/* This file is is generated by a shell script. DO NOT EDIT! */
/* AIX emulation code for ${EMULATION_NAME}
- Copyright 1991, 1993, 1995, 1996, 1997, 1998, 2000
- Free Software Foundation, Inc.
+ Copyright (C) 1991, 93, 95, 96, 97, 1998 Free Software Foundation, Inc.
Written by Steve Chamberlain <sac@cygnus.com>
AIX support by Ian Lance Taylor <ian@cygnus.com>
+ AIX 64 bit support by Tom Rix <trix@redhat.com>
This file is part of GLD, the Gnu Linker.
#include "ld.h"
#include "ldmain.h"
+#include "ldemul.h"
+#include "ldfile.h"
#include "ldmisc.h"
#include "ldexp.h"
#include "ldlang.h"
-#include "ldfile.h"
-#include "ldemul.h"
#include "ldctor.h"
#include "ldgram.h"
+#include "coff/internal.h"
+#include "coff/xcoff.h"
+
static void gld${EMULATION_NAME}_before_parse PARAMS ((void));
static int gld${EMULATION_NAME}_parse_args PARAMS ((int, char **));
static void gld${EMULATION_NAME}_after_open PARAMS ((void));
static struct export_symbol_list *export_symbols;
+/* Maintains the 32 or 64 bit mode state of import file */
+static unsigned int symbol_mode = 0x04;
+
/* This routine is called before anything else is done. */
static void
ldfile_output_architecture = bfd_arch_${ARCH};
#endif /* not TARGET_ */
config.has_shared = true;
+
+ /*
+ * The link_info.[init|fini]_functions are initialized in ld/lexsup.c.
+ * Override them here so we can use the link_info.init_function as a
+ * state flag that lets the backend know that -binitfini has been done.
+ */
+ link_info.init_function = NULL;
+ link_info.fini_function = NULL;
+
}
/* Handle AIX specific options. */
int indx;
int longind;
int optc;
- long val;
+ bfd_signed_vma val;
char *end;
-#define OPTION_IGNORE (300)
-#define OPTION_AUTOIMP (OPTION_IGNORE + 1)
-#define OPTION_ERNOTOK (OPTION_AUTOIMP + 1)
-#define OPTION_EROK (OPTION_ERNOTOK + 1)
-#define OPTION_EXPORT (OPTION_EROK + 1)
-#define OPTION_IMPORT (OPTION_EXPORT + 1)
-#define OPTION_LOADMAP (OPTION_IMPORT + 1)
-#define OPTION_MAXDATA (OPTION_LOADMAP + 1)
-#define OPTION_MAXSTACK (OPTION_MAXDATA + 1)
-#define OPTION_MODTYPE (OPTION_MAXSTACK + 1)
-#define OPTION_NOAUTOIMP (OPTION_MODTYPE + 1)
-#define OPTION_NOSTRCMPCT (OPTION_NOAUTOIMP + 1)
-#define OPTION_PD (OPTION_NOSTRCMPCT + 1)
-#define OPTION_PT (OPTION_PD + 1)
-#define OPTION_STRCMPCT (OPTION_PT + 1)
-#define OPTION_UNIX (OPTION_STRCMPCT + 1)
-
- static struct option longopts[] = {
+ enum {
+ OPTION_IGNORE = 300,
+ OPTION_AUTOIMP,
+ OPTION_ERNOTOK,
+ OPTION_EROK,
+ OPTION_EXPORT,
+ OPTION_IMPORT,
+ OPTION_INITFINI,
+ OPTION_LOADMAP,
+ OPTION_MAXDATA,
+ OPTION_MAXSTACK,
+ OPTION_MODTYPE,
+ OPTION_NOAUTOIMP,
+ OPTION_NOSTRCMPCT,
+ OPTION_PD,
+ OPTION_PT,
+ OPTION_STRCMPCT,
+ OPTION_UNIX
+ };
+
+ /*
+ b64 is an empty option. The native linker uses -b64 for xcoff64 support
+ Our linker uses -m aixppc64 for xcoff64 support. The choice for the
+ correct emulation is done in collect2.c via the environmental varible
+ LDEMULATION.
+
+ binitfini has special handling in the linker backend. The native linker
+ uses the arguemnts to generate a table of init and fini functions for
+ the executable. The important use for this option is to support aix 4.2+
+ c++ constructors and destructors. This is tied into gcc via collect2.c.
+ The function table is accessed by the runtime linker/loader by checking if
+ the first symbol in the loader symbol table is "__rtinit". The native
+ linker generates this table and the loader symbol. The gnu linker looks
+ for the symbol "__rtinit" and makes it the first loader symbol. It is the
+ responsiblity of the user to define the __rtinit symbol. The format for
+ __rtinit is given by the aix system file /usr/include/rtinit.h. You can
+ look at collect2.c to see an example of how this is done for 32 and 64 bit.
+ Below is an exmaple of a 32 bit assembly file that defines __rtinit.
+
+ .file "my_rtinit.s"
+
+ .csect .data[RW],3
+ .globl __rtinit
+ .extern init_function
+ .extern fini_function
+
+ __rtinit:
+ .long 0
+ .long f1i - __rtinit
+ .long f1f - __rtinit
+ .long f2i - f1i
+ .align 3
+ f1i: .long init_function
+ .long s1i - __rtinit
+ .long 0
+ f2i: .long 0
+ .long 0
+ .long 0
+ f1f: .long fini_function
+ .long s1f - __rtinit
+ .long 0
+ f2f: .long 0
+ .long 0
+ .long 0
+ .align 3
+ s1i: .string "init_function"
+ .align 3
+ s1f: .string "fini_function"
+
+ */
+
+ static const struct option longopts[] = {
{"basis", no_argument, NULL, OPTION_IGNORE},
{"bautoimp", no_argument, NULL, OPTION_AUTOIMP},
{"bcomprld", no_argument, NULL, OPTION_IGNORE},
{"bhalt", required_argument, NULL, OPTION_IGNORE},
{"bI", required_argument, NULL, OPTION_IMPORT},
{"bimport", required_argument, NULL, OPTION_IMPORT},
+ {"binitfini", required_argument, NULL, OPTION_INITFINI},
{"bl", required_argument, NULL, OPTION_LOADMAP},
{"bloadmap", required_argument, NULL, OPTION_LOADMAP},
{"bmaxdata", required_argument, NULL, OPTION_MAXDATA},
{"bso", no_argument, NULL, OPTION_AUTOIMP},
{"bstrcmpct", no_argument, NULL, OPTION_STRCMPCT},
{"btextro", no_argument, &textro, 1},
+ {"b64", no_argument, NULL, 0},
{"static", no_argument, NULL, OPTION_NOAUTOIMP},
{"unix", no_argument, NULL, OPTION_UNIX},
{NULL, no_argument, NULL, 0}
The AIX linker uses : to separate the option from the argument;
changing it to = lets us treat it as a getopt option. */
indx = optind;
- if (indx == 0)
+ if (indx == 0) {
indx = 1;
- if (indx < argc && strncmp (argv[indx], "-b", 2) == 0)
- {
- char *s;
-
- for (s = argv[indx]; *s != '\0'; s++)
- {
- if (*s == ':')
- {
- *s = '=';
- break;
- }
- }
+ }
+
+ if (indx < argc && strncmp (argv[indx], "-b", 2) == 0) {
+ char *s;
+
+ for (s = argv[indx]; *s != '\0'; s++) {
+ if (*s == ':') {
+ *s = '=';
+ break;
+ }
}
+ }
+
/* We add s and u so to the short options list so that -s and -u on
the command line do not match -static and -unix. */
break;
case 'D':
- val = strtol (optarg, &end, 0);
+ val = strtoll (optarg, &end, 0);
if (*end != '\0')
einfo ("%P: warning: ignoring invalid -D number %s\n", optarg);
else if (val != -1)
number, we assume the AIX option is intended. Otherwise, we
assume the usual GNU ld -T option is intended. We can't just
ignore the AIX option, because gcc passes it to the linker. */
- val = strtoul (optarg, &end, 0);
+ val = strtoull (optarg, &end, 0);
if (*end != '\0')
{
optind = prevoptind;
case OPTION_IGNORE:
break;
+ case OPTION_INITFINI: {
+ /*
+ * The aix linker init fini has the format :
+ *
+ * -binitfini:[ Initial][:Termination][:Priority]
+ *
+ * it allows the Termination and Priority to be optional.
+ *
+ * Since we support only one init/fini pair, we ignore the Priority.
+ *
+ * Define the special symbol __rtinit.
+ *
+ * strtok does not correctly handle the case of -binitfini::fini: so
+ * do it by hand
+ */
+ char *t, *i, *f;
+
+ i = t = optarg;
+ while (*t && ':' != *t) t++;
+ if (*t) *t++ = 0;
+
+ if (0 != strlen(i)) {
+ link_info.init_function = i;
+ }
+
+ f = t;
+ while (*t && ':' != *t) t++;
+ *t = 0;
+
+ if (0 != strlen(f)) {
+ link_info.fini_function = f;
+ }
+ }
+ break;
+
case OPTION_AUTOIMP:
link_info.static_link = false;
break;
break;
case OPTION_MAXDATA:
- val = strtoul (optarg, &end, 0);
+ val = strtoull (optarg, &end, 0);
if (*end != '\0')
einfo ("%P: warning: ignoring invalid -bmaxdata number %s\n",
optarg);
break;
case OPTION_MAXSTACK:
- val = strtoul (optarg, &end, 0);
+ val = strtoull (optarg, &end, 0);
if (*end != '\0')
einfo ("%P: warning: ignoring invalid -bmaxstack number %s\n",
optarg);
start on. The offset within the page should still be the
offset within the file, so we need to build an appropriate
expression. */
- val = strtoul (optarg, &end, 0);
+ val = strtoull (optarg, &end, 0);
if (*end != '\0')
einfo ("%P: warning: ignoring invalid -pD number %s\n", optarg);
else
/* This set the page that the .text section is supposed to start
on. The offset within the page should still be the offset
within the file. */
- val = strtoul (optarg, &end, 0);
+ val = strtoull (optarg, &end, 0);
if (*end != '\0')
einfo ("%P: warning: ignoring invalid -pT number %s\n", optarg);
else
struct filelist *fl;
struct export_symbol_list *el;
char *libpath;
- asection *special_sections[6];
+ asection *special_sections[XCOFF_NUMBER_OF_SPECIAL_SECTIONS];
int i;
/* Handle the import and export files, if any. */
for (fl = import_files; fl != NULL; fl = fl->next)
gld${EMULATION_NAME}_read_file (fl->name, true);
- for (el = export_symbols; el != NULL; el = el->next)
- {
- struct bfd_link_hash_entry *h;
-
- h = bfd_link_hash_lookup (link_info.hash, el->name, false, false, false);
- if (h == NULL)
- einfo ("%P%F: bfd_link_hash_lookup of export symbol failed: %E\n");
- if (! bfd_xcoff_export_symbol (output_bfd, &link_info, h, el->syscall))
- einfo ("%P%F: bfd_xcoff_export_symbol failed: %E\n");
- }
+ for (el = export_symbols; el != NULL; el = el->next) {
+ struct bfd_link_hash_entry *h;
+
+ h = bfd_link_hash_lookup (link_info.hash, el->name, false, false, false);
+ if (h == NULL)
+ einfo ("%P%F: bfd_link_hash_lookup of export symbol failed: %E\n");
+ if (! bfd_xcoff_export_symbol (output_bfd, &link_info, h, el->syscall))
+ einfo ("%P%F: bfd_xcoff_export_symbol failed: %E\n");
+ }
/* Track down all relocations called for by the linker script (these
are typically constructor/destructor entries created by
/* Look through the special sections, and put them in the right
place in the link ordering. This is especially magic. */
- for (i = 0; i < 6; i++)
- {
- asection *sec;
- lang_output_section_statement_type *os;
- lang_statement_union_type **pls;
- lang_input_section_type *is;
- const char *oname;
- boolean start;
-
- sec = special_sections[i];
- if (sec == NULL)
- continue;
+ for (i = 0; i < XCOFF_NUMBER_OF_SPECIAL_SECTIONS; i++) {
+ asection *sec;
+ lang_output_section_statement_type *os;
+ lang_statement_union_type **pls;
+ lang_input_section_type *is;
+ const char *oname;
+ boolean start;
+
+ sec = special_sections[i];
+ if (sec == NULL)
+ continue;
+
+ /* Remove this section from the list of the output section.
+ This assumes we know what the script looks like. */
+ is = NULL;
+ os = lang_output_section_find (sec->output_section->name);
+ if (os == NULL) {
+ einfo ("%P%F: can't find output section %s\n",
+ sec->output_section->name);
+ }
- /* Remove this section from the list of the output section.
- This assumes we know what the script looks like. */
- is = NULL;
- os = lang_output_section_find (sec->output_section->name);
- if (os == NULL)
- einfo ("%P%F: can't find output section %s\n",
- sec->output_section->name);
- for (pls = &os->children.head; *pls != NULL; pls = &(*pls)->next)
- {
- if ((*pls)->header.type == lang_input_section_enum
- && (*pls)->input_section.section == sec)
- {
- is = (lang_input_section_type *) *pls;
- *pls = (*pls)->next;
+ for (pls = &os->children.head; *pls != NULL; pls = &(*pls)->next) {
+
+ if ((*pls)->header.type == lang_input_section_enum &&
+ (*pls)->input_section.section == sec) {
+ is = (lang_input_section_type *) *pls;
+ *pls = (*pls)->next;
+ break;
+ }
+
+ if ((*pls)->header.type == lang_wild_statement_enum) {
+ lang_statement_union_type **pwls;
+
+ for (pwls = &(*pls)->wild_statement.children.head;
+ *pwls != NULL;
+ pwls = &(*pwls)->next) {
+
+ if ((*pwls)->header.type == lang_input_section_enum &&
+ (*pwls)->input_section.section == sec) {
+
+ is = (lang_input_section_type *) *pwls;
+ *pwls = (*pwls)->next;
break;
- }
- if ((*pls)->header.type == lang_wild_statement_enum)
- {
- lang_statement_union_type **pwls;
+ }
+ }
+
+ if (is != NULL)
+ break;
+ }
+ }
- for (pwls = &(*pls)->wild_statement.children.head;
- *pwls != NULL;
- pwls = &(*pwls)->next)
- {
- if ((*pwls)->header.type == lang_input_section_enum
- && (*pwls)->input_section.section == sec)
- {
- is = (lang_input_section_type *) *pwls;
- *pwls = (*pwls)->next;
- break;
- }
- }
- if (is != NULL)
- break;
- }
- }
+ if (is == NULL) {
+ einfo ("%P%F: can't find %s in output section\n",
+ bfd_get_section_name (sec->owner, sec));
+ }
+
+ /* Now figure out where the section should go. */
+ switch (i) {
+
+ default: /* to avoid warnings */
+ case XCOFF_SPECIAL_SECTION_TEXT:
+ /* _text */
+ oname = ".text";
+ start = true;
+ break;
+
+ case XCOFF_SPECIAL_SECTION_ETEXT:
+ /* _etext */
+ oname = ".text";
+ start = false;
+ break;
- if (is == NULL)
- einfo ("%P%F: can't find %s in output section\n",
- bfd_get_section_name (sec->owner, sec));
+ case XCOFF_SPECIAL_SECTION_DATA:
+ /* _data */
+ oname = ".data";
+ start = true;
+ break;
+
+ case XCOFF_SPECIAL_SECTION_EDATA:
+ /* _edata */
+ oname = ".data";
+ start = false;
+ break;
- /* Now figure out where the section should go. */
- switch (i)
- {
- default: /* to avoid warnings */
- case 0:
- /* _text */
- oname = ".text";
- start = true;
- break;
- case 1:
- /* _etext */
- oname = ".text";
- start = false;
- break;
- case 2:
- /* _data */
- oname = ".data";
- start = true;
- break;
- case 3:
- /* _edata */
- oname = ".data";
- start = false;
- break;
- case 4:
- case 5:
- /* _end and end */
- oname = ".bss";
- start = false;
- break;
- }
+ case XCOFF_SPECIAL_SECTION_END:
+ case XCOFF_SPECIAL_SECTION_END2:
+ /* _end and end */
+ oname = ".bss";
+ start = false;
+ break;
+ }
- os = lang_output_section_find (oname);
+ os = lang_output_section_find (oname);
- if (start)
- {
- is->header.next = os->children.head;
- os->children.head = (lang_statement_union_type *) is;
- }
- else
- {
- is->header.next = NULL;
- lang_statement_append (&os->children,
- (lang_statement_union_type *) is,
- &is->header.next);
- }
+ if (start) {
+ is->header.next = os->children.head;
+ os->children.head = (lang_statement_union_type *) is;
+ } else {
+ is->header.next = NULL;
+ lang_statement_append (&os->children,
+ (lang_statement_union_type *) is,
+ &is->header.next);
+ }
+ }
+}
+
+static int change_symbol_mode (char *input)
+{
+ /*
+ * 1 : state changed
+ * 0 : no change
+ */
+
+ char *symbol_mode_string[] = {
+ "# 32", /* 0x01 */
+ "# 64", /* 0x02 */
+ "# no32", /* 0x04 */
+ "# no64", /* 0x08 */
+ NULL,
+ };
+ unsigned int bit;
+ char *string;
+
+ for (bit = 0; ; bit++) {
+ string = symbol_mode_string[bit];
+ if (NULL == string)
+ return 0;
+
+ if (0 == strcmp(input, string)) {
+ symbol_mode = (1 << bit);
+ return 1;
+ }
+ }
+ /* should not be here */
+ return 0;
+}
+
+
+static int is_syscall(char *input)
+{
+ /*
+ * 1 : yes
+ * 0 : ignore
+ * -1 : error, try something else
+ */
+ unsigned int bit;
+ char *string;
+
+ char *syscall_string[] = {
+ "svc", /* 0x01 */
+ "svc32", /* 0x02 */
+ "svc3264", /* 0x04 */
+ "svc64", /* 0x08 */
+ "syscall", /* 0x10 */
+ "syscall32", /* 0x20 */
+ "syscall3264", /* 0x40 */
+ "syscall64", /* 0x80 */
+ NULL
+ };
+
+ for (bit = 0; ;bit++) {
+
+ string = syscall_string[bit];
+ if (NULL == string) {
+ return -1;
+ }
+
+ if (0 == strcmp(input, string)) {
+ if (1 << bit & ${SYSCALL_MASK}) {
+ return 1;
+ } else {
+ return 0;
+ }
}
+ }
+ /* should not be here */
+ return -1;
}
+
/* Read an import or export file. For an import file, this is called
by the before_allocation emulation routine. For an export file,
this is called by the parse_args emulation routine. */
impmember = NULL;
lineno = 0;
+
+ /*
+ * default to 32 and 64 bit mode
+ * symbols at top of /lib/syscalls.exp do not have a mode modifier and they
+ * are not repeated, assume 64 bit routines also want to use them.
+ * See the routine change_symbol_mode for more information.
+ */
+ symbol_mode = 0x04;
+
while ((c = getc (f)) != EOF)
{
char *s;
s = (char *) obstack_base (o);
while (isspace ((unsigned char) *s))
++s;
- if (*s == '\0'
- || *s == '*'
- || (*s == '#' && s[1] == ' ')
- || (! import && *s == '#' && s[1] == '!'))
+ if (*s == '\0' ||
+ *s == '*' ||
+ change_symbol_mode (s) ||
+ (*s == '#' && s[1] == ' ') ||
+ (! import && *s == '#' && s[1] == '!'))
{
obstack_free (o, obstack_base (o));
continue;
continue;
}
- /* This is a symbol to be imported or exported. */
- symname = s;
- syscall = false;
- address = (bfd_vma) -1;
-
- while (! isspace ((unsigned char) *s) && *s != '\0')
- ++s;
- if (*s != '\0')
- {
- char *se;
-
- *s++ = '\0';
-
- while (isspace ((unsigned char) *s))
- ++s;
-
- se = s;
- while (! isspace ((unsigned char) *se) && *se != '\0')
- ++se;
- if (*se != '\0')
- {
- *se++ = '\0';
- while (isspace ((unsigned char) *se))
- ++se;
- if (*se != '\0')
- einfo ("%s%d: warning: syntax error in import/export file\n",
- filename, lineno);
- }
-
- if (s == se)
- {
- /* There was no address after all. */
- }
- else if (strcasecmp (s, "svc") == 0
- || strcasecmp (s, "syscall") == 0)
- syscall = true;
- else
- {
+ if (symbol_mode & ${SYMBOL_MODE_MASK}) {
+
+ /* This is a symbol to be imported or exported. */
+ symname = s;
+ syscall = false;
+ address = (bfd_vma) -1;
+
+ while (! isspace ((unsigned char) *s) && *s != '\0')
+ ++s;
+ if (*s != '\0')
+ {
+ char *se;
+
+ *s++ = '\0';
+
+ while (isspace ((unsigned char) *s))
+ ++s;
+
+ se = s;
+ while (! isspace ((unsigned char) *se) && *se != '\0')
+ ++se;
+ if (*se != '\0')
+ {
+ *se++ = '\0';
+ while (isspace ((unsigned char) *se))
+ ++se;
+ if (*se != '\0')
+ einfo ("%s%d: warning: syntax error in import/export file\n",
+ filename, lineno);
+ }
+
+ if (s != se) {
+ int status;
char *end;
-
- address = strtoul (s, &end, 0);
- if (*end != '\0')
- einfo ("%s:%d: warning: syntax error in import/export file\n",
- filename, lineno);
- }
- }
-
- if (! import)
- {
- struct export_symbol_list *n;
-
- ldlang_add_undef (symname);
- n = ((struct export_symbol_list *)
- xmalloc (sizeof (struct export_symbol_list)));
- n->next = export_symbols;
- n->name = xstrdup (symname);
- n->syscall = syscall;
- export_symbols = n;
- }
- else
- {
- h = bfd_link_hash_lookup (link_info.hash, symname, false, false,
- true);
- if (h == NULL || h->type == bfd_link_hash_new)
- {
- /* We can just ignore attempts to import an unreferenced
- symbol. */
- }
- else
- {
- if (! bfd_xcoff_import_symbol (output_bfd, &link_info, h,
- address, imppath, impfile,
- impmember))
- einfo ("%X%s:%d: failed to import symbol %s: %E\n",
- filename, lineno, symname);
+
+ status = is_syscall(s);
+
+ switch (status) {
+ case 1:
+ /* this is a system call */
+ syscall = true;
+ break;
+
+ case 0:
+ /* ignore this system call */
+ break;
+
+ default:
+ /* not a system call, check for address */
+ address = strtoul (s, &end, 0);
+ if (*end != '\0') {
+ einfo ("%s:%d: warning: syntax error in import/export file\n", filename, lineno);
+
+ }
+ }
}
- }
-
+ }
+
+ if (! import)
+ {
+ struct export_symbol_list *n;
+
+ ldlang_add_undef (symname);
+ n = ((struct export_symbol_list *)
+ xmalloc (sizeof (struct export_symbol_list)));
+ n->next = export_symbols;
+ n->name = xstrdup (symname);
+ n->syscall = syscall;
+ export_symbols = n;
+ }
+ else
+ {
+ h = bfd_link_hash_lookup (link_info.hash, symname, false, false,
+ true);
+ if (h == NULL || h->type == bfd_link_hash_new)
+ {
+ /* We can just ignore attempts to import an unreferenced
+ symbol. */
+ }
+ else
+ {
+ if (! bfd_xcoff_import_symbol (output_bfd, &link_info, h,
+ address, imppath, impfile,
+ impmember))
+ einfo ("%X%s:%d: failed to import symbol %s: %E\n",
+ filename, lineno, symname);
+ }
+ }
+ }
obstack_free (o, obstack_base (o));
}
-
+
if (obstack_object_size (o) > 0)
{
- einfo ("%s:%d: warning: ignoring unterminated last line\n",
- filename, lineno);
- obstack_free (o, obstack_base (o));
- }
+ einfo ("%s:%d: warning: ignoring unterminated last line\n",
+ filename, lineno);
+ obstack_free (o, obstack_base (o));
+ }
- if (! keep)
- {
- obstack_free (o, NULL);
- free (o);
+ if (! keep)
+ {
+ obstack_free (o, NULL);
+ free (o);
+ }
}
-}
/* This routine saves us from worrying about declaring free. */
# Scripts compiled in.
# sed commands to quote an ld script as a C string.
-sc="-f stringify.sed"
+sc="-f ${srcdir}/emultempl/stringify.sed"
cat >>e${EMULATION_NAME}.c <<EOF
{
0, /* set_symbols */
gld${EMULATION_NAME}_parse_args,
gld${EMULATION_NAME}_unrecognized_file,
- NULL /* find_potential_libraries */
+ NULL, /* list_options */
+ NULL, /* recognized_file */
+ NULL, /* find potential_libraries */
};
EOF
projects / binutils-gdb.git / commitdiff