1 /* ----------------------------------------------------------------------- *
3 * Copyright 1996-2010 The NASM Authors - All Rights Reserved
4 * See the file AUTHORS included with the NASM distribution for
5 * the specific copyright holders.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following
11 * * Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * * Redistributions in binary form must reproduce the above
14 * copyright notice, this list of conditions and the following
15 * disclaimer in the documentation and/or other materials provided
16 * with the distribution.
18 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
19 * CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
20 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
21 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
22 * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
23 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
24 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
25 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
26 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
29 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
30 * EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32 * ----------------------------------------------------------------------- */
35 * assemble.c code generation for the Netwide Assembler
37 * the actual codes (C syntax, i.e. octal):
38 * \0 - terminates the code. (Unless it's a literal of course.)
39 * \1..\4 - that many literal bytes follow in the code stream
40 * \5 - add 4 to the primary operand number (b, low octdigit)
41 * \6 - add 4 to the secondary operand number (a, middle octdigit)
42 * \7 - add 4 to both the primary and the secondary operand number
43 * \10..\13 - a literal byte follows in the code stream, to be added
44 * to the register value of operand 0..3
45 * \14..\17 - a signed byte immediate operand, from operand 0..3
46 * \20..\23 - a byte immediate operand, from operand 0..3
47 * \24..\27 - an unsigned byte immediate operand, from operand 0..3
48 * \30..\33 - a word immediate operand, from operand 0..3
49 * \34..\37 - select between \3[0-3] and \4[0-3] depending on 16/32 bit
50 * assembly mode or the operand-size override on the operand
51 * \40..\43 - a long immediate operand, from operand 0..3
52 * \44..\47 - select between \3[0-3], \4[0-3] and \5[4-7]
53 * depending on the address size of the instruction.
54 * \50..\53 - a byte relative operand, from operand 0..3
55 * \54..\57 - a qword immediate operand, from operand 0..3
56 * \60..\63 - a word relative operand, from operand 0..3
57 * \64..\67 - select between \6[0-3] and \7[0-3] depending on 16/32 bit
58 * assembly mode or the operand-size override on the operand
59 * \70..\73 - a long relative operand, from operand 0..3
60 * \74..\77 - a word constant, from the _segment_ part of operand 0..3
61 * \1ab - a ModRM, calculated on EA in operand a, with the spare
62 * field the register value of operand b.
63 * \140..\143 - an immediate word or signed byte for operand 0..3
64 * \144..\147 - or 2 (s-field) into opcode byte if operand 0..3
65 * is a signed byte rather than a word. Opcode byte follows.
66 * \150..\153 - an immediate dword or signed byte for operand 0..3
67 * \154..\157 - or 2 (s-field) into opcode byte if operand 0..3
68 * is a signed byte rather than a dword. Opcode byte follows.
69 * \160..\163 - this instruction uses DREX rather than REX, with the
70 * OC0 field set to 0, and the dest field taken from
72 * \164..\167 - this instruction uses DREX rather than REX, with the
73 * OC0 field set to 1, and the dest field taken from
75 * \171 - placement of DREX suffix in the absence of an EA
76 * \172\ab - the register number from operand a in bits 7..4, with
77 * the 4-bit immediate from operand b in bits 3..0.
78 * \173\xab - the register number from operand a in bits 7..4, with
79 * the value b in bits 3..0.
80 * \174\a - the register number from operand a in bits 7..4, and
81 * an arbitrary value in bits 3..0 (assembled as zero.)
82 * \2ab - a ModRM, calculated on EA in operand a, with the spare
83 * field equal to digit b.
84 * \250..\253 - same as \150..\153, except warn if the 64-bit operand
85 * is not equal to the truncated and sign-extended 32-bit
86 * operand; used for 32-bit immediates in 64-bit mode.
87 * \254..\257 - a signed 32-bit operand to be extended to 64 bits.
88 * \260..\263 - this instruction uses VEX/XOP rather than REX, with the
89 * V field taken from operand 0..3.
90 * \270 - this instruction uses VEX/XOP rather than REX, with the
91 * V field set to 1111b.
93 * VEX/XOP prefixes are followed by the sequence:
94 * \tmm\wlp where mm is the M field; and wlp is:
96 * [w0] ww = 0 for W = 0
97 * [w1] ww = 1 for W = 1
98 * [wx] ww = 2 for W don't care (always assembled as 0)
99 * [ww] ww = 3 for W used as REX.W
101 * t = 0 for VEX (C4/C5), t = 1 for XOP (8F).
103 * \274..\277 - a signed byte immediate operand, from operand 0..3,
104 * which is to be extended to the operand size.
105 * \310 - indicates fixed 16-bit address size, i.e. optional 0x67.
106 * \311 - indicates fixed 32-bit address size, i.e. optional 0x67.
107 * \312 - (disassembler only) invalid with non-default address size.
108 * \313 - indicates fixed 64-bit address size, 0x67 invalid.
109 * \314 - (disassembler only) invalid with REX.B
110 * \315 - (disassembler only) invalid with REX.X
111 * \316 - (disassembler only) invalid with REX.R
112 * \317 - (disassembler only) invalid with REX.W
113 * \320 - indicates fixed 16-bit operand size, i.e. optional 0x66.
114 * \321 - indicates fixed 32-bit operand size, i.e. optional 0x66.
115 * \322 - indicates that this instruction is only valid when the
116 * operand size is the default (instruction to disassembler,
117 * generates no code in the assembler)
118 * \323 - indicates fixed 64-bit operand size, REX on extensions only.
119 * \324 - indicates 64-bit operand size requiring REX prefix.
120 * \325 - instruction which always uses spl/bpl/sil/dil
121 * \330 - a literal byte follows in the code stream, to be added
122 * to the condition code value of the instruction.
123 * \331 - instruction not valid with REP prefix. Hint for
124 * disassembler only; for SSE instructions.
125 * \332 - REP prefix (0xF2 byte) used as opcode extension.
126 * \333 - REP prefix (0xF3 byte) used as opcode extension.
127 * \334 - LOCK prefix used as REX.R (used in non-64-bit mode)
128 * \335 - disassemble a rep (0xF3 byte) prefix as repe not rep.
129 * \336 - force a REP(E) prefix (0xF2) even if not specified.
130 * \337 - force a REPNE prefix (0xF3) even if not specified.
131 * \336-\337 are still listed as prefixes in the disassembler.
132 * \340 - reserve <operand 0> bytes of uninitialized storage.
133 * Operand 0 had better be a segmentless constant.
134 * \341 - this instruction needs a WAIT "prefix"
135 * \344,\345 - the PUSH/POP (respectively) codes for CS, DS, ES, SS
136 * (POP is never used for CS) depending on operand 0
137 * \346,\347 - the second byte of PUSH/POP codes for FS, GS, depending
139 * \360 - no SSE prefix (== \364\331)
140 * \361 - 66 SSE prefix (== \366\331)
141 * \362 - F2 SSE prefix (== \364\332)
142 * \363 - F3 SSE prefix (== \364\333)
143 * \364 - operand-size prefix (0x66) not permitted
144 * \365 - address-size prefix (0x67) not permitted
145 * \366 - operand-size prefix (0x66) used as opcode extension
146 * \367 - address-size prefix (0x67) used as opcode extension
147 * \370,\371,\372 - match only if operand 0 meets byte jump criteria.
148 * 370 is used for Jcc, 371 is used for JMP.
149 * \373 - assemble 0x03 if bits==16, 0x05 if bits==32;
150 * used for conditional jump over longer jump
153 #include "compiler.h"
157 #include <inttypes.h>
161 #include "assemble.h"
167 * Matching errors. These should be sorted so that more specific
168 * errors come later in the sequence.
176 * Matching success; the conditional ones first
178 MOK_JUMP
, /* Matching OK but needs jmp_match() */
179 MOK_GOOD
/* Matching unconditionally OK */
183 int sib_present
; /* is a SIB byte necessary? */
184 int bytes
; /* # of bytes of offset needed */
185 int size
; /* lazy - this is sib+bytes+1 */
186 uint8_t modrm
, sib
, rex
, rip
; /* the bytes themselves */
189 static uint32_t cpu
; /* cpu level received from nasm.c */
190 static efunc errfunc
;
191 static struct ofmt
*outfmt
;
192 static ListGen
*list
;
194 static int64_t calcsize(int32_t, int64_t, int, insn
*, const uint8_t *);
195 static void gencode(int32_t segment
, int64_t offset
, int bits
,
196 insn
* ins
, const struct itemplate
*temp
,
198 static enum match_result
find_match(const struct itemplate
**tempp
,
200 int32_t segment
, int64_t offset
, int bits
);
201 static enum match_result
matches(const struct itemplate
*, insn
*, int bits
);
202 static opflags_t
regflag(const operand
*);
203 static int32_t regval(const operand
*);
204 static int rexflags(int, opflags_t
, int);
205 static int op_rexflags(const operand
*, int);
206 static ea
*process_ea(operand
*, ea
*, int, int, int, opflags_t
);
207 static void add_asp(insn
*, int);
209 static int has_prefix(insn
* ins
, enum prefix_pos pos
, enum prefixes prefix
)
211 return ins
->prefixes
[pos
] == prefix
;
214 static void assert_no_prefix(insn
* ins
, enum prefix_pos pos
)
216 if (ins
->prefixes
[pos
])
217 errfunc(ERR_NONFATAL
, "invalid %s prefix",
218 prefix_name(ins
->prefixes
[pos
]));
221 static const char *size_name(int size
)
243 static void warn_overflow(int pass
, int size
)
245 errfunc(ERR_WARNING
| pass
| ERR_WARN_NOV
,
246 "%s data exceeds bounds", size_name(size
));
249 static void warn_overflow_const(int64_t data
, int size
)
251 if (overflow_general(data
, size
))
252 warn_overflow(ERR_PASS1
, size
);
255 static void warn_overflow_opd(const struct operand
*o
, int size
)
257 if (o
->wrt
== NO_SEG
&& o
->segment
== NO_SEG
) {
258 if (overflow_general(o
->offset
, size
))
259 warn_overflow(ERR_PASS2
, size
);
264 * This routine wrappers the real output format's output routine,
265 * in order to pass a copy of the data off to the listing file
266 * generator at the same time.
268 static void out(int64_t offset
, int32_t segto
, const void *data
,
269 enum out_type type
, uint64_t size
,
270 int32_t segment
, int32_t wrt
)
272 static int32_t lineno
= 0; /* static!!! */
273 static char *lnfname
= NULL
;
276 if (type
== OUT_ADDRESS
&& segment
== NO_SEG
&& wrt
== NO_SEG
) {
278 * This is a non-relocated address, and we're going to
279 * convert it into RAWDATA format.
284 errfunc(ERR_PANIC
, "OUT_ADDRESS with size > 8");
288 WRITEADDR(q
, *(int64_t *)data
, size
);
293 list
->output(offset
, data
, type
, size
);
296 * this call to src_get determines when we call the
297 * debug-format-specific "linenum" function
298 * it updates lineno and lnfname to the current values
299 * returning 0 if "same as last time", -2 if lnfname
300 * changed, and the amount by which lineno changed,
301 * if it did. thus, these variables must be static
304 if (src_get(&lineno
, &lnfname
))
305 outfmt
->current_dfmt
->linenum(lnfname
, lineno
, segto
);
307 outfmt
->output(segto
, data
, type
, size
, segment
, wrt
);
310 static bool jmp_match(int32_t segment
, int64_t offset
, int bits
,
311 insn
* ins
, const uint8_t *code
)
316 if ((c
!= 0370 && c
!= 0371) || (ins
->oprs
[0].type
& STRICT
))
320 if (optimizing
< 0 && c
== 0371)
323 isize
= calcsize(segment
, offset
, bits
, ins
, code
);
325 if (ins
->oprs
[0].opflags
& OPFLAG_UNKNOWN
)
326 /* Be optimistic in pass 1 */
329 if (ins
->oprs
[0].segment
!= segment
)
332 isize
= ins
->oprs
[0].offset
- offset
- isize
; /* isize is delta */
333 return (isize
>= -128 && isize
<= 127); /* is it byte size? */
336 int64_t assemble(int32_t segment
, int64_t offset
, int bits
, uint32_t cp
,
337 insn
* instruction
, struct ofmt
*output
, efunc error
,
340 const struct itemplate
*temp
;
345 int64_t start
= offset
;
346 int64_t wsize
; /* size for DB etc. */
348 errfunc
= error
; /* to pass to other functions */
350 outfmt
= output
; /* likewise */
351 list
= listgen
; /* and again */
353 wsize
= idata_bytes(instruction
->opcode
);
359 int32_t t
= instruction
->times
;
362 "instruction->times < 0 (%ld) in assemble()", t
);
364 while (t
--) { /* repeat TIMES times */
365 list_for_each(e
, instruction
->eops
) {
366 if (e
->type
== EOT_DB_NUMBER
) {
368 errfunc(ERR_NONFATAL
,
369 "integer supplied to a DT, DO or DY"
372 out(offset
, segment
, &e
->offset
,
373 OUT_ADDRESS
, wsize
, e
->segment
, e
->wrt
);
376 } else if (e
->type
== EOT_DB_STRING
||
377 e
->type
== EOT_DB_STRING_FREE
) {
380 out(offset
, segment
, e
->stringval
,
381 OUT_RAWDATA
, e
->stringlen
, NO_SEG
, NO_SEG
);
382 align
= e
->stringlen
% wsize
;
385 align
= wsize
- align
;
386 out(offset
, segment
, zero_buffer
,
387 OUT_RAWDATA
, align
, NO_SEG
, NO_SEG
);
389 offset
+= e
->stringlen
+ align
;
392 if (t
> 0 && t
== instruction
->times
- 1) {
394 * Dummy call to list->output to give the offset to the
397 list
->output(offset
, NULL
, OUT_RAWDATA
, 0);
398 list
->uplevel(LIST_TIMES
);
401 if (instruction
->times
> 1)
402 list
->downlevel(LIST_TIMES
);
403 return offset
- start
;
406 if (instruction
->opcode
== I_INCBIN
) {
407 const char *fname
= instruction
->eops
->stringval
;
410 fp
= fopen(fname
, "rb");
412 error(ERR_NONFATAL
, "`incbin': unable to open file `%s'",
414 } else if (fseek(fp
, 0L, SEEK_END
) < 0) {
415 error(ERR_NONFATAL
, "`incbin': unable to seek on file `%s'",
418 static char buf
[4096];
419 size_t t
= instruction
->times
;
424 if (instruction
->eops
->next
) {
425 base
= instruction
->eops
->next
->offset
;
427 if (instruction
->eops
->next
->next
&&
428 len
> (size_t)instruction
->eops
->next
->next
->offset
)
429 len
= (size_t)instruction
->eops
->next
->next
->offset
;
432 * Dummy call to list->output to give the offset to the
435 list
->output(offset
, NULL
, OUT_RAWDATA
, 0);
436 list
->uplevel(LIST_INCBIN
);
440 fseek(fp
, base
, SEEK_SET
);
444 m
= fread(buf
, 1, l
> sizeof(buf
) ? sizeof(buf
) : l
, fp
);
447 * This shouldn't happen unless the file
448 * actually changes while we are reading
452 "`incbin': unexpected EOF while"
453 " reading file `%s'", fname
);
454 t
= 0; /* Try to exit cleanly */
457 out(offset
, segment
, buf
, OUT_RAWDATA
, m
,
462 list
->downlevel(LIST_INCBIN
);
463 if (instruction
->times
> 1) {
465 * Dummy call to list->output to give the offset to the
468 list
->output(offset
, NULL
, OUT_RAWDATA
, 0);
469 list
->uplevel(LIST_TIMES
);
470 list
->downlevel(LIST_TIMES
);
473 return instruction
->times
* len
;
475 return 0; /* if we're here, there's an error */
478 /* Check to see if we need an address-size prefix */
479 add_asp(instruction
, bits
);
481 m
= find_match(&temp
, instruction
, segment
, offset
, bits
);
485 int64_t insn_size
= calcsize(segment
, offset
, bits
,
486 instruction
, temp
->code
);
487 itimes
= instruction
->times
;
488 if (insn_size
< 0) /* shouldn't be, on pass two */
489 error(ERR_PANIC
, "errors made it through from pass one");
492 for (j
= 0; j
< MAXPREFIX
; j
++) {
494 switch (instruction
->prefixes
[j
]) {
512 error(ERR_WARNING
| ERR_PASS2
,
513 "cs segment base generated, but will be ignored in 64-bit mode");
519 error(ERR_WARNING
| ERR_PASS2
,
520 "ds segment base generated, but will be ignored in 64-bit mode");
526 error(ERR_WARNING
| ERR_PASS2
,
527 "es segment base generated, but will be ignored in 64-bit mode");
539 error(ERR_WARNING
| ERR_PASS2
,
540 "ss segment base generated, but will be ignored in 64-bit mode");
547 "segr6 and segr7 cannot be used as prefixes");
552 "16-bit addressing is not supported "
554 } else if (bits
!= 16)
564 "64-bit addressing is only supported "
588 error(ERR_PANIC
, "invalid instruction prefix");
591 out(offset
, segment
, &c
, OUT_RAWDATA
, 1,
596 insn_end
= offset
+ insn_size
;
597 gencode(segment
, offset
, bits
, instruction
,
600 if (itimes
> 0 && itimes
== instruction
->times
- 1) {
602 * Dummy call to list->output to give the offset to the
605 list
->output(offset
, NULL
, OUT_RAWDATA
, 0);
606 list
->uplevel(LIST_TIMES
);
609 if (instruction
->times
> 1)
610 list
->downlevel(LIST_TIMES
);
611 return offset
- start
;
615 case MERR_OPSIZEMISSING
:
616 error(ERR_NONFATAL
, "operation size not specified");
618 case MERR_OPSIZEMISMATCH
:
619 error(ERR_NONFATAL
, "mismatch in operand sizes");
622 error(ERR_NONFATAL
, "no instruction for this cpu level");
625 error(ERR_NONFATAL
, "instruction not supported in %d-bit mode",
630 "invalid combination of opcode and operands");
637 int64_t insn_size(int32_t segment
, int64_t offset
, int bits
, uint32_t cp
,
638 insn
* instruction
, efunc error
)
640 const struct itemplate
*temp
;
643 errfunc
= error
; /* to pass to other functions */
646 if (instruction
->opcode
== I_none
)
649 if (instruction
->opcode
== I_DB
|| instruction
->opcode
== I_DW
||
650 instruction
->opcode
== I_DD
|| instruction
->opcode
== I_DQ
||
651 instruction
->opcode
== I_DT
|| instruction
->opcode
== I_DO
||
652 instruction
->opcode
== I_DY
) {
654 int32_t isize
, osize
, wsize
;
657 wsize
= idata_bytes(instruction
->opcode
);
659 list_for_each(e
, instruction
->eops
) {
663 if (e
->type
== EOT_DB_NUMBER
) {
665 warn_overflow_const(e
->offset
, wsize
);
666 } else if (e
->type
== EOT_DB_STRING
||
667 e
->type
== EOT_DB_STRING_FREE
)
668 osize
= e
->stringlen
;
670 align
= (-osize
) % wsize
;
673 isize
+= osize
+ align
;
675 return isize
* instruction
->times
;
678 if (instruction
->opcode
== I_INCBIN
) {
679 const char *fname
= instruction
->eops
->stringval
;
684 fp
= fopen(fname
, "rb");
686 error(ERR_NONFATAL
, "`incbin': unable to open file `%s'",
688 else if (fseek(fp
, 0L, SEEK_END
) < 0)
689 error(ERR_NONFATAL
, "`incbin': unable to seek on file `%s'",
693 if (instruction
->eops
->next
) {
694 len
-= instruction
->eops
->next
->offset
;
695 if (instruction
->eops
->next
->next
&&
696 len
> (size_t)instruction
->eops
->next
->next
->offset
) {
697 len
= (size_t)instruction
->eops
->next
->next
->offset
;
700 val
= instruction
->times
* len
;
707 /* Check to see if we need an address-size prefix */
708 add_asp(instruction
, bits
);
710 m
= find_match(&temp
, instruction
, segment
, offset
, bits
);
712 /* we've matched an instruction. */
714 const uint8_t *codes
= temp
->code
;
717 isize
= calcsize(segment
, offset
, bits
, instruction
, codes
);
720 for (j
= 0; j
< MAXPREFIX
; j
++) {
721 switch (instruction
->prefixes
[j
]) {
747 return isize
* instruction
->times
;
749 return -1; /* didn't match any instruction */
753 static bool possible_sbyte(operand
*o
)
755 return o
->wrt
== NO_SEG
&& o
->segment
== NO_SEG
&&
756 !(o
->opflags
& OPFLAG_UNKNOWN
) &&
757 optimizing
>= 0 && !(o
->type
& STRICT
);
760 /* check that opn[op] is a signed byte of size 16 or 32 */
761 static bool is_sbyte16(operand
*o
)
765 if (!possible_sbyte(o
))
769 return v
>= -128 && v
<= 127;
772 static bool is_sbyte32(operand
*o
)
776 if (!possible_sbyte(o
))
780 return v
>= -128 && v
<= 127;
783 /* Common construct */
784 #define case4(x) case (x): case (x)+1: case (x)+2: case (x)+3
786 static int64_t calcsize(int32_t segment
, int64_t offset
, int bits
,
787 insn
* ins
, const uint8_t *codes
)
796 ins
->rex
= 0; /* Ensure REX is reset */
798 if (ins
->prefixes
[PPS_OSIZE
] == P_O64
)
801 (void)segment
; /* Don't warn that this parameter is unused */
802 (void)offset
; /* Don't warn that this parameter is unused */
806 op1
= (c
& 3) + ((opex
& 1) << 2);
807 op2
= ((c
>> 3) & 3) + ((opex
& 2) << 1);
808 opx
= &ins
->oprs
[op1
];
809 opex
= 0; /* For the next iteration */
816 codes
+= c
, length
+= c
;
827 op_rexflags(opx
, REX_B
|REX_H
|REX_P
|REX_W
);
842 if (opx
->type
& (BITS16
| BITS32
| BITS64
))
843 length
+= (opx
->type
& BITS16
) ? 2 : 4;
845 length
+= (bits
== 16) ? 2 : 4;
853 length
+= ins
->addr_size
>> 3;
861 length
+= 8; /* MOV reg64/imm */
869 if (opx
->type
& (BITS16
| BITS32
| BITS64
))
870 length
+= (opx
->type
& BITS16
) ? 2 : 4;
872 length
+= (bits
== 16) ? 2 : 4;
884 length
+= is_sbyte16(opx
) ? 1 : 2;
893 length
+= is_sbyte32(opx
) ? 1 : 4;
904 ins
->drexdst
= regval(opx
);
909 ins
->rex
|= REX_D
|REX_OC
;
910 ins
->drexdst
= regval(opx
);
924 length
+= is_sbyte32(opx
) ? 1 : 4;
933 ins
->drexdst
= regval(opx
);
934 ins
->vex_cm
= *codes
++;
935 ins
->vex_wlp
= *codes
++;
941 ins
->vex_cm
= *codes
++;
942 ins
->vex_wlp
= *codes
++;
955 length
+= (bits
!= 16) && !has_prefix(ins
, PPS_ASIZE
, P_A16
);
959 length
+= (bits
!= 32) && !has_prefix(ins
, PPS_ASIZE
, P_A32
);
966 if (bits
!= 64 || has_prefix(ins
, PPS_ASIZE
, P_A16
) ||
967 has_prefix(ins
, PPS_ASIZE
, P_A32
))
975 length
+= (bits
!= 16);
979 length
+= (bits
== 16);
1017 if (!ins
->prefixes
[PPS_LREP
])
1018 ins
->prefixes
[PPS_LREP
] = P_REP
;
1022 if (!ins
->prefixes
[PPS_LREP
])
1023 ins
->prefixes
[PPS_LREP
] = P_REPNE
;
1027 if (ins
->oprs
[0].segment
!= NO_SEG
)
1028 errfunc(ERR_NONFATAL
, "attempt to reserve non-constant"
1029 " quantity of BSS space");
1031 length
+= ins
->oprs
[0].offset
;
1035 if (!ins
->prefixes
[PPS_WAIT
])
1036 ins
->prefixes
[PPS_WAIT
] = P_WAIT
;
1086 struct operand
*opy
= &ins
->oprs
[op2
];
1088 ea_data
.rex
= 0; /* Ensure ea.REX is initially 0 */
1091 /* pick rfield from operand b (opx) */
1092 rflags
= regflag(opx
);
1093 rfield
= nasm_regvals
[opx
->basereg
];
1098 if (!process_ea(opy
, &ea_data
, bits
,
1099 ins
->addr_size
, rfield
, rflags
)) {
1100 errfunc(ERR_NONFATAL
, "invalid effective address");
1103 ins
->rex
|= ea_data
.rex
;
1104 length
+= ea_data
.size
;
1110 errfunc(ERR_PANIC
, "internal instruction table corrupt"
1111 ": instruction code \\%o (0x%02X) given", c
, c
);
1116 ins
->rex
&= rex_mask
;
1118 if (ins
->rex
& REX_NH
) {
1119 if (ins
->rex
& REX_H
) {
1120 errfunc(ERR_NONFATAL
, "instruction cannot use high registers");
1123 ins
->rex
&= ~REX_P
; /* Don't force REX prefix due to high reg */
1126 if (ins
->rex
& REX_V
) {
1127 int bad32
= REX_R
|REX_W
|REX_X
|REX_B
;
1129 if (ins
->rex
& REX_H
) {
1130 errfunc(ERR_NONFATAL
, "cannot use high register in vex instruction");
1133 switch (ins
->vex_wlp
& 030) {
1147 if (bits
!= 64 && ((ins
->rex
& bad32
) || ins
->drexdst
> 7)) {
1148 errfunc(ERR_NONFATAL
, "invalid operands in non-64-bit mode");
1151 if (ins
->vex_cm
!= 1 || (ins
->rex
& (REX_W
|REX_R
|REX_B
)))
1155 } else if (ins
->rex
& REX_D
) {
1156 if (ins
->rex
& REX_H
) {
1157 errfunc(ERR_NONFATAL
, "cannot use high register in drex instruction");
1160 if (bits
!= 64 && ((ins
->rex
& (REX_R
|REX_W
|REX_X
|REX_B
)) ||
1161 ins
->drexdst
> 7)) {
1162 errfunc(ERR_NONFATAL
, "invalid operands in non-64-bit mode");
1166 } else if (ins
->rex
& REX_REAL
) {
1167 if (ins
->rex
& REX_H
) {
1168 errfunc(ERR_NONFATAL
, "cannot use high register in rex instruction");
1170 } else if (bits
== 64) {
1172 } else if ((ins
->rex
& REX_L
) &&
1173 !(ins
->rex
& (REX_P
|REX_W
|REX_X
|REX_B
)) &&
1176 assert_no_prefix(ins
, PPS_LREP
);
1179 errfunc(ERR_NONFATAL
, "invalid operands in non-64-bit mode");
1187 #define EMIT_REX() \
1188 if (!(ins->rex & (REX_D|REX_V)) && (ins->rex & REX_REAL) && (bits == 64)) { \
1189 ins->rex = (ins->rex & REX_REAL)|REX_P; \
1190 out(offset, segment, &ins->rex, OUT_RAWDATA, 1, NO_SEG, NO_SEG); \
1195 static void gencode(int32_t segment
, int64_t offset
, int bits
,
1196 insn
* ins
, const struct itemplate
*temp
,
1199 static char condval
[] = { /* conditional opcodes */
1200 0x7, 0x3, 0x2, 0x6, 0x2, 0x4, 0xF, 0xD, 0xC, 0xE, 0x6, 0x2,
1201 0x3, 0x7, 0x3, 0x5, 0xE, 0xC, 0xD, 0xF, 0x1, 0xB, 0x9, 0x5,
1202 0x0, 0xA, 0xA, 0xB, 0x8, 0x4
1209 struct operand
*opx
;
1210 const uint8_t *codes
= temp
->code
;
1215 op1
= (c
& 3) + ((opex
& 1) << 2);
1216 op2
= ((c
>> 3) & 3) + ((opex
& 2) << 1);
1217 opx
= &ins
->oprs
[op1
];
1218 opex
= 0; /* For the next iteration */
1226 out(offset
, segment
, codes
, OUT_RAWDATA
, c
, NO_SEG
, NO_SEG
);
1239 bytes
[0] = *codes
++ + (regval(opx
) & 7);
1240 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1246 * The test for BITS8 and SBYTE here is intended to avoid
1247 * warning on optimizer actions due to SBYTE, while still
1248 * warn on explicit BYTE directives. Also warn, obviously,
1249 * if the optimizer isn't enabled.
1251 if (((opx
->type
& BITS8
) ||
1252 !(opx
->type
& temp
->opd
[op1
] & BYTENESS
)) &&
1253 (opx
->offset
< -128 || opx
->offset
> 127)) {
1254 errfunc(ERR_WARNING
| ERR_PASS2
| ERR_WARN_NOV
,
1255 "signed byte value exceeds bounds");
1257 if (opx
->segment
!= NO_SEG
) {
1259 out(offset
, segment
, &data
, OUT_ADDRESS
, 1,
1260 opx
->segment
, opx
->wrt
);
1262 bytes
[0] = opx
->offset
;
1263 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
,
1270 if (opx
->offset
< -256 || opx
->offset
> 255) {
1271 errfunc(ERR_WARNING
| ERR_PASS2
| ERR_WARN_NOV
,
1272 "byte value exceeds bounds");
1274 if (opx
->segment
!= NO_SEG
) {
1276 out(offset
, segment
, &data
, OUT_ADDRESS
, 1,
1277 opx
->segment
, opx
->wrt
);
1279 bytes
[0] = opx
->offset
;
1280 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
,
1287 if (opx
->offset
< 0 || opx
->offset
> 255)
1288 errfunc(ERR_WARNING
| ERR_PASS2
| ERR_WARN_NOV
,
1289 "unsigned byte value exceeds bounds");
1290 if (opx
->segment
!= NO_SEG
) {
1292 out(offset
, segment
, &data
, OUT_ADDRESS
, 1,
1293 opx
->segment
, opx
->wrt
);
1295 bytes
[0] = opx
->offset
;
1296 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
,
1303 warn_overflow_opd(opx
, 2);
1305 out(offset
, segment
, &data
, OUT_ADDRESS
, 2,
1306 opx
->segment
, opx
->wrt
);
1311 if (opx
->type
& (BITS16
| BITS32
))
1312 size
= (opx
->type
& BITS16
) ? 2 : 4;
1314 size
= (bits
== 16) ? 2 : 4;
1315 warn_overflow_opd(opx
, size
);
1317 out(offset
, segment
, &data
, OUT_ADDRESS
, size
,
1318 opx
->segment
, opx
->wrt
);
1323 warn_overflow_opd(opx
, 4);
1325 out(offset
, segment
, &data
, OUT_ADDRESS
, 4,
1326 opx
->segment
, opx
->wrt
);
1332 size
= ins
->addr_size
>> 3;
1333 warn_overflow_opd(opx
, size
);
1334 out(offset
, segment
, &data
, OUT_ADDRESS
, size
,
1335 opx
->segment
, opx
->wrt
);
1340 if (opx
->segment
!= segment
) {
1342 out(offset
, segment
, &data
,
1343 OUT_REL1ADR
, insn_end
- offset
,
1344 opx
->segment
, opx
->wrt
);
1346 data
= opx
->offset
- insn_end
;
1347 if (data
> 127 || data
< -128)
1348 errfunc(ERR_NONFATAL
, "short jump is out of range");
1349 out(offset
, segment
, &data
,
1350 OUT_ADDRESS
, 1, NO_SEG
, NO_SEG
);
1356 data
= (int64_t)opx
->offset
;
1357 out(offset
, segment
, &data
, OUT_ADDRESS
, 8,
1358 opx
->segment
, opx
->wrt
);
1363 if (opx
->segment
!= segment
) {
1365 out(offset
, segment
, &data
,
1366 OUT_REL2ADR
, insn_end
- offset
,
1367 opx
->segment
, opx
->wrt
);
1369 data
= opx
->offset
- insn_end
;
1370 out(offset
, segment
, &data
,
1371 OUT_ADDRESS
, 2, NO_SEG
, NO_SEG
);
1377 if (opx
->type
& (BITS16
| BITS32
| BITS64
))
1378 size
= (opx
->type
& BITS16
) ? 2 : 4;
1380 size
= (bits
== 16) ? 2 : 4;
1381 if (opx
->segment
!= segment
) {
1383 out(offset
, segment
, &data
,
1384 size
== 2 ? OUT_REL2ADR
: OUT_REL4ADR
,
1385 insn_end
- offset
, opx
->segment
, opx
->wrt
);
1387 data
= opx
->offset
- insn_end
;
1388 out(offset
, segment
, &data
,
1389 OUT_ADDRESS
, size
, NO_SEG
, NO_SEG
);
1395 if (opx
->segment
!= segment
) {
1397 out(offset
, segment
, &data
,
1398 OUT_REL4ADR
, insn_end
- offset
,
1399 opx
->segment
, opx
->wrt
);
1401 data
= opx
->offset
- insn_end
;
1402 out(offset
, segment
, &data
,
1403 OUT_ADDRESS
, 4, NO_SEG
, NO_SEG
);
1409 if (opx
->segment
== NO_SEG
)
1410 errfunc(ERR_NONFATAL
, "value referenced by FAR is not"
1413 out(offset
, segment
, &data
, OUT_ADDRESS
, 2,
1414 outfmt
->segbase(1 + opx
->segment
),
1421 warn_overflow_opd(opx
, 2);
1422 if (is_sbyte16(opx
)) {
1424 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
,
1428 out(offset
, segment
, &data
, OUT_ADDRESS
, 2,
1429 opx
->segment
, opx
->wrt
);
1436 bytes
[0] = *codes
++;
1437 if (is_sbyte16(opx
))
1438 bytes
[0] |= 2; /* s-bit */
1439 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1445 warn_overflow_opd(opx
, 4);
1446 if (is_sbyte32(opx
)) {
1448 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
,
1452 out(offset
, segment
, &data
, OUT_ADDRESS
, 4,
1453 opx
->segment
, opx
->wrt
);
1460 bytes
[0] = *codes
++;
1461 if (is_sbyte32(opx
))
1462 bytes
[0] |= 2; /* s-bit */
1463 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1473 (ins
->drexdst
<< 4) |
1474 (ins
->rex
& REX_OC
? 0x08 : 0) |
1475 (ins
->rex
& (REX_R
|REX_X
|REX_B
));
1477 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1483 opx
= &ins
->oprs
[c
>> 3];
1484 bytes
[0] = nasm_regvals
[opx
->basereg
] << 4;
1485 opx
= &ins
->oprs
[c
& 7];
1486 if (opx
->segment
!= NO_SEG
|| opx
->wrt
!= NO_SEG
) {
1487 errfunc(ERR_NONFATAL
,
1488 "non-absolute expression not permitted as argument %d",
1491 if (opx
->offset
& ~15) {
1492 errfunc(ERR_WARNING
| ERR_PASS2
| ERR_WARN_NOV
,
1493 "four-bit argument exceeds bounds");
1495 bytes
[0] |= opx
->offset
& 15;
1497 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1503 opx
= &ins
->oprs
[c
>> 4];
1504 bytes
[0] = nasm_regvals
[opx
->basereg
] << 4;
1506 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1512 opx
= &ins
->oprs
[c
];
1513 bytes
[0] = nasm_regvals
[opx
->basereg
] << 4;
1514 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1520 if (opx
->wrt
== NO_SEG
&& opx
->segment
== NO_SEG
&&
1521 (int32_t)data
!= (int64_t)data
) {
1522 errfunc(ERR_WARNING
| ERR_PASS2
| ERR_WARN_NOV
,
1523 "signed dword immediate exceeds bounds");
1525 if (is_sbyte32(opx
)) {
1527 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
,
1531 out(offset
, segment
, &data
, OUT_ADDRESS
, 4,
1532 opx
->segment
, opx
->wrt
);
1539 if (opx
->wrt
== NO_SEG
&& opx
->segment
== NO_SEG
&&
1540 (int32_t)data
!= (int64_t)data
) {
1541 errfunc(ERR_WARNING
| ERR_PASS2
| ERR_WARN_NOV
,
1542 "signed dword immediate exceeds bounds");
1544 out(offset
, segment
, &data
, OUT_ADDRESS
, 4,
1545 opx
->segment
, opx
->wrt
);
1552 if (ins
->vex_cm
!= 1 || (ins
->rex
& (REX_W
|REX_X
|REX_B
))) {
1553 bytes
[0] = (ins
->vex_cm
>> 6) ? 0x8f : 0xc4;
1554 bytes
[1] = (ins
->vex_cm
& 31) | ((~ins
->rex
& 7) << 5);
1555 bytes
[2] = ((ins
->rex
& REX_W
) << (7-3)) |
1556 ((~ins
->drexdst
& 15)<< 3) | (ins
->vex_wlp
& 07);
1557 out(offset
, segment
, &bytes
, OUT_RAWDATA
, 3, NO_SEG
, NO_SEG
);
1561 bytes
[1] = ((~ins
->rex
& REX_R
) << (7-2)) |
1562 ((~ins
->drexdst
& 15) << 3) | (ins
->vex_wlp
& 07);
1563 out(offset
, segment
, &bytes
, OUT_RAWDATA
, 2, NO_SEG
, NO_SEG
);
1573 if (ins
->rex
& REX_W
)
1575 else if (ins
->prefixes
[PPS_OSIZE
] == P_O16
)
1577 else if (ins
->prefixes
[PPS_OSIZE
] == P_O32
)
1582 um
= (uint64_t)2 << (s
-1);
1585 if (uv
> 127 && uv
< (uint64_t)-128 &&
1586 (uv
< um
-128 || uv
> um
-1)) {
1587 errfunc(ERR_WARNING
| ERR_PASS2
| ERR_WARN_NOV
,
1588 "signed byte value exceeds bounds");
1590 if (opx
->segment
!= NO_SEG
) {
1592 out(offset
, segment
, &data
, OUT_ADDRESS
, 1,
1593 opx
->segment
, opx
->wrt
);
1596 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
,
1607 if (bits
== 32 && !has_prefix(ins
, PPS_ASIZE
, P_A16
)) {
1609 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1616 if (bits
!= 32 && !has_prefix(ins
, PPS_ASIZE
, P_A32
)) {
1618 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1637 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1646 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1664 *bytes
= *codes
++ ^ condval
[ins
->condition
];
1665 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1674 *bytes
= c
- 0332 + 0xF2;
1675 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1680 if (ins
->rex
& REX_R
) {
1682 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1685 ins
->rex
&= ~(REX_L
|REX_R
);
1696 if (ins
->oprs
[0].segment
!= NO_SEG
)
1697 errfunc(ERR_PANIC
, "non-constant BSS size in pass two");
1699 int64_t size
= ins
->oprs
[0].offset
;
1701 out(offset
, segment
, NULL
,
1702 OUT_RESERVE
, size
, NO_SEG
, NO_SEG
);
1713 switch (ins
->oprs
[0].basereg
) {
1728 "bizarre 8086 segment register received");
1730 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1737 switch (ins
->oprs
[0].basereg
) {
1746 "bizarre 386 segment register received");
1748 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1757 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1763 bytes
[0] = c
- 0362 + 0xf2;
1764 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1774 *bytes
= c
- 0366 + 0x66;
1775 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1785 *bytes
= bits
== 16 ? 3 : 5;
1786 out(offset
, segment
, bytes
, OUT_RAWDATA
, 1, NO_SEG
, NO_SEG
);
1809 struct operand
*opy
= &ins
->oprs
[op2
];
1812 /* pick rfield from operand b (opx) */
1813 rflags
= regflag(opx
);
1814 rfield
= nasm_regvals
[opx
->basereg
];
1816 /* rfield is constant */
1821 if (!process_ea(opy
, &ea_data
, bits
, ins
->addr_size
,
1823 errfunc(ERR_NONFATAL
, "invalid effective address");
1828 *p
++ = ea_data
.modrm
;
1829 if (ea_data
.sib_present
)
1832 /* DREX suffixes come between the SIB and the displacement */
1833 if (ins
->rex
& REX_D
) {
1834 *p
++ = (ins
->drexdst
<< 4) |
1835 (ins
->rex
& REX_OC
? 0x08 : 0) |
1836 (ins
->rex
& (REX_R
|REX_X
|REX_B
));
1841 out(offset
, segment
, bytes
, OUT_RAWDATA
, s
, NO_SEG
, NO_SEG
);
1844 * Make sure the address gets the right offset in case
1845 * the line breaks in the .lst file (BR 1197827)
1850 switch (ea_data
.bytes
) {
1860 if (opy
->segment
== segment
) {
1862 if (overflow_signed(data
, ea_data
.bytes
))
1863 warn_overflow(ERR_PASS2
, ea_data
.bytes
);
1864 out(offset
, segment
, &data
, OUT_ADDRESS
,
1865 ea_data
.bytes
, NO_SEG
, NO_SEG
);
1867 /* overflow check in output/linker? */
1868 out(offset
, segment
, &data
, OUT_REL4ADR
,
1869 insn_end
- offset
, opy
->segment
, opy
->wrt
);
1872 if (overflow_general(opy
->offset
, ins
->addr_size
>> 3) ||
1873 signed_bits(opy
->offset
, ins
->addr_size
) !=
1874 signed_bits(opy
->offset
, ea_data
.bytes
* 8))
1875 warn_overflow(ERR_PASS2
, ea_data
.bytes
);
1878 out(offset
, segment
, &data
, OUT_ADDRESS
,
1879 ea_data
.bytes
, opy
->segment
, opy
->wrt
);
1885 "Invalid amount of bytes (%d) for offset?!",
1894 errfunc(ERR_PANIC
, "internal instruction table corrupt"
1895 ": instruction code \\%o (0x%02X) given", c
, c
);
1901 static opflags_t
regflag(const operand
* o
)
1903 if (!is_register(o
->basereg
))
1904 errfunc(ERR_PANIC
, "invalid operand passed to regflag()");
1905 return nasm_reg_flags
[o
->basereg
];
1908 static int32_t regval(const operand
* o
)
1910 if (!is_register(o
->basereg
))
1911 errfunc(ERR_PANIC
, "invalid operand passed to regval()");
1912 return nasm_regvals
[o
->basereg
];
1915 static int op_rexflags(const operand
* o
, int mask
)
1920 if (!is_register(o
->basereg
))
1921 errfunc(ERR_PANIC
, "invalid operand passed to op_rexflags()");
1923 flags
= nasm_reg_flags
[o
->basereg
];
1924 val
= nasm_regvals
[o
->basereg
];
1926 return rexflags(val
, flags
, mask
);
1929 static int rexflags(int val
, opflags_t flags
, int mask
)
1934 rex
|= REX_B
|REX_X
|REX_R
;
1937 if (!(REG_HIGH
& ~flags
)) /* AH, CH, DH, BH */
1939 else if (!(REG8
& ~flags
) && val
>= 4) /* SPL, BPL, SIL, DIL */
1945 static enum match_result
find_match(const struct itemplate
**tempp
,
1947 int32_t segment
, int64_t offset
, int bits
)
1949 const struct itemplate
*temp
;
1950 enum match_result m
, merr
;
1951 opflags_t xsizeflags
[MAX_OPERANDS
];
1952 bool opsizemissing
= false;
1955 for (i
= 0; i
< instruction
->operands
; i
++)
1956 xsizeflags
[i
] = instruction
->oprs
[i
].type
& SIZE_MASK
;
1958 merr
= MERR_INVALOP
;
1960 for (temp
= nasm_instructions
[instruction
->opcode
];
1961 temp
->opcode
!= I_none
; temp
++) {
1962 m
= matches(temp
, instruction
, bits
);
1963 if (m
== MOK_JUMP
) {
1964 if (jmp_match(segment
, offset
, bits
, instruction
, temp
->code
))
1968 } else if (m
== MERR_OPSIZEMISSING
&&
1969 (temp
->flags
& IF_SMASK
) != IF_SX
) {
1971 * Missing operand size and a candidate for fuzzy matching...
1973 for (i
= 0; i
< temp
->operands
; i
++) {
1974 if ((temp
->opd
[i
] & SAME_AS
) == 0)
1975 xsizeflags
[i
] |= temp
->opd
[i
] & SIZE_MASK
;
1977 opsizemissing
= true;
1981 if (merr
== MOK_GOOD
)
1985 /* No match, but see if we can get a fuzzy operand size match... */
1989 for (i
= 0; i
< instruction
->operands
; i
++) {
1991 * We ignore extrinsic operand sizes on registers, so we should
1992 * never try to fuzzy-match on them. This also resolves the case
1993 * when we have e.g. "xmmrm128" in two different positions.
1995 if (is_class(REGISTER
, instruction
->oprs
[i
].type
))
1998 /* This tests if xsizeflags[i] has more than one bit set */
1999 if ((xsizeflags
[i
] & (xsizeflags
[i
]-1)))
2000 goto done
; /* No luck */
2002 instruction
->oprs
[i
].type
|= xsizeflags
[i
]; /* Set the size */
2005 /* Try matching again... */
2006 for (temp
= nasm_instructions
[instruction
->opcode
];
2007 temp
->opcode
!= I_none
; temp
++) {
2008 m
= matches(temp
, instruction
, bits
);
2009 if (m
== MOK_JUMP
) {
2010 if (jmp_match(segment
, offset
, bits
, instruction
, temp
->code
))
2017 if (merr
== MOK_GOOD
)
2026 static enum match_result
matches(const struct itemplate
*itemp
,
2027 insn
*instruction
, int bits
)
2029 int i
, size
[MAX_OPERANDS
], asize
, oprs
;
2030 bool opsizemissing
= false;
2035 if (itemp
->opcode
!= instruction
->opcode
)
2036 return MERR_INVALOP
;
2039 * Count the operands
2041 if (itemp
->operands
!= instruction
->operands
)
2042 return MERR_INVALOP
;
2045 * Check that no spurious colons or TOs are present
2047 for (i
= 0; i
< itemp
->operands
; i
++)
2048 if (instruction
->oprs
[i
].type
& ~itemp
->opd
[i
] & (COLON
| TO
))
2049 return MERR_INVALOP
;
2052 * Process size flags
2054 switch (itemp
->flags
& IF_SMASK
) {
2094 if (itemp
->flags
& IF_ARMASK
) {
2095 /* S- flags only apply to a specific operand */
2096 i
= ((itemp
->flags
& IF_ARMASK
) >> IF_ARSHFT
) - 1;
2097 memset(size
, 0, sizeof size
);
2100 /* S- flags apply to all operands */
2101 for (i
= 0; i
< MAX_OPERANDS
; i
++)
2106 * Check that the operand flags all match up,
2107 * it's a bit tricky so lets be verbose:
2109 * 1) Find out the size of operand. If instruction
2110 * doesn't have one specified -- we're trying to
2111 * guess it either from template (IF_S* flag) or
2114 * 2) If template operand (i) has SAME_AS flag [used for registers only]
2115 * (ie the same operand as was specified somewhere in template, and
2116 * this referred operand index is being achieved via ~SAME_AS)
2117 * we are to be sure that both registers (in template and instruction)
2120 * 3) If template operand do not match the instruction OR
2121 * template has an operand size specified AND this size differ
2122 * from which instruction has (perhaps we got it from code bits)
2124 * a) Check that only size of instruction and operand is differ
2125 * other characteristics do match
2126 * b) Perhaps it's a register specified in instruction so
2127 * for such a case we just mark that operand as "size
2128 * missing" and this will turn on fuzzy operand size
2129 * logic facility (handled by a caller)
2131 for (i
= 0; i
< itemp
->operands
; i
++) {
2132 opflags_t type
= instruction
->oprs
[i
].type
;
2133 if (!(type
& SIZE_MASK
))
2136 if (itemp
->opd
[i
] & SAME_AS
) {
2137 int j
= itemp
->opd
[i
] & ~SAME_AS
;
2138 if (type
!= instruction
->oprs
[j
].type
||
2139 instruction
->oprs
[i
].basereg
!= instruction
->oprs
[j
].basereg
)
2140 return MERR_INVALOP
;
2141 } else if (itemp
->opd
[i
] & ~type
||
2142 ((itemp
->opd
[i
] & SIZE_MASK
) &&
2143 ((itemp
->opd
[i
] ^ type
) & SIZE_MASK
))) {
2144 if ((itemp
->opd
[i
] & ~type
& ~SIZE_MASK
) || (type
& SIZE_MASK
)) {
2145 return MERR_INVALOP
;
2146 } else if (!is_class(REGISTER
, type
)) {
2148 * Note: we don't honor extrinsic operand sizes for registers,
2149 * so "missing operand size" for a register should be
2150 * considered a wildcard match rather than an error.
2152 opsizemissing
= true;
2158 return MERR_OPSIZEMISSING
;
2161 * Check operand sizes
2163 if (itemp
->flags
& (IF_SM
| IF_SM2
)) {
2164 oprs
= (itemp
->flags
& IF_SM2
? 2 : itemp
->operands
);
2165 for (i
= 0; i
< oprs
; i
++) {
2166 asize
= itemp
->opd
[i
] & SIZE_MASK
;
2168 for (i
= 0; i
< oprs
; i
++)
2174 oprs
= itemp
->operands
;
2177 for (i
= 0; i
< itemp
->operands
; i
++) {
2178 if (!(itemp
->opd
[i
] & SIZE_MASK
) &&
2179 (instruction
->oprs
[i
].type
& SIZE_MASK
& ~size
[i
]))
2180 return MERR_OPSIZEMISMATCH
;
2184 * Check template is okay at the set cpu level
2186 if (((itemp
->flags
& IF_PLEVEL
) > cpu
))
2190 * Verify the appropriate long mode flag.
2192 if ((itemp
->flags
& (bits
== 64 ? IF_NOLONG
: IF_LONG
)))
2193 return MERR_BADMODE
;
2196 * Check if special handling needed for Jumps
2198 if ((itemp
->code
[0] & 0374) == 0370)
2204 static ea
*process_ea(operand
* input
, ea
* output
, int bits
,
2205 int addrbits
, int rfield
, opflags_t rflags
)
2207 bool forw_ref
= !!(input
->opflags
& OPFLAG_UNKNOWN
);
2209 output
->rip
= false;
2211 /* REX flags for the rfield operand */
2212 output
->rex
|= rexflags(rfield
, rflags
, REX_R
| REX_P
| REX_W
| REX_H
);
2214 if (is_class(REGISTER
, input
->type
)) { /* register direct */
2218 if (!is_register(input
->basereg
))
2221 i
= nasm_regvals
[input
->basereg
];
2224 return NULL
; /* Invalid EA register */
2226 output
->rex
|= op_rexflags(input
, REX_B
| REX_P
| REX_W
| REX_H
);
2228 output
->sib_present
= false; /* no SIB necessary */
2229 output
->bytes
= 0; /* no offset necessary either */
2230 output
->modrm
= 0xC0 | ((rfield
& 7) << 3) | (i
& 7);
2231 } else { /* it's a memory reference */
2232 if (input
->basereg
== -1 &&
2233 (input
->indexreg
== -1 || input
->scale
== 0)) {
2234 /* it's a pure offset */
2236 if (bits
== 64 && ((input
->type
& IP_REL
) == IP_REL
) &&
2237 input
->segment
== NO_SEG
) {
2238 nasm_error(ERR_WARNING
| ERR_PASS1
, "absolute address can not be RIP-relative");
2239 input
->type
&= ~IP_REL
;
2240 input
->type
|= MEMORY
;
2243 if (input
->eaflags
& EAF_BYTEOFFS
||
2244 (input
->eaflags
& EAF_WORDOFFS
&&
2245 input
->disp_size
!= (addrbits
!= 16 ? 32 : 16))) {
2246 nasm_error(ERR_WARNING
| ERR_PASS1
, "displacement size ignored on absolute address");
2249 if (bits
== 64 && (~input
->type
& IP_REL
)) {
2250 int scale
, index
, base
;
2251 output
->sib_present
= true;
2255 output
->sib
= (scale
<< 6) | (index
<< 3) | base
;
2257 output
->modrm
= 4 | ((rfield
& 7) << 3);
2258 output
->rip
= false;
2260 output
->sib_present
= false;
2261 output
->bytes
= (addrbits
!= 16 ? 4 : 2);
2262 output
->modrm
= (addrbits
!= 16 ? 5 : 6) | ((rfield
& 7) << 3);
2263 output
->rip
= bits
== 64;
2265 } else { /* it's an indirection */
2266 int i
= input
->indexreg
, b
= input
->basereg
, s
= input
->scale
;
2267 int32_t seg
= input
->segment
;
2268 int hb
= input
->hintbase
, ht
= input
->hinttype
;
2269 int t
, it
, bt
; /* register numbers */
2270 opflags_t x
, ix
, bx
; /* register flags */
2273 i
= -1; /* make this easy, at least */
2275 if (is_register(i
)) {
2276 it
= nasm_regvals
[i
];
2277 ix
= nasm_reg_flags
[i
];
2283 if (is_register(b
)) {
2284 bt
= nasm_regvals
[b
];
2285 bx
= nasm_reg_flags
[b
];
2291 /* check for a 32/64-bit memory reference... */
2292 if ((ix
|bx
) & (BITS32
|BITS64
)) {
2294 * it must be a 32/64-bit memory reference. Firstly we have
2295 * to check that all registers involved are type E/Rxx.
2297 int32_t sok
= BITS32
| BITS64
, o
= input
->offset
;
2300 if (!(REG64
& ~ix
) || !(REG32
& ~ix
))
2308 return NULL
; /* Invalid register */
2309 if (~sok
& bx
& SIZE_MASK
)
2310 return NULL
; /* Invalid size */
2315 * While we're here, ensure the user didn't specify
2318 if (input
->disp_size
== 16 || input
->disp_size
== 64)
2321 if (addrbits
== 16 ||
2322 (addrbits
== 32 && !(sok
& BITS32
)) ||
2323 (addrbits
== 64 && !(sok
& BITS64
)))
2326 /* now reorganize base/index */
2327 if (s
== 1 && bt
!= it
&& bt
!= -1 && it
!= -1 &&
2328 ((hb
== b
&& ht
== EAH_NOTBASE
) ||
2329 (hb
== i
&& ht
== EAH_MAKEBASE
))) {
2330 /* swap if hints say so */
2331 t
= bt
, bt
= it
, it
= t
;
2332 x
= bx
, bx
= ix
, ix
= x
;
2334 if (bt
== it
) /* convert EAX+2*EAX to 3*EAX */
2335 bt
= -1, bx
= 0, s
++;
2336 if (bt
== -1 && s
== 1 && !(hb
== it
&& ht
== EAH_NOTBASE
)) {
2337 /* make single reg base, unless hint */
2338 bt
= it
, bx
= ix
, it
= -1, ix
= 0;
2340 if (((s
== 2 && it
!= REG_NUM_ESP
&& !(input
->eaflags
& EAF_TIMESTWO
)) ||
2341 s
== 3 || s
== 5 || s
== 9) && bt
== -1)
2342 bt
= it
, bx
= ix
, s
--; /* convert 3*EAX to EAX+2*EAX */
2343 if (it
== -1 && (bt
& 7) != REG_NUM_ESP
&&
2344 (input
->eaflags
& EAF_TIMESTWO
))
2345 it
= bt
, ix
= bx
, bt
= -1, bx
= 0, s
= 1;
2346 /* convert [NOSPLIT EAX] to sib format with 0x0 displacement */
2347 if (s
== 1 && it
== REG_NUM_ESP
) {
2348 /* swap ESP into base if scale is 1 */
2349 t
= it
, it
= bt
, bt
= t
;
2350 x
= ix
, ix
= bx
, bx
= x
;
2352 if (it
== REG_NUM_ESP
||
2353 (s
!= 1 && s
!= 2 && s
!= 4 && s
!= 8 && it
!= -1))
2354 return NULL
; /* wrong, for various reasons */
2356 output
->rex
|= rexflags(it
, ix
, REX_X
);
2357 output
->rex
|= rexflags(bt
, bx
, REX_B
);
2359 if (it
== -1 && (bt
& 7) != REG_NUM_ESP
) {
2368 if (rm
!= REG_NUM_EBP
&& o
== 0 &&
2369 seg
== NO_SEG
&& !forw_ref
&&
2370 !(input
->eaflags
& (EAF_BYTEOFFS
| EAF_WORDOFFS
)))
2372 else if (input
->eaflags
& EAF_BYTEOFFS
||
2373 (o
>= -128 && o
<= 127 &&
2374 seg
== NO_SEG
&& !forw_ref
&&
2375 !(input
->eaflags
& EAF_WORDOFFS
)))
2381 output
->sib_present
= false;
2382 output
->bytes
= (bt
== -1 || mod
== 2 ? 4 : mod
);
2383 output
->modrm
= (mod
<< 6) | ((rfield
& 7) << 3) | rm
;
2386 int mod
, scale
, index
, base
;
2406 default: /* then what the smeg is it? */
2407 return NULL
; /* panic */
2415 if (base
!= REG_NUM_EBP
&& o
== 0 &&
2416 seg
== NO_SEG
&& !forw_ref
&&
2417 !(input
->eaflags
& (EAF_BYTEOFFS
| EAF_WORDOFFS
)))
2419 else if (input
->eaflags
& EAF_BYTEOFFS
||
2420 (o
>= -128 && o
<= 127 &&
2421 seg
== NO_SEG
&& !forw_ref
&&
2422 !(input
->eaflags
& EAF_WORDOFFS
)))
2428 output
->sib_present
= true;
2429 output
->bytes
= (bt
== -1 || mod
== 2 ? 4 : mod
);
2430 output
->modrm
= (mod
<< 6) | ((rfield
& 7) << 3) | 4;
2431 output
->sib
= (scale
<< 6) | (index
<< 3) | base
;
2433 } else { /* it's 16-bit */
2435 int16_t o
= input
->offset
;
2437 /* check for 64-bit long mode */
2441 /* check all registers are BX, BP, SI or DI */
2442 if ((b
!= -1 && b
!= R_BP
&& b
!= R_BX
&& b
!= R_SI
&& b
!= R_DI
) ||
2443 (i
!= -1 && i
!= R_BP
&& i
!= R_BX
&& i
!= R_SI
&& i
!= R_DI
))
2446 /* ensure the user didn't specify DWORD/QWORD */
2447 if (input
->disp_size
== 32 || input
->disp_size
== 64)
2450 if (s
!= 1 && i
!= -1)
2451 return NULL
; /* no can do, in 16-bit EA */
2452 if (b
== -1 && i
!= -1) {
2457 if ((b
== R_SI
|| b
== R_DI
) && i
!= -1) {
2462 /* have BX/BP as base, SI/DI index */
2464 return NULL
; /* shouldn't ever happen, in theory */
2465 if (i
!= -1 && b
!= -1 &&
2466 (i
== R_BP
|| i
== R_BX
|| b
== R_SI
|| b
== R_DI
))
2467 return NULL
; /* invalid combinations */
2468 if (b
== -1) /* pure offset: handled above */
2469 return NULL
; /* so if it gets to here, panic! */
2473 switch (i
* 256 + b
) {
2474 case R_SI
* 256 + R_BX
:
2477 case R_DI
* 256 + R_BX
:
2480 case R_SI
* 256 + R_BP
:
2483 case R_DI
* 256 + R_BP
:
2501 if (rm
== -1) /* can't happen, in theory */
2502 return NULL
; /* so panic if it does */
2504 if (o
== 0 && seg
== NO_SEG
&& !forw_ref
&& rm
!= 6 &&
2505 !(input
->eaflags
& (EAF_BYTEOFFS
| EAF_WORDOFFS
)))
2507 else if (input
->eaflags
& EAF_BYTEOFFS
||
2508 (o
>= -128 && o
<= 127 && seg
== NO_SEG
&&
2509 !forw_ref
&& !(input
->eaflags
& EAF_WORDOFFS
)))
2514 output
->sib_present
= false; /* no SIB - it's 16-bit */
2515 output
->bytes
= mod
; /* bytes of offset needed */
2516 output
->modrm
= (mod
<< 6) | ((rfield
& 7) << 3) | rm
;
2521 output
->size
= 1 + output
->sib_present
+ output
->bytes
;
2525 static void add_asp(insn
*ins
, int addrbits
)
2530 valid
= (addrbits
== 64) ? 64|32 : 32|16;
2532 switch (ins
->prefixes
[PPS_ASIZE
]) {
2543 valid
&= (addrbits
== 32) ? 16 : 32;
2549 for (j
= 0; j
< ins
->operands
; j
++) {
2550 if (is_class(MEMORY
, ins
->oprs
[j
].type
)) {
2553 /* Verify as Register */
2554 if (!is_register(ins
->oprs
[j
].indexreg
))
2557 i
= nasm_reg_flags
[ins
->oprs
[j
].indexreg
];
2559 /* Verify as Register */
2560 if (!is_register(ins
->oprs
[j
].basereg
))
2563 b
= nasm_reg_flags
[ins
->oprs
[j
].basereg
];
2565 if (ins
->oprs
[j
].scale
== 0)
2569 int ds
= ins
->oprs
[j
].disp_size
;
2570 if ((addrbits
!= 64 && ds
> 8) ||
2571 (addrbits
== 64 && ds
== 16))
2591 if (valid
& addrbits
) {
2592 ins
->addr_size
= addrbits
;
2593 } else if (valid
& ((addrbits
== 32) ? 16 : 32)) {
2594 /* Add an address size prefix */
2595 enum prefixes pref
= (addrbits
== 32) ? P_A16
: P_A32
;
2596 ins
->prefixes
[PPS_ASIZE
] = pref
;
2597 ins
->addr_size
= (addrbits
== 32) ? 16 : 32;
2600 errfunc(ERR_NONFATAL
, "impossible combination of address sizes");
2601 ins
->addr_size
= addrbits
; /* Error recovery */
2604 defdisp
= ins
->addr_size
== 16 ? 16 : 32;
2606 for (j
= 0; j
< ins
->operands
; j
++) {
2607 if (!(MEM_OFFS
& ~ins
->oprs
[j
].type
) &&
2608 (ins
->oprs
[j
].disp_size
? ins
->oprs
[j
].disp_size
: defdisp
) != ins
->addr_size
) {
2610 * mem_offs sizes must match the address size; if not,
2611 * strip the MEM_OFFS bit and match only EA instructions
2613 ins
->oprs
[j
].type
&= ~(MEM_OFFS
& ~MEMORY
);