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
|
//: Labels are defined by ending names with a ':'. This layer will compute
//: addresses for labels, and compute the offset for instructions using them.
:(scenarios transform)
:(scenario map_label)
== 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
loop:
05 0x0d0c0b0a/imm32 # add to EAX
+transform: label 'loop' is at address 1
:(before "End One-time Setup")
Transform.push_back(rewrite_labels);
:(code)
void rewrite_labels(program& p) {
trace(99, "transform") << "-- rewrite labels" << end();
if (p.segments.empty()) return;
segment& code = p.segments.at(0);
map<string, int32_t> address; // values are unsigned, but we're going to do subtractions on them so they need to fit in 31 bits
compute_addresses_for_labels(code, address);
if (trace_contains_errors()) return;
drop_labels(code);
if (trace_contains_errors()) return;
replace_labels_with_addresses(code, address);
}
void compute_addresses_for_labels(const segment& code, map<string, int32_t>& address) {
int current_byte = 0;
for (int i = 0; i < SIZE(code.lines); ++i) {
const line& inst = code.lines.at(i);
for (int j = 0; j < SIZE(inst.words); ++j) {
const word& curr = inst.words.at(j);
// hack: if we have any operand metadata left after previous transforms,
// deduce its size
// Maybe we should just move this transform to before instruction
// packing, and deduce the size of *all* operands. But then we'll also
// have to deal with bitfields.
if (has_metadata(curr, "disp32") || has_metadata(curr, "imm32")) {
if (*curr.data.rbegin() == ':')
raise << "'" << to_string(inst) << "': don't use ':' when jumping to labels\n" << end();
current_byte += 4;
}
// automatically handle /disp8 and /imm8 here
else if (*curr.data.rbegin() != ':') {
++current_byte;
}
else {
if (contains_any_operand_metadata(curr))
raise << "'" << to_string(inst) << "': label definition (':') not allowed in operand\n" << end();
if (j > 0)
raise << "'" << to_string(inst) << "': labels can only be the first word in a line.\n" << end();
string label = curr.data.substr(0, SIZE(curr.data)-1);
put(address, label, current_byte);
trace(99, "transform") << "label '" << label << "' is at address " << (current_byte+code.start) << end();
// no modifying current_byte; label definitions won't be in the final binary
}
}
}
}
void drop_labels(segment& code) {
for (int i = 0; i < SIZE(code.lines); ++i) {
line& inst = code.lines.at(i);
vector<word>::iterator new_end = remove_if(inst.words.begin(), inst.words.end(), is_label);
inst.words.erase(new_end, inst.words.end());
}
}
bool is_label(const word& w) {
return *w.data.rbegin() == ':';
}
void replace_labels_with_addresses(segment& code, const map<string, int32_t>& address) {
int32_t byte_next_instruction_starts_at = 0;
for (int i = 0; i < SIZE(code.lines); ++i) {
line& inst = code.lines.at(i);
byte_next_instruction_starts_at += num_bytes(inst);
line new_inst;
for (int j = 0; j < SIZE(inst.words); ++j) {
const word& curr = inst.words.at(j);
if (contains_key(address, curr.data)) {
int32_t offset = static_cast<int32_t>(get(address, curr.data)) - byte_next_instruction_starts_at;
if (has_metadata(curr, "disp8") || has_metadata(curr, "imm8")) {
if (offset > 0xff || offset < -0x7f)
raise << "'" << to_string(inst) << "': label too far away for distance " << std::hex << offset << " to fit in 8 bits\n" << end();
else
emit_hex_bytes(new_inst, offset, 1);
}
else if (has_metadata(curr, "disp16")) {
if (offset > 0xffff || offset < -0x7fff)
raise << "'" << to_string(inst) << "': label too far away for distance " << std::hex << offset << " to fit in 16 bits\n" << end();
else
emit_hex_bytes(new_inst, offset, 2);
}
else if (has_metadata(curr, "disp32") || has_metadata(curr, "imm32")) {
emit_hex_bytes(new_inst, offset, 4);
}
}
else {
new_inst.words.push_back(curr);
}
}
inst.words.swap(new_inst.words);
trace(99, "transform") << "instruction after transform: '" << data_to_string(inst) << "'" << end();
}
}
// Assumes all bitfields are packed.
uint32_t num_bytes(const line& inst) {
uint32_t sum = 0;
for (int i = 0; i < SIZE(inst.words); ++i) {
const word& curr = inst.words.at(i);
if (has_metadata(curr, "disp32") || has_metadata(curr, "imm32")) // only multi-byte operands
sum += 4;
else
sum++;
}
return sum;
}
string data_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).data;
}
return out.str();
}
//: Label definitions must be the first word on a line. No jumping inside
//: instructions.
//: They should also be the only word on a line.
//: However, you can absolutely have multiple labels map to the same address,
//: as long as they're on separate lines.
:(scenario multiple_labels_at)
== 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
# address 1
loop:
loop2:
# address 1 (labels take up no space)
05 0x0d0c0b0a/imm32 # add to EAX
# address 6
eb loop2/disp8
# address 8
eb loop3/disp8
# address 10
loop3:
+transform: label 'loop' is at address 1
+transform: label 'loop2' is at address 1
+transform: label 'loop3' is at address 10
# first jump is to -7
+transform: instruction after transform: 'eb f9'
# second jump is to 0 (fall through)
+transform: instruction after transform: 'eb 00'
|