Brian Silverman | 8649792 | 2018-02-10 19:28:39 -0500 | [diff] [blame] | 1 | /* ELF/DWARF string table handling. |
| 2 | Copyright (C) 2000, 2001, 2002, 2005, 2016 Red Hat, Inc. |
| 3 | This file is part of elfutils. |
| 4 | Written by Ulrich Drepper <drepper@redhat.com>, 2000. |
| 5 | |
| 6 | This file is free software; you can redistribute it and/or modify |
| 7 | it under the terms of either |
| 8 | |
| 9 | * the GNU Lesser General Public License as published by the Free |
| 10 | Software Foundation; either version 3 of the License, or (at |
| 11 | your option) any later version |
| 12 | |
| 13 | or |
| 14 | |
| 15 | * the GNU General Public License as published by the Free |
| 16 | Software Foundation; either version 2 of the License, or (at |
| 17 | your option) any later version |
| 18 | |
| 19 | or both in parallel, as here. |
| 20 | |
| 21 | elfutils is distributed in the hope that it will be useful, but |
| 22 | WITHOUT ANY WARRANTY; without even the implied warranty of |
| 23 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 24 | General Public License for more details. |
| 25 | |
| 26 | You should have received copies of the GNU General Public License and |
| 27 | the GNU Lesser General Public License along with this program. If |
| 28 | not, see <http://www.gnu.org/licenses/>. */ |
| 29 | |
| 30 | #ifdef HAVE_CONFIG_H |
| 31 | # include <config.h> |
| 32 | #endif |
| 33 | |
| 34 | #include <assert.h> |
| 35 | #include <inttypes.h> |
| 36 | #include <libelf.h> |
| 37 | #include <stddef.h> |
| 38 | #include <stdlib.h> |
| 39 | #include <string.h> |
| 40 | #include <unistd.h> |
| 41 | |
| 42 | #include "libdwelfP.h" |
| 43 | #include <system.h> |
| 44 | |
| 45 | |
| 46 | struct Dwelf_Strent |
| 47 | { |
| 48 | const char *string; |
| 49 | size_t len; |
| 50 | struct Dwelf_Strent *next; |
| 51 | struct Dwelf_Strent *left; |
| 52 | struct Dwelf_Strent *right; |
| 53 | size_t offset; |
| 54 | char reverse[0]; |
| 55 | }; |
| 56 | |
| 57 | |
| 58 | struct memoryblock |
| 59 | { |
| 60 | struct memoryblock *next; |
| 61 | char memory[0]; |
| 62 | }; |
| 63 | |
| 64 | |
| 65 | struct Dwelf_Strtab |
| 66 | { |
| 67 | struct Dwelf_Strent *root; |
| 68 | struct memoryblock *memory; |
| 69 | char *backp; |
| 70 | size_t left; |
| 71 | size_t total; |
| 72 | bool nullstr; |
| 73 | |
| 74 | struct Dwelf_Strent null; |
| 75 | }; |
| 76 | |
| 77 | |
| 78 | /* Cache for the pagesize. */ |
| 79 | static size_t ps; |
| 80 | /* We correct this value a bit so that `malloc' is not allocating more |
| 81 | than a page. */ |
| 82 | #define MALLOC_OVERHEAD (2 * sizeof (void *)) |
| 83 | |
| 84 | |
| 85 | Dwelf_Strtab * |
| 86 | dwelf_strtab_init (bool nullstr) |
| 87 | { |
| 88 | if (ps == 0) |
| 89 | { |
| 90 | ps = sysconf (_SC_PAGESIZE); |
| 91 | assert (sizeof (struct memoryblock) < ps - MALLOC_OVERHEAD); |
| 92 | } |
| 93 | |
| 94 | Dwelf_Strtab *ret |
| 95 | = (Dwelf_Strtab *) calloc (1, sizeof (struct Dwelf_Strtab)); |
| 96 | if (ret != NULL) |
| 97 | { |
| 98 | ret->nullstr = nullstr; |
| 99 | |
| 100 | if (nullstr) |
| 101 | { |
| 102 | ret->null.len = 1; |
| 103 | ret->null.string = ""; |
| 104 | } |
| 105 | } |
| 106 | |
| 107 | return ret; |
| 108 | } |
| 109 | |
| 110 | |
| 111 | static int |
| 112 | morememory (Dwelf_Strtab *st, size_t len) |
| 113 | { |
| 114 | size_t overhead = offsetof (struct memoryblock, memory); |
| 115 | len += overhead + MALLOC_OVERHEAD; |
| 116 | |
| 117 | /* Allocate nearest multiple of pagesize >= len. */ |
| 118 | len = ((len / ps) + (len % ps != 0)) * ps - MALLOC_OVERHEAD; |
| 119 | |
| 120 | struct memoryblock *newmem = (struct memoryblock *) malloc (len); |
| 121 | if (newmem == NULL) |
| 122 | return 1; |
| 123 | |
| 124 | newmem->next = st->memory; |
| 125 | st->memory = newmem; |
| 126 | st->backp = newmem->memory; |
| 127 | st->left = len - overhead; |
| 128 | |
| 129 | return 0; |
| 130 | } |
| 131 | |
| 132 | |
| 133 | void |
| 134 | dwelf_strtab_free (Dwelf_Strtab *st) |
| 135 | { |
| 136 | struct memoryblock *mb = st->memory; |
| 137 | |
| 138 | while (mb != NULL) |
| 139 | { |
| 140 | void *old = mb; |
| 141 | mb = mb->next; |
| 142 | free (old); |
| 143 | } |
| 144 | |
| 145 | free (st); |
| 146 | } |
| 147 | |
| 148 | |
| 149 | static Dwelf_Strent * |
| 150 | newstring (Dwelf_Strtab *st, const char *str, size_t len) |
| 151 | { |
| 152 | /* Compute the amount of padding needed to make the structure aligned. */ |
| 153 | size_t align = ((__alignof__ (struct Dwelf_Strent) |
| 154 | - (((uintptr_t) st->backp) |
| 155 | & (__alignof__ (struct Dwelf_Strent) - 1))) |
| 156 | & (__alignof__ (struct Dwelf_Strent) - 1)); |
| 157 | |
| 158 | /* Make sure there is enough room in the memory block. */ |
| 159 | if (st->left < align + sizeof (struct Dwelf_Strent) + len) |
| 160 | { |
| 161 | if (morememory (st, sizeof (struct Dwelf_Strent) + len)) |
| 162 | return NULL; |
| 163 | |
| 164 | align = 0; |
| 165 | } |
| 166 | |
| 167 | /* Create the reserved string. */ |
| 168 | Dwelf_Strent *newstr = (Dwelf_Strent *) (st->backp + align); |
| 169 | newstr->string = str; |
| 170 | newstr->len = len; |
| 171 | newstr->next = NULL; |
| 172 | newstr->left = NULL; |
| 173 | newstr->right = NULL; |
| 174 | newstr->offset = 0; |
| 175 | for (int i = len - 2; i >= 0; --i) |
| 176 | newstr->reverse[i] = str[len - 2 - i]; |
| 177 | newstr->reverse[len - 1] = '\0'; |
| 178 | st->backp += align + sizeof (struct Dwelf_Strent) + len; |
| 179 | st->left -= align + sizeof (struct Dwelf_Strent) + len; |
| 180 | |
| 181 | return newstr; |
| 182 | } |
| 183 | |
| 184 | |
| 185 | /* XXX This function should definitely be rewritten to use a balancing |
| 186 | tree algorith (AVL, red-black trees). For now a simple, correct |
| 187 | implementation is enough. */ |
| 188 | static Dwelf_Strent ** |
| 189 | searchstring (Dwelf_Strent **sep, Dwelf_Strent *newstr) |
| 190 | { |
| 191 | /* More strings? */ |
| 192 | if (*sep == NULL) |
| 193 | { |
| 194 | *sep = newstr; |
| 195 | return sep; |
| 196 | } |
| 197 | |
| 198 | /* Compare the strings. */ |
| 199 | int cmpres = memcmp ((*sep)->reverse, newstr->reverse, |
| 200 | MIN ((*sep)->len, newstr->len) - 1); |
| 201 | if (cmpres == 0) |
| 202 | /* We found a matching string. */ |
| 203 | return sep; |
| 204 | else if (cmpres > 0) |
| 205 | return searchstring (&(*sep)->left, newstr); |
| 206 | else |
| 207 | return searchstring (&(*sep)->right, newstr); |
| 208 | } |
| 209 | |
| 210 | |
| 211 | /* Add new string. The actual string is assumed to be permanent. */ |
| 212 | static Dwelf_Strent * |
| 213 | strtab_add (Dwelf_Strtab *st, const char *str, size_t len) |
| 214 | { |
| 215 | /* Make sure all "" strings get offset 0 but only if the table was |
| 216 | created with a special null entry in mind. */ |
| 217 | if (len == 1 && st->null.string != NULL) |
| 218 | return &st->null; |
| 219 | |
| 220 | /* Allocate memory for the new string and its associated information. */ |
| 221 | Dwelf_Strent *newstr = newstring (st, str, len); |
| 222 | if (newstr == NULL) |
| 223 | return NULL; |
| 224 | |
| 225 | /* Search in the array for the place to insert the string. If there |
| 226 | is no string with matching prefix and no string with matching |
| 227 | leading substring, create a new entry. */ |
| 228 | Dwelf_Strent **sep = searchstring (&st->root, newstr); |
| 229 | if (*sep != newstr) |
| 230 | { |
| 231 | /* This is not the same entry. This means we have a prefix match. */ |
| 232 | if ((*sep)->len > newstr->len) |
| 233 | { |
| 234 | /* Check whether we already know this string. */ |
| 235 | for (Dwelf_Strent *subs = (*sep)->next; subs != NULL; |
| 236 | subs = subs->next) |
| 237 | if (subs->len == newstr->len) |
| 238 | { |
| 239 | /* We have an exact match with a substring. Free the memory |
| 240 | we allocated. */ |
| 241 | st->left += st->backp - (char *) newstr; |
| 242 | st->backp = (char *) newstr; |
| 243 | |
| 244 | return subs; |
| 245 | } |
| 246 | |
| 247 | /* We have a new substring. This means we don't need the reverse |
| 248 | string of this entry anymore. */ |
| 249 | st->backp -= newstr->len; |
| 250 | st->left += newstr->len; |
| 251 | |
| 252 | newstr->next = (*sep)->next; |
| 253 | (*sep)->next = newstr; |
| 254 | } |
| 255 | else if ((*sep)->len != newstr->len) |
| 256 | { |
| 257 | /* When we get here it means that the string we are about to |
| 258 | add has a common prefix with a string we already have but |
| 259 | it is longer. In this case we have to put it first. */ |
| 260 | st->total += newstr->len - (*sep)->len; |
| 261 | newstr->next = *sep; |
| 262 | newstr->left = (*sep)->left; |
| 263 | newstr->right = (*sep)->right; |
| 264 | *sep = newstr; |
| 265 | } |
| 266 | else |
| 267 | { |
| 268 | /* We have an exact match. Free the memory we allocated. */ |
| 269 | st->left += st->backp - (char *) newstr; |
| 270 | st->backp = (char *) newstr; |
| 271 | |
| 272 | newstr = *sep; |
| 273 | } |
| 274 | } |
| 275 | else |
| 276 | st->total += newstr->len; |
| 277 | |
| 278 | return newstr; |
| 279 | } |
| 280 | |
| 281 | Dwelf_Strent * |
| 282 | dwelf_strtab_add (Dwelf_Strtab *st, const char *str) |
| 283 | { |
| 284 | return strtab_add (st, str, strlen (str) + 1); |
| 285 | } |
| 286 | |
| 287 | Dwelf_Strent * |
| 288 | dwelf_strtab_add_len (Dwelf_Strtab *st, const char *str, size_t len) |
| 289 | { |
| 290 | return strtab_add (st, str, len); |
| 291 | } |
| 292 | |
| 293 | static void |
| 294 | copystrings (Dwelf_Strent *nodep, char **freep, size_t *offsetp) |
| 295 | { |
| 296 | if (nodep->left != NULL) |
| 297 | copystrings (nodep->left, freep, offsetp); |
| 298 | |
| 299 | /* Process the current node. */ |
| 300 | nodep->offset = *offsetp; |
| 301 | *freep = (char *) mempcpy (*freep, nodep->string, nodep->len); |
| 302 | *offsetp += nodep->len; |
| 303 | |
| 304 | for (Dwelf_Strent *subs = nodep->next; subs != NULL; subs = subs->next) |
| 305 | { |
| 306 | assert (subs->len < nodep->len); |
| 307 | subs->offset = nodep->offset + nodep->len - subs->len; |
| 308 | assert (subs->offset != 0 || subs->string[0] == '\0'); |
| 309 | } |
| 310 | |
| 311 | if (nodep->right != NULL) |
| 312 | copystrings (nodep->right, freep, offsetp); |
| 313 | } |
| 314 | |
| 315 | |
| 316 | Elf_Data * |
| 317 | dwelf_strtab_finalize (Dwelf_Strtab *st, Elf_Data *data) |
| 318 | { |
| 319 | size_t nulllen = st->nullstr ? 1 : 0; |
| 320 | |
| 321 | /* Fill in the information. */ |
| 322 | data->d_buf = malloc (st->total + nulllen); |
| 323 | if (data->d_buf == NULL) |
| 324 | return NULL; |
| 325 | |
| 326 | /* The first byte must always be zero if we created the table with a |
| 327 | null string. */ |
| 328 | if (st->nullstr) |
| 329 | *((char *) data->d_buf) = '\0'; |
| 330 | |
| 331 | data->d_type = ELF_T_BYTE; |
| 332 | data->d_size = st->total + nulllen; |
| 333 | data->d_off = 0; |
| 334 | data->d_align = 1; |
| 335 | data->d_version = EV_CURRENT; |
| 336 | |
| 337 | /* Now run through the tree and add all the string while also updating |
| 338 | the offset members of the elfstrent records. */ |
| 339 | char *endp = (char *) data->d_buf + nulllen; |
| 340 | size_t copylen = nulllen; |
| 341 | if (st->root) |
| 342 | copystrings (st->root, &endp, ©len); |
| 343 | assert (copylen == st->total + nulllen); |
| 344 | |
| 345 | return data; |
| 346 | } |
| 347 | |
| 348 | |
| 349 | size_t |
| 350 | dwelf_strent_off (Dwelf_Strent *se) |
| 351 | { |
| 352 | return se->offset; |
| 353 | } |
| 354 | |
| 355 | |
| 356 | const char * |
| 357 | dwelf_strent_str (Dwelf_Strent *se) |
| 358 | { |
| 359 | return se->string; |
| 360 | } |