summaryrefslogtreecommitdiffstats
path: root/sandbox/win/src/sidestep/mini_disassembler.cpp
blob: 1e8e0bd97295cb652193251cad03af8365d9bf4c (plain)
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
// Copyright (c) 2012 The Chromium Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

// Implementation of MiniDisassembler.

#ifdef _WIN64
#error The code in this file should not be used on 64-bit Windows.
#endif

#include "sandbox/win/src/sidestep/mini_disassembler.h"

namespace sidestep {

MiniDisassembler::MiniDisassembler(bool operand_default_is_32_bits,
                                   bool address_default_is_32_bits)
    : operand_default_is_32_bits_(operand_default_is_32_bits),
      address_default_is_32_bits_(address_default_is_32_bits) {
  Initialize();
}

MiniDisassembler::MiniDisassembler()
    : operand_default_is_32_bits_(true),
      address_default_is_32_bits_(true) {
  Initialize();
}

InstructionType MiniDisassembler::Disassemble(
    unsigned char* start_byte,
    unsigned int* instruction_bytes) {
  // Clean up any state from previous invocations.
  Initialize();

  // Start by processing any prefixes.
  unsigned char* current_byte = start_byte;
  unsigned int size = 0;
  InstructionType instruction_type = ProcessPrefixes(current_byte, &size);

  if (IT_UNKNOWN == instruction_type)
    return instruction_type;

  current_byte += size;
  size = 0;

  // Invariant: We have stripped all prefixes, and the operand_is_32_bits_
  // and address_is_32_bits_ flags are correctly set.

  instruction_type = ProcessOpcode(current_byte, 0, &size);

  // Check for error processing instruction
  if ((IT_UNKNOWN == instruction_type_) || (IT_UNUSED == instruction_type_)) {
    return IT_UNKNOWN;
  }

  current_byte += size;

  // Invariant: operand_bytes_ indicates the total size of operands
  // specified by the opcode and/or ModR/M byte and/or SIB byte.
  // pCurrentByte points to the first byte after the ModR/M byte, or after
  // the SIB byte if it is present (i.e. the first byte of any operands
  // encoded in the instruction).

  // We get the total length of any prefixes, the opcode, and the ModR/M and
  // SIB bytes if present, by taking the difference of the original starting
  // address and the current byte (which points to the first byte of the
  // operands if present, or to the first byte of the next instruction if
  // they are not).  Adding the count of bytes in the operands encoded in
  // the instruction gives us the full length of the instruction in bytes.
  *instruction_bytes += operand_bytes_ + (current_byte - start_byte);

  // Return the instruction type, which was set by ProcessOpcode().
  return instruction_type_;
}

void MiniDisassembler::Initialize() {
  operand_is_32_bits_ = operand_default_is_32_bits_;
  address_is_32_bits_ = address_default_is_32_bits_;
  operand_bytes_ = 0;
  have_modrm_ = false;
  should_decode_modrm_ = false;
  instruction_type_ = IT_UNKNOWN;
  got_f2_prefix_ = false;
  got_f3_prefix_ = false;
  got_66_prefix_ = false;
}

InstructionType MiniDisassembler::ProcessPrefixes(unsigned char* start_byte,
                                                  unsigned int* size) {
  InstructionType instruction_type = IT_GENERIC;
  const Opcode& opcode = s_ia32_opcode_map_[0].table_[*start_byte];

  switch (opcode.type_) {
    case IT_PREFIX_ADDRESS:
      address_is_32_bits_ = !address_default_is_32_bits_;
      goto nochangeoperand;
    case IT_PREFIX_OPERAND:
      operand_is_32_bits_ = !operand_default_is_32_bits_;
      nochangeoperand:
    case IT_PREFIX:

      if (0xF2 == (*start_byte))
        got_f2_prefix_ = true;
      else if (0xF3 == (*start_byte))
        got_f3_prefix_ = true;
      else if (0x66 == (*start_byte))
        got_66_prefix_ = true;

      instruction_type = opcode.type_;
      (*size)++;
      // we got a prefix, so add one and check next byte
      ProcessPrefixes(start_byte + 1, size);
    default:
      break;   // not a prefix byte
  }

  return instruction_type;
}

InstructionType MiniDisassembler::ProcessOpcode(unsigned char* start_byte,
                                                unsigned int table_index,
                                                unsigned int* size) {
  const OpcodeTable& table = s_ia32_opcode_map_[table_index];   // Get our table
  unsigned char current_byte = (*start_byte) >> table.shift_;
  current_byte = current_byte & table.mask_;  // Mask out the bits we will use

  // Check whether the byte we have is inside the table we have.
  if (current_byte < table.min_lim_ || current_byte > table.max_lim_) {
    instruction_type_ = IT_UNKNOWN;
    return instruction_type_;
  }

  const Opcode& opcode = table.table_[current_byte];
  if (IT_UNUSED == opcode.type_) {
    // This instruction is not used by the IA-32 ISA, so we indicate
    // this to the user.  Probably means that we were pointed to
    // a byte in memory that was not the start of an instruction.
    instruction_type_ = IT_UNUSED;
    return instruction_type_;
  } else if (IT_REFERENCE == opcode.type_) {
    // We are looking at an opcode that has more bytes (or is continued
    // in the ModR/M byte).  Recursively find the opcode definition in
    // the table for the opcode's next byte.
    (*size)++;
    ProcessOpcode(start_byte + 1, opcode.table_index_, size);
    return instruction_type_;
  }

  const SpecificOpcode* specific_opcode = reinterpret_cast<
                                              const SpecificOpcode*>(&opcode);
  if (opcode.is_prefix_dependent_) {
    if (got_f2_prefix_ && opcode.opcode_if_f2_prefix_.mnemonic_ != 0) {
      specific_opcode = &opcode.opcode_if_f2_prefix_;
    } else if (got_f3_prefix_ && opcode.opcode_if_f3_prefix_.mnemonic_ != 0) {
      specific_opcode = &opcode.opcode_if_f3_prefix_;
    } else if (got_66_prefix_ && opcode.opcode_if_66_prefix_.mnemonic_ != 0) {
      specific_opcode = &opcode.opcode_if_66_prefix_;
    }
  }

  // Inv: The opcode type is known.
  instruction_type_ = specific_opcode->type_;

  // Let's process the operand types to see if we have any immediate
  // operands, and/or a ModR/M byte.

  ProcessOperand(specific_opcode->flag_dest_);
  ProcessOperand(specific_opcode->flag_source_);
  ProcessOperand(specific_opcode->flag_aux_);

  // Inv: We have processed the opcode and incremented operand_bytes_
  // by the number of bytes of any operands specified by the opcode
  // that are stored in the instruction (not registers etc.).  Now
  // we need to return the total number of bytes for the opcode and
  // for the ModR/M or SIB bytes if they are present.

  if (table.mask_ != 0xff) {
    if (have_modrm_) {
      // we're looking at a ModR/M byte so we're not going to
      // count that into the opcode size
      ProcessModrm(start_byte, size);
      return IT_GENERIC;
    } else {
      // need to count the ModR/M byte even if it's just being
      // used for opcode extension
      (*size)++;
      return IT_GENERIC;
    }
  } else {
    if (have_modrm_) {
      // The ModR/M byte is the next byte.
      (*size)++;
      ProcessModrm(start_byte + 1, size);
      return IT_GENERIC;
    } else {
      (*size)++;
      return IT_GENERIC;
    }
  }
}

bool MiniDisassembler::ProcessOperand(int flag_operand) {
  bool succeeded = true;
  if (AM_NOT_USED == flag_operand)
    return succeeded;

  // Decide what to do based on the addressing mode.
  switch (flag_operand & AM_MASK) {
    // No ModR/M byte indicated by these addressing modes, and no
    // additional (e.g. immediate) parameters.
    case AM_A:  // Direct address
    case AM_F:  // EFLAGS register
    case AM_X:  // Memory addressed by the DS:SI register pair
    case AM_Y:  // Memory addressed by the ES:DI register pair
    case AM_IMPLICIT:  // Parameter is implicit, occupies no space in
                       // instruction
      break;

    // There is a ModR/M byte but it does not necessarily need
    // to be decoded.
    case AM_C:  // reg field of ModR/M selects a control register
    case AM_D:  // reg field of ModR/M selects a debug register
    case AM_G:  // reg field of ModR/M selects a general register
    case AM_P:  // reg field of ModR/M selects an MMX register
    case AM_R:  // mod field of ModR/M may refer only to a general register
    case AM_S:  // reg field of ModR/M selects a segment register
    case AM_T:  // reg field of ModR/M selects a test register
    case AM_V:  // reg field of ModR/M selects a 128-bit XMM register
      have_modrm_ = true;
      break;

    // In these addressing modes, there is a ModR/M byte and it needs to be
    // decoded. No other (e.g. immediate) params than indicated in ModR/M.
    case AM_E:  // Operand is either a general-purpose register or memory,
                // specified by ModR/M byte
    case AM_M:  // ModR/M byte will refer only to memory
    case AM_Q:  // Operand is either an MMX register or memory (complex
                // evaluation), specified by ModR/M byte
    case AM_W:  // Operand is either a 128-bit XMM register or memory (complex
                // eval), specified by ModR/M byte
      have_modrm_ = true;
      should_decode_modrm_ = true;
      break;

    // These addressing modes specify an immediate or an offset value
    // directly, so we need to look at the operand type to see how many
    // bytes.
    case AM_I:  // Immediate data.
    case AM_J:  // Jump to offset.
    case AM_O:  // Operand is at offset.
      switch (flag_operand & OT_MASK) {
        case OT_B:  // Byte regardless of operand-size attribute.
          operand_bytes_ += OS_BYTE;
          break;
        case OT_C:  // Byte or word, depending on operand-size attribute.
          if (operand_is_32_bits_)
            operand_bytes_ += OS_WORD;
          else
            operand_bytes_ += OS_BYTE;
          break;
        case OT_D:  // Doubleword, regardless of operand-size attribute.
          operand_bytes_ += OS_DOUBLE_WORD;
          break;
        case OT_DQ:  // Double-quadword, regardless of operand-size attribute.
          operand_bytes_ += OS_DOUBLE_QUAD_WORD;
          break;
        case OT_P:  // 32-bit or 48-bit pointer, depending on operand-size
                    // attribute.
          if (operand_is_32_bits_)
            operand_bytes_ += OS_48_BIT_POINTER;
          else
            operand_bytes_ += OS_32_BIT_POINTER;
          break;
        case OT_PS:  // 128-bit packed single-precision floating-point data.
          operand_bytes_ += OS_128_BIT_PACKED_SINGLE_PRECISION_FLOATING;
          break;
        case OT_Q:  // Quadword, regardless of operand-size attribute.
          operand_bytes_ += OS_QUAD_WORD;
          break;
        case OT_S:  // 6-byte pseudo-descriptor.
          operand_bytes_ += OS_PSEUDO_DESCRIPTOR;
          break;
        case OT_SD:  // Scalar Double-Precision Floating-Point Value
        case OT_PD:  // Unaligned packed double-precision floating point value
          operand_bytes_ += OS_DOUBLE_PRECISION_FLOATING;
          break;
        case OT_SS:
          // Scalar element of a 128-bit packed single-precision
          // floating data.
          // We simply return enItUnknown since we don't have to support
          // floating point
          succeeded = false;
          break;
        case OT_V:  // Word or doubleword, depending on operand-size attribute.
          if (operand_is_32_bits_)
            operand_bytes_ += OS_DOUBLE_WORD;
          else
            operand_bytes_ += OS_WORD;
          break;
        case OT_W:  // Word, regardless of operand-size attribute.
          operand_bytes_ += OS_WORD;
          break;

        // Can safely ignore these.
        case OT_A:  // Two one-word operands in memory or two double-word
                    // operands in memory
        case OT_PI:  // Quadword MMX technology register (e.g. mm0)
        case OT_SI:  // Doubleword integer register (e.g., eax)
          break;

        default:
          break;
      }
      break;

    default:
      break;
  }

  return succeeded;
}

bool MiniDisassembler::ProcessModrm(unsigned char* start_byte,
                                    unsigned int* size) {
  // If we don't need to decode, we just return the size of the ModR/M
  // byte (there is never a SIB byte in this case).
  if (!should_decode_modrm_) {
    (*size)++;
    return true;
  }

  // We never care about the reg field, only the combination of the mod
  // and r/m fields, so let's start by packing those fields together into
  // 5 bits.
  unsigned char modrm = (*start_byte);
  unsigned char mod = modrm & 0xC0;  // mask out top two bits to get mod field
  modrm = modrm & 0x07;  // mask out bottom 3 bits to get r/m field
  mod = mod >> 3;  // shift the mod field to the right place
  modrm = mod | modrm;  // combine the r/m and mod fields as discussed
  mod = mod >> 3;  // shift the mod field to bits 2..0

  // Invariant: modrm contains the mod field in bits 4..3 and the r/m field
  // in bits 2..0, and mod contains the mod field in bits 2..0

  const ModrmEntry* modrm_entry = 0;
  if (address_is_32_bits_)
    modrm_entry = &s_ia32_modrm_map_[modrm];
  else
    modrm_entry = &s_ia16_modrm_map_[modrm];

  // Invariant: modrm_entry points to information that we need to decode
  // the ModR/M byte.

  // Add to the count of operand bytes, if the ModR/M byte indicates
  // that some operands are encoded in the instruction.
  if (modrm_entry->is_encoded_in_instruction_)
    operand_bytes_ += modrm_entry->operand_size_;

  // Process the SIB byte if necessary, and return the count
  // of ModR/M and SIB bytes.
  if (modrm_entry->use_sib_byte_) {
    (*size)++;
    return ProcessSib(start_byte + 1, mod, size);
  } else {
    (*size)++;
    return true;
  }
}

bool MiniDisassembler::ProcessSib(unsigned char* start_byte,
                                  unsigned char mod,
                                  unsigned int* size) {
  // get the mod field from the 2..0 bits of the SIB byte
  unsigned char sib_base = (*start_byte) & 0x07;
  if (0x05 == sib_base) {
    switch (mod) {
      case 0x00:  // mod == 00
      case 0x02:  // mod == 10
        operand_bytes_ += OS_DOUBLE_WORD;
        break;
      case 0x01:  // mod == 01
        operand_bytes_ += OS_BYTE;
        break;
      case 0x03:  // mod == 11
        // According to the IA-32 docs, there does not seem to be a disp
        // value for this value of mod
      default:
        break;
    }
  }

  (*size)++;
  return true;
}

};  // namespace sidestep