hide_elf32,	},
217 #endif
218 #ifdef NLIST_ELF64
219 	{	"ELF64",	check_elf64,	hide_elf64,	},
220 #endif
221 };
222 
223 int
224 hide_syms(const char *filename)
225 {
226 	int fd, i, n, rv;
227 
228 	fd = open(filename, O_RDWR, 0);
229 	if (fd == -1) {
230 		perror(filename);
231 		return 1;
232 	}
233 
234 	rv = 0;
235 
236         n = sizeof exec_formats / sizeof exec_formats[0];
237         for (i = 0; i < n; i++) {
238 		if (lseek(fd, 0, SEEK_SET) != 0) {
239 			perror(filename);
240 			goto err;
241 		}
242                 if ((*exec_formats[i].check)(fd, filename) != 0)
243                         break;
244 	}
245 	if (i == n) {
246 		fprintf(stderr, "%s: unknown executable format\n", filename);
247 		goto err;
248 	}
249 
250 	if (verbose)
251 		fprintf(stderr, "%s is an %s binary\n", filename,
252 		    exec_formats[i].name);
253 
254 	if (lseek(fd, 0, SEEK_SET) != 0) {
255 		perror(filename);
256 		goto err;
257 	}
258 	rv = (*exec_formats[i].hide)(fd, filename);
259 
260 out:
261 	close (fd);
262 	return (rv);
263 
264 err:
265 	rv = 1;
266 	goto out;
267 }
268