1 |
/* |
2 |
* Copyright (C) 2004-2005 Anders Gavare. All rights reserved. |
3 |
* |
4 |
* Redistribution and use in source and binary forms, with or without |
5 |
* modification, are permitted provided that the following conditions are met: |
6 |
* |
7 |
* 1. Redistributions of source code must retain the above copyright |
8 |
* notice, this list of conditions and the following disclaimer. |
9 |
* 2. Redistributions in binary form must reproduce the above copyright |
10 |
* notice, this list of conditions and the following disclaimer in the |
11 |
* documentation and/or other materials provided with the distribution. |
12 |
* 3. The name of the author may not be used to endorse or promote products |
13 |
* derived from this software without specific prior written permission. |
14 |
* |
15 |
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND |
16 |
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
17 |
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
18 |
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE |
19 |
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
20 |
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
21 |
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
22 |
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
23 |
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
24 |
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
25 |
* SUCH DAMAGE. |
26 |
* |
27 |
* |
28 |
* $Id: bintrans_i386.c,v 1.76 2005/06/22 10:12:25 debug Exp $ |
29 |
* |
30 |
* i386 specific code for dynamic binary translation. |
31 |
* See bintrans.c for more information. Included from bintrans.c. |
32 |
* |
33 |
* Translated code uses the following conventions at all time: |
34 |
* |
35 |
* esi points to the cpu struct |
36 |
* edi lowest 32 bits of cpu->pc |
37 |
* ebp contains cpu->bintrans_instructions_executed |
38 |
*/ |
39 |
|
40 |
|
41 |
struct cpu dummy_cpu; |
42 |
struct mips_coproc dummy_coproc; |
43 |
struct vth32_table dummy_vth32_table; |
44 |
|
45 |
|
46 |
/* |
47 |
* bintrans_host_cacheinvalidate() |
48 |
* |
49 |
* Invalidate the host's instruction cache. On i386, this isn't necessary, |
50 |
* so this is an empty function. |
51 |
*/ |
52 |
static void bintrans_host_cacheinvalidate(unsigned char *p, size_t len) |
53 |
{ |
54 |
/* Do nothing. */ |
55 |
} |
56 |
|
57 |
|
58 |
/* offsetof (in stdarg.h) could possibly be used, but I'm not sure |
59 |
if it will take care of the compiler problems... */ |
60 |
|
61 |
#define ofs_i (((size_t)&dummy_cpu.cd.mips.bintrans_instructions_executed) - ((size_t)&dummy_cpu)) |
62 |
#define ofs_pc (((size_t)&dummy_cpu.pc) - ((size_t)&dummy_cpu)) |
63 |
#define ofs_pc_last (((size_t)&dummy_cpu.cd.mips.pc_last) - ((size_t)&dummy_cpu)) |
64 |
#define ofs_tabl0 (((size_t)&dummy_cpu.cd.mips.vaddr_to_hostaddr_table0) - ((size_t)&dummy_cpu)) |
65 |
#define ofs_chunks ((size_t)&dummy_vth32_table.bintrans_chunks[0] - (size_t)&dummy_vth32_table) |
66 |
#define ofs_chunkbase ((size_t)&dummy_cpu.cd.mips.chunk_base_address - (size_t)&dummy_cpu) |
67 |
|
68 |
|
69 |
static void (*bintrans_runchunk)(struct cpu *, unsigned char *); |
70 |
static void (*bintrans_jump_to_32bit_pc)(struct cpu *); |
71 |
static void (*bintrans_load_32bit)(struct cpu *); |
72 |
static void (*bintrans_store_32bit)(struct cpu *); |
73 |
|
74 |
|
75 |
/* |
76 |
* bintrans_write_quickjump(): |
77 |
*/ |
78 |
static void bintrans_write_quickjump(struct memory *mem, |
79 |
unsigned char *quickjump_code, uint32_t chunkoffset) |
80 |
{ |
81 |
uint32_t i386_addr; |
82 |
unsigned char *a = quickjump_code; |
83 |
|
84 |
i386_addr = chunkoffset + (size_t)mem->translation_code_chunk_space; |
85 |
i386_addr = i386_addr - ((size_t)a + 5); |
86 |
|
87 |
/* printf("chunkoffset=%i, %08x %08x %i\n", |
88 |
chunkoffset, i386_addr, a, ofs); */ |
89 |
|
90 |
*a++ = 0xe9; |
91 |
*a++ = i386_addr; |
92 |
*a++ = i386_addr >> 8; |
93 |
*a++ = i386_addr >> 16; |
94 |
*a++ = i386_addr >> 24; |
95 |
} |
96 |
|
97 |
|
98 |
/* |
99 |
* bintrans_write_chunkreturn(): |
100 |
*/ |
101 |
static void bintrans_write_chunkreturn(unsigned char **addrp) |
102 |
{ |
103 |
unsigned char *a = *addrp; |
104 |
*a++ = 0xc3; /* ret */ |
105 |
*addrp = a; |
106 |
} |
107 |
|
108 |
|
109 |
/* |
110 |
* bintrans_write_chunkreturn_fail(): |
111 |
*/ |
112 |
static void bintrans_write_chunkreturn_fail(unsigned char **addrp) |
113 |
{ |
114 |
unsigned char *a = *addrp; |
115 |
|
116 |
/* 81 cd 00 00 00 01 orl $0x1000000,%ebp */ |
117 |
*a++ = 0x81; *a++ = 0xcd; |
118 |
*a++ = 0; *a++ = 0; *a++ = 0; *a++ = 0x01; /* TODO: not hardcoded */ |
119 |
|
120 |
*a++ = 0xc3; /* ret */ |
121 |
*addrp = a; |
122 |
} |
123 |
|
124 |
|
125 |
/* |
126 |
* bintrans_write_pc_inc(): |
127 |
*/ |
128 |
static void bintrans_write_pc_inc(unsigned char **addrp) |
129 |
{ |
130 |
unsigned char *a = *addrp; |
131 |
|
132 |
/* 83 c7 04 add $0x4,%edi */ |
133 |
*a++ = 0x83; *a++ = 0xc7; *a++ = 4; |
134 |
|
135 |
#if 0 |
136 |
if (!bintrans_32bit_only) { |
137 |
int ofs; |
138 |
/* 83 96 zz zz zz zz 00 adcl $0x0,zz(%esi) */ |
139 |
ofs = ((size_t)&dummy_cpu.pc) - (size_t)&dummy_cpu; |
140 |
ofs += 4; |
141 |
*a++ = 0x83; *a++ = 0x96; |
142 |
*a++ = ofs & 255; |
143 |
*a++ = (ofs >> 8) & 255; |
144 |
*a++ = (ofs >> 16) & 255; |
145 |
*a++ = (ofs >> 24) & 255; |
146 |
*a++ = 0; |
147 |
} |
148 |
#endif |
149 |
|
150 |
/* 45 inc %ebp */ |
151 |
*a++ = 0x45; |
152 |
|
153 |
*addrp = a; |
154 |
} |
155 |
|
156 |
|
157 |
/* |
158 |
* load_pc_into_eax_edx(): |
159 |
*/ |
160 |
static void load_pc_into_eax_edx(unsigned char **addrp) |
161 |
{ |
162 |
unsigned char *a; |
163 |
a = *addrp; |
164 |
|
165 |
/* 89 f8 mov %edi,%eax */ |
166 |
*a++ = 0x89; *a++ = 0xf8; |
167 |
|
168 |
#if 0 |
169 |
if (bintrans_32bit_only) { |
170 |
/* 99 cltd */ |
171 |
*a++ = 0x99; |
172 |
} else |
173 |
#endif |
174 |
{ |
175 |
int ofs = ((size_t)&dummy_cpu.pc) - (size_t)&dummy_cpu; |
176 |
/* 8b 96 3c 30 00 00 mov 0x303c(%esi),%edx */ |
177 |
ofs += 4; |
178 |
*a++ = 0x8b; *a++ = 0x96; |
179 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
180 |
} |
181 |
|
182 |
*addrp = a; |
183 |
} |
184 |
|
185 |
|
186 |
/* |
187 |
* store_eax_edx_into_pc(): |
188 |
*/ |
189 |
static void store_eax_edx_into_pc(unsigned char **addrp) |
190 |
{ |
191 |
unsigned char *a; |
192 |
int ofs = ((size_t)&dummy_cpu.pc) - (size_t)&dummy_cpu; |
193 |
a = *addrp; |
194 |
|
195 |
/* 89 c7 mov %eax,%edi */ |
196 |
*a++ = 0x89; *a++ = 0xc7; |
197 |
|
198 |
/* 89 96 3c 30 00 00 mov %edx,0x303c(%esi) */ |
199 |
ofs += 4; |
200 |
*a++ = 0x89; *a++ = 0x96; |
201 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
202 |
|
203 |
*addrp = a; |
204 |
} |
205 |
|
206 |
|
207 |
/* |
208 |
* load_into_eax_edx(): |
209 |
* |
210 |
* Usage: load_into_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); etc. |
211 |
*/ |
212 |
static void load_into_eax_edx(unsigned char **addrp, void *p) |
213 |
{ |
214 |
unsigned char *a; |
215 |
int ofs = (size_t)p - (size_t)&dummy_cpu; |
216 |
a = *addrp; |
217 |
|
218 |
/* 8b 86 38 30 00 00 mov 0x3038(%esi),%eax */ |
219 |
*a++ = 0x8b; *a++ = 0x86; |
220 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
221 |
|
222 |
#if 0 |
223 |
if (bintrans_32bit_only) { |
224 |
/* 99 cltd */ |
225 |
*a++ = 0x99; |
226 |
} else |
227 |
#endif |
228 |
{ |
229 |
/* 8b 96 3c 30 00 00 mov 0x303c(%esi),%edx */ |
230 |
ofs += 4; |
231 |
*a++ = 0x8b; *a++ = 0x96; |
232 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
233 |
} |
234 |
|
235 |
*addrp = a; |
236 |
} |
237 |
|
238 |
|
239 |
/* |
240 |
* load_into_eax_and_sign_extend_into_edx(): |
241 |
* |
242 |
* Usage: load_into_eax_and_sign_extend_into_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); etc. |
243 |
*/ |
244 |
static void load_into_eax_and_sign_extend_into_edx(unsigned char **addrp, void *p) |
245 |
{ |
246 |
unsigned char *a; |
247 |
int ofs = (size_t)p - (size_t)&dummy_cpu; |
248 |
a = *addrp; |
249 |
|
250 |
/* 8b 86 38 30 00 00 mov 0x3038(%esi),%eax */ |
251 |
*a++ = 0x8b; *a++ = 0x86; |
252 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
253 |
|
254 |
/* 99 cltd */ |
255 |
*a++ = 0x99; |
256 |
|
257 |
*addrp = a; |
258 |
} |
259 |
|
260 |
|
261 |
/* |
262 |
* load_into_eax_dont_care_about_edx(): |
263 |
* |
264 |
* Usage: load_into_eax_dont_care_about_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); etc. |
265 |
*/ |
266 |
static void load_into_eax_dont_care_about_edx(unsigned char **addrp, void *p) |
267 |
{ |
268 |
unsigned char *a; |
269 |
int ofs = (size_t)p - (size_t)&dummy_cpu; |
270 |
a = *addrp; |
271 |
|
272 |
/* 8b 86 38 30 00 00 mov 0x3038(%esi),%eax */ |
273 |
*a++ = 0x8b; *a++ = 0x86; |
274 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
275 |
|
276 |
*addrp = a; |
277 |
} |
278 |
|
279 |
|
280 |
/* |
281 |
* store_eax_edx(): |
282 |
* |
283 |
* Usage: store_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); etc. |
284 |
*/ |
285 |
static void store_eax_edx(unsigned char **addrp, void *p) |
286 |
{ |
287 |
unsigned char *a; |
288 |
int ofs = (size_t)p - (size_t)&dummy_cpu; |
289 |
a = *addrp; |
290 |
|
291 |
/* 89 86 38 30 00 00 mov %eax,0x3038(%esi) */ |
292 |
*a++ = 0x89; *a++ = 0x86; |
293 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
294 |
|
295 |
/* 89 96 3c 30 00 00 mov %edx,0x303c(%esi) */ |
296 |
ofs += 4; |
297 |
*a++ = 0x89; *a++ = 0x96; |
298 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
299 |
|
300 |
*addrp = a; |
301 |
} |
302 |
|
303 |
|
304 |
/* |
305 |
* bintrans_write_instruction__lui(): |
306 |
*/ |
307 |
static int bintrans_write_instruction__lui(unsigned char **addrp, int rt, int imm) |
308 |
{ |
309 |
unsigned char *a; |
310 |
|
311 |
a = *addrp; |
312 |
if (rt == 0) |
313 |
goto rt0; |
314 |
|
315 |
/* b8 00 00 dc fe mov $0xfedc0000,%eax */ |
316 |
*a++ = 0xb8; *a++ = 0; *a++ = 0; |
317 |
*a++ = imm & 255; *a++ = imm >> 8; |
318 |
|
319 |
/* 99 cltd */ |
320 |
*a++ = 0x99; |
321 |
|
322 |
store_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rt]); |
323 |
*addrp = a; |
324 |
|
325 |
rt0: |
326 |
bintrans_write_pc_inc(addrp); |
327 |
return 1; |
328 |
} |
329 |
|
330 |
|
331 |
/* |
332 |
* bintrans_write_instruction__jr(): |
333 |
*/ |
334 |
static int bintrans_write_instruction__jr(unsigned char **addrp, |
335 |
int rs, int rd, int special) |
336 |
{ |
337 |
unsigned char *a; |
338 |
int ofs; |
339 |
|
340 |
a = *addrp; |
341 |
|
342 |
/* |
343 |
* Perform the jump by setting cpu->delay_slot = TO_BE_DELAYED |
344 |
* and cpu->delay_jmpaddr = gpr[rs]. |
345 |
*/ |
346 |
|
347 |
/* c7 86 38 30 00 00 01 00 00 00 movl $0x1,0x3038(%esi) */ |
348 |
ofs = ((size_t)&dummy_cpu.cd.mips.delay_slot) - (size_t)&dummy_cpu; |
349 |
*a++ = 0xc7; *a++ = 0x86; |
350 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
351 |
*a++ = TO_BE_DELAYED; *a++ = 0; *a++ = 0; *a++ = 0; |
352 |
|
353 |
#if 0 |
354 |
if (bintrans_32bit_only) |
355 |
load_into_eax_and_sign_extend_into_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); |
356 |
else |
357 |
#endif |
358 |
load_into_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); |
359 |
|
360 |
store_eax_edx(&a, &dummy_cpu.cd.mips.delay_jmpaddr); |
361 |
|
362 |
if (special == SPECIAL_JALR && rd != 0) { |
363 |
/* gpr[rd] = retaddr (pc + 8) */ |
364 |
|
365 |
#if 0 |
366 |
if (bintrans_32bit_only) { |
367 |
load_pc_into_eax_edx(&a); |
368 |
/* 83 c0 08 add $0x8,%eax */ |
369 |
*a++ = 0x83; *a++ = 0xc0; *a++ = 0x08; |
370 |
} else |
371 |
#endif |
372 |
{ |
373 |
load_pc_into_eax_edx(&a); |
374 |
/* 83 c0 08 add $0x8,%eax */ |
375 |
/* 83 d2 00 adc $0x0,%edx */ |
376 |
*a++ = 0x83; *a++ = 0xc0; *a++ = 0x08; |
377 |
*a++ = 0x83; *a++ = 0xd2; *a++ = 0x00; |
378 |
} |
379 |
|
380 |
store_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rd]); |
381 |
} |
382 |
|
383 |
*addrp = a; |
384 |
bintrans_write_pc_inc(addrp); |
385 |
return 1; |
386 |
} |
387 |
|
388 |
|
389 |
/* |
390 |
* bintrans_write_instruction__mfmthilo(): |
391 |
*/ |
392 |
static int bintrans_write_instruction__mfmthilo(unsigned char **addrp, |
393 |
int rd, int from_flag, int hi_flag) |
394 |
{ |
395 |
unsigned char *a; |
396 |
|
397 |
a = *addrp; |
398 |
|
399 |
if (from_flag) { |
400 |
if (rd != 0) { |
401 |
/* mfhi or mflo */ |
402 |
if (hi_flag) |
403 |
load_into_eax_edx(&a, &dummy_cpu.cd.mips.hi); |
404 |
else |
405 |
load_into_eax_edx(&a, &dummy_cpu.cd.mips.lo); |
406 |
store_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rd]); |
407 |
} |
408 |
} else { |
409 |
/* mthi or mtlo */ |
410 |
load_into_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rd]); |
411 |
if (hi_flag) |
412 |
store_eax_edx(&a, &dummy_cpu.cd.mips.hi); |
413 |
else |
414 |
store_eax_edx(&a, &dummy_cpu.cd.mips.lo); |
415 |
} |
416 |
|
417 |
*addrp = a; |
418 |
bintrans_write_pc_inc(addrp); |
419 |
return 1; |
420 |
} |
421 |
|
422 |
|
423 |
/* |
424 |
* bintrans_write_instruction__addiu_etc(): |
425 |
*/ |
426 |
static int bintrans_write_instruction__addiu_etc(unsigned char **addrp, |
427 |
int rt, int rs, int imm, int instruction_type) |
428 |
{ |
429 |
unsigned char *a; |
430 |
unsigned int uimm; |
431 |
|
432 |
/* TODO: overflow detection for ADDI and DADDI */ |
433 |
switch (instruction_type) { |
434 |
case HI6_ADDI: |
435 |
case HI6_DADDI: |
436 |
return 0; |
437 |
} |
438 |
|
439 |
a = *addrp; |
440 |
|
441 |
if (rt == 0) |
442 |
goto rt0; |
443 |
|
444 |
uimm = imm & 0xffff; |
445 |
|
446 |
if (uimm == 0 && (instruction_type == HI6_ADDIU || |
447 |
instruction_type == HI6_ADDI)) { |
448 |
load_into_eax_and_sign_extend_into_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); |
449 |
store_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rt]); |
450 |
goto rt0; |
451 |
} |
452 |
|
453 |
if (uimm == 0 && (instruction_type == HI6_DADDIU || |
454 |
instruction_type == HI6_DADDI || instruction_type == HI6_ORI)) { |
455 |
load_into_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); |
456 |
store_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rt]); |
457 |
goto rt0; |
458 |
} |
459 |
|
460 |
#if 0 |
461 |
if (bintrans_32bit_only) |
462 |
load_into_eax_and_sign_extend_into_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); |
463 |
else |
464 |
#endif |
465 |
load_into_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); |
466 |
|
467 |
switch (instruction_type) { |
468 |
case HI6_ADDIU: |
469 |
case HI6_DADDIU: |
470 |
case HI6_ADDI: |
471 |
case HI6_DADDI: |
472 |
if (imm & 0x8000) { |
473 |
/* 05 39 fd ff ff add $0xfffffd39,%eax */ |
474 |
/* 83 d2 ff adc $0xffffffff,%edx */ |
475 |
*a++ = 0x05; *a++ = uimm; *a++ = uimm >> 8; *a++ = 0xff; *a++ = 0xff; |
476 |
if (instruction_type == HI6_DADDIU) { |
477 |
*a++ = 0x83; *a++ = 0xd2; *a++ = 0xff; |
478 |
} |
479 |
} else { |
480 |
/* 05 c7 02 00 00 add $0x2c7,%eax */ |
481 |
/* 83 d2 00 adc $0x0,%edx */ |
482 |
*a++ = 0x05; *a++ = uimm; *a++ = uimm >> 8; *a++ = 0; *a++ = 0; |
483 |
if (instruction_type == HI6_DADDIU) { |
484 |
*a++ = 0x83; *a++ = 0xd2; *a++ = 0; |
485 |
} |
486 |
} |
487 |
if (instruction_type == HI6_ADDIU) { |
488 |
/* 99 cltd */ |
489 |
*a++ = 0x99; |
490 |
} |
491 |
break; |
492 |
case HI6_ANDI: |
493 |
/* 25 34 12 00 00 and $0x1234,%eax */ |
494 |
/* 31 d2 xor %edx,%edx */ |
495 |
*a++ = 0x25; *a++ = uimm; *a++ = uimm >> 8; *a++ = 0; *a++ = 0; |
496 |
*a++ = 0x31; *a++ = 0xd2; |
497 |
break; |
498 |
case HI6_ORI: |
499 |
/* 0d 34 12 00 00 or $0x1234,%eax */ |
500 |
*a++ = 0xd; *a++ = uimm; *a++ = uimm >> 8; *a++ = 0; *a++ = 0; |
501 |
break; |
502 |
case HI6_XORI: |
503 |
/* 35 34 12 00 00 xor $0x1234,%eax */ |
504 |
*a++ = 0x35; *a++ = uimm; *a++ = uimm >> 8; *a++ = 0; *a++ = 0; |
505 |
break; |
506 |
case HI6_SLTIU: |
507 |
/* set if less than, unsigned. (compare edx:eax to ecx:ebx) */ |
508 |
/* ecx:ebx = the immediate value */ |
509 |
/* bb dc fe ff ff mov $0xfffffedc,%ebx */ |
510 |
/* b9 ff ff ff ff mov $0xffffffff,%ecx */ |
511 |
/* or */ |
512 |
/* 29 c9 sub %ecx,%ecx */ |
513 |
#if 0 |
514 |
if (bintrans_32bit_only) { |
515 |
/* 99 cltd */ |
516 |
*a++ = 0x99; |
517 |
} |
518 |
#endif |
519 |
*a++ = 0xbb; *a++ = uimm; *a++ = uimm >> 8; |
520 |
if (uimm & 0x8000) { |
521 |
*a++ = 0xff; *a++ = 0xff; |
522 |
*a++ = 0xb9; *a++ = 0xff; *a++ = 0xff; *a++ = 0xff; *a++ = 0xff; |
523 |
} else { |
524 |
*a++ = 0; *a++ = 0; |
525 |
*a++ = 0x29; *a++ = 0xc9; |
526 |
} |
527 |
|
528 |
/* if edx <= ecx and eax < ebx then 1, else 0. */ |
529 |
/* 39 ca cmp %ecx,%edx */ |
530 |
/* 77 0b ja <ret0> */ |
531 |
/* 39 d8 cmp %ebx,%eax */ |
532 |
/* 73 07 jae 58 <ret0> */ |
533 |
*a++ = 0x39; *a++ = 0xca; |
534 |
*a++ = 0x77; *a++ = 0x0b; |
535 |
*a++ = 0x39; *a++ = 0xd8; |
536 |
*a++ = 0x73; *a++ = 0x07; |
537 |
|
538 |
/* b8 01 00 00 00 mov $0x1,%eax */ |
539 |
/* eb 02 jmp <common> */ |
540 |
*a++ = 0xb8; *a++ = 1; *a++ = 0; *a++ = 0; *a++ = 0; |
541 |
*a++ = 0xeb; *a++ = 0x02; |
542 |
|
543 |
/* ret0: */ |
544 |
/* 29 c0 sub %eax,%eax */ |
545 |
*a++ = 0x29; *a++ = 0xc0; |
546 |
|
547 |
/* common: */ |
548 |
/* 99 cltd */ |
549 |
*a++ = 0x99; |
550 |
break; |
551 |
case HI6_SLTI: |
552 |
/* set if less than, signed. (compare edx:eax to ecx:ebx) */ |
553 |
/* ecx:ebx = the immediate value */ |
554 |
/* bb dc fe ff ff mov $0xfffffedc,%ebx */ |
555 |
/* b9 ff ff ff ff mov $0xffffffff,%ecx */ |
556 |
/* or */ |
557 |
/* 29 c9 sub %ecx,%ecx */ |
558 |
#if 0 |
559 |
if (bintrans_32bit_only) { |
560 |
/* 99 cltd */ |
561 |
*a++ = 0x99; |
562 |
} |
563 |
#endif |
564 |
*a++ = 0xbb; *a++ = uimm; *a++ = uimm >> 8; |
565 |
if (uimm & 0x8000) { |
566 |
*a++ = 0xff; *a++ = 0xff; |
567 |
*a++ = 0xb9; *a++ = 0xff; *a++ = 0xff; *a++ = 0xff; *a++ = 0xff; |
568 |
} else { |
569 |
*a++ = 0; *a++ = 0; |
570 |
*a++ = 0x29; *a++ = 0xc9; |
571 |
} |
572 |
|
573 |
/* if edx > ecx then 0. */ |
574 |
/* if edx < ecx then 1. */ |
575 |
/* if eax < ebx then 1, else 0. */ |
576 |
/* 39 ca cmp %ecx,%edx */ |
577 |
/* 7c 0a jl <ret1> */ |
578 |
/* 7f 04 jg <ret0> */ |
579 |
/* 39 d8 cmp %ebx,%eax */ |
580 |
/* 7c 04 jl <ret1> */ |
581 |
*a++ = 0x39; *a++ = 0xca; |
582 |
*a++ = 0x7c; *a++ = 0x0a; |
583 |
*a++ = 0x7f; *a++ = 0x04; |
584 |
*a++ = 0x39; *a++ = 0xd8; |
585 |
*a++ = 0x7c; *a++ = 0x04; |
586 |
|
587 |
/* ret0: */ |
588 |
/* 29 c0 sub %eax,%eax */ |
589 |
/* eb 05 jmp <common> */ |
590 |
*a++ = 0x29; *a++ = 0xc0; |
591 |
*a++ = 0xeb; *a++ = 0x05; |
592 |
|
593 |
/* ret1: */ |
594 |
/* b8 01 00 00 00 mov $0x1,%eax */ |
595 |
*a++ = 0xb8; *a++ = 1; *a++ = 0; *a++ = 0; *a++ = 0; |
596 |
|
597 |
/* common: */ |
598 |
/* 99 cltd */ |
599 |
*a++ = 0x99; |
600 |
break; |
601 |
} |
602 |
|
603 |
store_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rt]); |
604 |
|
605 |
rt0: |
606 |
*addrp = a; |
607 |
bintrans_write_pc_inc(addrp); |
608 |
return 1; |
609 |
} |
610 |
|
611 |
|
612 |
/* |
613 |
* bintrans_write_instruction__jal(): |
614 |
*/ |
615 |
static int bintrans_write_instruction__jal(unsigned char **addrp, |
616 |
int imm, int link) |
617 |
{ |
618 |
unsigned char *a; |
619 |
uint32_t subimm; |
620 |
int ofs; |
621 |
|
622 |
a = *addrp; |
623 |
|
624 |
load_pc_into_eax_edx(&a); |
625 |
|
626 |
if (link) { |
627 |
/* gpr[31] = pc + 8 */ |
628 |
#if 0 |
629 |
if (bintrans_32bit_only) { |
630 |
/* 50 push %eax */ |
631 |
/* 83 c0 08 add $0x8,%eax */ |
632 |
*a++ = 0x50; |
633 |
*a++ = 0x83; *a++ = 0xc0; *a++ = 0x08; |
634 |
} else |
635 |
#endif |
636 |
{ |
637 |
/* 50 push %eax */ |
638 |
/* 52 push %edx */ |
639 |
/* 83 c0 08 add $0x8,%eax */ |
640 |
/* 83 d2 00 adc $0x0,%edx */ |
641 |
*a++ = 0x50; |
642 |
*a++ = 0x52; |
643 |
*a++ = 0x83; *a++ = 0xc0; *a++ = 0x08; |
644 |
*a++ = 0x83; *a++ = 0xd2; *a++ = 0x00; |
645 |
} |
646 |
store_eax_edx(&a, &dummy_cpu.cd.mips.gpr[31]); |
647 |
#if 0 |
648 |
if (bintrans_32bit_only) { |
649 |
/* 58 pop %eax */ |
650 |
*a++ = 0x58; |
651 |
} else |
652 |
#endif |
653 |
{ |
654 |
/* 5a pop %edx */ |
655 |
/* 58 pop %eax */ |
656 |
*a++ = 0x5a; |
657 |
*a++ = 0x58; |
658 |
} |
659 |
} |
660 |
|
661 |
/* delay_jmpaddr = top 36 bits of pc together with lowest 28 bits of imm*4: */ |
662 |
imm *= 4; |
663 |
|
664 |
/* Add 4, because the jump is from the delay slot: */ |
665 |
/* 83 c0 04 add $0x4,%eax */ |
666 |
/* 83 d2 00 adc $0x0,%edx */ |
667 |
*a++ = 0x83; *a++ = 0xc0; *a++ = 0x04; |
668 |
*a++ = 0x83; *a++ = 0xd2; *a++ = 0x00; |
669 |
|
670 |
/* c1 e8 1c shr $0x1c,%eax */ |
671 |
/* c1 e0 1c shl $0x1c,%eax */ |
672 |
*a++ = 0xc1; *a++ = 0xe8; *a++ = 0x1c; |
673 |
*a++ = 0xc1; *a++ = 0xe0; *a++ = 0x1c; |
674 |
|
675 |
subimm = imm; |
676 |
subimm &= 0x0fffffff; |
677 |
|
678 |
/* 0d 78 56 34 12 or $0x12345678,%eax */ |
679 |
*a++ = 0x0d; *a++ = subimm; *a++ = subimm >> 8; |
680 |
*a++ = subimm >> 16; *a++ = subimm >> 24; |
681 |
|
682 |
store_eax_edx(&a, &dummy_cpu.cd.mips.delay_jmpaddr); |
683 |
|
684 |
/* c7 86 38 30 00 00 01 00 00 00 movl $0x1,0x3038(%esi) */ |
685 |
ofs = ((size_t)&dummy_cpu.cd.mips.delay_slot) - (size_t)&dummy_cpu; |
686 |
*a++ = 0xc7; *a++ = 0x86; |
687 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
688 |
*a++ = TO_BE_DELAYED; *a++ = 0; *a++ = 0; *a++ = 0; |
689 |
|
690 |
*addrp = a; |
691 |
bintrans_write_pc_inc(addrp); |
692 |
return 1; |
693 |
} |
694 |
|
695 |
|
696 |
/* |
697 |
* bintrans_write_instruction__addu_etc(): |
698 |
*/ |
699 |
static int bintrans_write_instruction__addu_etc(unsigned char **addrp, |
700 |
int rd, int rs, int rt, int sa, int instruction_type) |
701 |
{ |
702 |
unsigned char *a; |
703 |
int load64 = 0, do_store = 1; |
704 |
|
705 |
/* TODO: Not yet */ |
706 |
switch (instruction_type) { |
707 |
case SPECIAL_MULT: |
708 |
case SPECIAL_MULTU: |
709 |
case SPECIAL_DIV: |
710 |
case SPECIAL_DIVU: |
711 |
if (rd != 0) |
712 |
return 0; |
713 |
break; |
714 |
case SPECIAL_DSLL: |
715 |
case SPECIAL_DSLL32: |
716 |
case SPECIAL_DSRA: |
717 |
case SPECIAL_DSRA32: |
718 |
case SPECIAL_DSRL: |
719 |
case SPECIAL_DSRL32: |
720 |
case SPECIAL_MOVZ: |
721 |
case SPECIAL_MOVN: |
722 |
bintrans_write_chunkreturn_fail(addrp); |
723 |
return 0; |
724 |
} |
725 |
|
726 |
switch (instruction_type) { |
727 |
case SPECIAL_DADDU: |
728 |
case SPECIAL_DSUBU: |
729 |
case SPECIAL_OR: |
730 |
case SPECIAL_AND: |
731 |
case SPECIAL_NOR: |
732 |
case SPECIAL_XOR: |
733 |
case SPECIAL_DSLL: |
734 |
case SPECIAL_DSRL: |
735 |
case SPECIAL_DSRA: |
736 |
case SPECIAL_DSLL32: |
737 |
case SPECIAL_DSRL32: |
738 |
case SPECIAL_DSRA32: |
739 |
case SPECIAL_SLT: |
740 |
case SPECIAL_SLTU: |
741 |
load64 = 1; |
742 |
} |
743 |
|
744 |
switch (instruction_type) { |
745 |
case SPECIAL_MULT: |
746 |
case SPECIAL_MULTU: |
747 |
case SPECIAL_DIV: |
748 |
case SPECIAL_DIVU: |
749 |
break; |
750 |
default: |
751 |
if (rd == 0) |
752 |
goto rd0; |
753 |
} |
754 |
|
755 |
a = *addrp; |
756 |
|
757 |
if ((instruction_type == SPECIAL_ADDU || instruction_type == SPECIAL_DADDU |
758 |
|| instruction_type == SPECIAL_OR) && rt == 0) { |
759 |
if (load64) |
760 |
load_into_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); |
761 |
else |
762 |
load_into_eax_and_sign_extend_into_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); |
763 |
store_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rd]); |
764 |
*addrp = a; |
765 |
goto rd0; |
766 |
} |
767 |
|
768 |
/* edx:eax = rs, ecx:ebx = rt */ |
769 |
if (load64) { |
770 |
load_into_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rt]); |
771 |
/* 89 c3 mov %eax,%ebx */ |
772 |
/* 89 d1 mov %edx,%ecx */ |
773 |
*a++ = 0x89; *a++ = 0xc3; *a++ = 0x89; *a++ = 0xd1; |
774 |
load_into_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); |
775 |
} else { |
776 |
load_into_eax_and_sign_extend_into_edx(&a, &dummy_cpu.cd.mips.gpr[rt]); |
777 |
/* 89 c3 mov %eax,%ebx */ |
778 |
/* 89 d1 mov %edx,%ecx */ |
779 |
*a++ = 0x89; *a++ = 0xc3; *a++ = 0x89; *a++ = 0xd1; |
780 |
load_into_eax_and_sign_extend_into_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); |
781 |
} |
782 |
|
783 |
switch (instruction_type) { |
784 |
case SPECIAL_ADDU: |
785 |
/* 01 d8 add %ebx,%eax */ |
786 |
/* 99 cltd */ |
787 |
*a++ = 0x01; *a++ = 0xd8; |
788 |
*a++ = 0x99; |
789 |
break; |
790 |
case SPECIAL_DADDU: |
791 |
/* 01 d8 add %ebx,%eax */ |
792 |
/* 11 ca adc %ecx,%edx */ |
793 |
*a++ = 0x01; *a++ = 0xd8; |
794 |
*a++ = 0x11; *a++ = 0xca; |
795 |
break; |
796 |
case SPECIAL_SUBU: |
797 |
/* 29 d8 sub %ebx,%eax */ |
798 |
/* 99 cltd */ |
799 |
*a++ = 0x29; *a++ = 0xd8; |
800 |
*a++ = 0x99; |
801 |
break; |
802 |
case SPECIAL_DSUBU: |
803 |
/* 29 d8 sub %ebx,%eax */ |
804 |
/* 19 ca sbb %ecx,%edx */ |
805 |
*a++ = 0x29; *a++ = 0xd8; |
806 |
*a++ = 0x19; *a++ = 0xca; |
807 |
break; |
808 |
case SPECIAL_AND: |
809 |
/* 21 d8 and %ebx,%eax */ |
810 |
/* 21 ca and %ecx,%edx */ |
811 |
*a++ = 0x21; *a++ = 0xd8; |
812 |
*a++ = 0x21; *a++ = 0xca; |
813 |
break; |
814 |
case SPECIAL_OR: |
815 |
/* 09 d8 or %ebx,%eax */ |
816 |
/* 09 ca or %ecx,%edx */ |
817 |
*a++ = 0x09; *a++ = 0xd8; |
818 |
*a++ = 0x09; *a++ = 0xca; |
819 |
break; |
820 |
case SPECIAL_NOR: |
821 |
/* 09 d8 or %ebx,%eax */ |
822 |
/* 09 ca or %ecx,%edx */ |
823 |
/* f7 d0 not %eax */ |
824 |
/* f7 d2 not %edx */ |
825 |
*a++ = 0x09; *a++ = 0xd8; |
826 |
*a++ = 0x09; *a++ = 0xca; |
827 |
*a++ = 0xf7; *a++ = 0xd0; |
828 |
*a++ = 0xf7; *a++ = 0xd2; |
829 |
break; |
830 |
case SPECIAL_XOR: |
831 |
/* 31 d8 xor %ebx,%eax */ |
832 |
/* 31 ca xor %ecx,%edx */ |
833 |
*a++ = 0x31; *a++ = 0xd8; |
834 |
*a++ = 0x31; *a++ = 0xca; |
835 |
break; |
836 |
case SPECIAL_SLL: |
837 |
/* 89 d8 mov %ebx,%eax */ |
838 |
/* c1 e0 1f shl $0x1f,%eax */ |
839 |
/* 99 cltd */ |
840 |
*a++ = 0x89; *a++ = 0xd8; |
841 |
if (sa == 1) { |
842 |
*a++ = 0xd1; *a++ = 0xe0; |
843 |
} else { |
844 |
*a++ = 0xc1; *a++ = 0xe0; *a++ = sa; |
845 |
} |
846 |
*a++ = 0x99; |
847 |
break; |
848 |
case SPECIAL_SRA: |
849 |
/* 89 d8 mov %ebx,%eax */ |
850 |
/* c1 f8 1f sar $0x1f,%eax */ |
851 |
/* 99 cltd */ |
852 |
*a++ = 0x89; *a++ = 0xd8; |
853 |
if (sa == 1) { |
854 |
*a++ = 0xd1; *a++ = 0xf8; |
855 |
} else { |
856 |
*a++ = 0xc1; *a++ = 0xf8; *a++ = sa; |
857 |
} |
858 |
*a++ = 0x99; |
859 |
break; |
860 |
case SPECIAL_SRL: |
861 |
/* 89 d8 mov %ebx,%eax */ |
862 |
/* c1 e8 1f shr $0x1f,%eax */ |
863 |
/* 99 cltd */ |
864 |
*a++ = 0x89; *a++ = 0xd8; |
865 |
if (sa == 1) { |
866 |
*a++ = 0xd1; *a++ = 0xe8; |
867 |
} else { |
868 |
*a++ = 0xc1; *a++ = 0xe8; *a++ = sa; |
869 |
} |
870 |
*a++ = 0x99; |
871 |
break; |
872 |
case SPECIAL_SLTU: |
873 |
/* set if less than, unsigned. (compare edx:eax to ecx:ebx) */ |
874 |
/* if edx <= ecx and eax < ebx then 1, else 0. */ |
875 |
/* 39 ca cmp %ecx,%edx */ |
876 |
/* 77 0b ja <ret0> */ |
877 |
/* 39 d8 cmp %ebx,%eax */ |
878 |
/* 73 07 jae 58 <ret0> */ |
879 |
*a++ = 0x39; *a++ = 0xca; |
880 |
*a++ = 0x77; *a++ = 0x0b; |
881 |
*a++ = 0x39; *a++ = 0xd8; |
882 |
*a++ = 0x73; *a++ = 0x07; |
883 |
|
884 |
/* b8 01 00 00 00 mov $0x1,%eax */ |
885 |
/* eb 02 jmp <common> */ |
886 |
*a++ = 0xb8; *a++ = 1; *a++ = 0; *a++ = 0; *a++ = 0; |
887 |
*a++ = 0xeb; *a++ = 0x02; |
888 |
|
889 |
/* ret0: */ |
890 |
/* 29 c0 sub %eax,%eax */ |
891 |
*a++ = 0x29; *a++ = 0xc0; |
892 |
|
893 |
/* common: */ |
894 |
/* 99 cltd */ |
895 |
*a++ = 0x99; |
896 |
break; |
897 |
case SPECIAL_SLT: |
898 |
/* set if less than, signed. (compare edx:eax to ecx:ebx) */ |
899 |
/* if edx > ecx then 0. */ |
900 |
/* if edx < ecx then 1. */ |
901 |
/* if eax < ebx then 1, else 0. */ |
902 |
/* 39 ca cmp %ecx,%edx */ |
903 |
/* 7c 0a jl <ret1> */ |
904 |
/* 7f 04 jg <ret0> */ |
905 |
/* 39 d8 cmp %ebx,%eax */ |
906 |
/* 7c 04 jl <ret1> */ |
907 |
*a++ = 0x39; *a++ = 0xca; |
908 |
*a++ = 0x7c; *a++ = 0x0a; |
909 |
*a++ = 0x7f; *a++ = 0x04; |
910 |
*a++ = 0x39; *a++ = 0xd8; |
911 |
*a++ = 0x7c; *a++ = 0x04; |
912 |
|
913 |
/* ret0: */ |
914 |
/* 29 c0 sub %eax,%eax */ |
915 |
/* eb 05 jmp <common> */ |
916 |
*a++ = 0x29; *a++ = 0xc0; |
917 |
*a++ = 0xeb; *a++ = 0x05; |
918 |
|
919 |
/* ret1: */ |
920 |
/* b8 01 00 00 00 mov $0x1,%eax */ |
921 |
*a++ = 0xb8; *a++ = 1; *a++ = 0; *a++ = 0; *a++ = 0; |
922 |
|
923 |
/* common: */ |
924 |
/* 99 cltd */ |
925 |
*a++ = 0x99; |
926 |
break; |
927 |
case SPECIAL_SLLV: |
928 |
/* rd = rt << (rs&31) (logical) eax = ebx << (eax&31) */ |
929 |
/* xchg ebx,eax, then we can do eax = eax << (ebx&31) */ |
930 |
/* 93 xchg %eax,%ebx */ |
931 |
/* 89 d9 mov %ebx,%ecx */ |
932 |
/* 83 e1 1f and $0x1f,%ecx */ |
933 |
/* d3 e0 shl %cl,%eax */ |
934 |
*a++ = 0x93; |
935 |
*a++ = 0x89; *a++ = 0xd9; |
936 |
*a++ = 0x83; *a++ = 0xe1; *a++ = 0x1f; |
937 |
*a++ = 0xd3; *a++ = 0xe0; |
938 |
/* 99 cltd */ |
939 |
*a++ = 0x99; |
940 |
break; |
941 |
case SPECIAL_SRLV: |
942 |
/* rd = rt >> (rs&31) (logical) eax = ebx >> (eax&31) */ |
943 |
/* xchg ebx,eax, then we can do eax = eax >> (ebx&31) */ |
944 |
/* 93 xchg %eax,%ebx */ |
945 |
/* 89 d9 mov %ebx,%ecx */ |
946 |
/* 83 e1 1f and $0x1f,%ecx */ |
947 |
/* d3 e8 shr %cl,%eax */ |
948 |
*a++ = 0x93; |
949 |
*a++ = 0x89; *a++ = 0xd9; |
950 |
*a++ = 0x83; *a++ = 0xe1; *a++ = 0x1f; |
951 |
*a++ = 0xd3; *a++ = 0xe8; |
952 |
/* 99 cltd */ |
953 |
*a++ = 0x99; |
954 |
break; |
955 |
case SPECIAL_SRAV: |
956 |
/* rd = rt >> (rs&31) (arithmetic) eax = ebx >> (eax&31) */ |
957 |
/* xchg ebx,eax, then we can do eax = eax >> (ebx&31) */ |
958 |
/* 93 xchg %eax,%ebx */ |
959 |
/* 89 d9 mov %ebx,%ecx */ |
960 |
/* 83 e1 1f and $0x1f,%ecx */ |
961 |
/* d3 f8 sar %cl,%eax */ |
962 |
*a++ = 0x93; |
963 |
*a++ = 0x89; *a++ = 0xd9; |
964 |
*a++ = 0x83; *a++ = 0xe1; *a++ = 0x1f; |
965 |
*a++ = 0xd3; *a++ = 0xf8; |
966 |
/* 99 cltd */ |
967 |
*a++ = 0x99; |
968 |
break; |
969 |
case SPECIAL_MULT: |
970 |
case SPECIAL_MULTU: |
971 |
/* 57 push %edi */ |
972 |
*a++ = 0x57; |
973 |
if (instruction_type == SPECIAL_MULT) { |
974 |
/* f7 eb imul %ebx */ |
975 |
*a++ = 0xf7; *a++ = 0xeb; |
976 |
} else { |
977 |
/* f7 e3 mul %ebx */ |
978 |
*a++ = 0xf7; *a++ = 0xe3; |
979 |
} |
980 |
/* here: edx:eax = hi:lo */ |
981 |
/* 89 d7 mov %edx,%edi */ |
982 |
/* 99 cltd */ |
983 |
*a++ = 0x89; *a++ = 0xd7; |
984 |
*a++ = 0x99; |
985 |
/* here: edi=hi, edx:eax = sign-extended lo */ |
986 |
store_eax_edx(&a, &dummy_cpu.cd.mips.lo); |
987 |
/* 89 f8 mov %edi,%eax */ |
988 |
/* 99 cltd */ |
989 |
*a++ = 0x89; *a++ = 0xf8; |
990 |
*a++ = 0x99; |
991 |
/* here: edx:eax = sign-extended hi */ |
992 |
store_eax_edx(&a, &dummy_cpu.cd.mips.hi); |
993 |
/* 5f pop %edi */ |
994 |
*a++ = 0x5f; |
995 |
do_store = 0; |
996 |
break; |
997 |
case SPECIAL_DIV: |
998 |
case SPECIAL_DIVU: |
999 |
/* |
1000 |
* In: edx:eax = rs, ecx:ebx = rt |
1001 |
* Out: LO = rs / rt, HI = rs % rt |
1002 |
*/ |
1003 |
/* Division by zero on MIPS is undefined, but on |
1004 |
i386 it causes an exception, so we'll try to |
1005 |
avoid that. */ |
1006 |
*a++ = 0x83; *a++ = 0xfb; *a++ = 0x00; /* cmp $0x0,%ebx */ |
1007 |
*a++ = 0x75; *a++ = 0x01; /* jne skip_inc */ |
1008 |
*a++ = 0x43; /* inc %ebx */ |
1009 |
|
1010 |
/* 57 push %edi */ |
1011 |
*a++ = 0x57; |
1012 |
if (instruction_type == SPECIAL_DIV) { |
1013 |
*a++ = 0x99; /* cltd */ |
1014 |
*a++ = 0xf7; *a++ = 0xfb; /* idiv %ebx */ |
1015 |
} else { |
1016 |
*a++ = 0x29; *a++ = 0xd2; /* sub %edx,%edx */ |
1017 |
*a++ = 0xf7; *a++ = 0xf3; /* div %ebx */ |
1018 |
} |
1019 |
/* here: edx:eax = hi:lo */ |
1020 |
/* 89 d7 mov %edx,%edi */ |
1021 |
/* 99 cltd */ |
1022 |
*a++ = 0x89; *a++ = 0xd7; |
1023 |
*a++ = 0x99; |
1024 |
/* here: edi=hi, edx:eax = sign-extended lo */ |
1025 |
store_eax_edx(&a, &dummy_cpu.cd.mips.lo); |
1026 |
/* 89 f8 mov %edi,%eax */ |
1027 |
/* 99 cltd */ |
1028 |
*a++ = 0x89; *a++ = 0xf8; |
1029 |
*a++ = 0x99; |
1030 |
/* here: edx:eax = sign-extended hi */ |
1031 |
store_eax_edx(&a, &dummy_cpu.cd.mips.hi); |
1032 |
/* 5f pop %edi */ |
1033 |
*a++ = 0x5f; |
1034 |
do_store = 0; |
1035 |
break; |
1036 |
#if 0 |
1037 |
/* TODO: These are from bintrans_alpha.c. Translate them to i386. */ |
1038 |
|
1039 |
case SPECIAL_DSLL: |
1040 |
*a++ = 0x21; *a++ = 0x17 + ((sa & 7) << 5); *a++ = 0x40 + (sa >> 3); *a++ = 0x48; /* sll t1,sa,t0 */ |
1041 |
break; |
1042 |
case SPECIAL_DSLL32: |
1043 |
sa += 32; |
1044 |
*a++ = 0x21; *a++ = 0x17 + ((sa & 7) << 5); *a++ = 0x40 + (sa >> 3); *a++ = 0x48; /* sll t1,sa,t0 */ |
1045 |
break; |
1046 |
case SPECIAL_DSRA: |
1047 |
*a++ = 0x81; *a++ = 0x17 + ((sa & 7) << 5); *a++ = 0x40 + (sa >> 3); *a++ = 0x48; /* sra t1,sa,t0 */ |
1048 |
break; |
1049 |
case SPECIAL_DSRA32: |
1050 |
sa += 32; |
1051 |
*a++ = 0x81; *a++ = 0x17 + ((sa & 7) << 5); *a++ = 0x40 + (sa >> 3); *a++ = 0x48; /* sra t1,sa,t0 */ |
1052 |
break; |
1053 |
case SPECIAL_DSRL: |
1054 |
/* Note: bits of sa are distributed among two different bytes. */ |
1055 |
*a++ = 0x81; *a++ = 0x16 + ((sa & 7) << 5); *a++ = 0x40 + (sa >> 3); *a++ = 0x48; |
1056 |
break; |
1057 |
case SPECIAL_DSRL32: |
1058 |
/* Note: bits of sa are distributed among two different bytes. */ |
1059 |
sa += 32; |
1060 |
*a++ = 0x81; *a++ = 0x16 + ((sa & 7) << 5); *a++ = 0x40 + (sa >> 3); *a++ = 0x48; |
1061 |
break; |
1062 |
#endif |
1063 |
} |
1064 |
|
1065 |
if (do_store) |
1066 |
store_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rd]); |
1067 |
|
1068 |
*addrp = a; |
1069 |
rd0: |
1070 |
bintrans_write_pc_inc(addrp); |
1071 |
return 1; |
1072 |
} |
1073 |
|
1074 |
|
1075 |
/* |
1076 |
* bintrans_write_instruction__mfc_mtc(): |
1077 |
*/ |
1078 |
static int bintrans_write_instruction__mfc_mtc(struct memory *mem, |
1079 |
unsigned char **addrp, int coproc_nr, int flag64bit, int rt, |
1080 |
int rd, int mtcflag) |
1081 |
{ |
1082 |
unsigned char *a, *failskip; |
1083 |
int ofs; |
1084 |
|
1085 |
if (mtcflag && flag64bit) { |
1086 |
/* dmtc */ |
1087 |
return 0; |
1088 |
} |
1089 |
|
1090 |
/* |
1091 |
* NOTE: Only a few registers are readable without side effects. |
1092 |
*/ |
1093 |
if (rt == 0 && !mtcflag) |
1094 |
return 0; |
1095 |
|
1096 |
if (coproc_nr >= 1) |
1097 |
return 0; |
1098 |
|
1099 |
if (rd == COP0_RANDOM || rd == COP0_COUNT) |
1100 |
return 0; |
1101 |
|
1102 |
a = *addrp; |
1103 |
|
1104 |
/************************************************************* |
1105 |
* |
1106 |
* TODO: Check for kernel mode, or Coproc X usability bit! |
1107 |
* |
1108 |
*************************************************************/ |
1109 |
|
1110 |
/* 8b 96 3c 30 00 00 mov 0x303c(%esi),%edx */ |
1111 |
ofs = ((size_t)&dummy_cpu.cd.mips.coproc[0]) - (size_t)&dummy_cpu; |
1112 |
*a++ = 0x8b; *a++ = 0x96; |
1113 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
1114 |
|
1115 |
/* here, edx = cpu->coproc[0] */ |
1116 |
|
1117 |
if (mtcflag) { |
1118 |
/* mtc */ |
1119 |
|
1120 |
/* TODO: This code only works for mtc0, not dmtc0 */ |
1121 |
|
1122 |
/* 8b 9a 38 30 00 00 mov 0x3038(%edx),%ebx */ |
1123 |
ofs = ((size_t)&dummy_coproc.reg[rd]) - (size_t)&dummy_coproc; |
1124 |
*a++ = 0x8b; *a++ = 0x9a; |
1125 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
1126 |
|
1127 |
load_into_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rt]); |
1128 |
|
1129 |
/* |
1130 |
* Here: eax contains the value in register rt, |
1131 |
* ebx contains the coproc register rd value. |
1132 |
* |
1133 |
* In the general case, only allow mtc if it does not |
1134 |
* change the coprocessor register! |
1135 |
*/ |
1136 |
|
1137 |
switch (rd) { |
1138 |
|
1139 |
case COP0_INDEX: |
1140 |
break; |
1141 |
|
1142 |
case COP0_ENTRYLO0: |
1143 |
case COP0_ENTRYLO1: |
1144 |
/* TODO: Not all bits are writable! */ |
1145 |
break; |
1146 |
|
1147 |
case COP0_EPC: |
1148 |
break; |
1149 |
|
1150 |
case COP0_STATUS: |
1151 |
/* Only allow updates to the status register if |
1152 |
the interrupt enable bits were changed, but no |
1153 |
other bits! */ |
1154 |
/* 89 c1 mov %eax,%ecx */ |
1155 |
/* 89 da mov %ebx,%edx */ |
1156 |
/* 81 e1 00 00 e7 0f and $0x0fe70000,%ecx */ |
1157 |
/* 81 e2 00 00 e7 0f and $0x0fe70000,%edx */ |
1158 |
/* 39 ca cmp %ecx,%edx */ |
1159 |
/* 74 01 je <ok> */ |
1160 |
*a++ = 0x89; *a++ = 0xc1; |
1161 |
*a++ = 0x89; *a++ = 0xda; |
1162 |
*a++ = 0x81; *a++ = 0xe1; *a++ = 0x00; *a++ = 0x00; |
1163 |
if (mem->bintrans_32bit_only) { |
1164 |
*a++ = 0xe7; *a++ = 0x0f; |
1165 |
} else { |
1166 |
*a++ = 0xff; *a++ = 0xff; |
1167 |
} |
1168 |
*a++ = 0x81; *a++ = 0xe2; *a++ = 0x00; *a++ = 0x00; |
1169 |
if (mem->bintrans_32bit_only) { |
1170 |
*a++ = 0xe7; *a++ = 0x0f; |
1171 |
} else { |
1172 |
*a++ = 0xff; *a++ = 0xff; |
1173 |
} |
1174 |
*a++ = 0x39; *a++ = 0xca; |
1175 |
*a++ = 0x74; failskip = a; *a++ = 0x00; |
1176 |
bintrans_write_chunkreturn_fail(&a); |
1177 |
*failskip = (size_t)a - (size_t)failskip - 1; |
1178 |
|
1179 |
/* Only allow the update if it would NOT cause |
1180 |
an interrupt exception: */ |
1181 |
|
1182 |
/* 8b 96 3c 30 00 00 mov 0x303c(%esi),%edx */ |
1183 |
ofs = ((size_t)&dummy_cpu.cd.mips.coproc[0]) - (size_t)&dummy_cpu; |
1184 |
*a++ = 0x8b; *a++ = 0x96; |
1185 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
1186 |
|
1187 |
/* 8b 9a 38 30 00 00 mov 0x3038(%edx),%ebx */ |
1188 |
ofs = ((size_t)&dummy_coproc.reg[COP0_CAUSE]) - (size_t)&dummy_coproc; |
1189 |
*a++ = 0x8b; *a++ = 0x9a; |
1190 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
1191 |
|
1192 |
/* 21 c3 and %eax,%ebx */ |
1193 |
/* 81 e3 00 ff 00 00 and $0xff00,%ebx */ |
1194 |
/* 83 fb 00 cmp $0x0,%ebx */ |
1195 |
/* 74 01 je <ok> */ |
1196 |
*a++ = 0x21; *a++ = 0xc3; |
1197 |
*a++ = 0x81; *a++ = 0xe3; *a++ = 0x00; |
1198 |
*a++ = 0xff; *a++ = 0x00; *a++ = 0x00; |
1199 |
*a++ = 0x83; *a++ = 0xfb; *a++ = 0x00; |
1200 |
*a++ = 0x74; failskip = a; *a++ = 0x00; |
1201 |
bintrans_write_chunkreturn_fail(&a); |
1202 |
*failskip = (size_t)a - (size_t)failskip - 1; |
1203 |
|
1204 |
break; |
1205 |
|
1206 |
default: |
1207 |
/* 39 d8 cmp %ebx,%eax */ |
1208 |
/* 74 01 je <ok> */ |
1209 |
*a++ = 0x39; *a++ = 0xd8; |
1210 |
*a++ = 0x74; failskip = a; *a++ = 0x00; |
1211 |
bintrans_write_chunkreturn_fail(&a); |
1212 |
*failskip = (size_t)a - (size_t)failskip - 1; |
1213 |
} |
1214 |
|
1215 |
/* 8b 96 3c 30 00 00 mov 0x303c(%esi),%edx */ |
1216 |
ofs = ((size_t)&dummy_cpu.cd.mips.coproc[0]) - (size_t)&dummy_cpu; |
1217 |
*a++ = 0x8b; *a++ = 0x96; |
1218 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
1219 |
|
1220 |
/* 8d 9a 38 30 00 00 lea 0x3038(%edx),%ebx */ |
1221 |
ofs = ((size_t)&dummy_coproc.reg[rd]) - (size_t)&dummy_coproc; |
1222 |
*a++ = 0x8d; *a++ = 0x9a; |
1223 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
1224 |
|
1225 |
/* Sign-extend eax into edx:eax, and store it in |
1226 |
coprocessor register rd: */ |
1227 |
/* 99 cltd */ |
1228 |
*a++ = 0x99; |
1229 |
|
1230 |
/* 89 03 mov %eax,(%ebx) */ |
1231 |
/* 89 53 04 mov %edx,0x4(%ebx) */ |
1232 |
*a++ = 0x89; *a++ = 0x03; |
1233 |
*a++ = 0x89; *a++ = 0x53; *a++ = 0x04; |
1234 |
} else { |
1235 |
/* mfc */ |
1236 |
|
1237 |
/* 8b 82 38 30 00 00 mov 0x3038(%edx),%eax */ |
1238 |
ofs = ((size_t)&dummy_coproc.reg[rd]) - (size_t)&dummy_coproc; |
1239 |
*a++ = 0x8b; *a++ = 0x82; |
1240 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
1241 |
|
1242 |
if (flag64bit) { |
1243 |
/* Load high 32 bits: (note: edx gets overwritten) */ |
1244 |
/* 8b 92 3c 30 00 00 mov 0x303c(%edx),%edx */ |
1245 |
ofs += 4; |
1246 |
*a++ = 0x8b; *a++ = 0x92; |
1247 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
1248 |
} else { |
1249 |
/* 99 cltd */ |
1250 |
*a++ = 0x99; |
1251 |
} |
1252 |
|
1253 |
store_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rt]); |
1254 |
} |
1255 |
|
1256 |
*addrp = a; |
1257 |
bintrans_write_pc_inc(addrp); |
1258 |
return 1; |
1259 |
} |
1260 |
|
1261 |
|
1262 |
/* |
1263 |
* bintrans_write_instruction__branch(): |
1264 |
*/ |
1265 |
static int bintrans_write_instruction__branch(unsigned char **addrp, |
1266 |
int instruction_type, int regimm_type, int rt, int rs, int imm) |
1267 |
{ |
1268 |
unsigned char *a; |
1269 |
unsigned char *skip1 = NULL, *skip2 = NULL; |
1270 |
int ofs, likely = 0; |
1271 |
|
1272 |
switch (instruction_type) { |
1273 |
case HI6_BEQL: |
1274 |
case HI6_BNEL: |
1275 |
case HI6_BLEZL: |
1276 |
case HI6_BGTZL: |
1277 |
likely = 1; |
1278 |
} |
1279 |
|
1280 |
/* TODO: See the Alpha backend on how these could be implemented: */ |
1281 |
if (likely) |
1282 |
return 0; |
1283 |
|
1284 |
a = *addrp; |
1285 |
|
1286 |
/* |
1287 |
* edx:eax = gpr[rs]; ecx:ebx = gpr[rt]; |
1288 |
* |
1289 |
* Compare for equality (BEQ). |
1290 |
* If the result was zero, then it means equality; perform the |
1291 |
* delayed jump. Otherwise: skip. |
1292 |
*/ |
1293 |
|
1294 |
switch (instruction_type) { |
1295 |
case HI6_BEQ: |
1296 |
case HI6_BNE: |
1297 |
load_into_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rt]); |
1298 |
/* 89 c3 mov %eax,%ebx */ |
1299 |
/* 89 d1 mov %edx,%ecx */ |
1300 |
*a++ = 0x89; *a++ = 0xc3; *a++ = 0x89; *a++ = 0xd1; |
1301 |
} |
1302 |
load_into_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); |
1303 |
|
1304 |
if (instruction_type == HI6_BEQ && rt != rs) { |
1305 |
/* If rt != rs, then skip. */ |
1306 |
/* 39 c3 cmp %eax,%ebx */ |
1307 |
/* 75 05 jne 155 <skip> */ |
1308 |
/* 39 d1 cmp %edx,%ecx */ |
1309 |
/* 75 01 jne 155 <skip> */ |
1310 |
*a++ = 0x39; *a++ = 0xc3; |
1311 |
*a++ = 0x75; skip1 = a; *a++ = 0x00; |
1312 |
#if 0 |
1313 |
if (!bintrans_32bit_only) |
1314 |
#endif |
1315 |
{ |
1316 |
*a++ = 0x39; *a++ = 0xd1; |
1317 |
*a++ = 0x75; skip2 = a; *a++ = 0x00; |
1318 |
} |
1319 |
} |
1320 |
|
1321 |
if (instruction_type == HI6_BNE) { |
1322 |
/* If rt != rs, then ok. Otherwise skip. */ |
1323 |
#if 0 |
1324 |
if (bintrans_32bit_only) { |
1325 |
/* 39 c3 cmp %eax,%ebx */ |
1326 |
/* 74 xx je <skip> */ |
1327 |
*a++ = 0x39; *a++ = 0xc3; |
1328 |
*a++ = 0x74; skip2 = a; *a++ = 0x00; |
1329 |
} else |
1330 |
#endif |
1331 |
{ |
1332 |
/* 39 c3 cmp %eax,%ebx */ |
1333 |
/* 75 06 jne 156 <bra> */ |
1334 |
/* 39 d1 cmp %edx,%ecx */ |
1335 |
/* 75 02 jne 156 <bra> */ |
1336 |
/* eb 01 jmp 157 <skip> */ |
1337 |
*a++ = 0x39; *a++ = 0xc3; |
1338 |
*a++ = 0x75; *a++ = 0x06; |
1339 |
*a++ = 0x39; *a++ = 0xd1; |
1340 |
*a++ = 0x75; *a++ = 0x02; |
1341 |
*a++ = 0xeb; skip2 = a; *a++ = 0x00; |
1342 |
} |
1343 |
} |
1344 |
|
1345 |
if (instruction_type == HI6_BLEZ) { |
1346 |
/* If both eax and edx are zero, then do the branch. */ |
1347 |
/* 83 f8 00 cmp $0x0,%eax */ |
1348 |
/* 75 07 jne <nott> */ |
1349 |
/* 83 fa 00 cmp $0x0,%edx */ |
1350 |
/* 75 02 jne 23d <nott> */ |
1351 |
/* eb 01 jmp <branch> */ |
1352 |
*a++ = 0x83; *a++ = 0xf8; *a++ = 0x00; |
1353 |
*a++ = 0x75; *a++ = 0x07; |
1354 |
*a++ = 0x83; *a++ = 0xfa; *a++ = 0x00; |
1355 |
*a++ = 0x75; *a++ = 0x02; |
1356 |
*a++ = 0xeb; skip1 = a; *a++ = 0x00; |
1357 |
|
1358 |
/* If high bit of edx is set, then rs < 0. */ |
1359 |
/* f7 c2 00 00 00 80 test $0x80000000,%edx */ |
1360 |
/* 74 00 jz skip */ |
1361 |
*a++ = 0xf7; *a++ = 0xc2; *a++ = 0; *a++ = 0; *a++ = 0; *a++ = 0x80; |
1362 |
*a++ = 0x74; skip2 = a; *a++ = 0x00; |
1363 |
|
1364 |
if (skip1 != NULL) |
1365 |
*skip1 = (size_t)a - (size_t)skip1 - 1; |
1366 |
skip1 = NULL; |
1367 |
} |
1368 |
if (instruction_type == HI6_BGTZ) { |
1369 |
/* If both eax and edx are zero, then skip the branch. */ |
1370 |
/* 83 f8 00 cmp $0x0,%eax */ |
1371 |
/* 75 07 jne <nott> */ |
1372 |
/* 83 fa 00 cmp $0x0,%edx */ |
1373 |
/* 75 02 jne 23d <nott> */ |
1374 |
/* eb 01 jmp <skip> */ |
1375 |
*a++ = 0x83; *a++ = 0xf8; *a++ = 0x00; |
1376 |
*a++ = 0x75; *a++ = 0x07; |
1377 |
*a++ = 0x83; *a++ = 0xfa; *a++ = 0x00; |
1378 |
*a++ = 0x75; *a++ = 0x02; |
1379 |
*a++ = 0xeb; skip1 = a; *a++ = 0x00; |
1380 |
|
1381 |
/* If high bit of edx is set, then rs < 0. */ |
1382 |
/* f7 c2 00 00 00 80 test $0x80000000,%edx */ |
1383 |
/* 75 00 jnz skip */ |
1384 |
*a++ = 0xf7; *a++ = 0xc2; *a++ = 0; *a++ = 0; *a++ = 0; *a++ = 0x80; |
1385 |
*a++ = 0x75; skip2 = a; *a++ = 0x00; |
1386 |
} |
1387 |
if (instruction_type == HI6_REGIMM && regimm_type == REGIMM_BLTZ) { |
1388 |
/* If high bit of edx is set, then rs < 0. */ |
1389 |
/* f7 c2 00 00 00 80 test $0x80000000,%edx */ |
1390 |
/* 74 00 jz skip */ |
1391 |
*a++ = 0xf7; *a++ = 0xc2; *a++ = 0; *a++ = 0; *a++ = 0; *a++ = 0x80; |
1392 |
*a++ = 0x74; skip2 = a; *a++ = 0x00; |
1393 |
} |
1394 |
if (instruction_type == HI6_REGIMM && regimm_type == REGIMM_BGEZ) { |
1395 |
/* If high bit of edx is not set, then rs >= 0. */ |
1396 |
/* f7 c2 00 00 00 80 test $0x80000000,%edx */ |
1397 |
/* 75 00 jnz skip */ |
1398 |
*a++ = 0xf7; *a++ = 0xc2; *a++ = 0; *a++ = 0; *a++ = 0; *a++ = 0x80; |
1399 |
*a++ = 0x75; skip2 = a; *a++ = 0x00; |
1400 |
} |
1401 |
|
1402 |
/* |
1403 |
* Perform the jump by setting cpu->delay_slot = TO_BE_DELAYED |
1404 |
* and cpu->delay_jmpaddr = pc + 4 + (imm << 2). |
1405 |
*/ |
1406 |
|
1407 |
/* c7 86 38 30 00 00 01 00 00 00 movl $0x1,0x3038(%esi) */ |
1408 |
ofs = ((size_t)&dummy_cpu.cd.mips.delay_slot) - (size_t)&dummy_cpu; |
1409 |
*a++ = 0xc7; *a++ = 0x86; |
1410 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
1411 |
*a++ = TO_BE_DELAYED; *a++ = 0; *a++ = 0; *a++ = 0; |
1412 |
|
1413 |
load_pc_into_eax_edx(&a); |
1414 |
|
1415 |
/* 05 78 56 34 12 add $0x12345678,%eax */ |
1416 |
/* 83 d2 00 adc $0x0,%edx */ |
1417 |
/* or */ |
1418 |
/* 83 d2 ff adc $0xffffffff,%edx */ |
1419 |
imm = (imm << 2) + 4; |
1420 |
*a++ = 0x05; *a++ = imm; *a++ = imm >> 8; *a++ = imm >> 16; *a++ = imm >> 24; |
1421 |
if (imm >= 0) { |
1422 |
*a++ = 0x83; *a++ = 0xd2; *a++ = 0x00; |
1423 |
} else { |
1424 |
*a++ = 0x83; *a++ = 0xd2; *a++ = 0xff; |
1425 |
} |
1426 |
store_eax_edx(&a, &dummy_cpu.cd.mips.delay_jmpaddr); |
1427 |
|
1428 |
if (skip1 != NULL) |
1429 |
*skip1 = (size_t)a - (size_t)skip1 - 1; |
1430 |
if (skip2 != NULL) |
1431 |
*skip2 = (size_t)a - (size_t)skip2 - 1; |
1432 |
|
1433 |
*addrp = a; |
1434 |
bintrans_write_pc_inc(addrp); |
1435 |
return 1; |
1436 |
} |
1437 |
|
1438 |
|
1439 |
/* |
1440 |
* bintrans_write_instruction__delayedbranch(): |
1441 |
*/ |
1442 |
static int bintrans_write_instruction__delayedbranch(struct memory *mem, |
1443 |
unsigned char **addrp, uint32_t *potential_chunk_p, uint32_t *chunks, |
1444 |
int only_care_about_chunk_p, int p, int forward) |
1445 |
{ |
1446 |
unsigned char *a, *skip=NULL, *failskip; |
1447 |
int ofs; |
1448 |
uint32_t i386_addr; |
1449 |
|
1450 |
a = *addrp; |
1451 |
|
1452 |
if (only_care_about_chunk_p) |
1453 |
goto try_chunk_p; |
1454 |
|
1455 |
/* Skip all of this if there is no branch: */ |
1456 |
ofs = ((size_t)&dummy_cpu.cd.mips.delay_slot) - (size_t)&dummy_cpu; |
1457 |
|
1458 |
/* 8b 86 38 30 00 00 mov 0x3038(%esi),%eax */ |
1459 |
*a++ = 0x8b; *a++ = 0x86; |
1460 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
1461 |
|
1462 |
/* 83 f8 00 cmp $0x0,%eax */ |
1463 |
/* 74 01 je 16b <skippa> */ |
1464 |
*a++ = 0x83; *a++ = 0xf8; *a++ = 0x00; |
1465 |
*a++ = 0x74; skip = a; *a++ = 0; |
1466 |
|
1467 |
/* |
1468 |
* Perform the jump by setting cpu->delay_slot = 0 |
1469 |
* and pc = cpu->delay_jmpaddr. |
1470 |
*/ |
1471 |
|
1472 |
/* c7 86 38 30 00 00 00 00 00 00 movl $0x0,0x3038(%esi) */ |
1473 |
ofs = ((size_t)&dummy_cpu.cd.mips.delay_slot) - (size_t)&dummy_cpu; |
1474 |
*a++ = 0xc7; *a++ = 0x86; |
1475 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
1476 |
*a++ = 0; *a++ = 0; *a++ = 0; *a++ = 0; |
1477 |
|
1478 |
/* REMEMBER old pc: */ |
1479 |
load_pc_into_eax_edx(&a); |
1480 |
/* 89 c3 mov %eax,%ebx */ |
1481 |
/* 89 d1 mov %edx,%ecx */ |
1482 |
*a++ = 0x89; *a++ = 0xc3; |
1483 |
*a++ = 0x89; *a++ = 0xd1; |
1484 |
load_into_eax_edx(&a, &dummy_cpu.cd.mips.delay_jmpaddr); |
1485 |
store_eax_edx_into_pc(&a); |
1486 |
|
1487 |
try_chunk_p: |
1488 |
|
1489 |
if (potential_chunk_p == NULL) { |
1490 |
if (mem->bintrans_32bit_only) { |
1491 |
#if 1 |
1492 |
/* 8b 86 78 56 34 12 mov 0x12345678(%esi),%eax */ |
1493 |
/* ff e0 jmp *%eax */ |
1494 |
ofs = ((size_t)&dummy_cpu.cd.mips.bintrans_jump_to_32bit_pc) - (size_t)&dummy_cpu; |
1495 |
*a++ = 0x8b; *a++ = 0x86; |
1496 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
1497 |
*a++ = 0xff; *a++ = 0xe0; |
1498 |
|
1499 |
#else |
1500 |
/* Don't execute too many instructions. */ |
1501 |
/* 81 fd f0 1f 00 00 cmpl $0x1ff0,%ebp */ |
1502 |
/* 7c 01 jl <okk> */ |
1503 |
/* c3 ret */ |
1504 |
*a++ = 0x81; *a++ = 0xfd; |
1505 |
*a++ = (N_SAFE_BINTRANS_LIMIT-1) & 255; |
1506 |
*a++ = ((N_SAFE_BINTRANS_LIMIT-1) >> 8) & 255; *a++ = 0; *a++ = 0; |
1507 |
*a++ = 0x7c; failskip = a; *a++ = 0x01; |
1508 |
bintrans_write_chunkreturn_fail(&a); |
1509 |
*failskip = (size_t)a - (size_t)failskip - 1; |
1510 |
|
1511 |
/* |
1512 |
* ebx = ((vaddr >> 22) & 1023) * sizeof(void *) |
1513 |
* |
1514 |
* 89 c3 mov %eax,%ebx |
1515 |
* c1 eb 14 shr $20,%ebx |
1516 |
* 81 e3 fc 0f 00 00 and $0xffc,%ebx |
1517 |
*/ |
1518 |
*a++ = 0x89; *a++ = 0xc3; |
1519 |
*a++ = 0xc1; *a++ = 0xeb; *a++ = 0x14; |
1520 |
*a++ = 0x81; *a++ = 0xe3; *a++ = 0xfc; *a++ = 0x0f; *a++ = 0; *a++ = 0; |
1521 |
|
1522 |
/* |
1523 |
* ecx = vaddr_to_hostaddr_table0 |
1524 |
* |
1525 |
* 8b 8e 34 12 00 00 mov 0x1234(%esi),%ecx |
1526 |
*/ |
1527 |
ofs = ((size_t)&dummy_cpu.cd.mips.vaddr_to_hostaddr_table0) - (size_t)&dummy_cpu; |
1528 |
*a++ = 0x8b; *a++ = 0x8e; |
1529 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
1530 |
|
1531 |
/* |
1532 |
* ecx = vaddr_to_hostaddr_table0[a] |
1533 |
* |
1534 |
* 8b 0c 19 mov (%ecx,%ebx),%ecx |
1535 |
*/ |
1536 |
*a++ = 0x8b; *a++ = 0x0c; *a++ = 0x19; |
1537 |
|
1538 |
/* |
1539 |
* ebx = ((vaddr >> 12) & 1023) * sizeof(void *) |
1540 |
* |
1541 |
* 89 c3 mov %eax,%ebx |
1542 |
* c1 eb 0a shr $10,%ebx |
1543 |
* 81 e3 fc 0f 00 00 and $0xffc,%ebx |
1544 |
*/ |
1545 |
*a++ = 0x89; *a++ = 0xc3; |
1546 |
*a++ = 0xc1; *a++ = 0xeb; *a++ = 0x0a; |
1547 |
*a++ = 0x81; *a++ = 0xe3; *a++ = 0xfc; *a++ = 0x0f; *a++ = 0; *a++ = 0; |
1548 |
|
1549 |
/* |
1550 |
* ecx = vaddr_to_hostaddr_table0[a][b].cd.mips.chunks |
1551 |
* |
1552 |
* 8b 8c 19 56 34 12 00 mov 0x123456(%ecx,%ebx,1),%ecx |
1553 |
*/ |
1554 |
ofs = (size_t)&dummy_vth32_table.cd.mips.bintrans_chunks[0] |
1555 |
- (size_t)&dummy_vth32_table; |
1556 |
|
1557 |
*a++ = 0x8b; *a++ = 0x8c; *a++ = 0x19; |
1558 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
1559 |
|
1560 |
/* |
1561 |
* ecx = NULL? Then return with failure. |
1562 |
* |
1563 |
* 83 f9 00 cmp $0x0,%ecx |
1564 |
* 75 01 jne <okzzz> |
1565 |
*/ |
1566 |
*a++ = 0x83; *a++ = 0xf9; *a++ = 0x00; |
1567 |
*a++ = 0x75; fail = a; *a++ = 0x00; |
1568 |
bintrans_write_chunkreturn(&a); |
1569 |
*fail = (size_t)a - (size_t)fail - 1; |
1570 |
|
1571 |
/* |
1572 |
* 25 fc 0f 00 00 and $0xffc,%eax |
1573 |
* 01 c1 add %eax,%ecx |
1574 |
* |
1575 |
* 8b 01 mov (%ecx),%eax |
1576 |
* |
1577 |
* 83 f8 00 cmp $0x0,%eax |
1578 |
* 75 01 jne <ok> |
1579 |
* c3 ret |
1580 |
*/ |
1581 |
*a++ = 0x25; *a++ = 0xfc; *a++ = 0x0f; *a++ = 0; *a++ = 0; |
1582 |
*a++ = 0x01; *a++ = 0xc1; |
1583 |
|
1584 |
*a++ = 0x8b; *a++ = 0x01; |
1585 |
|
1586 |
*a++ = 0x83; *a++ = 0xf8; *a++ = 0x00; |
1587 |
*a++ = 0x75; fail = a; *a++ = 0x01; |
1588 |
bintrans_write_chunkreturn(&a); |
1589 |
*fail = (size_t)a - (size_t)fail - 1; |
1590 |
|
1591 |
/* 03 86 78 56 34 12 add 0x12345678(%esi),%eax */ |
1592 |
/* ff e0 jmp *%eax */ |
1593 |
ofs = ((size_t)&dummy_cpu.cd.mips.chunk_base_address) - (size_t)&dummy_cpu; |
1594 |
*a++ = 0x03; *a++ = 0x86; |
1595 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
1596 |
*a++ = 0xff; *a++ = 0xe0; |
1597 |
#endif |
1598 |
} else { |
1599 |
/* Not much we can do here if this wasn't to the same physical page... */ |
1600 |
|
1601 |
/* Don't execute too many instructions. */ |
1602 |
/* 81 fd f0 1f 00 00 cmpl $0x1ff0,%ebp */ |
1603 |
/* 7c 01 jl <okk> */ |
1604 |
/* c3 ret */ |
1605 |
*a++ = 0x81; *a++ = 0xfd; |
1606 |
*a++ = (N_SAFE_BINTRANS_LIMIT-1) & 255; |
1607 |
*a++ = ((N_SAFE_BINTRANS_LIMIT-1) >> 8) & 255; *a++ = 0; *a++ = 0; |
1608 |
*a++ = 0x7c; failskip = a; *a++ = 0x01; |
1609 |
bintrans_write_chunkreturn_fail(&a); |
1610 |
*failskip = (size_t)a - (size_t)failskip - 1; |
1611 |
|
1612 |
/* |
1613 |
* Compare the old pc (ecx:ebx) and the new pc (edx:eax). If they are on the |
1614 |
* same virtual page (which means that they are on the same physical |
1615 |
* page), then we can check the right chunk pointer, and if it |
1616 |
* is non-NULL, then we can jump there. Otherwise just return. |
1617 |
*/ |
1618 |
|
1619 |
/* Subtract 4 from the old pc first. (This is where the jump originated from.) */ |
1620 |
/* 83 eb 04 sub $0x4,%ebx */ |
1621 |
/* 83 d9 00 sbb $0x0,%ecx */ |
1622 |
*a++ = 0x83; *a++ = 0xeb; *a++ = 0x04; |
1623 |
*a++ = 0x83; *a++ = 0xd9; *a++ = 0x00; |
1624 |
|
1625 |
/* 39 d1 cmp %edx,%ecx */ |
1626 |
/* 74 01 je 1b9 <ok2> */ |
1627 |
/* c3 ret */ |
1628 |
*a++ = 0x39; *a++ = 0xd1; |
1629 |
*a++ = 0x74; *a++ = 0x01; |
1630 |
*a++ = 0xc3; |
1631 |
|
1632 |
/* Remember new pc: */ |
1633 |
/* 89 c1 mov %eax,%ecx */ |
1634 |
*a++ = 0x89; *a++ = 0xc1; |
1635 |
|
1636 |
/* 81 e3 00 f0 ff ff and $0xfffff000,%ebx */ |
1637 |
/* 25 00 f0 ff ff and $0xfffff000,%eax */ |
1638 |
*a++ = 0x81; *a++ = 0xe3; *a++ = 0x00; *a++ = 0xf0; *a++ = 0xff; *a++ = 0xff; |
1639 |
*a++ = 0x25; *a++ = 0x00; *a++ = 0xf0; *a++ = 0xff; *a++ = 0xff; |
1640 |
|
1641 |
/* 39 c3 cmp %eax,%ebx */ |
1642 |
/* 74 01 je <ok1> */ |
1643 |
/* c3 ret */ |
1644 |
*a++ = 0x39; *a++ = 0xc3; |
1645 |
*a++ = 0x74; *a++ = 0x01; |
1646 |
*a++ = 0xc3; |
1647 |
|
1648 |
/* 81 e1 ff 0f 00 00 and $0xfff,%ecx */ |
1649 |
*a++ = 0x81; *a++ = 0xe1; *a++ = 0xff; *a++ = 0x0f; *a++ = 0; *a++ = 0; |
1650 |
|
1651 |
/* 8b 81 78 56 34 12 mov 0x12345678(%ecx),%eax */ |
1652 |
ofs = (size_t)chunks; |
1653 |
*a++ = 0x8b; *a++ = 0x81; *a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
1654 |
|
1655 |
/* 83 f8 00 cmp $0x0,%eax */ |
1656 |
/* 75 01 jne 1cd <okjump> */ |
1657 |
/* c3 ret */ |
1658 |
*a++ = 0x83; *a++ = 0xf8; *a++ = 0x00; |
1659 |
*a++ = 0x75; *a++ = 0x01; |
1660 |
*a++ = 0xc3; |
1661 |
|
1662 |
/* 03 86 78 56 34 12 add 0x12345678(%esi),%eax */ |
1663 |
/* ff e0 jmp *%eax */ |
1664 |
ofs = ((size_t)&dummy_cpu.cd.mips.chunk_base_address) - (size_t)&dummy_cpu; |
1665 |
*a++ = 0x03; *a++ = 0x86; |
1666 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
1667 |
*a++ = 0xff; *a++ = 0xe0; |
1668 |
} |
1669 |
} else { |
1670 |
/* |
1671 |
* Just to make sure that we don't become too unreliant |
1672 |
* on the main program loop, we need to return every once |
1673 |
* in a while (interrupts etc). |
1674 |
* |
1675 |
* Load the "nr of instructions executed" (which is an int) |
1676 |
* and see if it is below a certain threshold. If so, then |
1677 |
* we go on with the fast path (bintrans), otherwise we |
1678 |
* abort by returning. |
1679 |
*/ |
1680 |
/* 81 fd f0 1f 00 00 cmpl $0x1ff0,%ebp */ |
1681 |
/* 7c 01 jl <okk> */ |
1682 |
/* c3 ret */ |
1683 |
if (!only_care_about_chunk_p && !forward) { |
1684 |
*a++ = 0x81; *a++ = 0xfd; |
1685 |
*a++ = (N_SAFE_BINTRANS_LIMIT-1) & 255; |
1686 |
*a++ = ((N_SAFE_BINTRANS_LIMIT-1) >> 8) & 255; *a++ = 0; *a++ = 0; |
1687 |
*a++ = 0x7c; failskip = a; *a++ = 0x01; |
1688 |
bintrans_write_chunkreturn_fail(&a); |
1689 |
*failskip = (size_t)a - (size_t)failskip - 1; |
1690 |
} |
1691 |
|
1692 |
/* |
1693 |
* potential_chunk_p points to an "uint32_t". |
1694 |
* If this value is non-NULL, then it is a piece of i386 |
1695 |
* machine language code corresponding to the address |
1696 |
* we're jumping to. Otherwise, those instructions haven't |
1697 |
* been translated yet, so we have to return to the main |
1698 |
* loop. (Actually, we have to add cpu->chunk_base_address.) |
1699 |
* |
1700 |
* Case 1: The value is non-NULL already at translation |
1701 |
* time. Then we can make a direct (fast) native |
1702 |
* i386 jump to the code chunk. |
1703 |
* |
1704 |
* Case 2: The value was NULL at translation time, then we |
1705 |
* have to check during runtime. |
1706 |
*/ |
1707 |
|
1708 |
/* Case 1: */ |
1709 |
/* printf("%08x ", *potential_chunk_p); */ |
1710 |
i386_addr = *potential_chunk_p + |
1711 |
(size_t)mem->translation_code_chunk_space; |
1712 |
i386_addr = i386_addr - ((size_t)a + 5); |
1713 |
if ((*potential_chunk_p) != 0) { |
1714 |
*a++ = 0xe9; |
1715 |
*a++ = i386_addr; |
1716 |
*a++ = i386_addr >> 8; |
1717 |
*a++ = i386_addr >> 16; |
1718 |
*a++ = i386_addr >> 24; |
1719 |
} else { |
1720 |
/* Case 2: */ |
1721 |
|
1722 |
bintrans_register_potential_quick_jump(mem, a, p); |
1723 |
|
1724 |
i386_addr = (size_t)potential_chunk_p; |
1725 |
|
1726 |
/* |
1727 |
* Load the chunk pointer into eax. |
1728 |
* If it is NULL (zero), then skip the following jump. |
1729 |
* Add chunk_base_address to eax, and jump to eax. |
1730 |
*/ |
1731 |
|
1732 |
/* a1 78 56 34 12 mov 0x12345678,%eax */ |
1733 |
/* 83 f8 00 cmp $0x0,%eax */ |
1734 |
/* 75 01 jne <okaa> */ |
1735 |
/* c3 ret */ |
1736 |
*a++ = 0xa1; |
1737 |
*a++ = i386_addr; *a++ = i386_addr >> 8; |
1738 |
*a++ = i386_addr >> 16; *a++ = i386_addr >> 24; |
1739 |
*a++ = 0x83; *a++ = 0xf8; *a++ = 0x00; |
1740 |
*a++ = 0x75; *a++ = 0x01; |
1741 |
*a++ = 0xc3; |
1742 |
|
1743 |
/* 03 86 78 56 34 12 add 0x12345678(%esi),%eax */ |
1744 |
/* ff e0 jmp *%eax */ |
1745 |
ofs = ((size_t)&dummy_cpu.cd.mips.chunk_base_address) - (size_t)&dummy_cpu; |
1746 |
*a++ = 0x03; *a++ = 0x86; |
1747 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
1748 |
*a++ = 0xff; *a++ = 0xe0; |
1749 |
} |
1750 |
} |
1751 |
|
1752 |
if (skip != NULL) |
1753 |
*skip = (size_t)a - (size_t)skip - 1; |
1754 |
|
1755 |
*addrp = a; |
1756 |
return 1; |
1757 |
} |
1758 |
|
1759 |
|
1760 |
/* |
1761 |
* bintrans_write_instruction__loadstore(): |
1762 |
*/ |
1763 |
static int bintrans_write_instruction__loadstore(struct memory *mem, |
1764 |
unsigned char **addrp, int rt, int imm, int rs, |
1765 |
int instruction_type, int bigendian) |
1766 |
{ |
1767 |
unsigned char *a, *retfail, *generic64bit, *doloadstore, |
1768 |
*okret0, *okret1, *okret2, *skip; |
1769 |
int ofs, alignment, load=0, unaligned=0; |
1770 |
|
1771 |
/* TODO: Not yet: */ |
1772 |
if (instruction_type == HI6_LQ_MDMX || instruction_type == HI6_SQ) |
1773 |
return 0; |
1774 |
|
1775 |
/* TODO: Not yet: */ |
1776 |
if (bigendian) |
1777 |
return 0; |
1778 |
|
1779 |
switch (instruction_type) { |
1780 |
case HI6_LQ_MDMX: |
1781 |
case HI6_LDL: |
1782 |
case HI6_LDR: |
1783 |
case HI6_LD: |
1784 |
case HI6_LWU: |
1785 |
case HI6_LWL: |
1786 |
case HI6_LWR: |
1787 |
case HI6_LW: |
1788 |
case HI6_LHU: |
1789 |
case HI6_LH: |
1790 |
case HI6_LBU: |
1791 |
case HI6_LB: |
1792 |
load = 1; |
1793 |
if (rt == 0) |
1794 |
return 0; |
1795 |
} |
1796 |
|
1797 |
switch (instruction_type) { |
1798 |
case HI6_LWL: |
1799 |
case HI6_LWR: |
1800 |
case HI6_LDL: |
1801 |
case HI6_LDR: |
1802 |
case HI6_SWL: |
1803 |
case HI6_SWR: |
1804 |
case HI6_SDL: |
1805 |
case HI6_SDR: |
1806 |
unaligned = 1; |
1807 |
} |
1808 |
|
1809 |
a = *addrp; |
1810 |
|
1811 |
if (mem->bintrans_32bit_only) |
1812 |
load_into_eax_dont_care_about_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); |
1813 |
else |
1814 |
load_into_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); |
1815 |
|
1816 |
if (imm & 0x8000) { |
1817 |
/* 05 34 f2 ff ff add $0xfffff234,%eax */ |
1818 |
/* 83 d2 ff adc $0xffffffff,%edx */ |
1819 |
*a++ = 5; |
1820 |
*a++ = imm; *a++ = imm >> 8; *a++ = 0xff; *a++ = 0xff; |
1821 |
if (!mem->bintrans_32bit_only) { |
1822 |
*a++ = 0x83; *a++ = 0xd2; *a++ = 0xff; |
1823 |
} |
1824 |
} else { |
1825 |
/* 05 34 12 00 00 add $0x1234,%eax */ |
1826 |
/* 83 d2 00 adc $0x0,%edx */ |
1827 |
*a++ = 5; |
1828 |
*a++ = imm; *a++ = imm >> 8; *a++ = 0; *a++ = 0; |
1829 |
if (!mem->bintrans_32bit_only) { |
1830 |
*a++ = 0x83; *a++ = 0xd2; *a++ = 0; |
1831 |
} |
1832 |
} |
1833 |
|
1834 |
alignment = 0; |
1835 |
switch (instruction_type) { |
1836 |
case HI6_LQ_MDMX: |
1837 |
case HI6_SQ: |
1838 |
alignment = 15; |
1839 |
break; |
1840 |
case HI6_LD: |
1841 |
case HI6_LDL: |
1842 |
case HI6_LDR: |
1843 |
case HI6_SD: |
1844 |
case HI6_SDL: |
1845 |
case HI6_SDR: |
1846 |
alignment = 7; |
1847 |
break; |
1848 |
case HI6_LW: |
1849 |
case HI6_LWL: |
1850 |
case HI6_LWR: |
1851 |
case HI6_LWU: |
1852 |
case HI6_SW: |
1853 |
case HI6_SWL: |
1854 |
case HI6_SWR: |
1855 |
alignment = 3; |
1856 |
break; |
1857 |
case HI6_LH: |
1858 |
case HI6_LHU: |
1859 |
case HI6_SH: |
1860 |
alignment = 1; |
1861 |
break; |
1862 |
} |
1863 |
|
1864 |
if (unaligned) { |
1865 |
/* |
1866 |
* Perform the actual load/store from an |
1867 |
* aligned address. |
1868 |
* |
1869 |
* 83 e0 fc and $0xfffffffc,%eax |
1870 |
*/ |
1871 |
*a++ = 0x83; *a++ = 0xe0; *a++ = 0xff - alignment; |
1872 |
} else if (alignment > 0) { |
1873 |
unsigned char *alignskip; |
1874 |
/* |
1875 |
* Check alignment: |
1876 |
* |
1877 |
* 89 c3 mov %eax,%ebx |
1878 |
* 83 e3 01 and $0x1,%ebx |
1879 |
* 74 01 jz <ok> |
1880 |
* c3 ret |
1881 |
*/ |
1882 |
*a++ = 0x89; *a++ = 0xc3; |
1883 |
*a++ = 0x83; *a++ = 0xe3; *a++ = alignment; |
1884 |
*a++ = 0x74; alignskip = a; *a++ = 0x00; |
1885 |
bintrans_write_chunkreturn_fail(&a); |
1886 |
*alignskip = (size_t)a - (size_t)alignskip - 1; |
1887 |
} |
1888 |
|
1889 |
|
1890 |
/* Here, edx:eax = vaddr */ |
1891 |
|
1892 |
if (mem->bintrans_32bit_only) { |
1893 |
/* Call the quick lookup routine: */ |
1894 |
if (load) |
1895 |
ofs = (size_t)bintrans_load_32bit; |
1896 |
else |
1897 |
ofs = (size_t)bintrans_store_32bit; |
1898 |
ofs = ofs - ((size_t)a + 5); |
1899 |
*a++ = 0xe8; *a++ = ofs; *a++ = ofs >> 8; |
1900 |
*a++ = ofs >> 16; *a++ = ofs >> 24; |
1901 |
|
1902 |
/* |
1903 |
* ecx = NULL? Then return with failure. |
1904 |
* |
1905 |
* 83 f9 00 cmp $0x0,%ecx |
1906 |
* 75 01 jne <okzzz> |
1907 |
*/ |
1908 |
*a++ = 0x83; *a++ = 0xf9; *a++ = 0x00; |
1909 |
*a++ = 0x75; retfail = a; *a++ = 0x00; |
1910 |
bintrans_write_chunkreturn_fail(&a); /* ret (and fail) */ |
1911 |
*retfail = (size_t)a - (size_t)retfail - 1; |
1912 |
|
1913 |
/* |
1914 |
* eax = offset within page = vaddr & 0xfff |
1915 |
* |
1916 |
* 25 ff 0f 00 00 and $0xfff,%eax |
1917 |
*/ |
1918 |
*a++ = 0x25; *a++ = 0xff; *a++ = 0x0f; *a++ = 0; *a++ = 0; |
1919 |
|
1920 |
/* |
1921 |
* ecx = host address ( = host page + offset) |
1922 |
* |
1923 |
* 83 e1 fe and $0xfffffffe,%ecx clear the lowest bit |
1924 |
* 01 c1 add %eax,%ecx |
1925 |
*/ |
1926 |
*a++ = 0x83; *a++ = 0xe1; *a++ = 0xfe; |
1927 |
*a++ = 0x01; *a++ = 0xc1; |
1928 |
} else { |
1929 |
/* |
1930 |
* If the load/store address has the top 32 bits set to |
1931 |
* 0x00000000 or 0xffffffff, then we can use the 32-bit |
1932 |
* lookup tables: |
1933 |
* |
1934 |
|
1935 |
TODO: top 33 bits!!!!!!! |
1936 |
|
1937 |
* 83 fa 00 cmp $0x0,%edx |
1938 |
* 74 05 je <ok32> |
1939 |
* 83 fa ff cmp $0xffffffff,%edx |
1940 |
* 75 01 jne <not32> |
1941 |
*/ |
1942 |
*a++ = 0x83; *a++ = 0xfa; *a++ = 0x00; |
1943 |
*a++ = 0x74; *a++ = 0x05; |
1944 |
*a++ = 0x83; *a++ = 0xfa; *a++ = 0xff; |
1945 |
*a++ = 0x75; generic64bit = a; *a++ = 0x01; |
1946 |
|
1947 |
/* Call the quick lookup routine: */ |
1948 |
if (load) |
1949 |
ofs = (size_t)bintrans_load_32bit; |
1950 |
else |
1951 |
ofs = (size_t)bintrans_store_32bit; |
1952 |
ofs = ofs - ((size_t)a + 5); |
1953 |
*a++ = 0xe8; *a++ = ofs; *a++ = ofs >> 8; |
1954 |
*a++ = ofs >> 16; *a++ = ofs >> 24; |
1955 |
|
1956 |
/* |
1957 |
* ecx = NULL? Then return with failure. |
1958 |
* |
1959 |
* 83 f9 00 cmp $0x0,%ecx |
1960 |
* 75 01 jne <okzzz> |
1961 |
*/ |
1962 |
*a++ = 0x83; *a++ = 0xf9; *a++ = 0x00; |
1963 |
*a++ = 0x75; retfail = a; *a++ = 0x00; |
1964 |
bintrans_write_chunkreturn_fail(&a); /* ret (and fail) */ |
1965 |
*retfail = (size_t)a - (size_t)retfail - 1; |
1966 |
|
1967 |
/* |
1968 |
* eax = offset within page = vaddr & 0xfff |
1969 |
* |
1970 |
* 25 ff 0f 00 00 and $0xfff,%eax |
1971 |
*/ |
1972 |
*a++ = 0x25; *a++ = 0xff; *a++ = 0x0f; *a++ = 0; *a++ = 0; |
1973 |
|
1974 |
/* |
1975 |
* ecx = host address ( = host page + offset) |
1976 |
* |
1977 |
* 83 e1 fe and $0xfffffffe,%ecx clear the lowest bit |
1978 |
* 01 c1 add %eax,%ecx |
1979 |
*/ |
1980 |
*a++ = 0x83; *a++ = 0xe1; *a++ = 0xfe; |
1981 |
*a++ = 0x01; *a++ = 0xc1; |
1982 |
|
1983 |
*a++ = 0xeb; doloadstore = a; *a++ = 0x01; |
1984 |
|
1985 |
|
1986 |
/* TODO: The stuff above is so similar to the pure 32-bit |
1987 |
case that it should be factored out. */ |
1988 |
|
1989 |
|
1990 |
*generic64bit = (size_t)a - (size_t)generic64bit - 1; |
1991 |
|
1992 |
/* |
1993 |
* 64-bit generic case: |
1994 |
*/ |
1995 |
|
1996 |
/* push writeflag */ |
1997 |
*a++ = 0x6a; *a++ = load? 0 : 1; |
1998 |
|
1999 |
/* push vaddr (edx:eax) */ |
2000 |
*a++ = 0x52; *a++ = 0x50; |
2001 |
|
2002 |
/* push cpu (esi) */ |
2003 |
*a++ = 0x56; |
2004 |
|
2005 |
/* eax = points to the right function */ |
2006 |
ofs = ((size_t)&dummy_cpu.cd.mips.fast_vaddr_to_hostaddr) - (size_t)&dummy_cpu; |
2007 |
*a++ = 0x8b; *a++ = 0x86; |
2008 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
2009 |
|
2010 |
/* ff d0 call *%eax */ |
2011 |
*a++ = 0xff; *a++ = 0xd0; |
2012 |
|
2013 |
/* 83 c4 08 add $0x10,%esp */ |
2014 |
*a++ = 0x83; *a++ = 0xc4; *a++ = 0x10; |
2015 |
|
2016 |
/* If eax is NULL, then return. */ |
2017 |
/* 83 f8 00 cmp $0x0,%eax */ |
2018 |
/* 75 01 jne 1cd <okjump> */ |
2019 |
/* c3 ret */ |
2020 |
*a++ = 0x83; *a++ = 0xf8; *a++ = 0x00; |
2021 |
*a++ = 0x75; retfail = a; *a++ = 0x00; |
2022 |
bintrans_write_chunkreturn_fail(&a); /* ret (and fail) */ |
2023 |
*retfail = (size_t)a - (size_t)retfail - 1; |
2024 |
|
2025 |
/* 89 c1 mov %eax,%ecx */ |
2026 |
*a++ = 0x89; *a++ = 0xc1; |
2027 |
|
2028 |
*doloadstore = (size_t)a - (size_t)doloadstore - 1; |
2029 |
} |
2030 |
|
2031 |
|
2032 |
if (!load) { |
2033 |
if (alignment >= 7) |
2034 |
load_into_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rt]); |
2035 |
else |
2036 |
load_into_eax_dont_care_about_edx(&a, &dummy_cpu.cd.mips.gpr[rt]); |
2037 |
} |
2038 |
|
2039 |
switch (instruction_type) { |
2040 |
case HI6_LD: |
2041 |
/* 8b 01 mov (%ecx),%eax */ |
2042 |
/* 8b 51 04 mov 0x4(%ecx),%edx */ |
2043 |
*a++ = 0x8b; *a++ = 0x01; |
2044 |
*a++ = 0x8b; *a++ = 0x51; *a++ = 0x04; |
2045 |
break; |
2046 |
case HI6_LWU: |
2047 |
/* 8b 01 mov (%ecx),%eax */ |
2048 |
/* 31 d2 xor %edx,%edx */ |
2049 |
*a++ = 0x8b; *a++ = 0x01; |
2050 |
*a++ = 0x31; *a++ = 0xd2; |
2051 |
break; |
2052 |
case HI6_LW: |
2053 |
/* 8b 01 mov (%ecx),%eax */ |
2054 |
/* 99 cltd */ |
2055 |
*a++ = 0x8b; *a++ = 0x01; |
2056 |
*a++ = 0x99; |
2057 |
break; |
2058 |
case HI6_LHU: |
2059 |
/* 31 c0 xor %eax,%eax */ |
2060 |
/* 66 8b 01 mov (%ecx),%ax */ |
2061 |
/* 99 cltd */ |
2062 |
*a++ = 0x31; *a++ = 0xc0; |
2063 |
*a++ = 0x66; *a++ = 0x8b; *a++ = 0x01; |
2064 |
*a++ = 0x99; |
2065 |
break; |
2066 |
case HI6_LH: |
2067 |
/* 66 8b 01 mov (%ecx),%ax */ |
2068 |
/* 98 cwtl */ |
2069 |
/* 99 cltd */ |
2070 |
*a++ = 0x66; *a++ = 0x8b; *a++ = 0x01; |
2071 |
*a++ = 0x98; |
2072 |
*a++ = 0x99; |
2073 |
break; |
2074 |
case HI6_LBU: |
2075 |
/* 31 c0 xor %eax,%eax */ |
2076 |
/* 8a 01 mov (%ecx),%al */ |
2077 |
/* 99 cltd */ |
2078 |
*a++ = 0x31; *a++ = 0xc0; |
2079 |
*a++ = 0x8a; *a++ = 0x01; |
2080 |
*a++ = 0x99; |
2081 |
break; |
2082 |
case HI6_LB: |
2083 |
/* 8a 01 mov (%ecx),%al */ |
2084 |
/* 66 98 cbtw */ |
2085 |
/* 98 cwtl */ |
2086 |
/* 99 cltd */ |
2087 |
*a++ = 0x8a; *a++ = 0x01; |
2088 |
*a++ = 0x66; *a++ = 0x98; |
2089 |
*a++ = 0x98; |
2090 |
*a++ = 0x99; |
2091 |
break; |
2092 |
|
2093 |
case HI6_LWL: |
2094 |
load_into_eax_dont_care_about_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); |
2095 |
/* 05 34 f2 ff ff add $0xfffff234,%eax */ |
2096 |
*a++ = 5; |
2097 |
*a++ = imm; *a++ = imm >> 8; *a++ = 0xff; *a++ = 0xff; |
2098 |
/* 83 e0 03 and $0x03,%eax */ |
2099 |
*a++ = 0x83; *a++ = 0xe0; *a++ = alignment; |
2100 |
/* 89 c3 mov %eax,%ebx */ |
2101 |
*a++ = 0x89; *a++ = 0xc3; |
2102 |
|
2103 |
load_into_eax_dont_care_about_edx(&a, &dummy_cpu.cd.mips.gpr[rt]); |
2104 |
|
2105 |
/* ALIGNED LOAD: */ |
2106 |
/* 8b 11 mov (%ecx),%edx */ |
2107 |
*a++ = 0x8b; *a++ = 0x11; |
2108 |
|
2109 |
/* |
2110 |
* CASE 0: |
2111 |
* memory = 0x12 0x34 0x56 0x78 |
2112 |
* register after lwl: 0x12 0x.. 0x.. 0x.. |
2113 |
*/ |
2114 |
/* 83 fb 00 cmp $0x0,%ebx */ |
2115 |
/* 75 01 jne <skip> */ |
2116 |
*a++ = 0x83; *a++ = 0xfb; *a++ = 0x00; |
2117 |
*a++ = 0x75; skip = a; *a++ = 0x01; |
2118 |
|
2119 |
/* c1 e2 18 shl $0x18,%edx */ |
2120 |
/* 25 ff ff ff 00 and $0xffffff,%eax */ |
2121 |
/* 09 d0 or %edx,%eax */ |
2122 |
*a++ = 0xc1; *a++ = 0xe2; *a++ = 0x18; |
2123 |
*a++ = 0x25; *a++ = 0xff; *a++ = 0xff; *a++ = 0xff; *a++ = 0x00; |
2124 |
*a++ = 0x09; *a++ = 0xd0; |
2125 |
|
2126 |
/* eb 00 jmp <okret> */ |
2127 |
*a++ = 0xeb; okret0 = a; *a++ = 0; |
2128 |
|
2129 |
*skip = (size_t)a - (size_t)skip - 1; |
2130 |
|
2131 |
/* |
2132 |
* CASE 1: |
2133 |
* memory = 0x12 0x34 0x56 0x78 |
2134 |
* register after lwl: 0x34 0x12 0x.. 0x.. |
2135 |
*/ |
2136 |
/* 83 fb 01 cmp $0x1,%ebx */ |
2137 |
/* 75 01 jne <skip> */ |
2138 |
*a++ = 0x83; *a++ = 0xfb; *a++ = 0x01; |
2139 |
*a++ = 0x75; skip = a; *a++ = 0x01; |
2140 |
|
2141 |
/* c1 e2 10 shl $0x10,%edx */ |
2142 |
/* 25 ff ff 00 00 and $0xffff,%eax */ |
2143 |
/* 09 d0 or %edx,%eax */ |
2144 |
*a++ = 0xc1; *a++ = 0xe2; *a++ = 0x10; |
2145 |
*a++ = 0x25; *a++ = 0xff; *a++ = 0xff; *a++ = 0x00; *a++ = 0x00; |
2146 |
*a++ = 0x09; *a++ = 0xd0; |
2147 |
|
2148 |
/* eb 00 jmp <okret> */ |
2149 |
*a++ = 0xeb; okret1 = a; *a++ = 0; |
2150 |
|
2151 |
*skip = (size_t)a - (size_t)skip - 1; |
2152 |
|
2153 |
/* |
2154 |
* CASE 2: |
2155 |
* memory = 0x12 0x34 0x56 0x78 |
2156 |
* register after lwl: 0x56 0x34 0x12 0x.. |
2157 |
*/ |
2158 |
/* 83 fb 02 cmp $0x2,%ebx */ |
2159 |
/* 75 01 jne <skip> */ |
2160 |
*a++ = 0x83; *a++ = 0xfb; *a++ = 0x02; |
2161 |
*a++ = 0x75; skip = a; *a++ = 0x01; |
2162 |
|
2163 |
/* c1 e2 08 shl $0x08,%edx */ |
2164 |
/* 25 ff 00 00 00 and $0xff,%eax */ |
2165 |
/* 09 d0 or %edx,%eax */ |
2166 |
*a++ = 0xc1; *a++ = 0xe2; *a++ = 0x08; |
2167 |
*a++ = 0x25; *a++ = 0xff; *a++ = 0x00; *a++ = 0x00; *a++ = 0x00; |
2168 |
*a++ = 0x09; *a++ = 0xd0; |
2169 |
|
2170 |
/* eb 00 jmp <okret> */ |
2171 |
*a++ = 0xeb; okret2 = a; *a++ = 0; |
2172 |
|
2173 |
*skip = (size_t)a - (size_t)skip - 1; |
2174 |
|
2175 |
/* |
2176 |
* CASE 3: |
2177 |
* memory = 0x12 0x34 0x56 0x78 |
2178 |
* register after lwl: 0x78 0x56 0x34 0x12 |
2179 |
*/ |
2180 |
/* 89 d0 mov %edx,%eax */ |
2181 |
*a++ = 0x89; *a++ = 0xd0; |
2182 |
|
2183 |
/* okret: */ |
2184 |
*okret0 = (size_t)a - (size_t)okret0 - 1; |
2185 |
*okret1 = (size_t)a - (size_t)okret1 - 1; |
2186 |
*okret2 = (size_t)a - (size_t)okret2 - 1; |
2187 |
|
2188 |
/* 99 cltd */ |
2189 |
*a++ = 0x99; |
2190 |
break; |
2191 |
|
2192 |
case HI6_LWR: |
2193 |
load_into_eax_dont_care_about_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); |
2194 |
/* 05 34 f2 ff ff add $0xfffff234,%eax */ |
2195 |
*a++ = 5; |
2196 |
*a++ = imm; *a++ = imm >> 8; *a++ = 0xff; *a++ = 0xff; |
2197 |
/* 83 e0 03 and $0x03,%eax */ |
2198 |
*a++ = 0x83; *a++ = 0xe0; *a++ = alignment; |
2199 |
/* 89 c3 mov %eax,%ebx */ |
2200 |
*a++ = 0x89; *a++ = 0xc3; |
2201 |
|
2202 |
load_into_eax_dont_care_about_edx(&a, &dummy_cpu.cd.mips.gpr[rt]); |
2203 |
|
2204 |
/* ALIGNED LOAD: */ |
2205 |
/* 8b 11 mov (%ecx),%edx */ |
2206 |
*a++ = 0x8b; *a++ = 0x11; |
2207 |
|
2208 |
/* |
2209 |
* CASE 0: |
2210 |
* memory = 0x12 0x34 0x56 0x78 |
2211 |
* register after lwr: 0x78 0x56 0x34 0x12 |
2212 |
*/ |
2213 |
/* 83 fb 00 cmp $0x0,%ebx */ |
2214 |
/* 75 01 jne <skip> */ |
2215 |
*a++ = 0x83; *a++ = 0xfb; *a++ = 0x00; |
2216 |
*a++ = 0x75; skip = a; *a++ = 0x01; |
2217 |
|
2218 |
/* 89 d0 mov %edx,%eax */ |
2219 |
*a++ = 0x89; *a++ = 0xd0; |
2220 |
|
2221 |
/* eb 00 jmp <okret> */ |
2222 |
*a++ = 0xeb; okret0 = a; *a++ = 0; |
2223 |
|
2224 |
*skip = (size_t)a - (size_t)skip - 1; |
2225 |
|
2226 |
/* |
2227 |
* CASE 1: |
2228 |
* memory = 0x12 0x34 0x56 0x78 |
2229 |
* register after lwr: 0x.. 0x78 0x56 0x34 |
2230 |
*/ |
2231 |
/* 83 fb 01 cmp $0x1,%ebx */ |
2232 |
/* 75 01 jne <skip> */ |
2233 |
*a++ = 0x83; *a++ = 0xfb; *a++ = 0x01; |
2234 |
*a++ = 0x75; skip = a; *a++ = 0x01; |
2235 |
|
2236 |
/* c1 ea 08 shr $0x8,%edx */ |
2237 |
/* 25 00 00 00 ff and $0xff000000,%eax */ |
2238 |
/* 09 d0 or %edx,%eax */ |
2239 |
*a++ = 0xc1; *a++ = 0xea; *a++ = 0x08; |
2240 |
*a++ = 0x25; *a++ = 0x00; *a++ = 0x00; *a++ = 0x00; *a++ = 0xff; |
2241 |
*a++ = 0x09; *a++ = 0xd0; |
2242 |
|
2243 |
/* eb 00 jmp <okret> */ |
2244 |
*a++ = 0xeb; okret1 = a; *a++ = 0; |
2245 |
|
2246 |
*skip = (size_t)a - (size_t)skip - 1; |
2247 |
|
2248 |
/* |
2249 |
* CASE 2: |
2250 |
* memory = 0x12 0x34 0x56 0x78 |
2251 |
* register after lwr: 0x.. 0x.. 0x78 0x56 |
2252 |
*/ |
2253 |
/* 83 fb 02 cmp $0x2,%ebx */ |
2254 |
/* 75 01 jne <skip> */ |
2255 |
*a++ = 0x83; *a++ = 0xfb; *a++ = 0x02; |
2256 |
*a++ = 0x75; skip = a; *a++ = 0x01; |
2257 |
|
2258 |
/* c1 ea 10 shr $0x10,%edx */ |
2259 |
/* 25 00 00 ff ff and $0xffff0000,%eax */ |
2260 |
/* 09 d0 or %edx,%eax */ |
2261 |
*a++ = 0xc1; *a++ = 0xea; *a++ = 0x10; |
2262 |
*a++ = 0x25; *a++ = 0x00; *a++ = 0x00; *a++ = 0xff; *a++ = 0xff; |
2263 |
*a++ = 0x09; *a++ = 0xd0; |
2264 |
|
2265 |
/* eb 00 jmp <okret> */ |
2266 |
*a++ = 0xeb; okret2 = a; *a++ = 0; |
2267 |
|
2268 |
*skip = (size_t)a - (size_t)skip - 1; |
2269 |
|
2270 |
/* |
2271 |
* CASE 3: |
2272 |
* memory = 0x12 0x34 0x56 0x78 |
2273 |
* register after lwr: 0x.. 0x.. 0x.. 0x78 |
2274 |
*/ |
2275 |
/* c1 ea 18 shr $0x18,%edx */ |
2276 |
/* 25 00 ff ff ff and $0xffffff00,%eax */ |
2277 |
/* 09 d0 or %edx,%eax */ |
2278 |
*a++ = 0xc1; *a++ = 0xea; *a++ = 0x18; |
2279 |
*a++ = 0x25; *a++ = 0x00; *a++ = 0xff; *a++ = 0xff; *a++ = 0xff; |
2280 |
*a++ = 0x09; *a++ = 0xd0; |
2281 |
|
2282 |
/* okret: */ |
2283 |
*okret0 = (size_t)a - (size_t)okret0 - 1; |
2284 |
*okret1 = (size_t)a - (size_t)okret1 - 1; |
2285 |
*okret2 = (size_t)a - (size_t)okret2 - 1; |
2286 |
|
2287 |
/* 99 cltd */ |
2288 |
*a++ = 0x99; |
2289 |
break; |
2290 |
|
2291 |
case HI6_SD: |
2292 |
/* 89 01 mov %eax,(%ecx) */ |
2293 |
/* 89 51 04 mov %edx,0x4(%ecx) */ |
2294 |
*a++ = 0x89; *a++ = 0x01; |
2295 |
*a++ = 0x89; *a++ = 0x51; *a++ = 0x04; |
2296 |
break; |
2297 |
case HI6_SW: |
2298 |
/* 89 01 mov %eax,(%ecx) */ |
2299 |
*a++ = 0x89; *a++ = 0x01; |
2300 |
break; |
2301 |
case HI6_SH: |
2302 |
/* 66 89 01 mov %ax,(%ecx) */ |
2303 |
*a++ = 0x66; *a++ = 0x89; *a++ = 0x01; |
2304 |
break; |
2305 |
case HI6_SB: |
2306 |
/* 88 01 mov %al,(%ecx) */ |
2307 |
*a++ = 0x88; *a++ = 0x01; |
2308 |
break; |
2309 |
|
2310 |
case HI6_SWL: |
2311 |
load_into_eax_dont_care_about_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); |
2312 |
/* 05 34 f2 ff ff add $0xfffff234,%eax */ |
2313 |
*a++ = 5; |
2314 |
*a++ = imm; *a++ = imm >> 8; *a++ = 0xff; *a++ = 0xff; |
2315 |
/* 83 e0 03 and $0x03,%eax */ |
2316 |
*a++ = 0x83; *a++ = 0xe0; *a++ = alignment; |
2317 |
/* 89 c3 mov %eax,%ebx */ |
2318 |
*a++ = 0x89; *a++ = 0xc3; |
2319 |
|
2320 |
load_into_eax_dont_care_about_edx(&a, &dummy_cpu.cd.mips.gpr[rt]); |
2321 |
|
2322 |
/* ALIGNED LOAD: */ |
2323 |
/* 8b 11 mov (%ecx),%edx */ |
2324 |
*a++ = 0x8b; *a++ = 0x11; |
2325 |
|
2326 |
/* |
2327 |
* CASE 0: |
2328 |
* memory (edx): 0x12 0x34 0x56 0x78 |
2329 |
* register (eax): 0x89abcdef |
2330 |
* mem after swl: 0x89 0x.. 0x.. 0x.. |
2331 |
*/ |
2332 |
/* 83 fb 00 cmp $0x0,%ebx */ |
2333 |
/* 75 01 jne <skip> */ |
2334 |
*a++ = 0x83; *a++ = 0xfb; *a++ = 0x00; |
2335 |
*a++ = 0x75; skip = a; *a++ = 0x01; |
2336 |
|
2337 |
/* 81 e2 00 ff ff ff and $0xffffff00,%edx */ |
2338 |
/* c1 e8 18 shr $0x18,%eax */ |
2339 |
/* 09 d0 or %edx,%eax */ |
2340 |
*a++ = 0x81; *a++ = 0xe2; *a++ = 0x00; *a++ = 0xff; *a++ = 0xff; *a++ = 0xff; |
2341 |
*a++ = 0xc1; *a++ = 0xe8; *a++ = 0x18; |
2342 |
*a++ = 0x09; *a++ = 0xd0; |
2343 |
|
2344 |
/* eb 00 jmp <okret> */ |
2345 |
*a++ = 0xeb; okret0 = a; *a++ = 0; |
2346 |
|
2347 |
*skip = (size_t)a - (size_t)skip - 1; |
2348 |
|
2349 |
/* |
2350 |
* CASE 1: |
2351 |
* memory (edx): 0x12 0x34 0x56 0x78 |
2352 |
* register (eax): 0x89abcdef |
2353 |
* mem after swl: 0xab 0x89 0x.. 0x.. |
2354 |
*/ |
2355 |
/* 83 fb 01 cmp $0x1,%ebx */ |
2356 |
/* 75 01 jne <skip> */ |
2357 |
*a++ = 0x83; *a++ = 0xfb; *a++ = 0x01; |
2358 |
*a++ = 0x75; skip = a; *a++ = 0x01; |
2359 |
|
2360 |
/* 81 e2 00 00 ff ff and $0xffff0000,%edx */ |
2361 |
/* c1 e8 10 shr $0x10,%eax */ |
2362 |
/* 09 d0 or %edx,%eax */ |
2363 |
*a++ = 0x81; *a++ = 0xe2; *a++ = 0x00; *a++ = 0x00; *a++ = 0xff; *a++ = 0xff; |
2364 |
*a++ = 0xc1; *a++ = 0xe8; *a++ = 0x10; |
2365 |
*a++ = 0x09; *a++ = 0xd0; |
2366 |
|
2367 |
/* eb 00 jmp <okret> */ |
2368 |
*a++ = 0xeb; okret1 = a; *a++ = 0; |
2369 |
|
2370 |
*skip = (size_t)a - (size_t)skip - 1; |
2371 |
|
2372 |
/* |
2373 |
* CASE 2: |
2374 |
* memory (edx): 0x12 0x34 0x56 0x78 |
2375 |
* register (eax): 0x89abcdef |
2376 |
* mem after swl: 0xcd 0xab 0x89 0x.. |
2377 |
*/ |
2378 |
/* 83 fb 02 cmp $0x2,%ebx */ |
2379 |
/* 75 01 jne <skip> */ |
2380 |
*a++ = 0x83; *a++ = 0xfb; *a++ = 0x02; |
2381 |
*a++ = 0x75; skip = a; *a++ = 0x01; |
2382 |
|
2383 |
/* 81 e2 00 00 00 ff and $0xff000000,%edx */ |
2384 |
/* c1 e8 08 shr $0x08,%eax */ |
2385 |
/* 09 d0 or %edx,%eax */ |
2386 |
*a++ = 0x81; *a++ = 0xe2; *a++ = 0x00; *a++ = 0x00; *a++ = 0x00; *a++ = 0xff; |
2387 |
*a++ = 0xc1; *a++ = 0xe8; *a++ = 0x08; |
2388 |
*a++ = 0x09; *a++ = 0xd0; |
2389 |
|
2390 |
/* eb 00 jmp <okret> */ |
2391 |
*a++ = 0xeb; okret2 = a; *a++ = 0; |
2392 |
|
2393 |
*skip = (size_t)a - (size_t)skip - 1; |
2394 |
|
2395 |
/* |
2396 |
* CASE 3: |
2397 |
* memory (edx): 0x12 0x34 0x56 0x78 |
2398 |
* register (eax): 0x89abcdef |
2399 |
* mem after swl: 0xef 0xcd 0xab 0x89 |
2400 |
*/ |
2401 |
/* eax = eax :-) */ |
2402 |
|
2403 |
/* okret: */ |
2404 |
*okret0 = (size_t)a - (size_t)okret0 - 1; |
2405 |
*okret1 = (size_t)a - (size_t)okret1 - 1; |
2406 |
*okret2 = (size_t)a - (size_t)okret2 - 1; |
2407 |
|
2408 |
/* Store back to memory: */ |
2409 |
/* 89 01 mov %eax,(%ecx) */ |
2410 |
*a++ = 0x89; *a++ = 0x01; |
2411 |
break; |
2412 |
|
2413 |
case HI6_SWR: |
2414 |
load_into_eax_dont_care_about_edx(&a, &dummy_cpu.cd.mips.gpr[rs]); |
2415 |
/* 05 34 f2 ff ff add $0xfffff234,%eax */ |
2416 |
*a++ = 5; |
2417 |
*a++ = imm; *a++ = imm >> 8; *a++ = 0xff; *a++ = 0xff; |
2418 |
/* 83 e0 03 and $0x03,%eax */ |
2419 |
*a++ = 0x83; *a++ = 0xe0; *a++ = alignment; |
2420 |
/* 89 c3 mov %eax,%ebx */ |
2421 |
*a++ = 0x89; *a++ = 0xc3; |
2422 |
|
2423 |
load_into_eax_dont_care_about_edx(&a, &dummy_cpu.cd.mips.gpr[rt]); |
2424 |
|
2425 |
/* ALIGNED LOAD: */ |
2426 |
/* 8b 11 mov (%ecx),%edx */ |
2427 |
*a++ = 0x8b; *a++ = 0x11; |
2428 |
|
2429 |
/* |
2430 |
* CASE 0: |
2431 |
* memory (edx): 0x12 0x34 0x56 0x78 |
2432 |
* register (eax): 0x89abcdef |
2433 |
* mem after swr: 0xef 0xcd 0xab 0x89 |
2434 |
*/ |
2435 |
/* 83 fb 00 cmp $0x0,%ebx */ |
2436 |
/* 75 01 jne <skip> */ |
2437 |
*a++ = 0x83; *a++ = 0xfb; *a++ = 0x00; |
2438 |
*a++ = 0x75; skip = a; *a++ = 0x01; |
2439 |
|
2440 |
/* eax = eax, so do nothing */ |
2441 |
|
2442 |
/* eb 00 jmp <okret> */ |
2443 |
*a++ = 0xeb; okret0 = a; *a++ = 0; |
2444 |
|
2445 |
*skip = (size_t)a - (size_t)skip - 1; |
2446 |
|
2447 |
/* |
2448 |
* CASE 1: |
2449 |
* memory (edx): 0x12 0x34 0x56 0x78 |
2450 |
* register (eax): 0x89abcdef |
2451 |
* mem after swr: 0x12 0xef 0xcd 0xab |
2452 |
*/ |
2453 |
/* 83 fb 01 cmp $0x1,%ebx */ |
2454 |
/* 75 01 jne <skip> */ |
2455 |
*a++ = 0x83; *a++ = 0xfb; *a++ = 0x01; |
2456 |
*a++ = 0x75; skip = a; *a++ = 0x01; |
2457 |
|
2458 |
/* 81 e2 ff 00 00 00 and $0x000000ff,%edx */ |
2459 |
/* c1 e0 08 shl $0x08,%eax */ |
2460 |
/* 09 d0 or %edx,%eax */ |
2461 |
*a++ = 0x81; *a++ = 0xe2; *a++ = 0xff; *a++ = 0x00; *a++ = 0x00; *a++ = 0x00; |
2462 |
*a++ = 0xc1; *a++ = 0xe0; *a++ = 0x08; |
2463 |
*a++ = 0x09; *a++ = 0xd0; |
2464 |
|
2465 |
/* eb 00 jmp <okret> */ |
2466 |
*a++ = 0xeb; okret1 = a; *a++ = 0; |
2467 |
|
2468 |
*skip = (size_t)a - (size_t)skip - 1; |
2469 |
|
2470 |
/* |
2471 |
* CASE 2: |
2472 |
* memory (edx): 0x12 0x34 0x56 0x78 |
2473 |
* register (eax): 0x89abcdef |
2474 |
* mem after swr: 0x12 0x34 0xef 0xcd |
2475 |
*/ |
2476 |
/* 83 fb 02 cmp $0x2,%ebx */ |
2477 |
/* 75 01 jne <skip> */ |
2478 |
*a++ = 0x83; *a++ = 0xfb; *a++ = 0x02; |
2479 |
*a++ = 0x75; skip = a; *a++ = 0x01; |
2480 |
|
2481 |
/* 81 e2 ff ff 00 00 and $0x0000ffff,%edx */ |
2482 |
/* c1 e0 10 shl $0x10,%eax */ |
2483 |
/* 09 d0 or %edx,%eax */ |
2484 |
*a++ = 0x81; *a++ = 0xe2; *a++ = 0xff; *a++ = 0xff; *a++ = 0x00; *a++ = 0x00; |
2485 |
*a++ = 0xc1; *a++ = 0xe0; *a++ = 0x10; |
2486 |
*a++ = 0x09; *a++ = 0xd0; |
2487 |
|
2488 |
/* eb 00 jmp <okret> */ |
2489 |
*a++ = 0xeb; okret2 = a; *a++ = 0; |
2490 |
|
2491 |
*skip = (size_t)a - (size_t)skip - 1; |
2492 |
|
2493 |
/* |
2494 |
* CASE 3: |
2495 |
* memory (edx): 0x12 0x34 0x56 0x78 |
2496 |
* register (eax): 0x89abcdef |
2497 |
* mem after swr: 0x12 0x34 0x56 0xef |
2498 |
*/ |
2499 |
/* 81 e2 ff ff ff 00 and $0x00ffffff,%edx */ |
2500 |
/* c1 e0 18 shl $0x18,%eax */ |
2501 |
/* 09 d0 or %edx,%eax */ |
2502 |
*a++ = 0x81; *a++ = 0xe2; *a++ = 0xff; *a++ = 0xff; *a++ = 0xff; *a++ = 0x00; |
2503 |
*a++ = 0xc1; *a++ = 0xe0; *a++ = 0x18; |
2504 |
*a++ = 0x09; *a++ = 0xd0; |
2505 |
|
2506 |
|
2507 |
/* okret: */ |
2508 |
*okret0 = (size_t)a - (size_t)okret0 - 1; |
2509 |
*okret1 = (size_t)a - (size_t)okret1 - 1; |
2510 |
*okret2 = (size_t)a - (size_t)okret2 - 1; |
2511 |
|
2512 |
/* Store back to memory: */ |
2513 |
/* 89 01 mov %eax,(%ecx) */ |
2514 |
*a++ = 0x89; *a++ = 0x01; |
2515 |
break; |
2516 |
|
2517 |
default: |
2518 |
bintrans_write_chunkreturn_fail(&a); /* ret (and fail) */ |
2519 |
} |
2520 |
|
2521 |
if (load && rt != 0) |
2522 |
store_eax_edx(&a, &dummy_cpu.cd.mips.gpr[rt]); |
2523 |
|
2524 |
*addrp = a; |
2525 |
bintrans_write_pc_inc(addrp); |
2526 |
return 1; |
2527 |
} |
2528 |
|
2529 |
|
2530 |
/* |
2531 |
* bintrans_write_instruction__tlb_rfe_etc(): |
2532 |
*/ |
2533 |
static int bintrans_write_instruction__tlb_rfe_etc(unsigned char **addrp, |
2534 |
int itype) |
2535 |
{ |
2536 |
unsigned char *a; |
2537 |
int ofs; |
2538 |
|
2539 |
switch (itype) { |
2540 |
case CALL_TLBP: |
2541 |
case CALL_TLBR: |
2542 |
case CALL_TLBWR: |
2543 |
case CALL_TLBWI: |
2544 |
case CALL_RFE: |
2545 |
case CALL_ERET: |
2546 |
case CALL_SYSCALL: |
2547 |
case CALL_BREAK: |
2548 |
break; |
2549 |
default: |
2550 |
return 0; |
2551 |
} |
2552 |
|
2553 |
a = *addrp; |
2554 |
|
2555 |
/* Put back PC into the cpu struct, both as pc and pc_last */ |
2556 |
*a++ = 0x89; *a++ = 0xbe; *a++ = ofs_pc&255; |
2557 |
*a++ = (ofs_pc>>8)&255; *a++ = (ofs_pc>>16)&255; |
2558 |
*a++ = (ofs_pc>>24)&255; /* mov %edi,pc(%esi) */ |
2559 |
|
2560 |
*a++ = 0x89; *a++ = 0xbe; *a++ = ofs_pc_last&255; |
2561 |
*a++ = (ofs_pc_last>>8)&255; *a++ = (ofs_pc_last>>16)&255; |
2562 |
*a++ = (ofs_pc_last>>24)&255; /* mov %edi,pc_last(%esi) */ |
2563 |
|
2564 |
/* ... and make sure that the high 32 bits are ALSO in pc_last: */ |
2565 |
/* 8b 86 38 12 00 00 mov 0x1238(%esi),%eax */ |
2566 |
ofs = ofs_pc + 4; |
2567 |
*a++ = 0x8b; *a++ = 0x86; *a++ = ofs&255; |
2568 |
*a++ = (ofs>>8)&255; *a++ = (ofs>>16)&255; |
2569 |
*a++ = (ofs>>24)&255; /* mov %edi,pc(%esi) */ |
2570 |
|
2571 |
/* 89 86 34 12 00 00 mov %eax,0x1234(%esi) */ |
2572 |
ofs = ofs_pc_last + 4; |
2573 |
*a++ = 0x89; *a++ = 0x86; *a++ = ofs&255; |
2574 |
*a++ = (ofs>>8)&255; *a++ = (ofs>>16)&255; |
2575 |
*a++ = (ofs>>24)&255; /* mov %edi,pc(%esi) */ |
2576 |
|
2577 |
switch (itype) { |
2578 |
case CALL_TLBP: |
2579 |
case CALL_TLBR: |
2580 |
/* push readflag */ |
2581 |
*a++ = 0x6a; *a++ = (itype == CALL_TLBR); |
2582 |
ofs = ((size_t)&dummy_cpu.cd.mips.bintrans_fast_tlbpr) - (size_t)&dummy_cpu; |
2583 |
break; |
2584 |
case CALL_TLBWR: |
2585 |
case CALL_TLBWI: |
2586 |
/* push randomflag */ |
2587 |
*a++ = 0x6a; *a++ = (itype == CALL_TLBWR); |
2588 |
ofs = ((size_t)&dummy_cpu.cd.mips.bintrans_fast_tlbwri) - (size_t)&dummy_cpu; |
2589 |
break; |
2590 |
case CALL_SYSCALL: |
2591 |
case CALL_BREAK: |
2592 |
/* push randomflag */ |
2593 |
*a++ = 0x6a; *a++ = (itype == CALL_BREAK? EXCEPTION_BP : EXCEPTION_SYS); |
2594 |
ofs = ((size_t)&dummy_cpu.cd.mips.bintrans_simple_exception) - (size_t)&dummy_cpu; |
2595 |
break; |
2596 |
case CALL_RFE: |
2597 |
ofs = ((size_t)&dummy_cpu.cd.mips.bintrans_fast_rfe) - (size_t)&dummy_cpu; |
2598 |
break; |
2599 |
case CALL_ERET: |
2600 |
ofs = ((size_t)&dummy_cpu.cd.mips.bintrans_fast_eret) - (size_t)&dummy_cpu; |
2601 |
break; |
2602 |
} |
2603 |
|
2604 |
/* push cpu (esi) */ |
2605 |
*a++ = 0x56; |
2606 |
|
2607 |
/* eax = points to the right function */ |
2608 |
*a++ = 0x8b; *a++ = 0x86; |
2609 |
*a++ = ofs; *a++ = ofs >> 8; *a++ = ofs >> 16; *a++ = ofs >> 24; |
2610 |
|
2611 |
/* ff d0 call *%eax */ |
2612 |
*a++ = 0xff; *a++ = 0xd0; |
2613 |
|
2614 |
switch (itype) { |
2615 |
case CALL_RFE: |
2616 |
case CALL_ERET: |
2617 |
/* 83 c4 04 add $4,%esp */ |
2618 |
*a++ = 0x83; *a++ = 0xc4; *a++ = 4; |
2619 |
break; |
2620 |
default: |
2621 |
/* 83 c4 08 add $8,%esp */ |
2622 |
*a++ = 0x83; *a++ = 0xc4; *a++ = 8; |
2623 |
break; |
2624 |
} |
2625 |
|
2626 |
/* Load PC from the cpu struct. */ |
2627 |
*a++ = 0x8b; *a++ = 0xbe; *a++ = ofs_pc&255; |
2628 |
*a++ = (ofs_pc>>8)&255; *a++ = (ofs_pc>>16)&255; |
2629 |
*a++ = (ofs_pc>>24)&255; /* mov pc(%esi),%edi */ |
2630 |
|
2631 |
*addrp = a; |
2632 |
|
2633 |
switch (itype) { |
2634 |
case CALL_ERET: |
2635 |
case CALL_SYSCALL: |
2636 |
case CALL_BREAK: |
2637 |
break; |
2638 |
default: |
2639 |
bintrans_write_pc_inc(addrp); |
2640 |
} |
2641 |
|
2642 |
return 1; |
2643 |
} |
2644 |
|
2645 |
|
2646 |
/* |
2647 |
* bintrans_backend_init(): |
2648 |
* |
2649 |
* This is neccessary for broken GCC 2.x. (For GCC 3.x, this wouldn't be |
2650 |
* neccessary, and the old code would have worked.) |
2651 |
*/ |
2652 |
static void bintrans_backend_init(void) |
2653 |
{ |
2654 |
int size; |
2655 |
unsigned char *p; |
2656 |
|
2657 |
|
2658 |
/* "runchunk": */ |
2659 |
size = 64; /* NOTE: This MUST be enough, or we fail */ |
2660 |
p = (unsigned char *)mmap(NULL, size, PROT_READ | PROT_WRITE | |
2661 |
PROT_EXEC, MAP_ANON | MAP_PRIVATE, -1, 0); |
2662 |
|
2663 |
/* If mmap() failed, try malloc(): */ |
2664 |
if (p == NULL) { |
2665 |
p = malloc(size); |
2666 |
if (p == NULL) { |
2667 |
fprintf(stderr, "bintrans_backend_init():" |
2668 |
" out of memory\n"); |
2669 |
exit(1); |
2670 |
} |
2671 |
} |
2672 |
|
2673 |
bintrans_runchunk = (void *)p; |
2674 |
|
2675 |
*p++ = 0x57; /* push %edi */ |
2676 |
*p++ = 0x56; /* push %esi */ |
2677 |
*p++ = 0x55; /* push %ebp */ |
2678 |
*p++ = 0x53; /* push %ebx */ |
2679 |
|
2680 |
/* |
2681 |
* In all translated code, esi points to the cpu struct, and |
2682 |
* ebp is the nr of executed (translated) instructions. |
2683 |
*/ |
2684 |
|
2685 |
/* 0=ebx, 4=ebp, 8=esi, 0xc=edi, 0x10=retaddr, 0x14=arg0, 0x18=arg1 */ |
2686 |
|
2687 |
/* mov 0x8(%esp,1),%esi */ |
2688 |
*p++ = 0x8b; *p++ = 0x74; *p++ = 0x24; *p++ = 0x14; |
2689 |
|
2690 |
/* mov nr_instr(%esi),%ebp */ |
2691 |
*p++ = 0x8b; *p++ = 0xae; *p++ = ofs_i&255; *p++ = (ofs_i>>8)&255; |
2692 |
*p++ = (ofs_i>>16)&255; *p++ = (ofs_i>>24)&255; |
2693 |
|
2694 |
/* mov pc(%esi),%edi */ |
2695 |
*p++ = 0x8b; *p++ = 0xbe; *p++ = ofs_pc&255; *p++ = (ofs_pc>>8)&255; |
2696 |
*p++ = (ofs_pc>>16)&255; *p++ = (ofs_pc>>24)&255; |
2697 |
|
2698 |
/* call *0x18(%esp,1) */ |
2699 |
*p++ = 0xff; *p++ = 0x54; *p++ = 0x24; *p++ = 0x18; |
2700 |
|
2701 |
/* mov %ebp,0x1234(%esi) */ |
2702 |
*p++ = 0x89; *p++ = 0xae; *p++ = ofs_i&255; *p++ = (ofs_i>>8)&255; |
2703 |
*p++ = (ofs_i>>16)&255; *p++ = (ofs_i>>24)&255; |
2704 |
|
2705 |
/* mov %edi,pc(%esi) */ |
2706 |
*p++ = 0x89; *p++ = 0xbe; *p++ = ofs_pc&255; *p++ = (ofs_pc>>8)&255; |
2707 |
*p++ = (ofs_pc>>16)&255; *p++ = (ofs_pc>>24)&255; |
2708 |
|
2709 |
*p++ = 0x5b; /* pop %ebx */ |
2710 |
*p++ = 0x5d; /* pop %ebp */ |
2711 |
*p++ = 0x5e; /* pop %esi */ |
2712 |
*p++ = 0x5f; /* pop %edi */ |
2713 |
*p++ = 0xc3; /* ret */ |
2714 |
|
2715 |
|
2716 |
|
2717 |
/* "jump_to_32bit_pc": */ |
2718 |
size = 128; /* NOTE: This MUST be enough, or we fail */ |
2719 |
p = (unsigned char *)mmap(NULL, size, PROT_READ | PROT_WRITE | |
2720 |
PROT_EXEC, MAP_ANON | MAP_PRIVATE, -1, 0); |
2721 |
|
2722 |
/* If mmap() failed, try malloc(): */ |
2723 |
if (p == NULL) { |
2724 |
p = malloc(size); |
2725 |
if (p == NULL) { |
2726 |
fprintf(stderr, "bintrans_backend_init():" |
2727 |
" out of memory\n"); |
2728 |
exit(1); |
2729 |
} |
2730 |
} |
2731 |
|
2732 |
bintrans_jump_to_32bit_pc = (void *)p; |
2733 |
|
2734 |
/* Don't execute too many instructions. */ |
2735 |
/* 81 fd f0 1f 00 00 cmpl $0x1ff0,%ebp */ |
2736 |
/* 7c 01 jl <okk> */ |
2737 |
/* c3 ret */ |
2738 |
*p++ = 0x81; *p++ = 0xfd; *p++ = (N_SAFE_BINTRANS_LIMIT-1) & 255; |
2739 |
*p++ = ((N_SAFE_BINTRANS_LIMIT-1) >> 8) & 255; *p++ = 0; *p++ = 0; |
2740 |
*p++ = 0x7c; *p++ = 0x01; |
2741 |
*p++ = 0xc3; |
2742 |
|
2743 |
/* |
2744 |
* ebx = ((vaddr >> 22) & 1023) * sizeof(void *) |
2745 |
* |
2746 |
* 89 c3 mov %eax,%ebx |
2747 |
* c1 eb 14 shr $20,%ebx |
2748 |
* 81 e3 fc 0f 00 00 and $0xffc,%ebx |
2749 |
*/ |
2750 |
*p++ = 0x89; *p++ = 0xc3; |
2751 |
*p++ = 0xc1; *p++ = 0xeb; *p++ = 0x14; |
2752 |
*p++ = 0x81; *p++ = 0xe3; *p++ = 0xfc; *p++ = 0x0f; *p++ = 0; *p++ = 0; |
2753 |
|
2754 |
/* |
2755 |
* ecx = vaddr_to_hostaddr_table0 |
2756 |
* |
2757 |
* 8b 8e 34 12 00 00 mov 0x1234(%esi),%ecx |
2758 |
*/ |
2759 |
*p++ = 0x8b; *p++ = 0x8e; |
2760 |
*p++ = ofs_tabl0 & 255; *p++ = (ofs_tabl0 >> 8) & 255; |
2761 |
*p++ = (ofs_tabl0 >> 16) & 255; *p++ = (ofs_tabl0 >> 24) & 255; |
2762 |
|
2763 |
/* |
2764 |
* ecx = vaddr_to_hostaddr_table0[a] |
2765 |
* |
2766 |
* 8b 0c 19 mov (%ecx,%ebx),%ecx |
2767 |
*/ |
2768 |
*p++ = 0x8b; *p++ = 0x0c; *p++ = 0x19; |
2769 |
|
2770 |
/* |
2771 |
* ebx = ((vaddr >> 12) & 1023) * sizeof(void *) |
2772 |
* |
2773 |
* 89 c3 mov %eax,%ebx |
2774 |
* c1 eb 0a shr $10,%ebx |
2775 |
* 81 e3 fc 0f 00 00 and $0xffc,%ebx |
2776 |
*/ |
2777 |
*p++ = 0x89; *p++ = 0xc3; |
2778 |
*p++ = 0xc1; *p++ = 0xeb; *p++ = 0x0a; |
2779 |
*p++ = 0x81; *p++ = 0xe3; *p++ = 0xfc; *p++ = 0x0f; *p++ = 0; *p++ = 0; |
2780 |
|
2781 |
/* |
2782 |
* ecx = vaddr_to_hostaddr_table0[a][b].cd.mips.chunks |
2783 |
* |
2784 |
* 8b 8c 19 56 34 12 00 mov 0x123456(%ecx,%ebx,1),%ecx |
2785 |
*/ |
2786 |
*p++ = 0x8b; *p++ = 0x8c; *p++ = 0x19; *p++ = ofs_chunks & 255; |
2787 |
*p++ = (ofs_chunks >> 8) & 255; *p++ = (ofs_chunks >> 16) & 255; |
2788 |
*p++ = (ofs_chunks >> 24) & 255; |
2789 |
|
2790 |
/* |
2791 |
* ecx = NULL? Then return with failure. |
2792 |
* |
2793 |
* 83 f9 00 cmp $0x0,%ecx |
2794 |
* 75 01 jne <okzzz> |
2795 |
*/ |
2796 |
*p++ = 0x83; *p++ = 0xf9; *p++ = 0x00; |
2797 |
*p++ = 0x75; *p++ = 0x01; |
2798 |
*p++ = 0xc3; /* TODO: failure? */ |
2799 |
|
2800 |
/* |
2801 |
* 25 fc 0f 00 00 and $0xffc,%eax |
2802 |
* 01 c1 add %eax,%ecx |
2803 |
* |
2804 |
* 8b 01 mov (%ecx),%eax |
2805 |
* |
2806 |
* 83 f8 00 cmp $0x0,%eax |
2807 |
* 75 01 jne <ok> |
2808 |
* c3 ret |
2809 |
*/ |
2810 |
*p++ = 0x25; *p++ = 0xfc; *p++ = 0x0f; *p++ = 0; *p++ = 0; |
2811 |
*p++ = 0x01; *p++ = 0xc1; |
2812 |
|
2813 |
*p++ = 0x8b; *p++ = 0x01; |
2814 |
|
2815 |
*p++ = 0x83; *p++ = 0xf8; *p++ = 0x00; |
2816 |
*p++ = 0x75; *p++ = 0x01; |
2817 |
*p++ = 0xc3; /* TODO: failure? */ |
2818 |
|
2819 |
/* 03 86 78 56 34 12 add 0x12345678(%esi),%eax */ |
2820 |
/* ff e0 jmp *%eax */ |
2821 |
*p++ = 0x03; *p++ = 0x86; *p++ = ofs_chunkbase & 255; |
2822 |
*p++ = (ofs_chunkbase >> 8) & 255; *p++ = (ofs_chunkbase >> 16) & 255; |
2823 |
*p++ = (ofs_chunkbase >> 24) & 255; |
2824 |
*p++ = 0xff; *p++ = 0xe0; |
2825 |
|
2826 |
|
2827 |
|
2828 |
/* "load_32bit": */ |
2829 |
size = 48; /* NOTE: This MUST be enough, or we fail */ |
2830 |
p = (unsigned char *)mmap(NULL, size, PROT_READ | PROT_WRITE | |
2831 |
PROT_EXEC, MAP_ANON | MAP_PRIVATE, -1, 0); |
2832 |
|
2833 |
/* If mmap() failed, try malloc(): */ |
2834 |
if (p == NULL) { |
2835 |
p = malloc(size); |
2836 |
if (p == NULL) { |
2837 |
fprintf(stderr, "bintrans_backend_init():" |
2838 |
" out of memory\n"); |
2839 |
exit(1); |
2840 |
} |
2841 |
} |
2842 |
|
2843 |
bintrans_load_32bit = (void *)p; |
2844 |
|
2845 |
/* |
2846 |
* ebx = ((vaddr >> 22) & 1023) * sizeof(void *) |
2847 |
* |
2848 |
* 89 c3 mov %eax,%ebx |
2849 |
* c1 eb 14 shr $20,%ebx |
2850 |
* 81 e3 fc 0f 00 00 and $0xffc,%ebx |
2851 |
*/ |
2852 |
*p++ = 0x89; *p++ = 0xc3; |
2853 |
*p++ = 0xc1; *p++ = 0xeb; *p++ = 0x14; |
2854 |
*p++ = 0x81; *p++ = 0xe3; *p++ = 0xfc; *p++ = 0x0f; *p++ = 0; *p++ = 0; |
2855 |
|
2856 |
/* |
2857 |
* ecx = vaddr_to_hostaddr_table0 |
2858 |
* |
2859 |
* 8b 8e 34 12 00 00 mov 0x1234(%esi),%ecx |
2860 |
*/ |
2861 |
*p++ = 0x8b; *p++ = 0x8e; *p++ = ofs_tabl0 & 255; |
2862 |
*p++ = (ofs_tabl0 >> 8) & 255; |
2863 |
*p++ = (ofs_tabl0 >> 16) & 255; *p++ = (ofs_tabl0 >> 24) & 255; |
2864 |
|
2865 |
/* |
2866 |
* ecx = vaddr_to_hostaddr_table0[a] |
2867 |
* |
2868 |
* 8b 0c 19 mov (%ecx,%ebx),%ecx |
2869 |
*/ |
2870 |
*p++ = 0x8b; *p++ = 0x0c; *p++ = 0x19; |
2871 |
|
2872 |
/* |
2873 |
* ebx = ((vaddr >> 12) & 1023) * sizeof(void *) |
2874 |
* |
2875 |
* 89 c3 mov %eax,%ebx |
2876 |
* c1 eb 0a shr $10,%ebx |
2877 |
* 81 e3 fc 0f 00 00 and $0xffc,%ebx |
2878 |
*/ |
2879 |
*p++ = 0x89; *p++ = 0xc3; |
2880 |
*p++ = 0xc1; *p++ = 0xeb; *p++ = 0x0a; |
2881 |
*p++ = 0x81; *p++ = 0xe3; *p++ = 0xfc; *p++ = 0x0f; *p++ = 0; *p++ = 0; |
2882 |
|
2883 |
/* |
2884 |
* ecx = vaddr_to_hostaddr_table0[a][b*2] |
2885 |
* |
2886 |
* 8b 0c 59 mov 0(%ecx,%ebx,2),%ecx |
2887 |
*/ |
2888 |
*p++ = 0x8b; *p++ = 0x0c; *p++ = 0x59; |
2889 |
|
2890 |
/* ret */ |
2891 |
*p++ = 0xc3; |
2892 |
|
2893 |
|
2894 |
|
2895 |
/* "store_32bit": */ |
2896 |
size = 48; /* NOTE: This MUST be enough, or we fail */ |
2897 |
p = (unsigned char *)mmap(NULL, size, PROT_READ | PROT_WRITE | |
2898 |
PROT_EXEC, MAP_ANON | MAP_PRIVATE, -1, 0); |
2899 |
|
2900 |
/* If mmap() failed, try malloc(): */ |
2901 |
if (p == NULL) { |
2902 |
p = malloc(size); |
2903 |
if (p == NULL) { |
2904 |
fprintf(stderr, "bintrans_backend_init():" |
2905 |
" out of memory\n"); |
2906 |
exit(1); |
2907 |
} |
2908 |
} |
2909 |
|
2910 |
bintrans_store_32bit = (void *)p; |
2911 |
|
2912 |
/* |
2913 |
* ebx = ((vaddr >> 22) & 1023) * sizeof(void *) |
2914 |
* |
2915 |
* 89 c3 mov %eax,%ebx |
2916 |
* c1 eb 14 shr $20,%ebx |
2917 |
* 81 e3 fc 0f 00 00 and $0xffc,%ebx |
2918 |
*/ |
2919 |
*p++ = 0x89; *p++ = 0xc3; |
2920 |
*p++ = 0xc1; *p++ = 0xeb; *p++ = 0x14; |
2921 |
*p++ = 0x81; *p++ = 0xe3; *p++ = 0xfc; *p++ = 0x0f; *p++ = 0; *p++ = 0; |
2922 |
|
2923 |
/* |
2924 |
* ecx = vaddr_to_hostaddr_table0 |
2925 |
* |
2926 |
* 8b 8e 34 12 00 00 mov 0x1234(%esi),%ecx |
2927 |
*/ |
2928 |
*p++ = 0x8b; *p++ = 0x8e; *p++ = ofs_tabl0 & 255; |
2929 |
*p++ = (ofs_tabl0 >> 8) & 255; |
2930 |
*p++ = (ofs_tabl0 >> 16) & 255; *p++ = (ofs_tabl0 >> 24) & 255; |
2931 |
|
2932 |
/* |
2933 |
* ecx = vaddr_to_hostaddr_table0[a] |
2934 |
* |
2935 |
* 8b 0c 19 mov (%ecx,%ebx),%ecx |
2936 |
*/ |
2937 |
*p++ = 0x8b; *p++ = 0x0c; *p++ = 0x19; |
2938 |
|
2939 |
/* |
2940 |
* ebx = ((vaddr >> 12) & 1023) * sizeof(void *) |
2941 |
* |
2942 |
* 89 c3 mov %eax,%ebx |
2943 |
* c1 eb 0a shr $10,%ebx |
2944 |
* 81 e3 fc 0f 00 00 and $0xffc,%ebx |
2945 |
*/ |
2946 |
*p++ = 0x89; *p++ = 0xc3; |
2947 |
*p++ = 0xc1; *p++ = 0xeb; *p++ = 0x0a; |
2948 |
*p++ = 0x81; *p++ = 0xe3; *p++ = 0xfc; *p++ = 0x0f; *p++ = 0; *p++ = 0; |
2949 |
|
2950 |
/* |
2951 |
* ecx = vaddr_to_hostaddr_table0[a][b*2] |
2952 |
* |
2953 |
* 8b 4c 59 04 mov 4(%ecx,%ebx,2),%ecx |
2954 |
*/ |
2955 |
*p++ = 0x8b; *p++ = 0x4c; *p++ = 0x59; *p++ = 0x04; |
2956 |
|
2957 |
/* ret */ |
2958 |
*p++ = 0xc3; |
2959 |
} |
2960 |
|