1 /* ELF strtab with GC and suffix merging support.
2    Copyright 2001, 2002, 2003, 2005, 2006 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 2 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, MA 02110-1301, USA.  */
20 
21 #include "bfd.h"
22 #include "sysdep.h"
23 #include "libbfd.h"
24 #include "elf-bfd.h"
25 #include "hashtab.h"
26 #include "libiberty.h"
27 
28 /* An entry in the strtab hash table.  */
29 
30 struct elf_strtab_hash_entry
31 {
32   struct bfd_hash_entry root;
33   /* Length of this entry.  This includes the zero terminator.  */
34   int len;
35   unsigned int refcount;
36   union {
37     /* Index within the merged section.  */
38     bfd_size_type index;
39     /* Entry this is a suffix of (if len < 0).  */
40     struct elf_strtab_hash_entry *suffix;
41   } u;
42 };
43 
44 /* The strtab hash table.  */
45 
46 struct elf_strtab_hash
47 {
48   struct bfd_hash_table table;
49   /* Next available index.  */
50   bfd_size_type size;
51   /* Number of array entries alloced.  */
52   bfd_size_type alloced;
53   /* Final strtab size.  */
54   bfd_size_type sec_size;
55   /* Array of pointers to strtab entries.  */
56   struct elf_strtab_hash_entry **array;
57 };
58 
59 /* Routine to create an entry in a section merge hashtab.  */
60 
61 static struct bfd_hash_entry *
62 elf_strtab_hash_newfunc (struct bfd_hash_entry *entry,
63 			 struct bfd_hash_table *table,
64 			 const char *string)
65 {
66   /* Allocate the structure if it has not already been allocated by a
67      subclass.  */
68   if (entry == NULL)
69     entry = bfd_hash_allocate (table, sizeof (struct elf_strtab_hash_entry));
70   if (entry == NULL)
71     return NULL;
72 
73   /* Call the allocation method of the superclass.  */
74   entry = bfd_hash_newfunc (entry, table, string);
75 
76   if (entry)
77     {
78       /* 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;
83       ret->refcount = 0;
84       ret->len = 0;
85     }
86 
87   return entry;
88 }
89 
90 /* Create a new hash table.  */
91 
92 struct elf_strtab_hash *
93 _bfd_elf_strtab_init (void)
94 {
95   struct elf_strtab_hash *table;
96   bfd_size_type amt = sizeof (struct elf_strtab_hash);
97 
98   table = bfd_malloc (amt);
99   if (table == NULL)
100     return NULL;
101 
102   if (!bfd_hash_table_init (&table->table, elf_strtab_hash_newfunc,
103 			    sizeof (struct elf_strtab_hash_entry)))
104     {
105       free (table);
106       return NULL;
107     }
108 
109   table->sec_size = 0;
110   table->size = 1;
111   table->alloced = 64;
112   amt = sizeof (struct elf_strtab_hasn_entry *);
113   table->array = bfd_malloc (table->alloced * amt);
114   if (table->array == NULL)
115     {
116       free (table);
117       return NULL;
118     }
119 
120   table->array[0] = NULL;
121 
122   return table;
123 }
124 
125 /* Free a strtab.  */
126 
127 void
128 _bfd_elf_strtab_free (struct elf_strtab_hash *tab)
129 {
130   bfd_hash_table_free (&tab->table);
131   free (tab->array);
132   free (tab);
133 }
134 
135 /* Get the index of an entity in a hash table, adding it if it is not
136    already present.  */
137 
138 bfd_size_type
139 _bfd_elf_strtab_add (struct elf_strtab_hash *tab,
140 		     const char *str,
141 		     bfd_boolean copy)
142 {
143   register struct elf_strtab_hash_entry *entry;
144 
145   /* We handle this specially, since we don't want to do refcounting
146      on it.  */
147   if (*str == '\0')
148     return 0;
149 
150   BFD_ASSERT (tab->sec_size == 0);
151   entry = (struct elf_strtab_hash_entry *)
152 	  bfd_hash_lookup (&tab->table, str, TRUE, copy);
153 
154   if (entry == NULL)
155     return (bfd_size_type) -1;
156 
157   entry->refcount++;
158   if (entry->len == 0)
159     {
160       entry->len = strlen (str) + 1;
161       /* 2G strings lose.  */
162       BFD_ASSERT (entry->len > 0);
163       if (tab->size == tab->alloced)
164 	{
165 	  bfd_size_type amt = sizeof (struct elf_strtab_hash_entry *);
166 	  tab->alloced *= 2;
167 	  tab->array = bfd_realloc (tab->array, tab->alloced * amt);
168 	  if (tab->array == NULL)
169 	    return (bfd_size_type) -1;
170 	}
171 
172       entry->u.index = tab->size++;
173       tab->array[entry->u.index] = entry;
174     }
175   return entry->u.index;
176 }
177 
178 void
179 _bfd_elf_strtab_addref (struct elf_strtab_hash *tab, bfd_size_type idx)
180 {
181   if (idx == 0 || idx == (bfd_size_type) -1)
182     return;
183   BFD_ASSERT (tab->sec_size == 0);
184   BFD_ASSERT (idx < tab->size);
185   ++tab->array[idx]->refcount;
186 }
187 
188 void
189 _bfd_elf_strtab_delref (struct elf_strtab_hash *tab, bfd_size_type idx)
190 {
191   if (idx == 0 || idx == (bfd_size_type) -1)
192     return;
193   BFD_ASSERT (tab->sec_size == 0);
194   BFD_ASSERT (idx < tab->size);
195   BFD_ASSERT (tab->array[idx]->refcount > 0);
196   --tab->array[idx]->refcount;
197 }
198 
199 void
200 _bfd_elf_strtab_clear_all_refs (struct elf_strtab_hash *tab)
201 {
202   bfd_size_type idx;
203 
204   for (idx = 1; idx < tab->size; ++idx)
205     tab->array[idx]->refcount = 0;
206 }
207 
208 bfd_size_type
209 _bfd_elf_strtab_size (struct elf_strtab_hash *tab)
210 {
211   return tab->sec_size ? tab->sec_size : tab->size;
212 }
213 
214 bfd_size_type
215 _bfd_elf_strtab_offset (struct elf_strtab_hash *tab, bfd_size_type idx)
216 {
217   struct elf_strtab_hash_entry *entry;
218 
219   if (idx == 0)
220     return 0;
221   BFD_ASSERT (idx < tab->size);
222   BFD_ASSERT (tab->sec_size);
223   entry = tab->array[idx];
224   BFD_ASSERT (entry->refcount > 0);
225   entry->refcount--;
226   return tab->array[idx]->u.index;
227 }
228 
229 bfd_boolean
230 _bfd_elf_strtab_emit (register bfd *abfd, struct elf_strtab_hash *tab)
231 {
232   bfd_size_type off = 1, i;
233 
234   if (bfd_bwrite ("", 1, abfd) != 1)
235     return FALSE;
236 
237   for (i = 1; i < tab->size; ++i)
238     {
239       register const char *str;
240       register unsigned int len;
241 
242       BFD_ASSERT (tab->array[i]->refcount == 0);
243       len = tab->array[i]->len;
244       if ((int) len < 0)
245 	continue;
246 
247       str = tab->array[i]->root.string;
248       if (bfd_bwrite (str, len, abfd) != len)
249 	return FALSE;
250 
251       off += len;
252     }
253 
254   BFD_ASSERT (off == tab->sec_size);
255   return TRUE;
256 }
257 
258 /* Compare two elf_strtab_hash_entry structures.  Called via qsort.  */
259 
260 static int
261 strrevcmp (const void *a, const void *b)
262 {
263   struct elf_strtab_hash_entry *A = *(struct elf_strtab_hash_entry **) a;
264   struct elf_strtab_hash_entry *B = *(struct elf_strtab_hash_entry **) b;
265   unsigned int lenA = A->len;
266   unsigned int lenB = B->len;
267   const unsigned char *s = (const unsigned char *) A->root.string + lenA - 1;
268   const unsigned char *t = (const unsigned char *) B->root.string + lenB - 1;
269   int l = lenA < lenB ? lenA : lenB;
270 
271   while (l)
272     {
273       if (*s != *t)
274 	return (int) *s - (int) *t;
275       s--;
276       t--;
277       l--;
278     }
279   return lenA - lenB;
280 }
281 
282 static inline int
283 is_suffix (const struct elf_strtab_hash_entry *A,
284 	   const struct elf_strtab_hash_entry *B)
285 {
286   if (A->len <= B->len)
287     /* B cannot be a suffix of A unless A is equal to B, which is guaranteed
288        not to be equal by the hash table.  */
289     return 0;
290 
291   return memcmp (A->root.string + (A->len - B->len),
292 		 B->root.string, B->len - 1) == 0;
293 }
294 
295 /* This function assigns final string table offsets for used strings,
296    merging strings matching suffixes of longer strings if possible.  */
297 
298 void
299 _bfd_elf_strtab_finalize (struct elf_strtab_hash *tab)
300 {
301   struct elf_strtab_hash_entry **array, **a, *e;
302   bfd_size_type size, amt;
303 
304   /* GCC 2.91.66 (egcs-1.1.2) on i386 miscompiles this function when i is
305      a 64-bit bfd_size_type: a 64-bit target or --enable-64-bit-bfd.
306      Besides, indexing with a long long wouldn't give anything but extra
307      cycles.  */
308   size_t i;
309 
310   /* Sort the strings by suffix and length.  */
311   amt = tab->size * sizeof (struct elf_strtab_hash_entry *);
312   array = bfd_malloc (amt);
313   if (array == NULL)
314     goto alloc_failure;
315 
316   for (i = 1, a = array; i < tab->size; ++i)
317     {
318       e = tab->array[i];
319       if (e->refcount)
320 	{
321 	  *a++ = e;
322 	  /* Adjust the length to not include the zero terminator.  */
323 	  e->len -= 1;
324 	}
325       else
326 	e->len = 0;
327     }
328 
329   size = a - array;
330   if (size != 0)
331     {
332       qsort (array, size, sizeof (struct elf_strtab_hash_entry *), strrevcmp);
333 
334       /* Loop over the sorted array and merge suffixes.  Start from the
335 	 end because we want eg.
336 
337 	 s1 -> "d"
338 	 s2 -> "bcd"
339 	 s3 -> "abcd"
340 
341 	 to end up as
342 
343 	 s3 -> "abcd"
344 	 s2 _____^
345 	 s1 _______^
346 
347 	 ie. we don't want s1 pointing into the old s2.  */
348       e = *--a;
349       e->len += 1;
350       while (--a >= array)
351 	{
352 	  struct elf_strtab_hash_entry *cmp = *a;
353 
354 	  cmp->len += 1;
355 	  if (is_suffix (e, cmp))
356 	    {
357 	      cmp->u.suffix = e;
358 	      cmp->len = -cmp->len;
359 	    }
360 	  else
361 	    e = cmp;
362 	}
363     }
364 
365 alloc_failure:
366   if (array)
367     free (array);
368 
369   /* Assign positions to the strings we want to keep.  */
370   size = 1;
371   for (i = 1; i < tab->size; ++i)
372     {
373       e = tab->array[i];
374       if (e->refcount && e->len > 0)
375 	{
376 	  e->u.index = size;
377 	  size += e->len;
378 	}
379     }
380 
381   tab->sec_size = size;
382 
383   /* Adjust the rest.  */
384   for (i = 1; i < tab->size; ++i)
385     {
386       e = tab->array[i];
387       if (e->refcount && e->len < 0)
388 	e->u.index = e->u.suffix->u.index + (e->u.suffix->len + e->len);
389     }
390 }
391