1 /* $NetBSD: crunchide.c,v 1.13 2009/08/20 17:39:51 he Exp $ */ 2 3 /* 4 * Copyright (c) 1997 Christopher G. Demetriou. All rights reserved. 5 * Copyright (c) 1994 University of Maryland 6 * All Rights Reserved. 7 * 8 * Permission to use, copy, modify, distribute, and sell this software and its 9 * documentation for any purpose is hereby granted without fee, provided that 10 * the above copyright notice appear in all copies and that both that 11 * copyright notice and this permission notice appear in supporting 12 * documentation, and that the name of U.M. not be used in advertising or 13 * publicity pertaining to distribution of the software without specific, 14 * written prior permission. U.M. makes no representations about the 15 * suitability of this software for any purpose. It is provided "as is" 16 * without express or implied warranty. 17 * 18 * U.M. DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL U.M. 20 * BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES 21 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION 22 * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN 23 * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. 24 * 25 * Author: James da Silva, Systems Design and Analysis Group 26 * Computer Science Department 27 * University of Maryland at College Park 28 */ 29 30 /* 31 * crunchide.c - tiptoes through a symbol table, hiding all defined 32 * global symbols. Allows the user to supply a "keep list" of symbols 33 * that are not to be hidden. This program relies on the use of the 34 * linker's -dc flag to actually put global bss data into the file's 35 * bss segment (rather than leaving it as undefined "common" data). 36 * 37 * The point of all this is to allow multiple programs to be linked 38 * together without getting multiple-defined errors. 39 * 40 * For example, consider a program "foo.c". It can be linked with a 41 * small stub routine, called "foostub.c", eg: 42 * int foo_main(int argc, char **argv){ return main(argc, argv); } 43 * like so: 44 * cc -c foo.c foostub.c 45 * ld -dc -r foo.o foostub.o -o foo.combined.o 46 * crunchide -k _foo_main foo.combined.o 47 * at this point, foo.combined.o can be linked with another program 48 * and invoked with "foo_main(argc, argv)". foo's main() and any 49 * other globals are hidden and will not conflict with other symbols. 50 * 51 * TODO: 52 * - resolve the theoretical hanging reloc problem (see check_reloc() 53 * below). I have yet to see this problem actually occur in any real 54 * program. In what cases will gcc/gas generate code that needs a 55 * relative reloc from a global symbol, other than PIC? The 56 * solution is to not hide the symbol from the linker in this case, 57 * but to generate some random name for it so that it doesn't link 58 * with anything but holds the place for the reloc. 59 * - arrange that all the BSS segments start at the same address, so 60 * that the final crunched binary BSS size is the max of all the 61 * component programs' BSS sizes, rather than their sum. 62 */ 63 64 #include <sys/cdefs.h> 65 #ifndef lint 66 __RCSID("$NetBSD: crunchide.c,v 1.13 2009/08/20 17:39:51 he Exp $"); 67 #endif 68 69 #include <unistd.h> 70 #include <stdio.h> 71 #include <stdlib.h> 72 #include <string.h> 73 #include <fcntl.h> 74 #include <errno.h> 75 #include <sys/types.h> 76 #include <sys/stat.h> 77 78 #include "extern.h" 79 80 void usage(void); 81 82 void add_to_keep_list(char *symbol); 83 void add_file_to_keep_list(char *filename); 84 85 int hide_syms(const char *filename); 86 87 int verbose; 88 89 int 90 main(int argc, char *argv[]) 91 { 92 int ch, errors; 93 94 setprogname(argv[0]); 95 96 while ((ch = getopt(argc, argv, "k:f:v")) != -1) 97 switch(ch) { 98 case 'k': 99 add_to_keep_list(optarg); 100 break; 101 case 'f': 102 add_file_to_keep_list(optarg); 103 break; 104 case 'v': 105 verbose = 1; 106 break; 107 default: 108 usage(); 109 } 110 111 argc -= optind; 112 argv += optind; 113 114 if (argc == 0) 115 usage(); 116 117 errors = 0; 118 while (argc) { 119 if (hide_syms(*argv)) 120 errors = 1; 121 argc--, argv++; 122 } 123 124 return errors; 125 } 126 127 void 128 usage(void) 129 { 130 fprintf(stderr, 131 "Usage: %s [-k keep-symbol] [-f keep-list-file] object-file\n" 132 "\t\t [object-file ...]\n", 133 getprogname()); 134 exit(1); 135 } 136 137 /* ---------------------------- */ 138 139 struct keep { 140 struct keep *next; 141 char *sym; 142 } *keep_list; 143 144 void 145 add_to_keep_list(char *symbol) 146 { 147 struct keep *newp, *prevp, *curp; 148 int cmp; 149 150 cmp = 0; 151 152 for (curp = keep_list, prevp = NULL; curp; prevp = curp, curp = curp->next) 153 if ((cmp = strcmp(symbol, curp->sym)) <= 0) 154 break; 155 156 if (curp && cmp == 0) 157 return; /* already in table */ 158 159 newp = (struct keep *) malloc(sizeof(struct keep)); 160 if (newp) 161 newp->sym = strdup(symbol); 162 if (newp == NULL || newp->sym == NULL) { 163 fprintf(stderr, "%s: out of memory for keep list\n", getprogname()); 164 exit(1); 165 } 166 167 newp->next = curp; 168 if (prevp) 169 prevp->next = newp; 170 else 171 keep_list = newp; 172 } 173 174 int 175 in_keep_list(const char *symbol) 176 { 177 struct keep *curp; 178 int cmp; 179 180 cmp = 0; 181 182 for (curp = keep_list; curp; curp = curp->next) 183 if((cmp = strcmp(symbol, curp->sym)) <= 0) 184 break; 185 186 return curp && cmp == 0; 187 } 188 189 void 190 add_file_to_keep_list(char *filename) 191 { 192 FILE *keepf; 193 char symbol[1024]; 194 int len; 195 196 if ((keepf = fopen(filename, "r")) == NULL) { 197 perror(filename); 198 usage(); 199 } 200 201 while (fgets(symbol, 1024, keepf)) { 202 len = strlen(symbol); 203 if (len && symbol[len-1] == '\n') 204 symbol[len-1] = '\0'; 205 206 add_to_keep_list(symbol); 207 } 208 fclose(keepf); 209 } 210 211 /* ---------------------------- */ 212 213 struct { 214 const char *name; 215 int (*check)(int, const char *); /* 1 if match, zero if not */ 216 int (*hide)(int, const char *); /* non-zero if error */ 217 } exec_formats[] = { 218 #ifdef NLIST_AOUT 219 { "a.out", check_aout, hide_aout, }, 220 #endif 221 #ifdef NLIST_COFF 222 { "COFF", check_coff, hide_coff, }, 223 #endif 224 #ifdef NLIST_ECOFF 225 { "ECOFF", check_ecoff, hide_ecoff, }, 226 #endif 227 #ifdef NLIST_ELF32 228 { "ELF32", check_elf32, hide_elf32, }, 229 #endif 230 #ifdef NLIST_ELF64 231 { "ELF64", check_elf64, hide_elf64, }, 232 #endif 233 }; 234 235 int 236 hide_syms(const char *filename) 237 { 238 int fd, i, n, rv; 239 240 fd = open(filename, O_RDWR, 0); 241 if (fd == -1) { 242 perror(filename); 243 return 1; 244 } 245 246 rv = 0; 247 248 n = sizeof exec_formats / sizeof exec_formats[0]; 249 for (i = 0; i < n; i++) { 250 if (lseek(fd, 0, SEEK_SET) != 0) { 251 perror(filename); 252 goto err; 253 } 254 if ((*exec_formats[i].check)(fd, filename) != 0) 255 break; 256 } 257 if (i == n) { 258 fprintf(stderr, "%s: unknown executable format\n", filename); 259 goto err; 260 } 261 262 if (verbose) 263 fprintf(stderr, "%s is an %s binary\n", filename, 264 exec_formats[i].name); 265 266 if (lseek(fd, 0, SEEK_SET) != 0) { 267 perror(filename); 268 goto err; 269 } 270 rv = (*exec_formats[i].hide)(fd, filename); 271 272 out: 273 close(fd); 274 return rv; 275 276 err: 277 rv = 1; 278 goto out; 279 } 280