1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
|
//: Beginning of "level 2": tagging bytes with metadata around what field of
//: an x86 instruction they're for.
//:
//: The x86 instruction set is variable-length, and how a byte is interpreted
//: affects later instruction boundaries. A lot of the pain in programming
//: machine code stems from computer and programmer going out of sync on what
//: a byte means. The miscommunication is usually not immediately caught, and
//: metastasizes at runtime into kilobytes of misinterpreted instructions.
//:
//: To mitigate these issues, we'll start programming in terms of logical
//: operands rather than physical bytes. Some operands are smaller than a
//: byte, and others may consist of multiple bytes. This layer will correctly
//: pack and order the bytes corresponding to the operands in an instruction.
:(before "End Help Texts")
put(Help, "instructions",
"Each x86 instruction consists of an instruction or opcode and some number\n"
"of operands.\n"
"Each operand has a type. An instruction won't have more than one operand of\n"
"any type.\n"
"Each instruction has some set of allowed operand types. It'll reject others.\n"
"The complete list of operand types: mod, subop, r32 (register), rm32\n"
"(register or memory), scale, index, base, disp8, disp16, disp32, imm8,\n"
"imm32.\n"
"Each of these has its own help page. Try reading 'subx help mod' next.\n"
);
:(before "End Help Contents")
cerr << " instructions\n";
:(scenario pack_immediate_constants)
== 0x1
# instruction effective address operand displacement immediate
# op subop mod rm32 base index scale r32
# 1-3 bytes 3 bits 2 bits 3 bits 3 bits 3 bits 2 bits 2 bits 0/1/2/4 bytes 0/1/2/4 bytes
bb 0x2a/imm32 # copy 42 to EBX
+transform: packing instruction 'bb 0x2a/imm32'
+transform: instruction after packing: 'bb 2a 00 00 00'
+run: copy imm32 0x0000002a to EBX
//: complete set of valid operand types
:(before "End Globals")
set<string> Instruction_operands;
:(before "End One-time Setup")
Instruction_operands.insert("subop");
Instruction_operands.insert("mod");
Instruction_operands.insert("rm32");
Instruction_operands.insert("base");
Instruction_operands.insert("index");
Instruction_operands.insert("scale");
Instruction_operands.insert("r32");
Instruction_operands.insert("disp8");
Instruction_operands.insert("disp16");
Instruction_operands.insert("disp32");
Instruction_operands.insert("imm8");
Instruction_operands.insert("imm32");
:(before "End Help Texts")
init_operand_type_help();
:(code)
void init_operand_type_help() {
put(Help, "mod",
"2-bit operand controlling the _addressing mode_ of many instructions,\n"
"to determine how to compute the _effective address_ to look up memory at\n"
"based on the 'rm32' operand and potentially others.\n"
"\n"
"If mod = 3, just operate on the contents of the register specified by rm32\n"
" (direct mode).\n"
"If mod = 2, effective address is usually* rm32 + disp32\n"
" (indirect mode with displacement).\n"
"If mod = 1, effective address is usually* rm32 + disp8\n"
" (indirect mode with displacement).\n"
"If mod = 0, effective address is usually* rm32 (indirect mode).\n"
"(* - The exception is when rm32 is '4'. Register 4 is the stack pointer (ESP).\n"
" Using it as an address gets more involved. For more details,\n"
" try reading the help pages for 'base', 'index' and 'scale'.)\n"
"\n"
"For complete details consult the IA-32 software developer's manual, table 2-2,\n"
"\"32-bit addressing forms with the ModR/M byte\".\n"
" https://www.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-software-developer-instruction-set-reference-manual-325383.pdf\n"
);
put(Help, "subop",
"Additional 3-bit operand for determining the instruction when the opcode is 81, 8f or ff.\n"
"Can't coexist with operand of type 'r32' in a single instruction, because the two use the same bits.\n"
);
put(Help, "r32",
"3-bit operand specifying a register operand used directly, without any further addressing modes.\n"
);
put(Help, "rm32",
"3-bit operand specifying a register operand whose precise interpretation interacts with 'mod'.\n"
"For complete details consult the IA-32 software developer's manual, table 2-2,\n"
"\"32-bit addressing forms with the ModR/M byte\".\n"
" https://www.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-software-developer-instruction-set-reference-manual-325383.pdf\n"
);
put(Help, "base",
"Additional 3-bit operand (when 'rm32' is 4 unless 'mod' is 3) specifying the register containing an address to look up.\n"
"This address may be further modified by 'index' and 'scale' operands.\n"
" effective address = base + index*scale + displacement (disp8 or disp32)\n"
"For complete details consult the IA-32 software developer's manual, table 2-3,\n"
"\"32-bit addressing forms with the SIB byte\".\n"
" https://www.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-software-developer-instruction-set-reference-manual-325383.pdf\n"
);
put(Help, "index",
"Optional 3-bit operand (when 'rm32' is 4 unless 'mod' is 3) that can be added to the 'base' operand to compute the 'effective address' at which to look up memory.\n"
" effective address = base + index*scale + displacement (disp8 or disp32)\n"
"For complete details consult the IA-32 software developer's manual, table 2-3,\n"
"\"32-bit addressing forms with the SIB byte\".\n"
" https://www.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-software-developer-instruction-set-reference-manual-325383.pdf\n"
);
put(Help, "scale",
"Optional 2-bit operand (when 'rm32' is 4 unless 'mod' is 3) that can be multiplied to the 'index' operand before adding the result to the 'base' operand to compute the _effective address_ to operate on.\n"
" effective address = base + index * scale + displacement (disp8 or disp32)\n"
"For complete details consult the IA-32 software developer's manual, table 2-3,\n"
"\"32-bit addressing forms with the SIB byte\".\n"
" https://www.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-software-developer-instruction-set-reference-manual-325383.pdf\n"
);
put(Help, "disp8",
"8-bit value to be added in many instructions.\n"
);
put(Help, "disp16",
"16-bit value to be added in many instructions.\n"
);
put(Help, "disp32",
"32-bit value to be added in many instructions.\n"
);
put(Help, "imm8",
"8-bit value for many instructions.\n"
);
put(Help, "imm32",
"32-bit value for many instructions.\n"
);
}
//:: transform packing operands into bytes in the right order
:(before "End Transforms")
// Begin Level-2 Transforms
Transform.push_back(pack_operands);
// End Level-2 Transforms
:(code)
void pack_operands(program& p) {
if (p.segments.empty()) return;
segment& code = p.segments.at(0);
// Pack Operands(segment code)
trace(99, "transform") << "-- pack operands" << end();
for (int i = 0; i < SIZE(code.lines); ++i) {
line& inst = code.lines.at(i);
if (all_hex_bytes(inst)) continue;
trace(99, "transform") << "packing instruction '" << to_string(/*with metadata*/inst) << "'" << end();
pack_operands(inst);
trace(99, "transform") << "instruction after packing: '" << to_string(/*without metadata*/inst.words) << "'" << end();
}
}
void pack_operands(line& inst) {
line new_inst;
add_opcodes(inst, new_inst);
add_modrm_byte(inst, new_inst);
add_sib_byte(inst, new_inst);
add_disp_bytes(inst, new_inst);
add_imm_bytes(inst, new_inst);
inst.words.swap(new_inst.words);
}
void add_opcodes(const line& in, line& out) {
out.words.push_back(in.words.at(0));
if (in.words.at(0).data == "0f" || in.words.at(0).data == "f3")
out.words.push_back(in.words.at(1));
if (in.words.at(0).data == "f3" && in.words.at(1).data == "0f")
out.words.push_back(in.words.at(2));
}
void add_modrm_byte(const line& in, line& out) {
uint8_t mod=0, reg_subop=0, rm32=0;
bool emit = false;
for (int i = 0; i < SIZE(in.words); ++i) {
const word& curr = in.words.at(i);
if (has_metadata(curr, "mod")) {
mod = hex_byte(curr.data);
emit = true;
}
else if (has_metadata(curr, "rm32")) {
rm32 = hex_byte(curr.data);
emit = true;
}
else if (has_metadata(curr, "r32")) {
reg_subop = hex_byte(curr.data);
emit = true;
}
else if (has_metadata(curr, "subop")) {
reg_subop = hex_byte(curr.data);
emit = true;
}
}
if (emit)
out.words.push_back(hex_byte_text((mod << 6) | (reg_subop << 3) | rm32));
}
void add_sib_byte(const line& in, line& out) {
uint8_t scale=0, index=0, base=0;
bool emit = false;
for (int i = 0; i < SIZE(in.words); ++i) {
const word& curr = in.words.at(i);
if (has_metadata(curr, "scale")) {
scale = hex_byte(curr.data);
emit = true;
}
else if (has_metadata(curr, "index")) {
index = hex_byte(curr.data);
emit = true;
}
else if (has_metadata(curr, "base")) {
base = hex_byte(curr.data);
emit = true;
}
}
if (emit)
out.words.push_back(hex_byte_text((scale << 6) | (index << 3) | base));
}
void add_disp_bytes(const line& in, line& out) {
for (int i = 0; i < SIZE(in.words); ++i) {
const word& curr = in.words.at(i);
if (has_metadata(curr, "disp8"))
emit_hex_bytes(out, curr, 1);
if (has_metadata(curr, "disp16"))
emit_hex_bytes(out, curr, 2);
else if (has_metadata(curr, "disp32"))
emit_hex_bytes(out, curr, 4);
}
}
void add_imm_bytes(const line& in, line& out) {
for (int i = 0; i < SIZE(in.words); ++i) {
const word& curr = in.words.at(i);
if (has_metadata(curr, "imm8"))
emit_hex_bytes(out, curr, 1);
else if (has_metadata(curr, "imm32"))
emit_hex_bytes(out, curr, 4);
}
}
void emit_hex_bytes(line& out, const word& w, int num) {
assert(num <= 4);
if (!is_hex_int(w.data)) {
out.words.push_back(w);
return;
}
emit_hex_bytes(out, static_cast<uint32_t>(parse_int(w.data)), num);
}
void emit_hex_bytes(line& out, uint32_t val, int num) {
assert(num <= 4);
for (int i = 0; i < num; ++i) {
out.words.push_back(hex_byte_text(val & 0xff));
val = val >> 8;
}
}
word hex_byte_text(uint8_t val) {
ostringstream out;
out << HEXBYTE << NUM(val);
word result;
result.data = out.str();
result.original = out.str()+"/auto";
return result;
}
string to_string(const vector<word>& in) {
ostringstream out;
for (int i = 0; i < SIZE(in); ++i) {
if (i > 0) out << ' ';
out << in.at(i).data;
}
return out.str();
}
:(scenario pack_disp8)
== 0x1
74 2/disp8 # jump 2 bytes away if ZF is set
+transform: packing instruction '74 2/disp8'
+transform: instruction after packing: '74 02'
:(scenarios transform)
:(scenario pack_disp8_negative)
== 0x1
# running this will cause an infinite loop
74 -1/disp8 # jump 1 byte before if ZF is set
+transform: packing instruction '74 -1/disp8'
+transform: instruction after packing: '74 ff'
:(scenarios run)
//: helper for scenario
:(code)
void transform(const string& text_bytes) {
program p;
istringstream in(text_bytes);
parse(in, p);
if (trace_contains_errors()) return;
transform(p);
}
:(scenario pack_modrm_imm32)
== 0x1
# instruction effective address operand displacement immediate
# op subop mod rm32 base index scale r32
# 1-3 bytes 3 bits 2 bits 3 bits 3 bits 3 bits 2 bits 2 bits 0/1/2/4 bytes 0/1/2/4 bytes
81 0/add/subop 3/mod/direct 3/ebx/rm32 1/imm32 # add 1 to EBX
+transform: packing instruction '81 0/add/subop 3/mod/direct 3/ebx/rm32 1/imm32'
+transform: instruction after packing: '81 c3 01 00 00 00'
:(scenario pack_imm32_large)
== 0x1
b9 0x080490a7/imm32 # copy to ECX
+transform: packing instruction 'b9 0x080490a7/imm32'
+transform: instruction after packing: 'b9 a7 90 04 08'
:(scenario pack_immediate_constants_hex)
== 0x1
# instruction effective address operand displacement immediate
# op subop mod rm32 base index scale r32
# 1-3 bytes 3 bits 2 bits 3 bits 3 bits 3 bits 2 bits 2 bits 0/1/2/4 bytes 0/1/2/4 bytes
bb 0x2a/imm32 # copy 42 to EBX
+transform: packing instruction 'bb 0x2a/imm32'
+transform: instruction after packing: 'bb 2a 00 00 00'
+run: copy imm32 0x0000002a to EBX
:(scenarios transform)
:(scenario pack_silently_ignores_non_hex)
== 0x1
# instruction effective address operand displacement immediate
# op subop mod rm32 base index scale r32
# 1-3 bytes 3 bits 2 bits 3 bits 3 bits 3 bits 2 bits 2 bits 0/1/2/4 bytes 0/1/2/4 bytes
bb foo/imm32 # copy foo to EBX
+transform: packing instruction 'bb foo/imm32'
# no change (we're just not printing metadata to the trace)
+transform: instruction after packing: 'bb foo'
$error: 0
:(scenarios run)
//:: helpers
:(code)
bool all_hex_bytes(const line& inst) {
for (int i = 0; i < SIZE(inst.words); ++i)
if (!is_hex_byte(inst.words.at(i)))
return false;
return true;
}
bool is_hex_byte(const word& curr) {
if (contains_any_operand_metadata(curr))
return false;
if (SIZE(curr.data) != 2)
return false;
if (curr.data.find_first_not_of("0123456789abcdefABCDEF") != string::npos)
return false;
return true;
}
bool contains_any_operand_metadata(const word& word) {
for (int i = 0; i < SIZE(word.metadata); ++i)
if (Instruction_operands.find(word.metadata.at(i)) != Instruction_operands.end())
return true;
return false;
}
bool has_metadata(const line& inst, const string& m) {
bool result = false;
for (int i = 0; i < SIZE(inst.words); ++i) {
if (!has_metadata(inst.words.at(i), m)) continue;
if (result) {
raise << "'" << to_string(inst) << "' has conflicting " << m << " operands\n" << end();
return false;
}
result = true;
}
return result;
}
bool has_metadata(const word& w, const string& m) {
bool result = false;
bool metadata_found = false;
for (int i = 0; i < SIZE(w.metadata); ++i) {
const string& curr = w.metadata.at(i);
if (!contains_key(Instruction_operands, curr)) continue; // ignore unrecognized metadata
if (metadata_found) {
raise << "'" << w.original << "' has conflicting operand types; it should have only one\n" << end();
return false;
}
metadata_found = true;
result = (curr == m);
}
return result;
}
word metadata(const line& inst, const string& m) {
for (int i = 0; i < SIZE(inst.words); ++i)
if (has_metadata(inst.words.at(i), m))
return inst.words.at(i);
assert(false);
}
bool is_hex_int(const string& s) {
if (s.empty()) return false;
size_t pos = 0;
if (s.at(0) == '-' || s.at(0) == '+') pos++;
if (s.substr(pos, pos+2) == "0x") pos += 2;
return s.find_first_not_of("0123456789abcdefABCDEF", pos) == string::npos;
}
int32_t parse_int(const string& s) {
istringstream in(s);
int32_t result = 0;
in >> std::hex >> result;
if (!in || !in.eof()) {
raise << "not a number: " << s << '\n' << end();
return 0;
}
return result;
}
string to_string(const line& inst) {
ostringstream out;
for (int i = 0; i < SIZE(inst.words); ++i) {
if (i > 0) out << ' ';
out << inst.words.at(i).original;
}
return out.str();
}
|