| 1 | /* ELF strtab with GC and suffix merging support. |
| 2 | Copyright (C) 2001-2020 Free Software Foundation, Inc. |
| 3 | Written by Jakub Jelinek <jakub@redhat.com>. |
| 4 | |
| 5 | This file is part of BFD, the Binary File Descriptor library. |
| 6 | |
| 7 | This program is free software; you can redistribute it and/or modify |
| 8 | it under the terms of the GNU General Public License as published by |
| 9 | the Free Software Foundation; either version 3 of the License, or |
| 10 | (at your option) any later version. |
| 11 | |
| 12 | This program is distributed in the hope that it will be useful, |
| 13 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 14 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 15 | GNU General Public License for more details. |
| 16 | |
| 17 | You should have received a copy of the GNU General Public License |
| 18 | along with this program; if not, write to the Free Software |
| 19 | Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, |
| 20 | MA 02110-1301, USA. */ |
| 21 | |
| 22 | #include "sysdep.h" |
| 23 | #include "bfd.h" |
| 24 | #include "libbfd.h" |
| 25 | #include "elf-bfd.h" |
| 26 | #include "hashtab.h" |
| 27 | #include "libiberty.h" |
| 28 | |
| 29 | /* An entry in the strtab hash table. */ |
| 30 | |
| 31 | struct elf_strtab_hash_entry |
| 32 | { |
| 33 | struct bfd_hash_entry root; |
| 34 | /* Length of this entry. This includes the zero terminator. */ |
| 35 | int len; |
| 36 | unsigned int refcount; |
| 37 | union { |
| 38 | /* Index within the merged section. */ |
| 39 | bfd_size_type index; |
| 40 | /* Entry this is a suffix of (if len < 0). */ |
| 41 | struct elf_strtab_hash_entry *suffix; |
| 42 | } u; |
| 43 | }; |
| 44 | |
| 45 | /* The strtab hash table. */ |
| 46 | |
| 47 | struct elf_strtab_hash |
| 48 | { |
| 49 | struct bfd_hash_table table; |
| 50 | /* Next available index. */ |
| 51 | size_t size; |
| 52 | /* Number of array entries alloced. */ |
| 53 | size_t alloced; |
| 54 | /* Final strtab size. */ |
| 55 | bfd_size_type sec_size; |
| 56 | /* Array of pointers to strtab entries. */ |
| 57 | struct elf_strtab_hash_entry **array; |
| 58 | }; |
| 59 | |
| 60 | /* Routine to create an entry in a section merge hashtab. */ |
| 61 | |
| 62 | static struct bfd_hash_entry * |
| 63 | elf_strtab_hash_newfunc (struct bfd_hash_entry *entry, |
| 64 | struct bfd_hash_table *table, |
| 65 | const char *string) |
| 66 | { |
| 67 | /* Allocate the structure if it has not already been allocated by a |
| 68 | subclass. */ |
| 69 | if (entry == NULL) |
| 70 | entry = (struct bfd_hash_entry *) |
| 71 | bfd_hash_allocate (table, sizeof (struct elf_strtab_hash_entry)); |
| 72 | if (entry == NULL) |
| 73 | return NULL; |
| 74 | |
| 75 | /* Call the allocation method of the superclass. */ |
| 76 | entry = bfd_hash_newfunc (entry, table, string); |
| 77 | |
| 78 | if (entry) |
| 79 | { |
| 80 | /* Initialize the local fields. */ |
| 81 | struct elf_strtab_hash_entry *ret; |
| 82 | |
| 83 | ret = (struct elf_strtab_hash_entry *) entry; |
| 84 | ret->u.index = -1; |
| 85 | ret->refcount = 0; |
| 86 | ret->len = 0; |
| 87 | } |
| 88 | |
| 89 | return entry; |
| 90 | } |
| 91 | |
| 92 | /* Create a new hash table. */ |
| 93 | |
| 94 | struct elf_strtab_hash * |
| 95 | _bfd_elf_strtab_init (void) |
| 96 | { |
| 97 | struct elf_strtab_hash *table; |
| 98 | bfd_size_type amt = sizeof (struct elf_strtab_hash); |
| 99 | |
| 100 | table = (struct elf_strtab_hash *) bfd_malloc (amt); |
| 101 | if (table == NULL) |
| 102 | return NULL; |
| 103 | |
| 104 | if (!bfd_hash_table_init (&table->table, elf_strtab_hash_newfunc, |
| 105 | sizeof (struct elf_strtab_hash_entry))) |
| 106 | { |
| 107 | free (table); |
| 108 | return NULL; |
| 109 | } |
| 110 | |
| 111 | table->sec_size = 0; |
| 112 | table->size = 1; |
| 113 | table->alloced = 64; |
| 114 | amt = sizeof (struct elf_strtab_hasn_entry *); |
| 115 | table->array = ((struct elf_strtab_hash_entry **) |
| 116 | bfd_malloc (table->alloced * amt)); |
| 117 | if (table->array == NULL) |
| 118 | { |
| 119 | free (table); |
| 120 | return NULL; |
| 121 | } |
| 122 | |
| 123 | table->array[0] = NULL; |
| 124 | |
| 125 | return table; |
| 126 | } |
| 127 | |
| 128 | /* Free a strtab. */ |
| 129 | |
| 130 | void |
| 131 | _bfd_elf_strtab_free (struct elf_strtab_hash *tab) |
| 132 | { |
| 133 | bfd_hash_table_free (&tab->table); |
| 134 | free (tab->array); |
| 135 | free (tab); |
| 136 | } |
| 137 | |
| 138 | /* Get the index of an entity in a hash table, adding it if it is not |
| 139 | already present. */ |
| 140 | |
| 141 | size_t |
| 142 | _bfd_elf_strtab_add (struct elf_strtab_hash *tab, |
| 143 | const char *str, |
| 144 | bfd_boolean copy) |
| 145 | { |
| 146 | register struct elf_strtab_hash_entry *entry; |
| 147 | |
| 148 | /* We handle this specially, since we don't want to do refcounting |
| 149 | on it. */ |
| 150 | if (*str == '\0') |
| 151 | return 0; |
| 152 | |
| 153 | BFD_ASSERT (tab->sec_size == 0); |
| 154 | entry = (struct elf_strtab_hash_entry *) |
| 155 | bfd_hash_lookup (&tab->table, str, TRUE, copy); |
| 156 | |
| 157 | if (entry == NULL) |
| 158 | return (size_t) -1; |
| 159 | |
| 160 | entry->refcount++; |
| 161 | if (entry->len == 0) |
| 162 | { |
| 163 | entry->len = strlen (str) + 1; |
| 164 | /* 2G strings lose. */ |
| 165 | BFD_ASSERT (entry->len > 0); |
| 166 | if (tab->size == tab->alloced) |
| 167 | { |
| 168 | bfd_size_type amt = sizeof (struct elf_strtab_hash_entry *); |
| 169 | tab->alloced *= 2; |
| 170 | tab->array = (struct elf_strtab_hash_entry **) |
| 171 | bfd_realloc_or_free (tab->array, tab->alloced * amt); |
| 172 | if (tab->array == NULL) |
| 173 | return (size_t) -1; |
| 174 | } |
| 175 | |
| 176 | entry->u.index = tab->size++; |
| 177 | tab->array[entry->u.index] = entry; |
| 178 | } |
| 179 | return entry->u.index; |
| 180 | } |
| 181 | |
| 182 | void |
| 183 | _bfd_elf_strtab_addref (struct elf_strtab_hash *tab, size_t idx) |
| 184 | { |
| 185 | if (idx == 0 || idx == (size_t) -1) |
| 186 | return; |
| 187 | BFD_ASSERT (tab->sec_size == 0); |
| 188 | BFD_ASSERT (idx < tab->size); |
| 189 | ++tab->array[idx]->refcount; |
| 190 | } |
| 191 | |
| 192 | void |
| 193 | _bfd_elf_strtab_delref (struct elf_strtab_hash *tab, size_t idx) |
| 194 | { |
| 195 | if (idx == 0 || idx == (size_t) -1) |
| 196 | return; |
| 197 | BFD_ASSERT (tab->sec_size == 0); |
| 198 | BFD_ASSERT (idx < tab->size); |
| 199 | BFD_ASSERT (tab->array[idx]->refcount > 0); |
| 200 | --tab->array[idx]->refcount; |
| 201 | } |
| 202 | |
| 203 | unsigned int |
| 204 | _bfd_elf_strtab_refcount (struct elf_strtab_hash *tab, size_t idx) |
| 205 | { |
| 206 | return tab->array[idx]->refcount; |
| 207 | } |
| 208 | |
| 209 | void |
| 210 | _bfd_elf_strtab_clear_all_refs (struct elf_strtab_hash *tab) |
| 211 | { |
| 212 | size_t idx; |
| 213 | |
| 214 | for (idx = 1; idx < tab->size; idx++) |
| 215 | tab->array[idx]->refcount = 0; |
| 216 | } |
| 217 | |
| 218 | /* Save strtab refcounts prior to adding --as-needed library. */ |
| 219 | |
| 220 | struct strtab_save |
| 221 | { |
| 222 | size_t size; |
| 223 | unsigned int refcount[1]; |
| 224 | }; |
| 225 | |
| 226 | void * |
| 227 | _bfd_elf_strtab_save (struct elf_strtab_hash *tab) |
| 228 | { |
| 229 | struct strtab_save *save; |
| 230 | size_t idx, size; |
| 231 | |
| 232 | size = sizeof (*save) + (tab->size - 1) * sizeof (save->refcount[0]); |
| 233 | save = bfd_malloc (size); |
| 234 | if (save == NULL) |
| 235 | return save; |
| 236 | |
| 237 | save->size = tab->size; |
| 238 | for (idx = 1; idx < tab->size; idx++) |
| 239 | save->refcount[idx] = tab->array[idx]->refcount; |
| 240 | return save; |
| 241 | } |
| 242 | |
| 243 | /* Restore strtab refcounts on finding --as-needed library not needed. */ |
| 244 | |
| 245 | void |
| 246 | _bfd_elf_strtab_restore (struct elf_strtab_hash *tab, void *buf) |
| 247 | { |
| 248 | size_t idx, curr_size = tab->size; |
| 249 | struct strtab_save *save = (struct strtab_save *) buf; |
| 250 | |
| 251 | BFD_ASSERT (tab->sec_size == 0); |
| 252 | BFD_ASSERT (save->size <= curr_size); |
| 253 | tab->size = save->size; |
| 254 | for (idx = 1; idx < save->size; ++idx) |
| 255 | tab->array[idx]->refcount = save->refcount[idx]; |
| 256 | |
| 257 | for (; idx < curr_size; ++idx) |
| 258 | { |
| 259 | /* We don't remove entries from the hash table, just set their |
| 260 | REFCOUNT to zero. Setting LEN zero will result in the size |
| 261 | growing if the entry is added again. See _bfd_elf_strtab_add. */ |
| 262 | tab->array[idx]->refcount = 0; |
| 263 | tab->array[idx]->len = 0; |
| 264 | } |
| 265 | } |
| 266 | |
| 267 | bfd_size_type |
| 268 | _bfd_elf_strtab_size (struct elf_strtab_hash *tab) |
| 269 | { |
| 270 | return tab->sec_size ? tab->sec_size : tab->size; |
| 271 | } |
| 272 | |
| 273 | bfd_size_type |
| 274 | _bfd_elf_strtab_len (struct elf_strtab_hash *tab) |
| 275 | { |
| 276 | return tab->size; |
| 277 | } |
| 278 | |
| 279 | bfd_size_type |
| 280 | _bfd_elf_strtab_offset (struct elf_strtab_hash *tab, size_t idx) |
| 281 | { |
| 282 | struct elf_strtab_hash_entry *entry; |
| 283 | |
| 284 | if (idx == 0) |
| 285 | return 0; |
| 286 | BFD_ASSERT (idx < tab->size); |
| 287 | BFD_ASSERT (tab->sec_size); |
| 288 | entry = tab->array[idx]; |
| 289 | BFD_ASSERT (entry->refcount > 0); |
| 290 | entry->refcount--; |
| 291 | return tab->array[idx]->u.index; |
| 292 | } |
| 293 | |
| 294 | const char * |
| 295 | _bfd_elf_strtab_str (struct elf_strtab_hash *tab, size_t idx, |
| 296 | bfd_size_type *offset) |
| 297 | { |
| 298 | if (idx == 0) |
| 299 | return 0; |
| 300 | BFD_ASSERT (idx < tab->size); |
| 301 | BFD_ASSERT (tab->sec_size); |
| 302 | if (offset) |
| 303 | *offset = tab->array[idx]->u.index; |
| 304 | return tab->array[idx]->root.string; |
| 305 | } |
| 306 | |
| 307 | bfd_boolean |
| 308 | _bfd_elf_strtab_emit (register bfd *abfd, struct elf_strtab_hash *tab) |
| 309 | { |
| 310 | bfd_size_type off = 1; |
| 311 | size_t i; |
| 312 | |
| 313 | if (bfd_bwrite ("", 1, abfd) != 1) |
| 314 | return FALSE; |
| 315 | |
| 316 | for (i = 1; i < tab->size; ++i) |
| 317 | { |
| 318 | register const char *str; |
| 319 | register unsigned int len; |
| 320 | |
| 321 | BFD_ASSERT (tab->array[i]->refcount == 0); |
| 322 | len = tab->array[i]->len; |
| 323 | if ((int) len < 0) |
| 324 | continue; |
| 325 | |
| 326 | str = tab->array[i]->root.string; |
| 327 | if (bfd_bwrite (str, len, abfd) != len) |
| 328 | return FALSE; |
| 329 | |
| 330 | off += len; |
| 331 | } |
| 332 | |
| 333 | BFD_ASSERT (off == tab->sec_size); |
| 334 | return TRUE; |
| 335 | } |
| 336 | |
| 337 | /* Compare two elf_strtab_hash_entry structures. Called via qsort. |
| 338 | Won't ever return zero as all entries differ, so there is no issue |
| 339 | with qsort stability here. */ |
| 340 | |
| 341 | static int |
| 342 | strrevcmp (const void *a, const void *b) |
| 343 | { |
| 344 | struct elf_strtab_hash_entry *A = *(struct elf_strtab_hash_entry **) a; |
| 345 | struct elf_strtab_hash_entry *B = *(struct elf_strtab_hash_entry **) b; |
| 346 | unsigned int lenA = A->len; |
| 347 | unsigned int lenB = B->len; |
| 348 | const unsigned char *s = (const unsigned char *) A->root.string + lenA - 1; |
| 349 | const unsigned char *t = (const unsigned char *) B->root.string + lenB - 1; |
| 350 | int l = lenA < lenB ? lenA : lenB; |
| 351 | |
| 352 | while (l) |
| 353 | { |
| 354 | if (*s != *t) |
| 355 | return (int) *s - (int) *t; |
| 356 | s--; |
| 357 | t--; |
| 358 | l--; |
| 359 | } |
| 360 | return lenA - lenB; |
| 361 | } |
| 362 | |
| 363 | static inline int |
| 364 | is_suffix (const struct elf_strtab_hash_entry *A, |
| 365 | const struct elf_strtab_hash_entry *B) |
| 366 | { |
| 367 | if (A->len <= B->len) |
| 368 | /* B cannot be a suffix of A unless A is equal to B, which is guaranteed |
| 369 | not to be equal by the hash table. */ |
| 370 | return 0; |
| 371 | |
| 372 | return memcmp (A->root.string + (A->len - B->len), |
| 373 | B->root.string, B->len - 1) == 0; |
| 374 | } |
| 375 | |
| 376 | /* This function assigns final string table offsets for used strings, |
| 377 | merging strings matching suffixes of longer strings if possible. */ |
| 378 | |
| 379 | void |
| 380 | _bfd_elf_strtab_finalize (struct elf_strtab_hash *tab) |
| 381 | { |
| 382 | struct elf_strtab_hash_entry **array, **a, *e; |
| 383 | bfd_size_type amt, sec_size; |
| 384 | size_t size, i; |
| 385 | |
| 386 | /* Sort the strings by suffix and length. */ |
| 387 | amt = tab->size; |
| 388 | amt *= sizeof (struct elf_strtab_hash_entry *); |
| 389 | array = (struct elf_strtab_hash_entry **) bfd_malloc (amt); |
| 390 | if (array == NULL) |
| 391 | goto alloc_failure; |
| 392 | |
| 393 | for (i = 1, a = array; i < tab->size; ++i) |
| 394 | { |
| 395 | e = tab->array[i]; |
| 396 | if (e->refcount) |
| 397 | { |
| 398 | *a++ = e; |
| 399 | /* Adjust the length to not include the zero terminator. */ |
| 400 | e->len -= 1; |
| 401 | } |
| 402 | else |
| 403 | e->len = 0; |
| 404 | } |
| 405 | |
| 406 | size = a - array; |
| 407 | if (size != 0) |
| 408 | { |
| 409 | qsort (array, size, sizeof (struct elf_strtab_hash_entry *), strrevcmp); |
| 410 | |
| 411 | /* Loop over the sorted array and merge suffixes. Start from the |
| 412 | end because we want eg. |
| 413 | |
| 414 | s1 -> "d" |
| 415 | s2 -> "bcd" |
| 416 | s3 -> "abcd" |
| 417 | |
| 418 | to end up as |
| 419 | |
| 420 | s3 -> "abcd" |
| 421 | s2 _____^ |
| 422 | s1 _______^ |
| 423 | |
| 424 | ie. we don't want s1 pointing into the old s2. */ |
| 425 | e = *--a; |
| 426 | e->len += 1; |
| 427 | while (--a >= array) |
| 428 | { |
| 429 | struct elf_strtab_hash_entry *cmp = *a; |
| 430 | |
| 431 | cmp->len += 1; |
| 432 | if (is_suffix (e, cmp)) |
| 433 | { |
| 434 | cmp->u.suffix = e; |
| 435 | cmp->len = -cmp->len; |
| 436 | } |
| 437 | else |
| 438 | e = cmp; |
| 439 | } |
| 440 | } |
| 441 | |
| 442 | alloc_failure: |
| 443 | if (array) |
| 444 | free (array); |
| 445 | |
| 446 | /* Assign positions to the strings we want to keep. */ |
| 447 | sec_size = 1; |
| 448 | for (i = 1; i < tab->size; ++i) |
| 449 | { |
| 450 | e = tab->array[i]; |
| 451 | if (e->refcount && e->len > 0) |
| 452 | { |
| 453 | e->u.index = sec_size; |
| 454 | sec_size += e->len; |
| 455 | } |
| 456 | } |
| 457 | |
| 458 | tab->sec_size = sec_size; |
| 459 | |
| 460 | /* Adjust the rest. */ |
| 461 | for (i = 1; i < tab->size; ++i) |
| 462 | { |
| 463 | e = tab->array[i]; |
| 464 | if (e->refcount && e->len < 0) |
| 465 | e->u.index = e->u.suffix->u.index + (e->u.suffix->len + e->len); |
| 466 | } |
| 467 | } |