X-Git-Url: http://drtracing.org/?a=blobdiff_plain;f=gas%2Fconfig%2Ftc-i386.c;h=5cf87bd0e2c4a85652cdf104d8b71e6bdf5f78ee;hb=bc805888417f16602aeb4bfd3f6e33d32ffcae24;hp=b06298ee8a2409ca0d12f9e3bf0a96318a72c97f;hpb=e1b283bbb7a3ff7364c7100bf7ac975155d16121;p=deliverable%2Fbinutils-gdb.git diff --git a/gas/config/tc-i386.c b/gas/config/tc-i386.c index b06298ee8a..5cf87bd0e2 100644 --- a/gas/config/tc-i386.c +++ b/gas/config/tc-i386.c @@ -1,6 +1,6 @@ /* i386.c -- Assemble code for the Intel 80386 - Copyright (C) 1989, 91, 92, 93, 94, 95, 96, 97, 98, 1999 - Free Software Foundation. + Copyright (C) 1989, 91, 92, 93, 94, 95, 96, 97, 98, 99, 2000 + Free Software Foundation, Inc. This file is part of GAS, the GNU Assembler. @@ -19,27 +19,26 @@ Software Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */ -/* - Intel 80386 machine specific gas. - Written by Eliot Dresselhaus (eliot@mgm.mit.edu). - Bugs & suggestions are completely welcome. This is free software. - Please help us make it better. - */ +/* Intel 80386 machine specific gas. + Written by Eliot Dresselhaus (eliot@mgm.mit.edu). + Bugs & suggestions are completely welcome. This is free software. + Please help us make it better. */ #include #include "as.h" #include "subsegs.h" +#include "dwarf2dbg.h" #include "opcode/i386.h" -#ifndef TC_RELOC -#define TC_RELOC(X,Y) (Y) -#endif - #ifndef REGISTER_WARNINGS #define REGISTER_WARNINGS 1 #endif +#ifndef INFER_ADDR_PREFIX +#define INFER_ADDR_PREFIX 1 +#endif + #ifndef SCALE1_WHEN_NO_INDEX /* Specifying a scale factor besides 1 when there is no index is futile. eg. `mov (%ebx,2),%al' does exactly the same as @@ -52,14 +51,17 @@ #define false 0 static unsigned int mode_from_disp_size PARAMS ((unsigned int)); -static int fits_in_signed_byte PARAMS ((long)); -static int fits_in_unsigned_byte PARAMS ((long)); -static int fits_in_unsigned_word PARAMS ((long)); -static int fits_in_signed_word PARAMS ((long)); -static int smallest_imm_type PARAMS ((long)); +static int fits_in_signed_byte PARAMS ((offsetT)); +static int fits_in_unsigned_byte PARAMS ((offsetT)); +static int fits_in_unsigned_word PARAMS ((offsetT)); +static int fits_in_signed_word PARAMS ((offsetT)); +static int smallest_imm_type PARAMS ((offsetT)); +static offsetT offset_in_range PARAMS ((offsetT, int)); static int add_prefix PARAMS ((unsigned int)); static void set_16bit_code_flag PARAMS ((int)); +static void set_16bit_gcc_code_flag PARAMS ((int)); static void set_intel_syntax PARAMS ((int)); +static void set_cpu_arch PARAMS ((int)); #ifdef BFD_ASSEMBLER static bfd_reloc_code_real_type reloc @@ -67,34 +69,39 @@ static bfd_reloc_code_real_type reloc #endif /* 'md_assemble ()' gathers together information and puts it into a - i386_insn. */ + i386_insn. */ + +union i386_op + { + expressionS *disps; + expressionS *imms; + const reg_entry *regs; + }; struct _i386_insn { - /* TM holds the template for the insn were currently assembling. */ + /* TM holds the template for the insn were currently assembling. */ template tm; /* SUFFIX holds the instruction mnemonic suffix if given. (e.g. 'l' for 'movl') */ char suffix; - /* Operands are coded with OPERANDS, TYPES, DISPS, IMMS, and REGS. */ - - /* OPERANDS gives the number of given operands. */ + /* OPERANDS gives the number of given operands. */ unsigned int operands; /* REG_OPERANDS, DISP_OPERANDS, MEM_OPERANDS, IMM_OPERANDS give the number of given register, displacement, memory operands and immediate - operands. */ + operands. */ unsigned int reg_operands, disp_operands, mem_operands, imm_operands; /* TYPES [i] is the type (see above #defines) which tells us how to - search through DISPS [i] & IMMS [i] & REGS [i] for the required - operand. */ + use OP[i] for the corresponding operand. */ unsigned int types[MAX_OPERANDS]; - /* Displacements (if given) for each operand. */ - expressionS *disps[MAX_OPERANDS]; + /* Displacement expression, immediate expression, or register for each + operand. */ + union i386_op op[MAX_OPERANDS]; /* Relocation type for operand */ #ifdef BFD_ASSEMBLER @@ -103,12 +110,6 @@ struct _i386_insn int disp_reloc[MAX_OPERANDS]; #endif - /* Immediate operands (if given) for each operand. */ - expressionS *imms[MAX_OPERANDS]; - - /* Register operands (if given) for each operand. */ - const reg_entry *regs[MAX_OPERANDS]; - /* BASE_REG, INDEX_REG, and LOG2_SCALE_FACTOR are used to encode the base index byte below. */ const reg_entry *base_reg; @@ -116,8 +117,8 @@ struct _i386_insn unsigned int log2_scale_factor; /* SEG gives the seg_entries of this insn. They are zero unless - explicit segment overrides are given. */ - const seg_entry *seg[2]; /* segments for memory operands (if given) */ + explicit segment overrides are given. */ + const seg_entry *seg[2]; /* PREFIX holds all the given prefix opcodes (usually null). PREFIXES is the number of prefix opcodes. */ @@ -142,8 +143,8 @@ const char extra_symbol_chars[] = "*%-("; #endif /* This array holds the chars that always start a comment. If the - pre-processor is disabled, these aren't very useful */ -#if defined (TE_I386AIX) || ((defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF)) && ! defined (TE_LINUX)) + pre-processor is disabled, these aren't very useful. */ +#if defined (TE_I386AIX) || ((defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF)) && ! defined (TE_LINUX) && !defined(TE_FreeBSD)) /* Putting '/' here makes it impossible to use the divide operator. However, we need it for compatibility with SVR4 systems. */ const char comment_chars[] = "#/"; @@ -155,36 +156,37 @@ const char comment_chars[] = "#"; /* This array holds the chars that only start a comment at the beginning of a line. If the line seems to have the form '# 123 filename' - .line and .file directives will appear in the pre-processed output */ -/* Note that input_file.c hand checks for '#' at the beginning of the + .line and .file directives will appear in the pre-processed output. + Note that input_file.c hand checks for '#' at the beginning of the first line of the input file. This is because the compiler outputs - #NO_APP at the beginning of its output. */ -/* Also note that comments started like this one will always work if + #NO_APP at the beginning of its output. + Also note that comments started like this one will always work if '/' isn't otherwise defined. */ -#if defined (TE_I386AIX) || ((defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF)) && ! defined (TE_LINUX)) +#if defined (TE_I386AIX) || ((defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF)) && ! defined (TE_LINUX) && !defined(TE_FreeBSD)) const char line_comment_chars[] = ""; #else const char line_comment_chars[] = "/"; #endif -const char line_separator_chars[] = ""; +const char line_separator_chars[] = ";"; -/* Chars that can be used to separate mant from exp in floating point nums */ +/* Chars that can be used to separate mant from exp in floating point + nums. */ const char EXP_CHARS[] = "eE"; -/* Chars that mean this number is a floating point constant */ -/* As in 0f12.456 */ -/* or 0d1.2345e12 */ +/* Chars that mean this number is a floating point constant + As in 0f12.456 + or 0d1.2345e12. */ const char FLT_CHARS[] = "fFdDxX"; -/* tables for lexical analysis */ +/* Tables for lexical analysis. */ static char mnemonic_chars[256]; static char register_chars[256]; static char operand_chars[256]; static char identifier_chars[256]; static char digit_chars[256]; -/* lexical macros */ +/* Lexical macros. */ #define is_mnemonic_char(x) (mnemonic_chars[(unsigned char) x]) #define is_operand_char(x) (operand_chars[(unsigned char) x]) #define is_register_char(x) (register_chars[(unsigned char) x]) @@ -192,49 +194,67 @@ static char digit_chars[256]; #define is_identifier_char(x) (identifier_chars[(unsigned char) x]) #define is_digit_char(x) (digit_chars[(unsigned char) x]) -/* put here all non-digit non-letter charcters that may occur in an operand */ +/* All non-digit non-letter charcters that may occur in an operand. */ static char operand_special_chars[] = "%$-+(,)*._~/<>|&^!:[@]"; /* md_assemble() always leaves the strings it's passed unaltered. To effect this we maintain a stack of saved characters that we've smashed with '\0's (indicating end of strings for various sub-fields of the - assembler instruction). */ + assembler instruction). */ static char save_stack[32]; -static char *save_stack_p; /* stack pointer */ +static char *save_stack_p; #define END_STRING_AND_SAVE(s) \ do { *save_stack_p++ = *(s); *(s) = '\0'; } while (0) #define RESTORE_END_STRING(s) \ do { *(s) = *--save_stack_p; } while (0) -/* The instruction we're assembling. */ +/* The instruction we're assembling. */ static i386_insn i; /* Possible templates for current insn. */ static const templates *current_templates; -/* Per instruction expressionS buffers: 2 displacements & 2 immediate max. */ +/* Per instruction expressionS buffers: 2 displacements & 2 immediate max. */ static expressionS disp_expressions[2], im_expressions[2]; -static int this_operand; /* current operand we are working on */ +/* Current operand we are working on. */ +static int this_operand; -static int flag_do_long_jump; /* FIXME what does this do? */ +/* 1 if we're writing 16-bit code, + 0 if 32-bit. */ +static int flag_16bit_code; -static int flag_16bit_code; /* 1 if we're writing 16-bit code, 0 if 32-bit */ +/* 1 for intel syntax, + 0 if att syntax. */ +static int intel_syntax = 0; -static int intel_syntax = 0; /* 1 for intel syntax, 0 if att syntax */ +/* 1 if register prefix % not required. */ +static int allow_naked_reg = 0; -static int allow_naked_reg = 0; /* 1 if register prefix % not required */ +/* Used in 16 bit gcc mode to add an l suffix to call, ret, enter, + leave, push, and pop instructions so that gcc has the same stack + frame as in 32 bit mode. */ +static char stackop_size = '\0'; + +/* Non-zero to quieten some warnings. */ +static int quiet_warnings = 0; + +/* CPU name. */ +static const char *cpu_arch_name = NULL; + +/* CPU feature flags. */ +static unsigned int cpu_arch_flags = 0; /* Interface to relax_segment. There are 2 relax states for 386 jump insns: one for conditional & - one for unconditional jumps. This is because the these two types - of jumps add different sizes to frags when we're figuring out what + one for unconditional jumps. This is because these two types of + jumps add different sizes to frags when we're figuring out what sort of jump to choose to reach a given label. */ -/* types */ -#define COND_JUMP 1 /* conditional jump */ -#define UNCOND_JUMP 2 /* unconditional jump */ -/* sizes */ +/* Types. */ +#define COND_JUMP 1 +#define UNCOND_JUMP 2 +/* Sizes. */ #define CODE16 1 #define SMALL 0 #define SMALL16 (SMALL|CODE16) @@ -250,7 +270,7 @@ static int allow_naked_reg = 0; /* 1 if register prefix % not required */ #endif #define ENCODE_RELAX_STATE(type,size) \ - ((relax_substateT)((type<<2) | (size))) + ((relax_substateT) ((type<<2) | (size))) #define SIZE_FROM_RELAX_STATE(s) \ ( (((s) & 0x3) == BIG ? 4 : (((s) & 0x3) == BIG16 ? 2 : 1)) ) @@ -264,12 +284,11 @@ static int allow_naked_reg = 0; /* 1 if register prefix % not required */ const relax_typeS md_relax_table[] = { -/* The fields are: - 1) most positive reach of this state, - 2) most negative reach of this state, - 3) how many bytes this mode will add to the size of the current frag - 4) which index into the table to try if we can't fit into this one. - */ + /* The fields are: + 1) most positive reach of this state, + 2) most negative reach of this state, + 3) how many bytes this mode will add to the size of the current frag + 4) which index into the table to try if we can't fit into this one. */ {1, 1, 0, 0}, {1, 1, 0, 0}, {1, 1, 0, 0}, @@ -295,15 +314,29 @@ const relax_typeS md_relax_table[] = }; +static const arch_entry cpu_arch[] = { + {"i8086", Cpu086 }, + {"i186", Cpu086|Cpu186 }, + {"i286", Cpu086|Cpu186|Cpu286 }, + {"i386", Cpu086|Cpu186|Cpu286|Cpu386 }, + {"i486", Cpu086|Cpu186|Cpu286|Cpu386|Cpu486 }, + {"i586", Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|CpuMMX }, + {"i686", Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuMMX|CpuSSE }, + {"pentium", Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|CpuMMX }, + {"pentiumpro",Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuMMX|CpuSSE }, + {"k6", Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|CpuMMX|Cpu3dnow }, + {"athlon", Cpu086|Cpu186|Cpu286|Cpu386|Cpu486|Cpu586|Cpu686|CpuMMX|Cpu3dnow }, + {NULL, 0 } +}; void i386_align_code (fragP, count) fragS *fragP; int count; { - /* Various efficient no-op patterns for aligning code labels. */ - /* Note: Don't try to assemble the instructions in the comments. */ - /* 0L and 0w are not legal */ + /* Various efficient no-op patterns for aligning code labels. + Note: Don't try to assemble the instructions in the comments. + 0L and 0w are not legal. */ static const char f32_1[] = {0x90}; /* nop */ static const char f32_2[] = @@ -343,6 +376,8 @@ i386_align_code (fragP, count) static const char f32_15[] = {0xeb,0x0d,0x90,0x90,0x90,0x90,0x90, /* jmp .+15; lotsa nops */ 0x90,0x90,0x90,0x90,0x90,0x90,0x90,0x90}; + static const char f16_3[] = + {0x8d,0x74,0x00}; /* lea 0(%esi),%esi */ static const char f16_4[] = {0x8d,0xb4,0x00,0x00}; /* lea 0w(%si),%si */ static const char f16_5[] = @@ -362,7 +397,7 @@ i386_align_code (fragP, count) f32_9, f32_10, f32_11, f32_12, f32_13, f32_14, f32_15 }; static const char *const f16_patt[] = { - f32_1, f32_2, f32_3, f16_4, f16_5, f16_6, f16_7, f16_8, + f32_1, f32_2, f16_3, f16_4, f16_5, f16_6, f16_7, f16_8, f32_15, f32_15, f32_15, f32_15, f32_15, f32_15, f32_15 }; @@ -370,14 +405,15 @@ i386_align_code (fragP, count) { if (flag_16bit_code) { - memcpy(fragP->fr_literal + fragP->fr_fix, - f16_patt[count - 1], count); - if (count > 8) /* adjust jump offset */ + memcpy (fragP->fr_literal + fragP->fr_fix, + f16_patt[count - 1], count); + if (count > 8) + /* Adjust jump offset. */ fragP->fr_literal[fragP->fr_fix + 1] = count - 2; } else - memcpy(fragP->fr_literal + fragP->fr_fix, - f32_patt[count - 1], count); + memcpy (fragP->fr_literal + fragP->fr_fix, + f32_patt[count - 1], count); fragP->fr_var = count; } } @@ -392,57 +428,58 @@ static const reg_entry *parse_register PARAMS ((char *reg_string, static void s_bss PARAMS ((int)); #endif -symbolS *GOT_symbol; /* Pre-defined "_GLOBAL_OFFSET_TABLE_" */ +symbolS *GOT_symbol; /* Pre-defined "_GLOBAL_OFFSET_TABLE_". */ static INLINE unsigned int mode_from_disp_size (t) unsigned int t; { - return (t & Disp8) ? 1 : (t & (Disp16|Disp32)) ? 2 : 0; + return (t & Disp8) ? 1 : (t & (Disp16 | Disp32)) ? 2 : 0; } static INLINE int fits_in_signed_byte (num) - long num; + offsetT num; { return (num >= -128) && (num <= 127); -} /* fits_in_signed_byte() */ +} static INLINE int fits_in_unsigned_byte (num) - long num; + offsetT num; { return (num & 0xff) == num; -} /* fits_in_unsigned_byte() */ +} static INLINE int fits_in_unsigned_word (num) - long num; + offsetT num; { return (num & 0xffff) == num; -} /* fits_in_unsigned_word() */ +} static INLINE int fits_in_signed_word (num) - long num; + offsetT num; { return (-32768 <= num) && (num <= 32767); -} /* fits_in_signed_word() */ +} static int smallest_imm_type (num) - long num; + offsetT num; { -#if 0 - /* This code is disabled because all the Imm1 forms in the opcode table - are slower on the i486, and they're the versions with the implicitly - specified single-position displacement, which has another syntax if - you really want to use that form. If you really prefer to have the - one-byte-shorter Imm1 form despite these problems, re-enable this - code. */ - if (num == 1) - return Imm1 | Imm8 | Imm8S | Imm16 | Imm32; -#endif + if (cpu_arch_flags != 0 + && cpu_arch_flags != (Cpu086 | Cpu186 | Cpu286 | Cpu386 | Cpu486)) + { + /* This code is disabled on the 486 because all the Imm1 forms + in the opcode table are slower on the i486. They're the + versions with the implicitly specified single-position + displacement, which has another syntax if you really want to + use that form. */ + if (num == 1) + return Imm1 | Imm8 | Imm8S | Imm16 | Imm32; + } return (fits_in_signed_byte (num) ? (Imm8S | Imm8 | Imm16 | Imm32) : fits_in_unsigned_byte (num) @@ -450,7 +487,37 @@ smallest_imm_type (num) : (fits_in_signed_word (num) || fits_in_unsigned_word (num)) ? (Imm16 | Imm32) : (Imm32)); -} /* smallest_imm_type() */ +} + +static offsetT +offset_in_range (val, size) + offsetT val; + int size; +{ + addressT mask; + + switch (size) + { + case 1: mask = ((addressT) 1 << 8) - 1; break; + case 2: mask = ((addressT) 1 << 16) - 1; break; + case 4: mask = ((addressT) 2 << 31) - 1; break; + default: abort (); + } + + /* If BFD64, sign extend val. */ + if ((val & ~(((addressT) 2 << 31) - 1)) == 0) + val = (val ^ ((addressT) 1 << 31)) - ((addressT) 1 << 31); + + if ((val & ~mask) != 0 && (val & ~mask) != ~mask) + { + char buf1[40], buf2[40]; + + sprint_value (buf1, val); + sprint_value (buf2, val & mask); + as_warn (_("%s shortened to %s"), buf1, buf2); + } + return val & mask; +} /* Returns 0 if attempting to add a prefix where one from the same class already exists, 1 if non rep/repne added, 2 if rep/repne @@ -510,14 +577,23 @@ add_prefix (prefix) static void set_16bit_code_flag (new_16bit_code_flag) - int new_16bit_code_flag; + int new_16bit_code_flag; +{ + flag_16bit_code = new_16bit_code_flag; + stackop_size = '\0'; +} + +static void +set_16bit_gcc_code_flag (new_16bit_code_flag) + int new_16bit_code_flag; { flag_16bit_code = new_16bit_code_flag; + stackop_size = new_16bit_code_flag ? 'l' : '\0'; } static void set_intel_syntax (syntax_flag) - int syntax_flag; + int syntax_flag; { /* Find out if register prefixing is specified. */ int ask_naked_reg = 0; @@ -528,40 +604,74 @@ set_intel_syntax (syntax_flag) char *string = input_line_pointer; int e = get_symbol_end (); - if (strcmp(string, "prefix") == 0) + if (strcmp (string, "prefix") == 0) ask_naked_reg = 1; - else if (strcmp(string, "noprefix") == 0) + else if (strcmp (string, "noprefix") == 0) ask_naked_reg = -1; else - as_bad (_("Bad argument to syntax directive.")); + as_bad (_("bad argument to syntax directive.")); *input_line_pointer = e; } demand_empty_rest_of_line (); - + intel_syntax = syntax_flag; if (ask_naked_reg == 0) { #ifdef BFD_ASSEMBLER allow_naked_reg = (intel_syntax - && (bfd_get_symbol_leading_char (stdoutput) != '\0')); + && (bfd_get_symbol_leading_char (stdoutput) != '\0')); #else - allow_naked_reg = 0; /* conservative default */ + /* Conservative default. */ + allow_naked_reg = 0; #endif } else allow_naked_reg = (ask_naked_reg < 0); } +static void +set_cpu_arch (dummy) + int dummy ATTRIBUTE_UNUSED; +{ + SKIP_WHITESPACE (); + + if (! is_end_of_line[(unsigned char) *input_line_pointer]) + { + char *string = input_line_pointer; + int e = get_symbol_end (); + int i; + + for (i = 0; cpu_arch[i].name; i++) + { + if (strcmp (string, cpu_arch[i].name) == 0) + { + cpu_arch_name = cpu_arch[i].name; + cpu_arch_flags = cpu_arch[i].flags; + break; + } + } + if (!cpu_arch[i].name) + as_bad (_("no such architecture: `%s'"), string); + + *input_line_pointer = e; + } + else + as_bad (_("missing cpu architecture")); + + demand_empty_rest_of_line (); +} + const pseudo_typeS md_pseudo_table[] = { -#ifndef I386COFF - {"bss", s_bss, 0}, -#endif #if !defined(OBJ_AOUT) && !defined(USE_ALIGN_PTWO) {"align", s_align_bytes, 0}, #else {"align", s_align_ptwo, 0}, +#endif + {"arch", set_cpu_arch, 0}, +#ifndef I386COFF + {"bss", s_bss, 0}, #endif {"ffloat", float_cons, 'f'}, {"dfloat", float_cons, 'd'}, @@ -569,35 +679,39 @@ const pseudo_typeS md_pseudo_table[] = {"value", cons, 2}, {"noopt", s_ignore, 0}, {"optim", s_ignore, 0}, + {"code16gcc", set_16bit_gcc_code_flag, 1}, {"code16", set_16bit_code_flag, 1}, {"code32", set_16bit_code_flag, 0}, {"intel_syntax", set_intel_syntax, 1}, {"att_syntax", set_intel_syntax, 0}, + {"file", dwarf2_directive_file, 0}, + {"loc", dwarf2_directive_loc, 0}, {0, 0, 0} }; -/* for interface with expression () */ +/* For interface with expression (). */ extern char *input_line_pointer; -/* hash table for instruction mnemonic lookup */ +/* Hash table for instruction mnemonic lookup. */ static struct hash_control *op_hash; -/* hash table for register lookup */ + +/* Hash table for register lookup. */ static struct hash_control *reg_hash; - void md_begin () { const char *hash_err; - /* initialize op_hash hash table */ + /* Initialize op_hash hash table. */ op_hash = hash_new (); { register const template *optab; register templates *core_optab; - optab = i386_optab; /* setup for loop */ + /* Setup for loop. */ + optab = i386_optab; core_optab = (templates *) xmalloc (sizeof (templates)); core_optab->start = optab; @@ -608,7 +722,7 @@ md_begin () || strcmp (optab->name, (optab - 1)->name) != 0) { /* different name --> ship out current template list; - add to hash table; & begin anew */ + add to hash table; & begin anew. */ core_optab->end = optab; hash_err = hash_insert (op_hash, (optab - 1)->name, @@ -628,7 +742,7 @@ md_begin () } } - /* initialize reg_hash hash table */ + /* Initialize reg_hash hash table. */ reg_hash = hash_new (); { register const reg_entry *regtab; @@ -643,7 +757,7 @@ md_begin () } } - /* fill in lexical tables: mnemonic_chars, operand_chars. */ + /* Fill in lexical tables: mnemonic_chars, operand_chars. */ { register int c; register char *p; @@ -682,8 +796,6 @@ md_begin () #ifdef LEX_AT identifier_chars['@'] = '@'; #endif - register_chars[')'] = ')'; - register_chars['('] = '('; digit_chars['-'] = '-'; identifier_chars['_'] = '_'; identifier_chars['.'] = '.'; @@ -710,10 +822,9 @@ i386_print_statistics (file) hash_print_statistics (file, "i386 register", reg_hash); } - #ifdef DEBUG386 -/* debugging routines for md_assemble */ +/* Debugging routines for md_assemble. */ static void pi PARAMS ((char *, i386_insn *)); static void pte PARAMS ((template *)); static void pt PARAMS ((unsigned int)); @@ -741,11 +852,11 @@ pi (line, x) fprintf (stdout, "\n"); if (x->types[i] & (Reg | SReg2 | SReg3 | Control | Debug | Test | RegMMX | RegXMM)) - fprintf (stdout, "%s\n", x->regs[i]->reg_name); + fprintf (stdout, "%s\n", x->op[i].regs->reg_name); if (x->types[i] & Imm) - pe (x->imms[i]); + pe (x->op[i].imms); if (x->types[i] & Disp) - pe (x->disps[i]); + pe (x->op[i].disps); } } @@ -755,8 +866,7 @@ pte (t) { int i; fprintf (stdout, " %d operands ", t->operands); - fprintf (stdout, "opcode %x ", - t->base_opcode); + fprintf (stdout, "opcode %x ", t->base_opcode); if (t->extension_opcode != None) fprintf (stdout, "ext %x ", t->extension_opcode); if (t->opcode_modifier & D) @@ -776,9 +886,9 @@ static void pe (e) expressionS *e; { - fprintf (stdout, " operation %d\n", e->X_op); - fprintf (stdout, " add_number %d (%x)\n", - e->X_add_number, e->X_add_number); + fprintf (stdout, " operation %d\n", e->X_op); + fprintf (stdout, " add_number %ld (%lx)\n", + (long) e->X_add_number, (long) e->X_add_number); if (e->X_add_symbol) { fprintf (stdout, " add_symbol "); @@ -871,8 +981,8 @@ tc_i386_force_relocation (fixp) return 1; return 0; #else - /* For COFF */ - return fixp->fx_r_type==7; + /* For COFF. */ + return fixp->fx_r_type == 7; #endif } @@ -886,7 +996,8 @@ reloc (size, pcrel, other) int pcrel; bfd_reloc_code_real_type other; { - if (other != NO_RELOC) return other; + if (other != NO_RELOC) + return other; if (pcrel) { @@ -896,7 +1007,7 @@ reloc (size, pcrel, other) case 2: return BFD_RELOC_16_PCREL; case 4: return BFD_RELOC_32_PCREL; } - as_bad (_("Can not do %d byte pc-relative relocation"), size); + as_bad (_("can not do %d byte pc-relative relocation"), size); } else { @@ -906,30 +1017,29 @@ reloc (size, pcrel, other) case 2: return BFD_RELOC_16; case 4: return BFD_RELOC_32; } - as_bad (_("Can not do %d byte relocation"), size); + as_bad (_("can not do %d byte relocation"), size); } return BFD_RELOC_NONE; } -/* - * Here we decide which fixups can be adjusted to make them relative to - * the beginning of the section instead of the symbol. Basically we need - * to make sure that the dynamic relocations are done correctly, so in - * some cases we force the original symbol to be used. - */ +/* Here we decide which fixups can be adjusted to make them relative to + the beginning of the section instead of the symbol. Basically we need + to make sure that the dynamic relocations are done correctly, so in + some cases we force the original symbol to be used. */ + int -tc_i386_fix_adjustable(fixP) - fixS * fixP; +tc_i386_fix_adjustable (fixP) + fixS *fixP; { -#ifdef OBJ_ELF - /* Prevent all adjustments to global symbols. */ - if (S_IS_EXTERN (fixP->fx_addsy)) - return 0; - if (S_IS_WEAK (fixP->fx_addsy)) +#if defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF) + /* Prevent all adjustments to global symbols, or else dynamic + linking will not work correctly. */ + if (S_IS_EXTERNAL (fixP->fx_addsy) + || S_IS_WEAK (fixP->fx_addsy)) return 0; #endif - /* adjust_reloc_syms doesn't know about the GOT */ + /* adjust_reloc_syms doesn't know about the GOT. */ if (fixP->fx_r_type == BFD_RELOC_386_GOTOFF || fixP->fx_r_type == BFD_RELOC_386_PLT32 || fixP->fx_r_type == BFD_RELOC_386_GOT32 @@ -949,12 +1059,14 @@ tc_i386_fix_adjustable(fixP) #define BFD_RELOC_386_GOTOFF 0 #endif -int +static int intel_float_operand PARAMS ((char *mnemonic)); + +static int intel_float_operand (mnemonic) char *mnemonic; { - if (mnemonic[0] == 'f' && mnemonic[1] =='i') - return 0; + if (mnemonic[0] == 'f' && mnemonic[1] == 'i') + return 2; if (mnemonic[0] == 'f') return 1; @@ -970,7 +1082,7 @@ void md_assemble (line) char *line; { - /* Points to template once we've found it. */ + /* Points to template once we've found it. */ const template *t; /* Count the size of the instruction generated. */ @@ -980,23 +1092,23 @@ md_assemble (line) char mnemonic[MAX_MNEM_SIZE]; - /* Initialize globals. */ + /* Initialize globals. */ memset (&i, '\0', sizeof (i)); for (j = 0; j < MAX_OPERANDS; j++) i.disp_reloc[j] = NO_RELOC; memset (disp_expressions, '\0', sizeof (disp_expressions)); memset (im_expressions, '\0', sizeof (im_expressions)); - save_stack_p = save_stack; /* reset stack pointer */ + save_stack_p = save_stack; /* First parse an instruction mnemonic & call i386_operand for the operands. We assume that the scrubber has arranged it so that line[0] is the valid - start of a (possibly prefixed) mnemonic. */ + start of a (possibly prefixed) mnemonic. */ { char *l = line; char *token_start = l; char *mnem_p; - /* Non-zero if we found a prefix only acceptable with string insns. */ + /* Non-zero if we found a prefix only acceptable with string insns. */ const char *expecting_string_instruction = NULL; while (1) @@ -1007,7 +1119,7 @@ md_assemble (line) mnem_p++; if (mnem_p >= mnemonic + sizeof (mnemonic)) { - as_bad (_("no such 386 instruction: `%s'"), token_start); + as_bad (_("no such instruction: `%s'"), token_start); return; } l++; @@ -1053,8 +1165,7 @@ md_assemble (line) case 0: return; case 2: - expecting_string_instruction = - current_templates->start->name; + expecting_string_instruction = current_templates->start->name; break; } /* Skip past PREFIX_SEPARATOR and reset token_start. */ @@ -1066,39 +1177,50 @@ md_assemble (line) if (!current_templates) { - /* See if we can get a match by trimming off a suffix. */ + /* See if we can get a match by trimming off a suffix. */ switch (mnem_p[-1]) { - case DWORD_MNEM_SUFFIX: case WORD_MNEM_SUFFIX: case BYTE_MNEM_SUFFIX: case SHORT_MNEM_SUFFIX: -#if LONG_MNEM_SUFFIX != DWORD_MNEM_SUFFIX case LONG_MNEM_SUFFIX: -#endif i.suffix = mnem_p[-1]; mnem_p[-1] = '\0'; current_templates = hash_find (op_hash, mnemonic); - break; - - /* Intel Syntax */ - case INTEL_DWORD_MNEM_SUFFIX: - if (intel_syntax) - { - i.suffix = mnem_p[-1]; - mnem_p[-1] = '\0'; - current_templates = hash_find (op_hash, mnemonic); - break; - } + break; + + /* Intel Syntax. */ + case DWORD_MNEM_SUFFIX: + if (intel_syntax) + { + i.suffix = mnem_p[-1]; + mnem_p[-1] = '\0'; + current_templates = hash_find (op_hash, mnemonic); + break; + } } if (!current_templates) { - as_bad (_("no such 386 instruction: `%s'"), token_start); + as_bad (_("no such instruction: `%s'"), token_start); return; } } - /* check for rep/repne without a string instruction */ + /* Check if instruction is supported on specified architecture. */ + if (cpu_arch_flags != 0) + { + if (current_templates->start->cpu_flags & ~cpu_arch_flags) + { + as_warn (_("`%s' is not supported on `%s'"), + current_templates->start->name, cpu_arch_name); + } + else if ((Cpu386 & ~cpu_arch_flags) && !flag_16bit_code) + { + as_warn (_("use .code16 to ensure correct addressing mode")); + } + } + + /* Check for rep/repne without a string instruction. */ if (expecting_string_instruction && !(current_templates->start->opcode_modifier & IsString)) { @@ -1107,20 +1229,18 @@ md_assemble (line) return; } - /* There may be operands to parse. */ + /* There may be operands to parse. */ if (*l != END_OF_INSN) { - /* parse operands */ - - /* 1 if operand is pending after ','. */ + /* 1 if operand is pending after ','. */ unsigned int expecting_operand = 0; - /* Non-zero if operand parens not balanced. */ + /* Non-zero if operand parens not balanced. */ unsigned int paren_not_balanced; do { - /* skip optional white space before operand */ + /* Skip optional white space before operand. */ if (is_space_char (*l)) ++l; if (!is_operand_char (*l) && *l != END_OF_INSN) @@ -1138,10 +1258,10 @@ md_assemble (line) { if (paren_not_balanced) { - if (!intel_syntax) + if (!intel_syntax) as_bad (_("unbalanced parenthesis in operand %d."), i.operands + 1); - else + else as_bad (_("unbalanced brackets in operand %d."), i.operands + 1); return; @@ -1156,24 +1276,24 @@ md_assemble (line) i.operands + 1); return; } - if (!intel_syntax) - { + if (!intel_syntax) + { if (*l == '(') ++paren_not_balanced; if (*l == ')') --paren_not_balanced; - } - else - { + } + else + { if (*l == '[') ++paren_not_balanced; if (*l == ']') --paren_not_balanced; - } + } l++; } if (l != token_start) - { /* yes, we've read in another operand */ + { /* Yes, we've read in another operand. */ unsigned int operand_ok; this_operand = i.operands++; if (i.operands > MAX_OPERANDS) @@ -1182,15 +1302,17 @@ md_assemble (line) MAX_OPERANDS); return; } - /* now parse operand adding info to 'i' as we go along */ + /* Now parse operand adding info to 'i' as we go along. */ END_STRING_AND_SAVE (l); - if (intel_syntax) - operand_ok = i386_intel_operand (token_start, intel_float_operand (mnemonic)); - else - operand_ok = i386_operand (token_start); + if (intel_syntax) + operand_ok = + i386_intel_operand (token_start, + intel_float_operand (mnemonic)); + else + operand_ok = i386_operand (token_start); - RESTORE_END_STRING (l); /* restore old contents */ + RESTORE_END_STRING (l); if (!operand_ok) return; } @@ -1209,17 +1331,18 @@ md_assemble (line) } } - /* now *l must be either ',' or END_OF_INSN */ + /* Now *l must be either ',' or END_OF_INSN. */ if (*l == ',') { if (*++l == END_OF_INSN) - { /* just skip it, if it's \n complain */ + { + /* Just skip it, if it's \n complain. */ goto expecting_operand_after_comma; } expecting_operand = 1; } } - while (*l != END_OF_INSN); /* until we get end of insn */ + while (*l != END_OF_INSN); } } @@ -1228,12 +1351,11 @@ md_assemble (line) Next, we find a template that matches the given insn, making sure the overlap of the given operands types is consistent - with the template operand types. */ + with the template operand types. */ #define MATCH(overlap, given, template) \ - ((overlap) \ - && ((given) & BaseIndex) == ((overlap) & BaseIndex) \ - && ((given) & JumpAbsolute) == ((template) & JumpAbsolute)) + ((overlap & ~JumpAbsolute) \ + && ((given) & (BaseIndex|JumpAbsolute)) == ((overlap) & (BaseIndex|JumpAbsolute))) /* If given types r0 and r1 are registers they must be of the same type unless the expected operand type register overlap is null. @@ -1245,90 +1367,141 @@ md_assemble (line) { register unsigned int overlap0, overlap1; - expressionS *exp; unsigned int overlap2; unsigned int found_reverse_match; int suffix_check; - /* All intel opcodes have reversed operands except for BOUND and ENTER */ - if (intel_syntax - && (strcmp (mnemonic, "enter") != 0) - && (strcmp (mnemonic, "bound") != 0) - && (strncmp (mnemonic, "fsub", 4) !=0) - && (strncmp (mnemonic, "fdiv", 4) !=0)) + /* All intel opcodes have reversed operands except for "bound" and + "enter". We also don't reverse intersegment "jmp" and "call" + instructions with 2 immediate operands so that the immediate segment + precedes the offset, as it does when in AT&T mode. "enter" and the + intersegment "jmp" and "call" instructions are the only ones that + have two immediate operands. */ + if (intel_syntax && i.operands > 1 + && (strcmp (mnemonic, "bound") != 0) + && !((i.types[0] & Imm) && (i.types[1] & Imm))) + { + union i386_op temp_op; + unsigned int temp_type; +#ifdef BFD_ASSEMBLER + enum bfd_reloc_code_real temp_disp_reloc; +#else + int temp_disp_reloc; +#endif + int xchg1 = 0; + int xchg2 = 0; + + if (i.operands == 2) + { + xchg1 = 0; + xchg2 = 1; + } + else if (i.operands == 3) + { + xchg1 = 0; + xchg2 = 2; + } + temp_type = i.types[xchg2]; + i.types[xchg2] = i.types[xchg1]; + i.types[xchg1] = temp_type; + temp_op = i.op[xchg2]; + i.op[xchg2] = i.op[xchg1]; + i.op[xchg1] = temp_op; + temp_disp_reloc = i.disp_reloc[xchg2]; + i.disp_reloc[xchg2] = i.disp_reloc[xchg1]; + i.disp_reloc[xchg1] = temp_disp_reloc; + + if (i.mem_operands == 2) + { + const seg_entry *temp_seg; + temp_seg = i.seg[0]; + i.seg[0] = i.seg[1]; + i.seg[1] = temp_seg; + } + } + + if (i.imm_operands) + { + /* Try to ensure constant immediates are represented in the smallest + opcode possible. */ + char guess_suffix = 0; + int op; + + if (i.suffix) + guess_suffix = i.suffix; + else if (i.reg_operands) + { + /* Figure out a suffix from the last register operand specified. + We can't do this properly yet, ie. excluding InOutPortReg, + but the following works for instructions with immediates. + In any case, we can't set i.suffix yet. */ + for (op = i.operands; --op >= 0;) + if (i.types[op] & Reg) + { + if (i.types[op] & Reg8) + guess_suffix = BYTE_MNEM_SUFFIX; + else if (i.types[op] & Reg16) + guess_suffix = WORD_MNEM_SUFFIX; + break; + } + } + else if (flag_16bit_code ^ (i.prefix[DATA_PREFIX] != 0)) + guess_suffix = WORD_MNEM_SUFFIX; + + for (op = i.operands; --op >= 0;) + if ((i.types[op] & Imm) + && i.op[op].imms->X_op == O_constant) + { + /* If a suffix is given, this operand may be shortened. */ + switch (guess_suffix) + { + case WORD_MNEM_SUFFIX: + i.types[op] |= Imm16; + break; + case BYTE_MNEM_SUFFIX: + i.types[op] |= Imm16 | Imm8 | Imm8S; + break; + } + + /* If this operand is at most 16 bits, convert it to a + signed 16 bit number before trying to see whether it will + fit in an even smaller size. This allows a 16-bit operand + such as $0xffe0 to be recognised as within Imm8S range. */ + if ((i.types[op] & Imm16) + && (i.op[op].imms->X_add_number & ~(offsetT)0xffff) == 0) + { + i.op[op].imms->X_add_number = + (((i.op[op].imms->X_add_number & 0xffff) ^ 0x8000) - 0x8000); + } + i.types[op] |= smallest_imm_type ((long) i.op[op].imms->X_add_number); + } + } + + if (i.disp_operands) { - const reg_entry *temp_reg; - expressionS *temp_disp; - expressionS *temp_imm; - unsigned int temp_type; - int xchg1, xchg2; - - if (i.operands == 2) - { - xchg1 = 0; - xchg2 = 1; - } - else if (i.operands == 3) - { - xchg1 = 0; - xchg2 = 2; - } - - if (i.operands > 1) - { - temp_type = i.types[xchg2]; - if (temp_type & (Reg | FloatReg)) - temp_reg = i.regs[xchg2]; - else if (temp_type & Imm) - temp_imm = i.imms[xchg2]; - else if (temp_type & Disp) - temp_disp = i.disps[xchg2]; - - i.types[xchg2] = i.types[xchg1]; - - if (i.types[xchg1] & (Reg | FloatReg)) - { - i.regs[xchg2] = i.regs[xchg1]; - i.regs[xchg1] = NULL; - } - else if (i.types[xchg2] & Imm) - { - i.imms[xchg2] = i.imms[xchg1]; - i.imms[xchg1] = NULL; - } - else if (i.types[xchg2] & Disp) - { - i.disps[xchg2] = i.disps[xchg1]; - i.disps[xchg1] = NULL; - } - - if (temp_type & (Reg | FloatReg)) - { - i.regs[xchg1] = temp_reg; - if (! (i.types[xchg1] & (Reg | FloatReg))) - i.regs[xchg2] = NULL; - } - else if (temp_type & Imm) - { - i.imms[xchg1] = temp_imm; - if (! (i.types[xchg1] & Imm)) - i.imms[xchg2] = NULL; - } - else if (temp_type & Disp) - { - i.disps[xchg1] = temp_disp; - if (! (i.types[xchg1] & Disp)) - i.disps[xchg2] = NULL; - } - - i.types[xchg1] = temp_type; - } - if (!strcmp(mnemonic,"jmp") - || !strcmp (mnemonic, "call")) - if ((i.types[0] & Reg) || i.types[0] & BaseIndex) - i.types[0] |= JumpAbsolute; - - } + /* Try to use the smallest displacement type too. */ + int op; + + for (op = i.operands; --op >= 0;) + if ((i.types[op] & Disp) + && i.op[op].imms->X_op == O_constant) + { + offsetT disp = i.op[op].disps->X_add_number; + + if (i.types[op] & Disp16) + { + /* We know this operand is at most 16 bits, so + convert to a signed 16 bit number before trying + to see whether it will fit in an even smaller + size. */ + + disp = (((disp & 0xffff) ^ 0x8000) - 0x8000); + } + if (fits_in_signed_byte (disp)) + i.types[op] |= Disp8; + } + } + overlap0 = 0; overlap1 = 0; overlap2 = 0; @@ -1340,33 +1513,32 @@ md_assemble (line) : (i.suffix == SHORT_MNEM_SUFFIX ? No_sSuf : (i.suffix == LONG_MNEM_SUFFIX - ? No_lSuf - : (i.suffix == INTEL_DWORD_MNEM_SUFFIX - ? No_dSuf - : (i.suffix == LONG_DOUBLE_MNEM_SUFFIX ? No_xSuf : 0)))))); + ? No_lSuf + : (i.suffix == DWORD_MNEM_SUFFIX + ? No_dSuf + : (i.suffix == LONG_DOUBLE_MNEM_SUFFIX ? No_xSuf : 0)))))); for (t = current_templates->start; t < current_templates->end; t++) { - /* Must have right number of operands. */ + /* Must have right number of operands. */ if (i.operands != t->operands) continue; - /* For some opcodes, don't check the suffix */ - if (intel_syntax) - { - if (strcmp (t->name, "fnstcw") - && strcmp (t->name, "fldcw") - && (t->opcode_modifier & suffix_check)) - continue; - } - /* Must not have disallowed suffix. */ - else if ((t->opcode_modifier & suffix_check)) - continue; + /* Check the suffix, except for some instructions in intel mode. */ + if ((t->opcode_modifier & suffix_check) + && !(intel_syntax + && (t->opcode_modifier & IgnoreSize)) + && !(intel_syntax + && t->base_opcode == 0xd9 + && (t->extension_opcode == 5 /* 0xd9,5 "fldcw" */ + || t->extension_opcode == 7))) /* 0xd9,7 "f{n}stcw" */ + continue; else if (!t->operands) - break; /* 0 operands always matches */ + /* 0 operands always matches. */ + break; overlap0 = i.types[0] & t->operand_types[0]; switch (t->operands) @@ -1385,12 +1557,11 @@ md_assemble (line) overlap1, i.types[1], t->operand_types[1])) { - - /* check if other direction is valid ... */ + /* Check if other direction is valid ... */ if ((t->opcode_modifier & (D|FloatD)) == 0) continue; - /* try reversing direction of operands */ + /* Try reversing direction of operands. */ overlap0 = i.types[0] & t->operand_types[1]; overlap1 = i.types[1] & t->operand_types[0]; if (!MATCH (overlap0, i.types[0], t->operand_types[1]) @@ -1400,7 +1571,7 @@ md_assemble (line) overlap1, i.types[1], t->operand_types[0])) { - /* does not match either direction */ + /* Does not match either direction. */ continue; } /* found_reverse_match holds which of D or FloatDR @@ -1408,7 +1579,7 @@ md_assemble (line) found_reverse_match = t->opcode_modifier & (D|FloatDR); break; } - /* found a forward 2 operand match here */ + /* Found a forward 2 operand match here. */ if (t->operands == 3) { /* Here we make use of the fact that there are no @@ -1420,43 +1591,67 @@ md_assemble (line) || !CONSISTENT_REGISTER_MATCH (overlap1, i.types[1], t->operand_types[1], overlap2, i.types[2], - t->operand_types[2])) + t->operand_types[2])) - continue; + continue; } - /* found either forward/reverse 2 or 3 operand match here: - slip through to break */ + /* Found either forward/reverse 2 or 3 operand match here: + slip through to break. */ } - break; /* we've found a match; break out of loop */ - } /* for (t = ... */ + /* We've found a match; break out of loop. */ + break; + } if (t == current_templates->end) - { /* we found no match */ + { + /* We found no match. */ as_bad (_("suffix or operands invalid for `%s'"), current_templates->start->name); return; } - if ((t->opcode_modifier & (IsPrefix|IgnoreSize)) == (IsPrefix|IgnoreSize)) + if (!quiet_warnings) { - /* Warn them that a data or address size prefix doesn't affect - assembly of the next line of code. */ - as_warn (_("stand-alone `%s' prefix"), t->name); + if (!intel_syntax + && ((i.types[0] & JumpAbsolute) + != (t->operand_types[0] & JumpAbsolute))) + { + as_warn (_("indirect %s without `*'"), t->name); + } + + if ((t->opcode_modifier & (IsPrefix|IgnoreSize)) + == (IsPrefix|IgnoreSize)) + { + /* Warn them that a data or address size prefix doesn't + affect assembly of the next line of code. */ + as_warn (_("stand-alone `%s' prefix"), t->name); + } } /* Copy the template we found. */ i.tm = *t; if (found_reverse_match) { + /* If we found a reverse match we must alter the opcode + direction bit. found_reverse_match holds bits to change + (different for int & float insns). */ + + i.tm.base_opcode ^= found_reverse_match; + i.tm.operand_types[0] = t->operand_types[1]; i.tm.operand_types[1] = t->operand_types[0]; } + /* Undo SYSV386_COMPAT brokenness when in Intel mode. See i386.h */ + if (SYSV386_COMPAT + && intel_syntax + && (i.tm.base_opcode & 0xfffffde0) == 0xdce0) + i.tm.base_opcode ^= FloatR; if (i.tm.opcode_modifier & FWait) if (! add_prefix (FWAIT_OPCODE)) return; - /* Check string instruction segment overrides */ + /* Check string instruction segment overrides. */ if ((i.tm.opcode_modifier & IsString) != 0 && i.mem_operands != 0) { int mem_op = (i.types[0] & AnyMem) ? 0 : 1; @@ -1494,47 +1689,48 @@ md_assemble (line) if (i.tm.opcode_modifier & Size16) i.suffix = WORD_MNEM_SUFFIX; else - i.suffix = DWORD_MNEM_SUFFIX; + i.suffix = LONG_MNEM_SUFFIX; } else if (i.reg_operands) { /* If there's no instruction mnemonic suffix we try to invent one - based on register operands. */ + based on register operands. */ if (!i.suffix) { /* We take i.suffix from the last register operand specified, Destination register type is more significant than source register type. */ int op; - for (op = i.operands; --op >= 0; ) - if (i.types[op] & Reg) + for (op = i.operands; --op >= 0;) + if ((i.types[op] & Reg) + && !(i.tm.operand_types[op] & InOutPortReg)) { i.suffix = ((i.types[op] & Reg8) ? BYTE_MNEM_SUFFIX : (i.types[op] & Reg16) ? WORD_MNEM_SUFFIX : - DWORD_MNEM_SUFFIX); + LONG_MNEM_SUFFIX); break; } } else if (i.suffix == BYTE_MNEM_SUFFIX) { int op; - for (op = i.operands; --op >= 0; ) + for (op = i.operands; --op >= 0;) { /* If this is an eight bit register, it's OK. If it's the 16 or 32 bit version of an eight bit register, - we will just use the low portion, and that's OK too. */ + we will just use the low portion, and that's OK too. */ if (i.types[op] & Reg8) continue; - /* movzx and movsx should not generate this warning. */ - if (intel_syntax - && (i.tm.base_opcode == 0xfb7 - || i.tm.base_opcode == 0xfb6 - || i.tm.base_opcode == 0xfbe - || i.tm.base_opcode == 0xfbf)) - continue; + /* movzx and movsx should not generate this warning. */ + if (intel_syntax + && (i.tm.base_opcode == 0xfb7 + || i.tm.base_opcode == 0xfb6 + || i.tm.base_opcode == 0xfbe + || i.tm.base_opcode == 0xfbf)) + continue; - if ((i.types[op] & WordReg) && i.regs[op]->reg_num < 4 + if ((i.types[op] & WordReg) && i.op[op].regs->reg_num < 4 #if 0 /* Check that the template allows eight bit regs This kills insns such as `orb $1,%edx', which @@ -1544,51 +1740,54 @@ md_assemble (line) ) { #if REGISTER_WARNINGS - if ((i.tm.operand_types[op] & InOutPortReg) == 0) + if (!quiet_warnings + && (i.tm.operand_types[op] & InOutPortReg) == 0) as_warn (_("using `%%%s' instead of `%%%s' due to `%c' suffix"), - (i.regs[op] - (i.types[op] & Reg16 ? 8 : 16))->reg_name, - i.regs[op]->reg_name, + (i.op[op].regs - (i.types[op] & Reg16 ? 8 : 16))->reg_name, + i.op[op].regs->reg_name, i.suffix); #endif continue; } - /* Any other register is bad */ + /* Any other register is bad. */ if (i.types[op] & (Reg | RegMMX | RegXMM | SReg2 | SReg3 | Control | Debug | Test | FloatReg | FloatAcc)) { as_bad (_("`%%%s' not allowed with `%s%c'"), - i.regs[op]->reg_name, + i.op[op].regs->reg_name, i.tm.name, i.suffix); return; } } } - else if (i.suffix == DWORD_MNEM_SUFFIX) + else if (i.suffix == LONG_MNEM_SUFFIX) { int op; - for (op = i.operands; --op >= 0; ) + + for (op = i.operands; --op >= 0;) /* Reject eight bit registers, except where the template requires them. (eg. movzb) */ if ((i.types[op] & Reg8) != 0 - && (i.tm.operand_types[op] & (Reg16|Reg32|Acc)) != 0) + && (i.tm.operand_types[op] & (Reg16 | Reg32 | Acc)) != 0) { as_bad (_("`%%%s' not allowed with `%s%c'"), - i.regs[op]->reg_name, + i.op[op].regs->reg_name, i.tm.name, i.suffix); return; } #if REGISTER_WARNINGS /* Warn if the e prefix on a general reg is missing. */ - else if ((i.types[op] & Reg16) != 0 + else if (!quiet_warnings + && (i.types[op] & Reg16) != 0 && (i.tm.operand_types[op] & (Reg32|Acc)) != 0) { as_warn (_("using `%%%s' instead of `%%%s' due to `%c' suffix"), - (i.regs[op] + 8)->reg_name, - i.regs[op]->reg_name, + (i.op[op].regs + 8)->reg_name, + i.op[op].regs->reg_name, i.suffix); } #endif @@ -1596,32 +1795,40 @@ md_assemble (line) else if (i.suffix == WORD_MNEM_SUFFIX) { int op; - for (op = i.operands; --op >= 0; ) + for (op = i.operands; --op >= 0;) /* Reject eight bit registers, except where the template requires them. (eg. movzb) */ if ((i.types[op] & Reg8) != 0 && (i.tm.operand_types[op] & (Reg16|Reg32|Acc)) != 0) { as_bad (_("`%%%s' not allowed with `%s%c'"), - i.regs[op]->reg_name, + i.op[op].regs->reg_name, i.tm.name, i.suffix); return; } #if REGISTER_WARNINGS /* Warn if the e prefix on a general reg is present. */ - else if ((i.types[op] & Reg32) != 0 + else if (!quiet_warnings + && (i.types[op] & Reg32) != 0 && (i.tm.operand_types[op] & (Reg16|Acc)) != 0) { as_warn (_("using `%%%s' instead of `%%%s' due to `%c' suffix"), - (i.regs[op] - 8)->reg_name, - i.regs[op]->reg_name, + (i.op[op].regs - 8)->reg_name, + i.op[op].regs->reg_name, i.suffix); } #endif } + else if (intel_syntax && (i.tm.opcode_modifier & IgnoreSize)) + /* Do nothing if the instruction is going to ignore the prefix. */ + ; else - abort(); + abort (); + } + else if ((i.tm.opcode_modifier & DefaultSize) && !i.suffix) + { + i.suffix = stackop_size; } /* Make still unresolved immediate matches conform to size of immediate @@ -1632,12 +1839,12 @@ md_assemble (line) { if (i.suffix) { - overlap0 &= (i.suffix == BYTE_MNEM_SUFFIX ? (Imm8 | Imm8S) : - (i.suffix == WORD_MNEM_SUFFIX ? Imm16 : Imm32)); + overlap0 &= (i.suffix == BYTE_MNEM_SUFFIX ? (Imm8 | Imm8S) : + (i.suffix == WORD_MNEM_SUFFIX ? Imm16 : Imm32)); } else if (overlap0 == (Imm16 | Imm32)) { - overlap0 = + overlap0 = (flag_16bit_code ^ (i.prefix[DATA_PREFIX] != 0)) ? Imm16 : Imm32; } else @@ -1652,12 +1859,12 @@ md_assemble (line) { if (i.suffix) { - overlap1 &= (i.suffix == BYTE_MNEM_SUFFIX ? (Imm8 | Imm8S) : - (i.suffix == WORD_MNEM_SUFFIX ? Imm16 : Imm32)); + overlap1 &= (i.suffix == BYTE_MNEM_SUFFIX ? (Imm8 | Imm8S) : + (i.suffix == WORD_MNEM_SUFFIX ? Imm16 : Imm32)); } else if (overlap1 == (Imm16 | Imm32)) { - overlap1 = + overlap1 = (flag_16bit_code ^ (i.prefix[DATA_PREFIX] != 0)) ? Imm16 : Imm32; } else @@ -1672,7 +1879,7 @@ md_assemble (line) if (overlap0 & ImplicitRegister) i.reg_operands--; if (overlap0 & Imm1) - i.imm_operands = 0; /* kludge for shift insns */ + i.imm_operands = 0; /* kludge for shift insns. */ i.types[1] = overlap1; if (overlap1 & ImplicitRegister) @@ -1691,17 +1898,17 @@ md_assemble (line) return; } - /* For movzx and movsx, need to check the register type */ + /* For movzx and movsx, need to check the register type. */ if (intel_syntax - && (i.tm.base_opcode == 0xfb6 || i.tm.base_opcode == 0xfbe)) + && (i.tm.base_opcode == 0xfb6 || i.tm.base_opcode == 0xfbe)) if (i.suffix && i.suffix == BYTE_MNEM_SUFFIX) - { - unsigned int prefix = DATA_PREFIX_OPCODE; + { + unsigned int prefix = DATA_PREFIX_OPCODE; - if ((i.regs[1]->reg_type & Reg16) != 0) - if (!add_prefix (prefix)) - return; - } + if ((i.op[1].regs->reg_type & Reg16) != 0) + if (!add_prefix (prefix)) + return; + } if (i.suffix && i.suffix != BYTE_MNEM_SUFFIX) { @@ -1716,8 +1923,7 @@ md_assemble (line) /* Now select between word & dword operations via the operand size prefix, except for instructions that will ignore this prefix anyway. */ - if (((intel_syntax && (i.suffix == INTEL_DWORD_MNEM_SUFFIX)) - || i.suffix == DWORD_MNEM_SUFFIX + if (((intel_syntax && (i.suffix == DWORD_MNEM_SUFFIX)) || i.suffix == LONG_MNEM_SUFFIX) == flag_16bit_code && !(i.tm.opcode_modifier & IgnoreSize)) { @@ -1730,17 +1936,11 @@ md_assemble (line) } /* Size floating point instruction. */ if (i.suffix == LONG_MNEM_SUFFIX - || (intel_syntax && i.suffix == INTEL_DWORD_MNEM_SUFFIX)) + || (intel_syntax && i.suffix == DWORD_MNEM_SUFFIX)) { if (i.tm.opcode_modifier & FloatMF) i.tm.base_opcode ^= 4; } - - if (intel_syntax && i.suffix == LONG_DOUBLE_MNEM_SUFFIX) - { - if (i.tm.opcode_modifier & FloatMF) - i.tm.base_opcode ^= 2; - } } if (i.tm.opcode_modifier & ImmExt) @@ -1752,63 +1952,59 @@ md_assemble (line) expressionS *exp; - assert(i.imm_operands == 0 && i.operands <= 2); + assert (i.imm_operands == 0 && i.operands <= 2 && 2 < MAX_OPERANDS); exp = &im_expressions[i.imm_operands++]; - i.imms[i.operands] = exp; + i.op[i.operands].imms = exp; i.types[i.operands++] = Imm8; exp->X_op = O_constant; exp->X_add_number = i.tm.extension_opcode; i.tm.extension_opcode = None; } - /* For insns with operands there are more diddles to do to the opcode. */ + /* For insns with operands there are more diddles to do to the opcode. */ if (i.operands) { - /* Default segment register this instruction will use + /* Default segment register this instruction will use for memory accesses. 0 means unknown. This is only for optimizing out unnecessary segment overrides. */ const seg_entry *default_seg = 0; - /* If we found a reverse match we must alter the opcode - direction bit. found_reverse_match holds bits to change - (different for int & float insns). */ - - i.tm.base_opcode ^= found_reverse_match; - /* The imul $imm, %reg instruction is converted into imul $imm, %reg, %reg, and the clr %reg instruction is converted into xor %reg, %reg. */ if (i.tm.opcode_modifier & regKludge) { unsigned int first_reg_op = (i.types[0] & Reg) ? 0 : 1; - /* Pretend we saw the extra register operand. */ - i.regs[first_reg_op+1] = i.regs[first_reg_op]; + /* Pretend we saw the extra register operand. */ + assert (i.op[first_reg_op + 1].regs == 0); + i.op[first_reg_op + 1].regs = i.op[first_reg_op].regs; + i.types[first_reg_op + 1] = i.types[first_reg_op]; i.reg_operands = 2; } if (i.tm.opcode_modifier & ShortForm) { - /* The register or float register operand is in operand 0 or 1. */ + /* The register or float register operand is in operand 0 or 1. */ unsigned int op = (i.types[0] & (Reg | FloatReg)) ? 0 : 1; - /* Register goes in low 3 bits of opcode. */ - i.tm.base_opcode |= i.regs[op]->reg_num; - if ((i.tm.opcode_modifier & Ugh) != 0) + /* Register goes in low 3 bits of opcode. */ + i.tm.base_opcode |= i.op[op].regs->reg_num; + if (!quiet_warnings && (i.tm.opcode_modifier & Ugh) != 0) { /* Warn about some common errors, but press on regardless. The first case can be generated by gcc (<= 2.8.1). */ if (i.operands == 2) { - /* reversed arguments on faddp, fsubp, etc. */ + /* Reversed arguments on faddp, fsubp, etc. */ as_warn (_("translating to `%s %%%s,%%%s'"), i.tm.name, - i.regs[1]->reg_name, - i.regs[0]->reg_name); + i.op[1].regs->reg_name, + i.op[0].regs->reg_name); } else { - /* extraneous `l' suffix on fp insn */ + /* Extraneous `l' suffix on fp insn. */ as_warn (_("translating to `%s %%%s'"), i.tm.name, - i.regs[0]->reg_name); + i.op[0].regs->reg_name); } } } @@ -1817,10 +2013,10 @@ md_assemble (line) /* The opcode is completed (modulo i.tm.extension_opcode which must be put into the modrm byte). Now, we make the modrm & index base bytes based on all the - info we've collected. */ + info we've collected. */ /* i.reg_operands MUST be the number of real register operands; - implicit registers do not count. */ + implicit registers do not count. */ if (i.reg_operands == 2) { unsigned int source, dest; @@ -1841,17 +2037,17 @@ md_assemble (line) destination in the i.rm.reg field. */ if ((i.tm.operand_types[dest] & AnyMem) == 0) { - i.rm.reg = i.regs[dest]->reg_num; - i.rm.regmem = i.regs[source]->reg_num; + i.rm.reg = i.op[dest].regs->reg_num; + i.rm.regmem = i.op[source].regs->reg_num; } else { - i.rm.reg = i.regs[source]->reg_num; - i.rm.regmem = i.regs[dest]->reg_num; + i.rm.reg = i.op[source].regs->reg_num; + i.rm.regmem = i.op[dest].regs->reg_num; } } else - { /* if it's not 2 reg operands... */ + { /* If it's not 2 reg operands... */ if (i.mem_operands) { unsigned int fake_zero_displacement = 0; @@ -1868,7 +2064,7 @@ md_assemble (line) fake_zero_displacement = 1; if (! i.index_reg) { - /* Operand is just */ + /* Operand is just */ if (flag_16bit_code ^ (i.prefix[ADDR_PREFIX] != 0)) { i.rm.regmem = NO_BASE_REGISTER_16; @@ -1882,47 +2078,47 @@ md_assemble (line) i.types[op] |= Disp32; } } - else /* ! i.base_reg && i.index_reg */ + else /* ! i.base_reg && i.index_reg */ { i.sib.index = i.index_reg->reg_num; i.sib.base = NO_BASE_REGISTER; i.sib.scale = i.log2_scale_factor; i.rm.regmem = ESCAPE_TO_TWO_BYTE_ADDRESSING; i.types[op] &= ~Disp; - i.types[op] |= Disp32; /* Must be 32 bit */ + i.types[op] |= Disp32; /* Must be 32 bit. */ } } else if (i.base_reg->reg_type & Reg16) { switch (i.base_reg->reg_num) { - case 3: /* (%bx) */ + case 3: /* (%bx) */ if (! i.index_reg) i.rm.regmem = 7; - else /* (%bx,%si) -> 0, or (%bx,%di) -> 1 */ + else /* (%bx,%si) -> 0, or (%bx,%di) -> 1 */ i.rm.regmem = i.index_reg->reg_num - 6; break; - case 5: /* (%bp) */ + case 5: /* (%bp) */ default_seg = &ss; if (! i.index_reg) { i.rm.regmem = 6; if ((i.types[op] & Disp) == 0) { - /* fake (%bp) into 0(%bp) */ + /* fake (%bp) into 0(%bp) */ i.types[op] |= Disp8; fake_zero_displacement = 1; } } - else /* (%bp,%si) -> 2, or (%bp,%di) -> 3 */ + else /* (%bp,%si) -> 2, or (%bp,%di) -> 3 */ i.rm.regmem = i.index_reg->reg_num - 6 + 2; break; - default: /* (%si) -> 4 or (%di) -> 5 */ + default: /* (%si) -> 4 or (%di) -> 5 */ i.rm.regmem = i.base_reg->reg_num - 6 + 4; } i.rm.mode = mode_from_disp_size (i.types[op]); } - else /* i.base_reg and 32 bit mode */ + else /* i.base_reg and 32 bit mode */ { i.rm.regmem = i.base_reg->reg_num; i.sib.base = i.base_reg->reg_num; @@ -1967,9 +2163,12 @@ md_assemble (line) if (fake_zero_displacement) { /* Fakes a zero displacement assuming that i.types[op] - holds the correct displacement size. */ + holds the correct displacement size. */ + expressionS *exp; + + assert (i.op[op].disps == 0); exp = &disp_expressions[i.disp_operands++]; - i.disps[op] = exp; + i.op[op].disps = exp; exp->X_op = O_constant; exp->X_add_number = 0; exp->X_add_symbol = (symbolS *) 0; @@ -1981,7 +2180,7 @@ md_assemble (line) operand (if any) based on i.tm.extension_opcode. Again, we must be careful to make sure that segment/control/debug/test/MMX registers are coded - into the i.rm.reg field. */ + into the i.rm.reg field. */ if (i.reg_operands) { unsigned int op = @@ -1997,11 +2196,11 @@ md_assemble (line) ? 1 : 2)); /* If there is an extension opcode to put here, the - register number must be put into the regmem field. */ + register number must be put into the regmem field. */ if (i.tm.extension_opcode != None) - i.rm.regmem = i.regs[op]->reg_num; + i.rm.regmem = i.op[op].regs->reg_num; else - i.rm.reg = i.regs[op]->reg_num; + i.rm.reg = i.op[op].regs->reg_num; /* Now, if no memory operand has set i.rm.mode = 0, 1, 2 we must set it to 3 to indicate this is a register @@ -2010,19 +2209,20 @@ md_assemble (line) i.rm.mode = 3; } - /* Fill in i.rm.reg field with extension opcode (if any). */ + /* Fill in i.rm.reg field with extension opcode (if any). */ if (i.tm.extension_opcode != None) i.rm.reg = i.tm.extension_opcode; } } else if (i.tm.opcode_modifier & (Seg2ShortForm | Seg3ShortForm)) { - if (i.tm.base_opcode == POP_SEG_SHORT && i.regs[0]->reg_num == 1) + if (i.tm.base_opcode == POP_SEG_SHORT + && i.op[0].regs->reg_num == 1) { as_bad (_("you can't `pop %%cs'")); return; } - i.tm.base_opcode |= (i.regs[0]->reg_num << 3); + i.tm.base_opcode |= (i.op[0].regs->reg_num << 3); } else if ((i.tm.base_opcode & ~(D|W)) == MOV_AX_DISP32) { @@ -2047,117 +2247,92 @@ md_assemble (line) return; } } - else if ((i.tm.opcode_modifier & Ugh) != 0) + else if (!quiet_warnings && (i.tm.opcode_modifier & Ugh) != 0) { - /* UnixWare fsub no args is alias for fsubp, fadd -> faddp, etc */ - as_warn (_("translating to `%sp'"), i.tm.name); + /* UnixWare fsub no args is alias for fsubp, fadd -> faddp, etc. */ + as_warn (_("translating to `%sp'"), i.tm.name); } } - /* Handle conversion of 'int $3' --> special int3 insn. */ - if (i.tm.base_opcode == INT_OPCODE && i.imms[0]->X_add_number == 3) + /* Handle conversion of 'int $3' --> special int3 insn. */ + if (i.tm.base_opcode == INT_OPCODE && i.op[0].imms->X_add_number == 3) { i.tm.base_opcode = INT3_OPCODE; i.imm_operands = 0; } - /* We are ready to output the insn. */ + if ((i.tm.opcode_modifier & (Jump | JumpByte | JumpDword)) + && i.op[0].disps->X_op == O_constant) + { + /* Convert "jmp constant" (and "call constant") to a jump (call) to + the absolute address given by the constant. Since ix86 jumps and + calls are pc relative, we need to generate a reloc. */ + i.op[0].disps->X_add_symbol = &abs_symbol; + i.op[0].disps->X_op = O_symbol; + } + + /* We are ready to output the insn. */ { register char *p; - /* Output jumps. */ + /* Output jumps. */ if (i.tm.opcode_modifier & Jump) { - long n = (long) i.disps[0]->X_add_number; - int prefix = (i.prefix[DATA_PREFIX] != 0); - int code16 = 0; + int size; + int code16; + int prefix; - if (prefix) + code16 = 0; + if (flag_16bit_code) + code16 = CODE16; + + prefix = 0; + if (i.prefix[DATA_PREFIX]) { + prefix = 1; i.prefixes -= 1; - code16 = CODE16; + code16 ^= CODE16; } - if (flag_16bit_code) - code16 ^= CODE16; - if (!intel_syntax && (i.prefixes != 0)) - as_warn (_("skipping prefixes on this instruction")); + size = 4; + if (code16) + size = 2; - if (i.disps[0]->X_op == O_constant) - { - if (fits_in_signed_byte (n)) - { - insn_size += 2; - p = frag_more (2); - p[0] = i.tm.base_opcode; - p[1] = n; - } - else - { - /* Use 16-bit jumps only for 16-bit code, - because text segments are limited to 64K anyway; - Use 32-bit jumps for 32-bit code, because they're faster, - and a 16-bit jump will clear the top 16 bits of %eip. */ - int jmp_size = code16 ? 2 : 4; - if (code16 && !fits_in_signed_word (n)) - { - as_bad (_("16-bit jump out of range")); - return; - } + if (i.prefixes != 0 && !intel_syntax) + as_warn (_("skipping prefixes on this instruction")); - if (i.tm.base_opcode == JUMP_PC_RELATIVE) - { /* pace */ - /* unconditional jump */ - insn_size += prefix + 1 + jmp_size; - p = frag_more (prefix + 1 + jmp_size); - if (prefix) - *p++ = DATA_PREFIX_OPCODE; - *p++ = (char) 0xe9; - md_number_to_chars (p, (valueT) n, jmp_size); - } - else - { - /* conditional jump */ - insn_size += prefix + 2 + jmp_size; - p = frag_more (prefix + 2 + jmp_size); - if (prefix) - *p++ = DATA_PREFIX_OPCODE; - *p++ = TWO_BYTE_OPCODE_ESCAPE; - *p++ = i.tm.base_opcode + 0x10; - md_number_to_chars (p, (valueT) n, jmp_size); - } - } - } - else - { - int size = code16 ? 2 : 4; - - /* It's a symbol; end frag & setup for relax. - Make sure there are more than 6 chars left in the current frag; - if not we'll have to start a new one. */ - frag_grow (prefix + 1 + 2 + size); - insn_size += 1 + prefix; - p = frag_more (1 + prefix); - if (prefix) - *p++ = DATA_PREFIX_OPCODE; - *p = i.tm.base_opcode; - frag_var (rs_machine_dependent, - prefix + 2 + size, /* 2 opcode/prefix + displacement */ - 1, - ((unsigned char) *p == JUMP_PC_RELATIVE - ? ENCODE_RELAX_STATE (UNCOND_JUMP, SMALL) | code16 - : ENCODE_RELAX_STATE (COND_JUMP, SMALL) | code16), - i.disps[0]->X_add_symbol, - (offsetT) n, p); - } + /* It's always a symbol; End frag & setup for relax. + Make sure there is enough room in this frag for the largest + instruction we may generate in md_convert_frag. This is 2 + bytes for the opcode and room for the prefix and largest + displacement. */ + frag_grow (prefix + 2 + size); + insn_size += prefix + 1; + /* Prefix and 1 opcode byte go in fr_fix. */ + p = frag_more (prefix + 1); + if (prefix) + *p++ = DATA_PREFIX_OPCODE; + *p = i.tm.base_opcode; + /* 1 possible extra opcode + displacement go in var part. + Pass reloc in fr_var. */ + frag_var (rs_machine_dependent, + 1 + size, + i.disp_reloc[0], + ((unsigned char) *p == JUMP_PC_RELATIVE + ? ENCODE_RELAX_STATE (UNCOND_JUMP, SMALL) | code16 + : ENCODE_RELAX_STATE (COND_JUMP, SMALL) | code16), + i.op[0].disps->X_add_symbol, + i.op[0].disps->X_add_number, + p); } else if (i.tm.opcode_modifier & (JumpByte | JumpDword)) { - int size = (i.tm.opcode_modifier & JumpByte) ? 1 : 4; - long n = (long) i.disps[0]->X_add_number; + int size; - if (size == 1) /* then this is a loop or jecxz type instruction */ + if (i.tm.opcode_modifier & JumpByte) { + /* This is a loop or jecxz type instruction. */ + size = 1; if (i.prefix[ADDR_PREFIX]) { insn_size += 1; @@ -2167,23 +2342,26 @@ md_assemble (line) } else { - int code16 = 0; + int code16; + + code16 = 0; + if (flag_16bit_code) + code16 = CODE16; if (i.prefix[DATA_PREFIX]) { insn_size += 1; FRAG_APPEND_1_CHAR (DATA_PREFIX_OPCODE); i.prefixes -= 1; - code16 = CODE16; + code16 ^= CODE16; } - if (flag_16bit_code) - code16 ^= CODE16; + size = 4; if (code16) size = 2; } - if (!intel_syntax && (i.prefixes != 0)) + if (i.prefixes != 0 && !intel_syntax) as_warn (_("skipping prefixes on this instruction")); if (fits_in_unsigned_byte (i.tm.base_opcode)) @@ -2193,89 +2371,74 @@ md_assemble (line) } else { - insn_size += 2 + size; /* opcode can be at most two bytes */ + /* Opcode can be at most two bytes. */ + insn_size += 2 + size; p = frag_more (2 + size); *p++ = (i.tm.base_opcode >> 8) & 0xff; } *p++ = i.tm.base_opcode & 0xff; - if (i.disps[0]->X_op == O_constant) - { - if (size == 1 && !fits_in_signed_byte (n)) - { - as_bad (_("`%s' only takes byte displacement; %ld shortened to %d"), - i.tm.name, n, *p); - } - else if (size == 2 && !fits_in_signed_word (n)) - { - as_bad (_("16-bit jump out of range")); - return; - } - md_number_to_chars (p, (valueT) n, size); - } - else - { - fix_new_exp (frag_now, p - frag_now->fr_literal, size, - i.disps[0], 1, reloc (size, 1, i.disp_reloc[0])); - - } + fix_new_exp (frag_now, p - frag_now->fr_literal, size, + i.op[0].disps, 1, reloc (size, 1, i.disp_reloc[0])); } else if (i.tm.opcode_modifier & JumpInterSegment) { int size; - int reloc_type; - int prefix = i.prefix[DATA_PREFIX] != 0; - int code16 = 0; + int prefix; + int code16; - if (prefix) + code16 = 0; + if (flag_16bit_code) + code16 = CODE16; + + prefix = 0; + if (i.prefix[DATA_PREFIX]) { - code16 = CODE16; + prefix = 1; i.prefixes -= 1; + code16 ^= CODE16; } - if (flag_16bit_code) - code16 ^= CODE16; size = 4; - reloc_type = BFD_RELOC_32; if (code16) - { - size = 2; - reloc_type = BFD_RELOC_16; - } + size = 2; - if (!intel_syntax && (i.prefixes != 0)) + if (i.prefixes != 0 && !intel_syntax) as_warn (_("skipping prefixes on this instruction")); - insn_size += prefix + 1 + 2 + size; /* 1 opcode; 2 segment; offset */ + /* 1 opcode; 2 segment; offset */ + insn_size += prefix + 1 + 2 + size; p = frag_more (prefix + 1 + 2 + size); if (prefix) *p++ = DATA_PREFIX_OPCODE; *p++ = i.tm.base_opcode; - if (i.imms[1]->X_op == O_constant) + if (i.op[1].imms->X_op == O_constant) { - long n = (long) i.imms[1]->X_add_number; + offsetT n = i.op[1].imms->X_add_number; - if (size == 2 && !fits_in_unsigned_word (n)) + if (size == 2 + && !fits_in_unsigned_word (n) + && !fits_in_signed_word (n)) { as_bad (_("16-bit jump out of range")); return; } - md_number_to_chars (p, (valueT) n, size); + md_number_to_chars (p, n, size); } else fix_new_exp (frag_now, p - frag_now->fr_literal, size, - i.imms[1], 0, reloc_type); - if (i.imms[0]->X_op != O_constant) + i.op[1].imms, 0, reloc (size, 0, i.disp_reloc[0])); + if (i.op[0].imms->X_op != O_constant) as_bad (_("can't handle non absolute segment in `%s'"), i.tm.name); - md_number_to_chars (p + size, (valueT) i.imms[0]->X_add_number, 2); + md_number_to_chars (p + size, (valueT) i.op[0].imms->X_add_number, 2); } else { - /* Output normal instructions here. */ + /* Output normal instructions here. */ unsigned char *q; - /* The prefix bytes. */ + /* The prefix bytes. */ for (q = i.prefix; q < i.prefix + sizeof (i.prefix) / sizeof (i.prefix[0]); q++) @@ -2288,7 +2451,7 @@ md_assemble (line) } } - /* Now the opcode; be careful about word order here! */ + /* Now the opcode; be careful about word order here! */ if (fits_in_unsigned_byte (i.tm.base_opcode)) { insn_size += 1; @@ -2298,12 +2461,12 @@ md_assemble (line) { insn_size += 2; p = frag_more (2); - /* put out high byte first: can't use md_number_to_chars! */ + /* Put out high byte first: can't use md_number_to_chars! */ *p++ = (i.tm.base_opcode >> 8) & 0xff; *p = i.tm.base_opcode & 0xff; } else - { /* opcode is either 3 or 4 bytes */ + { /* Opcode is either 3 or 4 bytes. */ if (i.tm.base_opcode & 0xff000000) { insn_size += 4; @@ -2354,163 +2517,124 @@ md_assemble (line) for (n = 0; n < i.operands; n++) { - if (i.disps[n]) + if (i.types[n] & Disp) { - if (i.disps[n]->X_op == O_constant) + if (i.op[n].disps->X_op == O_constant) { - if (i.types[n] & Disp8) - { - insn_size += 1; - p = frag_more (1); - md_number_to_chars (p, - (valueT) i.disps[n]->X_add_number, - 1); - } - else if (i.types[n] & Disp16) + int size; + offsetT val; + + size = 4; + if (i.types[n] & (Disp8 | Disp16)) { - insn_size += 2; - p = frag_more (2); - md_number_to_chars (p, - (valueT) i.disps[n]->X_add_number, - 2); - } - else - { /* Disp32 */ - insn_size += 4; - p = frag_more (4); - md_number_to_chars (p, - (valueT) i.disps[n]->X_add_number, - 4); + size = 2; + if (i.types[n] & Disp8) + size = 1; } - } - else if (i.types[n] & Disp32) - { - insn_size += 4; - p = frag_more (4); - fix_new_exp (frag_now, p - frag_now->fr_literal, 4, - i.disps[n], 0, - TC_RELOC (i.disp_reloc[n], BFD_RELOC_32)); + val = offset_in_range (i.op[n].disps->X_add_number, + size); + insn_size += size; + p = frag_more (size); + md_number_to_chars (p, val, size); } else - { /* must be Disp16 */ - insn_size += 2; - p = frag_more (2); - fix_new_exp (frag_now, p - frag_now->fr_literal, 2, - i.disps[n], 0, - TC_RELOC (i.disp_reloc[n], BFD_RELOC_16)); + { + int size = 4; + + if (i.types[n] & Disp16) + size = 2; + + insn_size += size; + p = frag_more (size); + fix_new_exp (frag_now, p - frag_now->fr_literal, size, + i.op[n].disps, 0, + reloc (size, 0, i.disp_reloc[n])); } } } - } /* end displacement output */ + } - /* output immediate */ + /* Output immediate. */ if (i.imm_operands) { register unsigned int n; for (n = 0; n < i.operands; n++) { - if (i.imms[n]) + if (i.types[n] & Imm) { - if (i.imms[n]->X_op == O_constant) + if (i.op[n].imms->X_op == O_constant) { - if (i.types[n] & (Imm8 | Imm8S)) - { - insn_size += 1; - p = frag_more (1); - md_number_to_chars (p, - (valueT) i.imms[n]->X_add_number, - 1); - } - else if (i.types[n] & Imm16) - { - insn_size += 2; - p = frag_more (2); - md_number_to_chars (p, - (valueT) i.imms[n]->X_add_number, - 2); - } - else + int size; + offsetT val; + + size = 4; + if (i.types[n] & (Imm8 | Imm8S | Imm16)) { - insn_size += 4; - p = frag_more (4); - md_number_to_chars (p, - (valueT) i.imms[n]->X_add_number, - 4); + size = 2; + if (i.types[n] & (Imm8 | Imm8S)) + size = 1; } + val = offset_in_range (i.op[n].imms->X_add_number, + size); + insn_size += size; + p = frag_more (size); + md_number_to_chars (p, val, size); } else - { /* not absolute_section */ - /* Need a 32-bit fixup (don't support 8bit - non-absolute ims). Try to support other - sizes ... */ - int r_type; - int size; - int pcrel = 0; + { + /* Not absolute_section. + Need a 32-bit fixup (don't support 8bit + non-absolute imms). Try to support other + sizes ... */ +#ifdef BFD_ASSEMBLER + enum bfd_reloc_code_real reloc_type; +#else + int reloc_type; +#endif + int size = 4; - if (i.types[n] & (Imm8 | Imm8S)) - size = 1; - else if (i.types[n] & Imm16) + if (i.types[n] & Imm16) size = 2; - else - size = 4; + else if (i.types[n] & (Imm8 | Imm8S)) + size = 1; + insn_size += size; p = frag_more (size); - r_type = reloc (size, 0, i.disp_reloc[0]); + reloc_type = reloc (size, 0, i.disp_reloc[0]); #ifdef BFD_ASSEMBLER - if (r_type == BFD_RELOC_32 + if (reloc_type == BFD_RELOC_32 && GOT_symbol - && GOT_symbol == i.imms[n]->X_add_symbol - && (i.imms[n]->X_op == O_symbol - || (i.imms[n]->X_op == O_add - && (i.imms[n]->X_op_symbol->sy_value.X_op + && GOT_symbol == i.op[n].imms->X_add_symbol + && (i.op[n].imms->X_op == O_symbol + || (i.op[n].imms->X_op == O_add + && ((symbol_get_value_expression + (i.op[n].imms->X_op_symbol)->X_op) == O_subtract)))) { - r_type = BFD_RELOC_386_GOTPC; - i.imms[n]->X_add_number += 3; + reloc_type = BFD_RELOC_386_GOTPC; + i.op[n].imms->X_add_number += 3; } #endif fix_new_exp (frag_now, p - frag_now->fr_literal, size, - i.imms[n], pcrel, r_type); + i.op[n].imms, 0, reloc_type); } } } - } /* end immediate output */ + } } + dwarf2_emit_insn (insn_size); + #ifdef DEBUG386 if (flag_debug) { pi (line, &i); } -#endif /* DEBUG386 */ +#endif /* DEBUG386 */ } } -static int i386_is_reg PARAMS ((char *)); - -static int -i386_is_reg (reg_string) - char *reg_string; -{ - register char *s = reg_string; - register char *p; - char reg_name_given[MAX_REG_NAME_SIZE + 1]; - - if (is_space_char (*s)) - ++s; - - p = reg_name_given; - while ((*p++ = register_chars[(unsigned char) *s++]) != '\0') - if (p >= reg_name_given + MAX_REG_NAME_SIZE) - return 0; - - if (!hash_find (reg_hash, reg_name_given)) - return 0; - else - return 1; -} - static int i386_immediate PARAMS ((char *)); static int @@ -2519,16 +2643,16 @@ i386_immediate (imm_start) { char *save_input_line_pointer; segT exp_seg = 0; - expressionS * exp; + expressionS *exp; if (i.imm_operands == MAX_IMMEDIATE_OPERANDS) { - as_bad (_("Only 1 or 2 immediate operands are allowed")); + as_bad (_("only 1 or 2 immediate operands are allowed")); return 0; } exp = &im_expressions[i.imm_operands++]; - i.imms[this_operand] = exp; + i.op[this_operand].imms = exp; if (is_space_char (*imm_start)) ++imm_start; @@ -2537,115 +2661,110 @@ i386_immediate (imm_start) input_line_pointer = imm_start; #ifndef LEX_AT - { - /* - * We can have operands of the form - * @GOTOFF+ - * Take the easy way out here and copy everything - * into a temporary buffer... - */ - register char *cp; - - cp = strchr (input_line_pointer, '@'); - if (cp != NULL) - { - char *tmpbuf; - int len, first; - - /* GOT relocations are not supported in 16 bit mode */ - if (flag_16bit_code) - as_bad (_("GOT relocations not supported in 16 bit mode")); - - if (GOT_symbol == NULL) - GOT_symbol = symbol_find_or_make (GLOBAL_OFFSET_TABLE_NAME); - - if (strncmp (cp + 1, "PLT", 3) == 0) - { - i.disp_reloc[this_operand] = BFD_RELOC_386_PLT32; - len = 3; - } - else if (strncmp (cp + 1, "GOTOFF", 6) == 0) - { - i.disp_reloc[this_operand] = BFD_RELOC_386_GOTOFF; - len = 6; - } - else if (strncmp (cp + 1, "GOT", 3) == 0) - { - i.disp_reloc[this_operand] = BFD_RELOC_386_GOT32; - len = 3; - } - else - as_bad (_("Bad reloc specifier in expression")); - - /* Replace the relocation token with ' ', so that errors like - foo@GOTOFF1 will be detected. */ - first = cp - input_line_pointer; - tmpbuf = (char *) alloca (strlen(input_line_pointer)); - memcpy (tmpbuf, input_line_pointer, first); - tmpbuf[first] = ' '; - strcpy (tmpbuf + first + 1, cp + 1 + len); - input_line_pointer = tmpbuf; - } - } + { + /* We can have operands of the form + @GOTOFF+ + Take the easy way out here and copy everything + into a temporary buffer... */ + register char *cp; + + cp = strchr (input_line_pointer, '@'); + if (cp != NULL) + { + char *tmpbuf; + int len = 0; + int first; + + /* GOT relocations are not supported in 16 bit mode. */ + if (flag_16bit_code) + as_bad (_("GOT relocations not supported in 16 bit mode")); + + if (GOT_symbol == NULL) + GOT_symbol = symbol_find_or_make (GLOBAL_OFFSET_TABLE_NAME); + + if (strncmp (cp + 1, "PLT", 3) == 0) + { + i.disp_reloc[this_operand] = BFD_RELOC_386_PLT32; + len = 3; + } + else if (strncmp (cp + 1, "GOTOFF", 6) == 0) + { + i.disp_reloc[this_operand] = BFD_RELOC_386_GOTOFF; + len = 6; + } + else if (strncmp (cp + 1, "GOT", 3) == 0) + { + i.disp_reloc[this_operand] = BFD_RELOC_386_GOT32; + len = 3; + } + else + as_bad (_("bad reloc specifier in expression")); + + /* Replace the relocation token with ' ', so that errors like + foo@GOTOFF1 will be detected. */ + first = cp - input_line_pointer; + tmpbuf = (char *) alloca (strlen (input_line_pointer)); + memcpy (tmpbuf, input_line_pointer, first); + tmpbuf[first] = ' '; + strcpy (tmpbuf + first + 1, cp + 1 + len); + input_line_pointer = tmpbuf; + } + } #endif exp_seg = expression (exp); SKIP_WHITESPACE (); if (*input_line_pointer) - as_bad (_("Ignoring junk `%s' after expression"), input_line_pointer); + as_bad (_("ignoring junk `%s' after expression"), input_line_pointer); input_line_pointer = save_input_line_pointer; - if (exp->X_op == O_absent) + if (exp->X_op == O_absent || exp->X_op == O_big) { - /* missing or bad expr becomes absolute 0 */ - as_bad (_("Missing or invalid immediate expression `%s' taken as 0"), - imm_start); + /* Missing or bad expr becomes absolute 0. */ + as_bad (_("missing or invalid immediate expression `%s' taken as 0"), + imm_start); exp->X_op = O_constant; exp->X_add_number = 0; exp->X_add_symbol = (symbolS *) 0; exp->X_op_symbol = (symbolS *) 0; - i.types[this_operand] |= Imm; } - else if (exp->X_op == O_constant) + + if (exp->X_op == O_constant) { - i.types[this_operand] |= - smallest_imm_type ((long) exp->X_add_number); - - /* If a suffix is given, this operand may be shortended. */ - switch (i.suffix) - { - case WORD_MNEM_SUFFIX: - i.types[this_operand] |= Imm16; - break; - case BYTE_MNEM_SUFFIX: - i.types[this_operand] |= Imm16 | Imm8 | Imm8S; - break; - } + /* Size it properly later. */ + i.types[this_operand] |= Imm32; } -#ifdef OBJ_AOUT - else if (exp_seg != text_section - && exp_seg != data_section - && exp_seg != bss_section - && exp_seg != undefined_section +#if (defined (OBJ_AOUT) || defined (OBJ_MAYBE_AOUT)) + else if (1 +#ifdef BFD_ASSEMBLER + && OUTPUT_FLAVOR == bfd_target_aout_flavour +#endif + && exp_seg != text_section + && exp_seg != data_section + && exp_seg != bss_section + && exp_seg != undefined_section #ifdef BFD_ASSEMBLER - && !bfd_is_com_section (exp_seg) + && !bfd_is_com_section (exp_seg) #endif - ) + ) { - seg_unimplemented: - as_bad (_("Unimplemented segment type %d in operand"), exp_seg); +#ifdef BFD_ASSEMBLER + as_bad (_("unimplemented segment %s in operand"), exp_seg->name); +#else + as_bad (_("unimplemented segment type %d in operand"), exp_seg); +#endif return 0; } #endif else { /* This is an address. The size of the address will be - determined later, depending on destination register, - suffix, or the default for the section. We exclude - Imm8S here so that `push $foo' and other instructions - with an Imm8S form will use Imm16 or Imm32. */ + determined later, depending on destination register, + suffix, or the default for the section. We exclude + Imm8S here so that `push $foo' and other instructions + with an Imm8S form will use Imm16 or Imm32. */ i.types[this_operand] |= (Imm8 | Imm16 | Imm32); } @@ -2679,13 +2798,13 @@ i386_scale (scale) default: bad_scale: as_bad (_("expecting scale factor of 1, 2, 4, or 8: got `%s'"), - scale); + scale); return 0; } if (i.log2_scale_factor != 0 && ! i.index_reg) { as_warn (_("scale factor of %d without an index register"), - 1 << i.log2_scale_factor); + 1 << i.log2_scale_factor); #if SCALE1_WHEN_NO_INDEX i.log2_scale_factor = 0; #endif @@ -2705,17 +2824,12 @@ i386_displacement (disp_start, disp_end) char *save_input_line_pointer; int bigdisp = Disp32; - /* All of the pieces of the displacement expression are handled together. */ - if (intel_syntax && i.disp_operands != 0) - return 1; - if (flag_16bit_code ^ (i.prefix[ADDR_PREFIX] != 0)) bigdisp = Disp16; i.types[this_operand] |= bigdisp; exp = &disp_expressions[i.disp_operands]; - i.disps[this_operand] = exp; - i.disp_reloc[this_operand] = NO_RELOC; + i.op[this_operand].disps = exp; i.disp_operands++; save_input_line_pointer = input_line_pointer; input_line_pointer = disp_start; @@ -2727,670 +2841,255 @@ i386_displacement (disp_start, disp_end) #if GCC_ASM_O_HACK END_STRING_AND_SAVE (disp_end + 1); if ((i.types[this_operand] & BaseIndex) != 0 - && displacement_string_end[-1] == '+') + && displacement_string_end[-1] == '+') { /* This hack is to avoid a warning when using the "o" - constraint within gcc asm statements. - For instance: - - #define _set_tssldt_desc(n,addr,limit,type) \ - __asm__ __volatile__ ( \ - "movw %w2,%0\n\t" \ - "movw %w1,2+%0\n\t" \ - "rorl $16,%1\n\t" \ - "movb %b1,4+%0\n\t" \ - "movb %4,5+%0\n\t" \ - "movb $0,6+%0\n\t" \ - "movb %h1,7+%0\n\t" \ - "rorl $16,%1" \ - : "=o"(*(n)) : "q" (addr), "ri"(limit), "i"(type)) - - This works great except that the output assembler ends - up looking a bit weird if it turns out that there is - no offset. You end up producing code that looks like: - - #APP - movw $235,(%eax) - movw %dx,2+(%eax) - rorl $16,%edx - movb %dl,4+(%eax) - movb $137,5+(%eax) - movb $0,6+(%eax) - movb %dh,7+(%eax) - rorl $16,%edx - #NO_APP - - So here we provide the missing zero. - */ - - *displacement_string_end = '0'; + constraint within gcc asm statements. + For instance: + + #define _set_tssldt_desc(n,addr,limit,type) \ + __asm__ __volatile__ ( \ + "movw %w2,%0\n\t" \ + "movw %w1,2+%0\n\t" \ + "rorl $16,%1\n\t" \ + "movb %b1,4+%0\n\t" \ + "movb %4,5+%0\n\t" \ + "movb $0,6+%0\n\t" \ + "movb %h1,7+%0\n\t" \ + "rorl $16,%1" \ + : "=o"(*(n)) : "q" (addr), "ri"(limit), "i"(type)) + + This works great except that the output assembler ends + up looking a bit weird if it turns out that there is + no offset. You end up producing code that looks like: + + #APP + movw $235,(%eax) + movw %dx,2+(%eax) + rorl $16,%edx + movb %dl,4+(%eax) + movb $137,5+(%eax) + movb $0,6+(%eax) + movb %dh,7+(%eax) + rorl $16,%edx + #NO_APP + + So here we provide the missing zero. */ + + *displacement_string_end = '0'; } #endif #ifndef LEX_AT - { - /* - * We can have operands of the form - * @GOTOFF+ - * Take the easy way out here and copy everything - * into a temporary buffer... - */ - register char *cp; - - cp = strchr (input_line_pointer, '@'); - if (cp != NULL) - { - char *tmpbuf; - int len, first; - - /* GOT relocations are not supported in 16 bit mode */ - if (flag_16bit_code) - as_bad (_("GOT relocations not supported in 16 bit mode")); - - if (GOT_symbol == NULL) - GOT_symbol = symbol_find_or_make (GLOBAL_OFFSET_TABLE_NAME); - - if (strncmp (cp + 1, "PLT", 3) == 0) - { - i.disp_reloc[this_operand] = BFD_RELOC_386_PLT32; - len = 3; - } - else if (strncmp (cp + 1, "GOTOFF", 6) == 0) - { - i.disp_reloc[this_operand] = BFD_RELOC_386_GOTOFF; - len = 6; - } - else if (strncmp (cp + 1, "GOT", 3) == 0) - { - i.disp_reloc[this_operand] = BFD_RELOC_386_GOT32; - len = 3; - } - else - as_bad (_("Bad reloc specifier in expression")); - - /* Replace the relocation token with ' ', so that errors like - foo@GOTOFF1 will be detected. */ - first = cp - input_line_pointer; - tmpbuf = (char *) alloca (strlen(input_line_pointer)); - memcpy (tmpbuf, input_line_pointer, first); - tmpbuf[first] = ' '; - strcpy (tmpbuf + first + 1, cp + 1 + len); - input_line_pointer = tmpbuf; - } - } -#endif - - exp_seg = expression (exp); - -#ifdef BFD_ASSEMBLER - /* We do this to make sure that the section symbol is in - the symbol table. We will ultimately change the relocation - to be relative to the beginning of the section */ - if (i.disp_reloc[this_operand] == BFD_RELOC_386_GOTOFF) - { - if (S_IS_LOCAL(exp->X_add_symbol) - && S_GET_SEGMENT (exp->X_add_symbol) != undefined_section) - section_symbol(exp->X_add_symbol->bsym->section); - assert (exp->X_op == O_symbol); - exp->X_op = O_subtract; - exp->X_op_symbol = GOT_symbol; - i.disp_reloc[this_operand] = BFD_RELOC_32; - } -#endif - - SKIP_WHITESPACE (); - if (*input_line_pointer) - as_bad (_("Ignoring junk `%s' after expression"), - input_line_pointer); -#if GCC_ASM_O_HACK - RESTORE_END_STRING (disp_end + 1); -#endif - RESTORE_END_STRING (disp_end); - input_line_pointer = save_input_line_pointer; - - if (exp->X_op == O_constant) - { - if (fits_in_signed_byte (exp->X_add_number)) - i.types[this_operand] |= Disp8; - } -#ifdef OBJ_AOUT - else if (exp_seg != text_section - && exp_seg != data_section - && exp_seg != bss_section - && exp_seg != undefined_section) - { - as_bad (_ ("Unimplemented segment type %d in operand"), exp_seg); - return 0; - } -#endif - return 1; -} - -static int i386_operand_modifier PARAMS ((char **, int)); - -static int -i386_operand_modifier (op_string, got_a_float) - char **op_string; - int got_a_float; -{ - if (!strncasecmp (*op_string, "BYTE PTR", 8)) - { - i.suffix = BYTE_MNEM_SUFFIX; - *op_string += 8; - return BYTE_PTR; - - } - else if (!strncasecmp (*op_string, "WORD PTR", 8)) - { - i.suffix = WORD_MNEM_SUFFIX; - *op_string += 8; - return WORD_PTR; - } - - else if (!strncasecmp (*op_string, "DWORD PTR", 9)) - { - if (got_a_float) - i.suffix = SHORT_MNEM_SUFFIX; - else - i.suffix = DWORD_MNEM_SUFFIX; - *op_string += 9; - return DWORD_PTR; - } - - else if (!strncasecmp (*op_string, "QWORD PTR", 9)) - { - i.suffix = INTEL_DWORD_MNEM_SUFFIX; - *op_string += 9; - return QWORD_PTR; - } - - else if (!strncasecmp (*op_string, "XWORD PTR", 9)) + { + /* We can have operands of the form + @GOTOFF+ + Take the easy way out here and copy everything + into a temporary buffer... */ + register char *cp; + + cp = strchr (input_line_pointer, '@'); + if (cp != NULL) { - i.suffix = LONG_DOUBLE_MNEM_SUFFIX; - *op_string += 9; - return XWORD_PTR; - } + char *tmpbuf; + int len = 0; + int first; - else if (!strncasecmp (*op_string, "SHORT", 5)) - { - *op_string += 5; - return SHORT; - } + /* GOT relocations are not supported in 16 bit mode. */ + if (flag_16bit_code) + as_bad (_("GOT relocations not supported in 16 bit mode")); - else if (!strncasecmp (*op_string, "OFFSET FLAT:", 12)) - { - *op_string += 12; - return OFFSET_FLAT; - } + if (GOT_symbol == NULL) + GOT_symbol = symbol_find_or_make (GLOBAL_OFFSET_TABLE_NAME); - else if (!strncasecmp (*op_string, "FLAT", 4)) - { - *op_string += 4; - return FLAT; + if (strncmp (cp + 1, "PLT", 3) == 0) + { + i.disp_reloc[this_operand] = BFD_RELOC_386_PLT32; + len = 3; + } + else if (strncmp (cp + 1, "GOTOFF", 6) == 0) + { + i.disp_reloc[this_operand] = BFD_RELOC_386_GOTOFF; + len = 6; + } + else if (strncmp (cp + 1, "GOT", 3) == 0) + { + i.disp_reloc[this_operand] = BFD_RELOC_386_GOT32; + len = 3; + } + else + as_bad (_("bad reloc specifier in expression")); + + /* Replace the relocation token with ' ', so that errors like + foo@GOTOFF1 will be detected. */ + first = cp - input_line_pointer; + tmpbuf = (char *) alloca (strlen (input_line_pointer)); + memcpy (tmpbuf, input_line_pointer, first); + tmpbuf[first] = ' '; + strcpy (tmpbuf + first + 1, cp + 1 + len); + input_line_pointer = tmpbuf; } + } +#endif - else return NONE_FOUND; -} - -static char * build_displacement_string PARAMS ((int, char *)); + exp_seg = expression (exp); -static char * -build_displacement_string (initial_disp, op_string) - int initial_disp; - char *op_string; -{ - char *temp_string = (char *) malloc (strlen (op_string) + 1); - char *end_of_operand_string; - char *tc; - char *temp_disp; - - temp_string[0] = '\0'; - tc = end_of_operand_string = strchr (op_string, '['); - if ( initial_disp && !end_of_operand_string) +#ifdef BFD_ASSEMBLER + /* We do this to make sure that the section symbol is in + the symbol table. We will ultimately change the relocation + to be relative to the beginning of the section. */ + if (i.disp_reloc[this_operand] == BFD_RELOC_386_GOTOFF) { - strcpy (temp_string, op_string); - return (temp_string); + if (S_IS_LOCAL(exp->X_add_symbol) + && S_GET_SEGMENT (exp->X_add_symbol) != undefined_section) + section_symbol (S_GET_SEGMENT (exp->X_add_symbol)); + assert (exp->X_op == O_symbol); + exp->X_op = O_subtract; + exp->X_op_symbol = GOT_symbol; + i.disp_reloc[this_operand] = BFD_RELOC_32; } +#endif - /* Build the whole displacement string */ - if (initial_disp) - { - strncpy (temp_string, op_string, end_of_operand_string - op_string); - temp_string[end_of_operand_string - op_string] = '\0'; - temp_disp = tc; - } - else - temp_disp = op_string; + SKIP_WHITESPACE (); + if (*input_line_pointer) + as_bad (_("ignoring junk `%s' after expression"), + input_line_pointer); +#if GCC_ASM_O_HACK + RESTORE_END_STRING (disp_end + 1); +#endif + RESTORE_END_STRING (disp_end); + input_line_pointer = save_input_line_pointer; - while (*temp_disp != '\0') + if (exp->X_op == O_absent || exp->X_op == O_big) { - int add_minus = (*temp_disp == '-'); - - if (*temp_disp == '+' || *temp_disp == '-' || *temp_disp == '[') - temp_disp++; - - if (is_space_char (*temp_disp)) - temp_disp++; - - /* Don't consider registers */ - if (*temp_disp != REGISTER_PREFIX - && !(allow_naked_reg && i386_is_reg (temp_disp))) - { - char *string_start = temp_disp; - - while (*temp_disp != ']' - && *temp_disp != '+' - && *temp_disp != '-' - && *temp_disp != '*') - ++temp_disp; - - if (add_minus) - strcat (temp_string, "-"); - else - strcat (temp_string, "+"); - - strncat (temp_string, string_start, temp_disp - string_start); - if (*temp_disp == '+' || *temp_disp == '-') - --temp_disp; - } - - while (*temp_disp != '\0' - && *temp_disp != '+' - && *temp_disp != '-') - ++temp_disp; + /* Missing or bad expr becomes absolute 0. */ + as_bad (_("missing or invalid displacement expression `%s' taken as 0"), + disp_start); + exp->X_op = O_constant; + exp->X_add_number = 0; + exp->X_add_symbol = (symbolS *) 0; + exp->X_op_symbol = (symbolS *) 0; } - return temp_string; -} - -static int i386_parse_seg PARAMS ((char *)); - -static int -i386_parse_seg (op_string) - char *op_string; -{ - if (is_space_char (*op_string)) - ++op_string; - - /* Should be one of es, cs, ss, ds fs or gs */ - switch (*op_string++) +#if (defined (OBJ_AOUT) || defined (OBJ_MAYBE_AOUT)) + if (exp->X_op != O_constant +#ifdef BFD_ASSEMBLER + && OUTPUT_FLAVOR == bfd_target_aout_flavour +#endif + && exp_seg != text_section + && exp_seg != data_section + && exp_seg != bss_section + && exp_seg != undefined_section) { - case 'e': - i.seg[i.mem_operands] = &es; - break; - case 'c': - i.seg[i.mem_operands] = &cs; - break; - case 's': - i.seg[i.mem_operands] = &ss; - break; - case 'd': - i.seg[i.mem_operands] = &ds; - break; - case 'f': - i.seg[i.mem_operands] = &fs; - break; - case 'g': - i.seg[i.mem_operands] = &gs; - break; - default: - as_bad (_("bad segment name `%s'"), op_string); +#ifdef BFD_ASSEMBLER + as_bad (_("unimplemented segment %s in operand"), exp_seg->name); +#else + as_bad (_("unimplemented segment type %d in operand"), exp_seg); +#endif return 0; } - - if (*op_string++ != 's') - { - as_bad (_("bad segment name `%s'"), op_string); - return 0; - } - - if (is_space_char (*op_string)) - ++op_string; - - if (*op_string != ':') - { - as_bad (_("bad segment name `%s'"), op_string); - return 0; - } - - return 1; - -} - -static int i386_intel_memory_operand PARAMS ((char *)); - -static int -i386_intel_memory_operand (op_string) - char *op_string; -{ - - char *end_of_operand_string; - - if (is_digit_char (*op_string) - && strchr (op_string, '[') == 0) - { - if (!i386_immediate (op_string)) - return 0; - else - return 1; - } - - /* Look for displacement preceding open bracket */ - if (*op_string != '[') - { - char *end_seg; - char *temp_string; - - end_seg = strchr (op_string, ':'); - if (end_seg) - { - if (!i386_parse_seg (op_string)) - return 0; - op_string = end_seg + 1; - } - - temp_string = build_displacement_string (true, op_string); - if (!i386_displacement (temp_string, temp_string + strlen (temp_string))) - return 0; - - end_of_operand_string = strchr (op_string, '['); - if (!end_of_operand_string) - end_of_operand_string = op_string + strlen (op_string); - - if (is_space_char (*end_of_operand_string)) - --end_of_operand_string; - - op_string = end_of_operand_string; - } - - if (*op_string == '[') - { - ++op_string; - - /* Pick off each component and figure out where it belongs */ - - end_of_operand_string = op_string; - - while (*op_string != ']') - { - - while (*end_of_operand_string != '+' - && *end_of_operand_string != '-' - && *end_of_operand_string != '*' - && *end_of_operand_string != ']') - end_of_operand_string++; - - if (*op_string == '+') - { - char *temp_string = op_string + 1; - if (is_space_char (*temp_string)) - ++temp_string; - if (*temp_string == REGISTER_PREFIX - || (allow_naked_reg && i386_is_reg (temp_string))) - ++op_string; - } - - if (*op_string == REGISTER_PREFIX - || (allow_naked_reg && i386_is_reg (op_string))) - { - const reg_entry *temp_reg; - char *end_op; - - END_STRING_AND_SAVE (end_of_operand_string); - temp_reg = parse_register (op_string, &end_op); - RESTORE_END_STRING (end_of_operand_string); - - if (temp_reg == NULL) - return 0; - - if (i.base_reg == NULL) - i.base_reg = temp_reg; - else - i.index_reg = temp_reg; - - i.types[this_operand] |= BaseIndex; - - } - else if (is_digit_char (*op_string) || *op_string == '+' || *op_string == '-') - { - - char *temp_string = build_displacement_string (false, op_string); - - if (*temp_string == '+') - ++temp_string; - - if (!i386_displacement (temp_string, temp_string + strlen (temp_string))) - return 0; - - ++op_string; - end_of_operand_string = op_string; - while (*end_of_operand_string != ']' - && *end_of_operand_string != '+' - && *end_of_operand_string != '-' - && *end_of_operand_string != '*') - ++end_of_operand_string; - } - else if (*op_string == '*') - { - ++op_string; - - if (i.base_reg && !i.index_reg) - { - i.index_reg = i.base_reg; - i.base_reg = 0; - } - - if (!i386_scale (op_string)) - return 0; - } - op_string = end_of_operand_string; - ++end_of_operand_string; - } - } - +#endif return 1; } -static int i386_intel_operand PARAMS ((char *, int)); +static int i386_index_check PARAMS((const char *)); + +/* Make sure the memory operand we've been dealt is valid. + Return 1 on success, 0 on a failure. */ static int -i386_intel_operand (operand_string, got_a_float) - char *operand_string; - int got_a_float; +i386_index_check (operand_string) + const char *operand_string; { - char *op_string = operand_string; - - int operand_modifier = i386_operand_modifier (&op_string, got_a_float); - if (is_space_char (*op_string)) - ++op_string; - - switch (operand_modifier) - { - case BYTE_PTR: - case WORD_PTR: - case DWORD_PTR: - case QWORD_PTR: - case XWORD_PTR: - if ((i.mem_operands == 1 - && (current_templates->start->opcode_modifier & IsString) == 0) - || i.mem_operands == 2) - { - as_bad (_("too many memory references for `%s'"), - current_templates->start->name); - return 0; - } - - if (!i386_intel_memory_operand (op_string)) - return 0; - - i.mem_operands++; - break; - - case FLAT: - - case OFFSET_FLAT: - if (!i386_immediate (op_string)) - return 0; - break; - - case SHORT: - - case NONE_FOUND: - /* Should be register or immediate */ - if (is_digit_char (*op_string) - && strchr (op_string, '[') == 0) - { - if (!i386_immediate (op_string)) - return 0; - } - else if (*op_string == REGISTER_PREFIX - || (allow_naked_reg - && i386_is_reg (op_string))) - { - - register const reg_entry * r; - char *end_op; - - r = parse_register (op_string, &end_op); - if (r == NULL) - return 0; - - /* Check for a segment override by searching for ':' after a - segment register. */ - op_string = end_op; - if (is_space_char (*op_string)) - ++op_string; - if (*op_string == ':' && (r->reg_type & (SReg2 | SReg3))) - { - switch (r->reg_num) - { - case 0: - i.seg[i.mem_operands] = &es; - break; - case 1: - i.seg[i.mem_operands] = &cs; - break; - case 2: - i.seg[i.mem_operands] = &ss; - break; - case 3: - i.seg[i.mem_operands] = &ds; - break; - case 4: - i.seg[i.mem_operands] = &fs; - break; - case 5: - i.seg[i.mem_operands] = &gs; - break; - } - - } - i.types[this_operand] |= r->reg_type & ~BaseIndex; - i.regs[this_operand] = r; - i.reg_operands++; - } - - else - { - - if (!i386_intel_memory_operand (op_string)) - return 0; +#if INFER_ADDR_PREFIX + int fudged = 0; - i.mem_operands++; - } - break; - - } /* end switch */ - /* Special case for (%dx) while doing input/output op. */ - if (i.base_reg - && i.base_reg->reg_type == (Reg16 | InOutPortReg) - && i.index_reg == 0 - && i.log2_scale_factor == 0 - && i.seg[i.mem_operands] == 0 - && (i.types[this_operand] & Disp) == 0) - { - i.types[this_operand] = InOutPortReg; - return 1; - } - /* Make sure the memory operand we've been dealt is valid. */ - if (flag_16bit_code ^ (i.prefix[ADDR_PREFIX] != 0)) + tryprefix: +#endif + if (flag_16bit_code ^ (i.prefix[ADDR_PREFIX] != 0) + /* 16 bit mode checks. */ + ? ((i.base_reg + && ((i.base_reg->reg_type & (Reg16|BaseIndex)) + != (Reg16|BaseIndex))) + || (i.index_reg + && (((i.index_reg->reg_type & (Reg16|BaseIndex)) + != (Reg16|BaseIndex)) + || ! (i.base_reg + && i.base_reg->reg_num < 6 + && i.index_reg->reg_num >= 6 + && i.log2_scale_factor == 0)))) + /* 32 bit mode checks. */ + : ((i.base_reg + && (i.base_reg->reg_type & Reg32) == 0) + || (i.index_reg + && ((i.index_reg->reg_type & (Reg32|BaseIndex)) + != (Reg32|BaseIndex))))) { - if ((i.base_reg - && ((i.base_reg->reg_type & (Reg16|BaseIndex)) - != (Reg16|BaseIndex))) - || (i.index_reg - && (((i.index_reg->reg_type & (Reg16|BaseIndex)) - != (Reg16|BaseIndex)) - || ! (i.base_reg - && i.base_reg->reg_num < 6 - && i.index_reg->reg_num >= 6 - && i.log2_scale_factor == 0)))) - { - as_bad (_("`%s' is not a valid %s bit base/index expression"), - operand_string, "16"); - return 0; - } - } - else +#if INFER_ADDR_PREFIX + if (i.prefix[ADDR_PREFIX] == 0 && stackop_size != '\0') { - if ((i.base_reg - && (i.base_reg->reg_type & Reg32) == 0) - || (i.index_reg - && ((i.index_reg->reg_type & (Reg32|BaseIndex)) - != (Reg32|BaseIndex)))) - { - as_bad (_("`%s' is not a valid %s bit base/index expression"), - operand_string, "32"); - return 0; - } + i.prefix[ADDR_PREFIX] = ADDR_PREFIX_OPCODE; + i.prefixes += 1; + /* Change the size of any displacement too. At most one of + Disp16 or Disp32 is set. + FIXME. There doesn't seem to be any real need for separate + Disp16 and Disp32 flags. The same goes for Imm16 and Imm32. + Removing them would probably clean up the code quite a lot. */ + if (i.types[this_operand] & (Disp16|Disp32)) + i.types[this_operand] ^= (Disp16|Disp32); + fudged = 1; + goto tryprefix; } + if (fudged) + as_bad (_("`%s' is not a valid base/index expression"), + operand_string); + else +#endif + as_bad (_("`%s' is not a valid %s bit base/index expression"), + operand_string, + flag_16bit_code ^ (i.prefix[ADDR_PREFIX] != 0) ? "16" : "32"); + return 0; + } return 1; } /* Parse OPERAND_STRING into the i386_insn structure I. Returns non-zero - on error. */ - -static int i386_operand PARAMS ((char *)); + on error. */ static int i386_operand (operand_string) char *operand_string; { + const reg_entry *r; + char *end_op; char *op_string = operand_string; if (is_space_char (*op_string)) ++op_string; /* We check for an absolute prefix (differentiating, - for example, 'jmp pc_relative_label' from 'jmp *absolute_label'. */ + for example, 'jmp pc_relative_label' from 'jmp *absolute_label'. */ if (*op_string == ABSOLUTE_PREFIX) { ++op_string; if (is_space_char (*op_string)) - ++op_string; + ++op_string; i.types[this_operand] |= JumpAbsolute; } - /* Check if operand is a register. */ - if (*op_string == REGISTER_PREFIX - || (allow_naked_reg && i386_is_reg (op_string))) + /* Check if operand is a register. */ + if ((*op_string == REGISTER_PREFIX || allow_naked_reg) + && (r = parse_register (op_string, &end_op)) != NULL) { - register const reg_entry *r; - char *end_op; - - r = parse_register (op_string, &end_op); - if (r == NULL) - return 0; - /* Check for a segment override by searching for ':' after a segment register. */ op_string = end_op; if (is_space_char (*op_string)) - ++op_string; + ++op_string; if (*op_string == ':' && (r->reg_type & (SReg2 | SReg3))) - { - switch (r->reg_num) - { - case 0: + { + switch (r->reg_num) + { + case 0: i.seg[i.mem_operands] = &es; break; case 1: @@ -3415,8 +3114,6 @@ i386_operand (operand_string) if (is_space_char (*op_string)) ++op_string; - /* Pretend given string starts here. */ - operand_string = op_string; if (!is_digit_char (*op_string) && !is_identifier_char (*op_string) && *op_string != '(' @@ -3425,7 +3122,7 @@ i386_operand (operand_string) as_bad (_("bad memory operand `%s'"), op_string); return 0; } - /* Handle case of %es:*foo. */ + /* Handle case of %es:*foo. */ if (*op_string == ABSOLUTE_PREFIX) { ++op_string; @@ -3437,34 +3134,41 @@ i386_operand (operand_string) } if (*op_string) { - as_bad (_("Junk `%s' after register"), op_string); + as_bad (_("junk `%s' after register"), op_string); return 0; } i.types[this_operand] |= r->reg_type & ~BaseIndex; - i.regs[this_operand] = r; + i.op[this_operand].regs = r; i.reg_operands++; } + else if (*op_string == REGISTER_PREFIX) + { + as_bad (_("bad register name `%s'"), op_string); + return 0; + } else if (*op_string == IMMEDIATE_PREFIX) - { /* ... or an immediate */ + { ++op_string; - if (!i386_immediate (op_string)) - return 0; - } - else if (is_digit_char (*op_string) - || is_identifier_char (*op_string) + if (i.types[this_operand] & JumpAbsolute) + { + as_bad (_("immediate operand illegal with absolute jump")); + return 0; + } + if (!i386_immediate (op_string)) + return 0; + } + else if (is_digit_char (*op_string) + || is_identifier_char (*op_string) || *op_string == '(' ) { - /* This is a memory reference of some sort. */ - char *end_of_operand_string; - register char *base_string; - int found_base_index_form; + /* This is a memory reference of some sort. */ + char *base_string; - /* Start and end of displacement string expression (if found). */ + /* Start and end of displacement string expression (if found). */ char *displacement_string_start; char *displacement_string_end; do_memory_reference: - if ((i.mem_operands == 1 && (current_templates->start->opcode_modifier & IsString) == 0) || i.mem_operands == 2) @@ -3478,20 +3182,22 @@ i386_operand (operand_string) looking for an ')' at the end of the operand, searching for the '(' matching it, and finding a REGISTER_PREFIX or ',' after the '('. */ - found_base_index_form = 0; - end_of_operand_string = op_string + strlen (op_string); + base_string = op_string + strlen (op_string); - --end_of_operand_string; - if (is_space_char (*end_of_operand_string)) - --end_of_operand_string; + --base_string; + if (is_space_char (*base_string)) + --base_string; - base_string = end_of_operand_string; + /* If we only have a displacement, set-up for it to be parsed later. */ + displacement_string_start = op_string; + displacement_string_end = base_string + 1; if (*base_string == ')') { + char *temp_string; unsigned int parens_balanced = 1; /* We've already checked that the number of left & right ()'s are - equal, so this loop will not be infinite. */ + equal, so this loop will not be infinite. */ do { base_string--; @@ -3502,115 +3208,93 @@ i386_operand (operand_string) } while (parens_balanced); - /* If there is a displacement set-up for it to be parsed later. */ - displacement_string_start = op_string; - displacement_string_end = base_string; + temp_string = base_string; /* Skip past '(' and whitespace. */ ++base_string; if (is_space_char (*base_string)) ++base_string; - if (*base_string == REGISTER_PREFIX - || (allow_naked_reg && i386_is_reg (base_string)) - || *base_string == ',') - found_base_index_form = 1; - } - - /* If we can't parse a base index register expression, we've found - a pure displacement expression. We set up displacement_string_start - and displacement_string_end for the code below. */ - if (!found_base_index_form) - { - displacement_string_start = op_string; - displacement_string_end = end_of_operand_string + 1; - } - else - { - i.types[this_operand] |= BaseIndex; - - /* Find base register (if any). */ - if (*base_string != ',') + if (*base_string == ',' + || ((*base_string == REGISTER_PREFIX || allow_naked_reg) + && (i.base_reg = parse_register (base_string, &end_op)) != NULL)) { - char *end_op; - - /* Trim off the closing ')' so that parse_register won't - see it. */ - END_STRING_AND_SAVE (end_of_operand_string); - i.base_reg = parse_register (base_string, &end_op); - RESTORE_END_STRING (end_of_operand_string); - - if (i.base_reg == NULL) - return 0; - - base_string = end_op; - if (is_space_char (*base_string)) - ++base_string; - } + displacement_string_end = temp_string; - /* There may be an index reg or scale factor here. */ - if (*base_string == ',') - { - ++base_string; - if (is_space_char (*base_string)) - ++base_string; + i.types[this_operand] |= BaseIndex; - if (*base_string == REGISTER_PREFIX - || (allow_naked_reg && i386_is_reg (base_string))) + if (i.base_reg) { - char *end_op; - - END_STRING_AND_SAVE (end_of_operand_string); - i.index_reg = parse_register (base_string, &end_op); - RESTORE_END_STRING (end_of_operand_string); - - if (i.index_reg == NULL) - return 0; - base_string = end_op; if (is_space_char (*base_string)) ++base_string; - if (*base_string == ',') + } + + /* There may be an index reg or scale factor here. */ + if (*base_string == ',') + { + ++base_string; + if (is_space_char (*base_string)) + ++base_string; + + if ((*base_string == REGISTER_PREFIX || allow_naked_reg) + && (i.index_reg = parse_register (base_string, &end_op)) != NULL) { - ++base_string; + base_string = end_op; if (is_space_char (*base_string)) ++base_string; + if (*base_string == ',') + { + ++base_string; + if (is_space_char (*base_string)) + ++base_string; + } + else if (*base_string != ')' ) + { + as_bad (_("expecting `,' or `)' after index register in `%s'"), + operand_string); + return 0; + } } - else if (*base_string != ')' ) + else if (*base_string == REGISTER_PREFIX) { - as_bad (_("expecting `,' or `)' after index register in `%s'"), - operand_string); + as_bad (_("bad register name `%s'"), base_string); return 0; } - } - /* Check for scale factor. */ - if (isdigit ((unsigned char) *base_string)) - { - if (!i386_scale (base_string)) - return 0; + /* Check for scale factor. */ + if (isdigit ((unsigned char) *base_string)) + { + if (!i386_scale (base_string)) + return 0; - ++base_string; - if (is_space_char (*base_string)) - ++base_string; - if (*base_string != ')') + ++base_string; + if (is_space_char (*base_string)) + ++base_string; + if (*base_string != ')') + { + as_bad (_("expecting `)' after scale factor in `%s'"), + operand_string); + return 0; + } + } + else if (!i.index_reg) { - as_bad (_("expecting `)' after scale factor in `%s'"), - operand_string); + as_bad (_("expecting index register or scale factor after `,'; got '%c'"), + *base_string); return 0; } } - else if (!i.index_reg) + else if (*base_string != ')') { - as_bad (_("expecting index register or scale factor after `,'; got '%c'"), - *base_string); + as_bad (_("expecting `,' or `)' after base register in `%s'"), + operand_string); return 0; } } - else if (*base_string != ')') + else if (*base_string == REGISTER_PREFIX) { - as_bad (_("expecting `,' or `)' after base register in `%s'"), - operand_string); + as_bad (_("bad register name `%s'"), base_string); return 0; } } @@ -3620,9 +3304,9 @@ i386_operand (operand_string) displacement_string_end are meaningful. */ if (displacement_string_start != displacement_string_end) { - if (!i386_displacement (displacement_string_start, - displacement_string_end)) - return 0; + if (!i386_displacement (displacement_string_start, + displacement_string_end)) + return 0; } /* Special case for (%dx) while doing input/output op. */ @@ -3636,166 +3320,148 @@ i386_operand (operand_string) i.types[this_operand] = InOutPortReg; return 1; } - /* Make sure the memory operand we've been dealt is valid. */ - if (flag_16bit_code ^ (i.prefix[ADDR_PREFIX] != 0)) - { - if ((i.base_reg - && ((i.base_reg->reg_type & (Reg16|BaseIndex)) - != (Reg16|BaseIndex))) - || (i.index_reg - && (((i.index_reg->reg_type & (Reg16|BaseIndex)) - != (Reg16|BaseIndex)) - || ! (i.base_reg - && i.base_reg->reg_num < 6 - && i.index_reg->reg_num >= 6 - && i.log2_scale_factor == 0)))) - { - as_bad (_("`%s' is not a valid %s bit base/index expression"), - operand_string, "16"); - return 0; - } - } - else - { - if ((i.base_reg - && (i.base_reg->reg_type & Reg32) == 0) - || (i.index_reg - && ((i.index_reg->reg_type & (Reg32|BaseIndex)) - != (Reg32|BaseIndex)))) - { - as_bad (_("`%s' is not a valid %s bit base/index expression"), - operand_string, "32"); - return 0; - } - } + + if (i386_index_check (operand_string) == 0) + return 0; i.mem_operands++; } else - { /* it's not a memory operand; argh! */ + { + /* It's not a memory operand; argh! */ as_bad (_("invalid char %s beginning operand %d `%s'"), output_invalid (*op_string), this_operand + 1, op_string); return 0; } - return 1; /* normal return */ + return 1; /* Normal return. */ } -/* - * md_estimate_size_before_relax() - * - * Called just before relax(). - * Any symbol that is now undefined will not become defined. - * Return the correct fr_subtype in the frag. - * Return the initial "guess for fr_var" to caller. - * The guess for fr_var is ACTUALLY the growth beyond fr_fix. - * Whatever we do to grow fr_fix or fr_var contributes to our returned value. - * Although it may not be explicit in the frag, pretend fr_var starts with a - * 0 value. - */ +/* md_estimate_size_before_relax() + + Called just before relax() for rs_machine_dependent frags. The x86 + assembler uses these frags to handle variable size jump + instructions. + + Any symbol that is now undefined will not become defined. + Return the correct fr_subtype in the frag. + Return the initial "guess for variable size of frag" to caller. + The guess is actually the growth beyond the fixed part. Whatever + we do to grow the fixed or variable part contributes to our + returned value. */ + int md_estimate_size_before_relax (fragP, segment) register fragS *fragP; register segT segment; { - register unsigned char *opcode; - register int old_fr_fix; - - old_fr_fix = fragP->fr_fix; - opcode = (unsigned char *) fragP->fr_opcode; /* We've already got fragP->fr_subtype right; all we have to do is - check for un-relaxable symbols. */ - if (S_GET_SEGMENT (fragP->fr_symbol) != segment) + check for un-relaxable symbols. On an ELF system, we can't relax + an externally visible symbol, because it may be overridden by a + shared library. */ + if (S_GET_SEGMENT (fragP->fr_symbol) != segment +#if defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF) + || S_IS_EXTERNAL (fragP->fr_symbol) + || S_IS_WEAK (fragP->fr_symbol) +#endif + ) { - /* symbol is undefined in this segment */ - int code16 = fragP->fr_subtype & CODE16; - int size = code16 ? 2 : 4; - int pcrel_reloc = code16 ? BFD_RELOC_16_PCREL : BFD_RELOC_32_PCREL; + /* Symbol is undefined in this segment, or we need to keep a + reloc so that weak symbols can be overridden. */ + int size = (fragP->fr_subtype & CODE16) ? 2 : 4; +#ifdef BFD_ASSEMBLER + enum bfd_reloc_code_real reloc_type; +#else + int reloc_type; +#endif + unsigned char *opcode; + int old_fr_fix; + + if (fragP->fr_var != NO_RELOC) + reloc_type = fragP->fr_var; + else if (size == 2) + reloc_type = BFD_RELOC_16_PCREL; + else + reloc_type = BFD_RELOC_32_PCREL; + + old_fr_fix = fragP->fr_fix; + opcode = (unsigned char *) fragP->fr_opcode; switch (opcode[0]) { - case JUMP_PC_RELATIVE: /* make jmp (0xeb) a dword displacement jump */ - opcode[0] = 0xe9; /* dword disp jmp */ + case JUMP_PC_RELATIVE: + /* Make jmp (0xeb) a dword displacement jump. */ + opcode[0] = 0xe9; fragP->fr_fix += size; fix_new (fragP, old_fr_fix, size, fragP->fr_symbol, fragP->fr_offset, 1, - (GOT_symbol && /* Not quite right - we should switch on - presence of @PLT, but I cannot see how - to get to that from here. We should have - done this in md_assemble to really - get it right all of the time, but I - think it does not matter that much, as - this will be right most of the time. ERY*/ - S_GET_SEGMENT(fragP->fr_symbol) == undefined_section) - ? BFD_RELOC_386_PLT32 : pcrel_reloc); + reloc_type); break; default: - /* This changes the byte-displacement jump 0x7N --> - the dword-displacement jump 0x0f8N */ + /* This changes the byte-displacement jump 0x7N + to the dword-displacement jump 0x0f,0x8N. */ opcode[1] = opcode[0] + 0x10; - opcode[0] = TWO_BYTE_OPCODE_ESCAPE; /* two-byte escape */ - fragP->fr_fix += 1 + size; /* we've added an opcode byte */ + opcode[0] = TWO_BYTE_OPCODE_ESCAPE; + /* We've added an opcode byte. */ + fragP->fr_fix += 1 + size; fix_new (fragP, old_fr_fix + 1, size, fragP->fr_symbol, fragP->fr_offset, 1, - (GOT_symbol && /* Not quite right - we should switch on - presence of @PLT, but I cannot see how - to get to that from here. ERY */ - S_GET_SEGMENT(fragP->fr_symbol) == undefined_section) - ? BFD_RELOC_386_PLT32 : pcrel_reloc); + reloc_type); break; } frag_wane (fragP); + return fragP->fr_fix - old_fr_fix; } - return (fragP->fr_var + fragP->fr_fix - old_fr_fix); -} /* md_estimate_size_before_relax() */ - -/* - * md_convert_frag(); - * - * Called after relax() is finished. - * In: Address of frag. - * fr_type == rs_machine_dependent. - * fr_subtype is what the address relaxed to. - * - * Out: Any fixSs and constants are set up. - * Caller will turn frag into a ".space 0". - */ + /* Guess a short jump. */ + return 1; +} + +/* Called after relax() is finished. + + In: Address of frag. + fr_type == rs_machine_dependent. + fr_subtype is what the address relaxed to. + + Out: Any fixSs and constants are set up. + Caller will turn frag into a ".space 0". */ + #ifndef BFD_ASSEMBLER void md_convert_frag (headers, sec, fragP) - object_headers *headers; - segT sec; + object_headers *headers ATTRIBUTE_UNUSED; + segT sec ATTRIBUTE_UNUSED; register fragS *fragP; #else void md_convert_frag (abfd, sec, fragP) - bfd *abfd; - segT sec; + bfd *abfd ATTRIBUTE_UNUSED; + segT sec ATTRIBUTE_UNUSED; register fragS *fragP; #endif { register unsigned char *opcode; unsigned char *where_to_put_displacement = NULL; - unsigned int target_address; - unsigned int opcode_address; + offsetT target_address; + offsetT opcode_address; unsigned int extension = 0; - int displacement_from_opcode_start; + offsetT displacement_from_opcode_start; opcode = (unsigned char *) fragP->fr_opcode; - /* Address we want to reach in file space. */ + /* Address we want to reach in file space. */ target_address = S_GET_VALUE (fragP->fr_symbol) + fragP->fr_offset; -#ifdef BFD_ASSEMBLER /* not needed otherwise? */ - target_address += fragP->fr_symbol->sy_frag->fr_address; +#ifdef BFD_ASSEMBLER + /* Not needed otherwise? */ + target_address += symbol_get_frag (fragP->fr_symbol)->fr_address; #endif - /* Address opcode resides at in file space. */ + /* Address opcode resides at in file space. */ opcode_address = fragP->fr_address + fragP->fr_fix; - /* Displacement from opcode start to fill into instruction. */ + /* Displacement from opcode start to fill into instruction. */ displacement_from_opcode_start = target_address - opcode_address; switch (fragP->fr_subtype) @@ -3804,33 +3470,33 @@ md_convert_frag (abfd, sec, fragP) case ENCODE_RELAX_STATE (COND_JUMP, SMALL16): case ENCODE_RELAX_STATE (UNCOND_JUMP, SMALL): case ENCODE_RELAX_STATE (UNCOND_JUMP, SMALL16): - /* don't have to change opcode */ - extension = 1; /* 1 opcode + 1 displacement */ + /* Don't have to change opcode. */ + extension = 1; /* 1 opcode + 1 displacement */ where_to_put_displacement = &opcode[1]; break; case ENCODE_RELAX_STATE (COND_JUMP, BIG): - extension = 5; /* 2 opcode + 4 displacement */ + extension = 5; /* 2 opcode + 4 displacement */ opcode[1] = opcode[0] + 0x10; opcode[0] = TWO_BYTE_OPCODE_ESCAPE; where_to_put_displacement = &opcode[2]; break; case ENCODE_RELAX_STATE (UNCOND_JUMP, BIG): - extension = 4; /* 1 opcode + 4 displacement */ + extension = 4; /* 1 opcode + 4 displacement */ opcode[0] = 0xe9; where_to_put_displacement = &opcode[1]; break; case ENCODE_RELAX_STATE (COND_JUMP, BIG16): - extension = 3; /* 2 opcode + 2 displacement */ + extension = 3; /* 2 opcode + 2 displacement */ opcode[1] = opcode[0] + 0x10; opcode[0] = TWO_BYTE_OPCODE_ESCAPE; where_to_put_displacement = &opcode[2]; break; case ENCODE_RELAX_STATE (UNCOND_JUMP, BIG16): - extension = 2; /* 1 opcode + 2 displacement */ + extension = 2; /* 1 opcode + 2 displacement */ opcode[0] = 0xe9; where_to_put_displacement = &opcode[1]; break; @@ -3839,29 +3505,34 @@ md_convert_frag (abfd, sec, fragP) BAD_CASE (fragP->fr_subtype); break; } - /* now put displacement after opcode */ + /* Now put displacement after opcode. */ md_number_to_chars ((char *) where_to_put_displacement, (valueT) (displacement_from_opcode_start - extension), SIZE_FROM_RELAX_STATE (fragP->fr_subtype)); fragP->fr_fix += extension; } +/* Size of byte displacement jmp. */ +int md_short_jump_size = 2; + +/* Size of dword displacement jmp. */ +int md_long_jump_size = 5; -int md_short_jump_size = 2; /* size of byte displacement jmp */ -int md_long_jump_size = 5; /* size of dword displacement jmp */ -const int md_reloc_size = 8; /* Size of relocation record */ +/* Size of relocation record. */ +const int md_reloc_size = 8; void md_create_short_jump (ptr, from_addr, to_addr, frag, to_symbol) char *ptr; addressT from_addr, to_addr; - fragS *frag; - symbolS *to_symbol; + fragS *frag ATTRIBUTE_UNUSED; + symbolS *to_symbol ATTRIBUTE_UNUSED; { - long offset; + offsetT offset; offset = to_addr - (from_addr + 2); - md_number_to_chars (ptr, (valueT) 0xeb, 1); /* opcode for byte-disp jump */ + /* Opcode for byte-disp jump. */ + md_number_to_chars (ptr, (valueT) 0xeb, 1); md_number_to_chars (ptr + 1, (valueT) offset, 1); } @@ -3869,25 +3540,14 @@ void md_create_long_jump (ptr, from_addr, to_addr, frag, to_symbol) char *ptr; addressT from_addr, to_addr; - fragS *frag; - symbolS *to_symbol; + fragS *frag ATTRIBUTE_UNUSED; + symbolS *to_symbol ATTRIBUTE_UNUSED; { - long offset; + offsetT offset; - if (flag_do_long_jump) - { - offset = to_addr - S_GET_VALUE (to_symbol); - md_number_to_chars (ptr, (valueT) 0xe9, 1);/* opcode for long jmp */ - md_number_to_chars (ptr + 1, (valueT) offset, 4); - fix_new (frag, (ptr + 1) - frag->fr_literal, 4, - to_symbol, (offsetT) 0, 0, BFD_RELOC_32); - } - else - { - offset = to_addr - (from_addr + 5); - md_number_to_chars (ptr, (valueT) 0xe9, 1); - md_number_to_chars (ptr + 1, (valueT) offset, 4); - } + offset = to_addr - (from_addr + 5); + md_number_to_chars (ptr, (valueT) 0xe9, 1); + md_number_to_chars (ptr + 1, (valueT) offset, 4); } /* Apply a fixup (fixS) to segment data, once it has been determined @@ -3899,9 +3559,14 @@ md_create_long_jump (ptr, from_addr, to_addr, frag, to_symbol) int md_apply_fix3 (fixP, valp, seg) - fixS *fixP; /* The fix we're to put in. */ - valueT *valp; /* Pointer to the value of the bits. */ - segT seg; /* Segment fix is from. */ + /* The fix we're to put in. */ + fixS *fixP; + + /* Pointer to the value of the bits. */ + valueT *valp; + + /* Segment fix is from. */ + segT seg ATTRIBUTE_UNUSED; { register char *p = fixP->fx_where + fixP->fx_frag->fr_literal; valueT value = *valp; @@ -3911,6 +3576,9 @@ md_apply_fix3 (fixP, valp, seg) { switch (fixP->fx_r_type) { + default: + break; + case BFD_RELOC_32: fixP->fx_r_type = BFD_RELOC_32_PCREL; break; @@ -3923,10 +3591,10 @@ md_apply_fix3 (fixP, valp, seg) } } - /* - * This is a hack. There should be a better way to - * handle this. - */ + /* This is a hack. There should be a better way to handle this. + This covers for the fact that bfd_install_relocation will + subtract the current location (for partial_inplace, PC relative + relocations); see more below. */ if ((fixP->fx_r_type == BFD_RELOC_32_PCREL || fixP->fx_r_type == BFD_RELOC_16_PCREL || fixP->fx_r_type == BFD_RELOC_8_PCREL) @@ -3941,55 +3609,61 @@ md_apply_fix3 (fixP, valp, seg) value += fixP->fx_where + fixP->fx_frag->fr_address; #endif #if defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF) - if (OUTPUT_FLAVOR == bfd_target_elf_flavour - && (S_GET_SEGMENT (fixP->fx_addsy) == seg - || (fixP->fx_addsy->bsym->flags & BSF_SECTION_SYM) != 0) - && ! S_IS_EXTERNAL (fixP->fx_addsy) - && ! S_IS_WEAK (fixP->fx_addsy) - && S_IS_DEFINED (fixP->fx_addsy) - && ! S_IS_COMMON (fixP->fx_addsy)) + if (OUTPUT_FLAVOR == bfd_target_elf_flavour) { - /* Yes, we add the values in twice. This is because - bfd_perform_relocation subtracts them out again. I think - bfd_perform_relocation is broken, but I don't dare change - it. FIXME. */ - value += fixP->fx_where + fixP->fx_frag->fr_address; + segT fseg = S_GET_SEGMENT (fixP->fx_addsy); + + if ((fseg == seg + || (symbol_section_p (fixP->fx_addsy) + && fseg != absolute_section)) + && ! S_IS_EXTERNAL (fixP->fx_addsy) + && ! S_IS_WEAK (fixP->fx_addsy) + && S_IS_DEFINED (fixP->fx_addsy) + && ! S_IS_COMMON (fixP->fx_addsy)) + { + /* Yes, we add the values in twice. This is because + bfd_perform_relocation subtracts them out again. I think + bfd_perform_relocation is broken, but I don't dare change + it. FIXME. */ + value += fixP->fx_where + fixP->fx_frag->fr_address; + } } #endif #if defined (OBJ_COFF) && defined (TE_PE) /* For some reason, the PE format does not store a section - address offset for a PC relative symbol. */ + address offset for a PC relative symbol. */ if (S_GET_SEGMENT (fixP->fx_addsy) != seg) value += md_pcrel_from (fixP); #endif } /* Fix a few things - the dynamic linker expects certain values here, - and we must not dissappoint it. */ + and we must not dissappoint it. */ #if defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF) if (OUTPUT_FLAVOR == bfd_target_elf_flavour && fixP->fx_addsy) - switch (fixP->fx_r_type) { - case BFD_RELOC_386_PLT32: - /* Make the jump instruction point to the address of the operand. At - runtime we merely add the offset to the actual PLT entry. */ - value = 0xfffffffc; - break; - case BFD_RELOC_386_GOTPC: -/* - * This is tough to explain. We end up with this one if we have + switch (fixP->fx_r_type) + { + case BFD_RELOC_386_PLT32: + /* Make the jump instruction point to the address of the operand. At + runtime we merely add the offset to the actual PLT entry. */ + value = -4; + break; + case BFD_RELOC_386_GOTPC: + +/* This is tough to explain. We end up with this one if we have * operands that look like "_GLOBAL_OFFSET_TABLE_+[.-.L284]". The goal * here is to obtain the absolute address of the GOT, and it is strongly * preferable from a performance point of view to avoid using a runtime - * relocation for this. The actual sequence of instructions often look + * relocation for this. The actual sequence of instructions often look * something like: - * - * call .L66 + * + * call .L66 * .L66: - * popl %ebx - * addl $_GLOBAL_OFFSET_TABLE_+[.-.L66],%ebx - * - * The call and pop essentially return the absolute address of + * popl %ebx + * addl $_GLOBAL_OFFSET_TABLE_+[.-.L66],%ebx + * + * The call and pop essentially return the absolute address of * the label .L66 and store it in %ebx. The linker itself will * ultimately change the first operand of the addl so that %ebx points to * the GOT, but to keep things simple, the .o file must have this operand @@ -3998,72 +3672,57 @@ md_apply_fix3 (fixP, valp, seg) * treat a GOTPC relocation as asking for a pcrel offset to the GOT to be * added in, and the addend of the relocation is stored in the operand * field for the instruction itself. - * - * Our job here is to fix the operand so that it would add the correct + * + * Our job here is to fix the operand so that it would add the correct * offset so that %ebx would point to itself. The thing that is tricky is * that .-.L66 will point to the beginning of the instruction, so we need * to further modify the operand so that it will point to itself. * There are other cases where you have something like: - * - * .long $_GLOBAL_OFFSET_TABLE_+[.-.L66] - * + * + * .long $_GLOBAL_OFFSET_TABLE_+[.-.L66] + * * and here no correction would be required. Internally in the assembler - * we treat operands of this form as not being pcrel since the '.' is + * we treat operands of this form as not being pcrel since the '.' is * explicitly mentioned, and I wonder whether it would simplify matters * to do it this way. Who knows. In earlier versions of the PIC patches, * the pcrel_adjust field was used to store the correction, but since the - * expression is not pcrel, I felt it would be confusing to do it this way. - */ - value -= 1; - break; - case BFD_RELOC_386_GOT32: - value = 0; /* Fully resolved at runtime. No addend. */ - break; - case BFD_RELOC_386_GOTOFF: - break; - - case BFD_RELOC_VTABLE_INHERIT: - case BFD_RELOC_VTABLE_ENTRY: - fixP->fx_done = 0; - return 1; - - default: - break; - } -#endif /* defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF) */ + * expression is not pcrel, I felt it would be confusing to do it this + * way. */ + + value -= 1; + break; + case BFD_RELOC_386_GOT32: + value = 0; /* Fully resolved at runtime. No addend. */ + break; + case BFD_RELOC_386_GOTOFF: + break; + + case BFD_RELOC_VTABLE_INHERIT: + case BFD_RELOC_VTABLE_ENTRY: + fixP->fx_done = 0; + return 1; + + default: + break; + } +#endif /* defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF) */ *valp = value; -#endif /* defined (BFD_ASSEMBLER) && !defined (TE_Mach) */ +#endif /* defined (BFD_ASSEMBLER) && !defined (TE_Mach) */ md_number_to_chars (p, value, fixP->fx_size); return 1; } - -#if 0 -/* This is never used. */ -long /* Knows about the byte order in a word. */ -md_chars_to_number (con, nbytes) - unsigned char con[]; /* Low order byte 1st. */ - int nbytes; /* Number of bytes in the input. */ -{ - long retval; - for (retval = 0, con += nbytes - 1; nbytes--; con--) - { - retval <<= BITS_PER_CHAR; - retval |= *con; - } - return retval; -} -#endif /* 0 */ - #define MAX_LITTLENUMS 6 -/* Turn the string pointed to by litP into a floating point constant of type - type, and emit the appropriate bytes. The number of LITTLENUMS emitted - is stored in *sizeP . An error message is returned, or NULL on OK. */ +/* Turn the string pointed to by litP into a floating point constant + of type TYPE, and emit the appropriate bytes. The number of + LITTLENUMS emitted is stored in *SIZEP. An error message is + returned, or NULL on OK. */ + char * md_atof (type, litP, sizeP) - char type; + int type; char *litP; int *sizeP; { @@ -4110,8 +3769,6 @@ md_atof (type, litP, sizeP) char output_invalid_buf[8]; -static char * output_invalid PARAMS ((int)); - static char * output_invalid (c) int c; @@ -4125,15 +3782,13 @@ output_invalid (c) /* REG_STRING starts *before* REGISTER_PREFIX. */ -static const reg_entry * parse_register PARAMS ((char *, char **)); - static const reg_entry * parse_register (reg_string, end_op) char *reg_string; char **end_op; { - register char *s = reg_string; - register char *p; + char *s = reg_string; + char *p; char reg_name_given[MAX_REG_NAME_SIZE + 1]; const reg_entry *r; @@ -4145,37 +3800,57 @@ parse_register (reg_string, end_op) ++s; p = reg_name_given; - while ((*p++ = register_chars[(unsigned char) *s++]) != '\0') + while ((*p++ = register_chars[(unsigned char) *s]) != '\0') { if (p >= reg_name_given + MAX_REG_NAME_SIZE) - { - if (!allow_naked_reg) - { - *p = '\0'; - as_bad (_("bad register name `%s'"), reg_name_given); - } - return (const reg_entry *) NULL; - } + return (const reg_entry *) NULL; + s++; } - *end_op = s - 1; + /* For naked regs, make sure that we are not dealing with an identifier. + This prevents confusing an identifier like `eax_var' with register + `eax'. */ + if (allow_naked_reg && identifier_chars[(unsigned char) *s]) + return (const reg_entry *) NULL; + + *end_op = s; r = (const reg_entry *) hash_find (reg_hash, reg_name_given); - if (r == NULL) + /* Handle floating point regs, allowing spaces in the (i) part. */ + if (r == i386_regtab /* %st is first entry of table */) { - if (!allow_naked_reg) - as_bad (_("bad register name `%s'"), reg_name_given); - return (const reg_entry *) NULL; + if (is_space_char (*s)) + ++s; + if (*s == '(') + { + ++s; + if (is_space_char (*s)) + ++s; + if (*s >= '0' && *s <= '7') + { + r = &i386_float_regtab[*s - '0']; + ++s; + if (is_space_char (*s)) + ++s; + if (*s == ')') + { + *end_op = s + 1; + return r; + } + } + /* We have "%st(" then garbage. */ + return (const reg_entry *) NULL; + } } return r; } -#ifdef OBJ_ELF -CONST char *md_shortopts = "kmVQ:"; +#if defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF) +const char *md_shortopts = "kVQ:sq"; #else -CONST char *md_shortopts = "m"; +const char *md_shortopts = "q"; #endif struct option md_longopts[] = { {NULL, no_argument, NULL, 0} @@ -4185,17 +3860,18 @@ size_t md_longopts_size = sizeof (md_longopts); int md_parse_option (c, arg) int c; - char *arg; + char *arg ATTRIBUTE_UNUSED; { switch (c) { - case 'm': - flag_do_long_jump = 1; + case 'q': + quiet_warnings = 1; break; #if defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF) - /* -k: Ignore for FreeBSD compatibility. */ - case 'k': + /* -Qy, -Qn: SVR4 arguments controlling whether a .comment section + should be emitted or not. FIXME: Not implemented. */ + case 'Q': break; /* -V: SVR4 argument to print version ID. */ @@ -4203,9 +3879,13 @@ md_parse_option (c, arg) print_version_id (); break; - /* -Qy, -Qn: SVR4 arguments controlling whether a .comment section - should be emitted or not. FIXME: Not implemented. */ - case 'Q': + /* -k: Ignore for FreeBSD compatibility. */ + case 'k': + break; + + case 's': + /* -s: On i386 Solaris, this tells the native assembler to use + .stab instead of .stab.excl. We always use .stab anyhow. */ break; #endif @@ -4219,72 +3899,94 @@ void md_show_usage (stream) FILE *stream; { +#if defined (OBJ_ELF) || defined (OBJ_MAYBE_ELF) + fprintf (stream, _("\ + -Q ignored\n\ + -V print assembler version number\n\ + -k ignored\n\ + -q quieten some warnings\n\ + -s ignored\n")); +#else fprintf (stream, _("\ --m do long jump\n")); + -q quieten some warnings\n")); +#endif } #ifdef BFD_ASSEMBLER -#ifdef OBJ_MAYBE_ELF -#ifdef OBJ_MAYBE_COFF +#if ((defined (OBJ_MAYBE_ELF) && defined (OBJ_MAYBE_COFF)) \ + || (defined (OBJ_MAYBE_ELF) && defined (OBJ_MAYBE_AOUT)) \ + || (defined (OBJ_MAYBE_COFF) && defined (OBJ_MAYBE_AOUT))) /* Pick the target format to use. */ -const char * +const char * i386_target_format () { switch (OUTPUT_FLAVOR) { +#ifdef OBJ_MAYBE_AOUT + case bfd_target_aout_flavour: + return AOUT_TARGET_FORMAT; +#endif +#ifdef OBJ_MAYBE_COFF case bfd_target_coff_flavour: return "coff-i386"; +#endif +#ifdef OBJ_MAYBE_ELF case bfd_target_elf_flavour: return "elf32-i386"; +#endif default: abort (); return NULL; } } -#endif /* OBJ_MAYBE_COFF */ -#endif /* OBJ_MAYBE_ELF */ -#endif /* BFD_ASSEMBLER */ +#endif /* OBJ_MAYBE_ more than one */ +#endif /* BFD_ASSEMBLER */ -/* ARGSUSED */ symbolS * md_undefined_symbol (name) char *name; { - if (*name == '_' && *(name+1) == 'G' - && strcmp(name, GLOBAL_OFFSET_TABLE_NAME) == 0) - { - if (!GOT_symbol) - { - if (symbol_find (name)) - as_bad (_("GOT already in symbol table")); - GOT_symbol = symbol_new (name, undefined_section, - (valueT) 0, &zero_address_frag); - }; - return GOT_symbol; - } + if (name[0] == GLOBAL_OFFSET_TABLE_NAME[0] + && name[1] == GLOBAL_OFFSET_TABLE_NAME[1] + && name[2] == GLOBAL_OFFSET_TABLE_NAME[2] + && strcmp (name, GLOBAL_OFFSET_TABLE_NAME) == 0) + { + if (!GOT_symbol) + { + if (symbol_find (name)) + as_bad (_("GOT already in symbol table")); + GOT_symbol = symbol_new (name, undefined_section, + (valueT) 0, &zero_address_frag); + }; + return GOT_symbol; + } return 0; } /* Round up a section size to the appropriate boundary. */ + valueT md_section_align (segment, size) - segT segment; + segT segment ATTRIBUTE_UNUSED; valueT size; { -#ifdef OBJ_AOUT #ifdef BFD_ASSEMBLER - /* For a.out, force the section size to be aligned. If we don't do - this, BFD will align it for us, but it will not write out the - final bytes of the section. This may be a bug in BFD, but it is - easier to fix it here since that is how the other a.out targets - work. */ - int align; - - align = bfd_get_section_alignment (stdoutput, segment); - size = ((size + (1 << align) - 1) & ((valueT) -1 << align)); +#if (defined (OBJ_AOUT) || defined (OBJ_MAYBE_AOUT)) + if (OUTPUT_FLAVOR == bfd_target_aout_flavour) + { + /* For a.out, force the section size to be aligned. If we don't do + this, BFD will align it for us, but it will not write out the + final bytes of the section. This may be a bug in BFD, but it is + easier to fix it here since that is how the other a.out targets + work. */ + int align; + + align = bfd_get_section_alignment (stdoutput, segment); + size = ((size + (1 << align) - 1) & ((valueT) -1 << align)); + } #endif #endif @@ -4306,7 +4008,7 @@ md_pcrel_from (fixP) static void s_bss (ignore) - int ignore; + int ignore ATTRIBUTE_UNUSED; { register int temp; @@ -4317,7 +4019,6 @@ s_bss (ignore) #endif - #ifdef BFD_ASSEMBLER void @@ -4333,7 +4034,7 @@ i386_validate_fix (fixp) arelent * tc_gen_reloc (section, fixp) - asection *section; + asection *section ATTRIBUTE_UNUSED; fixS *fixp; { arelent *rel; @@ -4356,7 +4057,7 @@ tc_gen_reloc (section, fixp) switch (fixp->fx_size) { default: - as_bad (_("Can not do %d byte pc-relative relocation"), + as_bad (_("can not do %d byte pc-relative relocation"), fixp->fx_size); code = BFD_RELOC_32_PCREL; break; @@ -4370,7 +4071,7 @@ tc_gen_reloc (section, fixp) switch (fixp->fx_size) { default: - as_bad (_("Can not do %d byte relocation"), fixp->fx_size); + as_bad (_("can not do %d byte relocation"), fixp->fx_size); code = BFD_RELOC_32; break; case 1: code = BFD_RELOC_8; break; @@ -4387,7 +4088,8 @@ tc_gen_reloc (section, fixp) code = BFD_RELOC_386_GOTPC; rel = (arelent *) xmalloc (sizeof (arelent)); - rel->sym_ptr_ptr = &fixp->fx_addsy->bsym; + rel->sym_ptr_ptr = (asymbol **) xmalloc (sizeof (asymbol *)); + *rel->sym_ptr_ptr = symbol_get_bfdsym (fixp->fx_addsy); rel->address = fixp->fx_frag->fr_address + fixp->fx_where; /* HACK: Since i386 ELF uses Rel instead of Rela, encode the @@ -4404,7 +4106,7 @@ tc_gen_reloc (section, fixp) if (rel->howto == NULL) { as_bad_where (fixp->fx_file, fixp->fx_line, - _("Cannot represent relocation type %s"), + _("cannot represent relocation type %s"), bfd_get_reloc_code_name (code)); /* Set howto to a garbage value so that we can keep going. */ rel->howto = bfd_reloc_type_lookup (stdoutput, BFD_RELOC_32); @@ -4414,7 +4116,7 @@ tc_gen_reloc (section, fixp) return rel; } -#else /* ! BFD_ASSEMBLER */ +#else /* ! BFD_ASSEMBLER */ #if (defined(OBJ_AOUT) | defined(OBJ_BOUT)) void @@ -4423,12 +4125,10 @@ tc_aout_fix_to_chars (where, fixP, segment_address_in_file) fixS *fixP; relax_addressT segment_address_in_file; { - /* - * In: length of relocation (or of address) in chars: 1, 2 or 4. - * Out: GNU LD relocation length code: 0, 1, or 2. - */ + /* In: length of relocation (or of address) in chars: 1, 2 or 4. + Out: GNU LD relocation length code: 0, 1, or 2. */ - static const unsigned char nbytes_r_length[] = {42, 0, 1, 42, 2}; + static const unsigned char nbytes_r_length[] = { 42, 0, 1, 42, 2 }; long r_symbolnum; know (fixP->fx_addsy != NULL); @@ -4450,7 +4150,7 @@ tc_aout_fix_to_chars (where, fixP, segment_address_in_file) | (((fixP->fx_pcrel << 0) & 0x01) & 0x0f)); } -#endif /* OBJ_AOUT or OBJ_BOUT */ +#endif /* OBJ_AOUT or OBJ_BOUT. */ #if defined (I386COFF) @@ -4480,8 +4180,909 @@ tc_coff_sizemachdep (frag) return 0; } -#endif /* I386COFF */ +#endif /* I386COFF */ -#endif /* ! BFD_ASSEMBLER */ +#endif /* ! BFD_ASSEMBLER */ -/* end of tc-i386.c */ +/* Parse operands using Intel syntax. This implements a recursive descent + parser based on the BNF grammar published in Appendix B of the MASM 6.1 + Programmer's Guide. + + FIXME: We do not recognize the full operand grammar defined in the MASM + documentation. In particular, all the structure/union and + high-level macro operands are missing. + + Uppercase words are terminals, lower case words are non-terminals. + Objects surrounded by double brackets '[[' ']]' are optional. Vertical + bars '|' denote choices. Most grammar productions are implemented in + functions called 'intel_'. + + Initial production is 'expr'. + + addOp + | - + + alpha [a-zA-Z] + + byteRegister AL | AH | BL | BH | CL | CH | DL | DH + + constant digits [[ radixOverride ]] + + dataType BYTE | WORD | DWORD | QWORD | XWORD + + digits decdigit + | digits decdigit + | digits hexdigit + + decdigit [0-9] + + e05 e05 addOp e06 + | e06 + + e06 e06 mulOp e09 + | e09 + + e09 OFFSET e10 + | e09 PTR e10 + | e09 : e10 + | e10 + + e10 e10 [ expr ] + | e11 + + e11 ( expr ) + | [ expr ] + | constant + | dataType + | id + | $ + | register + + => expr SHORT e05 + | e05 + + gpRegister AX | EAX | BX | EBX | CX | ECX | DX | EDX + | BP | EBP | SP | ESP | DI | EDI | SI | ESI + + hexdigit a | b | c | d | e | f + | A | B | C | D | E | F + + id alpha + | id alpha + | id decdigit + + mulOp * | / | MOD + + quote " | ' + + register specialRegister + | gpRegister + | byteRegister + + segmentRegister CS | DS | ES | FS | GS | SS + + specialRegister CR0 | CR2 | CR3 + | DR0 | DR1 | DR2 | DR3 | DR6 | DR7 + | TR3 | TR4 | TR5 | TR6 | TR7 + + We simplify the grammar in obvious places (e.g., register parsing is + done by calling parse_register) and eliminate immediate left recursion + to implement a recursive-descent parser. + + expr SHORT e05 + | e05 + + e05 e06 e05' + + e05' addOp e06 e05' + | Empty + + e06 e09 e06' + + e06' mulOp e09 e06' + | Empty + + e09 OFFSET e10 e09' + | e10 e09' + + e09' PTR e10 e09' + | : e10 e09' + | Empty + + e10 e11 e10' + + e10' [ expr ] e10' + | Empty + + e11 ( expr ) + | [ expr ] + | BYTE + | WORD + | DWORD + | QWORD + | XWORD + | . + | $ + | register + | id + | constant */ + +/* Parsing structure for the intel syntax parser. Used to implement the + semantic actions for the operand grammar. */ +struct intel_parser_s + { + char *op_string; /* The string being parsed. */ + int got_a_float; /* Whether the operand is a float. */ + int op_modifier; /* Operand modifier. */ + int is_mem; /* 1 if operand is memory reference. */ + const reg_entry *reg; /* Last register reference found. */ + char *disp; /* Displacement string being built. */ + }; + +static struct intel_parser_s intel_parser; + +/* Token structure for parsing intel syntax. */ +struct intel_token + { + int code; /* Token code. */ + const reg_entry *reg; /* Register entry for register tokens. */ + char *str; /* String representation. */ + }; + +static struct intel_token cur_token, prev_token; + +/* Token codes for the intel parser. */ +#define T_NIL -1 +#define T_CONST 1 +#define T_REG 2 +#define T_BYTE 3 +#define T_WORD 4 +#define T_DWORD 5 +#define T_QWORD 6 +#define T_XWORD 7 +#define T_SHORT 8 +#define T_OFFSET 9 +#define T_PTR 10 +#define T_ID 11 + +/* Prototypes for intel parser functions. */ +static int intel_match_token PARAMS ((int code)); +static void intel_get_token PARAMS ((void)); +static void intel_putback_token PARAMS ((void)); +static int intel_expr PARAMS ((void)); +static int intel_e05 PARAMS ((void)); +static int intel_e05_1 PARAMS ((void)); +static int intel_e06 PARAMS ((void)); +static int intel_e06_1 PARAMS ((void)); +static int intel_e09 PARAMS ((void)); +static int intel_e09_1 PARAMS ((void)); +static int intel_e10 PARAMS ((void)); +static int intel_e10_1 PARAMS ((void)); +static int intel_e11 PARAMS ((void)); + +static int +i386_intel_operand (operand_string, got_a_float) + char *operand_string; + int got_a_float; +{ + int ret; + char *p; + + /* Initialize token holders. */ + cur_token.code = prev_token.code = T_NIL; + cur_token.reg = prev_token.reg = NULL; + cur_token.str = prev_token.str = NULL; + + /* Initialize parser structure. */ + p = intel_parser.op_string = (char *)malloc (strlen (operand_string) + 1); + if (p == NULL) + abort (); + strcpy (intel_parser.op_string, operand_string); + intel_parser.got_a_float = got_a_float; + intel_parser.op_modifier = -1; + intel_parser.is_mem = 0; + intel_parser.reg = NULL; + intel_parser.disp = (char *)malloc (strlen (operand_string) + 1); + if (intel_parser.disp == NULL) + abort (); + intel_parser.disp[0] = '\0'; + + /* Read the first token and start the parser. */ + intel_get_token (); + ret = intel_expr (); + + if (ret) + { + /* If we found a memory reference, hand it over to i386_displacement + to fill in the rest of the operand fields. */ + if (intel_parser.is_mem) + { + if ((i.mem_operands == 1 + && (current_templates->start->opcode_modifier & IsString) == 0) + || i.mem_operands == 2) + { + as_bad (_("too many memory references for '%s'"), + current_templates->start->name); + ret = 0; + } + else + { + char *s = intel_parser.disp; + i.mem_operands++; + + /* Add the displacement expression. */ + if (*s != '\0') + ret = i386_displacement (s, s + strlen (s)) + && i386_index_check (s); + } + } + + /* Constant and OFFSET expressions are handled by i386_immediate. */ + else if (intel_parser.op_modifier == OFFSET_FLAT + || intel_parser.reg == NULL) + ret = i386_immediate (intel_parser.disp); + } + + free (p); + free (intel_parser.disp); + + return ret; +} + +/* expr SHORT e05 + | e05 */ +static int +intel_expr () +{ + /* expr SHORT e05 */ + if (cur_token.code == T_SHORT) + { + intel_parser.op_modifier = SHORT; + intel_match_token (T_SHORT); + + return (intel_e05 ()); + } + + /* expr e05 */ + else + return intel_e05 (); +} + +/* e05 e06 e05' + + e05' addOp e06 e05' + | Empty */ +static int +intel_e05 () +{ + return (intel_e06 () && intel_e05_1 ()); +} + +static int +intel_e05_1 () +{ + /* e05' addOp e06 e05' */ + if (cur_token.code == '+' || cur_token.code == '-') + { + strcat (intel_parser.disp, cur_token.str); + intel_match_token (cur_token.code); + + return (intel_e06 () && intel_e05_1 ()); + } + + /* e05' Empty */ + else + return 1; +} + +/* e06 e09 e06' + + e06' mulOp e09 e06' + | Empty */ +static int +intel_e06 () +{ + return (intel_e09 () && intel_e06_1 ()); +} + +static int +intel_e06_1 () +{ + /* e06' mulOp e09 e06' */ + if (cur_token.code == '*' || cur_token.code == '/') + { + strcat (intel_parser.disp, cur_token.str); + intel_match_token (cur_token.code); + + return (intel_e09 () && intel_e06_1 ()); + } + + /* e06' Empty */ + else + return 1; +} + +/* e09 OFFSET e10 e09' + | e10 e09' + + e09' PTR e10 e09' + | : e10 e09' + | Empty */ +static int +intel_e09 () +{ + /* e09 OFFSET e10 e09' */ + if (cur_token.code == T_OFFSET) + { + intel_parser.is_mem = 0; + intel_parser.op_modifier = OFFSET_FLAT; + intel_match_token (T_OFFSET); + + return (intel_e10 () && intel_e09_1 ()); + } + + /* e09 e10 e09' */ + else + return (intel_e10 () && intel_e09_1 ()); +} + +static int +intel_e09_1 () +{ + /* e09' PTR e10 e09' */ + if (cur_token.code == T_PTR) + { + if (prev_token.code == T_BYTE) + i.suffix = BYTE_MNEM_SUFFIX; + + else if (prev_token.code == T_WORD) + { + if (intel_parser.got_a_float == 2) /* "fi..." */ + i.suffix = SHORT_MNEM_SUFFIX; + else + i.suffix = WORD_MNEM_SUFFIX; + } + + else if (prev_token.code == T_DWORD) + { + if (intel_parser.got_a_float == 1) /* "f..." */ + i.suffix = SHORT_MNEM_SUFFIX; + else + i.suffix = LONG_MNEM_SUFFIX; + } + + else if (prev_token.code == T_QWORD) + i.suffix = DWORD_MNEM_SUFFIX; + + else if (prev_token.code == T_XWORD) + i.suffix = LONG_DOUBLE_MNEM_SUFFIX; + + else + { + as_bad (_("Unknown operand modifier `%s'\n"), prev_token.str); + return 0; + } + + intel_match_token (T_PTR); + + return (intel_e10 () && intel_e09_1 ()); + } + + /* e09 : e10 e09' */ + else if (cur_token.code == ':') + { + intel_parser.is_mem = 1; + + return (intel_match_token (':') && intel_e10 () && intel_e09_1 ()); + } + + /* e09' Empty */ + else + return 1; +} + +/* e10 e11 e10' + + e10' [ expr ] e10' + | Empty */ +static int +intel_e10 () +{ + return (intel_e11 () && intel_e10_1 ()); +} + +static int +intel_e10_1 () +{ + /* e10' [ expr ] e10' */ + if (cur_token.code == '[') + { + intel_match_token ('['); + intel_parser.is_mem = 1; + + /* Add a '+' to the displacement string if necessary. */ + if (*intel_parser.disp != '\0') + strcat (intel_parser.disp, "+"); + + return (intel_expr () && intel_match_token (']') && intel_e10_1 ()); + } + + /* e10' Empty */ + else + return 1; +} + +/* e11 ( expr ) + | [ expr ] + | BYTE + | WORD + | DWORD + | QWORD + | XWORD + | $ + | . + | register + | id + | constant */ +static int +intel_e11 () +{ + /* e11 ( expr ) */ + if (cur_token.code == '(') + { + intel_match_token ('('); + strcat (intel_parser.disp, "("); + + if (intel_expr () && intel_match_token (')')) + { + strcat (intel_parser.disp, ")"); + return 1; + } + else + return 0; + } + + /* e11 [ expr ] */ + else if (cur_token.code == '[') + { + intel_match_token ('['); + intel_parser.is_mem = 1; + + /* Operands for jump/call inside brackets denote absolute addresses. */ + if (current_templates->start->opcode_modifier & Jump + || current_templates->start->opcode_modifier & JumpDword + || current_templates->start->opcode_modifier & JumpByte + || current_templates->start->opcode_modifier & JumpInterSegment) + i.types[this_operand] |= JumpAbsolute; + + /* Add a '+' to the displacement string if necessary. */ + if (*intel_parser.disp != '\0') + strcat (intel_parser.disp, "+"); + + return (intel_expr () && intel_match_token (']')); + } + + /* e11 BYTE + | WORD + | DWORD + | QWORD + | XWORD */ + else if (cur_token.code == T_BYTE + || cur_token.code == T_WORD + || cur_token.code == T_DWORD + || cur_token.code == T_QWORD + || cur_token.code == T_XWORD) + { + intel_match_token (cur_token.code); + + return 1; + } + + /* e11 $ + | . */ + else if (cur_token.code == '$' || cur_token.code == '.') + { + strcat (intel_parser.disp, cur_token.str); + intel_match_token (cur_token.code); + intel_parser.is_mem = 1; + + return 1; + } + + /* e11 register */ + else if (cur_token.code == T_REG) + { + const reg_entry *reg = intel_parser.reg = cur_token.reg; + + intel_match_token (T_REG); + + /* Check for segment change. */ + if (cur_token.code == ':') + { + if (reg->reg_type & (SReg2 | SReg3)) + { + switch (reg->reg_num) + { + case 0: + i.seg[i.mem_operands] = &es; + break; + case 1: + i.seg[i.mem_operands] = &cs; + break; + case 2: + i.seg[i.mem_operands] = &ss; + break; + case 3: + i.seg[i.mem_operands] = &ds; + break; + case 4: + i.seg[i.mem_operands] = &fs; + break; + case 5: + i.seg[i.mem_operands] = &gs; + break; + } + } + else + { + as_bad (_("`%s' is not a valid segment register"), reg->reg_name); + return 0; + } + } + + /* Not a segment register. Check for register scaling. */ + else if (cur_token.code == '*') + { + if (!intel_parser.is_mem) + { + as_bad (_("Register scaling only allowed in memory operands.")); + return 0; + } + + /* What follows must be a valid scale. */ + if (intel_match_token ('*') + && strchr ("01248", *cur_token.str)) + { + i.index_reg = reg; + i.types[this_operand] |= BaseIndex; + + /* Set the scale after setting the register (otherwise, + i386_scale will complain) */ + i386_scale (cur_token.str); + intel_match_token (T_CONST); + } + else + { + as_bad (_("expecting scale factor of 1, 2, 4, or 8: got `%s'"), + cur_token.str); + return 0; + } + } + + /* No scaling. If this is a memory operand, the register is either a + base register (first occurrence) or an index register (second + occurrence). */ + else if (intel_parser.is_mem && !(reg->reg_type & (SReg2 | SReg3))) + { + if (i.base_reg && i.index_reg) + { + as_bad (_("Too many register references in memory operand.\n")); + return 0; + } + + if (i.base_reg == NULL) + i.base_reg = reg; + else + i.index_reg = reg; + + i.types[this_operand] |= BaseIndex; + } + + /* Offset modifier. Add the register to the displacement string to be + parsed as an immediate expression after we're done. */ + else if (intel_parser.op_modifier == OFFSET_FLAT) + strcat (intel_parser.disp, reg->reg_name); + + /* It's neither base nor index nor offset. */ + else + { + i.types[this_operand] |= reg->reg_type & ~BaseIndex; + i.op[this_operand].regs = reg; + i.reg_operands++; + } + + /* Since registers are not part of the displacement string (except + when we're parsing offset operands), we may need to remove any + preceding '+' from the displacement string. */ + if (*intel_parser.disp != '\0' + && intel_parser.op_modifier != OFFSET_FLAT) + { + char *s = intel_parser.disp; + s += strlen (s) - 1; + if (*s == '+') + *s = '\0'; + } + + return 1; + } + + /* e11 id */ + else if (cur_token.code == T_ID) + { + /* Add the identifier to the displacement string. */ + strcat (intel_parser.disp, cur_token.str); + intel_match_token (T_ID); + + /* The identifier represents a memory reference only if it's not + preceded by an offset modifier. */ + if (intel_parser.op_modifier != OFFSET_FLAT + && intel_parser.op_modifier != FLAT) + intel_parser.is_mem = 1; + + return 1; + } + + /* e11 constant */ + else if (cur_token.code == T_CONST + || cur_token.code == '-' + || cur_token.code == '+') + { + char *save_str; + + /* Allow constants that start with `+' or `-'. */ + if (cur_token.code == '-' || cur_token.code == '+') + { + strcat (intel_parser.disp, cur_token.str); + intel_match_token (cur_token.code); + if (cur_token.code != T_CONST) + { + as_bad (_("Syntax error. Expecting a constant. Got `%s'.\n"), + cur_token.str); + return 0; + } + } + + save_str = (char *)malloc (strlen (cur_token.str) + 1); + if (save_str == NULL) + abort (); + strcpy (save_str, cur_token.str); + + /* Get the next token to check for register scaling. */ + intel_match_token (cur_token.code); + + /* Check if this constant is a scaling factor for an index register. */ + if (cur_token.code == '*') + { + if (intel_match_token ('*') && cur_token.code == T_REG) + { + if (!intel_parser.is_mem) + { + as_bad (_("Register scaling only allowed in memory operands.")); + return 0; + } + + /* The constant is followed by `* reg', so it must be + a valid scale. */ + if (strchr ("01248", *save_str)) + { + i.index_reg = cur_token.reg; + i.types[this_operand] |= BaseIndex; + + /* Set the scale after setting the register (otherwise, + i386_scale will complain) */ + i386_scale (save_str); + intel_match_token (T_REG); + + /* Since registers are not part of the displacement + string, we may need to remove any preceding '+' from + the displacement string. */ + if (*intel_parser.disp != '\0') + { + char *s = intel_parser.disp; + s += strlen (s) - 1; + if (*s == '+') + *s = '\0'; + } + + free (save_str); + + return 1; + } + else + return 0; + } + + /* The constant was not used for register scaling. Since we have + already consumed the token following `*' we now need to put it + back in the stream. */ + else + intel_putback_token (); + } + + /* Add the constant to the displacement string. */ + strcat (intel_parser.disp, save_str); + free (save_str); + + return 1; + } + + as_bad (_("Unrecognized token '%s'"), cur_token.str); + return 0; +} + +/* Match the given token against cur_token. If they match, read the next + token from the operand string. */ +static int +intel_match_token (code) + int code; +{ + if (cur_token.code == code) + { + intel_get_token (); + return 1; + } + else + { + as_bad (_("Unexpected token `%s'\n"), cur_token.str); + return 0; + } +} + +/* Read a new token from intel_parser.op_string and store it in cur_token. */ +static void +intel_get_token () +{ + char *end_op; + const reg_entry *reg; + struct intel_token new_token; + + new_token.code = T_NIL; + new_token.reg = NULL; + new_token.str = NULL; + + /* Free the memory allocated to the previous token and move + cur_token to prev_token. */ + if (prev_token.str) + free (prev_token.str); + + prev_token = cur_token; + + /* Skip whitespace. */ + while (is_space_char (*intel_parser.op_string)) + intel_parser.op_string++; + + /* Return an empty token if we find nothing else on the line. */ + if (*intel_parser.op_string == '\0') + { + cur_token = new_token; + return; + } + + /* The new token cannot be larger than the remainder of the operand + string. */ + new_token.str = (char *)malloc (strlen (intel_parser.op_string) + 1); + if (new_token.str == NULL) + abort (); + new_token.str[0] = '\0'; + + if (strchr ("0123456789", *intel_parser.op_string)) + { + char *p = new_token.str; + char *q = intel_parser.op_string; + new_token.code = T_CONST; + + /* Allow any kind of identifier char to encompass floating point and + hexadecimal numbers. */ + while (is_identifier_char (*q)) + *p++ = *q++; + *p = '\0'; + + /* Recognize special symbol names [0-9][bf]. */ + if (strlen (intel_parser.op_string) == 2 + && (intel_parser.op_string[1] == 'b' + || intel_parser.op_string[1] == 'f')) + new_token.code = T_ID; + } + + else if (strchr ("+-/*:[]()", *intel_parser.op_string)) + { + new_token.code = *intel_parser.op_string; + new_token.str[0] = *intel_parser.op_string; + new_token.str[1] = '\0'; + } + + else if ((*intel_parser.op_string == REGISTER_PREFIX || allow_naked_reg) + && ((reg = parse_register (intel_parser.op_string, &end_op)) != NULL)) + { + new_token.code = T_REG; + new_token.reg = reg; + + if (*intel_parser.op_string == REGISTER_PREFIX) + { + new_token.str[0] = REGISTER_PREFIX; + new_token.str[1] = '\0'; + } + + strcat (new_token.str, reg->reg_name); + } + + else if (is_identifier_char (*intel_parser.op_string)) + { + char *p = new_token.str; + char *q = intel_parser.op_string; + + /* A '.' or '$' followed by an identifier char is an identifier. + Otherwise, it's operator '.' followed by an expression. */ + if ((*q == '.' || *q == '$') && !is_identifier_char (*(q + 1))) + { + new_token.code = *q; + new_token.str[0] = *q; + new_token.str[1] = '\0'; + } + else + { + while (is_identifier_char (*q) || *q == '@') + *p++ = *q++; + *p = '\0'; + + if (strcasecmp (new_token.str, "BYTE") == 0) + new_token.code = T_BYTE; + + else if (strcasecmp (new_token.str, "WORD") == 0) + new_token.code = T_WORD; + + else if (strcasecmp (new_token.str, "DWORD") == 0) + new_token.code = T_DWORD; + + else if (strcasecmp (new_token.str, "QWORD") == 0) + new_token.code = T_QWORD; + + else if (strcasecmp (new_token.str, "XWORD") == 0) + new_token.code = T_XWORD; + + else if (strcasecmp (new_token.str, "PTR") == 0) + new_token.code = T_PTR; + + else if (strcasecmp (new_token.str, "SHORT") == 0) + new_token.code = T_SHORT; + + else if (strcasecmp (new_token.str, "OFFSET") == 0) + { + new_token.code = T_OFFSET; + + /* ??? This is not mentioned in the MASM grammar but gcc + makes use of it with -mintel-syntax. OFFSET may be + followed by FLAT: */ + if (strncasecmp (q, " FLAT:", 6) == 0) + strcat (new_token.str, " FLAT:"); + } + + /* ??? This is not mentioned in the MASM grammar. */ + else if (strcasecmp (new_token.str, "FLAT") == 0) + new_token.code = T_OFFSET; + + else + new_token.code = T_ID; + } + } + + else + as_bad (_("Unrecognized token `%s'\n"), intel_parser.op_string); + + intel_parser.op_string += strlen (new_token.str); + cur_token = new_token; +} + +/* Put cur_token back into the token stream and make cur_token point to + prev_token. */ +static void +intel_putback_token () +{ + intel_parser.op_string -= strlen (cur_token.str); + free (cur_token.str); + cur_token = prev_token; + + /* Forget prev_token. */ + prev_token.code = T_NIL; + prev_token.reg = NULL; + prev_token.str = NULL; +}