1 //:: simulated x86 registers
  2 
  3 :(before "End Types")
  4 enum {
  5   EAX,
  6   ECX,
  7   EDX,
  8   EBX,
  9   ESP,
 10   EBP,
 11   ESI,
 12   EDI,
 13   NUM_INT_REGISTERS,
 14 };
 15 union reg {
 16   int32_t i;
 17   uint32_t u;
 18 };
 19 :(before "End Globals")
 20 reg Reg[NUM_INT_REGISTERS] = { {0} };
 21 uint32_t EIP = 0;
 22 :(before "End Reset")
 23 bzero(Reg, sizeof(Reg));
 24 EIP = 0;
 25 
 26 //:: simulated flag registers; just a subset that we care about
 27 
 28 :(before "End Globals")
 29 bool SF = false;  // sign flag
 30 bool ZF = false;  // zero flag
 31 bool OF = false;  // overflow flag
 32 :(before "End Reset")
 33 SF = ZF = OF = false;
 34 
 35 //: how the flag registers are updated after each instruction
 36 
 37 :(before "End Includes")
 38 // beware: no side-effects in args
 39 #define BINARY_ARITHMETIC_OP(op, arg1, arg2) { \
 40   /* arg1 and arg2 must be signed */ \
 41   int64_t tmp = arg1 op arg2; \
 42   arg1 = arg1 op arg2; \
 43   trace(2, "run") << "storing 0x" << HEXWORD << arg1 << end(); \
 44   SF = (arg1 < 0); \
 45   ZF = (arg1 == 0); \
 46   OF = (arg1 != tmp); \
 47 }
 48 
 49 #define BINARY_BITWISE_OP(op, arg1, arg2) { \
 50   /* arg1 and arg2 must be unsigned */ \
 51   arg1 = arg1 op arg2; \
 52   trace(2, "run") << "storing 0x" << HEXWORD << arg1 << end(); \
 53   SF = (arg1 >> 31); \
 54   ZF = (arg1 == 0); \
 55   OF = false; \
 56 }
 57 
 58 //:: simulated RAM
 59 
 60 :(before "End Globals")
 61 vector<uint8_t> Mem;
 62 uint32_t End_of_program = 0;
 63 :(before "End Reset")
 64 Mem.clear();
 65 Mem.resize(1024);
 66 End_of_program = 0;
 67 
 68 //:: core interpreter loop
 69 
 70 :(scenario add_imm32_to_eax)
 71 # In scenarios, programs are a series of hex bytes, each (variable-length)
 72 # instruction on one line.
 73 #
 74 # x86 instructions consist of the following parts (see cheatsheet.pdf):
 75 #   opcode        ModR/M                    SIB                   displacement    immediate
 76 #   instruction   mod, reg, Reg/Mem bits    scale, index, base
 77 #   1-3 bytes     0/1 byte                  0/1 byte              0/1/2/4 bytes   0/1/2/4 bytes
 78   ¦ 05                                                                            0a 0b 0c 0d  # add 0x0d0c0b0a to EAX
 79 # All hex bytes must be exactly 2 characters each. No '0x' prefixes.
 80 +load: 1 -> 05
 81 +load: 2 -> 0a
 82 +load: 3 -> 0b
 83 +load: 4 -> 0c
 84 +load: 5 -> 0d
 85 +run: add imm32 0x0d0c0b0a to reg EAX
 86 +run: storing 0x0d0c0b0a
 87 
 88 :(code)
 89 // helper for tests: load a program into memory from a textual representation
 90 // of its bytes, and run it
 91 void run(const string& text_bytes) {
 92   load_program(text_bytes);
 93   EIP = 1;  // preserve null pointer
 94   while (EIP < End_of_program)
 95   ¦ run_one_instruction();
 96 }
 97 
 98 // skeleton of how x86 instructions are decoded
 99 void run_one_instruction() {
100   uint8_t op=0, op2=0, op3=0;
101   switch (op = next()) {
102   case 0xf4:  // hlt
103   ¦ EIP = End_of_program;
104   ¦ break;
105   // our first opcode
106   case 0x05: {  // add imm32 to EAX
107   ¦ int32_t arg2 = imm32();
108   ¦ trace(2, "run") << "add imm32 0x" << HEXWORD << arg2 << " to reg EAX" << end();
109   ¦ BINARY_ARITHMETIC_OP(+, Reg[EAX].i, arg2);
110   ¦ break;
111   }
112   // End Single-Byte Opcodes
113   case 0x0f:
114   ¦ switch(op2 = next()) {
115   ¦ // End Two-Byte Opcodes Starting With 0f
116   ¦ default:
117   ¦ ¦ cerr << "unrecognized second opcode after 0f: " << HEXBYTE << NUM(op2) << '\n';
118   ¦ ¦ exit(1);
119   ¦ }
120   ¦ break;
121   case 0xf3:
122   ¦ switch(op2 = next()) {
123   ¦ // End Two-Byte Opcodes Starting With f3
124   ¦ case 0x0f:
125   ¦ ¦ switch(op3 = next()) {
126   ¦ ¦ // End Three-Byte Opcodes Starting With f3 0f
127   ¦ ¦ default:
128   ¦ ¦ ¦ cerr << "unrecognized third opcode after f3 0f: " << HEXBYTE << NUM(op3) << '\n';
129   ¦ ¦ ¦ exit(1);
130   ¦ ¦ }
131   ¦ ¦ break;
132   ¦ default:
133   ¦ ¦ cerr << "unrecognized second opcode after f3: " << HEXBYTE << NUM(op2) << '\n';
134   ¦ ¦ exit(1);
135   ¦ }
136   ¦ break;
137   default:
138   ¦ cerr << "unrecognized opcode: " << HEXBYTE << NUM(op) << '\n';
139   ¦ exit(1);
140   }
141 }
142 
143 void load_program(const string& text_bytes) {
144   uint32_t addr = 1;
145   istringstream in(text_bytes);
146   in >> std::noskipws;
147   while (has_data(in)) {
148   ¦ char c1 = next_hex_byte(in);
149   ¦ if (c1 == '\0') break;
150   ¦ if (!has_data(in)) {
151   ¦ ¦ raise << "input program truncated mid-byte\n" << end();
152   ¦ ¦ return;
153   ¦ }
154   ¦ char c2 = next_hex_byte(in);
155   ¦ if (c2 == '\0') {
156   ¦ ¦ raise << "input program truncated mid-byte\n" << end();
157   ¦ ¦ return;
158   ¦ }
159   ¦ Mem.at(addr) = to_byte(c1, c2);
160   ¦ trace(99, "load") << addr << " -> " << HEXBYTE << NUM(Mem.at(addr)) << end();
161   ¦ addr++;
162   }
163   End_of_program = addr;
164 }
165 
166 char next_hex_byte(istream& in) {
167   while (has_data(in)) {
168   ¦ char c = '\0';
169   ¦ in >> c;
170   ¦ if (c == ' ' || c == '\n') continue;
171   ¦ while (c == '#') {
172   ¦ ¦ while (has_data(in)) {
173   ¦ ¦ ¦ in >> c;
174   ¦ ¦ ¦ if (c == '\n') {
175   ¦ ¦ ¦ ¦ in >> c;
176   ¦ ¦ ¦ ¦ break;
177   ¦ ¦ ¦ }
178   ¦ ¦ }
179   ¦ }
180   ¦ if (c >= '0' && c <= '9') return c;
181   ¦ else if (c >= 'a' && c <= 'f') return c;
182   ¦ else if (c >= 'A' && c <= 'F') return tolower(c);
183   ¦ // disallow any non-hex characters, including a '0x' prefix
184   ¦ if (!isspace(c)) {
185   ¦ ¦ raise << "invalid non-hex character '" << c << "'\n" << end();
186   ¦ ¦ break;
187   ¦ }
188   }
189   return '\0';
190 }
191 
192 uint8_t to_byte(char hex_byte1, char hex_byte2) {
193   return to_hex_num(hex_byte1)*16 + to_hex_num(hex_byte2);
194 }
195 uint8_t to_hex_num(char c) {
196   if (c >= '0' && c <= '9') return c - '0';
197   if (c >= 'a' && c <= 'f') return c - 'a' + 10;
198   assert(false);
199   return 0;
200 }
201 
202 inline uint8_t next() {
203   return Mem.at(EIP++);
204 }
205 
206 // read a 32-bit immediate in little-endian order from the instruction stream
207 int32_t imm32() {
208   int32_t result = next();
209   result |= (next()<<8);
210   result |= (next()<<16);
211   result |= (next()<<24);
212   return result;
213 }
214 
215 :(before "End Includes")
216 #include <iomanip>
217 #define HEXBYTE  std::hex << std::setw(2) << std::setfill('0')
218 #define HEXWORD  std::hex << std::setw(8) << std::setfill('0')
219 // ugly that iostream doesn't print uint8_t as an integer
220 #define NUM(X) static_cast<int>(X)
221 #include <stdint.h>