1 |
/* ELF strtab with GC and suffix merging support. |
/* ELF strtab with GC and suffix merging support. |
2 |
Copyright 2001, 2002 Free Software Foundation, Inc. |
Copyright 2001, 2002, 2003, 2005 Free Software Foundation, Inc. |
3 |
Written by Jakub Jelinek <jakub@redhat.com>. |
Written by Jakub Jelinek <jakub@redhat.com>. |
4 |
|
|
5 |
This file is part of BFD, the Binary File Descriptor library. |
This file is part of BFD, the Binary File Descriptor library. |
30 |
struct elf_strtab_hash_entry |
struct elf_strtab_hash_entry |
31 |
{ |
{ |
32 |
struct bfd_hash_entry root; |
struct bfd_hash_entry root; |
33 |
/* Length of this entry. */ |
/* Length of this entry. This includes the zero terminator. */ |
34 |
unsigned int len; |
int len; |
35 |
unsigned int refcount; |
unsigned int refcount; |
36 |
union { |
union { |
37 |
/* Index within the merged section. */ |
/* Index within the merged section. */ |
38 |
bfd_size_type index; |
bfd_size_type index; |
39 |
/* Entry this is a suffix of (if len is 0). */ |
/* Entry this is a suffix of (if len < 0). */ |
40 |
struct elf_strtab_hash_entry *suffix; |
struct elf_strtab_hash_entry *suffix; |
|
struct elf_strtab_hash_entry *next; |
|
41 |
} u; |
} u; |
42 |
}; |
}; |
43 |
|
|
56 |
struct elf_strtab_hash_entry **array; |
struct elf_strtab_hash_entry **array; |
57 |
}; |
}; |
58 |
|
|
|
static struct bfd_hash_entry *elf_strtab_hash_newfunc |
|
|
PARAMS ((struct bfd_hash_entry *, struct bfd_hash_table *, const char *)); |
|
|
static int cmplengthentry PARAMS ((const PTR, const PTR)); |
|
|
static int last4_eq PARAMS ((const PTR, const PTR)); |
|
|
|
|
59 |
/* Routine to create an entry in a section merge hashtab. */ |
/* Routine to create an entry in a section merge hashtab. */ |
60 |
|
|
61 |
static struct bfd_hash_entry * |
static struct bfd_hash_entry * |
62 |
elf_strtab_hash_newfunc (entry, table, string) |
elf_strtab_hash_newfunc (struct bfd_hash_entry *entry, |
63 |
struct bfd_hash_entry *entry; |
struct bfd_hash_table *table, |
64 |
struct bfd_hash_table *table; |
const char *string) |
|
const char *string; |
|
65 |
{ |
{ |
|
struct elf_strtab_hash_entry *ret = (struct elf_strtab_hash_entry *) entry; |
|
|
|
|
66 |
/* Allocate the structure if it has not already been allocated by a |
/* Allocate the structure if it has not already been allocated by a |
67 |
subclass. */ |
subclass. */ |
68 |
if (ret == (struct elf_strtab_hash_entry *) NULL) |
if (entry == NULL) |
69 |
ret = ((struct elf_strtab_hash_entry *) |
entry = bfd_hash_allocate (table, sizeof (struct elf_strtab_hash_entry)); |
70 |
bfd_hash_allocate (table, sizeof (struct elf_strtab_hash_entry))); |
if (entry == NULL) |
|
if (ret == (struct elf_strtab_hash_entry *) NULL) |
|
71 |
return NULL; |
return NULL; |
72 |
|
|
73 |
/* Call the allocation method of the superclass. */ |
/* Call the allocation method of the superclass. */ |
74 |
ret = ((struct elf_strtab_hash_entry *) |
entry = bfd_hash_newfunc (entry, table, string); |
|
bfd_hash_newfunc ((struct bfd_hash_entry *) ret, table, string)); |
|
75 |
|
|
76 |
if (ret) |
if (entry) |
77 |
{ |
{ |
78 |
/* Initialize the local fields. */ |
/* Initialize the local fields. */ |
79 |
|
struct elf_strtab_hash_entry *ret; |
80 |
|
|
81 |
|
ret = (struct elf_strtab_hash_entry *) entry; |
82 |
ret->u.index = -1; |
ret->u.index = -1; |
83 |
ret->refcount = 0; |
ret->refcount = 0; |
84 |
ret->len = 0; |
ret->len = 0; |
85 |
} |
} |
86 |
|
|
87 |
return (struct bfd_hash_entry *)ret; |
return entry; |
88 |
} |
} |
89 |
|
|
90 |
/* Create a new hash table. */ |
/* Create a new hash table. */ |
91 |
|
|
92 |
struct elf_strtab_hash * |
struct elf_strtab_hash * |
93 |
_bfd_elf_strtab_init () |
_bfd_elf_strtab_init (void) |
94 |
{ |
{ |
95 |
struct elf_strtab_hash *table; |
struct elf_strtab_hash *table; |
96 |
bfd_size_type amt = sizeof (struct elf_strtab_hash); |
bfd_size_type amt = sizeof (struct elf_strtab_hash); |
97 |
|
|
98 |
table = (struct elf_strtab_hash *) bfd_malloc (amt); |
table = bfd_malloc (amt); |
99 |
if (table == NULL) |
if (table == NULL) |
100 |
return NULL; |
return NULL; |
101 |
|
|
109 |
table->size = 1; |
table->size = 1; |
110 |
table->alloced = 64; |
table->alloced = 64; |
111 |
amt = sizeof (struct elf_strtab_hasn_entry *); |
amt = sizeof (struct elf_strtab_hasn_entry *); |
112 |
table->array = (struct elf_strtab_hash_entry **) |
table->array = bfd_malloc (table->alloced * amt); |
|
bfd_malloc (table->alloced * amt); |
|
113 |
if (table->array == NULL) |
if (table->array == NULL) |
114 |
{ |
{ |
115 |
free (table); |
free (table); |
124 |
/* Free a strtab. */ |
/* Free a strtab. */ |
125 |
|
|
126 |
void |
void |
127 |
_bfd_elf_strtab_free (tab) |
_bfd_elf_strtab_free (struct elf_strtab_hash *tab) |
|
struct elf_strtab_hash *tab; |
|
128 |
{ |
{ |
129 |
bfd_hash_table_free (&tab->table); |
bfd_hash_table_free (&tab->table); |
130 |
free (tab->array); |
free (tab->array); |
135 |
already present. */ |
already present. */ |
136 |
|
|
137 |
bfd_size_type |
bfd_size_type |
138 |
_bfd_elf_strtab_add (tab, str, copy) |
_bfd_elf_strtab_add (struct elf_strtab_hash *tab, |
139 |
struct elf_strtab_hash *tab; |
const char *str, |
140 |
const char *str; |
bfd_boolean copy) |
|
boolean copy; |
|
141 |
{ |
{ |
142 |
register struct elf_strtab_hash_entry *entry; |
register struct elf_strtab_hash_entry *entry; |
143 |
|
|
148 |
|
|
149 |
BFD_ASSERT (tab->sec_size == 0); |
BFD_ASSERT (tab->sec_size == 0); |
150 |
entry = (struct elf_strtab_hash_entry *) |
entry = (struct elf_strtab_hash_entry *) |
151 |
bfd_hash_lookup (&tab->table, str, true, copy); |
bfd_hash_lookup (&tab->table, str, TRUE, copy); |
152 |
|
|
153 |
if (entry == NULL) |
if (entry == NULL) |
154 |
return (bfd_size_type) -1; |
return (bfd_size_type) -1; |
157 |
if (entry->len == 0) |
if (entry->len == 0) |
158 |
{ |
{ |
159 |
entry->len = strlen (str) + 1; |
entry->len = strlen (str) + 1; |
160 |
|
/* 2G strings lose. */ |
161 |
|
BFD_ASSERT (entry->len > 0); |
162 |
if (tab->size == tab->alloced) |
if (tab->size == tab->alloced) |
163 |
{ |
{ |
164 |
bfd_size_type amt = sizeof (struct elf_strtab_hash_entry *); |
bfd_size_type amt = sizeof (struct elf_strtab_hash_entry *); |
165 |
tab->alloced *= 2; |
tab->alloced *= 2; |
166 |
tab->array = (struct elf_strtab_hash_entry **) |
tab->array = bfd_realloc (tab->array, tab->alloced * amt); |
|
bfd_realloc (tab->array, tab->alloced * amt); |
|
167 |
if (tab->array == NULL) |
if (tab->array == NULL) |
168 |
return (bfd_size_type) -1; |
return (bfd_size_type) -1; |
169 |
} |
} |
175 |
} |
} |
176 |
|
|
177 |
void |
void |
178 |
_bfd_elf_strtab_addref (tab, idx) |
_bfd_elf_strtab_addref (struct elf_strtab_hash *tab, bfd_size_type idx) |
|
struct elf_strtab_hash *tab; |
|
|
bfd_size_type idx; |
|
179 |
{ |
{ |
180 |
if (idx == 0 || idx == (bfd_size_type) -1) |
if (idx == 0 || idx == (bfd_size_type) -1) |
181 |
return; |
return; |
185 |
} |
} |
186 |
|
|
187 |
void |
void |
188 |
_bfd_elf_strtab_delref (tab, idx) |
_bfd_elf_strtab_delref (struct elf_strtab_hash *tab, bfd_size_type idx) |
|
struct elf_strtab_hash *tab; |
|
|
bfd_size_type idx; |
|
189 |
{ |
{ |
190 |
if (idx == 0 || idx == (bfd_size_type) -1) |
if (idx == 0 || idx == (bfd_size_type) -1) |
191 |
return; |
return; |
196 |
} |
} |
197 |
|
|
198 |
void |
void |
199 |
_bfd_elf_strtab_clear_all_refs (tab) |
_bfd_elf_strtab_clear_all_refs (struct elf_strtab_hash *tab) |
|
struct elf_strtab_hash *tab; |
|
200 |
{ |
{ |
201 |
bfd_size_type idx; |
bfd_size_type idx; |
202 |
|
|
205 |
} |
} |
206 |
|
|
207 |
bfd_size_type |
bfd_size_type |
208 |
_bfd_elf_strtab_size (tab) |
_bfd_elf_strtab_size (struct elf_strtab_hash *tab) |
|
struct elf_strtab_hash *tab; |
|
209 |
{ |
{ |
210 |
return tab->sec_size ? tab->sec_size : tab->size; |
return tab->sec_size ? tab->sec_size : tab->size; |
211 |
} |
} |
212 |
|
|
213 |
bfd_size_type |
bfd_size_type |
214 |
_bfd_elf_strtab_offset (tab, idx) |
_bfd_elf_strtab_offset (struct elf_strtab_hash *tab, bfd_size_type idx) |
|
struct elf_strtab_hash *tab; |
|
|
bfd_size_type idx; |
|
215 |
{ |
{ |
216 |
struct elf_strtab_hash_entry *entry; |
struct elf_strtab_hash_entry *entry; |
217 |
|
|
225 |
return tab->array[idx]->u.index; |
return tab->array[idx]->u.index; |
226 |
} |
} |
227 |
|
|
228 |
boolean |
bfd_boolean |
229 |
_bfd_elf_strtab_emit (abfd, tab) |
_bfd_elf_strtab_emit (register bfd *abfd, struct elf_strtab_hash *tab) |
|
register bfd *abfd; |
|
|
struct elf_strtab_hash *tab; |
|
230 |
{ |
{ |
231 |
bfd_size_type off = 1, i; |
bfd_size_type off = 1, i; |
232 |
|
|
233 |
if (bfd_bwrite ("", 1, abfd) != 1) |
if (bfd_bwrite ("", 1, abfd) != 1) |
234 |
return false; |
return FALSE; |
235 |
|
|
236 |
for (i = 1; i < tab->size; ++i) |
for (i = 1; i < tab->size; ++i) |
237 |
{ |
{ |
238 |
register const char *str; |
register const char *str; |
239 |
register size_t len; |
register unsigned int len; |
240 |
|
|
|
str = tab->array[i]->root.string; |
|
|
len = tab->array[i]->len; |
|
241 |
BFD_ASSERT (tab->array[i]->refcount == 0); |
BFD_ASSERT (tab->array[i]->refcount == 0); |
242 |
if (len == 0) |
len = tab->array[i]->len; |
243 |
|
if ((int) len < 0) |
244 |
continue; |
continue; |
245 |
|
|
246 |
if (bfd_bwrite ((PTR) str, (bfd_size_type) len, abfd) != len) |
str = tab->array[i]->root.string; |
247 |
return false; |
if (bfd_bwrite (str, len, abfd) != len) |
248 |
|
return FALSE; |
249 |
|
|
250 |
off += len; |
off += len; |
251 |
} |
} |
252 |
|
|
253 |
BFD_ASSERT (off == tab->sec_size); |
BFD_ASSERT (off == tab->sec_size); |
254 |
return true; |
return TRUE; |
255 |
} |
} |
256 |
|
|
257 |
/* Compare two elf_strtab_hash_entry structures. This is called via qsort. */ |
/* Compare two elf_strtab_hash_entry structures. Called via qsort. */ |
258 |
|
|
259 |
static int |
static int |
260 |
cmplengthentry (a, b) |
strrevcmp (const void *a, const void *b) |
261 |
const PTR a; |
{ |
262 |
const PTR b; |
struct elf_strtab_hash_entry *A = *(struct elf_strtab_hash_entry **) a; |
263 |
{ |
struct elf_strtab_hash_entry *B = *(struct elf_strtab_hash_entry **) b; |
264 |
struct elf_strtab_hash_entry * A = *(struct elf_strtab_hash_entry **) a; |
unsigned int lenA = A->len; |
265 |
struct elf_strtab_hash_entry * B = *(struct elf_strtab_hash_entry **) b; |
unsigned int lenB = B->len; |
266 |
|
const unsigned char *s = (const unsigned char *) A->root.string + lenA - 1; |
267 |
if (A->len < B->len) |
const unsigned char *t = (const unsigned char *) B->root.string + lenB - 1; |
268 |
return 1; |
int l = lenA < lenB ? lenA : lenB; |
|
else if (A->len > B->len) |
|
|
return -1; |
|
269 |
|
|
270 |
return memcmp (A->root.string, B->root.string, A->len); |
while (l) |
271 |
|
{ |
272 |
|
if (*s != *t) |
273 |
|
return (int) *s - (int) *t; |
274 |
|
s--; |
275 |
|
t--; |
276 |
|
l--; |
277 |
|
} |
278 |
|
return lenA - lenB; |
279 |
} |
} |
280 |
|
|
281 |
static int |
static inline int |
282 |
last4_eq (a, b) |
is_suffix (const struct elf_strtab_hash_entry *A, |
283 |
const PTR a; |
const struct elf_strtab_hash_entry *B) |
284 |
const PTR b; |
{ |
|
{ |
|
|
struct elf_strtab_hash_entry * A = (struct elf_strtab_hash_entry *) a; |
|
|
struct elf_strtab_hash_entry * B = (struct elf_strtab_hash_entry *) b; |
|
|
|
|
|
if (memcmp (A->root.string + A->len - 5, B->root.string + B->len - 5, 4) |
|
|
!= 0) |
|
|
/* This was a hashtable collision. */ |
|
|
return 0; |
|
|
|
|
285 |
if (A->len <= B->len) |
if (A->len <= B->len) |
286 |
/* B cannot be a suffix of A unless A is equal to B, which is guaranteed |
/* B cannot be a suffix of A unless A is equal to B, which is guaranteed |
287 |
not to be equal by the hash table. */ |
not to be equal by the hash table. */ |
288 |
return 0; |
return 0; |
289 |
|
|
290 |
return memcmp (A->root.string + (A->len - B->len), |
return memcmp (A->root.string + (A->len - B->len), |
291 |
B->root.string, B->len - 5) == 0; |
B->root.string, B->len - 1) == 0; |
292 |
} |
} |
293 |
|
|
294 |
/* This function assigns final string table offsets for used strings, |
/* This function assigns final string table offsets for used strings, |
295 |
merging strings matching suffixes of longer strings if possible. */ |
merging strings matching suffixes of longer strings if possible. */ |
296 |
|
|
297 |
void |
void |
298 |
_bfd_elf_strtab_finalize (tab) |
_bfd_elf_strtab_finalize (struct elf_strtab_hash *tab) |
|
struct elf_strtab_hash *tab; |
|
299 |
{ |
{ |
300 |
struct elf_strtab_hash_entry **array, **a, **end, *e; |
struct elf_strtab_hash_entry **array, **a, *e; |
|
htab_t last4tab = NULL; |
|
301 |
bfd_size_type size, amt; |
bfd_size_type size, amt; |
|
struct elf_strtab_hash_entry *last[256], **last_ptr[256]; |
|
302 |
|
|
303 |
/* GCC 2.91.66 (egcs-1.1.2) on i386 miscompiles this function when i is |
/* GCC 2.91.66 (egcs-1.1.2) on i386 miscompiles this function when i is |
304 |
a 64-bit bfd_size_type: a 64-bit target or --enable-64-bit-bfd. |
a 64-bit bfd_size_type: a 64-bit target or --enable-64-bit-bfd. |
306 |
cycles. */ |
cycles. */ |
307 |
size_t i; |
size_t i; |
308 |
|
|
309 |
/* Now sort the strings by length, longest first. */ |
/* Sort the strings by suffix and length. */ |
|
array = NULL; |
|
310 |
amt = tab->size * sizeof (struct elf_strtab_hash_entry *); |
amt = tab->size * sizeof (struct elf_strtab_hash_entry *); |
311 |
array = (struct elf_strtab_hash_entry **) bfd_malloc (amt); |
array = bfd_malloc (amt); |
312 |
if (array == NULL) |
if (array == NULL) |
313 |
goto alloc_failure; |
goto alloc_failure; |
314 |
|
|
|
memset (last, 0, sizeof (last)); |
|
|
for (i = 0; i < 256; ++i) |
|
|
last_ptr[i] = &last[i]; |
|
315 |
for (i = 1, a = array; i < tab->size; ++i) |
for (i = 1, a = array; i < tab->size; ++i) |
316 |
if (tab->array[i]->refcount) |
{ |
317 |
*a++ = tab->array[i]; |
e = tab->array[i]; |
318 |
else |
if (e->refcount) |
319 |
tab->array[i]->len = 0; |
{ |
320 |
|
*a++ = e; |
321 |
|
/* Adjust the length to not include the zero terminator. */ |
322 |
|
e->len -= 1; |
323 |
|
} |
324 |
|
else |
325 |
|
e->len = 0; |
326 |
|
} |
327 |
|
|
328 |
size = a - array; |
size = a - array; |
329 |
|
if (size != 0) |
330 |
|
{ |
331 |
|
qsort (array, size, sizeof (struct elf_strtab_hash_entry *), strrevcmp); |
332 |
|
|
333 |
qsort (array, size, sizeof (struct elf_strtab_hash_entry *), cmplengthentry); |
/* Loop over the sorted array and merge suffixes. Start from the |
334 |
|
end because we want eg. |
335 |
|
|
336 |
last4tab = htab_create_alloc (size * 4, NULL, last4_eq, NULL, calloc, free); |
s1 -> "d" |
337 |
if (last4tab == NULL) |
s2 -> "bcd" |
338 |
goto alloc_failure; |
s3 -> "abcd" |
339 |
|
|
340 |
/* Now insert the strings into hash tables (strings with last 4 characters |
to end up as |
|
and strings with last character equal), look for longer strings which |
|
|
we're suffix of. */ |
|
|
for (a = array, end = array + size; a < end; a++) |
|
|
{ |
|
|
register hashval_t hash; |
|
|
unsigned int c; |
|
|
unsigned int j; |
|
|
const unsigned char *s; |
|
|
PTR *p; |
|
341 |
|
|
342 |
e = *a; |
s3 -> "abcd" |
343 |
if (e->len > 4) |
s2 _____^ |
344 |
{ |
s1 _______^ |
|
s = e->root.string + e->len - 1; |
|
|
hash = 0; |
|
|
for (j = 0; j < 4; j++) |
|
|
{ |
|
|
c = *--s; |
|
|
hash += c + (c << 17); |
|
|
hash ^= hash >> 2; |
|
|
} |
|
|
p = htab_find_slot_with_hash (last4tab, e, hash, INSERT); |
|
|
if (p == NULL) |
|
|
goto alloc_failure; |
|
|
if (*p) |
|
|
{ |
|
|
struct elf_strtab_hash_entry *ent; |
|
345 |
|
|
346 |
ent = (struct elf_strtab_hash_entry *) *p; |
ie. we don't want s1 pointing into the old s2. */ |
347 |
e->u.suffix = ent; |
e = *--a; |
348 |
e->len = 0; |
e->len += 1; |
349 |
continue; |
while (--a >= array) |
|
} |
|
|
else |
|
|
*p = (PTR) e; |
|
|
} |
|
|
else |
|
350 |
{ |
{ |
351 |
struct elf_strtab_hash_entry *tem; |
struct elf_strtab_hash_entry *cmp = *a; |
|
|
|
|
c = e->root.string[e->len - 2] & 0xff; |
|
352 |
|
|
353 |
for (tem = last[c]; tem; tem = tem->u.next) |
cmp->len += 1; |
354 |
if (tem->len > e->len |
if (is_suffix (e, cmp)) |
|
&& memcmp (tem->root.string + (tem->len - e->len), |
|
|
e->root.string, e->len - 1) == 0) |
|
|
break; |
|
|
if (tem) |
|
355 |
{ |
{ |
356 |
e->u.suffix = tem; |
cmp->u.suffix = e; |
357 |
e->len = 0; |
cmp->len = -cmp->len; |
|
continue; |
|
358 |
} |
} |
359 |
|
else |
360 |
|
e = cmp; |
361 |
} |
} |
|
|
|
|
c = e->root.string[e->len - 2] & 0xff; |
|
|
/* Put longest strings first. */ |
|
|
*last_ptr[c] = e; |
|
|
last_ptr[c] = &e->u.next; |
|
|
e->u.next = NULL; |
|
362 |
} |
} |
363 |
|
|
364 |
alloc_failure: |
alloc_failure: |
365 |
if (array) |
if (array) |
366 |
free (array); |
free (array); |
|
if (last4tab) |
|
|
htab_delete (last4tab); |
|
367 |
|
|
368 |
/* Now assign positions to the strings we want to keep. */ |
/* Assign positions to the strings we want to keep. */ |
369 |
size = 1; |
size = 1; |
370 |
for (i = 1; i < tab->size; ++i) |
for (i = 1; i < tab->size; ++i) |
371 |
{ |
{ |
372 |
e = tab->array[i]; |
e = tab->array[i]; |
373 |
if (e->refcount && e->len) |
if (e->refcount && e->len > 0) |
374 |
{ |
{ |
375 |
e->u.index = size; |
e->u.index = size; |
376 |
size += e->len; |
size += e->len; |
379 |
|
|
380 |
tab->sec_size = size; |
tab->sec_size = size; |
381 |
|
|
382 |
/* And now adjust the rest. */ |
/* Adjust the rest. */ |
383 |
for (i = 1; i < tab->size; ++i) |
for (i = 1; i < tab->size; ++i) |
384 |
{ |
{ |
385 |
e = tab->array[i]; |
e = tab->array[i]; |
386 |
if (e->refcount && ! e->len) |
if (e->refcount && e->len < 0) |
387 |
e->u.index = e->u.suffix->u.index |
e->u.index = e->u.suffix->u.index + (e->u.suffix->len + e->len); |
|
+ (e->u.suffix->len - strlen (e->root.string) - 1); |
|
388 |
} |
} |
389 |
} |
} |