1 /* hash.c -- hash table routines for BFD
2 Copyright 1993, 1994, 1995, 1997, 1999, 2001, 2002, 2003, 2004, 2005,
3 2006, 2007, 2009, 2010, 2011 Free Software Foundation, Inc.
4 Written by Steve Chamberlain <sac@cygnus.com>
6 This file is part of BFD, the Binary File Descriptor library.
8 This program is free software; you can redistribute it and/or modify
9 it under the terms of the GNU General Public License as published by
10 the Free Software Foundation; either version 3 of the License, or
11 (at your option) any later version.
13 This program is distributed in the hope that it will be useful,
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 GNU General Public License for more details.
18 You should have received a copy of the GNU General Public License
19 along with this program; if not, write to the Free Software
20 Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston,
21 MA 02110-1301, USA. */
27 #include "libiberty.h"
34 BFD provides a simple set of hash table functions. Routines
35 are provided to initialize a hash table, to free a hash table,
36 to look up a string in a hash table and optionally create an
37 entry for it, and to traverse a hash table. There is
38 currently no routine to delete an string from a hash table.
40 The basic hash table does not permit any data to be stored
41 with a string. However, a hash table is designed to present a
42 base class from which other types of hash tables may be
43 derived. These derived types may store additional information
44 with the string. Hash tables were implemented in this way,
45 rather than simply providing a data pointer in a hash table
46 entry, because they were designed for use by the linker back
47 ends. The linker may create thousands of hash table entries,
48 and the overhead of allocating private data and storing and
49 following pointers becomes noticeable.
51 The basic hash table code is in <<hash.c>>.
54 @* Creating and Freeing a Hash Table::
55 @* Looking Up or Entering a String::
56 @* Traversing a Hash Table::
57 @* Deriving a New Hash Table Type::
61 Creating and Freeing a Hash Table, Looking Up or Entering a String, Hash Tables, Hash Tables
63 Creating and freeing a hash table
65 @findex bfd_hash_table_init
66 @findex bfd_hash_table_init_n
67 To create a hash table, create an instance of a <<struct
68 bfd_hash_table>> (defined in <<bfd.h>>) and call
69 <<bfd_hash_table_init>> (if you know approximately how many
70 entries you will need, the function <<bfd_hash_table_init_n>>,
71 which takes a @var{size} argument, may be used).
72 <<bfd_hash_table_init>> returns <<FALSE>> if some sort of
75 @findex bfd_hash_newfunc
76 The function <<bfd_hash_table_init>> take as an argument a
77 function to use to create new entries. For a basic hash
78 table, use the function <<bfd_hash_newfunc>>. @xref{Deriving
79 a New Hash Table Type}, for why you would want to use a
80 different value for this argument.
82 @findex bfd_hash_allocate
83 <<bfd_hash_table_init>> will create an objalloc which will be
84 used to allocate new entries. You may allocate memory on this
85 objalloc using <<bfd_hash_allocate>>.
87 @findex bfd_hash_table_free
88 Use <<bfd_hash_table_free>> to free up all the memory that has
89 been allocated for a hash table. This will not free up the
90 <<struct bfd_hash_table>> itself, which you must provide.
92 @findex bfd_hash_set_default_size
93 Use <<bfd_hash_set_default_size>> to set the default size of
97 Looking Up or Entering a String, Traversing a Hash Table, Creating and Freeing a Hash Table, Hash Tables
99 Looking up or entering a string
101 @findex bfd_hash_lookup
102 The function <<bfd_hash_lookup>> is used both to look up a
103 string in the hash table and to create a new entry.
105 If the @var{create} argument is <<FALSE>>, <<bfd_hash_lookup>>
106 will look up a string. If the string is found, it will
107 returns a pointer to a <<struct bfd_hash_entry>>. If the
108 string is not found in the table <<bfd_hash_lookup>> will
109 return <<NULL>>. You should not modify any of the fields in
110 the returns <<struct bfd_hash_entry>>.
112 If the @var{create} argument is <<TRUE>>, the string will be
113 entered into the hash table if it is not already there.
114 Either way a pointer to a <<struct bfd_hash_entry>> will be
115 returned, either to the existing structure or to a newly
116 created one. In this case, a <<NULL>> return means that an
119 If the @var{create} argument is <<TRUE>>, and a new entry is
120 created, the @var{copy} argument is used to decide whether to
121 copy the string onto the hash table objalloc or not. If
122 @var{copy} is passed as <<FALSE>>, you must be careful not to
123 deallocate or modify the string as long as the hash table
127 Traversing a Hash Table, Deriving a New Hash Table Type, Looking Up or Entering a String, Hash Tables
129 Traversing a hash table
131 @findex bfd_hash_traverse
132 The function <<bfd_hash_traverse>> may be used to traverse a
133 hash table, calling a function on each element. The traversal
134 is done in a random order.
136 <<bfd_hash_traverse>> takes as arguments a function and a
137 generic <<void *>> pointer. The function is called with a
138 hash table entry (a <<struct bfd_hash_entry *>>) and the
139 generic pointer passed to <<bfd_hash_traverse>>. The function
140 must return a <<boolean>> value, which indicates whether to
141 continue traversing the hash table. If the function returns
142 <<FALSE>>, <<bfd_hash_traverse>> will stop the traversal and
146 Deriving a New Hash Table Type, , Traversing a Hash Table, Hash Tables
148 Deriving a new hash table type
150 Many uses of hash tables want to store additional information
151 which each entry in the hash table. Some also find it
152 convenient to store additional information with the hash table
153 itself. This may be done using a derived hash table.
155 Since C is not an object oriented language, creating a derived
156 hash table requires sticking together some boilerplate
157 routines with a few differences specific to the type of hash
158 table you want to create.
160 An example of a derived hash table is the linker hash table.
161 The structures for this are defined in <<bfdlink.h>>. The
162 functions are in <<linker.c>>.
164 You may also derive a hash table from an already derived hash
165 table. For example, the a.out linker backend code uses a hash
166 table derived from the linker hash table.
169 @* Define the Derived Structures::
170 @* Write the Derived Creation Routine::
171 @* Write Other Derived Routines::
175 Define the Derived Structures, Write the Derived Creation Routine, Deriving a New Hash Table Type, Deriving a New Hash Table Type
177 Define the derived structures
179 You must define a structure for an entry in the hash table,
180 and a structure for the hash table itself.
182 The first field in the structure for an entry in the hash
183 table must be of the type used for an entry in the hash table
184 you are deriving from. If you are deriving from a basic hash
185 table this is <<struct bfd_hash_entry>>, which is defined in
186 <<bfd.h>>. The first field in the structure for the hash
187 table itself must be of the type of the hash table you are
188 deriving from itself. If you are deriving from a basic hash
189 table, this is <<struct bfd_hash_table>>.
191 For example, the linker hash table defines <<struct
192 bfd_link_hash_entry>> (in <<bfdlink.h>>). The first field,
193 <<root>>, is of type <<struct bfd_hash_entry>>. Similarly,
194 the first field in <<struct bfd_link_hash_table>>, <<table>>,
195 is of type <<struct bfd_hash_table>>.
198 Write the Derived Creation Routine, Write Other Derived Routines, Define the Derived Structures, Deriving a New Hash Table Type
200 Write the derived creation routine
202 You must write a routine which will create and initialize an
203 entry in the hash table. This routine is passed as the
204 function argument to <<bfd_hash_table_init>>.
206 In order to permit other hash tables to be derived from the
207 hash table you are creating, this routine must be written in a
210 The first argument to the creation routine is a pointer to a
211 hash table entry. This may be <<NULL>>, in which case the
212 routine should allocate the right amount of space. Otherwise
213 the space has already been allocated by a hash table type
214 derived from this one.
216 After allocating space, the creation routine must call the
217 creation routine of the hash table type it is derived from,
218 passing in a pointer to the space it just allocated. This
219 will initialize any fields used by the base hash table.
221 Finally the creation routine must initialize any local fields
222 for the new hash table type.
224 Here is a boilerplate example of a creation routine.
225 @var{function_name} is the name of the routine.
226 @var{entry_type} is the type of an entry in the hash table you
227 are creating. @var{base_newfunc} is the name of the creation
228 routine of the hash table type your hash table is derived
233 .struct bfd_hash_entry *
234 .@var{function_name} (struct bfd_hash_entry *entry,
235 . struct bfd_hash_table *table,
236 . const char *string)
238 . struct @var{entry_type} *ret = (@var{entry_type} *) entry;
240 . {* Allocate the structure if it has not already been allocated by a
244 . ret = bfd_hash_allocate (table, sizeof (* ret));
249 . {* Call the allocation method of the base class. *}
250 . ret = ((@var{entry_type} *)
251 . @var{base_newfunc} ((struct bfd_hash_entry *) ret, table, string));
253 . {* Initialize the local fields here. *}
255 . return (struct bfd_hash_entry *) ret;
259 The creation routine for the linker hash table, which is in
260 <<linker.c>>, looks just like this example.
261 @var{function_name} is <<_bfd_link_hash_newfunc>>.
262 @var{entry_type} is <<struct bfd_link_hash_entry>>.
263 @var{base_newfunc} is <<bfd_hash_newfunc>>, the creation
264 routine for a basic hash table.
266 <<_bfd_link_hash_newfunc>> also initializes the local fields
267 in a linker hash table entry: <<type>>, <<written>> and
271 Write Other Derived Routines, , Write the Derived Creation Routine, Deriving a New Hash Table Type
273 Write other derived routines
275 You will want to write other routines for your new hash table,
278 You will want an initialization routine which calls the
279 initialization routine of the hash table you are deriving from
280 and initializes any other local fields. For the linker hash
281 table, this is <<_bfd_link_hash_table_init>> in <<linker.c>>.
283 You will want a lookup routine which calls the lookup routine
284 of the hash table you are deriving from and casts the result.
285 The linker hash table uses <<bfd_link_hash_lookup>> in
286 <<linker.c>> (this actually takes an additional argument which
287 it uses to decide how to return the looked up value).
289 You may want a traversal routine. This should just call the
290 traversal routine of the hash table you are deriving from with
291 appropriate casts. The linker hash table uses
292 <<bfd_link_hash_traverse>> in <<linker.c>>.
294 These routines may simply be defined as macros. For example,
295 the a.out backend linker hash table, which is derived from the
296 linker hash table, uses macros for the lookup and traversal
297 routines. These are <<aout_link_hash_lookup>> and
298 <<aout_link_hash_traverse>> in aoutx.h.
301 /* The default number of entries to use when creating a hash table. */
302 #define DEFAULT_SIZE 4051
304 /* The following function returns a nearest prime number which is
305 greater than N, and near a power of two. Copied from libiberty.
306 Returns zero for ridiculously large N to signify an error. */
309 higher_prime_number (unsigned long n)
311 /* These are primes that are near, but slightly smaller than, a
313 static const unsigned long primes[] = {
315 (unsigned long) 2039,
316 (unsigned long) 32749,
317 (unsigned long) 65521,
318 (unsigned long) 131071,
319 (unsigned long) 262139,
320 (unsigned long) 524287,
321 (unsigned long) 1048573,
322 (unsigned long) 2097143,
323 (unsigned long) 4194301,
324 (unsigned long) 8388593,
325 (unsigned long) 16777213,
326 (unsigned long) 33554393,
327 (unsigned long) 67108859,
328 (unsigned long) 134217689,
329 (unsigned long) 268435399,
330 (unsigned long) 536870909,
331 (unsigned long) 1073741789,
332 (unsigned long) 2147483647,
334 ((unsigned long) 2147483647) + ((unsigned long) 2147483644),
337 const unsigned long *low = &primes[0];
338 const unsigned long *high = &primes[sizeof (primes) / sizeof (primes[0])];
342 const unsigned long *mid = low + (high - low) / 2;
355 static unsigned long bfd_default_hash_table_size = DEFAULT_SIZE;
357 /* Create a new hash table, given a number of entries. */
360 bfd_hash_table_init_n (struct bfd_hash_table *table,
361 struct bfd_hash_entry *(*newfunc) (struct bfd_hash_entry *,
362 struct bfd_hash_table *,
364 unsigned int entsize,
370 alloc *= sizeof (struct bfd_hash_entry *);
371 if (alloc / sizeof (struct bfd_hash_entry *) != size)
373 bfd_set_error (bfd_error_no_memory);
377 table->memory = (void *) objalloc_create ();
378 if (table->memory == NULL)
380 bfd_set_error (bfd_error_no_memory);
383 table->table = (struct bfd_hash_entry **)
384 objalloc_alloc ((struct objalloc *) table->memory, alloc);
385 if (table->table == NULL)
387 bfd_set_error (bfd_error_no_memory);
390 memset ((void *) table->table, 0, alloc);
392 table->entsize = entsize;
395 table->newfunc = newfunc;
399 /* Create a new hash table with the default number of entries. */
402 bfd_hash_table_init (struct bfd_hash_table *table,
403 struct bfd_hash_entry *(*newfunc) (struct bfd_hash_entry *,
404 struct bfd_hash_table *,
406 unsigned int entsize)
408 return bfd_hash_table_init_n (table, newfunc, entsize,
409 bfd_default_hash_table_size);
412 /* Free a hash table. */
415 bfd_hash_table_free (struct bfd_hash_table *table)
417 objalloc_free ((struct objalloc *) table->memory);
418 table->memory = NULL;
421 static inline unsigned long
422 bfd_hash_hash (const char *string, unsigned int *lenp)
424 const unsigned char *s;
431 s = (const unsigned char *) string;
432 while ((c = *s++) != '\0')
434 hash += c + (c << 17);
437 len = (s - (const unsigned char *) string) - 1;
438 hash += len + (len << 17);
445 /* Look up a string in a hash table. */
447 struct bfd_hash_entry *
448 bfd_hash_lookup (struct bfd_hash_table *table,
454 struct bfd_hash_entry *hashp;
458 hash = bfd_hash_hash (string, &len);
459 _index = hash % table->size;
460 for (hashp = table->table[_index];
464 if (hashp->hash == hash
465 && strcmp (hashp->string, string) == 0)
476 new_string = (char *) objalloc_alloc ((struct objalloc *) table->memory,
480 bfd_set_error (bfd_error_no_memory);
483 memcpy (new_string, string, len + 1);
487 return bfd_hash_insert (table, string, hash);
490 /* Insert an entry in a hash table. */
492 struct bfd_hash_entry *
493 bfd_hash_insert (struct bfd_hash_table *table,
497 struct bfd_hash_entry *hashp;
500 hashp = (*table->newfunc) (NULL, table, string);
503 hashp->string = string;
505 _index = hash % table->size;
506 hashp->next = table->table[_index];
507 table->table[_index] = hashp;
510 if (!table->frozen && table->count > table->size * 3 / 4)
512 unsigned long newsize = higher_prime_number (table->size);
513 struct bfd_hash_entry **newtable;
515 unsigned long alloc = newsize * sizeof (struct bfd_hash_entry *);
517 /* If we can't find a higher prime, or we can't possibly alloc
518 that much memory, don't try to grow the table. */
519 if (newsize == 0 || alloc / sizeof (struct bfd_hash_entry *) != newsize)
525 newtable = ((struct bfd_hash_entry **)
526 objalloc_alloc ((struct objalloc *) table->memory, alloc));
527 if (newtable == NULL)
532 memset ((PTR) newtable, 0, alloc);
534 for (hi = 0; hi < table->size; hi ++)
535 while (table->table[hi])
537 struct bfd_hash_entry *chain = table->table[hi];
538 struct bfd_hash_entry *chain_end = chain;
540 while (chain_end->next && chain_end->next->hash == chain->hash)
541 chain_end = chain_end->next;
543 table->table[hi] = chain_end->next;
544 _index = chain->hash % newsize;
545 chain_end->next = newtable[_index];
546 newtable[_index] = chain;
548 table->table = newtable;
549 table->size = newsize;
555 /* Rename an entry in a hash table. */
558 bfd_hash_rename (struct bfd_hash_table *table,
560 struct bfd_hash_entry *ent)
563 struct bfd_hash_entry **pph;
565 _index = ent->hash % table->size;
566 for (pph = &table->table[_index]; *pph != NULL; pph = &(*pph)->next)
573 ent->string = string;
574 ent->hash = bfd_hash_hash (string, NULL);
575 _index = ent->hash % table->size;
576 ent->next = table->table[_index];
577 table->table[_index] = ent;
580 /* Replace an entry in a hash table. */
583 bfd_hash_replace (struct bfd_hash_table *table,
584 struct bfd_hash_entry *old,
585 struct bfd_hash_entry *nw)
588 struct bfd_hash_entry **pph;
590 _index = old->hash % table->size;
591 for (pph = &table->table[_index];
605 /* Allocate space in a hash table. */
608 bfd_hash_allocate (struct bfd_hash_table *table,
613 ret = objalloc_alloc ((struct objalloc *) table->memory, size);
614 if (ret == NULL && size != 0)
615 bfd_set_error (bfd_error_no_memory);
619 /* Base method for creating a new hash table entry. */
621 struct bfd_hash_entry *
622 bfd_hash_newfunc (struct bfd_hash_entry *entry,
623 struct bfd_hash_table *table,
624 const char *string ATTRIBUTE_UNUSED)
627 entry = (struct bfd_hash_entry *) bfd_hash_allocate (table,
632 /* Traverse a hash table. */
635 bfd_hash_traverse (struct bfd_hash_table *table,
636 bfd_boolean (*func) (struct bfd_hash_entry *, void *),
642 for (i = 0; i < table->size; i++)
644 struct bfd_hash_entry *p;
646 for (p = table->table[i]; p != NULL; p = p->next)
647 if (! (*func) (p, info))
655 bfd_hash_set_default_size (unsigned long hash_size)
657 /* Extend this prime list if you want more granularity of hash table size. */
658 static const unsigned long hash_size_primes[] =
660 251, 509, 1021, 2039, 4051, 8599, 16699, 32749, 65537
664 /* Work out best prime number near the hash_size. */
665 for (_index = 0; _index < ARRAY_SIZE (hash_size_primes) - 1; ++_index)
666 if (hash_size <= hash_size_primes[_index])
669 bfd_default_hash_table_size = hash_size_primes[_index];
670 return bfd_default_hash_table_size;
673 /* A few different object file formats (a.out, COFF, ELF) use a string
674 table. These functions support adding strings to a string table,
675 returning the byte offset, and writing out the table.
677 Possible improvements:
678 + look for strings matching trailing substrings of other strings
679 + better data structures? balanced trees?
680 + look at reducing memory use elsewhere -- maybe if we didn't have
681 to construct the entire symbol table at once, we could get by
682 with smaller amounts of VM? (What effect does that have on the
683 string table reductions?) */
685 /* An entry in the strtab hash table. */
687 struct strtab_hash_entry
689 struct bfd_hash_entry root;
690 /* Index in string table. */
692 /* Next string in strtab. */
693 struct strtab_hash_entry *next;
696 /* The strtab hash table. */
698 struct bfd_strtab_hash
700 struct bfd_hash_table table;
701 /* Size of strtab--also next available index. */
703 /* First string in strtab. */
704 struct strtab_hash_entry *first;
705 /* Last string in strtab. */
706 struct strtab_hash_entry *last;
707 /* Whether to precede strings with a two byte length, as in the
708 XCOFF .debug section. */
712 /* Routine to create an entry in a strtab. */
714 static struct bfd_hash_entry *
715 strtab_hash_newfunc (struct bfd_hash_entry *entry,
716 struct bfd_hash_table *table,
719 struct strtab_hash_entry *ret = (struct strtab_hash_entry *) entry;
721 /* Allocate the structure if it has not already been allocated by a
724 ret = (struct strtab_hash_entry *) bfd_hash_allocate (table,
729 /* Call the allocation method of the superclass. */
730 ret = (struct strtab_hash_entry *)
731 bfd_hash_newfunc ((struct bfd_hash_entry *) ret, table, string);
735 /* Initialize the local fields. */
736 ret->index = (bfd_size_type) -1;
740 return (struct bfd_hash_entry *) ret;
743 /* Look up an entry in an strtab. */
745 #define strtab_hash_lookup(t, string, create, copy) \
746 ((struct strtab_hash_entry *) \
747 bfd_hash_lookup (&(t)->table, (string), (create), (copy)))
749 /* Create a new strtab. */
751 struct bfd_strtab_hash *
752 _bfd_stringtab_init (void)
754 struct bfd_strtab_hash *table;
755 bfd_size_type amt = sizeof (* table);
757 table = (struct bfd_strtab_hash *) bfd_malloc (amt);
761 if (!bfd_hash_table_init (&table->table, strtab_hash_newfunc,
762 sizeof (struct strtab_hash_entry)))
771 table->xcoff = FALSE;
776 /* Create a new strtab in which the strings are output in the format
777 used in the XCOFF .debug section: a two byte length precedes each
780 struct bfd_strtab_hash *
781 _bfd_xcoff_stringtab_init (void)
783 struct bfd_strtab_hash *ret;
785 ret = _bfd_stringtab_init ();
794 _bfd_stringtab_free (struct bfd_strtab_hash *table)
796 bfd_hash_table_free (&table->table);
800 /* Get the index of a string in a strtab, adding it if it is not
801 already present. If HASH is FALSE, we don't really use the hash
802 table, and we don't eliminate duplicate strings. */
805 _bfd_stringtab_add (struct bfd_strtab_hash *tab,
810 struct strtab_hash_entry *entry;
814 entry = strtab_hash_lookup (tab, str, TRUE, copy);
816 return (bfd_size_type) -1;
820 entry = (struct strtab_hash_entry *) bfd_hash_allocate (&tab->table,
823 return (bfd_size_type) -1;
825 entry->root.string = str;
830 n = (char *) bfd_hash_allocate (&tab->table, strlen (str) + 1);
832 return (bfd_size_type) -1;
833 entry->root.string = n;
835 entry->index = (bfd_size_type) -1;
839 if (entry->index == (bfd_size_type) -1)
841 entry->index = tab->size;
842 tab->size += strlen (str) + 1;
848 if (tab->first == NULL)
851 tab->last->next = entry;
858 /* Get the number of bytes in a strtab. */
861 _bfd_stringtab_size (struct bfd_strtab_hash *tab)
866 /* Write out a strtab. ABFD must already be at the right location in
870 _bfd_stringtab_emit (bfd *abfd, struct bfd_strtab_hash *tab)
873 struct strtab_hash_entry *entry;
877 for (entry = tab->first; entry != NULL; entry = entry->next)
882 str = entry->root.string;
883 len = strlen (str) + 1;
889 /* The output length includes the null byte. */
890 bfd_put_16 (abfd, (bfd_vma) len, buf);
891 if (bfd_bwrite ((void *) buf, (bfd_size_type) 2, abfd) != 2)
895 if (bfd_bwrite ((void *) str, (bfd_size_type) len, abfd) != len)