]> git.ipfire.org Git - thirdparty/binutils-gdb.git/blob - opcodes/or1k-asm.c
opcodes: add a .gitattributes file for aarch64 autogenerated file exceptions
[thirdparty/binutils-gdb.git] / opcodes / or1k-asm.c
1 /* DO NOT EDIT! -*- buffer-read-only: t -*- vi:set ro: */
2 /* Assembler interface for targets using CGEN. -*- C -*-
3 CGEN: Cpu tools GENerator
4
5 THIS FILE IS MACHINE GENERATED WITH CGEN.
6 - the resultant file is machine generated, cgen-asm.in isn't
7
8 Copyright (C) 1996-2024 Free Software Foundation, Inc.
9
10 This file is part of libopcodes.
11
12 This library is free software; you can redistribute it and/or modify
13 it under the terms of the GNU General Public License as published by
14 the Free Software Foundation; either version 3, or (at your option)
15 any later version.
16
17 It is distributed in the hope that it will be useful, but WITHOUT
18 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
19 or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public
20 License for more details.
21
22 You should have received a copy of the GNU General Public License
23 along with this program; if not, write to the Free Software Foundation, Inc.,
24 51 Franklin Street - Fifth Floor, Boston, MA 02110-1301, USA. */
25
26
27 /* ??? Eventually more and more of this stuff can go to cpu-independent files.
28 Keep that in mind. */
29
30 #include "sysdep.h"
31 #include <stdio.h>
32 #include "ansidecl.h"
33 #include "bfd.h"
34 #include "symcat.h"
35 #include "or1k-desc.h"
36 #include "or1k-opc.h"
37 #include "opintl.h"
38 #include "xregex.h"
39 #include "libiberty.h"
40 #include "safe-ctype.h"
41
42 #undef min
43 #define min(a,b) ((a) < (b) ? (a) : (b))
44 #undef max
45 #define max(a,b) ((a) > (b) ? (a) : (b))
46
47 static const char * parse_insn_normal
48 (CGEN_CPU_DESC, const CGEN_INSN *, const char **, CGEN_FIELDS *);
49 \f
50 /* -- assembler routines inserted here. */
51
52 /* -- asm.c */
53
54 static const char * MISSING_CLOSING_PARENTHESIS = N_("missing `)'");
55 static const char * INVALID_STORE_RELOC = N_("relocation invalid for store");
56 static const char * INVALID_RELOC_TYPE = N_("internal relocation type invalid");
57
58 #define CGEN_VERBOSE_ASSEMBLER_ERRORS
59
60 static const char *
61 parse_disp26 (CGEN_CPU_DESC cd,
62 const char ** strp,
63 int opindex,
64 int opinfo ATTRIBUTE_UNUSED,
65 enum cgen_parse_operand_result * resultp,
66 bfd_vma * valuep)
67 {
68 const char *str = *strp;
69 const char *errmsg = NULL;
70 bfd_reloc_code_real_type reloc = BFD_RELOC_OR1K_REL_26;
71
72 if (strncasecmp (str, "plta(", 5) == 0)
73 {
74 *strp = str + 5;
75 reloc = BFD_RELOC_OR1K_PLTA26;
76 }
77 else if (strncasecmp (str, "plt(", 4) == 0)
78 {
79 *strp = str + 4;
80 reloc = BFD_RELOC_OR1K_PLT26;
81 }
82
83 errmsg = cgen_parse_address (cd, strp, opindex, reloc, resultp, valuep);
84
85 if (reloc != BFD_RELOC_OR1K_REL_26)
86 {
87 if (**strp != ')')
88 errmsg = MISSING_CLOSING_PARENTHESIS;
89 else
90 ++*strp;
91 }
92
93 return errmsg;
94 }
95
96 static const char *
97 parse_disp21 (CGEN_CPU_DESC cd,
98 const char ** strp,
99 int opindex,
100 int opinfo ATTRIBUTE_UNUSED,
101 enum cgen_parse_operand_result * resultp,
102 bfd_vma * valuep)
103 {
104 const char *str = *strp;
105 const char *errmsg = NULL;
106 bfd_reloc_code_real_type reloc = BFD_RELOC_OR1K_PCREL_PG21;
107
108 if (strncasecmp (str, "got(", 4) == 0)
109 {
110 *strp = str + 4;
111 reloc = BFD_RELOC_OR1K_GOT_PG21;
112 }
113 else if (strncasecmp (str, "tlsgd(", 6) == 0)
114 {
115 *strp = str + 6;
116 reloc = BFD_RELOC_OR1K_TLS_GD_PG21;
117 }
118 else if (strncasecmp (str, "tlsldm(", 7) == 0)
119 {
120 *strp = str + 7;
121 reloc = BFD_RELOC_OR1K_TLS_LDM_PG21;
122 }
123 else if (strncasecmp (str, "gottp(", 6) == 0)
124 {
125 *strp = str + 6;
126 reloc = BFD_RELOC_OR1K_TLS_IE_PG21;
127 }
128
129 errmsg = cgen_parse_address (cd, strp, opindex, reloc, resultp, valuep);
130
131 if (reloc != BFD_RELOC_OR1K_PCREL_PG21)
132 {
133 if (**strp != ')')
134 errmsg = MISSING_CLOSING_PARENTHESIS;
135 else
136 ++*strp;
137 }
138
139 return errmsg;
140 }
141
142 enum or1k_rclass
143 {
144 RCLASS_DIRECT = 0,
145 RCLASS_GOT = 1,
146 RCLASS_GOTPC = 2,
147 RCLASS_GOTOFF = 3,
148 RCLASS_TLSGD = 4,
149 RCLASS_TLSLDM = 5,
150 RCLASS_DTPOFF = 6,
151 RCLASS_GOTTPOFF = 7,
152 RCLASS_TPOFF = 8,
153 };
154
155 enum or1k_rtype
156 {
157 RTYPE_LO = 0,
158 RTYPE_SLO = 1,
159 RTYPE_PO = 2,
160 RTYPE_SPO = 3,
161 RTYPE_HI = 4,
162 RTYPE_AHI = 5,
163 };
164
165 #define RCLASS_SHIFT 3
166 #define RTYPE_MASK 7
167
168 static const bfd_reloc_code_real_type or1k_imm16_relocs[][6] = {
169 { BFD_RELOC_LO16,
170 BFD_RELOC_OR1K_SLO16,
171 BFD_RELOC_OR1K_LO13,
172 BFD_RELOC_OR1K_SLO13,
173 BFD_RELOC_HI16,
174 BFD_RELOC_HI16_S, },
175 { BFD_RELOC_OR1K_GOT16,
176 BFD_RELOC_UNUSED,
177 BFD_RELOC_OR1K_GOT_LO13,
178 BFD_RELOC_UNUSED,
179 BFD_RELOC_UNUSED,
180 BFD_RELOC_OR1K_GOT_AHI16 },
181 { BFD_RELOC_OR1K_GOTPC_LO16,
182 BFD_RELOC_UNUSED,
183 BFD_RELOC_UNUSED,
184 BFD_RELOC_UNUSED,
185 BFD_RELOC_OR1K_GOTPC_HI16,
186 BFD_RELOC_UNUSED },
187 { BFD_RELOC_LO16_GOTOFF,
188 BFD_RELOC_OR1K_GOTOFF_SLO16,
189 BFD_RELOC_UNUSED,
190 BFD_RELOC_UNUSED,
191 BFD_RELOC_HI16_GOTOFF,
192 BFD_RELOC_HI16_S_GOTOFF },
193 { BFD_RELOC_OR1K_TLS_GD_LO16,
194 BFD_RELOC_UNUSED,
195 BFD_RELOC_OR1K_TLS_GD_LO13,
196 BFD_RELOC_UNUSED,
197 BFD_RELOC_OR1K_TLS_GD_HI16,
198 BFD_RELOC_UNUSED },
199 { BFD_RELOC_OR1K_TLS_LDM_LO16,
200 BFD_RELOC_UNUSED,
201 BFD_RELOC_OR1K_TLS_LDM_LO13,
202 BFD_RELOC_UNUSED,
203 BFD_RELOC_OR1K_TLS_LDM_HI16,
204 BFD_RELOC_UNUSED },
205 { BFD_RELOC_OR1K_TLS_LDO_LO16,
206 BFD_RELOC_UNUSED,
207 BFD_RELOC_UNUSED,
208 BFD_RELOC_UNUSED,
209 BFD_RELOC_OR1K_TLS_LDO_HI16,
210 BFD_RELOC_UNUSED },
211 { BFD_RELOC_OR1K_TLS_IE_LO16,
212 BFD_RELOC_UNUSED,
213 BFD_RELOC_OR1K_TLS_IE_LO13,
214 BFD_RELOC_UNUSED,
215 BFD_RELOC_OR1K_TLS_IE_HI16,
216 BFD_RELOC_OR1K_TLS_IE_AHI16 },
217 { BFD_RELOC_OR1K_TLS_LE_LO16,
218 BFD_RELOC_OR1K_TLS_LE_SLO16,
219 BFD_RELOC_UNUSED,
220 BFD_RELOC_UNUSED,
221 BFD_RELOC_OR1K_TLS_LE_HI16,
222 BFD_RELOC_OR1K_TLS_LE_AHI16 },
223 };
224
225 static int
226 parse_reloc (const char **strp)
227 {
228 const char *str = *strp;
229 enum or1k_rclass cls = RCLASS_DIRECT;
230 enum or1k_rtype typ;
231
232 if (strncasecmp (str, "got(", 4) == 0)
233 {
234 *strp = str + 4;
235 return (RCLASS_GOT << RCLASS_SHIFT) | RTYPE_LO;
236 }
237 if (strncasecmp (str, "gotpo(", 6) == 0)
238 {
239 *strp = str + 6;
240 return (RCLASS_GOT << RCLASS_SHIFT) | RTYPE_PO;
241 }
242 if (strncasecmp (str, "gottppo(", 8) == 0)
243 {
244 *strp = str + 8;
245 return (RCLASS_GOTTPOFF << RCLASS_SHIFT) | RTYPE_PO;
246 }
247
248 if (strncasecmp (str, "gotpc", 5) == 0)
249 {
250 str += 5;
251 cls = RCLASS_GOTPC;
252 }
253 else if (strncasecmp (str, "gotoff", 6) == 0)
254 {
255 str += 6;
256 cls = RCLASS_GOTOFF;
257 }
258 else if (strncasecmp (str, "tlsgd", 5) == 0)
259 {
260 str += 5;
261 cls = RCLASS_TLSGD;
262 }
263 else if (strncasecmp (str, "tlsldm", 6) == 0)
264 {
265 str += 6;
266 cls = RCLASS_TLSLDM;
267 }
268 else if (strncasecmp (str, "dtpoff", 6) == 0)
269 {
270 str += 6;
271 cls = RCLASS_DTPOFF;
272 }
273 else if (strncasecmp (str, "gottpoff", 8) == 0)
274 {
275 str += 8;
276 cls = RCLASS_GOTTPOFF;
277 }
278 else if (strncasecmp (str, "tpoff", 5) == 0)
279 {
280 str += 5;
281 cls = RCLASS_TPOFF;
282 }
283 else if (strncasecmp (str, "got", 3) == 0)
284 {
285 str += 3;
286 cls = RCLASS_GOT;
287 }
288
289 if (strncasecmp (str, "hi(", 3) == 0)
290 {
291 str += 3;
292 typ = RTYPE_HI;
293 }
294 else if (strncasecmp (str, "lo(", 3) == 0)
295 {
296 str += 3;
297 typ = RTYPE_LO;
298 }
299 else if (strncasecmp (str, "ha(", 3) == 0)
300 {
301 str += 3;
302 typ = RTYPE_AHI;
303 }
304 else if (strncasecmp (str, "po(", 3) == 0 && cls != RCLASS_GOTTPOFF)
305 {
306 str += 3;
307 typ = RTYPE_PO;
308 }
309 else
310 return -1;
311
312 *strp = str;
313 return (cls << RCLASS_SHIFT) | typ;
314 }
315
316 static const char *
317 parse_imm16 (CGEN_CPU_DESC cd, const char **strp, int opindex,
318 long *valuep, int splitp)
319 {
320 const char *errmsg;
321 enum cgen_parse_operand_result result_type;
322 bfd_reloc_code_real_type reloc = BFD_RELOC_UNUSED;
323 enum or1k_rtype reloc_type;
324 int reloc_code;
325 bfd_vma ret;
326
327 if (**strp == '#')
328 ++*strp;
329
330 reloc_code = parse_reloc (strp);
331 reloc_type = reloc_code & RTYPE_MASK;
332 if (reloc_code >= 0)
333 {
334 enum or1k_rclass reloc_class = reloc_code >> RCLASS_SHIFT;
335 if (splitp)
336 {
337 if ((reloc_type == RTYPE_LO || reloc_type == RTYPE_PO)
338 && reloc_class != RCLASS_GOT)
339 /* If split we or up the type to RTYPE_SLO or RTYPE_SPO. */
340 reloc_type |= 1;
341 else
342 return INVALID_STORE_RELOC;
343 }
344 reloc = or1k_imm16_relocs[reloc_class][reloc_type];
345 }
346
347 if (reloc != BFD_RELOC_UNUSED)
348 {
349 bfd_vma value;
350
351 errmsg = cgen_parse_address (cd, strp, opindex, reloc,
352 &result_type, &value);
353 if (**strp != ')')
354 errmsg = MISSING_CLOSING_PARENTHESIS;
355 ++*strp;
356
357 ret = value;
358
359 if (errmsg == NULL && result_type == CGEN_PARSE_OPERAND_RESULT_NUMBER)
360 switch (reloc_type)
361 {
362 case RTYPE_AHI:
363 ret += 0x8000;
364 /* FALLTHRU */
365 case RTYPE_HI:
366 ret >>= 16;
367 /* FALLTHRU */
368 case RTYPE_LO:
369 case RTYPE_SLO:
370 ret &= 0xffff;
371 ret = (ret ^ 0x8000) - 0x8000;
372 break;
373 case RTYPE_PO:
374 case RTYPE_SPO:
375 ret &= 0x1fff;
376 break;
377 default:
378 errmsg = INVALID_RELOC_TYPE;
379 }
380 }
381 else
382 {
383 long value;
384 errmsg = cgen_parse_signed_integer (cd, strp, opindex, &value);
385 ret = value;
386 }
387
388 if (errmsg == NULL)
389 *valuep = ret;
390
391 return errmsg;
392 }
393
394 static const char *
395 parse_simm16 (CGEN_CPU_DESC cd, const char **strp, int opindex, long *valuep)
396 {
397 return parse_imm16(cd, strp, opindex, (long *) valuep, 0);
398 }
399
400 static const char *
401 parse_simm16_split (CGEN_CPU_DESC cd, const char **strp, int opindex,
402 long *valuep)
403 {
404 return parse_imm16(cd, strp, opindex, (long *) valuep, 1);
405 }
406
407 static const char *
408 parse_uimm16 (CGEN_CPU_DESC cd, const char **strp, int opindex,
409 unsigned long *valuep)
410 {
411 const char *errmsg = parse_imm16(cd, strp, opindex, (long *) valuep, 0);
412 if (errmsg == NULL)
413 *valuep &= 0xffff;
414 return errmsg;
415 }
416
417 static const char *
418 parse_uimm16_split (CGEN_CPU_DESC cd, const char **strp, int opindex,
419 unsigned long *valuep)
420 {
421 const char *errmsg = parse_imm16(cd, strp, opindex, (long *) valuep, 1);
422 if (errmsg == NULL)
423 *valuep &= 0xffff;
424 return errmsg;
425 }
426
427 /* Parse register pairs with syntax rA,rB to a flag + rA value. */
428
429 static const char *
430 parse_regpair (CGEN_CPU_DESC cd, const char **strp,
431 int opindex ATTRIBUTE_UNUSED, unsigned long *valuep)
432 {
433 long reg1_index;
434 long reg2_index;
435 const char *errmsg;
436
437 /* The first part should just be a register. */
438 errmsg = cgen_parse_keyword (cd, strp, &or1k_cgen_opval_h_gpr,
439 &reg1_index);
440
441 /* If that worked skip the comma separator. */
442 if (errmsg == NULL)
443 {
444 if (**strp == ',')
445 ++*strp;
446 else
447 errmsg = "Unexpected character, expected ','";
448 }
449
450 /* If that worked the next part is just another register. */
451 if (errmsg == NULL)
452 errmsg = cgen_parse_keyword (cd, strp, &or1k_cgen_opval_h_gpr,
453 &reg2_index);
454
455 /* Validate the register pair is valid and create the output value. */
456 if (errmsg == NULL)
457 {
458 int regoffset = reg2_index - reg1_index;
459
460 if (regoffset == 1 || regoffset == 2)
461 {
462 unsigned short offsetmask;
463 unsigned short value;
464
465 offsetmask = ((regoffset == 2 ? 1 : 0) << 5);
466 value = offsetmask | reg1_index;
467
468 *valuep = value;
469 }
470 else
471 errmsg = "Invalid register pair, offset not 1 or 2.";
472 }
473
474 return errmsg;
475 }
476
477 /* -- */
478
479 const char * or1k_cgen_parse_operand
480 (CGEN_CPU_DESC, int, const char **, CGEN_FIELDS *);
481
482 /* Main entry point for operand parsing.
483
484 This function is basically just a big switch statement. Earlier versions
485 used tables to look up the function to use, but
486 - if the table contains both assembler and disassembler functions then
487 the disassembler contains much of the assembler and vice-versa,
488 - there's a lot of inlining possibilities as things grow,
489 - using a switch statement avoids the function call overhead.
490
491 This function could be moved into `parse_insn_normal', but keeping it
492 separate makes clear the interface between `parse_insn_normal' and each of
493 the handlers. */
494
495 const char *
496 or1k_cgen_parse_operand (CGEN_CPU_DESC cd,
497 int opindex,
498 const char ** strp,
499 CGEN_FIELDS * fields)
500 {
501 const char * errmsg = NULL;
502 /* Used by scalar operands that still need to be parsed. */
503 long junk ATTRIBUTE_UNUSED;
504
505 switch (opindex)
506 {
507 case OR1K_OPERAND_DISP21 :
508 {
509 bfd_vma value = 0;
510 errmsg = parse_disp21 (cd, strp, OR1K_OPERAND_DISP21, 0, NULL, & value);
511 fields->f_disp21 = value;
512 }
513 break;
514 case OR1K_OPERAND_DISP26 :
515 {
516 bfd_vma value = 0;
517 errmsg = parse_disp26 (cd, strp, OR1K_OPERAND_DISP26, 0, NULL, & value);
518 fields->f_disp26 = value;
519 }
520 break;
521 case OR1K_OPERAND_RA :
522 errmsg = cgen_parse_keyword (cd, strp, & or1k_cgen_opval_h_gpr, & fields->f_r2);
523 break;
524 case OR1K_OPERAND_RAD32F :
525 errmsg = parse_regpair (cd, strp, OR1K_OPERAND_RAD32F, (unsigned long *) (& fields->f_rad32));
526 break;
527 case OR1K_OPERAND_RADI :
528 errmsg = parse_regpair (cd, strp, OR1K_OPERAND_RADI, (unsigned long *) (& fields->f_rad32));
529 break;
530 case OR1K_OPERAND_RASF :
531 errmsg = cgen_parse_keyword (cd, strp, & or1k_cgen_opval_h_fsr, & fields->f_r2);
532 break;
533 case OR1K_OPERAND_RB :
534 errmsg = cgen_parse_keyword (cd, strp, & or1k_cgen_opval_h_gpr, & fields->f_r3);
535 break;
536 case OR1K_OPERAND_RBD32F :
537 errmsg = parse_regpair (cd, strp, OR1K_OPERAND_RBD32F, (unsigned long *) (& fields->f_rbd32));
538 break;
539 case OR1K_OPERAND_RBDI :
540 errmsg = parse_regpair (cd, strp, OR1K_OPERAND_RBDI, (unsigned long *) (& fields->f_rbd32));
541 break;
542 case OR1K_OPERAND_RBSF :
543 errmsg = cgen_parse_keyword (cd, strp, & or1k_cgen_opval_h_fsr, & fields->f_r3);
544 break;
545 case OR1K_OPERAND_RD :
546 errmsg = cgen_parse_keyword (cd, strp, & or1k_cgen_opval_h_gpr, & fields->f_r1);
547 break;
548 case OR1K_OPERAND_RDD32F :
549 errmsg = parse_regpair (cd, strp, OR1K_OPERAND_RDD32F, (unsigned long *) (& fields->f_rdd32));
550 break;
551 case OR1K_OPERAND_RDDI :
552 errmsg = parse_regpair (cd, strp, OR1K_OPERAND_RDDI, (unsigned long *) (& fields->f_rdd32));
553 break;
554 case OR1K_OPERAND_RDSF :
555 errmsg = cgen_parse_keyword (cd, strp, & or1k_cgen_opval_h_fsr, & fields->f_r1);
556 break;
557 case OR1K_OPERAND_SIMM16 :
558 errmsg = parse_simm16 (cd, strp, OR1K_OPERAND_SIMM16, (long *) (& fields->f_simm16));
559 break;
560 case OR1K_OPERAND_SIMM16_SPLIT :
561 errmsg = parse_simm16_split (cd, strp, OR1K_OPERAND_SIMM16_SPLIT, (long *) (& fields->f_simm16_split));
562 break;
563 case OR1K_OPERAND_UIMM16 :
564 errmsg = parse_uimm16 (cd, strp, OR1K_OPERAND_UIMM16, (unsigned long *) (& fields->f_uimm16));
565 break;
566 case OR1K_OPERAND_UIMM16_SPLIT :
567 errmsg = parse_uimm16_split (cd, strp, OR1K_OPERAND_UIMM16_SPLIT, (unsigned long *) (& fields->f_uimm16_split));
568 break;
569 case OR1K_OPERAND_UIMM6 :
570 errmsg = cgen_parse_unsigned_integer (cd, strp, OR1K_OPERAND_UIMM6, (unsigned long *) (& fields->f_uimm6));
571 break;
572
573 default :
574 /* xgettext:c-format */
575 opcodes_error_handler
576 (_("internal error: unrecognized field %d while parsing"),
577 opindex);
578 abort ();
579 }
580
581 return errmsg;
582 }
583
584 cgen_parse_fn * const or1k_cgen_parse_handlers[] =
585 {
586 parse_insn_normal,
587 };
588
589 void
590 or1k_cgen_init_asm (CGEN_CPU_DESC cd)
591 {
592 or1k_cgen_init_opcode_table (cd);
593 or1k_cgen_init_ibld_table (cd);
594 cd->parse_handlers = & or1k_cgen_parse_handlers[0];
595 cd->parse_operand = or1k_cgen_parse_operand;
596 #ifdef CGEN_ASM_INIT_HOOK
597 CGEN_ASM_INIT_HOOK
598 #endif
599 }
600
601 \f
602
603 /* Regex construction routine.
604
605 This translates an opcode syntax string into a regex string,
606 by replacing any non-character syntax element (such as an
607 opcode) with the pattern '.*'
608
609 It then compiles the regex and stores it in the opcode, for
610 later use by or1k_cgen_assemble_insn
611
612 Returns NULL for success, an error message for failure. */
613
614 char *
615 or1k_cgen_build_insn_regex (CGEN_INSN *insn)
616 {
617 CGEN_OPCODE *opc = (CGEN_OPCODE *) CGEN_INSN_OPCODE (insn);
618 const char *mnem = CGEN_INSN_MNEMONIC (insn);
619 char rxbuf[CGEN_MAX_RX_ELEMENTS];
620 char *rx = rxbuf;
621 const CGEN_SYNTAX_CHAR_TYPE *syn;
622 int reg_err;
623
624 syn = CGEN_SYNTAX_STRING (CGEN_OPCODE_SYNTAX (opc));
625
626 /* Mnemonics come first in the syntax string. */
627 if (! CGEN_SYNTAX_MNEMONIC_P (* syn))
628 return _("missing mnemonic in syntax string");
629 ++syn;
630
631 /* Generate a case sensitive regular expression that emulates case
632 insensitive matching in the "C" locale. We cannot generate a case
633 insensitive regular expression because in Turkish locales, 'i' and 'I'
634 are not equal modulo case conversion. */
635
636 /* Copy the literal mnemonic out of the insn. */
637 for (; *mnem; mnem++)
638 {
639 char c = *mnem;
640
641 if (ISALPHA (c))
642 {
643 *rx++ = '[';
644 *rx++ = TOLOWER (c);
645 *rx++ = TOUPPER (c);
646 *rx++ = ']';
647 }
648 else
649 *rx++ = c;
650 }
651
652 /* Copy any remaining literals from the syntax string into the rx. */
653 for(; * syn != 0 && rx <= rxbuf + (CGEN_MAX_RX_ELEMENTS - 7 - 4); ++syn)
654 {
655 if (CGEN_SYNTAX_CHAR_P (* syn))
656 {
657 char c = CGEN_SYNTAX_CHAR (* syn);
658
659 switch (c)
660 {
661 /* Escape any regex metacharacters in the syntax. */
662 case '.': case '[': case '\\':
663 case '*': case '^': case '$':
664
665 #ifdef CGEN_ESCAPE_EXTENDED_REGEX
666 case '?': case '{': case '}':
667 case '(': case ')': case '*':
668 case '|': case '+': case ']':
669 #endif
670 *rx++ = '\\';
671 *rx++ = c;
672 break;
673
674 default:
675 if (ISALPHA (c))
676 {
677 *rx++ = '[';
678 *rx++ = TOLOWER (c);
679 *rx++ = TOUPPER (c);
680 *rx++ = ']';
681 }
682 else
683 *rx++ = c;
684 break;
685 }
686 }
687 else
688 {
689 /* Replace non-syntax fields with globs. */
690 *rx++ = '.';
691 *rx++ = '*';
692 }
693 }
694
695 /* Trailing whitespace ok. */
696 * rx++ = '[';
697 * rx++ = ' ';
698 * rx++ = '\t';
699 * rx++ = ']';
700 * rx++ = '*';
701
702 /* But anchor it after that. */
703 * rx++ = '$';
704 * rx = '\0';
705
706 CGEN_INSN_RX (insn) = xmalloc (sizeof (regex_t));
707 reg_err = regcomp ((regex_t *) CGEN_INSN_RX (insn), rxbuf, REG_NOSUB);
708
709 if (reg_err == 0)
710 return NULL;
711 else
712 {
713 static char msg[80];
714
715 regerror (reg_err, (regex_t *) CGEN_INSN_RX (insn), msg, 80);
716 regfree ((regex_t *) CGEN_INSN_RX (insn));
717 free (CGEN_INSN_RX (insn));
718 (CGEN_INSN_RX (insn)) = NULL;
719 return msg;
720 }
721 }
722
723 \f
724 /* Default insn parser.
725
726 The syntax string is scanned and operands are parsed and stored in FIELDS.
727 Relocs are queued as we go via other callbacks.
728
729 ??? Note that this is currently an all-or-nothing parser. If we fail to
730 parse the instruction, we return 0 and the caller will start over from
731 the beginning. Backtracking will be necessary in parsing subexpressions,
732 but that can be handled there. Not handling backtracking here may get
733 expensive in the case of the m68k. Deal with later.
734
735 Returns NULL for success, an error message for failure. */
736
737 static const char *
738 parse_insn_normal (CGEN_CPU_DESC cd,
739 const CGEN_INSN *insn,
740 const char **strp,
741 CGEN_FIELDS *fields)
742 {
743 /* ??? Runtime added insns not handled yet. */
744 const CGEN_SYNTAX *syntax = CGEN_INSN_SYNTAX (insn);
745 const char *str = *strp;
746 const char *errmsg;
747 const char *p;
748 const CGEN_SYNTAX_CHAR_TYPE * syn;
749 #ifdef CGEN_MNEMONIC_OPERANDS
750 /* FIXME: wip */
751 int past_opcode_p;
752 #endif
753
754 /* For now we assume the mnemonic is first (there are no leading operands).
755 We can parse it without needing to set up operand parsing.
756 GAS's input scrubber will ensure mnemonics are lowercase, but we may
757 not be called from GAS. */
758 p = CGEN_INSN_MNEMONIC (insn);
759 while (*p && TOLOWER (*p) == TOLOWER (*str))
760 ++p, ++str;
761
762 if (* p)
763 return _("unrecognized instruction");
764
765 #ifndef CGEN_MNEMONIC_OPERANDS
766 if (* str && ! ISSPACE (* str))
767 return _("unrecognized instruction");
768 #endif
769
770 CGEN_INIT_PARSE (cd);
771 cgen_init_parse_operand (cd);
772 #ifdef CGEN_MNEMONIC_OPERANDS
773 past_opcode_p = 0;
774 #endif
775
776 /* We don't check for (*str != '\0') here because we want to parse
777 any trailing fake arguments in the syntax string. */
778 syn = CGEN_SYNTAX_STRING (syntax);
779
780 /* Mnemonics come first for now, ensure valid string. */
781 if (! CGEN_SYNTAX_MNEMONIC_P (* syn))
782 abort ();
783
784 ++syn;
785
786 while (* syn != 0)
787 {
788 /* Non operand chars must match exactly. */
789 if (CGEN_SYNTAX_CHAR_P (* syn))
790 {
791 /* FIXME: While we allow for non-GAS callers above, we assume the
792 first char after the mnemonic part is a space. */
793 /* FIXME: We also take inappropriate advantage of the fact that
794 GAS's input scrubber will remove extraneous blanks. */
795 if (TOLOWER (*str) == TOLOWER (CGEN_SYNTAX_CHAR (* syn)))
796 {
797 #ifdef CGEN_MNEMONIC_OPERANDS
798 if (CGEN_SYNTAX_CHAR(* syn) == ' ')
799 past_opcode_p = 1;
800 #endif
801 ++ syn;
802 ++ str;
803 }
804 else if (*str)
805 {
806 /* Syntax char didn't match. Can't be this insn. */
807 static char msg [80];
808
809 /* xgettext:c-format */
810 sprintf (msg, _("syntax error (expected char `%c', found `%c')"),
811 CGEN_SYNTAX_CHAR(*syn), *str);
812 return msg;
813 }
814 else
815 {
816 /* Ran out of input. */
817 static char msg [80];
818
819 /* xgettext:c-format */
820 sprintf (msg, _("syntax error (expected char `%c', found end of instruction)"),
821 CGEN_SYNTAX_CHAR(*syn));
822 return msg;
823 }
824 continue;
825 }
826
827 #ifdef CGEN_MNEMONIC_OPERANDS
828 (void) past_opcode_p;
829 #endif
830 /* We have an operand of some sort. */
831 errmsg = cd->parse_operand (cd, CGEN_SYNTAX_FIELD (*syn), &str, fields);
832 if (errmsg)
833 return errmsg;
834
835 /* Done with this operand, continue with next one. */
836 ++ syn;
837 }
838
839 /* If we're at the end of the syntax string, we're done. */
840 if (* syn == 0)
841 {
842 /* FIXME: For the moment we assume a valid `str' can only contain
843 blanks now. IE: We needn't try again with a longer version of
844 the insn and it is assumed that longer versions of insns appear
845 before shorter ones (eg: lsr r2,r3,1 vs lsr r2,r3). */
846 while (ISSPACE (* str))
847 ++ str;
848
849 if (* str != '\0')
850 return _("junk at end of line"); /* FIXME: would like to include `str' */
851
852 return NULL;
853 }
854
855 /* We couldn't parse it. */
856 return _("unrecognized instruction");
857 }
858 \f
859 /* Main entry point.
860 This routine is called for each instruction to be assembled.
861 STR points to the insn to be assembled.
862 We assume all necessary tables have been initialized.
863 The assembled instruction, less any fixups, is stored in BUF.
864 Remember that if CGEN_INT_INSN_P then BUF is an int and thus the value
865 still needs to be converted to target byte order, otherwise BUF is an array
866 of bytes in target byte order.
867 The result is a pointer to the insn's entry in the opcode table,
868 or NULL if an error occured (an error message will have already been
869 printed).
870
871 Note that when processing (non-alias) macro-insns,
872 this function recurses.
873
874 ??? It's possible to make this cpu-independent.
875 One would have to deal with a few minor things.
876 At this point in time doing so would be more of a curiosity than useful
877 [for example this file isn't _that_ big], but keeping the possibility in
878 mind helps keep the design clean. */
879
880 const CGEN_INSN *
881 or1k_cgen_assemble_insn (CGEN_CPU_DESC cd,
882 const char *str,
883 CGEN_FIELDS *fields,
884 CGEN_INSN_BYTES_PTR buf,
885 char **errmsg)
886 {
887 const char *start;
888 CGEN_INSN_LIST *ilist;
889 const char *parse_errmsg = NULL;
890 const char *insert_errmsg = NULL;
891 int recognized_mnemonic = 0;
892
893 /* Skip leading white space. */
894 while (ISSPACE (* str))
895 ++ str;
896
897 /* The instructions are stored in hashed lists.
898 Get the first in the list. */
899 ilist = CGEN_ASM_LOOKUP_INSN (cd, str);
900
901 /* Keep looking until we find a match. */
902 start = str;
903 for ( ; ilist != NULL ; ilist = CGEN_ASM_NEXT_INSN (ilist))
904 {
905 const CGEN_INSN *insn = ilist->insn;
906 recognized_mnemonic = 1;
907
908 #ifdef CGEN_VALIDATE_INSN_SUPPORTED
909 /* Not usually needed as unsupported opcodes
910 shouldn't be in the hash lists. */
911 /* Is this insn supported by the selected cpu? */
912 if (! or1k_cgen_insn_supported (cd, insn))
913 continue;
914 #endif
915 /* If the RELAXED attribute is set, this is an insn that shouldn't be
916 chosen immediately. Instead, it is used during assembler/linker
917 relaxation if possible. */
918 if (CGEN_INSN_ATTR_VALUE (insn, CGEN_INSN_RELAXED) != 0)
919 continue;
920
921 str = start;
922
923 /* Skip this insn if str doesn't look right lexically. */
924 if (CGEN_INSN_RX (insn) != NULL &&
925 regexec ((regex_t *) CGEN_INSN_RX (insn), str, 0, NULL, 0) == REG_NOMATCH)
926 continue;
927
928 /* Allow parse/insert handlers to obtain length of insn. */
929 CGEN_FIELDS_BITSIZE (fields) = CGEN_INSN_BITSIZE (insn);
930
931 parse_errmsg = CGEN_PARSE_FN (cd, insn) (cd, insn, & str, fields);
932 if (parse_errmsg != NULL)
933 continue;
934
935 /* ??? 0 is passed for `pc'. */
936 insert_errmsg = CGEN_INSERT_FN (cd, insn) (cd, insn, fields, buf,
937 (bfd_vma) 0);
938 if (insert_errmsg != NULL)
939 continue;
940
941 /* It is up to the caller to actually output the insn and any
942 queued relocs. */
943 return insn;
944 }
945
946 {
947 static char errbuf[150];
948 const char *tmp_errmsg;
949 #ifdef CGEN_VERBOSE_ASSEMBLER_ERRORS
950 #define be_verbose 1
951 #else
952 #define be_verbose 0
953 #endif
954
955 if (be_verbose)
956 {
957 /* If requesting verbose error messages, use insert_errmsg.
958 Failing that, use parse_errmsg. */
959 tmp_errmsg = (insert_errmsg ? insert_errmsg :
960 parse_errmsg ? parse_errmsg :
961 recognized_mnemonic ?
962 _("unrecognized form of instruction") :
963 _("unrecognized instruction"));
964
965 if (strlen (start) > 50)
966 /* xgettext:c-format */
967 sprintf (errbuf, "%s `%.50s...'", tmp_errmsg, start);
968 else
969 /* xgettext:c-format */
970 sprintf (errbuf, "%s `%.50s'", tmp_errmsg, start);
971 }
972 else
973 {
974 if (strlen (start) > 50)
975 /* xgettext:c-format */
976 sprintf (errbuf, _("bad instruction `%.50s...'"), start);
977 else
978 /* xgettext:c-format */
979 sprintf (errbuf, _("bad instruction `%.50s'"), start);
980 }
981
982 *errmsg = errbuf;
983 return NULL;
984 }
985 }