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/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 
71 #include "extern.h"
72 
73 static char *pname = "crunchide";
74 
75 static void usage(void) __dead2;
76 
77 static void add_to_keep_list(char *symbol);
78 static void add_file_to_keep_list(char *filename);
79 
80 static int hide_syms(const char *filename);
81 
82 static int verbose;
83 
84 int
85 main(int argc, char **argv)
86 {
87     int ch, errors;
88 
89     if(argc > 0) pname = argv[0];
90 
91     while ((ch = getopt(argc, argv, "k:f:v")) != -1)
92 	switch(ch) {
93 	case 'k':
94 	    add_to_keep_list(optarg);
95 	    break;
96 	case 'f':
97 	    add_file_to_keep_list(optarg);
98 	    break;
99 	case 'v':
100 	    verbose = 1;
101 	    break;
102 	default:
103 	    usage();
104 	}
105 
106     argc -= optind;
107     argv += optind;
108 
109     if(argc == 0) usage();
110 
111     errors = 0;
112     while(argc) {
113 	if (hide_syms(*argv))
114 		errors = 1;
115 	argc--, argv++;
116     }
117 
118     return errors;
119 }
120 
121 static void
122 usage(void)
123 {
124     fprintf(stderr,
125 	    "usage: %s [-k <symbol-name>] [-f <keep-list-file>] <files> ...\n",
126 	    pname);
127     exit(1);
128 }
129 
130 /* ---------------------------- */
131 
132 static struct keep {
133     struct keep *next;
134     char *sym;
135 } *keep_list;
136 
137 static void
138 add_to_keep_list(char *symbol)
139 {
140     struct keep *newp, *prevp, *curp;
141     int cmp;
142 
143     cmp = 0;
144 
145     for(curp = keep_list, prevp = NULL; curp; prevp = curp, curp = curp->next)
146 	if((cmp = strcmp(symbol, curp->sym)) <= 0) break;
147 
148     if(curp && cmp == 0)
149 	return;	/* already in table */
150 
151     newp = (struct keep *) malloc(sizeof(struct keep));
152     if(newp) newp->sym = strdup(symbol);
153     if(newp == NULL || newp->sym == NULL) {
154 	fprintf(stderr, "%s: out of memory for keep list\n", pname);
155 	exit(1);
156     }
157 
158     newp->next = curp;
159     if(prevp) prevp->next = newp;
160     else keep_list = newp;
161 }
162 
163 int
164 in_keep_list(const char *symbol)
165 {
166     struct keep *curp;
167     int cmp;
168 
169     cmp = 0;
170 
171     for(curp = keep_list; curp; curp = curp->next)
172 	if((cmp = strcmp(symbol, curp->sym)) <= 0) break;
173 
174     return curp && cmp == 0;
175 }
176 
177 static void
178 add_file_to_keep_list(char *filename)
179 {
180     FILE *keepf;
181     char symbol[1024];
182     int len;
183 
184     if((keepf = fopen(filename, "r")) == NULL) {
185 	perror(filename);
186 	usage();
187     }
188 
189     while(fgets(symbol, sizeof(symbol), keepf)) {
190 	len = strlen(symbol);
191 	if(len && symbol[len-1] == '\n')
192 	    symbol[len-1] = '\0';
193 
194 	add_to_keep_list(symbol);
195     }
196     fclose(keepf);
197 }
198 
199 /* ---------------------------- */
200 
201 static struct {
202 	const char *name;
203 	int	(*check)(int, const char *);	/* 1 if match, zero if not */
204 	int	(*hide)(int, const char *);	/* non-zero if error */
205 } exec_formats[] = {
206 #ifdef NLIST_ELF64
207 	{	"ELF64",	check_elf64,	hide_elf64,	},
208 #endif
209 };
210 
211 static int
212 hide_syms(const char *filename)
213 {
214 	int fd, i, n, rv;
215 
216 	fd = open(filename, O_RDWR, 0);
217 	if (fd == -1) {
218 		perror(filename);
219 		return 1;
220 	}
221 
222 	rv = 0;
223 
224         n = sizeof exec_formats / sizeof exec_formats[0];
225         for (i = 0; i < n; i++) {
226 		if (lseek(fd, 0, SEEK_SET) != 0) {
227 			perror(filename);
228 			goto err;
229 		}
230                 if ((*exec_formats[i].check)(fd, filename) != 0)
231                         break;
232 	}
233 	if (i == n) {
234 		fprintf(stderr, "%s: unknown executable format\n", filename);
235 		goto err;
236 	}
237 
238 	if (verbose)
239 		fprintf(stderr, "%s is an %s binary\n", filename,
240 		    exec_formats[i].name);
241 
242 	if (lseek(fd, 0, SEEK_SET) != 0) {
243 		perror(filename);
244 		goto err;
245 	}
246 	rv = (*exec_formats[i].hide)(fd, filename);
247 
248 out:
249 	close (fd);
250 	return (rv);
251 
252 err:
253 	rv = 1;
254 	goto out;
255 }
256