| 1 | /* CTF string table management. |
| 2 | Copyright (C) 2019 Free Software Foundation, Inc. |
| 3 | |
| 4 | This file is part of libctf. |
| 5 | |
| 6 | libctf is free software; you can redistribute it and/or modify it under |
| 7 | the terms of the GNU General Public License as published by the Free |
| 8 | Software Foundation; either version 3, or (at your option) any later |
| 9 | version. |
| 10 | |
| 11 | This program is distributed in the hope that it will be useful, but |
| 12 | WITHOUT ANY WARRANTY; without even the implied warranty of |
| 13 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. |
| 14 | See the GNU General Public License for more details. |
| 15 | |
| 16 | You should have received a copy of the GNU General Public License |
| 17 | along with this program; see the file COPYING. If not see |
| 18 | <http://www.gnu.org/licenses/>. */ |
| 19 | |
| 20 | #include <ctf-impl.h> |
| 21 | #include <string.h> |
| 22 | |
| 23 | /* Convert an encoded CTF string name into a pointer to a C string, using an |
| 24 | explicit internal strtab rather than the fp-based one. */ |
| 25 | const char * |
| 26 | ctf_strraw_explicit (ctf_file_t *fp, uint32_t name, ctf_strs_t *strtab) |
| 27 | { |
| 28 | ctf_strs_t *ctsp = &fp->ctf_str[CTF_NAME_STID (name)]; |
| 29 | |
| 30 | if ((CTF_NAME_STID (name) == CTF_STRTAB_0) && (strtab != NULL)) |
| 31 | ctsp = strtab; |
| 32 | |
| 33 | /* If this name is in the external strtab, and there is a synthetic strtab, |
| 34 | use it in preference. */ |
| 35 | |
| 36 | if (CTF_NAME_STID (name) == CTF_STRTAB_1 |
| 37 | && fp->ctf_syn_ext_strtab != NULL) |
| 38 | return ctf_dynhash_lookup (fp->ctf_syn_ext_strtab, |
| 39 | (void *) (uintptr_t) name); |
| 40 | |
| 41 | if (ctsp->cts_strs != NULL && CTF_NAME_OFFSET (name) < ctsp->cts_len) |
| 42 | return (ctsp->cts_strs + CTF_NAME_OFFSET (name)); |
| 43 | |
| 44 | /* String table not loaded or corrupt offset. */ |
| 45 | return NULL; |
| 46 | } |
| 47 | |
| 48 | /* Convert an encoded CTF string name into a pointer to a C string by looking |
| 49 | up the appropriate string table buffer and then adding the offset. */ |
| 50 | const char * |
| 51 | ctf_strraw (ctf_file_t *fp, uint32_t name) |
| 52 | { |
| 53 | return ctf_strraw_explicit (fp, name, NULL); |
| 54 | } |
| 55 | |
| 56 | /* Return a guaranteed-non-NULL pointer to the string with the given CTF |
| 57 | name. */ |
| 58 | const char * |
| 59 | ctf_strptr (ctf_file_t *fp, uint32_t name) |
| 60 | { |
| 61 | const char *s = ctf_strraw (fp, name); |
| 62 | return (s != NULL ? s : "(?)"); |
| 63 | } |
| 64 | |
| 65 | /* Remove all refs to a given atom. */ |
| 66 | static void |
| 67 | ctf_str_purge_atom_refs (ctf_str_atom_t *atom) |
| 68 | { |
| 69 | ctf_str_atom_ref_t *ref, *next; |
| 70 | |
| 71 | for (ref = ctf_list_next (&atom->csa_refs); ref != NULL; ref = next) |
| 72 | { |
| 73 | next = ctf_list_next (ref); |
| 74 | ctf_list_delete (&atom->csa_refs, ref); |
| 75 | ctf_free (ref); |
| 76 | } |
| 77 | } |
| 78 | |
| 79 | /* Free an atom (only called on ctf_close().) */ |
| 80 | static void |
| 81 | ctf_str_free_atom (void *a) |
| 82 | { |
| 83 | ctf_str_atom_t *atom = a; |
| 84 | |
| 85 | ctf_str_purge_atom_refs (atom); |
| 86 | ctf_free (atom); |
| 87 | } |
| 88 | |
| 89 | /* Create the atoms table. There is always at least one atom in it, the null |
| 90 | string. */ |
| 91 | int |
| 92 | ctf_str_create_atoms (ctf_file_t *fp) |
| 93 | { |
| 94 | fp->ctf_str_atoms = ctf_dynhash_create (ctf_hash_string, ctf_hash_eq_string, |
| 95 | ctf_free, ctf_str_free_atom); |
| 96 | if (fp->ctf_str_atoms == NULL) |
| 97 | return -ENOMEM; |
| 98 | |
| 99 | ctf_str_add (fp, ""); |
| 100 | return 0; |
| 101 | } |
| 102 | |
| 103 | /* Destroy the atoms table. */ |
| 104 | void |
| 105 | ctf_str_free_atoms (ctf_file_t *fp) |
| 106 | { |
| 107 | ctf_dynhash_destroy (fp->ctf_str_atoms); |
| 108 | } |
| 109 | |
| 110 | /* Add a string to the atoms table, copying the passed-in string. Return the |
| 111 | atom added. Return NULL only when out of memory (and do not touch the |
| 112 | passed-in string in that case). Possibly augment the ref list with the |
| 113 | passed-in ref. */ |
| 114 | static ctf_str_atom_t * |
| 115 | ctf_str_add_ref_internal (ctf_file_t *fp, const char *str, |
| 116 | int add_ref, uint32_t *ref) |
| 117 | { |
| 118 | char *newstr = NULL; |
| 119 | ctf_str_atom_t *atom = NULL; |
| 120 | ctf_str_atom_ref_t *aref = NULL; |
| 121 | |
| 122 | atom = ctf_dynhash_lookup (fp->ctf_str_atoms, str); |
| 123 | |
| 124 | if (add_ref) |
| 125 | { |
| 126 | if ((aref = ctf_alloc (sizeof (struct ctf_str_atom_ref))) == NULL) |
| 127 | return NULL; |
| 128 | aref->caf_ref = ref; |
| 129 | } |
| 130 | |
| 131 | if (atom) |
| 132 | { |
| 133 | if (add_ref) |
| 134 | { |
| 135 | ctf_list_append (&atom->csa_refs, aref); |
| 136 | fp->ctf_str_num_refs++; |
| 137 | } |
| 138 | return atom; |
| 139 | } |
| 140 | |
| 141 | if ((atom = ctf_alloc (sizeof (struct ctf_str_atom))) == NULL) |
| 142 | goto oom; |
| 143 | memset (atom, 0, sizeof (struct ctf_str_atom)); |
| 144 | |
| 145 | if ((newstr = ctf_strdup (str)) == NULL) |
| 146 | goto oom; |
| 147 | |
| 148 | if (ctf_dynhash_insert (fp->ctf_str_atoms, newstr, atom) < 0) |
| 149 | goto oom; |
| 150 | |
| 151 | atom->csa_str = newstr; |
| 152 | atom->csa_snapshot_id = fp->ctf_snapshots; |
| 153 | if (add_ref) |
| 154 | { |
| 155 | ctf_list_append (&atom->csa_refs, aref); |
| 156 | fp->ctf_str_num_refs++; |
| 157 | } |
| 158 | return atom; |
| 159 | |
| 160 | oom: |
| 161 | ctf_free (atom); |
| 162 | ctf_free (aref); |
| 163 | ctf_free (newstr); |
| 164 | return NULL; |
| 165 | } |
| 166 | |
| 167 | /* Add a string to the atoms table and return it, without augmenting the ref |
| 168 | list for this string. */ |
| 169 | const char * |
| 170 | ctf_str_add (ctf_file_t *fp, const char *str) |
| 171 | { |
| 172 | ctf_str_atom_t *atom; |
| 173 | if (!str) |
| 174 | return NULL; |
| 175 | |
| 176 | atom = ctf_str_add_ref_internal (fp, str, FALSE, 0); |
| 177 | if (!atom) |
| 178 | return NULL; |
| 179 | |
| 180 | return atom->csa_str; |
| 181 | } |
| 182 | |
| 183 | /* Like ctf_str_add(), but additionally augment the atom's refs list with the |
| 184 | passed-in ref, whether or not the string is already present. There is no |
| 185 | attempt to deduplicate the refs list (but duplicates are harmless). */ |
| 186 | const char * |
| 187 | ctf_str_add_ref (ctf_file_t *fp, const char *str, uint32_t *ref) |
| 188 | { |
| 189 | ctf_str_atom_t *atom; |
| 190 | if (!str) |
| 191 | return NULL; |
| 192 | |
| 193 | atom = ctf_str_add_ref_internal (fp, str, TRUE, ref); |
| 194 | if (!atom) |
| 195 | return NULL; |
| 196 | |
| 197 | return atom->csa_str; |
| 198 | } |
| 199 | |
| 200 | /* Add an external strtab reference at OFFSET. */ |
| 201 | const char * |
| 202 | ctf_str_add_external (ctf_file_t *fp, const char *str, uint32_t offset) |
| 203 | { |
| 204 | ctf_str_atom_t *atom; |
| 205 | if (!str) |
| 206 | return NULL; |
| 207 | |
| 208 | atom = ctf_str_add_ref_internal (fp, str, FALSE, 0); |
| 209 | if (!atom) |
| 210 | return NULL; |
| 211 | |
| 212 | atom->csa_offset = CTF_SET_STID (offset, CTF_STRTAB_1); |
| 213 | return atom->csa_str; |
| 214 | } |
| 215 | |
| 216 | /* A ctf_dynhash_iter_remove() callback that removes atoms later than a given |
| 217 | snapshot ID. */ |
| 218 | static int |
| 219 | ctf_str_rollback_atom (void *key _libctf_unused_, void *value, void *arg) |
| 220 | { |
| 221 | ctf_str_atom_t *atom = (ctf_str_atom_t *) value; |
| 222 | ctf_snapshot_id_t *id = (ctf_snapshot_id_t *) arg; |
| 223 | |
| 224 | return (atom->csa_snapshot_id > id->snapshot_id); |
| 225 | } |
| 226 | |
| 227 | /* Roll back, deleting all atoms created after a particular ID. */ |
| 228 | void |
| 229 | ctf_str_rollback (ctf_file_t *fp, ctf_snapshot_id_t id) |
| 230 | { |
| 231 | ctf_dynhash_iter_remove (fp->ctf_str_atoms, ctf_str_rollback_atom, &id); |
| 232 | } |
| 233 | |
| 234 | /* An adaptor around ctf_purge_atom_refs. */ |
| 235 | static void |
| 236 | ctf_str_purge_one_atom_refs (void *key _libctf_unused_, void *value, |
| 237 | void *arg _libctf_unused_) |
| 238 | { |
| 239 | ctf_str_atom_t *atom = (ctf_str_atom_t *) value; |
| 240 | ctf_str_purge_atom_refs (atom); |
| 241 | } |
| 242 | |
| 243 | /* Remove all the recorded refs from the atoms table. */ |
| 244 | void |
| 245 | ctf_str_purge_refs (ctf_file_t *fp) |
| 246 | { |
| 247 | if (fp->ctf_str_num_refs > 0) |
| 248 | ctf_dynhash_iter (fp->ctf_str_atoms, ctf_str_purge_one_atom_refs, NULL); |
| 249 | fp->ctf_str_num_refs = 0; |
| 250 | } |
| 251 | |
| 252 | /* Update a list of refs to the specified value. */ |
| 253 | static void |
| 254 | ctf_str_update_refs (ctf_str_atom_t *refs, uint32_t value) |
| 255 | { |
| 256 | ctf_str_atom_ref_t *ref; |
| 257 | |
| 258 | for (ref = ctf_list_next (&refs->csa_refs); ref != NULL; |
| 259 | ref = ctf_list_next (ref)) |
| 260 | *(ref->caf_ref) = value; |
| 261 | } |
| 262 | |
| 263 | /* State shared across the strtab write process. */ |
| 264 | typedef struct ctf_strtab_write_state |
| 265 | { |
| 266 | /* Strtab we are writing, and the number of strings in it. */ |
| 267 | ctf_strs_writable_t *strtab; |
| 268 | size_t strtab_count; |
| 269 | |
| 270 | /* Pointers to (existing) atoms in the atoms table, for qsorting. */ |
| 271 | ctf_str_atom_t **sorttab; |
| 272 | |
| 273 | /* Loop counter for sorttab population. */ |
| 274 | size_t i; |
| 275 | |
| 276 | /* The null-string atom (skipped during population). */ |
| 277 | ctf_str_atom_t *nullstr; |
| 278 | } ctf_strtab_write_state_t; |
| 279 | |
| 280 | /* Count the number of entries in the strtab, and its length. */ |
| 281 | static void |
| 282 | ctf_str_count_strtab (void *key _libctf_unused_, void *value, |
| 283 | void *arg) |
| 284 | { |
| 285 | ctf_str_atom_t *atom = (ctf_str_atom_t *) value; |
| 286 | ctf_strtab_write_state_t *s = (ctf_strtab_write_state_t *) arg; |
| 287 | |
| 288 | /* We only factor in the length of items that have no offset: |
| 289 | other items are in the external strtab. They still contribute to the |
| 290 | total count, though, because we still have to sort them. */ |
| 291 | if (!atom->csa_offset) |
| 292 | s->strtab->cts_len += strlen (atom->csa_str) + 1; |
| 293 | s->strtab_count++; |
| 294 | } |
| 295 | |
| 296 | /* Populate the sorttab with pointers to the strtab atoms. */ |
| 297 | static void |
| 298 | ctf_str_populate_sorttab (void *key _libctf_unused_, void *value, |
| 299 | void *arg) |
| 300 | { |
| 301 | ctf_str_atom_t *atom = (ctf_str_atom_t *) value; |
| 302 | ctf_strtab_write_state_t *s = (ctf_strtab_write_state_t *) arg; |
| 303 | |
| 304 | /* Skip the null string. */ |
| 305 | if (s->nullstr == atom) |
| 306 | return; |
| 307 | |
| 308 | s->sorttab[s->i++] = atom; |
| 309 | } |
| 310 | |
| 311 | /* Sort the strtab. */ |
| 312 | static int |
| 313 | ctf_str_sort_strtab (const void *a, const void *b) |
| 314 | { |
| 315 | ctf_str_atom_t **one = (ctf_str_atom_t **) a; |
| 316 | ctf_str_atom_t **two = (ctf_str_atom_t **) b; |
| 317 | |
| 318 | return (strcmp ((*one)->csa_str, (*two)->csa_str)); |
| 319 | } |
| 320 | |
| 321 | /* Write out and return a strtab containing all strings with recorded refs, |
| 322 | adjusting the refs to refer to the corresponding string. The returned strtab |
| 323 | may be NULL on error. Also populate the synthetic strtab with mappings from |
| 324 | external strtab offsets to names, so we can look them up with ctf_strptr(). |
| 325 | Only external strtab offsets with references are added. */ |
| 326 | ctf_strs_writable_t |
| 327 | ctf_str_write_strtab (ctf_file_t *fp) |
| 328 | { |
| 329 | ctf_strs_writable_t strtab; |
| 330 | ctf_str_atom_t *nullstr; |
| 331 | uint32_t cur_stroff = 0; |
| 332 | ctf_strtab_write_state_t s; |
| 333 | ctf_str_atom_t **sorttab; |
| 334 | size_t i; |
| 335 | int any_external = 0; |
| 336 | |
| 337 | memset (&strtab, 0, sizeof (struct ctf_strs_writable)); |
| 338 | memset (&s, 0, sizeof (struct ctf_strtab_write_state)); |
| 339 | s.strtab = &strtab; |
| 340 | |
| 341 | nullstr = ctf_dynhash_lookup (fp->ctf_str_atoms, ""); |
| 342 | if (!nullstr) |
| 343 | { |
| 344 | ctf_dprintf ("Internal error: null string not found in strtab.\n"); |
| 345 | strtab.cts_strs = NULL; |
| 346 | return strtab; |
| 347 | } |
| 348 | |
| 349 | ctf_dynhash_iter (fp->ctf_str_atoms, ctf_str_count_strtab, &s); |
| 350 | |
| 351 | ctf_dprintf ("%lu bytes of strings in strtab.\n", |
| 352 | (unsigned long) strtab.cts_len); |
| 353 | |
| 354 | /* Sort the strtab. Force the null string to be first. */ |
| 355 | sorttab = calloc (s.strtab_count, sizeof (ctf_str_atom_t *)); |
| 356 | if (!sorttab) |
| 357 | goto oom; |
| 358 | |
| 359 | sorttab[0] = nullstr; |
| 360 | s.i = 1; |
| 361 | s.sorttab = sorttab; |
| 362 | s.nullstr = nullstr; |
| 363 | ctf_dynhash_iter (fp->ctf_str_atoms, ctf_str_populate_sorttab, &s); |
| 364 | |
| 365 | qsort (&sorttab[1], s.strtab_count - 1, sizeof (ctf_str_atom_t *), |
| 366 | ctf_str_sort_strtab); |
| 367 | |
| 368 | if ((strtab.cts_strs = ctf_alloc (strtab.cts_len)) == NULL) |
| 369 | goto oom_sorttab; |
| 370 | |
| 371 | if (!fp->ctf_syn_ext_strtab) |
| 372 | fp->ctf_syn_ext_strtab = ctf_dynhash_create (ctf_hash_integer, |
| 373 | ctf_hash_eq_integer, |
| 374 | NULL, NULL); |
| 375 | if (!fp->ctf_syn_ext_strtab) |
| 376 | goto oom_strtab; |
| 377 | |
| 378 | /* Update all refs: also update the strtab appropriately. */ |
| 379 | for (i = 0; i < s.strtab_count; i++) |
| 380 | { |
| 381 | if (sorttab[i]->csa_offset) |
| 382 | { |
| 383 | /* External strtab entry: populate the synthetic external strtab. |
| 384 | |
| 385 | This is safe because you cannot ctf_rollback to before the point |
| 386 | when a ctf_update is done, and the strtab is written at ctf_update |
| 387 | time. So any atoms we reference here are sure to stick around |
| 388 | until ctf_file_close. */ |
| 389 | |
| 390 | any_external = 1; |
| 391 | ctf_str_update_refs (sorttab[i], sorttab[i]->csa_offset); |
| 392 | if (ctf_dynhash_insert (fp->ctf_syn_ext_strtab, |
| 393 | (void *) (uintptr_t) sorttab[i]->csa_offset, |
| 394 | (void *) sorttab[i]->csa_str) < 0) |
| 395 | goto oom_strtab; |
| 396 | } |
| 397 | else |
| 398 | { |
| 399 | /* Internal strtab entry: actually add to the string table. */ |
| 400 | |
| 401 | ctf_str_update_refs (sorttab[i], cur_stroff); |
| 402 | strcpy (&strtab.cts_strs[cur_stroff], sorttab[i]->csa_str); |
| 403 | cur_stroff += strlen (sorttab[i]->csa_str) + 1; |
| 404 | } |
| 405 | } |
| 406 | free (sorttab); |
| 407 | |
| 408 | if (!any_external) |
| 409 | { |
| 410 | ctf_dynhash_destroy (fp->ctf_syn_ext_strtab); |
| 411 | fp->ctf_syn_ext_strtab = NULL; |
| 412 | } |
| 413 | |
| 414 | return strtab; |
| 415 | |
| 416 | oom_strtab: |
| 417 | free (strtab.cts_strs); |
| 418 | strtab.cts_strs = NULL; |
| 419 | oom_sorttab: |
| 420 | free (sorttab); |
| 421 | oom: |
| 422 | return strtab; |
| 423 | } |