/* ELF strtab with GC and suffix merging support.
- Copyright 2001, 2002 Free Software Foundation, Inc.
+ Copyright (C) 2001-2019 Free Software Foundation, Inc.
Written by Jakub Jelinek <jakub@redhat.com>.
This file is part of BFD, the Binary File Descriptor library.
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
- the Free Software Foundation; either version 2 of the License, or
+ the Free Software Foundation; either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
- Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */
+ Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston,
+ MA 02110-1301, USA. */
-#include "bfd.h"
#include "sysdep.h"
+#include "bfd.h"
#include "libbfd.h"
#include "elf-bfd.h"
#include "hashtab.h"
struct elf_strtab_hash_entry
{
struct bfd_hash_entry root;
- /* Length of this entry. */
- unsigned int len;
+ /* Length of this entry. This includes the zero terminator. */
+ int len;
unsigned int refcount;
union {
/* Index within the merged section. */
bfd_size_type index;
- /* Entry this is a suffix of (if len is 0). */
+ /* Entry this is a suffix of (if len < 0). */
struct elf_strtab_hash_entry *suffix;
- struct elf_strtab_hash_entry *next;
} u;
};
{
struct bfd_hash_table table;
/* Next available index. */
- bfd_size_type size;
+ size_t size;
/* Number of array entries alloced. */
- bfd_size_type alloced;
+ size_t alloced;
/* Final strtab size. */
bfd_size_type sec_size;
/* Array of pointers to strtab entries. */
/* Allocate the structure if it has not already been allocated by a
subclass. */
if (entry == NULL)
- entry = bfd_hash_allocate (table, sizeof (struct elf_strtab_hash_entry));
+ entry = (struct bfd_hash_entry *)
+ bfd_hash_allocate (table, sizeof (struct elf_strtab_hash_entry));
if (entry == NULL)
return NULL;
struct elf_strtab_hash *table;
bfd_size_type amt = sizeof (struct elf_strtab_hash);
- table = bfd_malloc (amt);
+ table = (struct elf_strtab_hash *) bfd_malloc (amt);
if (table == NULL)
return NULL;
- if (! bfd_hash_table_init (&table->table, elf_strtab_hash_newfunc))
+ if (!bfd_hash_table_init (&table->table, elf_strtab_hash_newfunc,
+ sizeof (struct elf_strtab_hash_entry)))
{
free (table);
return NULL;
table->size = 1;
table->alloced = 64;
amt = sizeof (struct elf_strtab_hasn_entry *);
- table->array = bfd_malloc (table->alloced * amt);
+ table->array = ((struct elf_strtab_hash_entry **)
+ bfd_malloc (table->alloced * amt));
if (table->array == NULL)
{
free (table);
/* Get the index of an entity in a hash table, adding it if it is not
already present. */
-bfd_size_type
+size_t
_bfd_elf_strtab_add (struct elf_strtab_hash *tab,
const char *str,
bfd_boolean copy)
bfd_hash_lookup (&tab->table, str, TRUE, copy);
if (entry == NULL)
- return (bfd_size_type) -1;
+ return (size_t) -1;
entry->refcount++;
if (entry->len == 0)
{
entry->len = strlen (str) + 1;
+ /* 2G strings lose. */
+ BFD_ASSERT (entry->len > 0);
if (tab->size == tab->alloced)
{
bfd_size_type amt = sizeof (struct elf_strtab_hash_entry *);
tab->alloced *= 2;
- tab->array = bfd_realloc (tab->array, tab->alloced * amt);
+ tab->array = (struct elf_strtab_hash_entry **)
+ bfd_realloc_or_free (tab->array, tab->alloced * amt);
if (tab->array == NULL)
- return (bfd_size_type) -1;
+ return (size_t) -1;
}
entry->u.index = tab->size++;
}
void
-_bfd_elf_strtab_addref (struct elf_strtab_hash *tab, bfd_size_type idx)
+_bfd_elf_strtab_addref (struct elf_strtab_hash *tab, size_t idx)
{
- if (idx == 0 || idx == (bfd_size_type) -1)
+ if (idx == 0 || idx == (size_t) -1)
return;
BFD_ASSERT (tab->sec_size == 0);
BFD_ASSERT (idx < tab->size);
}
void
-_bfd_elf_strtab_delref (struct elf_strtab_hash *tab, bfd_size_type idx)
+_bfd_elf_strtab_delref (struct elf_strtab_hash *tab, size_t idx)
{
- if (idx == 0 || idx == (bfd_size_type) -1)
+ if (idx == 0 || idx == (size_t) -1)
return;
BFD_ASSERT (tab->sec_size == 0);
BFD_ASSERT (idx < tab->size);
--tab->array[idx]->refcount;
}
+unsigned int
+_bfd_elf_strtab_refcount (struct elf_strtab_hash *tab, size_t idx)
+{
+ return tab->array[idx]->refcount;
+}
+
void
_bfd_elf_strtab_clear_all_refs (struct elf_strtab_hash *tab)
{
- bfd_size_type idx;
+ size_t idx;
- for (idx = 1; idx < tab->size; ++idx)
+ for (idx = 1; idx < tab->size; idx++)
tab->array[idx]->refcount = 0;
}
+/* Save strtab refcounts prior to adding --as-needed library. */
+
+struct strtab_save
+{
+ size_t size;
+ unsigned int refcount[1];
+};
+
+void *
+_bfd_elf_strtab_save (struct elf_strtab_hash *tab)
+{
+ struct strtab_save *save;
+ size_t idx, size;
+
+ size = sizeof (*save) + (tab->size - 1) * sizeof (save->refcount[0]);
+ save = bfd_malloc (size);
+ if (save == NULL)
+ return save;
+
+ save->size = tab->size;
+ for (idx = 1; idx < tab->size; idx++)
+ save->refcount[idx] = tab->array[idx]->refcount;
+ return save;
+}
+
+/* Restore strtab refcounts on finding --as-needed library not needed. */
+
+void
+_bfd_elf_strtab_restore (struct elf_strtab_hash *tab, void *buf)
+{
+ size_t idx, curr_size = tab->size;
+ struct strtab_save *save = (struct strtab_save *) buf;
+
+ BFD_ASSERT (tab->sec_size == 0);
+ BFD_ASSERT (save->size <= curr_size);
+ tab->size = save->size;
+ for (idx = 1; idx < save->size; ++idx)
+ tab->array[idx]->refcount = save->refcount[idx];
+
+ for (; idx < curr_size; ++idx)
+ {
+ /* We don't remove entries from the hash table, just set their
+ REFCOUNT to zero. Setting LEN zero will result in the size
+ growing if the entry is added again. See _bfd_elf_strtab_add. */
+ tab->array[idx]->refcount = 0;
+ tab->array[idx]->len = 0;
+ }
+}
+
bfd_size_type
_bfd_elf_strtab_size (struct elf_strtab_hash *tab)
{
}
bfd_size_type
-_bfd_elf_strtab_offset (struct elf_strtab_hash *tab, bfd_size_type idx)
+_bfd_elf_strtab_len (struct elf_strtab_hash *tab)
+{
+ return tab->size;
+}
+
+bfd_size_type
+_bfd_elf_strtab_offset (struct elf_strtab_hash *tab, size_t idx)
{
struct elf_strtab_hash_entry *entry;
return tab->array[idx]->u.index;
}
+const char *
+_bfd_elf_strtab_str (struct elf_strtab_hash *tab, size_t idx,
+ bfd_size_type *offset)
+{
+ if (idx == 0)
+ return 0;
+ BFD_ASSERT (idx < tab->size);
+ BFD_ASSERT (tab->sec_size);
+ if (offset)
+ *offset = tab->array[idx]->u.index;
+ return tab->array[idx]->root.string;
+}
+
bfd_boolean
_bfd_elf_strtab_emit (register bfd *abfd, struct elf_strtab_hash *tab)
{
- bfd_size_type off = 1, i;
+ bfd_size_type off = 1;
+ size_t i;
if (bfd_bwrite ("", 1, abfd) != 1)
return FALSE;
for (i = 1; i < tab->size; ++i)
{
register const char *str;
- register size_t len;
+ register unsigned int len;
- str = tab->array[i]->root.string;
- len = tab->array[i]->len;
BFD_ASSERT (tab->array[i]->refcount == 0);
- if (len == 0)
+ len = tab->array[i]->len;
+ if ((int) len < 0)
continue;
+ str = tab->array[i]->root.string;
if (bfd_bwrite (str, len, abfd) != len)
return FALSE;
return TRUE;
}
-/* Compare two elf_strtab_hash_entry structures. This is called via qsort. */
+/* Compare two elf_strtab_hash_entry structures. Called via qsort.
+ Won't ever return zero as all entries differ, so there is no issue
+ with qsort stability here. */
static int
-cmplengthentry (const void *a, const void *b)
+strrevcmp (const void *a, const void *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;
+ unsigned int lenA = A->len;
+ unsigned int lenB = B->len;
+ const unsigned char *s = (const unsigned char *) A->root.string + lenA - 1;
+ const unsigned char *t = (const unsigned char *) B->root.string + lenB - 1;
+ int l = lenA < lenB ? lenA : lenB;
- if (A->len < B->len)
- return 1;
- else if (A->len > B->len)
- return -1;
-
- return memcmp (A->root.string, B->root.string, A->len);
+ while (l)
+ {
+ if (*s != *t)
+ return (int) *s - (int) *t;
+ s--;
+ t--;
+ l--;
+ }
+ return lenA - lenB;
}
-static int
-last4_eq (const void *a, const void *b)
+static inline int
+is_suffix (const struct elf_strtab_hash_entry *A,
+ const struct elf_strtab_hash_entry *B)
{
- const struct elf_strtab_hash_entry *A = a;
- const struct elf_strtab_hash_entry *B = b;
-
- if (memcmp (A->root.string + A->len - 5, B->root.string + B->len - 5, 4)
- != 0)
- /* This was a hashtable collision. */
- return 0;
-
if (A->len <= B->len)
/* B cannot be a suffix of A unless A is equal to B, which is guaranteed
not to be equal by the hash table. */
return 0;
return memcmp (A->root.string + (A->len - B->len),
- B->root.string, B->len - 5) == 0;
+ B->root.string, B->len - 1) == 0;
}
/* This function assigns final string table offsets for used strings,
void
_bfd_elf_strtab_finalize (struct elf_strtab_hash *tab)
{
- struct elf_strtab_hash_entry **array, **a, **end, *e;
- htab_t last4tab = NULL;
- bfd_size_type size, amt;
- struct elf_strtab_hash_entry *last[256], **last_ptr[256];
-
- /* GCC 2.91.66 (egcs-1.1.2) on i386 miscompiles this function when i is
- a 64-bit bfd_size_type: a 64-bit target or --enable-64-bit-bfd.
- Besides, indexing with a long long wouldn't give anything but extra
- cycles. */
- size_t i;
-
- /* Now sort the strings by length, longest first. */
- array = NULL;
- amt = tab->size * sizeof (struct elf_strtab_hash_entry *);
- array = bfd_malloc (amt);
+ struct elf_strtab_hash_entry **array, **a, *e;
+ bfd_size_type amt, sec_size;
+ size_t size, i;
+
+ /* Sort the strings by suffix and length. */
+ amt = tab->size;
+ amt *= sizeof (struct elf_strtab_hash_entry *);
+ array = (struct elf_strtab_hash_entry **) bfd_malloc (amt);
if (array == NULL)
goto alloc_failure;
- memset (last, 0, sizeof (last));
- for (i = 0; i < 256; ++i)
- last_ptr[i] = &last[i];
for (i = 1, a = array; i < tab->size; ++i)
- if (tab->array[i]->refcount)
- *a++ = tab->array[i];
- else
- tab->array[i]->len = 0;
+ {
+ e = tab->array[i];
+ if (e->refcount)
+ {
+ *a++ = e;
+ /* Adjust the length to not include the zero terminator. */
+ e->len -= 1;
+ }
+ else
+ e->len = 0;
+ }
size = a - array;
+ if (size != 0)
+ {
+ qsort (array, size, sizeof (struct elf_strtab_hash_entry *), strrevcmp);
- qsort (array, size, sizeof (struct elf_strtab_hash_entry *), cmplengthentry);
+ /* Loop over the sorted array and merge suffixes. Start from the
+ end because we want eg.
- last4tab = htab_create_alloc (size * 4, NULL, last4_eq, NULL, calloc, free);
- if (last4tab == NULL)
- goto alloc_failure;
+ s1 -> "d"
+ s2 -> "bcd"
+ s3 -> "abcd"
- /* Now insert the strings into hash tables (strings with last 4 characters
- 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;
- void **p;
-
- e = *a;
- if (e->len > 4)
- {
- 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;
+ to end up as
- ent = *p;
- e->u.suffix = ent;
- e->len = 0;
- continue;
- }
- else
- *p = e;
- }
- else
- {
- struct elf_strtab_hash_entry *tem;
+ s3 -> "abcd"
+ s2 _____^
+ s1 _______^
- c = e->root.string[e->len - 2] & 0xff;
+ ie. we don't want s1 pointing into the old s2. */
+ e = *--a;
+ e->len += 1;
+ while (--a >= array)
+ {
+ struct elf_strtab_hash_entry *cmp = *a;
- for (tem = last[c]; tem; tem = tem->u.next)
- if (tem->len > e->len
- && memcmp (tem->root.string + (tem->len - e->len),
- e->root.string, e->len - 1) == 0)
- break;
- if (tem)
+ cmp->len += 1;
+ if (is_suffix (e, cmp))
{
- e->u.suffix = tem;
- e->len = 0;
- continue;
+ cmp->u.suffix = e;
+ cmp->len = -cmp->len;
}
+ else
+ e = cmp;
}
-
- 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;
}
alloc_failure:
if (array)
free (array);
- if (last4tab)
- htab_delete (last4tab);
- /* Now assign positions to the strings we want to keep. */
- size = 1;
+ /* Assign positions to the strings we want to keep. */
+ sec_size = 1;
for (i = 1; i < tab->size; ++i)
{
e = tab->array[i];
- if (e->refcount && e->len)
+ if (e->refcount && e->len > 0)
{
- e->u.index = size;
- size += e->len;
+ e->u.index = sec_size;
+ sec_size += e->len;
}
}
- tab->sec_size = size;
+ tab->sec_size = sec_size;
- /* And now adjust the rest. */
+ /* Adjust the rest. */
for (i = 1; i < tab->size; ++i)
{
e = tab->array[i];
- if (e->refcount && ! e->len)
- e->u.index = e->u.suffix->u.index
- + (e->u.suffix->len - strlen (e->root.string) - 1);
+ if (e->refcount && e->len < 0)
+ e->u.index = e->u.suffix->u.index + (e->u.suffix->len + e->len);
}
}