diff options
Diffstat (limited to 'gas/config/obj-elf.c')
| -rw-r--r-- | gas/config/obj-elf.c | 1764 |
1 files changed, 1764 insertions, 0 deletions
diff --git a/gas/config/obj-elf.c b/gas/config/obj-elf.c new file mode 100644 index 0000000000..f50c4b7257 --- /dev/null +++ b/gas/config/obj-elf.c @@ -0,0 +1,1764 @@ +/* ELF object file format + Copyright (C) 1992, 93, 94, 95, 96, 97, 98, 1999 Free Software Foundation, Inc. + + This file is part of GAS, the GNU Assembler. + + GAS 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, + or (at your option) any later version. + + GAS is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See + the GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with GAS; see the file COPYING. If not, write to the Free + Software Foundation, 59 Temple Place - Suite 330, Boston, MA + 02111-1307, USA. */ + +#define OBJ_HEADER "obj-elf.h" +#include "as.h" +#include "subsegs.h" +#include "obstack.h" + +#ifndef ECOFF_DEBUGGING +#define ECOFF_DEBUGGING 0 +#else +#define NEED_ECOFF_DEBUG +#endif + +#ifdef NEED_ECOFF_DEBUG +#include "ecoff.h" +#endif + +#ifdef TC_ALPHA +#include "elf/alpha.h" +#endif + +#ifdef TC_MIPS +#include "elf/mips.h" +#endif + +#ifdef TC_PPC +#include "elf/ppc.h" +#endif + +static bfd_vma elf_s_get_size PARAMS ((symbolS *)); +static void elf_s_set_size PARAMS ((symbolS *, bfd_vma)); +static bfd_vma elf_s_get_align PARAMS ((symbolS *)); +static void elf_s_set_align PARAMS ((symbolS *, bfd_vma)); +static void elf_copy_symbol_attributes PARAMS ((symbolS *, symbolS *)); +static int elf_sec_sym_ok_for_reloc PARAMS ((asection *)); +static void adjust_stab_sections PARAMS ((bfd *, asection *, PTR)); + +#ifdef NEED_ECOFF_DEBUG +static boolean elf_get_extr PARAMS ((asymbol *, EXTR *)); +static void elf_set_index PARAMS ((asymbol *, bfd_size_type)); +#endif + +static void obj_elf_line PARAMS ((int)); +void obj_elf_version PARAMS ((int)); +static void obj_elf_size PARAMS ((int)); +static void obj_elf_type PARAMS ((int)); +static void obj_elf_ident PARAMS ((int)); +static void obj_elf_weak PARAMS ((int)); +static void obj_elf_local PARAMS ((int)); +static void obj_elf_common PARAMS ((int)); +static void obj_elf_symver PARAMS ((int)); +static void obj_elf_vtable_inherit PARAMS ((int)); +static void obj_elf_vtable_entry PARAMS ((int)); +static void obj_elf_data PARAMS ((int)); +static void obj_elf_text PARAMS ((int)); +static void obj_elf_subsection PARAMS ((int)); + +static const pseudo_typeS elf_pseudo_table[] = +{ + {"comm", obj_elf_common, 0}, + {"ident", obj_elf_ident, 0}, + {"local", obj_elf_local, 0}, + {"previous", obj_elf_previous, 0}, + {"section", obj_elf_section, 0}, + {"section.s", obj_elf_section, 0}, + {"sect", obj_elf_section, 0}, + {"sect.s", obj_elf_section, 0}, + {"size", obj_elf_size, 0}, + {"type", obj_elf_type, 0}, + {"version", obj_elf_version, 0}, + {"weak", obj_elf_weak, 0}, + + /* These are used for stabs-in-elf configurations. */ + {"line", obj_elf_line, 0}, + + /* This is a GNU extension to handle symbol versions. */ + {"symver", obj_elf_symver, 0}, + + /* A GNU extension to change subsection only. */ + {"subsection", obj_elf_subsection, 0}, + + /* These are GNU extensions to aid in garbage collecting C++ vtables. */ + {"vtable_inherit", obj_elf_vtable_inherit, 0}, + {"vtable_entry", obj_elf_vtable_entry, 0}, + + /* These are used for dwarf. */ + {"2byte", cons, 2}, + {"4byte", cons, 4}, + {"8byte", cons, 8}, + + /* We need to trap the section changing calls to handle .previous. */ + {"data", obj_elf_data, 0}, + {"text", obj_elf_text, 0}, + + /* End sentinel. */ + {NULL}, +}; + +static const pseudo_typeS ecoff_debug_pseudo_table[] = +{ +#ifdef NEED_ECOFF_DEBUG + /* COFF style debugging information for ECOFF. .ln is not used; .loc + is used instead. */ + { "def", ecoff_directive_def, 0 }, + { "dim", ecoff_directive_dim, 0 }, + { "endef", ecoff_directive_endef, 0 }, + { "file", ecoff_directive_file, 0 }, + { "scl", ecoff_directive_scl, 0 }, + { "tag", ecoff_directive_tag, 0 }, + { "val", ecoff_directive_val, 0 }, + + /* COFF debugging requires pseudo-ops .size and .type, but ELF + already has meanings for those. We use .esize and .etype + instead. These are only generated by gcc anyhow. */ + { "esize", ecoff_directive_size, 0 }, + { "etype", ecoff_directive_type, 0 }, + + /* ECOFF specific debugging information. */ + { "begin", ecoff_directive_begin, 0 }, + { "bend", ecoff_directive_bend, 0 }, + { "end", ecoff_directive_end, 0 }, + { "ent", ecoff_directive_ent, 0 }, + { "fmask", ecoff_directive_fmask, 0 }, + { "frame", ecoff_directive_frame, 0 }, + { "loc", ecoff_directive_loc, 0 }, + { "mask", ecoff_directive_mask, 0 }, + + /* Other ECOFF directives. */ + { "extern", ecoff_directive_extern, 0 }, + + /* These are used on Irix. I don't know how to implement them. */ + { "alias", s_ignore, 0 }, + { "bgnb", s_ignore, 0 }, + { "endb", s_ignore, 0 }, + { "lab", s_ignore, 0 }, + { "noalias", s_ignore, 0 }, + { "verstamp", s_ignore, 0 }, + { "vreg", s_ignore, 0 }, +#endif + + {NULL} /* end sentinel */ +}; + +#undef NO_RELOC +#include "aout/aout64.h" + +/* This is called when the assembler starts. */ + +void +elf_begin () +{ + /* Add symbols for the known sections to the symbol table. */ + symbol_table_insert (section_symbol (bfd_get_section_by_name (stdoutput, + TEXT_SECTION_NAME))); + symbol_table_insert (section_symbol (bfd_get_section_by_name (stdoutput, + DATA_SECTION_NAME))); + symbol_table_insert (section_symbol (bfd_get_section_by_name (stdoutput, + BSS_SECTION_NAME))); +} + +void +elf_pop_insert () +{ + pop_insert (elf_pseudo_table); + if (ECOFF_DEBUGGING) + pop_insert (ecoff_debug_pseudo_table); +} + +static bfd_vma +elf_s_get_size (sym) + symbolS *sym; +{ + return S_GET_SIZE (sym); +} + +static void +elf_s_set_size (sym, sz) + symbolS *sym; + bfd_vma sz; +{ + S_SET_SIZE (sym, sz); +} + +static bfd_vma +elf_s_get_align (sym) + symbolS *sym; +{ + return S_GET_ALIGN (sym); +} + +static void +elf_s_set_align (sym, align) + symbolS *sym; + bfd_vma align; +{ + S_SET_ALIGN (sym, align); +} + +static void +elf_copy_symbol_attributes (dest, src) + symbolS *dest, *src; +{ + OBJ_COPY_SYMBOL_ATTRIBUTES (dest, src); +} + +static int +elf_sec_sym_ok_for_reloc (sec) + asection *sec; +{ + return obj_sec_sym_ok_for_reloc (sec); +} + +void +elf_file_symbol (s) + char *s; +{ + symbolS *sym; + + sym = symbol_new (s, absolute_section, (valueT) 0, (struct frag *) 0); + sym->sy_frag = &zero_address_frag; + sym->bsym->flags |= BSF_FILE; + + if (symbol_rootP != sym) + { + symbol_remove (sym, &symbol_rootP, &symbol_lastP); + symbol_insert (sym, symbol_rootP, &symbol_rootP, &symbol_lastP); +#ifdef DEBUG + verify_symbol_chain (symbol_rootP, symbol_lastP); +#endif + } + +#ifdef NEED_ECOFF_DEBUG + ecoff_new_file (s); +#endif +} + +static void +obj_elf_common (ignore) + int ignore; +{ + char *name; + char c; + char *p; + int temp, size; + symbolS *symbolP; + int have_align; + + name = input_line_pointer; + c = get_symbol_end (); + /* just after name is now '\0' */ + p = input_line_pointer; + *p = c; + SKIP_WHITESPACE (); + if (*input_line_pointer != ',') + { + as_bad (_("Expected comma after symbol-name")); + ignore_rest_of_line (); + return; + } + input_line_pointer++; /* skip ',' */ + if ((temp = get_absolute_expression ()) < 0) + { + as_bad (_(".COMMon length (%d.) <0! Ignored."), temp); + ignore_rest_of_line (); + return; + } + size = temp; + *p = 0; + symbolP = symbol_find_or_make (name); + *p = c; + if (S_IS_DEFINED (symbolP) && ! S_IS_COMMON (symbolP)) + { + as_bad (_("Ignoring attempt to re-define symbol")); + ignore_rest_of_line (); + return; + } + if (S_GET_VALUE (symbolP) != 0) + { + if (S_GET_VALUE (symbolP) != size) + { + as_warn (_("Length of .comm \"%s\" is already %ld. Not changed to %d."), + S_GET_NAME (symbolP), (long) S_GET_VALUE (symbolP), size); + } + } + know (symbolP->sy_frag == &zero_address_frag); + if (*input_line_pointer != ',') + have_align = 0; + else + { + have_align = 1; + input_line_pointer++; + SKIP_WHITESPACE (); + } + if (! have_align || *input_line_pointer != '"') + { + if (! have_align) + temp = 0; + else + { + temp = get_absolute_expression (); + if (temp < 0) + { + temp = 0; + as_warn (_("Common alignment negative; 0 assumed")); + } + } + if (symbolP->local) + { + segT old_sec; + int old_subsec; + char *pfrag; + int align; + + /* allocate_bss: */ + old_sec = now_seg; + old_subsec = now_subseg; + if (temp) + { + /* convert to a power of 2 alignment */ + for (align = 0; (temp & 1) == 0; temp >>= 1, ++align); + if (temp != 1) + { + as_bad (_("Common alignment not a power of 2")); + ignore_rest_of_line (); + return; + } + } + else + align = 0; + record_alignment (bss_section, align); + subseg_set (bss_section, 0); + if (align) + frag_align (align, 0, 0); + if (S_GET_SEGMENT (symbolP) == bss_section) + symbolP->sy_frag->fr_symbol = 0; + symbolP->sy_frag = frag_now; + pfrag = frag_var (rs_org, 1, 1, (relax_substateT) 0, symbolP, + (offsetT) size, (char *) 0); + *pfrag = 0; + S_SET_SIZE (symbolP, size); + S_SET_SEGMENT (symbolP, bss_section); + S_CLEAR_EXTERNAL (symbolP); + subseg_set (old_sec, old_subsec); + } + else + { + allocate_common: + S_SET_VALUE (symbolP, (valueT) size); + S_SET_ALIGN (symbolP, temp); + S_SET_EXTERNAL (symbolP); + S_SET_SEGMENT (symbolP, bfd_com_section_ptr); + } + } + else + { + input_line_pointer++; + /* @@ Some use the dot, some don't. Can we get some consistency?? */ + if (*input_line_pointer == '.') + input_line_pointer++; + /* @@ Some say data, some say bss. */ + if (strncmp (input_line_pointer, "bss\"", 4) + && strncmp (input_line_pointer, "data\"", 5)) + { + while (*--input_line_pointer != '"') + ; + input_line_pointer--; + goto bad_common_segment; + } + while (*input_line_pointer++ != '"') + ; + goto allocate_common; + } + + symbolP->bsym->flags |= BSF_OBJECT; + + demand_empty_rest_of_line (); + return; + + { + bad_common_segment: + p = input_line_pointer; + while (*p && *p != '\n') + p++; + c = *p; + *p = '\0'; + as_bad (_("bad .common segment %s"), input_line_pointer + 1); + *p = c; + input_line_pointer = p; + ignore_rest_of_line (); + return; + } +} + +static void +obj_elf_local (ignore) + int ignore; +{ + char *name; + int c; + symbolS *symbolP; + + do + { + name = input_line_pointer; + c = get_symbol_end (); + symbolP = symbol_find_or_make (name); + *input_line_pointer = c; + SKIP_WHITESPACE (); + S_CLEAR_EXTERNAL (symbolP); + symbolP->local = 1; + if (c == ',') + { + input_line_pointer++; + SKIP_WHITESPACE (); + if (*input_line_pointer == '\n') + c = '\n'; + } + } + while (c == ','); + demand_empty_rest_of_line (); +} + +static void +obj_elf_weak (ignore) + int ignore; +{ + char *name; + int c; + symbolS *symbolP; + + do + { + name = input_line_pointer; + c = get_symbol_end (); + symbolP = symbol_find_or_make (name); + *input_line_pointer = c; + SKIP_WHITESPACE (); + S_SET_WEAK (symbolP); + symbolP->local = 1; + if (c == ',') + { + input_line_pointer++; + SKIP_WHITESPACE (); + if (*input_line_pointer == '\n') + c = '\n'; + } + } + while (c == ','); + demand_empty_rest_of_line (); +} + +static segT previous_section; +static int previous_subsection; + +/* Handle the .section pseudo-op. This code supports two different + syntaxes. + + The first is found on Solaris, and looks like + .section ".sec1",#alloc,#execinstr,#write + Here the names after '#' are the SHF_* flags to turn on for the + section. I'm not sure how it determines the SHT_* type (BFD + doesn't really give us control over the type, anyhow). + + The second format is found on UnixWare, and probably most SVR4 + machines, and looks like + .section .sec1,"a",@progbits + The quoted string may contain any combination of a, w, x, and + represents the SHF_* flags to turn on for the section. The string + beginning with '@' can be progbits or nobits. There should be + other possibilities, but I don't know what they are. In any case, + BFD doesn't really let us set the section type. */ + +/* Certain named sections have particular defined types, listed on p. + 4-19 of the ABI. */ +struct special_section +{ + const char *name; + int type; + int attributes; +}; + +static struct special_section special_sections[] = +{ + { ".bss", SHT_NOBITS, SHF_ALLOC + SHF_WRITE }, + { ".comment", SHT_PROGBITS, 0 }, + { ".data", SHT_PROGBITS, SHF_ALLOC + SHF_WRITE }, + { ".data1", SHT_PROGBITS, SHF_ALLOC + SHF_WRITE }, + { ".debug", SHT_PROGBITS, 0 }, + { ".fini", SHT_PROGBITS, SHF_ALLOC + SHF_EXECINSTR }, + { ".init", SHT_PROGBITS, SHF_ALLOC + SHF_EXECINSTR }, + { ".line", SHT_PROGBITS, 0 }, + { ".note", SHT_NOTE, 0 }, + { ".rodata", SHT_PROGBITS, SHF_ALLOC }, + { ".rodata1", SHT_PROGBITS, SHF_ALLOC }, + { ".text", SHT_PROGBITS, SHF_ALLOC + SHF_EXECINSTR }, + +#ifdef ELF_TC_SPECIAL_SECTIONS + ELF_TC_SPECIAL_SECTIONS +#endif + +#if 0 + /* The following section names are special, but they can not + reasonably appear in assembler code. Some of the attributes are + processor dependent. */ + { ".dynamic", SHT_DYNAMIC, SHF_ALLOC /* + SHF_WRITE */ }, + { ".dynstr", SHT_STRTAB, SHF_ALLOC }, + { ".dynsym", SHT_DYNSYM, SHF_ALLOC }, + { ".got", SHT_PROGBITS, 0 }, + { ".hash", SHT_HASH, SHF_ALLOC }, + { ".interp", SHT_PROGBITS, /* SHF_ALLOC */ }, + { ".plt", SHT_PROGBITS, 0 }, + { ".shstrtab",SHT_STRTAB, 0 }, + { ".strtab", SHT_STRTAB, /* SHF_ALLOC */ }, + { ".symtab", SHT_SYMTAB, /* SHF_ALLOC */ }, +#endif + + { NULL, 0, 0 } +}; + +void +obj_elf_section (xxx) + int xxx; +{ + char *string; + int new_sec; + segT sec; + int type, attr; + int i; + flagword flags; + symbolS *secsym; + +#ifdef md_flush_pending_output + md_flush_pending_output (); +#endif + + if (flag_mri) + { + char mri_type; + + previous_section = now_seg; + previous_subsection = now_subseg; + + s_mri_sect (&mri_type); + +#ifdef md_elf_section_change_hook + md_elf_section_change_hook (); +#endif + + return; + } + + /* Get name of section. */ + SKIP_WHITESPACE (); + if (*input_line_pointer == '"') + { + string = demand_copy_C_string (&xxx); + if (string == NULL) + { + ignore_rest_of_line (); + return; + } + } + else + { + char *p = input_line_pointer; + char c; + while (0 == strchr ("\n\t,; ", *p)) + p++; + if (p == input_line_pointer) + { + as_warn (_("Missing section name")); + ignore_rest_of_line (); + return; + } + c = *p; + *p = 0; + string = xmalloc ((unsigned long) (p - input_line_pointer + 1)); + strcpy (string, input_line_pointer); + *p = c; + input_line_pointer = p; + } + + /* Switch to the section, creating it if necessary. */ + previous_section = now_seg; + previous_subsection = now_subseg; + + new_sec = bfd_get_section_by_name (stdoutput, string) == NULL; + sec = subseg_new (string, 0); + + /* If this section already existed, we don't bother to change the + flag values. */ + if (! new_sec) + { + while (! is_end_of_line[(unsigned char) *input_line_pointer]) + ++input_line_pointer; + ++input_line_pointer; + +#ifdef md_elf_section_change_hook + md_elf_section_change_hook (); +#endif + + return; + } + + SKIP_WHITESPACE (); + + type = SHT_NULL; + attr = 0; + + if (*input_line_pointer == ',') + { + /* Skip the comma. */ + ++input_line_pointer; + + SKIP_WHITESPACE (); + + if (*input_line_pointer == '"') + { + /* Pick up a string with a combination of a, w, x. */ + ++input_line_pointer; + while (*input_line_pointer != '"') + { + switch (*input_line_pointer) + { + case 'a': + attr |= SHF_ALLOC; + break; + case 'w': + attr |= SHF_WRITE; + break; + case 'x': + attr |= SHF_EXECINSTR; + break; + default: + { + char *bad_msg = _("Bad .section directive: want a,w,x in string"); +#ifdef md_elf_section_letter + int md_attr = md_elf_section_letter (*input_line_pointer, &bad_msg); + if (md_attr) + attr |= md_attr; + else +#endif + { + as_warn (bad_msg); + ignore_rest_of_line (); + return; + } + } + } + ++input_line_pointer; + } + + /* Skip the closing quote. */ + ++input_line_pointer; + + SKIP_WHITESPACE (); + if (*input_line_pointer == ',') + { + ++input_line_pointer; + SKIP_WHITESPACE (); + if (*input_line_pointer == '@' || *input_line_pointer == '%') + { + ++input_line_pointer; + if (strncmp (input_line_pointer, "progbits", + sizeof "progbits" - 1) == 0) + { + type = SHT_PROGBITS; + input_line_pointer += sizeof "progbits" - 1; + } + else if (strncmp (input_line_pointer, "nobits", + sizeof "nobits" - 1) == 0) + { + type = SHT_NOBITS; + input_line_pointer += sizeof "nobits" - 1; + } + else + { +#ifdef md_elf_section_type + int md_type = md_elf_section_type (&input_line_pointer); + if (md_type) + type = md_type; + else +#endif + { + as_warn (_("Unrecognized section type")); + ignore_rest_of_line (); + } + } + } + } + } + else + { + do + { + SKIP_WHITESPACE (); + if (*input_line_pointer != '#') + { + as_warn (_("Bad .section directive - character following name is not '#'")); + ignore_rest_of_line (); + return; + } + ++input_line_pointer; + if (strncmp (input_line_pointer, "write", + sizeof "write" - 1) == 0) + { + attr |= SHF_WRITE; + input_line_pointer += sizeof "write" - 1; + } + else if (strncmp (input_line_pointer, "alloc", + sizeof "alloc" - 1) == 0) + { + attr |= SHF_ALLOC; + input_line_pointer += sizeof "alloc" - 1; + } + else if (strncmp (input_line_pointer, "execinstr", + sizeof "execinstr" - 1) == 0) + { + attr |= SHF_EXECINSTR; + input_line_pointer += sizeof "execinstr" - 1; + } + else + { +#ifdef md_elf_section_word + int md_attr = md_elf_section_word (&input_line_pointer); + if (md_attr) + attr |= md_attr; + else +#endif + { + as_warn (_("Unrecognized section attribute")); + ignore_rest_of_line (); + return; + } + } + SKIP_WHITESPACE (); + } + while (*input_line_pointer++ == ','); + --input_line_pointer; + } + } + + /* See if this is one of the special sections. */ + for (i = 0; special_sections[i].name != NULL; i++) + { + if (string[1] == special_sections[i].name[1] + && strcmp (string, special_sections[i].name) == 0) + { + if (type == SHT_NULL) + type = special_sections[i].type; + else if (type != special_sections[i].type) + as_warn (_("Setting incorrect section type for %s"), string); + + if ((attr &~ special_sections[i].attributes) != 0) + { + /* As a GNU extension, we permit a .note section to be + allocatable. If the linker sees an allocateable + .note section, it will create a PT_NOTE segment in + the output file. */ + if (strcmp (string, ".note") != 0 + || attr != SHF_ALLOC) + as_warn (_("Setting incorrect section attributes for %s"), + string); + } + attr |= special_sections[i].attributes; + + break; + } + } + + flags = (SEC_RELOC + | ((attr & SHF_WRITE) ? 0 : SEC_READONLY) + | ((attr & SHF_ALLOC) ? SEC_ALLOC : 0) + | (((attr & SHF_ALLOC) && type != SHT_NOBITS) ? SEC_LOAD : 0) + | ((attr & SHF_EXECINSTR) ? SEC_CODE : 0)); + if (special_sections[i].name == NULL) + { + if (type == SHT_PROGBITS) + flags |= SEC_ALLOC | SEC_LOAD; + else if (type == SHT_NOBITS) + { + flags |= SEC_ALLOC; + flags &=~ SEC_LOAD; + } + +#ifdef md_elf_section_flags + flags = md_elf_section_flags (flags, attr, type); +#endif + } + + /* Prevent SEC_HAS_CONTENTS from being inadvertently set. */ + if (type == SHT_NOBITS) + seg_info (sec)->bss = 1; + + bfd_set_section_flags (stdoutput, sec, flags); + + /* Add a symbol for this section to the symbol table. */ + secsym = symbol_find (string); + if (secsym != NULL) + secsym->bsym = sec->symbol; + else + symbol_table_insert (section_symbol (sec)); + +#ifdef md_elf_section_change_hook + md_elf_section_change_hook (); +#endif + + demand_empty_rest_of_line (); +} + +/* Change to the .data section. */ + +static void +obj_elf_data (i) + int i; +{ +#ifdef md_flush_pending_output + md_flush_pending_output (); +#endif + + previous_section = now_seg; + previous_subsection = now_subseg; + s_data (i); + +#ifdef md_elf_section_change_hook + md_elf_section_change_hook (); +#endif +} + +/* Change to the .text section. */ + +static void +obj_elf_text (i) + int i; +{ +#ifdef md_flush_pending_output + md_flush_pending_output (); +#endif + + previous_section = now_seg; + previous_subsection = now_subseg; + s_text (i); + +#ifdef md_elf_section_change_hook + md_elf_section_change_hook (); +#endif +} + +static void +obj_elf_subsection (ignore) + int ignore; +{ + register int temp; + +#ifdef md_flush_pending_output + md_flush_pending_output (); +#endif + + previous_section = now_seg; + previous_subsection = now_subseg; + + temp = get_absolute_expression (); + subseg_set (now_seg, (subsegT) temp); + demand_empty_rest_of_line (); + +#ifdef md_elf_section_change_hook + md_elf_section_change_hook (); +#endif +} + +/* This can be called from the processor backends if they change + sections. */ + +void +obj_elf_section_change_hook () +{ + previous_section = now_seg; + previous_subsection = now_subseg; +} + +void +obj_elf_previous (ignore) + int ignore; +{ + if (previous_section == 0) + { + as_bad (_(".previous without corresponding .section; ignored")); + return; + } + +#ifdef md_flush_pending_output + md_flush_pending_output (); +#endif + + subseg_set (previous_section, previous_subsection); + previous_section = 0; + +#ifdef md_elf_section_change_hook + md_elf_section_change_hook (); +#endif +} + +static void +obj_elf_line (ignore) + int ignore; +{ + /* Assume delimiter is part of expression. BSD4.2 as fails with + delightful bug, so we are not being incompatible here. */ + new_logical_line ((char *) NULL, (int) (get_absolute_expression ())); + demand_empty_rest_of_line (); +} + +/* This handles the .symver pseudo-op, which is used to specify a + symbol version. The syntax is ``.symver NAME,SYMVERNAME''. + SYMVERNAME may contain ELF_VER_CHR ('@') characters. This + pseudo-op causes the assembler to emit a symbol named SYMVERNAME + with the same value as the symbol NAME. */ + +static void +obj_elf_symver (ignore) + int ignore; +{ + char *name; + char c; + symbolS *sym; + + name = input_line_pointer; + c = get_symbol_end (); + + sym = symbol_find_or_make (name); + + *input_line_pointer = c; + + if (sym->sy_obj.versioned_name != NULL) + { + as_bad (_("multiple .symver directives for symbol `%s'"), + S_GET_NAME (sym)); + ignore_rest_of_line (); + return; + } + + SKIP_WHITESPACE (); + if (*input_line_pointer != ',') + { + as_bad (_("expected comma after name in .symver")); + ignore_rest_of_line (); + return; + } + + ++input_line_pointer; + name = input_line_pointer; + while (1) + { + c = get_symbol_end (); + if (c != ELF_VER_CHR) + break; + *input_line_pointer++ = c; + } + + sym->sy_obj.versioned_name = xstrdup (name); + + *input_line_pointer = c; + + if (strchr (sym->sy_obj.versioned_name, ELF_VER_CHR) == NULL) + { + as_bad (_("missing version name in `%s' for symbol `%s'"), + sym->sy_obj.versioned_name, S_GET_NAME (sym)); + ignore_rest_of_line (); + return; + } + + demand_empty_rest_of_line (); +} + +/* This handles the .vtable_inherit pseudo-op, which is used to indicate + to the linker the hierarchy in which a particular table resides. The + syntax is ".vtable_inherit CHILDNAME, PARENTNAME". */ + +static void +obj_elf_vtable_inherit (ignore) +{ + char *cname, *pname; + symbolS *csym, *psym; + char c, bad = 0; + + if (*input_line_pointer == '#') + ++input_line_pointer; + + cname = input_line_pointer; + c = get_symbol_end (); + csym = symbol_find (cname); + + /* GCFIXME: should check that we don't have two .vtable_inherits for + the same child symbol. Also, we can currently only do this if the + child symbol is already exists and is placed in a fragment. */ + + if (csym == NULL || csym->sy_frag == NULL) + { + as_bad ("expected `%s' to have already been set for .vtable_inherit", + cname); + bad = 1; + } + + *input_line_pointer = c; + + SKIP_WHITESPACE (); + if (*input_line_pointer != ',') + { + as_bad ("expected comma after name in .vtable_inherit"); + ignore_rest_of_line (); + return; + } + + ++input_line_pointer; + SKIP_WHITESPACE (); + + if (*input_line_pointer == '#') + ++input_line_pointer; + + if (input_line_pointer[0] == '0' + && (input_line_pointer[1] == '\0' + || isspace(input_line_pointer[1]))) + { + psym = section_symbol (absolute_section); + ++input_line_pointer; + } + else + { + pname = input_line_pointer; + c = get_symbol_end (); + psym = symbol_find_or_make (pname); + *input_line_pointer = c; + } + + demand_empty_rest_of_line (); + + if (bad) + return; + + assert (csym->sy_value.X_op == O_constant); + fix_new (csym->sy_frag, csym->sy_value.X_add_number, 0, psym, 0, 0, + BFD_RELOC_VTABLE_INHERIT); +} + +/* This handles the .vtable_entry pseudo-op, which is used to indicate + to the linker that a vtable slot was used. The syntax is + ".vtable_entry tablename, offset". */ + +static void +obj_elf_vtable_entry (ignore) +{ + char *name; + symbolS *sym; + offsetT offset; + char c; + + if (*input_line_pointer == '#') + ++input_line_pointer; + + name = input_line_pointer; + c = get_symbol_end (); + sym = symbol_find_or_make (name); + *input_line_pointer = c; + + SKIP_WHITESPACE (); + if (*input_line_pointer != ',') + { + as_bad ("expected comma after name in .vtable_entry"); + ignore_rest_of_line (); + return; + } + + ++input_line_pointer; + if (*input_line_pointer == '#') + ++input_line_pointer; + + offset = get_absolute_expression (); + + fix_new (frag_now, frag_now_fix (), 0, sym, offset, 0, + BFD_RELOC_VTABLE_ENTRY); + + demand_empty_rest_of_line (); +} + +void +obj_read_begin_hook () +{ +#ifdef NEED_ECOFF_DEBUG + if (ECOFF_DEBUGGING) + ecoff_read_begin_hook (); +#endif +} + +void +obj_symbol_new_hook (symbolP) + symbolS *symbolP; +{ + symbolP->sy_obj.size = NULL; + symbolP->sy_obj.versioned_name = NULL; + +#ifdef NEED_ECOFF_DEBUG + if (ECOFF_DEBUGGING) + ecoff_symbol_new_hook (symbolP); +#endif +} + +void +obj_elf_version (ignore) + int ignore; +{ + char *name; + unsigned int c; + char ch; + char *p; + asection *seg = now_seg; + subsegT subseg = now_subseg; + Elf_Internal_Note i_note; + Elf_External_Note e_note; + asection *note_secp = (asection *) NULL; + int i, len; + + SKIP_WHITESPACE (); + if (*input_line_pointer == '\"') + { + ++input_line_pointer; /* -> 1st char of string. */ + name = input_line_pointer; + + while (is_a_char (c = next_char_of_string ())) + ; + c = *input_line_pointer; + *input_line_pointer = '\0'; + *(input_line_pointer - 1) = '\0'; + *input_line_pointer = c; + + /* create the .note section */ + + note_secp = subseg_new (".note", 0); + bfd_set_section_flags (stdoutput, + note_secp, + SEC_HAS_CONTENTS | SEC_READONLY); + + /* process the version string */ + + len = strlen (name); + + i_note.namesz = ((len + 1) + 3) & ~3; /* round this to word boundary */ + i_note.descsz = 0; /* no description */ + i_note.type = NT_VERSION; + p = frag_more (sizeof (e_note.namesz)); + md_number_to_chars (p, (valueT) i_note.namesz, 4); + p = frag_more (sizeof (e_note.descsz)); + md_number_to_chars (p, (valueT) i_note.descsz, 4); + p = frag_more (sizeof (e_note.type)); + md_number_to_chars (p, (valueT) i_note.type, 4); + + for (i = 0; i < len; i++) + { + ch = *(name + i); + { + FRAG_APPEND_1_CHAR (ch); + } + } + frag_align (2, 0, 0); + + subseg_set (seg, subseg); + } + else + { + as_bad (_("Expected quoted string")); + } + demand_empty_rest_of_line (); +} + +static void +obj_elf_size (ignore) + int ignore; +{ + char *name = input_line_pointer; + char c = get_symbol_end (); + char *p; + expressionS exp; + symbolS *sym; + + p = input_line_pointer; + *p = c; + SKIP_WHITESPACE (); + if (*input_line_pointer != ',') + { + *p = 0; + as_bad (_("expected comma after name `%s' in .size directive"), name); + *p = c; + ignore_rest_of_line (); + return; + } + input_line_pointer++; + expression (&exp); + if (exp.X_op == O_absent) + { + as_bad (_("missing expression in .size directive")); + exp.X_op = O_constant; + exp.X_add_number = 0; + } + *p = 0; + sym = symbol_find_or_make (name); + *p = c; + if (exp.X_op == O_constant) + S_SET_SIZE (sym, exp.X_add_number); + else + { + sym->sy_obj.size = (expressionS *) xmalloc (sizeof (expressionS)); + *sym->sy_obj.size = exp; + } + demand_empty_rest_of_line (); +} + +/* Handle the ELF .type pseudo-op. This sets the type of a symbol. + There are four syntaxes: + + The first (used on Solaris) is + .type SYM,#function + The second (used on UnixWare) is + .type SYM,@function + The third (reportedly to be used on Irix 6.0) is + .type SYM STT_FUNC + The fourth (used on NetBSD/Arm and Linux/ARM) is + .type SYM,%function + */ + +static void +obj_elf_type (ignore) + int ignore; +{ + char *name; + char c; + int type; + const char *typename; + symbolS *sym; + + name = input_line_pointer; + c = get_symbol_end (); + sym = symbol_find_or_make (name); + *input_line_pointer = c; + + SKIP_WHITESPACE (); + if (*input_line_pointer == ',') + ++input_line_pointer; + + SKIP_WHITESPACE (); + if ( *input_line_pointer == '#' + || *input_line_pointer == '@' + || *input_line_pointer == '%') + ++input_line_pointer; + + typename = input_line_pointer; + c = get_symbol_end (); + + type = 0; + if (strcmp (typename, "function") == 0 + || strcmp (typename, "STT_FUNC") == 0) + type = BSF_FUNCTION; + else if (strcmp (typename, "object") == 0 + || strcmp (typename, "STT_OBJECT") == 0) + type = BSF_OBJECT; + else + as_bad (_("ignoring unrecognized symbol type \"%s\""), typename); + + *input_line_pointer = c; + + sym->bsym->flags |= type; + + demand_empty_rest_of_line (); +} + +static void +obj_elf_ident (ignore) + int ignore; +{ + static segT comment_section; + segT old_section = now_seg; + int old_subsection = now_subseg; + + if (!comment_section) + { + char *p; + comment_section = subseg_new (".comment", 0); + bfd_set_section_flags (stdoutput, comment_section, + SEC_READONLY | SEC_HAS_CONTENTS); + p = frag_more (1); + *p = 0; + } + else + subseg_set (comment_section, 0); + stringer (1); + subseg_set (old_section, old_subsection); +} + +#ifdef INIT_STAB_SECTION + +/* The first entry in a .stabs section is special. */ + +void +obj_elf_init_stab_section (seg) + segT seg; +{ + char *file; + char *p; + char *stabstr_name; + unsigned int stroff; + + /* Force the section to align to a longword boundary. Without this, + UnixWare ar crashes. */ + bfd_set_section_alignment (stdoutput, seg, 2); + + /* Make space for this first symbol. */ + p = frag_more (12); + /* Zero it out. */ + memset (p, 0, 12); + as_where (&file, (unsigned int *) NULL); + stabstr_name = (char *) alloca (strlen (segment_name (seg)) + 4); + strcpy (stabstr_name, segment_name (seg)); + strcat (stabstr_name, "str"); + stroff = get_stab_string_offset (file, stabstr_name); + know (stroff == 1); + md_number_to_chars (p, stroff, 4); + seg_info (seg)->stabu.p = p; +} + +#endif + +/* Fill in the counts in the first entry in a .stabs section. */ + +static void +adjust_stab_sections (abfd, sec, xxx) + bfd *abfd; + asection *sec; + PTR xxx; +{ + char *name; + asection *strsec; + char *p; + int strsz, nsyms; + + if (strncmp (".stab", sec->name, 5)) + return; + if (!strcmp ("str", sec->name + strlen (sec->name) - 3)) + return; + + name = (char *) alloca (strlen (sec->name) + 4); + strcpy (name, sec->name); + strcat (name, "str"); + strsec = bfd_get_section_by_name (abfd, name); + if (strsec) + strsz = bfd_section_size (abfd, strsec); + else + strsz = 0; + nsyms = bfd_section_size (abfd, sec) / 12 - 1; + + p = seg_info (sec)->stabu.p; + assert (p != 0); + + bfd_h_put_16 (abfd, (bfd_vma) nsyms, (bfd_byte *) p + 6); + bfd_h_put_32 (abfd, (bfd_vma) strsz, (bfd_byte *) p + 8); +} + +#ifdef NEED_ECOFF_DEBUG + +/* This function is called by the ECOFF code. It is supposed to + record the external symbol information so that the backend can + write it out correctly. The ELF backend doesn't actually handle + this at the moment, so we do it ourselves. We save the information + in the symbol. */ + +void +elf_ecoff_set_ext (sym, ext) + symbolS *sym; + struct ecoff_extr *ext; +{ + sym->bsym->udata.p = (PTR) ext; +} + +/* This function is called by bfd_ecoff_debug_externals. It is + supposed to *EXT to the external symbol information, and return + whether the symbol should be used at all. */ + +static boolean +elf_get_extr (sym, ext) + asymbol *sym; + EXTR *ext; +{ + if (sym->udata.p == NULL) + return false; + *ext = *(EXTR *) sym->udata.p; + return true; +} + +/* This function is called by bfd_ecoff_debug_externals. It has + nothing to do for ELF. */ + +/*ARGSUSED*/ +static void +elf_set_index (sym, indx) + asymbol *sym; + bfd_size_type indx; +{ +} + +#endif /* NEED_ECOFF_DEBUG */ + +void +elf_frob_symbol (symp, puntp) + symbolS *symp; + int *puntp; +{ +#ifdef NEED_ECOFF_DEBUG + if (ECOFF_DEBUGGING) + ecoff_frob_symbol (symp); +#endif + + if (symp->sy_obj.size != NULL) + { + switch (symp->sy_obj.size->X_op) + { + case O_subtract: + S_SET_SIZE (symp, + (S_GET_VALUE (symp->sy_obj.size->X_add_symbol) + + symp->sy_obj.size->X_add_number + - S_GET_VALUE (symp->sy_obj.size->X_op_symbol))); + break; + case O_constant: + S_SET_SIZE (symp, + (S_GET_VALUE (symp->sy_obj.size->X_add_symbol) + + symp->sy_obj.size->X_add_number)); + break; + default: + as_bad (_(".size expression too complicated to fix up")); + break; + } + free (symp->sy_obj.size); + symp->sy_obj.size = NULL; + } + + if (symp->sy_obj.versioned_name != NULL) + { + /* This symbol was given a new name with the .symver directive. + + If this is an external reference, just rename the symbol to + include the version string. This will make the relocs be + against the correct versioned symbol. + + If this is a definition, add an alias. FIXME: Using an alias + will permit the debugging information to refer to the right + symbol. However, it's not clear whether it is the best + approach. */ + + if (! S_IS_DEFINED (symp)) + { + char *p; + + /* Verify that the name isn't using the @@ syntax--this is + reserved for definitions of the default version to link + against. */ + p = strchr (symp->sy_obj.versioned_name, ELF_VER_CHR); + know (p != NULL); + if (p[1] == ELF_VER_CHR) + { + as_bad (_("invalid attempt to declare external version name as default in symbol `%s'"), + symp->sy_obj.versioned_name); + *puntp = true; + } + S_SET_NAME (symp, symp->sy_obj.versioned_name); + } + else + { + symbolS *symp2; + + /* FIXME: Creating a new symbol here is risky. We're in the + final loop over the symbol table. We can get away with + it only because the symbol goes to the end of the list, + where the loop will still see it. It would probably be + better to do this in obj_frob_file_before_adjust. */ + + symp2 = symbol_find_or_make (symp->sy_obj.versioned_name); + + /* Now we act as though we saw symp2 = sym. */ + + S_SET_SEGMENT (symp2, S_GET_SEGMENT (symp)); + + /* Subtracting out the frag address here is a hack because + we are in the middle of the final loop. */ + S_SET_VALUE (symp2, S_GET_VALUE (symp) - symp->sy_frag->fr_address); + + symp2->sy_frag = symp->sy_frag; + + /* This will copy over the size information. */ + copy_symbol_attributes (symp2, symp); + + if (S_IS_WEAK (symp)) + S_SET_WEAK (symp2); + + if (S_IS_EXTERNAL (symp)) + S_SET_EXTERNAL (symp2); + } + } + + /* Double check weak symbols. */ + if (symp->bsym->flags & BSF_WEAK) + { + if (S_IS_COMMON (symp)) + as_bad (_("Symbol `%s' can not be both weak and common"), + S_GET_NAME (symp)); + } + +#ifdef TC_MIPS + /* The Irix 5 and 6 assemblers set the type of any common symbol and + any undefined non-function symbol to STT_OBJECT. We try to be + compatible, since newer Irix 5 and 6 linkers care. However, we + only set undefined symbols to be STT_OBJECT if we are on Irix, + because that is the only time gcc will generate the necessary + .global directives to mark functions. */ + + if (S_IS_COMMON (symp)) + symp->bsym->flags |= BSF_OBJECT; + + if (strstr (TARGET_OS, "irix") != NULL + && (! S_IS_DEFINED (symp) && ((symp->bsym->flags & BSF_FUNCTION) == 0))) + symp->bsym->flags |= BSF_OBJECT; +#endif + +#ifdef TC_PPC + /* Frob the PowerPC, so that the symbol always has object type + if it is not some other type. VxWorks needs this. */ + if ((symp->bsym->flags & (BSF_FUNCTION | BSF_FILE | BSF_SECTION_SYM)) == 0 + && S_IS_DEFINED (symp)) + symp->bsym->flags |= BSF_OBJECT; +#endif +} + +void +elf_frob_file () +{ + bfd_map_over_sections (stdoutput, adjust_stab_sections, (PTR) 0); + +#ifdef elf_tc_final_processing + elf_tc_final_processing (); +#endif +} + +/* It is required that we let write_relocs have the opportunity to + optimize away fixups before output has begun, since it is possible + to eliminate all fixups for a section and thus we never should + have generated the relocation section. */ + +void +elf_frob_file_after_relocs () +{ +#ifdef NEED_ECOFF_DEBUG + if (ECOFF_DEBUGGING) + /* Generate the ECOFF debugging information. */ + { + const struct ecoff_debug_swap *debug_swap; + struct ecoff_debug_info debug; + char *buf; + asection *sec; + + debug_swap + = get_elf_backend_data (stdoutput)->elf_backend_ecoff_debug_swap; + know (debug_swap != (const struct ecoff_debug_swap *) NULL); + ecoff_build_debug (&debug.symbolic_header, &buf, debug_swap); + + /* Set up the pointers in debug. */ +#define SET(ptr, offset, type) \ + debug.ptr = (type) (buf + debug.symbolic_header.offset) + + SET (line, cbLineOffset, unsigned char *); + SET (external_dnr, cbDnOffset, PTR); + SET (external_pdr, cbPdOffset, PTR); + SET (external_sym, cbSymOffset, PTR); + SET (external_opt, cbOptOffset, PTR); + SET (external_aux, cbAuxOffset, union aux_ext *); + SET (ss, cbSsOffset, char *); + SET (external_fdr, cbFdOffset, PTR); + SET (external_rfd, cbRfdOffset, PTR); + /* ssext and external_ext are set up just below. */ + +#undef SET + + /* Set up the external symbols. */ + debug.ssext = debug.ssext_end = NULL; + debug.external_ext = debug.external_ext_end = NULL; + if (! bfd_ecoff_debug_externals (stdoutput, &debug, debug_swap, true, + elf_get_extr, elf_set_index)) + as_fatal (_("Failed to set up debugging information: %s"), + bfd_errmsg (bfd_get_error ())); + + sec = bfd_get_section_by_name (stdoutput, ".mdebug"); + assert (sec != NULL); + + know (stdoutput->output_has_begun == false); + + /* We set the size of the section, call bfd_set_section_contents + to force the ELF backend to allocate a file position, and then + write out the data. FIXME: Is this really the best way to do + this? */ + sec->_raw_size = bfd_ecoff_debug_size (stdoutput, &debug, debug_swap); + + if (! bfd_set_section_contents (stdoutput, sec, (PTR) NULL, + (file_ptr) 0, (bfd_size_type) 0)) + as_fatal (_("Can't start writing .mdebug section: %s"), + bfd_errmsg (bfd_get_error ())); + + know (stdoutput->output_has_begun == true); + know (sec->filepos != 0); + + if (! bfd_ecoff_write_debug (stdoutput, &debug, debug_swap, + sec->filepos)) + as_fatal (_("Could not write .mdebug section: %s"), + bfd_errmsg (bfd_get_error ())); + } +#endif /* NEED_ECOFF_DEBUG */ +} + +#ifdef SCO_ELF + +/* Heavily plagarized from obj_elf_version. The idea is to emit the + SCO specific identifier in the .notes section to satisfy the SCO + linker. + + This looks more complicated than it really is. As opposed to the + "obvious" solution, this should handle the cross dev cases + correctly. (i.e, hosting on a 64 bit big endian processor, but + generating SCO Elf code) Efficiency isn't a concern, as there + should be exactly one of these sections per object module. + + SCO OpenServer 5 identifies it's ELF modules with a standard ELF + .note section. + + int_32 namesz = 4 ; Name size + int_32 descsz = 12 ; Descriptive information + int_32 type = 1 ; + char name[4] = "SCO" ; Originator name ALWAYS SCO + NULL + int_32 version = (major ver # << 16) | version of tools ; + int_32 source = (tool_id << 16 ) | 1 ; + int_32 info = 0 ; These are set by the SCO tools, but we + don't know enough about the source + environment to set them. SCO ld currently + ignores them, and recommends we set them + to zero. */ + +#define SCO_MAJOR_VERSION 0x1 +#define SCO_MINOR_VERSION 0x1 + +void +sco_id () +{ + + char *name; + unsigned int c; + char ch; + char *p; + asection *seg = now_seg; + subsegT subseg = now_subseg; + Elf_Internal_Note i_note; + Elf_External_Note e_note; + asection *note_secp = (asection *) NULL; + int i, len; + + /* create the .note section */ + + note_secp = subseg_new (".note", 0); + bfd_set_section_flags (stdoutput, + note_secp, + SEC_HAS_CONTENTS | SEC_READONLY); + + /* process the version string */ + + i_note.namesz = 4; + i_note.descsz = 12; /* 12 descriptive bytes */ + i_note.type = NT_VERSION; /* Contains a version string */ + + p = frag_more (sizeof (i_note.namesz)); + md_number_to_chars (p, (valueT) i_note.namesz, 4); + + p = frag_more (sizeof (i_note.descsz)); + md_number_to_chars (p, (valueT) i_note.descsz, 4); + + p = frag_more (sizeof (i_note.type)); + md_number_to_chars (p, (valueT) i_note.type, 4); + + p = frag_more (4); + strcpy (p, "SCO"); + + /* Note: this is the version number of the ELF we're representing */ + p = frag_more (4); + md_number_to_chars (p, (SCO_MAJOR_VERSION << 16) | (SCO_MINOR_VERSION), 4); + + /* Here, we pick a magic number for ourselves (yes, I "registered" + it with SCO. The bottom bit shows that we are compat with the + SCO ABI. */ + p = frag_more (4); + md_number_to_chars (p, 0x4c520000 | 0x0001, 4); + + /* If we knew (or cared) what the source language options were, we'd + fill them in here. SCO has given us permission to ignore these + and just set them to zero. */ + p = frag_more (4); + md_number_to_chars (p, 0x0000, 4); + + frag_align (2, 0, 0); + + /* We probably can't restore the current segment, for there likely + isn't one yet... */ + if (seg && subseg) + subseg_set (seg, subseg); + +} + +#endif /* SCO_ELF */ + +const struct format_ops elf_format_ops = +{ + bfd_target_elf_flavour, + 0, + 1, + elf_frob_symbol, + elf_frob_file, + elf_frob_file_after_relocs, + elf_s_get_size, elf_s_set_size, + elf_s_get_align, elf_s_set_align, + elf_copy_symbol_attributes, +#ifdef NEED_ECOFF_DEBUG + ecoff_generate_asm_lineno, + ecoff_stab, +#else + 0, + 0, /* process_stab */ +#endif + elf_sec_sym_ok_for_reloc, + elf_pop_insert, +#ifdef NEED_ECOFF_DEBUG + elf_ecoff_set_ext, +#else + 0, +#endif + obj_read_begin_hook, + obj_symbol_new_hook, +}; |