Linux Kernel  3.7.1
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
emulate.c
Go to the documentation of this file.
1 /******************************************************************************
2  * emulate.c
3  *
4  * Generic x86 (32-bit and 64-bit) instruction decoder and emulator.
5  *
6  * Copyright (c) 2005 Keir Fraser
7  *
8  * Linux coding style, mod r/m decoder, segment base fixes, real-mode
9  * privileged instructions:
10  *
11  * Copyright (C) 2006 Qumranet
12  * Copyright 2010 Red Hat, Inc. and/or its affiliates.
13  *
14  * Avi Kivity <[email protected]>
15  * Yaniv Kamay <[email protected]>
16  *
17  * This work is licensed under the terms of the GNU GPL, version 2. See
18  * the COPYING file in the top-level directory.
19  *
20  * From: xen-unstable 10676:af9809f51f81a3c43f276f00c81a52ef558afda4
21  */
22 
23 #include <linux/kvm_host.h>
24 #include "kvm_cache_regs.h"
25 #include <linux/module.h>
26 #include <asm/kvm_emulate.h>
27 
28 #include "x86.h"
29 #include "tss.h"
30 
31 /*
32  * Operand types
33  */
34 #define OpNone 0ull
35 #define OpImplicit 1ull /* No generic decode */
36 #define OpReg 2ull /* Register */
37 #define OpMem 3ull /* Memory */
38 #define OpAcc 4ull /* Accumulator: AL/AX/EAX/RAX */
39 #define OpDI 5ull /* ES:DI/EDI/RDI */
40 #define OpMem64 6ull /* Memory, 64-bit */
41 #define OpImmUByte 7ull /* Zero-extended 8-bit immediate */
42 #define OpDX 8ull /* DX register */
43 #define OpCL 9ull /* CL register (for shifts) */
44 #define OpImmByte 10ull /* 8-bit sign extended immediate */
45 #define OpOne 11ull /* Implied 1 */
46 #define OpImm 12ull /* Sign extended immediate */
47 #define OpMem16 13ull /* Memory operand (16-bit). */
48 #define OpMem32 14ull /* Memory operand (32-bit). */
49 #define OpImmU 15ull /* Immediate operand, zero extended */
50 #define OpSI 16ull /* SI/ESI/RSI */
51 #define OpImmFAddr 17ull /* Immediate far address */
52 #define OpMemFAddr 18ull /* Far address in memory */
53 #define OpImmU16 19ull /* Immediate operand, 16 bits, zero extended */
54 #define OpES 20ull /* ES */
55 #define OpCS 21ull /* CS */
56 #define OpSS 22ull /* SS */
57 #define OpDS 23ull /* DS */
58 #define OpFS 24ull /* FS */
59 #define OpGS 25ull /* GS */
60 #define OpMem8 26ull /* 8-bit zero extended memory operand */
61 
62 #define OpBits 5 /* Width of operand field */
63 #define OpMask ((1ull << OpBits) - 1)
64 
65 /*
66  * Opcode effective-address decode tables.
67  * Note that we only emulate instructions that have at least one memory
68  * operand (excluding implicit stack references). We assume that stack
69  * references and instruction fetches will never occur in special memory
70  * areas that require emulation. So, for example, 'mov <imm>,<reg>' need
71  * not be handled.
72  */
73 
74 /* Operand sizes: 8-bit operands or specified/overridden size. */
75 #define ByteOp (1<<0) /* 8-bit operands. */
76 /* Destination operand type. */
77 #define DstShift 1
78 #define ImplicitOps (OpImplicit << DstShift)
79 #define DstReg (OpReg << DstShift)
80 #define DstMem (OpMem << DstShift)
81 #define DstAcc (OpAcc << DstShift)
82 #define DstDI (OpDI << DstShift)
83 #define DstMem64 (OpMem64 << DstShift)
84 #define DstImmUByte (OpImmUByte << DstShift)
85 #define DstDX (OpDX << DstShift)
86 #define DstMask (OpMask << DstShift)
87 /* Source operand type. */
88 #define SrcShift 6
89 #define SrcNone (OpNone << SrcShift)
90 #define SrcReg (OpReg << SrcShift)
91 #define SrcMem (OpMem << SrcShift)
92 #define SrcMem16 (OpMem16 << SrcShift)
93 #define SrcMem32 (OpMem32 << SrcShift)
94 #define SrcImm (OpImm << SrcShift)
95 #define SrcImmByte (OpImmByte << SrcShift)
96 #define SrcOne (OpOne << SrcShift)
97 #define SrcImmUByte (OpImmUByte << SrcShift)
98 #define SrcImmU (OpImmU << SrcShift)
99 #define SrcSI (OpSI << SrcShift)
100 #define SrcImmFAddr (OpImmFAddr << SrcShift)
101 #define SrcMemFAddr (OpMemFAddr << SrcShift)
102 #define SrcAcc (OpAcc << SrcShift)
103 #define SrcImmU16 (OpImmU16 << SrcShift)
104 #define SrcDX (OpDX << SrcShift)
105 #define SrcMem8 (OpMem8 << SrcShift)
106 #define SrcMask (OpMask << SrcShift)
107 #define BitOp (1<<11)
108 #define MemAbs (1<<12) /* Memory operand is absolute displacement */
109 #define String (1<<13) /* String instruction (rep capable) */
110 #define Stack (1<<14) /* Stack instruction (push/pop) */
111 #define GroupMask (7<<15) /* Opcode uses one of the group mechanisms */
112 #define Group (1<<15) /* Bits 3:5 of modrm byte extend opcode */
113 #define GroupDual (2<<15) /* Alternate decoding of mod == 3 */
114 #define Prefix (3<<15) /* Instruction varies with 66/f2/f3 prefix */
115 #define RMExt (4<<15) /* Opcode extension in ModRM r/m if mod == 3 */
116 #define Sse (1<<18) /* SSE Vector instruction */
117 /* Generic ModRM decode. */
118 #define ModRM (1<<19)
119 /* Destination is only written; never read. */
120 #define Mov (1<<20)
121 /* Misc flags */
122 #define Prot (1<<21) /* instruction generates #UD if not in prot-mode */
123 #define VendorSpecific (1<<22) /* Vendor specific instruction */
124 #define NoAccess (1<<23) /* Don't access memory (lea/invlpg/verr etc) */
125 #define Op3264 (1<<24) /* Operand is 64b in long mode, 32b otherwise */
126 #define Undefined (1<<25) /* No Such Instruction */
127 #define Lock (1<<26) /* lock prefix is allowed for the instruction */
128 #define Priv (1<<27) /* instruction generates #GP if current CPL != 0 */
129 #define No64 (1<<28)
130 #define PageTable (1 << 29) /* instruction used to write page table */
131 /* Source 2 operand type */
132 #define Src2Shift (30)
133 #define Src2None (OpNone << Src2Shift)
134 #define Src2CL (OpCL << Src2Shift)
135 #define Src2ImmByte (OpImmByte << Src2Shift)
136 #define Src2One (OpOne << Src2Shift)
137 #define Src2Imm (OpImm << Src2Shift)
138 #define Src2ES (OpES << Src2Shift)
139 #define Src2CS (OpCS << Src2Shift)
140 #define Src2SS (OpSS << Src2Shift)
141 #define Src2DS (OpDS << Src2Shift)
142 #define Src2FS (OpFS << Src2Shift)
143 #define Src2GS (OpGS << Src2Shift)
144 #define Src2Mask (OpMask << Src2Shift)
145 #define Mmx ((u64)1 << 40) /* MMX Vector instruction */
146 #define Aligned ((u64)1 << 41) /* Explicitly aligned (e.g. MOVDQA) */
147 #define Unaligned ((u64)1 << 42) /* Explicitly unaligned (e.g. MOVDQU) */
148 #define Avx ((u64)1 << 43) /* Advanced Vector Extensions */
149 
150 #define X2(x...) x, x
151 #define X3(x...) X2(x), x
152 #define X4(x...) X2(x), X2(x)
153 #define X5(x...) X4(x), x
154 #define X6(x...) X4(x), X2(x)
155 #define X7(x...) X4(x), X3(x)
156 #define X8(x...) X4(x), X4(x)
157 #define X16(x...) X8(x), X8(x)
158 
159 struct opcode {
160  u64 flags : 56;
162  union {
163  int (*execute)(struct x86_emulate_ctxt *ctxt);
164  const struct opcode *group;
165  const struct group_dual *gdual;
166  const struct gprefix *gprefix;
167  } u;
168  int (*check_perm)(struct x86_emulate_ctxt *ctxt);
169 };
170 
171 struct group_dual {
172  struct opcode mod012[8];
173  struct opcode mod3[8];
174 };
175 
176 struct gprefix {
177  struct opcode pfx_no;
178  struct opcode pfx_66;
179  struct opcode pfx_f2;
180  struct opcode pfx_f3;
181 };
182 
183 /* EFLAGS bit definitions. */
184 #define EFLG_ID (1<<21)
185 #define EFLG_VIP (1<<20)
186 #define EFLG_VIF (1<<19)
187 #define EFLG_AC (1<<18)
188 #define EFLG_VM (1<<17)
189 #define EFLG_RF (1<<16)
190 #define EFLG_IOPL (3<<12)
191 #define EFLG_NT (1<<14)
192 #define EFLG_OF (1<<11)
193 #define EFLG_DF (1<<10)
194 #define EFLG_IF (1<<9)
195 #define EFLG_TF (1<<8)
196 #define EFLG_SF (1<<7)
197 #define EFLG_ZF (1<<6)
198 #define EFLG_AF (1<<4)
199 #define EFLG_PF (1<<2)
200 #define EFLG_CF (1<<0)
201 
202 #define EFLG_RESERVED_ZEROS_MASK 0xffc0802a
203 #define EFLG_RESERVED_ONE_MASK 2
204 
205 static ulong reg_read(struct x86_emulate_ctxt *ctxt, unsigned nr)
206 {
207  if (!(ctxt->regs_valid & (1 << nr))) {
208  ctxt->regs_valid |= 1 << nr;
209  ctxt->_regs[nr] = ctxt->ops->read_gpr(ctxt, nr);
210  }
211  return ctxt->_regs[nr];
212 }
213 
214 static ulong *reg_write(struct x86_emulate_ctxt *ctxt, unsigned nr)
215 {
216  ctxt->regs_valid |= 1 << nr;
217  ctxt->regs_dirty |= 1 << nr;
218  return &ctxt->_regs[nr];
219 }
220 
221 static ulong *reg_rmw(struct x86_emulate_ctxt *ctxt, unsigned nr)
222 {
223  reg_read(ctxt, nr);
224  return reg_write(ctxt, nr);
225 }
226 
227 static void writeback_registers(struct x86_emulate_ctxt *ctxt)
228 {
229  unsigned reg;
230 
231  for_each_set_bit(reg, (ulong *)&ctxt->regs_dirty, 16)
232  ctxt->ops->write_gpr(ctxt, reg, ctxt->_regs[reg]);
233 }
234 
235 static void invalidate_registers(struct x86_emulate_ctxt *ctxt)
236 {
237  ctxt->regs_dirty = 0;
238  ctxt->regs_valid = 0;
239 }
240 
241 /*
242  * Instruction emulation:
243  * Most instructions are emulated directly via a fragment of inline assembly
244  * code. This allows us to save/restore EFLAGS and thus very easily pick up
245  * any modified flags.
246  */
247 
248 #if defined(CONFIG_X86_64)
249 #define _LO32 "k" /* force 32-bit operand */
250 #define _STK "%%rsp" /* stack pointer */
251 #elif defined(__i386__)
252 #define _LO32 "" /* force 32-bit operand */
253 #define _STK "%%esp" /* stack pointer */
254 #endif
255 
256 /*
257  * These EFLAGS bits are restored from saved value during emulation, and
258  * any changes are written back to the saved value after emulation.
259  */
260 #define EFLAGS_MASK (EFLG_OF|EFLG_SF|EFLG_ZF|EFLG_AF|EFLG_PF|EFLG_CF)
261 
262 /* Before executing instruction: restore necessary bits in EFLAGS. */
263 #define _PRE_EFLAGS(_sav, _msk, _tmp) \
264  /* EFLAGS = (_sav & _msk) | (EFLAGS & ~_msk); _sav &= ~_msk; */ \
265  "movl %"_sav",%"_LO32 _tmp"; " \
266  "push %"_tmp"; " \
267  "push %"_tmp"; " \
268  "movl %"_msk",%"_LO32 _tmp"; " \
269  "andl %"_LO32 _tmp",("_STK"); " \
270  "pushf; " \
271  "notl %"_LO32 _tmp"; " \
272  "andl %"_LO32 _tmp",("_STK"); " \
273  "andl %"_LO32 _tmp","__stringify(BITS_PER_LONG/4)"("_STK"); " \
274  "pop %"_tmp"; " \
275  "orl %"_LO32 _tmp",("_STK"); " \
276  "popf; " \
277  "pop %"_sav"; "
278 
279 /* After executing instruction: write-back necessary bits in EFLAGS. */
280 #define _POST_EFLAGS(_sav, _msk, _tmp) \
281  /* _sav |= EFLAGS & _msk; */ \
282  "pushf; " \
283  "pop %"_tmp"; " \
284  "andl %"_msk",%"_LO32 _tmp"; " \
285  "orl %"_LO32 _tmp",%"_sav"; "
286 
287 #ifdef CONFIG_X86_64
288 #define ON64(x) x
289 #else
290 #define ON64(x)
291 #endif
292 
293 #define ____emulate_2op(ctxt, _op, _x, _y, _suffix, _dsttype) \
294  do { \
295  __asm__ __volatile__ ( \
296  _PRE_EFLAGS("0", "4", "2") \
297  _op _suffix " %"_x"3,%1; " \
298  _POST_EFLAGS("0", "4", "2") \
299  : "=m" ((ctxt)->eflags), \
300  "+q" (*(_dsttype*)&(ctxt)->dst.val), \
301  "=&r" (_tmp) \
302  : _y ((ctxt)->src.val), "i" (EFLAGS_MASK)); \
303  } while (0)
304 
305 
306 /* Raw emulation: instruction has two explicit operands. */
307 #define __emulate_2op_nobyte(ctxt,_op,_wx,_wy,_lx,_ly,_qx,_qy) \
308  do { \
309  unsigned long _tmp; \
310  \
311  switch ((ctxt)->dst.bytes) { \
312  case 2: \
313  ____emulate_2op(ctxt,_op,_wx,_wy,"w",u16); \
314  break; \
315  case 4: \
316  ____emulate_2op(ctxt,_op,_lx,_ly,"l",u32); \
317  break; \
318  case 8: \
319  ON64(____emulate_2op(ctxt,_op,_qx,_qy,"q",u64)); \
320  break; \
321  } \
322  } while (0)
323 
324 #define __emulate_2op(ctxt,_op,_bx,_by,_wx,_wy,_lx,_ly,_qx,_qy) \
325  do { \
326  unsigned long _tmp; \
327  switch ((ctxt)->dst.bytes) { \
328  case 1: \
329  ____emulate_2op(ctxt,_op,_bx,_by,"b",u8); \
330  break; \
331  default: \
332  __emulate_2op_nobyte(ctxt, _op, \
333  _wx, _wy, _lx, _ly, _qx, _qy); \
334  break; \
335  } \
336  } while (0)
337 
338 /* Source operand is byte-sized and may be restricted to just %cl. */
339 #define emulate_2op_SrcB(ctxt, _op) \
340  __emulate_2op(ctxt, _op, "b", "c", "b", "c", "b", "c", "b", "c")
341 
342 /* Source operand is byte, word, long or quad sized. */
343 #define emulate_2op_SrcV(ctxt, _op) \
344  __emulate_2op(ctxt, _op, "b", "q", "w", "r", _LO32, "r", "", "r")
345 
346 /* Source operand is word, long or quad sized. */
347 #define emulate_2op_SrcV_nobyte(ctxt, _op) \
348  __emulate_2op_nobyte(ctxt, _op, "w", "r", _LO32, "r", "", "r")
349 
350 /* Instruction has three operands and one operand is stored in ECX register */
351 #define __emulate_2op_cl(ctxt, _op, _suffix, _type) \
352  do { \
353  unsigned long _tmp; \
354  _type _clv = (ctxt)->src2.val; \
355  _type _srcv = (ctxt)->src.val; \
356  _type _dstv = (ctxt)->dst.val; \
357  \
358  __asm__ __volatile__ ( \
359  _PRE_EFLAGS("0", "5", "2") \
360  _op _suffix " %4,%1 \n" \
361  _POST_EFLAGS("0", "5", "2") \
362  : "=m" ((ctxt)->eflags), "+r" (_dstv), "=&r" (_tmp) \
363  : "c" (_clv) , "r" (_srcv), "i" (EFLAGS_MASK) \
364  ); \
365  \
366  (ctxt)->src2.val = (unsigned long) _clv; \
367  (ctxt)->src2.val = (unsigned long) _srcv; \
368  (ctxt)->dst.val = (unsigned long) _dstv; \
369  } while (0)
370 
371 #define emulate_2op_cl(ctxt, _op) \
372  do { \
373  switch ((ctxt)->dst.bytes) { \
374  case 2: \
375  __emulate_2op_cl(ctxt, _op, "w", u16); \
376  break; \
377  case 4: \
378  __emulate_2op_cl(ctxt, _op, "l", u32); \
379  break; \
380  case 8: \
381  ON64(__emulate_2op_cl(ctxt, _op, "q", ulong)); \
382  break; \
383  } \
384  } while (0)
385 
386 #define __emulate_1op(ctxt, _op, _suffix) \
387  do { \
388  unsigned long _tmp; \
389  \
390  __asm__ __volatile__ ( \
391  _PRE_EFLAGS("0", "3", "2") \
392  _op _suffix " %1; " \
393  _POST_EFLAGS("0", "3", "2") \
394  : "=m" ((ctxt)->eflags), "+m" ((ctxt)->dst.val), \
395  "=&r" (_tmp) \
396  : "i" (EFLAGS_MASK)); \
397  } while (0)
398 
399 /* Instruction has only one explicit operand (no source operand). */
400 #define emulate_1op(ctxt, _op) \
401  do { \
402  switch ((ctxt)->dst.bytes) { \
403  case 1: __emulate_1op(ctxt, _op, "b"); break; \
404  case 2: __emulate_1op(ctxt, _op, "w"); break; \
405  case 4: __emulate_1op(ctxt, _op, "l"); break; \
406  case 8: ON64(__emulate_1op(ctxt, _op, "q")); break; \
407  } \
408  } while (0)
409 
410 #define __emulate_1op_rax_rdx(ctxt, _op, _suffix, _ex) \
411  do { \
412  unsigned long _tmp; \
413  ulong *rax = reg_rmw((ctxt), VCPU_REGS_RAX); \
414  ulong *rdx = reg_rmw((ctxt), VCPU_REGS_RDX); \
415  \
416  __asm__ __volatile__ ( \
417  _PRE_EFLAGS("0", "5", "1") \
418  "1: \n\t" \
419  _op _suffix " %6; " \
420  "2: \n\t" \
421  _POST_EFLAGS("0", "5", "1") \
422  ".pushsection .fixup,\"ax\" \n\t" \
423  "3: movb $1, %4 \n\t" \
424  "jmp 2b \n\t" \
425  ".popsection \n\t" \
426  _ASM_EXTABLE(1b, 3b) \
427  : "=m" ((ctxt)->eflags), "=&r" (_tmp), \
428  "+a" (*rax), "+d" (*rdx), "+qm"(_ex) \
429  : "i" (EFLAGS_MASK), "m" ((ctxt)->src.val)); \
430  } while (0)
431 
432 /* instruction has only one source operand, destination is implicit (e.g. mul, div, imul, idiv) */
433 #define emulate_1op_rax_rdx(ctxt, _op, _ex) \
434  do { \
435  switch((ctxt)->src.bytes) { \
436  case 1: \
437  __emulate_1op_rax_rdx(ctxt, _op, "b", _ex); \
438  break; \
439  case 2: \
440  __emulate_1op_rax_rdx(ctxt, _op, "w", _ex); \
441  break; \
442  case 4: \
443  __emulate_1op_rax_rdx(ctxt, _op, "l", _ex); \
444  break; \
445  case 8: ON64( \
446  __emulate_1op_rax_rdx(ctxt, _op, "q", _ex)); \
447  break; \
448  } \
449  } while (0)
450 
451 static int emulator_check_intercept(struct x86_emulate_ctxt *ctxt,
453  enum x86_intercept_stage stage)
454 {
455  struct x86_instruction_info info = {
456  .intercept = intercept,
457  .rep_prefix = ctxt->rep_prefix,
458  .modrm_mod = ctxt->modrm_mod,
459  .modrm_reg = ctxt->modrm_reg,
460  .modrm_rm = ctxt->modrm_rm,
461  .src_val = ctxt->src.val64,
462  .src_bytes = ctxt->src.bytes,
463  .dst_bytes = ctxt->dst.bytes,
464  .ad_bytes = ctxt->ad_bytes,
465  .next_rip = ctxt->eip,
466  };
467 
468  return ctxt->ops->intercept(ctxt, &info, stage);
469 }
470 
471 static void assign_masked(ulong *dest, ulong src, ulong mask)
472 {
473  *dest = (*dest & ~mask) | (src & mask);
474 }
475 
476 static inline unsigned long ad_mask(struct x86_emulate_ctxt *ctxt)
477 {
478  return (1UL << (ctxt->ad_bytes << 3)) - 1;
479 }
480 
481 static ulong stack_mask(struct x86_emulate_ctxt *ctxt)
482 {
483  u16 sel;
484  struct desc_struct ss;
485 
486  if (ctxt->mode == X86EMUL_MODE_PROT64)
487  return ~0UL;
488  ctxt->ops->get_segment(ctxt, &sel, &ss, NULL, VCPU_SREG_SS);
489  return ~0U >> ((ss.d ^ 1) * 16); /* d=0: 0xffff; d=1: 0xffffffff */
490 }
491 
492 static int stack_size(struct x86_emulate_ctxt *ctxt)
493 {
494  return (__fls(stack_mask(ctxt)) + 1) >> 3;
495 }
496 
497 /* Access/update address held in a register, based on addressing mode. */
498 static inline unsigned long
499 address_mask(struct x86_emulate_ctxt *ctxt, unsigned long reg)
500 {
501  if (ctxt->ad_bytes == sizeof(unsigned long))
502  return reg;
503  else
504  return reg & ad_mask(ctxt);
505 }
506 
507 static inline unsigned long
508 register_address(struct x86_emulate_ctxt *ctxt, unsigned long reg)
509 {
510  return address_mask(ctxt, reg);
511 }
512 
513 static void masked_increment(ulong *reg, ulong mask, int inc)
514 {
515  assign_masked(reg, *reg + inc, mask);
516 }
517 
518 static inline void
519 register_address_increment(struct x86_emulate_ctxt *ctxt, unsigned long *reg, int inc)
520 {
521  ulong mask;
522 
523  if (ctxt->ad_bytes == sizeof(unsigned long))
524  mask = ~0UL;
525  else
526  mask = ad_mask(ctxt);
527  masked_increment(reg, mask, inc);
528 }
529 
530 static void rsp_increment(struct x86_emulate_ctxt *ctxt, int inc)
531 {
532  masked_increment(reg_rmw(ctxt, VCPU_REGS_RSP), stack_mask(ctxt), inc);
533 }
534 
535 static inline void jmp_rel(struct x86_emulate_ctxt *ctxt, int rel)
536 {
537  register_address_increment(ctxt, &ctxt->_eip, rel);
538 }
539 
540 static u32 desc_limit_scaled(struct desc_struct *desc)
541 {
542  u32 limit = get_desc_limit(desc);
543 
544  return desc->g ? (limit << 12) | 0xfff : limit;
545 }
546 
547 static void set_seg_override(struct x86_emulate_ctxt *ctxt, int seg)
548 {
549  ctxt->has_seg_override = true;
550  ctxt->seg_override = seg;
551 }
552 
553 static unsigned long seg_base(struct x86_emulate_ctxt *ctxt, int seg)
554 {
555  if (ctxt->mode == X86EMUL_MODE_PROT64 && seg < VCPU_SREG_FS)
556  return 0;
557 
558  return ctxt->ops->get_cached_segment_base(ctxt, seg);
559 }
560 
561 static unsigned seg_override(struct x86_emulate_ctxt *ctxt)
562 {
563  if (!ctxt->has_seg_override)
564  return 0;
565 
566  return ctxt->seg_override;
567 }
568 
569 static int emulate_exception(struct x86_emulate_ctxt *ctxt, int vec,
570  u32 error, bool valid)
571 {
572  ctxt->exception.vector = vec;
573  ctxt->exception.error_code = error;
574  ctxt->exception.error_code_valid = valid;
576 }
577 
578 static int emulate_db(struct x86_emulate_ctxt *ctxt)
579 {
580  return emulate_exception(ctxt, DB_VECTOR, 0, false);
581 }
582 
583 static int emulate_gp(struct x86_emulate_ctxt *ctxt, int err)
584 {
585  return emulate_exception(ctxt, GP_VECTOR, err, true);
586 }
587 
588 static int emulate_ss(struct x86_emulate_ctxt *ctxt, int err)
589 {
590  return emulate_exception(ctxt, SS_VECTOR, err, true);
591 }
592 
593 static int emulate_ud(struct x86_emulate_ctxt *ctxt)
594 {
595  return emulate_exception(ctxt, UD_VECTOR, 0, false);
596 }
597 
598 static int emulate_ts(struct x86_emulate_ctxt *ctxt, int err)
599 {
600  return emulate_exception(ctxt, TS_VECTOR, err, true);
601 }
602 
603 static int emulate_de(struct x86_emulate_ctxt *ctxt)
604 {
605  return emulate_exception(ctxt, DE_VECTOR, 0, false);
606 }
607 
608 static int emulate_nm(struct x86_emulate_ctxt *ctxt)
609 {
610  return emulate_exception(ctxt, NM_VECTOR, 0, false);
611 }
612 
613 static u16 get_segment_selector(struct x86_emulate_ctxt *ctxt, unsigned seg)
614 {
615  u16 selector;
616  struct desc_struct desc;
617 
618  ctxt->ops->get_segment(ctxt, &selector, &desc, NULL, seg);
619  return selector;
620 }
621 
622 static void set_segment_selector(struct x86_emulate_ctxt *ctxt, u16 selector,
623  unsigned seg)
624 {
625  u16 dummy;
626  u32 base3;
627  struct desc_struct desc;
628 
629  ctxt->ops->get_segment(ctxt, &dummy, &desc, &base3, seg);
630  ctxt->ops->set_segment(ctxt, selector, &desc, base3, seg);
631 }
632 
633 /*
634  * x86 defines three classes of vector instructions: explicitly
635  * aligned, explicitly unaligned, and the rest, which change behaviour
636  * depending on whether they're AVX encoded or not.
637  *
638  * Also included is CMPXCHG16B which is not a vector instruction, yet it is
639  * subject to the same check.
640  */
641 static bool insn_aligned(struct x86_emulate_ctxt *ctxt, unsigned size)
642 {
643  if (likely(size < 16))
644  return false;
645 
646  if (ctxt->d & Aligned)
647  return true;
648  else if (ctxt->d & Unaligned)
649  return false;
650  else if (ctxt->d & Avx)
651  return false;
652  else
653  return true;
654 }
655 
656 static int __linearize(struct x86_emulate_ctxt *ctxt,
657  struct segmented_address addr,
658  unsigned size, bool write, bool fetch,
659  ulong *linear)
660 {
661  struct desc_struct desc;
662  bool usable;
663  ulong la;
664  u32 lim;
665  u16 sel;
666  unsigned cpl, rpl;
667 
668  la = seg_base(ctxt, addr.seg) + addr.ea;
669  switch (ctxt->mode) {
670  case X86EMUL_MODE_PROT64:
671  if (((signed long)la << 16) >> 16 != la)
672  return emulate_gp(ctxt, 0);
673  break;
674  default:
675  usable = ctxt->ops->get_segment(ctxt, &sel, &desc, NULL,
676  addr.seg);
677  if (!usable)
678  goto bad;
679  /* code segment or read-only data segment */
680  if (((desc.type & 8) || !(desc.type & 2)) && write)
681  goto bad;
682  /* unreadable code segment */
683  if (!fetch && (desc.type & 8) && !(desc.type & 2))
684  goto bad;
685  lim = desc_limit_scaled(&desc);
686  if ((desc.type & 8) || !(desc.type & 4)) {
687  /* expand-up segment */
688  if (addr.ea > lim || (u32)(addr.ea + size - 1) > lim)
689  goto bad;
690  } else {
691  /* expand-down segment */
692  if (addr.ea <= lim || (u32)(addr.ea + size - 1) <= lim)
693  goto bad;
694  lim = desc.d ? 0xffffffff : 0xffff;
695  if (addr.ea > lim || (u32)(addr.ea + size - 1) > lim)
696  goto bad;
697  }
698  cpl = ctxt->ops->cpl(ctxt);
699  if (ctxt->mode == X86EMUL_MODE_REAL)
700  rpl = 0;
701  else
702  rpl = sel & 3;
703  cpl = max(cpl, rpl);
704  if (!(desc.type & 8)) {
705  /* data segment */
706  if (cpl > desc.dpl)
707  goto bad;
708  } else if ((desc.type & 8) && !(desc.type & 4)) {
709  /* nonconforming code segment */
710  if (cpl != desc.dpl)
711  goto bad;
712  } else if ((desc.type & 8) && (desc.type & 4)) {
713  /* conforming code segment */
714  if (cpl < desc.dpl)
715  goto bad;
716  }
717  break;
718  }
719  if (fetch ? ctxt->mode != X86EMUL_MODE_PROT64 : ctxt->ad_bytes != 8)
720  la &= (u32)-1;
721  if (insn_aligned(ctxt, size) && ((la & (size - 1)) != 0))
722  return emulate_gp(ctxt, 0);
723  *linear = la;
724  return X86EMUL_CONTINUE;
725 bad:
726  if (addr.seg == VCPU_SREG_SS)
727  return emulate_ss(ctxt, sel);
728  else
729  return emulate_gp(ctxt, sel);
730 }
731 
732 static int linearize(struct x86_emulate_ctxt *ctxt,
733  struct segmented_address addr,
734  unsigned size, bool write,
735  ulong *linear)
736 {
737  return __linearize(ctxt, addr, size, write, false, linear);
738 }
739 
740 
741 static int segmented_read_std(struct x86_emulate_ctxt *ctxt,
742  struct segmented_address addr,
743  void *data,
744  unsigned size)
745 {
746  int rc;
747  ulong linear;
748 
749  rc = linearize(ctxt, addr, size, false, &linear);
750  if (rc != X86EMUL_CONTINUE)
751  return rc;
752  return ctxt->ops->read_std(ctxt, linear, data, size, &ctxt->exception);
753 }
754 
755 /*
756  * Fetch the next byte of the instruction being emulated which is pointed to
757  * by ctxt->_eip, then increment ctxt->_eip.
758  *
759  * Also prefetch the remaining bytes of the instruction without crossing page
760  * boundary if they are not in fetch_cache yet.
761  */
762 static int do_insn_fetch_byte(struct x86_emulate_ctxt *ctxt, u8 *dest)
763 {
764  struct fetch_cache *fc = &ctxt->fetch;
765  int rc;
766  int size, cur_size;
767 
768  if (ctxt->_eip == fc->end) {
769  unsigned long linear;
770  struct segmented_address addr = { .seg = VCPU_SREG_CS,
771  .ea = ctxt->_eip };
772  cur_size = fc->end - fc->start;
773  size = min(15UL - cur_size,
774  PAGE_SIZE - offset_in_page(ctxt->_eip));
775  rc = __linearize(ctxt, addr, size, false, true, &linear);
776  if (unlikely(rc != X86EMUL_CONTINUE))
777  return rc;
778  rc = ctxt->ops->fetch(ctxt, linear, fc->data + cur_size,
779  size, &ctxt->exception);
780  if (unlikely(rc != X86EMUL_CONTINUE))
781  return rc;
782  fc->end += size;
783  }
784  *dest = fc->data[ctxt->_eip - fc->start];
785  ctxt->_eip++;
786  return X86EMUL_CONTINUE;
787 }
788 
789 static int do_insn_fetch(struct x86_emulate_ctxt *ctxt,
790  void *dest, unsigned size)
791 {
792  int rc;
793 
794  /* x86 instructions are limited to 15 bytes. */
795  if (unlikely(ctxt->_eip + size - ctxt->eip > 15))
796  return X86EMUL_UNHANDLEABLE;
797  while (size--) {
798  rc = do_insn_fetch_byte(ctxt, dest++);
799  if (rc != X86EMUL_CONTINUE)
800  return rc;
801  }
802  return X86EMUL_CONTINUE;
803 }
804 
805 /* Fetch next part of the instruction being emulated. */
806 #define insn_fetch(_type, _ctxt) \
807 ({ unsigned long _x; \
808  rc = do_insn_fetch(_ctxt, &_x, sizeof(_type)); \
809  if (rc != X86EMUL_CONTINUE) \
810  goto done; \
811  (_type)_x; \
812 })
813 
814 #define insn_fetch_arr(_arr, _size, _ctxt) \
815 ({ rc = do_insn_fetch(_ctxt, _arr, (_size)); \
816  if (rc != X86EMUL_CONTINUE) \
817  goto done; \
818 })
819 
820 /*
821  * Given the 'reg' portion of a ModRM byte, and a register block, return a
822  * pointer into the block that addresses the relevant register.
823  * @highbyte_regs specifies whether to decode AH,CH,DH,BH.
824  */
825 static void *decode_register(struct x86_emulate_ctxt *ctxt, u8 modrm_reg,
826  int highbyte_regs)
827 {
828  void *p;
829 
830  if (highbyte_regs && modrm_reg >= 4 && modrm_reg < 8)
831  p = (unsigned char *)reg_rmw(ctxt, modrm_reg & 3) + 1;
832  else
833  p = reg_rmw(ctxt, modrm_reg);
834  return p;
835 }
836 
837 static int read_descriptor(struct x86_emulate_ctxt *ctxt,
838  struct segmented_address addr,
839  u16 *size, unsigned long *address, int op_bytes)
840 {
841  int rc;
842 
843  if (op_bytes == 2)
844  op_bytes = 3;
845  *address = 0;
846  rc = segmented_read_std(ctxt, addr, size, 2);
847  if (rc != X86EMUL_CONTINUE)
848  return rc;
849  addr.ea += 2;
850  rc = segmented_read_std(ctxt, addr, address, op_bytes);
851  return rc;
852 }
853 
854 static int test_cc(unsigned int condition, unsigned int flags)
855 {
856  int rc = 0;
857 
858  switch ((condition & 15) >> 1) {
859  case 0: /* o */
860  rc |= (flags & EFLG_OF);
861  break;
862  case 1: /* b/c/nae */
863  rc |= (flags & EFLG_CF);
864  break;
865  case 2: /* z/e */
866  rc |= (flags & EFLG_ZF);
867  break;
868  case 3: /* be/na */
869  rc |= (flags & (EFLG_CF|EFLG_ZF));
870  break;
871  case 4: /* s */
872  rc |= (flags & EFLG_SF);
873  break;
874  case 5: /* p/pe */
875  rc |= (flags & EFLG_PF);
876  break;
877  case 7: /* le/ng */
878  rc |= (flags & EFLG_ZF);
879  /* fall through */
880  case 6: /* l/nge */
881  rc |= (!(flags & EFLG_SF) != !(flags & EFLG_OF));
882  break;
883  }
884 
885  /* Odd condition identifiers (lsb == 1) have inverted sense. */
886  return (!!rc ^ (condition & 1));
887 }
888 
889 static void fetch_register_operand(struct operand *op)
890 {
891  switch (op->bytes) {
892  case 1:
893  op->val = *(u8 *)op->addr.reg;
894  break;
895  case 2:
896  op->val = *(u16 *)op->addr.reg;
897  break;
898  case 4:
899  op->val = *(u32 *)op->addr.reg;
900  break;
901  case 8:
902  op->val = *(u64 *)op->addr.reg;
903  break;
904  }
905 }
906 
907 static void read_sse_reg(struct x86_emulate_ctxt *ctxt, sse128_t *data, int reg)
908 {
909  ctxt->ops->get_fpu(ctxt);
910  switch (reg) {
911  case 0: asm("movdqa %%xmm0, %0" : "=m"(*data)); break;
912  case 1: asm("movdqa %%xmm1, %0" : "=m"(*data)); break;
913  case 2: asm("movdqa %%xmm2, %0" : "=m"(*data)); break;
914  case 3: asm("movdqa %%xmm3, %0" : "=m"(*data)); break;
915  case 4: asm("movdqa %%xmm4, %0" : "=m"(*data)); break;
916  case 5: asm("movdqa %%xmm5, %0" : "=m"(*data)); break;
917  case 6: asm("movdqa %%xmm6, %0" : "=m"(*data)); break;
918  case 7: asm("movdqa %%xmm7, %0" : "=m"(*data)); break;
919 #ifdef CONFIG_X86_64
920  case 8: asm("movdqa %%xmm8, %0" : "=m"(*data)); break;
921  case 9: asm("movdqa %%xmm9, %0" : "=m"(*data)); break;
922  case 10: asm("movdqa %%xmm10, %0" : "=m"(*data)); break;
923  case 11: asm("movdqa %%xmm11, %0" : "=m"(*data)); break;
924  case 12: asm("movdqa %%xmm12, %0" : "=m"(*data)); break;
925  case 13: asm("movdqa %%xmm13, %0" : "=m"(*data)); break;
926  case 14: asm("movdqa %%xmm14, %0" : "=m"(*data)); break;
927  case 15: asm("movdqa %%xmm15, %0" : "=m"(*data)); break;
928 #endif
929  default: BUG();
930  }
931  ctxt->ops->put_fpu(ctxt);
932 }
933 
934 static void write_sse_reg(struct x86_emulate_ctxt *ctxt, sse128_t *data,
935  int reg)
936 {
937  ctxt->ops->get_fpu(ctxt);
938  switch (reg) {
939  case 0: asm("movdqa %0, %%xmm0" : : "m"(*data)); break;
940  case 1: asm("movdqa %0, %%xmm1" : : "m"(*data)); break;
941  case 2: asm("movdqa %0, %%xmm2" : : "m"(*data)); break;
942  case 3: asm("movdqa %0, %%xmm3" : : "m"(*data)); break;
943  case 4: asm("movdqa %0, %%xmm4" : : "m"(*data)); break;
944  case 5: asm("movdqa %0, %%xmm5" : : "m"(*data)); break;
945  case 6: asm("movdqa %0, %%xmm6" : : "m"(*data)); break;
946  case 7: asm("movdqa %0, %%xmm7" : : "m"(*data)); break;
947 #ifdef CONFIG_X86_64
948  case 8: asm("movdqa %0, %%xmm8" : : "m"(*data)); break;
949  case 9: asm("movdqa %0, %%xmm9" : : "m"(*data)); break;
950  case 10: asm("movdqa %0, %%xmm10" : : "m"(*data)); break;
951  case 11: asm("movdqa %0, %%xmm11" : : "m"(*data)); break;
952  case 12: asm("movdqa %0, %%xmm12" : : "m"(*data)); break;
953  case 13: asm("movdqa %0, %%xmm13" : : "m"(*data)); break;
954  case 14: asm("movdqa %0, %%xmm14" : : "m"(*data)); break;
955  case 15: asm("movdqa %0, %%xmm15" : : "m"(*data)); break;
956 #endif
957  default: BUG();
958  }
959  ctxt->ops->put_fpu(ctxt);
960 }
961 
962 static void read_mmx_reg(struct x86_emulate_ctxt *ctxt, u64 *data, int reg)
963 {
964  ctxt->ops->get_fpu(ctxt);
965  switch (reg) {
966  case 0: asm("movq %%mm0, %0" : "=m"(*data)); break;
967  case 1: asm("movq %%mm1, %0" : "=m"(*data)); break;
968  case 2: asm("movq %%mm2, %0" : "=m"(*data)); break;
969  case 3: asm("movq %%mm3, %0" : "=m"(*data)); break;
970  case 4: asm("movq %%mm4, %0" : "=m"(*data)); break;
971  case 5: asm("movq %%mm5, %0" : "=m"(*data)); break;
972  case 6: asm("movq %%mm6, %0" : "=m"(*data)); break;
973  case 7: asm("movq %%mm7, %0" : "=m"(*data)); break;
974  default: BUG();
975  }
976  ctxt->ops->put_fpu(ctxt);
977 }
978 
979 static void write_mmx_reg(struct x86_emulate_ctxt *ctxt, u64 *data, int reg)
980 {
981  ctxt->ops->get_fpu(ctxt);
982  switch (reg) {
983  case 0: asm("movq %0, %%mm0" : : "m"(*data)); break;
984  case 1: asm("movq %0, %%mm1" : : "m"(*data)); break;
985  case 2: asm("movq %0, %%mm2" : : "m"(*data)); break;
986  case 3: asm("movq %0, %%mm3" : : "m"(*data)); break;
987  case 4: asm("movq %0, %%mm4" : : "m"(*data)); break;
988  case 5: asm("movq %0, %%mm5" : : "m"(*data)); break;
989  case 6: asm("movq %0, %%mm6" : : "m"(*data)); break;
990  case 7: asm("movq %0, %%mm7" : : "m"(*data)); break;
991  default: BUG();
992  }
993  ctxt->ops->put_fpu(ctxt);
994 }
995 
996 static void decode_register_operand(struct x86_emulate_ctxt *ctxt,
997  struct operand *op)
998 {
999  unsigned reg = ctxt->modrm_reg;
1000  int highbyte_regs = ctxt->rex_prefix == 0;
1001 
1002  if (!(ctxt->d & ModRM))
1003  reg = (ctxt->b & 7) | ((ctxt->rex_prefix & 1) << 3);
1004 
1005  if (ctxt->d & Sse) {
1006  op->type = OP_XMM;
1007  op->bytes = 16;
1008  op->addr.xmm = reg;
1009  read_sse_reg(ctxt, &op->vec_val, reg);
1010  return;
1011  }
1012  if (ctxt->d & Mmx) {
1013  reg &= 7;
1014  op->type = OP_MM;
1015  op->bytes = 8;
1016  op->addr.mm = reg;
1017  return;
1018  }
1019 
1020  op->type = OP_REG;
1021  if (ctxt->d & ByteOp) {
1022  op->addr.reg = decode_register(ctxt, reg, highbyte_regs);
1023  op->bytes = 1;
1024  } else {
1025  op->addr.reg = decode_register(ctxt, reg, 0);
1026  op->bytes = ctxt->op_bytes;
1027  }
1028  fetch_register_operand(op);
1029  op->orig_val = op->val;
1030 }
1031 
1032 static void adjust_modrm_seg(struct x86_emulate_ctxt *ctxt, int base_reg)
1033 {
1034  if (base_reg == VCPU_REGS_RSP || base_reg == VCPU_REGS_RBP)
1035  ctxt->modrm_seg = VCPU_SREG_SS;
1036 }
1037 
1038 static int decode_modrm(struct x86_emulate_ctxt *ctxt,
1039  struct operand *op)
1040 {
1041  u8 sib;
1042  int index_reg = 0, base_reg = 0, scale;
1043  int rc = X86EMUL_CONTINUE;
1044  ulong modrm_ea = 0;
1045 
1046  if (ctxt->rex_prefix) {
1047  ctxt->modrm_reg = (ctxt->rex_prefix & 4) << 1; /* REX.R */
1048  index_reg = (ctxt->rex_prefix & 2) << 2; /* REX.X */
1049  ctxt->modrm_rm = base_reg = (ctxt->rex_prefix & 1) << 3; /* REG.B */
1050  }
1051 
1052  ctxt->modrm_mod |= (ctxt->modrm & 0xc0) >> 6;
1053  ctxt->modrm_reg |= (ctxt->modrm & 0x38) >> 3;
1054  ctxt->modrm_rm |= (ctxt->modrm & 0x07);
1055  ctxt->modrm_seg = VCPU_SREG_DS;
1056 
1057  if (ctxt->modrm_mod == 3) {
1058  op->type = OP_REG;
1059  op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
1060  op->addr.reg = decode_register(ctxt, ctxt->modrm_rm, ctxt->d & ByteOp);
1061  if (ctxt->d & Sse) {
1062  op->type = OP_XMM;
1063  op->bytes = 16;
1064  op->addr.xmm = ctxt->modrm_rm;
1065  read_sse_reg(ctxt, &op->vec_val, ctxt->modrm_rm);
1066  return rc;
1067  }
1068  if (ctxt->d & Mmx) {
1069  op->type = OP_MM;
1070  op->bytes = 8;
1071  op->addr.xmm = ctxt->modrm_rm & 7;
1072  return rc;
1073  }
1074  fetch_register_operand(op);
1075  return rc;
1076  }
1077 
1078  op->type = OP_MEM;
1079 
1080  if (ctxt->ad_bytes == 2) {
1081  unsigned bx = reg_read(ctxt, VCPU_REGS_RBX);
1082  unsigned bp = reg_read(ctxt, VCPU_REGS_RBP);
1083  unsigned si = reg_read(ctxt, VCPU_REGS_RSI);
1084  unsigned di = reg_read(ctxt, VCPU_REGS_RDI);
1085 
1086  /* 16-bit ModR/M decode. */
1087  switch (ctxt->modrm_mod) {
1088  case 0:
1089  if (ctxt->modrm_rm == 6)
1090  modrm_ea += insn_fetch(u16, ctxt);
1091  break;
1092  case 1:
1093  modrm_ea += insn_fetch(s8, ctxt);
1094  break;
1095  case 2:
1096  modrm_ea += insn_fetch(u16, ctxt);
1097  break;
1098  }
1099  switch (ctxt->modrm_rm) {
1100  case 0:
1101  modrm_ea += bx + si;
1102  break;
1103  case 1:
1104  modrm_ea += bx + di;
1105  break;
1106  case 2:
1107  modrm_ea += bp + si;
1108  break;
1109  case 3:
1110  modrm_ea += bp + di;
1111  break;
1112  case 4:
1113  modrm_ea += si;
1114  break;
1115  case 5:
1116  modrm_ea += di;
1117  break;
1118  case 6:
1119  if (ctxt->modrm_mod != 0)
1120  modrm_ea += bp;
1121  break;
1122  case 7:
1123  modrm_ea += bx;
1124  break;
1125  }
1126  if (ctxt->modrm_rm == 2 || ctxt->modrm_rm == 3 ||
1127  (ctxt->modrm_rm == 6 && ctxt->modrm_mod != 0))
1128  ctxt->modrm_seg = VCPU_SREG_SS;
1129  modrm_ea = (u16)modrm_ea;
1130  } else {
1131  /* 32/64-bit ModR/M decode. */
1132  if ((ctxt->modrm_rm & 7) == 4) {
1133  sib = insn_fetch(u8, ctxt);
1134  index_reg |= (sib >> 3) & 7;
1135  base_reg |= sib & 7;
1136  scale = sib >> 6;
1137 
1138  if ((base_reg & 7) == 5 && ctxt->modrm_mod == 0)
1139  modrm_ea += insn_fetch(s32, ctxt);
1140  else {
1141  modrm_ea += reg_read(ctxt, base_reg);
1142  adjust_modrm_seg(ctxt, base_reg);
1143  }
1144  if (index_reg != 4)
1145  modrm_ea += reg_read(ctxt, index_reg) << scale;
1146  } else if ((ctxt->modrm_rm & 7) == 5 && ctxt->modrm_mod == 0) {
1147  if (ctxt->mode == X86EMUL_MODE_PROT64)
1148  ctxt->rip_relative = 1;
1149  } else {
1150  base_reg = ctxt->modrm_rm;
1151  modrm_ea += reg_read(ctxt, base_reg);
1152  adjust_modrm_seg(ctxt, base_reg);
1153  }
1154  switch (ctxt->modrm_mod) {
1155  case 0:
1156  if (ctxt->modrm_rm == 5)
1157  modrm_ea += insn_fetch(s32, ctxt);
1158  break;
1159  case 1:
1160  modrm_ea += insn_fetch(s8, ctxt);
1161  break;
1162  case 2:
1163  modrm_ea += insn_fetch(s32, ctxt);
1164  break;
1165  }
1166  }
1167  op->addr.mem.ea = modrm_ea;
1168 done:
1169  return rc;
1170 }
1171 
1172 static int decode_abs(struct x86_emulate_ctxt *ctxt,
1173  struct operand *op)
1174 {
1175  int rc = X86EMUL_CONTINUE;
1176 
1177  op->type = OP_MEM;
1178  switch (ctxt->ad_bytes) {
1179  case 2:
1180  op->addr.mem.ea = insn_fetch(u16, ctxt);
1181  break;
1182  case 4:
1183  op->addr.mem.ea = insn_fetch(u32, ctxt);
1184  break;
1185  case 8:
1186  op->addr.mem.ea = insn_fetch(u64, ctxt);
1187  break;
1188  }
1189 done:
1190  return rc;
1191 }
1192 
1193 static void fetch_bit_operand(struct x86_emulate_ctxt *ctxt)
1194 {
1195  long sv = 0, mask;
1196 
1197  if (ctxt->dst.type == OP_MEM && ctxt->src.type == OP_REG) {
1198  mask = ~(ctxt->dst.bytes * 8 - 1);
1199 
1200  if (ctxt->src.bytes == 2)
1201  sv = (s16)ctxt->src.val & (s16)mask;
1202  else if (ctxt->src.bytes == 4)
1203  sv = (s32)ctxt->src.val & (s32)mask;
1204 
1205  ctxt->dst.addr.mem.ea += (sv >> 3);
1206  }
1207 
1208  /* only subword offset */
1209  ctxt->src.val &= (ctxt->dst.bytes << 3) - 1;
1210 }
1211 
1212 static int read_emulated(struct x86_emulate_ctxt *ctxt,
1213  unsigned long addr, void *dest, unsigned size)
1214 {
1215  int rc;
1216  struct read_cache *mc = &ctxt->mem_read;
1217 
1218  if (mc->pos < mc->end)
1219  goto read_cached;
1220 
1221  WARN_ON((mc->end + size) >= sizeof(mc->data));
1222 
1223  rc = ctxt->ops->read_emulated(ctxt, addr, mc->data + mc->end, size,
1224  &ctxt->exception);
1225  if (rc != X86EMUL_CONTINUE)
1226  return rc;
1227 
1228  mc->end += size;
1229 
1230 read_cached:
1231  memcpy(dest, mc->data + mc->pos, size);
1232  mc->pos += size;
1233  return X86EMUL_CONTINUE;
1234 }
1235 
1236 static int segmented_read(struct x86_emulate_ctxt *ctxt,
1237  struct segmented_address addr,
1238  void *data,
1239  unsigned size)
1240 {
1241  int rc;
1242  ulong linear;
1243 
1244  rc = linearize(ctxt, addr, size, false, &linear);
1245  if (rc != X86EMUL_CONTINUE)
1246  return rc;
1247  return read_emulated(ctxt, linear, data, size);
1248 }
1249 
1250 static int segmented_write(struct x86_emulate_ctxt *ctxt,
1251  struct segmented_address addr,
1252  const void *data,
1253  unsigned size)
1254 {
1255  int rc;
1256  ulong linear;
1257 
1258  rc = linearize(ctxt, addr, size, true, &linear);
1259  if (rc != X86EMUL_CONTINUE)
1260  return rc;
1261  return ctxt->ops->write_emulated(ctxt, linear, data, size,
1262  &ctxt->exception);
1263 }
1264 
1265 static int segmented_cmpxchg(struct x86_emulate_ctxt *ctxt,
1266  struct segmented_address addr,
1267  const void *orig_data, const void *data,
1268  unsigned size)
1269 {
1270  int rc;
1271  ulong linear;
1272 
1273  rc = linearize(ctxt, addr, size, true, &linear);
1274  if (rc != X86EMUL_CONTINUE)
1275  return rc;
1276  return ctxt->ops->cmpxchg_emulated(ctxt, linear, orig_data, data,
1277  size, &ctxt->exception);
1278 }
1279 
1280 static int pio_in_emulated(struct x86_emulate_ctxt *ctxt,
1281  unsigned int size, unsigned short port,
1282  void *dest)
1283 {
1284  struct read_cache *rc = &ctxt->io_read;
1285 
1286  if (rc->pos == rc->end) { /* refill pio read ahead */
1287  unsigned int in_page, n;
1288  unsigned int count = ctxt->rep_prefix ?
1289  address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) : 1;
1290  in_page = (ctxt->eflags & EFLG_DF) ?
1293  n = min(min(in_page, (unsigned int)sizeof(rc->data)) / size,
1294  count);
1295  if (n == 0)
1296  n = 1;
1297  rc->pos = rc->end = 0;
1298  if (!ctxt->ops->pio_in_emulated(ctxt, size, port, rc->data, n))
1299  return 0;
1300  rc->end = n * size;
1301  }
1302 
1303  if (ctxt->rep_prefix && !(ctxt->eflags & EFLG_DF)) {
1304  ctxt->dst.data = rc->data + rc->pos;
1305  ctxt->dst.type = OP_MEM_STR;
1306  ctxt->dst.count = (rc->end - rc->pos) / size;
1307  rc->pos = rc->end;
1308  } else {
1309  memcpy(dest, rc->data + rc->pos, size);
1310  rc->pos += size;
1311  }
1312  return 1;
1313 }
1314 
1315 static int read_interrupt_descriptor(struct x86_emulate_ctxt *ctxt,
1316  u16 index, struct desc_struct *desc)
1317 {
1318  struct desc_ptr dt;
1319  ulong addr;
1320 
1321  ctxt->ops->get_idt(ctxt, &dt);
1322 
1323  if (dt.size < index * 8 + 7)
1324  return emulate_gp(ctxt, index << 3 | 0x2);
1325 
1326  addr = dt.address + index * 8;
1327  return ctxt->ops->read_std(ctxt, addr, desc, sizeof *desc,
1328  &ctxt->exception);
1329 }
1330 
1331 static void get_descriptor_table_ptr(struct x86_emulate_ctxt *ctxt,
1332  u16 selector, struct desc_ptr *dt)
1333 {
1334  const struct x86_emulate_ops *ops = ctxt->ops;
1335 
1336  if (selector & 1 << 2) {
1337  struct desc_struct desc;
1338  u16 sel;
1339 
1340  memset (dt, 0, sizeof *dt);
1341  if (!ops->get_segment(ctxt, &sel, &desc, NULL, VCPU_SREG_LDTR))
1342  return;
1343 
1344  dt->size = desc_limit_scaled(&desc); /* what if limit > 65535? */
1345  dt->address = get_desc_base(&desc);
1346  } else
1347  ops->get_gdt(ctxt, dt);
1348 }
1349 
1350 /* allowed just for 8 bytes segments */
1351 static int read_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1352  u16 selector, struct desc_struct *desc,
1353  ulong *desc_addr_p)
1354 {
1355  struct desc_ptr dt;
1356  u16 index = selector >> 3;
1357  ulong addr;
1358 
1359  get_descriptor_table_ptr(ctxt, selector, &dt);
1360 
1361  if (dt.size < index * 8 + 7)
1362  return emulate_gp(ctxt, selector & 0xfffc);
1363 
1364  *desc_addr_p = addr = dt.address + index * 8;
1365  return ctxt->ops->read_std(ctxt, addr, desc, sizeof *desc,
1366  &ctxt->exception);
1367 }
1368 
1369 /* allowed just for 8 bytes segments */
1370 static int write_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1371  u16 selector, struct desc_struct *desc)
1372 {
1373  struct desc_ptr dt;
1374  u16 index = selector >> 3;
1375  ulong addr;
1376 
1377  get_descriptor_table_ptr(ctxt, selector, &dt);
1378 
1379  if (dt.size < index * 8 + 7)
1380  return emulate_gp(ctxt, selector & 0xfffc);
1381 
1382  addr = dt.address + index * 8;
1383  return ctxt->ops->write_std(ctxt, addr, desc, sizeof *desc,
1384  &ctxt->exception);
1385 }
1386 
1387 /* Does not support long mode */
1388 static int load_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1389  u16 selector, int seg)
1390 {
1391  struct desc_struct seg_desc, old_desc;
1392  u8 dpl, rpl, cpl;
1393  unsigned err_vec = GP_VECTOR;
1394  u32 err_code = 0;
1395  bool null_selector = !(selector & ~0x3); /* 0000-0003 are null */
1396  ulong desc_addr;
1397  int ret;
1398  u16 dummy;
1399 
1400  memset(&seg_desc, 0, sizeof seg_desc);
1401 
1402  if ((seg <= VCPU_SREG_GS && ctxt->mode == X86EMUL_MODE_VM86)
1403  || ctxt->mode == X86EMUL_MODE_REAL) {
1404  /* set real mode segment descriptor */
1405  ctxt->ops->get_segment(ctxt, &dummy, &seg_desc, NULL, seg);
1406  set_desc_base(&seg_desc, selector << 4);
1407  goto load;
1408  }
1409 
1410  rpl = selector & 3;
1411  cpl = ctxt->ops->cpl(ctxt);
1412 
1413  /* NULL selector is not valid for TR, CS and SS (except for long mode) */
1414  if ((seg == VCPU_SREG_CS
1415  || (seg == VCPU_SREG_SS
1416  && (ctxt->mode != X86EMUL_MODE_PROT64 || rpl != cpl))
1417  || seg == VCPU_SREG_TR)
1418  && null_selector)
1419  goto exception;
1420 
1421  /* TR should be in GDT only */
1422  if (seg == VCPU_SREG_TR && (selector & (1 << 2)))
1423  goto exception;
1424 
1425  if (null_selector) /* for NULL selector skip all following checks */
1426  goto load;
1427 
1428  ret = read_segment_descriptor(ctxt, selector, &seg_desc, &desc_addr);
1429  if (ret != X86EMUL_CONTINUE)
1430  return ret;
1431 
1432  err_code = selector & 0xfffc;
1433  err_vec = GP_VECTOR;
1434 
1435  /* can't load system descriptor into segment selector */
1436  if (seg <= VCPU_SREG_GS && !seg_desc.s)
1437  goto exception;
1438 
1439  if (!seg_desc.p) {
1440  err_vec = (seg == VCPU_SREG_SS) ? SS_VECTOR : NP_VECTOR;
1441  goto exception;
1442  }
1443 
1444  dpl = seg_desc.dpl;
1445 
1446  switch (seg) {
1447  case VCPU_SREG_SS:
1448  /*
1449  * segment is not a writable data segment or segment
1450  * selector's RPL != CPL or segment selector's RPL != CPL
1451  */
1452  if (rpl != cpl || (seg_desc.type & 0xa) != 0x2 || dpl != cpl)
1453  goto exception;
1454  break;
1455  case VCPU_SREG_CS:
1456  if (!(seg_desc.type & 8))
1457  goto exception;
1458 
1459  if (seg_desc.type & 4) {
1460  /* conforming */
1461  if (dpl > cpl)
1462  goto exception;
1463  } else {
1464  /* nonconforming */
1465  if (rpl > cpl || dpl != cpl)
1466  goto exception;
1467  }
1468  /* CS(RPL) <- CPL */
1469  selector = (selector & 0xfffc) | cpl;
1470  break;
1471  case VCPU_SREG_TR:
1472  if (seg_desc.s || (seg_desc.type != 1 && seg_desc.type != 9))
1473  goto exception;
1474  old_desc = seg_desc;
1475  seg_desc.type |= 2; /* busy */
1476  ret = ctxt->ops->cmpxchg_emulated(ctxt, desc_addr, &old_desc, &seg_desc,
1477  sizeof(seg_desc), &ctxt->exception);
1478  if (ret != X86EMUL_CONTINUE)
1479  return ret;
1480  break;
1481  case VCPU_SREG_LDTR:
1482  if (seg_desc.s || seg_desc.type != 2)
1483  goto exception;
1484  break;
1485  default: /* DS, ES, FS, or GS */
1486  /*
1487  * segment is not a data or readable code segment or
1488  * ((segment is a data or nonconforming code segment)
1489  * and (both RPL and CPL > DPL))
1490  */
1491  if ((seg_desc.type & 0xa) == 0x8 ||
1492  (((seg_desc.type & 0xc) != 0xc) &&
1493  (rpl > dpl && cpl > dpl)))
1494  goto exception;
1495  break;
1496  }
1497 
1498  if (seg_desc.s) {
1499  /* mark segment as accessed */
1500  seg_desc.type |= 1;
1501  ret = write_segment_descriptor(ctxt, selector, &seg_desc);
1502  if (ret != X86EMUL_CONTINUE)
1503  return ret;
1504  }
1505 load:
1506  ctxt->ops->set_segment(ctxt, selector, &seg_desc, 0, seg);
1507  return X86EMUL_CONTINUE;
1508 exception:
1509  emulate_exception(ctxt, err_vec, err_code, true);
1510  return X86EMUL_PROPAGATE_FAULT;
1511 }
1512 
1513 static void write_register_operand(struct operand *op)
1514 {
1515  /* The 4-byte case *is* correct: in 64-bit mode we zero-extend. */
1516  switch (op->bytes) {
1517  case 1:
1518  *(u8 *)op->addr.reg = (u8)op->val;
1519  break;
1520  case 2:
1521  *(u16 *)op->addr.reg = (u16)op->val;
1522  break;
1523  case 4:
1524  *op->addr.reg = (u32)op->val;
1525  break; /* 64b: zero-extend */
1526  case 8:
1527  *op->addr.reg = op->val;
1528  break;
1529  }
1530 }
1531 
1532 static int writeback(struct x86_emulate_ctxt *ctxt)
1533 {
1534  int rc;
1535 
1536  switch (ctxt->dst.type) {
1537  case OP_REG:
1538  write_register_operand(&ctxt->dst);
1539  break;
1540  case OP_MEM:
1541  if (ctxt->lock_prefix)
1542  rc = segmented_cmpxchg(ctxt,
1543  ctxt->dst.addr.mem,
1544  &ctxt->dst.orig_val,
1545  &ctxt->dst.val,
1546  ctxt->dst.bytes);
1547  else
1548  rc = segmented_write(ctxt,
1549  ctxt->dst.addr.mem,
1550  &ctxt->dst.val,
1551  ctxt->dst.bytes);
1552  if (rc != X86EMUL_CONTINUE)
1553  return rc;
1554  break;
1555  case OP_MEM_STR:
1556  rc = segmented_write(ctxt,
1557  ctxt->dst.addr.mem,
1558  ctxt->dst.data,
1559  ctxt->dst.bytes * ctxt->dst.count);
1560  if (rc != X86EMUL_CONTINUE)
1561  return rc;
1562  break;
1563  case OP_XMM:
1564  write_sse_reg(ctxt, &ctxt->dst.vec_val, ctxt->dst.addr.xmm);
1565  break;
1566  case OP_MM:
1567  write_mmx_reg(ctxt, &ctxt->dst.mm_val, ctxt->dst.addr.mm);
1568  break;
1569  case OP_NONE:
1570  /* no writeback */
1571  break;
1572  default:
1573  break;
1574  }
1575  return X86EMUL_CONTINUE;
1576 }
1577 
1578 static int push(struct x86_emulate_ctxt *ctxt, void *data, int bytes)
1579 {
1580  struct segmented_address addr;
1581 
1582  rsp_increment(ctxt, -bytes);
1583  addr.ea = reg_read(ctxt, VCPU_REGS_RSP) & stack_mask(ctxt);
1584  addr.seg = VCPU_SREG_SS;
1585 
1586  return segmented_write(ctxt, addr, data, bytes);
1587 }
1588 
1589 static int em_push(struct x86_emulate_ctxt *ctxt)
1590 {
1591  /* Disable writeback. */
1592  ctxt->dst.type = OP_NONE;
1593  return push(ctxt, &ctxt->src.val, ctxt->op_bytes);
1594 }
1595 
1596 static int emulate_pop(struct x86_emulate_ctxt *ctxt,
1597  void *dest, int len)
1598 {
1599  int rc;
1600  struct segmented_address addr;
1601 
1602  addr.ea = reg_read(ctxt, VCPU_REGS_RSP) & stack_mask(ctxt);
1603  addr.seg = VCPU_SREG_SS;
1604  rc = segmented_read(ctxt, addr, dest, len);
1605  if (rc != X86EMUL_CONTINUE)
1606  return rc;
1607 
1608  rsp_increment(ctxt, len);
1609  return rc;
1610 }
1611 
1612 static int em_pop(struct x86_emulate_ctxt *ctxt)
1613 {
1614  return emulate_pop(ctxt, &ctxt->dst.val, ctxt->op_bytes);
1615 }
1616 
1617 static int emulate_popf(struct x86_emulate_ctxt *ctxt,
1618  void *dest, int len)
1619 {
1620  int rc;
1621  unsigned long val, change_mask;
1622  int iopl = (ctxt->eflags & X86_EFLAGS_IOPL) >> IOPL_SHIFT;
1623  int cpl = ctxt->ops->cpl(ctxt);
1624 
1625  rc = emulate_pop(ctxt, &val, len);
1626  if (rc != X86EMUL_CONTINUE)
1627  return rc;
1628 
1629  change_mask = EFLG_CF | EFLG_PF | EFLG_AF | EFLG_ZF | EFLG_SF | EFLG_OF
1631 
1632  switch(ctxt->mode) {
1633  case X86EMUL_MODE_PROT64:
1634  case X86EMUL_MODE_PROT32:
1635  case X86EMUL_MODE_PROT16:
1636  if (cpl == 0)
1637  change_mask |= EFLG_IOPL;
1638  if (cpl <= iopl)
1639  change_mask |= EFLG_IF;
1640  break;
1641  case X86EMUL_MODE_VM86:
1642  if (iopl < 3)
1643  return emulate_gp(ctxt, 0);
1644  change_mask |= EFLG_IF;
1645  break;
1646  default: /* real mode */
1647  change_mask |= (EFLG_IOPL | EFLG_IF);
1648  break;
1649  }
1650 
1651  *(unsigned long *)dest =
1652  (ctxt->eflags & ~change_mask) | (val & change_mask);
1653 
1654  return rc;
1655 }
1656 
1657 static int em_popf(struct x86_emulate_ctxt *ctxt)
1658 {
1659  ctxt->dst.type = OP_REG;
1660  ctxt->dst.addr.reg = &ctxt->eflags;
1661  ctxt->dst.bytes = ctxt->op_bytes;
1662  return emulate_popf(ctxt, &ctxt->dst.val, ctxt->op_bytes);
1663 }
1664 
1665 static int em_enter(struct x86_emulate_ctxt *ctxt)
1666 {
1667  int rc;
1668  unsigned frame_size = ctxt->src.val;
1669  unsigned nesting_level = ctxt->src2.val & 31;
1670  ulong rbp;
1671 
1672  if (nesting_level)
1673  return X86EMUL_UNHANDLEABLE;
1674 
1675  rbp = reg_read(ctxt, VCPU_REGS_RBP);
1676  rc = push(ctxt, &rbp, stack_size(ctxt));
1677  if (rc != X86EMUL_CONTINUE)
1678  return rc;
1679  assign_masked(reg_rmw(ctxt, VCPU_REGS_RBP), reg_read(ctxt, VCPU_REGS_RSP),
1680  stack_mask(ctxt));
1681  assign_masked(reg_rmw(ctxt, VCPU_REGS_RSP),
1682  reg_read(ctxt, VCPU_REGS_RSP) - frame_size,
1683  stack_mask(ctxt));
1684  return X86EMUL_CONTINUE;
1685 }
1686 
1687 static int em_leave(struct x86_emulate_ctxt *ctxt)
1688 {
1689  assign_masked(reg_rmw(ctxt, VCPU_REGS_RSP), reg_read(ctxt, VCPU_REGS_RBP),
1690  stack_mask(ctxt));
1691  return emulate_pop(ctxt, reg_rmw(ctxt, VCPU_REGS_RBP), ctxt->op_bytes);
1692 }
1693 
1694 static int em_push_sreg(struct x86_emulate_ctxt *ctxt)
1695 {
1696  int seg = ctxt->src2.val;
1697 
1698  ctxt->src.val = get_segment_selector(ctxt, seg);
1699 
1700  return em_push(ctxt);
1701 }
1702 
1703 static int em_pop_sreg(struct x86_emulate_ctxt *ctxt)
1704 {
1705  int seg = ctxt->src2.val;
1706  unsigned long selector;
1707  int rc;
1708 
1709  rc = emulate_pop(ctxt, &selector, ctxt->op_bytes);
1710  if (rc != X86EMUL_CONTINUE)
1711  return rc;
1712 
1713  rc = load_segment_descriptor(ctxt, (u16)selector, seg);
1714  return rc;
1715 }
1716 
1717 static int em_pusha(struct x86_emulate_ctxt *ctxt)
1718 {
1719  unsigned long old_esp = reg_read(ctxt, VCPU_REGS_RSP);
1720  int rc = X86EMUL_CONTINUE;
1721  int reg = VCPU_REGS_RAX;
1722 
1723  while (reg <= VCPU_REGS_RDI) {
1724  (reg == VCPU_REGS_RSP) ?
1725  (ctxt->src.val = old_esp) : (ctxt->src.val = reg_read(ctxt, reg));
1726 
1727  rc = em_push(ctxt);
1728  if (rc != X86EMUL_CONTINUE)
1729  return rc;
1730 
1731  ++reg;
1732  }
1733 
1734  return rc;
1735 }
1736 
1737 static int em_pushf(struct x86_emulate_ctxt *ctxt)
1738 {
1739  ctxt->src.val = (unsigned long)ctxt->eflags;
1740  return em_push(ctxt);
1741 }
1742 
1743 static int em_popa(struct x86_emulate_ctxt *ctxt)
1744 {
1745  int rc = X86EMUL_CONTINUE;
1746  int reg = VCPU_REGS_RDI;
1747 
1748  while (reg >= VCPU_REGS_RAX) {
1749  if (reg == VCPU_REGS_RSP) {
1750  rsp_increment(ctxt, ctxt->op_bytes);
1751  --reg;
1752  }
1753 
1754  rc = emulate_pop(ctxt, reg_rmw(ctxt, reg), ctxt->op_bytes);
1755  if (rc != X86EMUL_CONTINUE)
1756  break;
1757  --reg;
1758  }
1759  return rc;
1760 }
1761 
1762 static int __emulate_int_real(struct x86_emulate_ctxt *ctxt, int irq)
1763 {
1764  const struct x86_emulate_ops *ops = ctxt->ops;
1765  int rc;
1766  struct desc_ptr dt;
1767  gva_t cs_addr;
1768  gva_t eip_addr;
1769  u16 cs, eip;
1770 
1771  /* TODO: Add limit checks */
1772  ctxt->src.val = ctxt->eflags;
1773  rc = em_push(ctxt);
1774  if (rc != X86EMUL_CONTINUE)
1775  return rc;
1776 
1777  ctxt->eflags &= ~(EFLG_IF | EFLG_TF | EFLG_AC);
1778 
1779  ctxt->src.val = get_segment_selector(ctxt, VCPU_SREG_CS);
1780  rc = em_push(ctxt);
1781  if (rc != X86EMUL_CONTINUE)
1782  return rc;
1783 
1784  ctxt->src.val = ctxt->_eip;
1785  rc = em_push(ctxt);
1786  if (rc != X86EMUL_CONTINUE)
1787  return rc;
1788 
1789  ops->get_idt(ctxt, &dt);
1790 
1791  eip_addr = dt.address + (irq << 2);
1792  cs_addr = dt.address + (irq << 2) + 2;
1793 
1794  rc = ops->read_std(ctxt, cs_addr, &cs, 2, &ctxt->exception);
1795  if (rc != X86EMUL_CONTINUE)
1796  return rc;
1797 
1798  rc = ops->read_std(ctxt, eip_addr, &eip, 2, &ctxt->exception);
1799  if (rc != X86EMUL_CONTINUE)
1800  return rc;
1801 
1802  rc = load_segment_descriptor(ctxt, cs, VCPU_SREG_CS);
1803  if (rc != X86EMUL_CONTINUE)
1804  return rc;
1805 
1806  ctxt->_eip = eip;
1807 
1808  return rc;
1809 }
1810 
1811 int emulate_int_real(struct x86_emulate_ctxt *ctxt, int irq)
1812 {
1813  int rc;
1814 
1815  invalidate_registers(ctxt);
1816  rc = __emulate_int_real(ctxt, irq);
1817  if (rc == X86EMUL_CONTINUE)
1818  writeback_registers(ctxt);
1819  return rc;
1820 }
1821 
1822 static int emulate_int(struct x86_emulate_ctxt *ctxt, int irq)
1823 {
1824  switch(ctxt->mode) {
1825  case X86EMUL_MODE_REAL:
1826  return __emulate_int_real(ctxt, irq);
1827  case X86EMUL_MODE_VM86:
1828  case X86EMUL_MODE_PROT16:
1829  case X86EMUL_MODE_PROT32:
1830  case X86EMUL_MODE_PROT64:
1831  default:
1832  /* Protected mode interrupts unimplemented yet */
1833  return X86EMUL_UNHANDLEABLE;
1834  }
1835 }
1836 
1837 static int emulate_iret_real(struct x86_emulate_ctxt *ctxt)
1838 {
1839  int rc = X86EMUL_CONTINUE;
1840  unsigned long temp_eip = 0;
1841  unsigned long temp_eflags = 0;
1842  unsigned long cs = 0;
1843  unsigned long mask = EFLG_CF | EFLG_PF | EFLG_AF | EFLG_ZF | EFLG_SF | EFLG_TF |
1845  EFLG_AC | EFLG_ID | (1 << 1); /* Last one is the reserved bit */
1846  unsigned long vm86_mask = EFLG_VM | EFLG_VIF | EFLG_VIP;
1847 
1848  /* TODO: Add stack limit check */
1849 
1850  rc = emulate_pop(ctxt, &temp_eip, ctxt->op_bytes);
1851 
1852  if (rc != X86EMUL_CONTINUE)
1853  return rc;
1854 
1855  if (temp_eip & ~0xffff)
1856  return emulate_gp(ctxt, 0);
1857 
1858  rc = emulate_pop(ctxt, &cs, ctxt->op_bytes);
1859 
1860  if (rc != X86EMUL_CONTINUE)
1861  return rc;
1862 
1863  rc = emulate_pop(ctxt, &temp_eflags, ctxt->op_bytes);
1864 
1865  if (rc != X86EMUL_CONTINUE)
1866  return rc;
1867 
1868  rc = load_segment_descriptor(ctxt, (u16)cs, VCPU_SREG_CS);
1869 
1870  if (rc != X86EMUL_CONTINUE)
1871  return rc;
1872 
1873  ctxt->_eip = temp_eip;
1874 
1875 
1876  if (ctxt->op_bytes == 4)
1877  ctxt->eflags = ((temp_eflags & mask) | (ctxt->eflags & vm86_mask));
1878  else if (ctxt->op_bytes == 2) {
1879  ctxt->eflags &= ~0xffff;
1880  ctxt->eflags |= temp_eflags;
1881  }
1882 
1883  ctxt->eflags &= ~EFLG_RESERVED_ZEROS_MASK; /* Clear reserved zeros */
1884  ctxt->eflags |= EFLG_RESERVED_ONE_MASK;
1885 
1886  return rc;
1887 }
1888 
1889 static int em_iret(struct x86_emulate_ctxt *ctxt)
1890 {
1891  switch(ctxt->mode) {
1892  case X86EMUL_MODE_REAL:
1893  return emulate_iret_real(ctxt);
1894  case X86EMUL_MODE_VM86:
1895  case X86EMUL_MODE_PROT16:
1896  case X86EMUL_MODE_PROT32:
1897  case X86EMUL_MODE_PROT64:
1898  default:
1899  /* iret from protected mode unimplemented yet */
1900  return X86EMUL_UNHANDLEABLE;
1901  }
1902 }
1903 
1904 static int em_jmp_far(struct x86_emulate_ctxt *ctxt)
1905 {
1906  int rc;
1907  unsigned short sel;
1908 
1909  memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
1910 
1911  rc = load_segment_descriptor(ctxt, sel, VCPU_SREG_CS);
1912  if (rc != X86EMUL_CONTINUE)
1913  return rc;
1914 
1915  ctxt->_eip = 0;
1916  memcpy(&ctxt->_eip, ctxt->src.valptr, ctxt->op_bytes);
1917  return X86EMUL_CONTINUE;
1918 }
1919 
1920 static int em_grp2(struct x86_emulate_ctxt *ctxt)
1921 {
1922  switch (ctxt->modrm_reg) {
1923  case 0: /* rol */
1924  emulate_2op_SrcB(ctxt, "rol");
1925  break;
1926  case 1: /* ror */
1927  emulate_2op_SrcB(ctxt, "ror");
1928  break;
1929  case 2: /* rcl */
1930  emulate_2op_SrcB(ctxt, "rcl");
1931  break;
1932  case 3: /* rcr */
1933  emulate_2op_SrcB(ctxt, "rcr");
1934  break;
1935  case 4: /* sal/shl */
1936  case 6: /* sal/shl */
1937  emulate_2op_SrcB(ctxt, "sal");
1938  break;
1939  case 5: /* shr */
1940  emulate_2op_SrcB(ctxt, "shr");
1941  break;
1942  case 7: /* sar */
1943  emulate_2op_SrcB(ctxt, "sar");
1944  break;
1945  }
1946  return X86EMUL_CONTINUE;
1947 }
1948 
1949 static int em_not(struct x86_emulate_ctxt *ctxt)
1950 {
1951  ctxt->dst.val = ~ctxt->dst.val;
1952  return X86EMUL_CONTINUE;
1953 }
1954 
1955 static int em_neg(struct x86_emulate_ctxt *ctxt)
1956 {
1957  emulate_1op(ctxt, "neg");
1958  return X86EMUL_CONTINUE;
1959 }
1960 
1961 static int em_mul_ex(struct x86_emulate_ctxt *ctxt)
1962 {
1963  u8 ex = 0;
1964 
1965  emulate_1op_rax_rdx(ctxt, "mul", ex);
1966  return X86EMUL_CONTINUE;
1967 }
1968 
1969 static int em_imul_ex(struct x86_emulate_ctxt *ctxt)
1970 {
1971  u8 ex = 0;
1972 
1973  emulate_1op_rax_rdx(ctxt, "imul", ex);
1974  return X86EMUL_CONTINUE;
1975 }
1976 
1977 static int em_div_ex(struct x86_emulate_ctxt *ctxt)
1978 {
1979  u8 de = 0;
1980 
1981  emulate_1op_rax_rdx(ctxt, "div", de);
1982  if (de)
1983  return emulate_de(ctxt);
1984  return X86EMUL_CONTINUE;
1985 }
1986 
1987 static int em_idiv_ex(struct x86_emulate_ctxt *ctxt)
1988 {
1989  u8 de = 0;
1990 
1991  emulate_1op_rax_rdx(ctxt, "idiv", de);
1992  if (de)
1993  return emulate_de(ctxt);
1994  return X86EMUL_CONTINUE;
1995 }
1996 
1997 static int em_grp45(struct x86_emulate_ctxt *ctxt)
1998 {
1999  int rc = X86EMUL_CONTINUE;
2000 
2001  switch (ctxt->modrm_reg) {
2002  case 0: /* inc */
2003  emulate_1op(ctxt, "inc");
2004  break;
2005  case 1: /* dec */
2006  emulate_1op(ctxt, "dec");
2007  break;
2008  case 2: /* call near abs */ {
2009  long int old_eip;
2010  old_eip = ctxt->_eip;
2011  ctxt->_eip = ctxt->src.val;
2012  ctxt->src.val = old_eip;
2013  rc = em_push(ctxt);
2014  break;
2015  }
2016  case 4: /* jmp abs */
2017  ctxt->_eip = ctxt->src.val;
2018  break;
2019  case 5: /* jmp far */
2020  rc = em_jmp_far(ctxt);
2021  break;
2022  case 6: /* push */
2023  rc = em_push(ctxt);
2024  break;
2025  }
2026  return rc;
2027 }
2028 
2029 static int em_cmpxchg8b(struct x86_emulate_ctxt *ctxt)
2030 {
2031  u64 old = ctxt->dst.orig_val64;
2032 
2033  if (((u32) (old >> 0) != (u32) reg_read(ctxt, VCPU_REGS_RAX)) ||
2034  ((u32) (old >> 32) != (u32) reg_read(ctxt, VCPU_REGS_RDX))) {
2035  *reg_write(ctxt, VCPU_REGS_RAX) = (u32) (old >> 0);
2036  *reg_write(ctxt, VCPU_REGS_RDX) = (u32) (old >> 32);
2037  ctxt->eflags &= ~EFLG_ZF;
2038  } else {
2039  ctxt->dst.val64 = ((u64)reg_read(ctxt, VCPU_REGS_RCX) << 32) |
2040  (u32) reg_read(ctxt, VCPU_REGS_RBX);
2041 
2042  ctxt->eflags |= EFLG_ZF;
2043  }
2044  return X86EMUL_CONTINUE;
2045 }
2046 
2047 static int em_ret(struct x86_emulate_ctxt *ctxt)
2048 {
2049  ctxt->dst.type = OP_REG;
2050  ctxt->dst.addr.reg = &ctxt->_eip;
2051  ctxt->dst.bytes = ctxt->op_bytes;
2052  return em_pop(ctxt);
2053 }
2054 
2055 static int em_ret_far(struct x86_emulate_ctxt *ctxt)
2056 {
2057  int rc;
2058  unsigned long cs;
2059 
2060  rc = emulate_pop(ctxt, &ctxt->_eip, ctxt->op_bytes);
2061  if (rc != X86EMUL_CONTINUE)
2062  return rc;
2063  if (ctxt->op_bytes == 4)
2064  ctxt->_eip = (u32)ctxt->_eip;
2065  rc = emulate_pop(ctxt, &cs, ctxt->op_bytes);
2066  if (rc != X86EMUL_CONTINUE)
2067  return rc;
2068  rc = load_segment_descriptor(ctxt, (u16)cs, VCPU_SREG_CS);
2069  return rc;
2070 }
2071 
2072 static int em_cmpxchg(struct x86_emulate_ctxt *ctxt)
2073 {
2074  /* Save real source value, then compare EAX against destination. */
2075  ctxt->src.orig_val = ctxt->src.val;
2076  ctxt->src.val = reg_read(ctxt, VCPU_REGS_RAX);
2077  emulate_2op_SrcV(ctxt, "cmp");
2078 
2079  if (ctxt->eflags & EFLG_ZF) {
2080  /* Success: write back to memory. */
2081  ctxt->dst.val = ctxt->src.orig_val;
2082  } else {
2083  /* Failure: write the value we saw to EAX. */
2084  ctxt->dst.type = OP_REG;
2085  ctxt->dst.addr.reg = reg_rmw(ctxt, VCPU_REGS_RAX);
2086  }
2087  return X86EMUL_CONTINUE;
2088 }
2089 
2090 static int em_lseg(struct x86_emulate_ctxt *ctxt)
2091 {
2092  int seg = ctxt->src2.val;
2093  unsigned short sel;
2094  int rc;
2095 
2096  memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
2097 
2098  rc = load_segment_descriptor(ctxt, sel, seg);
2099  if (rc != X86EMUL_CONTINUE)
2100  return rc;
2101 
2102  ctxt->dst.val = ctxt->src.val;
2103  return rc;
2104 }
2105 
2106 static void
2107 setup_syscalls_segments(struct x86_emulate_ctxt *ctxt,
2108  struct desc_struct *cs, struct desc_struct *ss)
2109 {
2110  cs->l = 0; /* will be adjusted later */
2111  set_desc_base(cs, 0); /* flat segment */
2112  cs->g = 1; /* 4kb granularity */
2113  set_desc_limit(cs, 0xfffff); /* 4GB limit */
2114  cs->type = 0x0b; /* Read, Execute, Accessed */
2115  cs->s = 1;
2116  cs->dpl = 0; /* will be adjusted later */
2117  cs->p = 1;
2118  cs->d = 1;
2119  cs->avl = 0;
2120 
2121  set_desc_base(ss, 0); /* flat segment */
2122  set_desc_limit(ss, 0xfffff); /* 4GB limit */
2123  ss->g = 1; /* 4kb granularity */
2124  ss->s = 1;
2125  ss->type = 0x03; /* Read/Write, Accessed */
2126  ss->d = 1; /* 32bit stack segment */
2127  ss->dpl = 0;
2128  ss->p = 1;
2129  ss->l = 0;
2130  ss->avl = 0;
2131 }
2132 
2133 static bool vendor_intel(struct x86_emulate_ctxt *ctxt)
2134 {
2135  u32 eax, ebx, ecx, edx;
2136 
2137  eax = ecx = 0;
2138  ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
2142 }
2143 
2144 static bool em_syscall_is_enabled(struct x86_emulate_ctxt *ctxt)
2145 {
2146  const struct x86_emulate_ops *ops = ctxt->ops;
2147  u32 eax, ebx, ecx, edx;
2148 
2149  /*
2150  * syscall should always be enabled in longmode - so only become
2151  * vendor specific (cpuid) if other modes are active...
2152  */
2153  if (ctxt->mode == X86EMUL_MODE_PROT64)
2154  return true;
2155 
2156  eax = 0x00000000;
2157  ecx = 0x00000000;
2158  ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
2159  /*
2160  * Intel ("GenuineIntel")
2161  * remark: Intel CPUs only support "syscall" in 64bit
2162  * longmode. Also an 64bit guest with a
2163  * 32bit compat-app running will #UD !! While this
2164  * behaviour can be fixed (by emulating) into AMD
2165  * response - CPUs of AMD can't behave like Intel.
2166  */
2170  return false;
2171 
2172  /* AMD ("AuthenticAMD") */
2176  return true;
2177 
2178  /* AMD ("AMDisbetter!") */
2182  return true;
2183 
2184  /* default: (not Intel, not AMD), apply Intel's stricter rules... */
2185  return false;
2186 }
2187 
2188 static int em_syscall(struct x86_emulate_ctxt *ctxt)
2189 {
2190  const struct x86_emulate_ops *ops = ctxt->ops;
2191  struct desc_struct cs, ss;
2192  u64 msr_data;
2193  u16 cs_sel, ss_sel;
2194  u64 efer = 0;
2195 
2196  /* syscall is not available in real mode */
2197  if (ctxt->mode == X86EMUL_MODE_REAL ||
2198  ctxt->mode == X86EMUL_MODE_VM86)
2199  return emulate_ud(ctxt);
2200 
2201  if (!(em_syscall_is_enabled(ctxt)))
2202  return emulate_ud(ctxt);
2203 
2204  ops->get_msr(ctxt, MSR_EFER, &efer);
2205  setup_syscalls_segments(ctxt, &cs, &ss);
2206 
2207  if (!(efer & EFER_SCE))
2208  return emulate_ud(ctxt);
2209 
2210  ops->get_msr(ctxt, MSR_STAR, &msr_data);
2211  msr_data >>= 32;
2212  cs_sel = (u16)(msr_data & 0xfffc);
2213  ss_sel = (u16)(msr_data + 8);
2214 
2215  if (efer & EFER_LMA) {
2216  cs.d = 0;
2217  cs.l = 1;
2218  }
2219  ops->set_segment(ctxt, cs_sel, &cs, 0, VCPU_SREG_CS);
2220  ops->set_segment(ctxt, ss_sel, &ss, 0, VCPU_SREG_SS);
2221 
2222  *reg_write(ctxt, VCPU_REGS_RCX) = ctxt->_eip;
2223  if (efer & EFER_LMA) {
2224 #ifdef CONFIG_X86_64
2225  *reg_write(ctxt, VCPU_REGS_R11) = ctxt->eflags & ~EFLG_RF;
2226 
2227  ops->get_msr(ctxt,
2228  ctxt->mode == X86EMUL_MODE_PROT64 ?
2229  MSR_LSTAR : MSR_CSTAR, &msr_data);
2230  ctxt->_eip = msr_data;
2231 
2232  ops->get_msr(ctxt, MSR_SYSCALL_MASK, &msr_data);
2233  ctxt->eflags &= ~(msr_data | EFLG_RF);
2234 #endif
2235  } else {
2236  /* legacy mode */
2237  ops->get_msr(ctxt, MSR_STAR, &msr_data);
2238  ctxt->_eip = (u32)msr_data;
2239 
2240  ctxt->eflags &= ~(EFLG_VM | EFLG_IF | EFLG_RF);
2241  }
2242 
2243  return X86EMUL_CONTINUE;
2244 }
2245 
2246 static int em_sysenter(struct x86_emulate_ctxt *ctxt)
2247 {
2248  const struct x86_emulate_ops *ops = ctxt->ops;
2249  struct desc_struct cs, ss;
2250  u64 msr_data;
2251  u16 cs_sel, ss_sel;
2252  u64 efer = 0;
2253 
2254  ops->get_msr(ctxt, MSR_EFER, &efer);
2255  /* inject #GP if in real mode */
2256  if (ctxt->mode == X86EMUL_MODE_REAL)
2257  return emulate_gp(ctxt, 0);
2258 
2259  /*
2260  * Not recognized on AMD in compat mode (but is recognized in legacy
2261  * mode).
2262  */
2263  if ((ctxt->mode == X86EMUL_MODE_PROT32) && (efer & EFER_LMA)
2264  && !vendor_intel(ctxt))
2265  return emulate_ud(ctxt);
2266 
2267  /* XXX sysenter/sysexit have not been tested in 64bit mode.
2268  * Therefore, we inject an #UD.
2269  */
2270  if (ctxt->mode == X86EMUL_MODE_PROT64)
2271  return emulate_ud(ctxt);
2272 
2273  setup_syscalls_segments(ctxt, &cs, &ss);
2274 
2275  ops->get_msr(ctxt, MSR_IA32_SYSENTER_CS, &msr_data);
2276  switch (ctxt->mode) {
2277  case X86EMUL_MODE_PROT32:
2278  if ((msr_data & 0xfffc) == 0x0)
2279  return emulate_gp(ctxt, 0);
2280  break;
2281  case X86EMUL_MODE_PROT64:
2282  if (msr_data == 0x0)
2283  return emulate_gp(ctxt, 0);
2284  break;
2285  default:
2286  break;
2287  }
2288 
2289  ctxt->eflags &= ~(EFLG_VM | EFLG_IF | EFLG_RF);
2290  cs_sel = (u16)msr_data;
2291  cs_sel &= ~SELECTOR_RPL_MASK;
2292  ss_sel = cs_sel + 8;
2293  ss_sel &= ~SELECTOR_RPL_MASK;
2294  if (ctxt->mode == X86EMUL_MODE_PROT64 || (efer & EFER_LMA)) {
2295  cs.d = 0;
2296  cs.l = 1;
2297  }
2298 
2299  ops->set_segment(ctxt, cs_sel, &cs, 0, VCPU_SREG_CS);
2300  ops->set_segment(ctxt, ss_sel, &ss, 0, VCPU_SREG_SS);
2301 
2302  ops->get_msr(ctxt, MSR_IA32_SYSENTER_EIP, &msr_data);
2303  ctxt->_eip = msr_data;
2304 
2305  ops->get_msr(ctxt, MSR_IA32_SYSENTER_ESP, &msr_data);
2306  *reg_write(ctxt, VCPU_REGS_RSP) = msr_data;
2307 
2308  return X86EMUL_CONTINUE;
2309 }
2310 
2311 static int em_sysexit(struct x86_emulate_ctxt *ctxt)
2312 {
2313  const struct x86_emulate_ops *ops = ctxt->ops;
2314  struct desc_struct cs, ss;
2315  u64 msr_data;
2316  int usermode;
2317  u16 cs_sel = 0, ss_sel = 0;
2318 
2319  /* inject #GP if in real mode or Virtual 8086 mode */
2320  if (ctxt->mode == X86EMUL_MODE_REAL ||
2321  ctxt->mode == X86EMUL_MODE_VM86)
2322  return emulate_gp(ctxt, 0);
2323 
2324  setup_syscalls_segments(ctxt, &cs, &ss);
2325 
2326  if ((ctxt->rex_prefix & 0x8) != 0x0)
2327  usermode = X86EMUL_MODE_PROT64;
2328  else
2329  usermode = X86EMUL_MODE_PROT32;
2330 
2331  cs.dpl = 3;
2332  ss.dpl = 3;
2333  ops->get_msr(ctxt, MSR_IA32_SYSENTER_CS, &msr_data);
2334  switch (usermode) {
2335  case X86EMUL_MODE_PROT32:
2336  cs_sel = (u16)(msr_data + 16);
2337  if ((msr_data & 0xfffc) == 0x0)
2338  return emulate_gp(ctxt, 0);
2339  ss_sel = (u16)(msr_data + 24);
2340  break;
2341  case X86EMUL_MODE_PROT64:
2342  cs_sel = (u16)(msr_data + 32);
2343  if (msr_data == 0x0)
2344  return emulate_gp(ctxt, 0);
2345  ss_sel = cs_sel + 8;
2346  cs.d = 0;
2347  cs.l = 1;
2348  break;
2349  }
2350  cs_sel |= SELECTOR_RPL_MASK;
2351  ss_sel |= SELECTOR_RPL_MASK;
2352 
2353  ops->set_segment(ctxt, cs_sel, &cs, 0, VCPU_SREG_CS);
2354  ops->set_segment(ctxt, ss_sel, &ss, 0, VCPU_SREG_SS);
2355 
2356  ctxt->_eip = reg_read(ctxt, VCPU_REGS_RDX);
2357  *reg_write(ctxt, VCPU_REGS_RSP) = reg_read(ctxt, VCPU_REGS_RCX);
2358 
2359  return X86EMUL_CONTINUE;
2360 }
2361 
2362 static bool emulator_bad_iopl(struct x86_emulate_ctxt *ctxt)
2363 {
2364  int iopl;
2365  if (ctxt->mode == X86EMUL_MODE_REAL)
2366  return false;
2367  if (ctxt->mode == X86EMUL_MODE_VM86)
2368  return true;
2369  iopl = (ctxt->eflags & X86_EFLAGS_IOPL) >> IOPL_SHIFT;
2370  return ctxt->ops->cpl(ctxt) > iopl;
2371 }
2372 
2373 static bool emulator_io_port_access_allowed(struct x86_emulate_ctxt *ctxt,
2374  u16 port, u16 len)
2375 {
2376  const struct x86_emulate_ops *ops = ctxt->ops;
2377  struct desc_struct tr_seg;
2378  u32 base3;
2379  int r;
2380  u16 tr, io_bitmap_ptr, perm, bit_idx = port & 0x7;
2381  unsigned mask = (1 << len) - 1;
2382  unsigned long base;
2383 
2384  ops->get_segment(ctxt, &tr, &tr_seg, &base3, VCPU_SREG_TR);
2385  if (!tr_seg.p)
2386  return false;
2387  if (desc_limit_scaled(&tr_seg) < 103)
2388  return false;
2389  base = get_desc_base(&tr_seg);
2390 #ifdef CONFIG_X86_64
2391  base |= ((u64)base3) << 32;
2392 #endif
2393  r = ops->read_std(ctxt, base + 102, &io_bitmap_ptr, 2, NULL);
2394  if (r != X86EMUL_CONTINUE)
2395  return false;
2396  if (io_bitmap_ptr + port/8 > desc_limit_scaled(&tr_seg))
2397  return false;
2398  r = ops->read_std(ctxt, base + io_bitmap_ptr + port/8, &perm, 2, NULL);
2399  if (r != X86EMUL_CONTINUE)
2400  return false;
2401  if ((perm >> bit_idx) & mask)
2402  return false;
2403  return true;
2404 }
2405 
2406 static bool emulator_io_permited(struct x86_emulate_ctxt *ctxt,
2407  u16 port, u16 len)
2408 {
2409  if (ctxt->perm_ok)
2410  return true;
2411 
2412  if (emulator_bad_iopl(ctxt))
2413  if (!emulator_io_port_access_allowed(ctxt, port, len))
2414  return false;
2415 
2416  ctxt->perm_ok = true;
2417 
2418  return true;
2419 }
2420 
2421 static void save_state_to_tss16(struct x86_emulate_ctxt *ctxt,
2422  struct tss_segment_16 *tss)
2423 {
2424  tss->ip = ctxt->_eip;
2425  tss->flag = ctxt->eflags;
2426  tss->ax = reg_read(ctxt, VCPU_REGS_RAX);
2427  tss->cx = reg_read(ctxt, VCPU_REGS_RCX);
2428  tss->dx = reg_read(ctxt, VCPU_REGS_RDX);
2429  tss->bx = reg_read(ctxt, VCPU_REGS_RBX);
2430  tss->sp = reg_read(ctxt, VCPU_REGS_RSP);
2431  tss->bp = reg_read(ctxt, VCPU_REGS_RBP);
2432  tss->si = reg_read(ctxt, VCPU_REGS_RSI);
2433  tss->di = reg_read(ctxt, VCPU_REGS_RDI);
2434 
2435  tss->es = get_segment_selector(ctxt, VCPU_SREG_ES);
2436  tss->cs = get_segment_selector(ctxt, VCPU_SREG_CS);
2437  tss->ss = get_segment_selector(ctxt, VCPU_SREG_SS);
2438  tss->ds = get_segment_selector(ctxt, VCPU_SREG_DS);
2439  tss->ldt = get_segment_selector(ctxt, VCPU_SREG_LDTR);
2440 }
2441 
2442 static int load_state_from_tss16(struct x86_emulate_ctxt *ctxt,
2443  struct tss_segment_16 *tss)
2444 {
2445  int ret;
2446 
2447  ctxt->_eip = tss->ip;
2448  ctxt->eflags = tss->flag | 2;
2449  *reg_write(ctxt, VCPU_REGS_RAX) = tss->ax;
2450  *reg_write(ctxt, VCPU_REGS_RCX) = tss->cx;
2451  *reg_write(ctxt, VCPU_REGS_RDX) = tss->dx;
2452  *reg_write(ctxt, VCPU_REGS_RBX) = tss->bx;
2453  *reg_write(ctxt, VCPU_REGS_RSP) = tss->sp;
2454  *reg_write(ctxt, VCPU_REGS_RBP) = tss->bp;
2455  *reg_write(ctxt, VCPU_REGS_RSI) = tss->si;
2456  *reg_write(ctxt, VCPU_REGS_RDI) = tss->di;
2457 
2458  /*
2459  * SDM says that segment selectors are loaded before segment
2460  * descriptors
2461  */
2462  set_segment_selector(ctxt, tss->ldt, VCPU_SREG_LDTR);
2463  set_segment_selector(ctxt, tss->es, VCPU_SREG_ES);
2464  set_segment_selector(ctxt, tss->cs, VCPU_SREG_CS);
2465  set_segment_selector(ctxt, tss->ss, VCPU_SREG_SS);
2466  set_segment_selector(ctxt, tss->ds, VCPU_SREG_DS);
2467 
2468  /*
2469  * Now load segment descriptors. If fault happens at this stage
2470  * it is handled in a context of new task
2471  */
2472  ret = load_segment_descriptor(ctxt, tss->ldt, VCPU_SREG_LDTR);
2473  if (ret != X86EMUL_CONTINUE)
2474  return ret;
2475  ret = load_segment_descriptor(ctxt, tss->es, VCPU_SREG_ES);
2476  if (ret != X86EMUL_CONTINUE)
2477  return ret;
2478  ret = load_segment_descriptor(ctxt, tss->cs, VCPU_SREG_CS);
2479  if (ret != X86EMUL_CONTINUE)
2480  return ret;
2481  ret = load_segment_descriptor(ctxt, tss->ss, VCPU_SREG_SS);
2482  if (ret != X86EMUL_CONTINUE)
2483  return ret;
2484  ret = load_segment_descriptor(ctxt, tss->ds, VCPU_SREG_DS);
2485  if (ret != X86EMUL_CONTINUE)
2486  return ret;
2487 
2488  return X86EMUL_CONTINUE;
2489 }
2490 
2491 static int task_switch_16(struct x86_emulate_ctxt *ctxt,
2492  u16 tss_selector, u16 old_tss_sel,
2493  ulong old_tss_base, struct desc_struct *new_desc)
2494 {
2495  const struct x86_emulate_ops *ops = ctxt->ops;
2496  struct tss_segment_16 tss_seg;
2497  int ret;
2498  u32 new_tss_base = get_desc_base(new_desc);
2499 
2500  ret = ops->read_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2501  &ctxt->exception);
2502  if (ret != X86EMUL_CONTINUE)
2503  /* FIXME: need to provide precise fault address */
2504  return ret;
2505 
2506  save_state_to_tss16(ctxt, &tss_seg);
2507 
2508  ret = ops->write_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2509  &ctxt->exception);
2510  if (ret != X86EMUL_CONTINUE)
2511  /* FIXME: need to provide precise fault address */
2512  return ret;
2513 
2514  ret = ops->read_std(ctxt, new_tss_base, &tss_seg, sizeof tss_seg,
2515  &ctxt->exception);
2516  if (ret != X86EMUL_CONTINUE)
2517  /* FIXME: need to provide precise fault address */
2518  return ret;
2519 
2520  if (old_tss_sel != 0xffff) {
2521  tss_seg.prev_task_link = old_tss_sel;
2522 
2523  ret = ops->write_std(ctxt, new_tss_base,
2524  &tss_seg.prev_task_link,
2525  sizeof tss_seg.prev_task_link,
2526  &ctxt->exception);
2527  if (ret != X86EMUL_CONTINUE)
2528  /* FIXME: need to provide precise fault address */
2529  return ret;
2530  }
2531 
2532  return load_state_from_tss16(ctxt, &tss_seg);
2533 }
2534 
2535 static void save_state_to_tss32(struct x86_emulate_ctxt *ctxt,
2536  struct tss_segment_32 *tss)
2537 {
2538  tss->cr3 = ctxt->ops->get_cr(ctxt, 3);
2539  tss->eip = ctxt->_eip;
2540  tss->eflags = ctxt->eflags;
2541  tss->eax = reg_read(ctxt, VCPU_REGS_RAX);
2542  tss->ecx = reg_read(ctxt, VCPU_REGS_RCX);
2543  tss->edx = reg_read(ctxt, VCPU_REGS_RDX);
2544  tss->ebx = reg_read(ctxt, VCPU_REGS_RBX);
2545  tss->esp = reg_read(ctxt, VCPU_REGS_RSP);
2546  tss->ebp = reg_read(ctxt, VCPU_REGS_RBP);
2547  tss->esi = reg_read(ctxt, VCPU_REGS_RSI);
2548  tss->edi = reg_read(ctxt, VCPU_REGS_RDI);
2549 
2550  tss->es = get_segment_selector(ctxt, VCPU_SREG_ES);
2551  tss->cs = get_segment_selector(ctxt, VCPU_SREG_CS);
2552  tss->ss = get_segment_selector(ctxt, VCPU_SREG_SS);
2553  tss->ds = get_segment_selector(ctxt, VCPU_SREG_DS);
2554  tss->fs = get_segment_selector(ctxt, VCPU_SREG_FS);
2555  tss->gs = get_segment_selector(ctxt, VCPU_SREG_GS);
2556  tss->ldt_selector = get_segment_selector(ctxt, VCPU_SREG_LDTR);
2557 }
2558 
2559 static int load_state_from_tss32(struct x86_emulate_ctxt *ctxt,
2560  struct tss_segment_32 *tss)
2561 {
2562  int ret;
2563 
2564  if (ctxt->ops->set_cr(ctxt, 3, tss->cr3))
2565  return emulate_gp(ctxt, 0);
2566  ctxt->_eip = tss->eip;
2567  ctxt->eflags = tss->eflags | 2;
2568 
2569  /* General purpose registers */
2570  *reg_write(ctxt, VCPU_REGS_RAX) = tss->eax;
2571  *reg_write(ctxt, VCPU_REGS_RCX) = tss->ecx;
2572  *reg_write(ctxt, VCPU_REGS_RDX) = tss->edx;
2573  *reg_write(ctxt, VCPU_REGS_RBX) = tss->ebx;
2574  *reg_write(ctxt, VCPU_REGS_RSP) = tss->esp;
2575  *reg_write(ctxt, VCPU_REGS_RBP) = tss->ebp;
2576  *reg_write(ctxt, VCPU_REGS_RSI) = tss->esi;
2577  *reg_write(ctxt, VCPU_REGS_RDI) = tss->edi;
2578 
2579  /*
2580  * SDM says that segment selectors are loaded before segment
2581  * descriptors
2582  */
2583  set_segment_selector(ctxt, tss->ldt_selector, VCPU_SREG_LDTR);
2584  set_segment_selector(ctxt, tss->es, VCPU_SREG_ES);
2585  set_segment_selector(ctxt, tss->cs, VCPU_SREG_CS);
2586  set_segment_selector(ctxt, tss->ss, VCPU_SREG_SS);
2587  set_segment_selector(ctxt, tss->ds, VCPU_SREG_DS);
2588  set_segment_selector(ctxt, tss->fs, VCPU_SREG_FS);
2589  set_segment_selector(ctxt, tss->gs, VCPU_SREG_GS);
2590 
2591  /*
2592  * If we're switching between Protected Mode and VM86, we need to make
2593  * sure to update the mode before loading the segment descriptors so
2594  * that the selectors are interpreted correctly.
2595  *
2596  * Need to get rflags to the vcpu struct immediately because it
2597  * influences the CPL which is checked at least when loading the segment
2598  * descriptors and when pushing an error code to the new kernel stack.
2599  *
2600  * TODO Introduce a separate ctxt->ops->set_cpl callback
2601  */
2602  if (ctxt->eflags & X86_EFLAGS_VM)
2603  ctxt->mode = X86EMUL_MODE_VM86;
2604  else
2605  ctxt->mode = X86EMUL_MODE_PROT32;
2606 
2607  ctxt->ops->set_rflags(ctxt, ctxt->eflags);
2608 
2609  /*
2610  * Now load segment descriptors. If fault happenes at this stage
2611  * it is handled in a context of new task
2612  */
2613  ret = load_segment_descriptor(ctxt, tss->ldt_selector, VCPU_SREG_LDTR);
2614  if (ret != X86EMUL_CONTINUE)
2615  return ret;
2616  ret = load_segment_descriptor(ctxt, tss->es, VCPU_SREG_ES);
2617  if (ret != X86EMUL_CONTINUE)
2618  return ret;
2619  ret = load_segment_descriptor(ctxt, tss->cs, VCPU_SREG_CS);
2620  if (ret != X86EMUL_CONTINUE)
2621  return ret;
2622  ret = load_segment_descriptor(ctxt, tss->ss, VCPU_SREG_SS);
2623  if (ret != X86EMUL_CONTINUE)
2624  return ret;
2625  ret = load_segment_descriptor(ctxt, tss->ds, VCPU_SREG_DS);
2626  if (ret != X86EMUL_CONTINUE)
2627  return ret;
2628  ret = load_segment_descriptor(ctxt, tss->fs, VCPU_SREG_FS);
2629  if (ret != X86EMUL_CONTINUE)
2630  return ret;
2631  ret = load_segment_descriptor(ctxt, tss->gs, VCPU_SREG_GS);
2632  if (ret != X86EMUL_CONTINUE)
2633  return ret;
2634 
2635  return X86EMUL_CONTINUE;
2636 }
2637 
2638 static int task_switch_32(struct x86_emulate_ctxt *ctxt,
2639  u16 tss_selector, u16 old_tss_sel,
2640  ulong old_tss_base, struct desc_struct *new_desc)
2641 {
2642  const struct x86_emulate_ops *ops = ctxt->ops;
2643  struct tss_segment_32 tss_seg;
2644  int ret;
2645  u32 new_tss_base = get_desc_base(new_desc);
2646 
2647  ret = ops->read_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2648  &ctxt->exception);
2649  if (ret != X86EMUL_CONTINUE)
2650  /* FIXME: need to provide precise fault address */
2651  return ret;
2652 
2653  save_state_to_tss32(ctxt, &tss_seg);
2654 
2655  ret = ops->write_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2656  &ctxt->exception);
2657  if (ret != X86EMUL_CONTINUE)
2658  /* FIXME: need to provide precise fault address */
2659  return ret;
2660 
2661  ret = ops->read_std(ctxt, new_tss_base, &tss_seg, sizeof tss_seg,
2662  &ctxt->exception);
2663  if (ret != X86EMUL_CONTINUE)
2664  /* FIXME: need to provide precise fault address */
2665  return ret;
2666 
2667  if (old_tss_sel != 0xffff) {
2668  tss_seg.prev_task_link = old_tss_sel;
2669 
2670  ret = ops->write_std(ctxt, new_tss_base,
2671  &tss_seg.prev_task_link,
2672  sizeof tss_seg.prev_task_link,
2673  &ctxt->exception);
2674  if (ret != X86EMUL_CONTINUE)
2675  /* FIXME: need to provide precise fault address */
2676  return ret;
2677  }
2678 
2679  return load_state_from_tss32(ctxt, &tss_seg);
2680 }
2681 
2682 static int emulator_do_task_switch(struct x86_emulate_ctxt *ctxt,
2683  u16 tss_selector, int idt_index, int reason,
2684  bool has_error_code, u32 error_code)
2685 {
2686  const struct x86_emulate_ops *ops = ctxt->ops;
2687  struct desc_struct curr_tss_desc, next_tss_desc;
2688  int ret;
2689  u16 old_tss_sel = get_segment_selector(ctxt, VCPU_SREG_TR);
2690  ulong old_tss_base =
2692  u32 desc_limit;
2693  ulong desc_addr;
2694 
2695  /* FIXME: old_tss_base == ~0 ? */
2696 
2697  ret = read_segment_descriptor(ctxt, tss_selector, &next_tss_desc, &desc_addr);
2698  if (ret != X86EMUL_CONTINUE)
2699  return ret;
2700  ret = read_segment_descriptor(ctxt, old_tss_sel, &curr_tss_desc, &desc_addr);
2701  if (ret != X86EMUL_CONTINUE)
2702  return ret;
2703 
2704  /* FIXME: check that next_tss_desc is tss */
2705 
2706  /*
2707  * Check privileges. The three cases are task switch caused by...
2708  *
2709  * 1. jmp/call/int to task gate: Check against DPL of the task gate
2710  * 2. Exception/IRQ/iret: No check is performed
2711  * 3. jmp/call to TSS: Check against DPL of the TSS
2712  */
2713  if (reason == TASK_SWITCH_GATE) {
2714  if (idt_index != -1) {
2715  /* Software interrupts */
2716  struct desc_struct task_gate_desc;
2717  int dpl;
2718 
2719  ret = read_interrupt_descriptor(ctxt, idt_index,
2720  &task_gate_desc);
2721  if (ret != X86EMUL_CONTINUE)
2722  return ret;
2723 
2724  dpl = task_gate_desc.dpl;
2725  if ((tss_selector & 3) > dpl || ops->cpl(ctxt) > dpl)
2726  return emulate_gp(ctxt, (idt_index << 3) | 0x2);
2727  }
2728  } else if (reason != TASK_SWITCH_IRET) {
2729  int dpl = next_tss_desc.dpl;
2730  if ((tss_selector & 3) > dpl || ops->cpl(ctxt) > dpl)
2731  return emulate_gp(ctxt, tss_selector);
2732  }
2733 
2734 
2735  desc_limit = desc_limit_scaled(&next_tss_desc);
2736  if (!next_tss_desc.p ||
2737  ((desc_limit < 0x67 && (next_tss_desc.type & 8)) ||
2738  desc_limit < 0x2b)) {
2739  emulate_ts(ctxt, tss_selector & 0xfffc);
2740  return X86EMUL_PROPAGATE_FAULT;
2741  }
2742 
2743  if (reason == TASK_SWITCH_IRET || reason == TASK_SWITCH_JMP) {
2744  curr_tss_desc.type &= ~(1 << 1); /* clear busy flag */
2745  write_segment_descriptor(ctxt, old_tss_sel, &curr_tss_desc);
2746  }
2747 
2748  if (reason == TASK_SWITCH_IRET)
2749  ctxt->eflags = ctxt->eflags & ~X86_EFLAGS_NT;
2750 
2751  /* set back link to prev task only if NT bit is set in eflags
2752  note that old_tss_sel is not used after this point */
2753  if (reason != TASK_SWITCH_CALL && reason != TASK_SWITCH_GATE)
2754  old_tss_sel = 0xffff;
2755 
2756  if (next_tss_desc.type & 8)
2757  ret = task_switch_32(ctxt, tss_selector, old_tss_sel,
2758  old_tss_base, &next_tss_desc);
2759  else
2760  ret = task_switch_16(ctxt, tss_selector, old_tss_sel,
2761  old_tss_base, &next_tss_desc);
2762  if (ret != X86EMUL_CONTINUE)
2763  return ret;
2764 
2765  if (reason == TASK_SWITCH_CALL || reason == TASK_SWITCH_GATE)
2766  ctxt->eflags = ctxt->eflags | X86_EFLAGS_NT;
2767 
2768  if (reason != TASK_SWITCH_IRET) {
2769  next_tss_desc.type |= (1 << 1); /* set busy flag */
2770  write_segment_descriptor(ctxt, tss_selector, &next_tss_desc);
2771  }
2772 
2773  ops->set_cr(ctxt, 0, ops->get_cr(ctxt, 0) | X86_CR0_TS);
2774  ops->set_segment(ctxt, tss_selector, &next_tss_desc, 0, VCPU_SREG_TR);
2775 
2776  if (has_error_code) {
2777  ctxt->op_bytes = ctxt->ad_bytes = (next_tss_desc.type & 8) ? 4 : 2;
2778  ctxt->lock_prefix = 0;
2779  ctxt->src.val = (unsigned long) error_code;
2780  ret = em_push(ctxt);
2781  }
2782 
2783  return ret;
2784 }
2785 
2787  u16 tss_selector, int idt_index, int reason,
2788  bool has_error_code, u32 error_code)
2789 {
2790  int rc;
2791 
2792  invalidate_registers(ctxt);
2793  ctxt->_eip = ctxt->eip;
2794  ctxt->dst.type = OP_NONE;
2795 
2796  rc = emulator_do_task_switch(ctxt, tss_selector, idt_index, reason,
2797  has_error_code, error_code);
2798 
2799  if (rc == X86EMUL_CONTINUE) {
2800  ctxt->eip = ctxt->_eip;
2801  writeback_registers(ctxt);
2802  }
2803 
2805 }
2806 
2807 static void string_addr_inc(struct x86_emulate_ctxt *ctxt, int reg,
2808  struct operand *op)
2809 {
2810  int df = (ctxt->eflags & EFLG_DF) ? -op->count : op->count;
2811 
2812  register_address_increment(ctxt, reg_rmw(ctxt, reg), df * op->bytes);
2813  op->addr.mem.ea = register_address(ctxt, reg_read(ctxt, reg));
2814 }
2815 
2816 static int em_das(struct x86_emulate_ctxt *ctxt)
2817 {
2818  u8 al, old_al;
2819  bool af, cf, old_cf;
2820 
2821  cf = ctxt->eflags & X86_EFLAGS_CF;
2822  al = ctxt->dst.val;
2823 
2824  old_al = al;
2825  old_cf = cf;
2826  cf = false;
2827  af = ctxt->eflags & X86_EFLAGS_AF;
2828  if ((al & 0x0f) > 9 || af) {
2829  al -= 6;
2830  cf = old_cf | (al >= 250);
2831  af = true;
2832  } else {
2833  af = false;
2834  }
2835  if (old_al > 0x99 || old_cf) {
2836  al -= 0x60;
2837  cf = true;
2838  }
2839 
2840  ctxt->dst.val = al;
2841  /* Set PF, ZF, SF */
2842  ctxt->src.type = OP_IMM;
2843  ctxt->src.val = 0;
2844  ctxt->src.bytes = 1;
2845  emulate_2op_SrcV(ctxt, "or");
2846  ctxt->eflags &= ~(X86_EFLAGS_AF | X86_EFLAGS_CF);
2847  if (cf)
2848  ctxt->eflags |= X86_EFLAGS_CF;
2849  if (af)
2850  ctxt->eflags |= X86_EFLAGS_AF;
2851  return X86EMUL_CONTINUE;
2852 }
2853 
2854 static int em_call(struct x86_emulate_ctxt *ctxt)
2855 {
2856  long rel = ctxt->src.val;
2857 
2858  ctxt->src.val = (unsigned long)ctxt->_eip;
2859  jmp_rel(ctxt, rel);
2860  return em_push(ctxt);
2861 }
2862 
2863 static int em_call_far(struct x86_emulate_ctxt *ctxt)
2864 {
2865  u16 sel, old_cs;
2866  ulong old_eip;
2867  int rc;
2868 
2869  old_cs = get_segment_selector(ctxt, VCPU_SREG_CS);
2870  old_eip = ctxt->_eip;
2871 
2872  memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
2873  if (load_segment_descriptor(ctxt, sel, VCPU_SREG_CS))
2874  return X86EMUL_CONTINUE;
2875 
2876  ctxt->_eip = 0;
2877  memcpy(&ctxt->_eip, ctxt->src.valptr, ctxt->op_bytes);
2878 
2879  ctxt->src.val = old_cs;
2880  rc = em_push(ctxt);
2881  if (rc != X86EMUL_CONTINUE)
2882  return rc;
2883 
2884  ctxt->src.val = old_eip;
2885  return em_push(ctxt);
2886 }
2887 
2888 static int em_ret_near_imm(struct x86_emulate_ctxt *ctxt)
2889 {
2890  int rc;
2891 
2892  ctxt->dst.type = OP_REG;
2893  ctxt->dst.addr.reg = &ctxt->_eip;
2894  ctxt->dst.bytes = ctxt->op_bytes;
2895  rc = emulate_pop(ctxt, &ctxt->dst.val, ctxt->op_bytes);
2896  if (rc != X86EMUL_CONTINUE)
2897  return rc;
2898  rsp_increment(ctxt, ctxt->src.val);
2899  return X86EMUL_CONTINUE;
2900 }
2901 
2902 static int em_add(struct x86_emulate_ctxt *ctxt)
2903 {
2904  emulate_2op_SrcV(ctxt, "add");
2905  return X86EMUL_CONTINUE;
2906 }
2907 
2908 static int em_or(struct x86_emulate_ctxt *ctxt)
2909 {
2910  emulate_2op_SrcV(ctxt, "or");
2911  return X86EMUL_CONTINUE;
2912 }
2913 
2914 static int em_adc(struct x86_emulate_ctxt *ctxt)
2915 {
2916  emulate_2op_SrcV(ctxt, "adc");
2917  return X86EMUL_CONTINUE;
2918 }
2919 
2920 static int em_sbb(struct x86_emulate_ctxt *ctxt)
2921 {
2922  emulate_2op_SrcV(ctxt, "sbb");
2923  return X86EMUL_CONTINUE;
2924 }
2925 
2926 static int em_and(struct x86_emulate_ctxt *ctxt)
2927 {
2928  emulate_2op_SrcV(ctxt, "and");
2929  return X86EMUL_CONTINUE;
2930 }
2931 
2932 static int em_sub(struct x86_emulate_ctxt *ctxt)
2933 {
2934  emulate_2op_SrcV(ctxt, "sub");
2935  return X86EMUL_CONTINUE;
2936 }
2937 
2938 static int em_xor(struct x86_emulate_ctxt *ctxt)
2939 {
2940  emulate_2op_SrcV(ctxt, "xor");
2941  return X86EMUL_CONTINUE;
2942 }
2943 
2944 static int em_cmp(struct x86_emulate_ctxt *ctxt)
2945 {
2946  emulate_2op_SrcV(ctxt, "cmp");
2947  /* Disable writeback. */
2948  ctxt->dst.type = OP_NONE;
2949  return X86EMUL_CONTINUE;
2950 }
2951 
2952 static int em_test(struct x86_emulate_ctxt *ctxt)
2953 {
2954  emulate_2op_SrcV(ctxt, "test");
2955  /* Disable writeback. */
2956  ctxt->dst.type = OP_NONE;
2957  return X86EMUL_CONTINUE;
2958 }
2959 
2960 static int em_xchg(struct x86_emulate_ctxt *ctxt)
2961 {
2962  /* Write back the register source. */
2963  ctxt->src.val = ctxt->dst.val;
2964  write_register_operand(&ctxt->src);
2965 
2966  /* Write back the memory destination with implicit LOCK prefix. */
2967  ctxt->dst.val = ctxt->src.orig_val;
2968  ctxt->lock_prefix = 1;
2969  return X86EMUL_CONTINUE;
2970 }
2971 
2972 static int em_imul(struct x86_emulate_ctxt *ctxt)
2973 {
2974  emulate_2op_SrcV_nobyte(ctxt, "imul");
2975  return X86EMUL_CONTINUE;
2976 }
2977 
2978 static int em_imul_3op(struct x86_emulate_ctxt *ctxt)
2979 {
2980  ctxt->dst.val = ctxt->src2.val;
2981  return em_imul(ctxt);
2982 }
2983 
2984 static int em_cwd(struct x86_emulate_ctxt *ctxt)
2985 {
2986  ctxt->dst.type = OP_REG;
2987  ctxt->dst.bytes = ctxt->src.bytes;
2988  ctxt->dst.addr.reg = reg_rmw(ctxt, VCPU_REGS_RDX);
2989  ctxt->dst.val = ~((ctxt->src.val >> (ctxt->src.bytes * 8 - 1)) - 1);
2990 
2991  return X86EMUL_CONTINUE;
2992 }
2993 
2994 static int em_rdtsc(struct x86_emulate_ctxt *ctxt)
2995 {
2996  u64 tsc = 0;
2997 
2998  ctxt->ops->get_msr(ctxt, MSR_IA32_TSC, &tsc);
2999  *reg_write(ctxt, VCPU_REGS_RAX) = (u32)tsc;
3000  *reg_write(ctxt, VCPU_REGS_RDX) = tsc >> 32;
3001  return X86EMUL_CONTINUE;
3002 }
3003 
3004 static int em_rdpmc(struct x86_emulate_ctxt *ctxt)
3005 {
3006  u64 pmc;
3007 
3008  if (ctxt->ops->read_pmc(ctxt, reg_read(ctxt, VCPU_REGS_RCX), &pmc))
3009  return emulate_gp(ctxt, 0);
3010  *reg_write(ctxt, VCPU_REGS_RAX) = (u32)pmc;
3011  *reg_write(ctxt, VCPU_REGS_RDX) = pmc >> 32;
3012  return X86EMUL_CONTINUE;
3013 }
3014 
3015 static int em_mov(struct x86_emulate_ctxt *ctxt)
3016 {
3017  memcpy(ctxt->dst.valptr, ctxt->src.valptr, ctxt->op_bytes);
3018  return X86EMUL_CONTINUE;
3019 }
3020 
3021 static int em_cr_write(struct x86_emulate_ctxt *ctxt)
3022 {
3023  if (ctxt->ops->set_cr(ctxt, ctxt->modrm_reg, ctxt->src.val))
3024  return emulate_gp(ctxt, 0);
3025 
3026  /* Disable writeback. */
3027  ctxt->dst.type = OP_NONE;
3028  return X86EMUL_CONTINUE;
3029 }
3030 
3031 static int em_dr_write(struct x86_emulate_ctxt *ctxt)
3032 {
3033  unsigned long val;
3034 
3035  if (ctxt->mode == X86EMUL_MODE_PROT64)
3036  val = ctxt->src.val & ~0ULL;
3037  else
3038  val = ctxt->src.val & ~0U;
3039 
3040  /* #UD condition is already handled. */
3041  if (ctxt->ops->set_dr(ctxt, ctxt->modrm_reg, val) < 0)
3042  return emulate_gp(ctxt, 0);
3043 
3044  /* Disable writeback. */
3045  ctxt->dst.type = OP_NONE;
3046  return X86EMUL_CONTINUE;
3047 }
3048 
3049 static int em_wrmsr(struct x86_emulate_ctxt *ctxt)
3050 {
3051  u64 msr_data;
3052 
3053  msr_data = (u32)reg_read(ctxt, VCPU_REGS_RAX)
3054  | ((u64)reg_read(ctxt, VCPU_REGS_RDX) << 32);
3055  if (ctxt->ops->set_msr(ctxt, reg_read(ctxt, VCPU_REGS_RCX), msr_data))
3056  return emulate_gp(ctxt, 0);
3057 
3058  return X86EMUL_CONTINUE;
3059 }
3060 
3061 static int em_rdmsr(struct x86_emulate_ctxt *ctxt)
3062 {
3063  u64 msr_data;
3064 
3065  if (ctxt->ops->get_msr(ctxt, reg_read(ctxt, VCPU_REGS_RCX), &msr_data))
3066  return emulate_gp(ctxt, 0);
3067 
3068  *reg_write(ctxt, VCPU_REGS_RAX) = (u32)msr_data;
3069  *reg_write(ctxt, VCPU_REGS_RDX) = msr_data >> 32;
3070  return X86EMUL_CONTINUE;
3071 }
3072 
3073 static int em_mov_rm_sreg(struct x86_emulate_ctxt *ctxt)
3074 {
3075  if (ctxt->modrm_reg > VCPU_SREG_GS)
3076  return emulate_ud(ctxt);
3077 
3078  ctxt->dst.val = get_segment_selector(ctxt, ctxt->modrm_reg);
3079  return X86EMUL_CONTINUE;
3080 }
3081 
3082 static int em_mov_sreg_rm(struct x86_emulate_ctxt *ctxt)
3083 {
3084  u16 sel = ctxt->src.val;
3085 
3086  if (ctxt->modrm_reg == VCPU_SREG_CS || ctxt->modrm_reg > VCPU_SREG_GS)
3087  return emulate_ud(ctxt);
3088 
3089  if (ctxt->modrm_reg == VCPU_SREG_SS)
3091 
3092  /* Disable writeback. */
3093  ctxt->dst.type = OP_NONE;
3094  return load_segment_descriptor(ctxt, sel, ctxt->modrm_reg);
3095 }
3096 
3097 static int em_lldt(struct x86_emulate_ctxt *ctxt)
3098 {
3099  u16 sel = ctxt->src.val;
3100 
3101  /* Disable writeback. */
3102  ctxt->dst.type = OP_NONE;
3103  return load_segment_descriptor(ctxt, sel, VCPU_SREG_LDTR);
3104 }
3105 
3106 static int em_ltr(struct x86_emulate_ctxt *ctxt)
3107 {
3108  u16 sel = ctxt->src.val;
3109 
3110  /* Disable writeback. */
3111  ctxt->dst.type = OP_NONE;
3112  return load_segment_descriptor(ctxt, sel, VCPU_SREG_TR);
3113 }
3114 
3115 static int em_invlpg(struct x86_emulate_ctxt *ctxt)
3116 {
3117  int rc;
3118  ulong linear;
3119 
3120  rc = linearize(ctxt, ctxt->src.addr.mem, 1, false, &linear);
3121  if (rc == X86EMUL_CONTINUE)
3122  ctxt->ops->invlpg(ctxt, linear);
3123  /* Disable writeback. */
3124  ctxt->dst.type = OP_NONE;
3125  return X86EMUL_CONTINUE;
3126 }
3127 
3128 static int em_clts(struct x86_emulate_ctxt *ctxt)
3129 {
3130  ulong cr0;
3131 
3132  cr0 = ctxt->ops->get_cr(ctxt, 0);
3133  cr0 &= ~X86_CR0_TS;
3134  ctxt->ops->set_cr(ctxt, 0, cr0);
3135  return X86EMUL_CONTINUE;
3136 }
3137 
3138 static int em_vmcall(struct x86_emulate_ctxt *ctxt)
3139 {
3140  int rc;
3141 
3142  if (ctxt->modrm_mod != 3 || ctxt->modrm_rm != 1)
3143  return X86EMUL_UNHANDLEABLE;
3144 
3145  rc = ctxt->ops->fix_hypercall(ctxt);
3146  if (rc != X86EMUL_CONTINUE)
3147  return rc;
3148 
3149  /* Let the processor re-execute the fixed hypercall */
3150  ctxt->_eip = ctxt->eip;
3151  /* Disable writeback. */
3152  ctxt->dst.type = OP_NONE;
3153  return X86EMUL_CONTINUE;
3154 }
3155 
3156 static int emulate_store_desc_ptr(struct x86_emulate_ctxt *ctxt,
3157  void (*get)(struct x86_emulate_ctxt *ctxt,
3158  struct desc_ptr *ptr))
3159 {
3160  struct desc_ptr desc_ptr;
3161 
3162  if (ctxt->mode == X86EMUL_MODE_PROT64)
3163  ctxt->op_bytes = 8;
3164  get(ctxt, &desc_ptr);
3165  if (ctxt->op_bytes == 2) {
3166  ctxt->op_bytes = 4;
3167  desc_ptr.address &= 0x00ffffff;
3168  }
3169  /* Disable writeback. */
3170  ctxt->dst.type = OP_NONE;
3171  return segmented_write(ctxt, ctxt->dst.addr.mem,
3172  &desc_ptr, 2 + ctxt->op_bytes);
3173 }
3174 
3175 static int em_sgdt(struct x86_emulate_ctxt *ctxt)
3176 {
3177  return emulate_store_desc_ptr(ctxt, ctxt->ops->get_gdt);
3178 }
3179 
3180 static int em_sidt(struct x86_emulate_ctxt *ctxt)
3181 {
3182  return emulate_store_desc_ptr(ctxt, ctxt->ops->get_idt);
3183 }
3184 
3185 static int em_lgdt(struct x86_emulate_ctxt *ctxt)
3186 {
3187  struct desc_ptr desc_ptr;
3188  int rc;
3189 
3190  if (ctxt->mode == X86EMUL_MODE_PROT64)
3191  ctxt->op_bytes = 8;
3192  rc = read_descriptor(ctxt, ctxt->src.addr.mem,
3194  ctxt->op_bytes);
3195  if (rc != X86EMUL_CONTINUE)
3196  return rc;
3197  ctxt->ops->set_gdt(ctxt, &desc_ptr);
3198  /* Disable writeback. */
3199  ctxt->dst.type = OP_NONE;
3200  return X86EMUL_CONTINUE;
3201 }
3202 
3203 static int em_vmmcall(struct x86_emulate_ctxt *ctxt)
3204 {
3205  int rc;
3206 
3207  rc = ctxt->ops->fix_hypercall(ctxt);
3208 
3209  /* Disable writeback. */
3210  ctxt->dst.type = OP_NONE;
3211  return rc;
3212 }
3213 
3214 static int em_lidt(struct x86_emulate_ctxt *ctxt)
3215 {
3216  struct desc_ptr desc_ptr;
3217  int rc;
3218 
3219  if (ctxt->mode == X86EMUL_MODE_PROT64)
3220  ctxt->op_bytes = 8;
3221  rc = read_descriptor(ctxt, ctxt->src.addr.mem,
3223  ctxt->op_bytes);
3224  if (rc != X86EMUL_CONTINUE)
3225  return rc;
3226  ctxt->ops->set_idt(ctxt, &desc_ptr);
3227  /* Disable writeback. */
3228  ctxt->dst.type = OP_NONE;
3229  return X86EMUL_CONTINUE;
3230 }
3231 
3232 static int em_smsw(struct x86_emulate_ctxt *ctxt)
3233 {
3234  ctxt->dst.bytes = 2;
3235  ctxt->dst.val = ctxt->ops->get_cr(ctxt, 0);
3236  return X86EMUL_CONTINUE;
3237 }
3238 
3239 static int em_lmsw(struct x86_emulate_ctxt *ctxt)
3240 {
3241  ctxt->ops->set_cr(ctxt, 0, (ctxt->ops->get_cr(ctxt, 0) & ~0x0eul)
3242  | (ctxt->src.val & 0x0f));
3243  ctxt->dst.type = OP_NONE;
3244  return X86EMUL_CONTINUE;
3245 }
3246 
3247 static int em_loop(struct x86_emulate_ctxt *ctxt)
3248 {
3249  register_address_increment(ctxt, reg_rmw(ctxt, VCPU_REGS_RCX), -1);
3250  if ((address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) != 0) &&
3251  (ctxt->b == 0xe2 || test_cc(ctxt->b ^ 0x5, ctxt->eflags)))
3252  jmp_rel(ctxt, ctxt->src.val);
3253 
3254  return X86EMUL_CONTINUE;
3255 }
3256 
3257 static int em_jcxz(struct x86_emulate_ctxt *ctxt)
3258 {
3259  if (address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) == 0)
3260  jmp_rel(ctxt, ctxt->src.val);
3261 
3262  return X86EMUL_CONTINUE;
3263 }
3264 
3265 static int em_in(struct x86_emulate_ctxt *ctxt)
3266 {
3267  if (!pio_in_emulated(ctxt, ctxt->dst.bytes, ctxt->src.val,
3268  &ctxt->dst.val))
3269  return X86EMUL_IO_NEEDED;
3270 
3271  return X86EMUL_CONTINUE;
3272 }
3273 
3274 static int em_out(struct x86_emulate_ctxt *ctxt)
3275 {
3276  ctxt->ops->pio_out_emulated(ctxt, ctxt->src.bytes, ctxt->dst.val,
3277  &ctxt->src.val, 1);
3278  /* Disable writeback. */
3279  ctxt->dst.type = OP_NONE;
3280  return X86EMUL_CONTINUE;
3281 }
3282 
3283 static int em_cli(struct x86_emulate_ctxt *ctxt)
3284 {
3285  if (emulator_bad_iopl(ctxt))
3286  return emulate_gp(ctxt, 0);
3287 
3288  ctxt->eflags &= ~X86_EFLAGS_IF;
3289  return X86EMUL_CONTINUE;
3290 }
3291 
3292 static int em_sti(struct x86_emulate_ctxt *ctxt)
3293 {
3294  if (emulator_bad_iopl(ctxt))
3295  return emulate_gp(ctxt, 0);
3296 
3298  ctxt->eflags |= X86_EFLAGS_IF;
3299  return X86EMUL_CONTINUE;
3300 }
3301 
3302 static int em_bt(struct x86_emulate_ctxt *ctxt)
3303 {
3304  /* Disable writeback. */
3305  ctxt->dst.type = OP_NONE;
3306  /* only subword offset */
3307  ctxt->src.val &= (ctxt->dst.bytes << 3) - 1;
3308 
3309  emulate_2op_SrcV_nobyte(ctxt, "bt");
3310  return X86EMUL_CONTINUE;
3311 }
3312 
3313 static int em_bts(struct x86_emulate_ctxt *ctxt)
3314 {
3315  emulate_2op_SrcV_nobyte(ctxt, "bts");
3316  return X86EMUL_CONTINUE;
3317 }
3318 
3319 static int em_btr(struct x86_emulate_ctxt *ctxt)
3320 {
3321  emulate_2op_SrcV_nobyte(ctxt, "btr");
3322  return X86EMUL_CONTINUE;
3323 }
3324 
3325 static int em_btc(struct x86_emulate_ctxt *ctxt)
3326 {
3327  emulate_2op_SrcV_nobyte(ctxt, "btc");
3328  return X86EMUL_CONTINUE;
3329 }
3330 
3331 static int em_bsf(struct x86_emulate_ctxt *ctxt)
3332 {
3333  emulate_2op_SrcV_nobyte(ctxt, "bsf");
3334  return X86EMUL_CONTINUE;
3335 }
3336 
3337 static int em_bsr(struct x86_emulate_ctxt *ctxt)
3338 {
3339  emulate_2op_SrcV_nobyte(ctxt, "bsr");
3340  return X86EMUL_CONTINUE;
3341 }
3342 
3343 static int em_cpuid(struct x86_emulate_ctxt *ctxt)
3344 {
3345  u32 eax, ebx, ecx, edx;
3346 
3347  eax = reg_read(ctxt, VCPU_REGS_RAX);
3348  ecx = reg_read(ctxt, VCPU_REGS_RCX);
3349  ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
3350  *reg_write(ctxt, VCPU_REGS_RAX) = eax;
3351  *reg_write(ctxt, VCPU_REGS_RBX) = ebx;
3352  *reg_write(ctxt, VCPU_REGS_RCX) = ecx;
3353  *reg_write(ctxt, VCPU_REGS_RDX) = edx;
3354  return X86EMUL_CONTINUE;
3355 }
3356 
3357 static int em_lahf(struct x86_emulate_ctxt *ctxt)
3358 {
3359  *reg_rmw(ctxt, VCPU_REGS_RAX) &= ~0xff00UL;
3360  *reg_rmw(ctxt, VCPU_REGS_RAX) |= (ctxt->eflags & 0xff) << 8;
3361  return X86EMUL_CONTINUE;
3362 }
3363 
3364 static int em_bswap(struct x86_emulate_ctxt *ctxt)
3365 {
3366  switch (ctxt->op_bytes) {
3367 #ifdef CONFIG_X86_64
3368  case 8:
3369  asm("bswap %0" : "+r"(ctxt->dst.val));
3370  break;
3371 #endif
3372  default:
3373  asm("bswap %0" : "+r"(*(u32 *)&ctxt->dst.val));
3374  break;
3375  }
3376  return X86EMUL_CONTINUE;
3377 }
3378 
3379 static bool valid_cr(int nr)
3380 {
3381  switch (nr) {
3382  case 0:
3383  case 2 ... 4:
3384  case 8:
3385  return true;
3386  default:
3387  return false;
3388  }
3389 }
3390 
3391 static int check_cr_read(struct x86_emulate_ctxt *ctxt)
3392 {
3393  if (!valid_cr(ctxt->modrm_reg))
3394  return emulate_ud(ctxt);
3395 
3396  return X86EMUL_CONTINUE;
3397 }
3398 
3399 static int check_cr_write(struct x86_emulate_ctxt *ctxt)
3400 {
3401  u64 new_val = ctxt->src.val64;
3402  int cr = ctxt->modrm_reg;
3403  u64 efer = 0;
3404 
3405  static u64 cr_reserved_bits[] = {
3406  0xffffffff00000000ULL,
3407  0, 0, 0, /* CR3 checked later */
3409  0, 0, 0,
3411  };
3412 
3413  if (!valid_cr(cr))
3414  return emulate_ud(ctxt);
3415 
3416  if (new_val & cr_reserved_bits[cr])
3417  return emulate_gp(ctxt, 0);
3418 
3419  switch (cr) {
3420  case 0: {
3421  u64 cr4;
3422  if (((new_val & X86_CR0_PG) && !(new_val & X86_CR0_PE)) ||
3423  ((new_val & X86_CR0_NW) && !(new_val & X86_CR0_CD)))
3424  return emulate_gp(ctxt, 0);
3425 
3426  cr4 = ctxt->ops->get_cr(ctxt, 4);
3427  ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3428 
3429  if ((new_val & X86_CR0_PG) && (efer & EFER_LME) &&
3430  !(cr4 & X86_CR4_PAE))
3431  return emulate_gp(ctxt, 0);
3432 
3433  break;
3434  }
3435  case 3: {
3436  u64 rsvd = 0;
3437 
3438  ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3439  if (efer & EFER_LMA)
3440  rsvd = CR3_L_MODE_RESERVED_BITS;
3441  else if (ctxt->ops->get_cr(ctxt, 4) & X86_CR4_PAE)
3442  rsvd = CR3_PAE_RESERVED_BITS;
3443  else if (ctxt->ops->get_cr(ctxt, 0) & X86_CR0_PG)
3444  rsvd = CR3_NONPAE_RESERVED_BITS;
3445 
3446  if (new_val & rsvd)
3447  return emulate_gp(ctxt, 0);
3448 
3449  break;
3450  }
3451  case 4: {
3452  ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3453 
3454  if ((efer & EFER_LMA) && !(new_val & X86_CR4_PAE))
3455  return emulate_gp(ctxt, 0);
3456 
3457  break;
3458  }
3459  }
3460 
3461  return X86EMUL_CONTINUE;
3462 }
3463 
3464 static int check_dr7_gd(struct x86_emulate_ctxt *ctxt)
3465 {
3466  unsigned long dr7;
3467 
3468  ctxt->ops->get_dr(ctxt, 7, &dr7);
3469 
3470  /* Check if DR7.Global_Enable is set */
3471  return dr7 & (1 << 13);
3472 }
3473 
3474 static int check_dr_read(struct x86_emulate_ctxt *ctxt)
3475 {
3476  int dr = ctxt->modrm_reg;
3477  u64 cr4;
3478 
3479  if (dr > 7)
3480  return emulate_ud(ctxt);
3481 
3482  cr4 = ctxt->ops->get_cr(ctxt, 4);
3483  if ((cr4 & X86_CR4_DE) && (dr == 4 || dr == 5))
3484  return emulate_ud(ctxt);
3485 
3486  if (check_dr7_gd(ctxt))
3487  return emulate_db(ctxt);
3488 
3489  return X86EMUL_CONTINUE;
3490 }
3491 
3492 static int check_dr_write(struct x86_emulate_ctxt *ctxt)
3493 {
3494  u64 new_val = ctxt->src.val64;
3495  int dr = ctxt->modrm_reg;
3496 
3497  if ((dr == 6 || dr == 7) && (new_val & 0xffffffff00000000ULL))
3498  return emulate_gp(ctxt, 0);
3499 
3500  return check_dr_read(ctxt);
3501 }
3502 
3503 static int check_svme(struct x86_emulate_ctxt *ctxt)
3504 {
3505  u64 efer;
3506 
3507  ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3508 
3509  if (!(efer & EFER_SVME))
3510  return emulate_ud(ctxt);
3511 
3512  return X86EMUL_CONTINUE;
3513 }
3514 
3515 static int check_svme_pa(struct x86_emulate_ctxt *ctxt)
3516 {
3517  u64 rax = reg_read(ctxt, VCPU_REGS_RAX);
3518 
3519  /* Valid physical address? */
3520  if (rax & 0xffff000000000000ULL)
3521  return emulate_gp(ctxt, 0);
3522 
3523  return check_svme(ctxt);
3524 }
3525 
3526 static int check_rdtsc(struct x86_emulate_ctxt *ctxt)
3527 {
3528  u64 cr4 = ctxt->ops->get_cr(ctxt, 4);
3529 
3530  if (cr4 & X86_CR4_TSD && ctxt->ops->cpl(ctxt))
3531  return emulate_ud(ctxt);
3532 
3533  return X86EMUL_CONTINUE;
3534 }
3535 
3536 static int check_rdpmc(struct x86_emulate_ctxt *ctxt)
3537 {
3538  u64 cr4 = ctxt->ops->get_cr(ctxt, 4);
3539  u64 rcx = reg_read(ctxt, VCPU_REGS_RCX);
3540 
3541  if ((!(cr4 & X86_CR4_PCE) && ctxt->ops->cpl(ctxt)) ||
3542  (rcx > 3))
3543  return emulate_gp(ctxt, 0);
3544 
3545  return X86EMUL_CONTINUE;
3546 }
3547 
3548 static int check_perm_in(struct x86_emulate_ctxt *ctxt)
3549 {
3550  ctxt->dst.bytes = min(ctxt->dst.bytes, 4u);
3551  if (!emulator_io_permited(ctxt, ctxt->src.val, ctxt->dst.bytes))
3552  return emulate_gp(ctxt, 0);
3553 
3554  return X86EMUL_CONTINUE;
3555 }
3556 
3557 static int check_perm_out(struct x86_emulate_ctxt *ctxt)
3558 {
3559  ctxt->src.bytes = min(ctxt->src.bytes, 4u);
3560  if (!emulator_io_permited(ctxt, ctxt->dst.val, ctxt->src.bytes))
3561  return emulate_gp(ctxt, 0);
3562 
3563  return X86EMUL_CONTINUE;
3564 }
3565 
3566 #define D(_y) { .flags = (_y) }
3567 #define DI(_y, _i) { .flags = (_y), .intercept = x86_intercept_##_i }
3568 #define DIP(_y, _i, _p) { .flags = (_y), .intercept = x86_intercept_##_i, \
3569  .check_perm = (_p) }
3570 #define N D(0)
3571 #define EXT(_f, _e) { .flags = ((_f) | RMExt), .u.group = (_e) }
3572 #define G(_f, _g) { .flags = ((_f) | Group | ModRM), .u.group = (_g) }
3573 #define GD(_f, _g) { .flags = ((_f) | GroupDual | ModRM), .u.gdual = (_g) }
3574 #define I(_f, _e) { .flags = (_f), .u.execute = (_e) }
3575 #define II(_f, _e, _i) \
3576  { .flags = (_f), .u.execute = (_e), .intercept = x86_intercept_##_i }
3577 #define IIP(_f, _e, _i, _p) \
3578  { .flags = (_f), .u.execute = (_e), .intercept = x86_intercept_##_i, \
3579  .check_perm = (_p) }
3580 #define GP(_f, _g) { .flags = ((_f) | Prefix), .u.gprefix = (_g) }
3581 
3582 #define D2bv(_f) D((_f) | ByteOp), D(_f)
3583 #define D2bvIP(_f, _i, _p) DIP((_f) | ByteOp, _i, _p), DIP(_f, _i, _p)
3584 #define I2bv(_f, _e) I((_f) | ByteOp, _e), I(_f, _e)
3585 #define I2bvIP(_f, _e, _i, _p) \
3586  IIP((_f) | ByteOp, _e, _i, _p), IIP(_f, _e, _i, _p)
3587 
3588 #define I6ALU(_f, _e) I2bv((_f) | DstMem | SrcReg | ModRM, _e), \
3589  I2bv(((_f) | DstReg | SrcMem | ModRM) & ~Lock, _e), \
3590  I2bv(((_f) & ~Lock) | DstAcc | SrcImm, _e)
3591 
3592 static const struct opcode group7_rm1[] = {
3593  DI(SrcNone | Priv, monitor),
3594  DI(SrcNone | Priv, mwait),
3595  N, N, N, N, N, N,
3596 };
3597 
3598 static const struct opcode group7_rm3[] = {
3599  DIP(SrcNone | Prot | Priv, vmrun, check_svme_pa),
3600  II(SrcNone | Prot | VendorSpecific, em_vmmcall, vmmcall),
3601  DIP(SrcNone | Prot | Priv, vmload, check_svme_pa),
3602  DIP(SrcNone | Prot | Priv, vmsave, check_svme_pa),
3603  DIP(SrcNone | Prot | Priv, stgi, check_svme),
3604  DIP(SrcNone | Prot | Priv, clgi, check_svme),
3605  DIP(SrcNone | Prot | Priv, skinit, check_svme),
3606  DIP(SrcNone | Prot | Priv, invlpga, check_svme),
3607 };
3608 
3609 static const struct opcode group7_rm7[] = {
3610  N,
3611  DIP(SrcNone, rdtscp, check_rdtsc),
3612  N, N, N, N, N, N,
3613 };
3614 
3615 static const struct opcode group1[] = {
3616  I(Lock, em_add),
3617  I(Lock | PageTable, em_or),
3618  I(Lock, em_adc),
3619  I(Lock, em_sbb),
3620  I(Lock | PageTable, em_and),
3621  I(Lock, em_sub),
3622  I(Lock, em_xor),
3623  I(0, em_cmp),
3624 };
3625 
3626 static const struct opcode group1A[] = {
3627  I(DstMem | SrcNone | Mov | Stack, em_pop), N, N, N, N, N, N, N,
3628 };
3629 
3630 static const struct opcode group3[] = {
3631  I(DstMem | SrcImm, em_test),
3632  I(DstMem | SrcImm, em_test),
3633  I(DstMem | SrcNone | Lock, em_not),
3634  I(DstMem | SrcNone | Lock, em_neg),
3635  I(SrcMem, em_mul_ex),
3636  I(SrcMem, em_imul_ex),
3637  I(SrcMem, em_div_ex),
3638  I(SrcMem, em_idiv_ex),
3639 };
3640 
3641 static const struct opcode group4[] = {
3642  I(ByteOp | DstMem | SrcNone | Lock, em_grp45),
3643  I(ByteOp | DstMem | SrcNone | Lock, em_grp45),
3644  N, N, N, N, N, N,
3645 };
3646 
3647 static const struct opcode group5[] = {
3648  I(DstMem | SrcNone | Lock, em_grp45),
3649  I(DstMem | SrcNone | Lock, em_grp45),
3650  I(SrcMem | Stack, em_grp45),
3651  I(SrcMemFAddr | ImplicitOps | Stack, em_call_far),
3652  I(SrcMem | Stack, em_grp45),
3653  I(SrcMemFAddr | ImplicitOps, em_grp45),
3654  I(SrcMem | Stack, em_grp45), N,
3655 };
3656 
3657 static const struct opcode group6[] = {
3658  DI(Prot, sldt),
3659  DI(Prot, str),
3660  II(Prot | Priv | SrcMem16, em_lldt, lldt),
3661  II(Prot | Priv | SrcMem16, em_ltr, ltr),
3662  N, N, N, N,
3663 };
3664 
3665 static const struct group_dual group7 = { {
3666  II(Mov | DstMem | Priv, em_sgdt, sgdt),
3667  II(Mov | DstMem | Priv, em_sidt, sidt),
3668  II(SrcMem | Priv, em_lgdt, lgdt),
3669  II(SrcMem | Priv, em_lidt, lidt),
3670  II(SrcNone | DstMem | Mov, em_smsw, smsw), N,
3671  II(SrcMem16 | Mov | Priv, em_lmsw, lmsw),
3672  II(SrcMem | ByteOp | Priv | NoAccess, em_invlpg, invlpg),
3673 }, {
3674  I(SrcNone | Priv | VendorSpecific, em_vmcall),
3675  EXT(0, group7_rm1),
3676  N, EXT(0, group7_rm3),
3677  II(SrcNone | DstMem | Mov, em_smsw, smsw), N,
3678  II(SrcMem16 | Mov | Priv, em_lmsw, lmsw),
3679  EXT(0, group7_rm7),
3680 } };
3681 
3682 static const struct opcode group8[] = {
3683  N, N, N, N,
3684  I(DstMem | SrcImmByte, em_bt),
3685  I(DstMem | SrcImmByte | Lock | PageTable, em_bts),
3686  I(DstMem | SrcImmByte | Lock, em_btr),
3687  I(DstMem | SrcImmByte | Lock | PageTable, em_btc),
3688 };
3689 
3690 static const struct group_dual group9 = { {
3691  N, I(DstMem64 | Lock | PageTable, em_cmpxchg8b), N, N, N, N, N, N,
3692 }, {
3693  N, N, N, N, N, N, N, N,
3694 } };
3695 
3696 static const struct opcode group11[] = {
3697  I(DstMem | SrcImm | Mov | PageTable, em_mov),
3698  X7(D(Undefined)),
3699 };
3700 
3701 static const struct gprefix pfx_0f_6f_0f_7f = {
3702  I(Mmx, em_mov), I(Sse | Aligned, em_mov), N, I(Sse | Unaligned, em_mov),
3703 };
3704 
3705 static const struct gprefix pfx_vmovntpx = {
3706  I(0, em_mov), N, N, N,
3707 };
3708 
3709 static const struct opcode opcode_table[256] = {
3710  /* 0x00 - 0x07 */
3711  I6ALU(Lock, em_add),
3712  I(ImplicitOps | Stack | No64 | Src2ES, em_push_sreg),
3713  I(ImplicitOps | Stack | No64 | Src2ES, em_pop_sreg),
3714  /* 0x08 - 0x0F */
3715  I6ALU(Lock | PageTable, em_or),
3716  I(ImplicitOps | Stack | No64 | Src2CS, em_push_sreg),
3717  N,
3718  /* 0x10 - 0x17 */
3719  I6ALU(Lock, em_adc),
3720  I(ImplicitOps | Stack | No64 | Src2SS, em_push_sreg),
3721  I(ImplicitOps | Stack | No64 | Src2SS, em_pop_sreg),
3722  /* 0x18 - 0x1F */
3723  I6ALU(Lock, em_sbb),
3724  I(ImplicitOps | Stack | No64 | Src2DS, em_push_sreg),
3725  I(ImplicitOps | Stack | No64 | Src2DS, em_pop_sreg),
3726  /* 0x20 - 0x27 */
3727  I6ALU(Lock | PageTable, em_and), N, N,
3728  /* 0x28 - 0x2F */
3729  I6ALU(Lock, em_sub), N, I(ByteOp | DstAcc | No64, em_das),
3730  /* 0x30 - 0x37 */
3731  I6ALU(Lock, em_xor), N, N,
3732  /* 0x38 - 0x3F */
3733  I6ALU(0, em_cmp), N, N,
3734  /* 0x40 - 0x4F */
3735  X16(D(DstReg)),
3736  /* 0x50 - 0x57 */
3737  X8(I(SrcReg | Stack, em_push)),
3738  /* 0x58 - 0x5F */
3739  X8(I(DstReg | Stack, em_pop)),
3740  /* 0x60 - 0x67 */
3741  I(ImplicitOps | Stack | No64, em_pusha),
3742  I(ImplicitOps | Stack | No64, em_popa),
3743  N, D(DstReg | SrcMem32 | ModRM | Mov) /* movsxd (x86/64) */ ,
3744  N, N, N, N,
3745  /* 0x68 - 0x6F */
3746  I(SrcImm | Mov | Stack, em_push),
3747  I(DstReg | SrcMem | ModRM | Src2Imm, em_imul_3op),
3748  I(SrcImmByte | Mov | Stack, em_push),
3749  I(DstReg | SrcMem | ModRM | Src2ImmByte, em_imul_3op),
3750  I2bvIP(DstDI | SrcDX | Mov | String | Unaligned, em_in, ins, check_perm_in), /* insb, insw/insd */
3751  I2bvIP(SrcSI | DstDX | String, em_out, outs, check_perm_out), /* outsb, outsw/outsd */
3752  /* 0x70 - 0x7F */
3753  X16(D(SrcImmByte)),
3754  /* 0x80 - 0x87 */
3755  G(ByteOp | DstMem | SrcImm, group1),
3756  G(DstMem | SrcImm, group1),
3757  G(ByteOp | DstMem | SrcImm | No64, group1),
3758  G(DstMem | SrcImmByte, group1),
3759  I2bv(DstMem | SrcReg | ModRM, em_test),
3760  I2bv(DstMem | SrcReg | ModRM | Lock | PageTable, em_xchg),
3761  /* 0x88 - 0x8F */
3762  I2bv(DstMem | SrcReg | ModRM | Mov | PageTable, em_mov),
3763  I2bv(DstReg | SrcMem | ModRM | Mov, em_mov),
3764  I(DstMem | SrcNone | ModRM | Mov | PageTable, em_mov_rm_sreg),
3765  D(ModRM | SrcMem | NoAccess | DstReg),
3766  I(ImplicitOps | SrcMem16 | ModRM, em_mov_sreg_rm),
3767  G(0, group1A),
3768  /* 0x90 - 0x97 */
3769  DI(SrcAcc | DstReg, pause), X7(D(SrcAcc | DstReg)),
3770  /* 0x98 - 0x9F */
3771  D(DstAcc | SrcNone), I(ImplicitOps | SrcAcc, em_cwd),
3772  I(SrcImmFAddr | No64, em_call_far), N,
3773  II(ImplicitOps | Stack, em_pushf, pushf),
3774  II(ImplicitOps | Stack, em_popf, popf), N, I(ImplicitOps, em_lahf),
3775  /* 0xA0 - 0xA7 */
3776  I2bv(DstAcc | SrcMem | Mov | MemAbs, em_mov),
3777  I2bv(DstMem | SrcAcc | Mov | MemAbs | PageTable, em_mov),
3778  I2bv(SrcSI | DstDI | Mov | String, em_mov),
3779  I2bv(SrcSI | DstDI | String, em_cmp),
3780  /* 0xA8 - 0xAF */
3781  I2bv(DstAcc | SrcImm, em_test),
3782  I2bv(SrcAcc | DstDI | Mov | String, em_mov),
3783  I2bv(SrcSI | DstAcc | Mov | String, em_mov),
3784  I2bv(SrcAcc | DstDI | String, em_cmp),
3785  /* 0xB0 - 0xB7 */
3786  X8(I(ByteOp | DstReg | SrcImm | Mov, em_mov)),
3787  /* 0xB8 - 0xBF */
3788  X8(I(DstReg | SrcImm | Mov, em_mov)),
3789  /* 0xC0 - 0xC7 */
3790  D2bv(DstMem | SrcImmByte | ModRM),
3791  I(ImplicitOps | Stack | SrcImmU16, em_ret_near_imm),
3792  I(ImplicitOps | Stack, em_ret),
3793  I(DstReg | SrcMemFAddr | ModRM | No64 | Src2ES, em_lseg),
3794  I(DstReg | SrcMemFAddr | ModRM | No64 | Src2DS, em_lseg),
3795  G(ByteOp, group11), G(0, group11),
3796  /* 0xC8 - 0xCF */
3797  I(Stack | SrcImmU16 | Src2ImmByte, em_enter), I(Stack, em_leave),
3798  N, I(ImplicitOps | Stack, em_ret_far),
3799  D(ImplicitOps), DI(SrcImmByte, intn),
3800  D(ImplicitOps | No64), II(ImplicitOps, em_iret, iret),
3801  /* 0xD0 - 0xD7 */
3802  D2bv(DstMem | SrcOne | ModRM), D2bv(DstMem | ModRM),
3803  N, N, N, N,
3804  /* 0xD8 - 0xDF */
3805  N, N, N, N, N, N, N, N,
3806  /* 0xE0 - 0xE7 */
3807  X3(I(SrcImmByte, em_loop)),
3808  I(SrcImmByte, em_jcxz),
3809  I2bvIP(SrcImmUByte | DstAcc, em_in, in, check_perm_in),
3810  I2bvIP(SrcAcc | DstImmUByte, em_out, out, check_perm_out),
3811  /* 0xE8 - 0xEF */
3812  I(SrcImm | Stack, em_call), D(SrcImm | ImplicitOps),
3813  I(SrcImmFAddr | No64, em_jmp_far), D(SrcImmByte | ImplicitOps),
3814  I2bvIP(SrcDX | DstAcc, em_in, in, check_perm_in),
3815  I2bvIP(SrcAcc | DstDX, em_out, out, check_perm_out),
3816  /* 0xF0 - 0xF7 */
3817  N, DI(ImplicitOps, icebp), N, N,
3818  DI(ImplicitOps | Priv, hlt), D(ImplicitOps),
3819  G(ByteOp, group3), G(0, group3),
3820  /* 0xF8 - 0xFF */
3822  I(ImplicitOps, em_cli), I(ImplicitOps, em_sti),
3823  D(ImplicitOps), D(ImplicitOps), G(0, group4), G(0, group5),
3824 };
3825 
3826 static const struct opcode twobyte_table[256] = {
3827  /* 0x00 - 0x0F */
3828  G(0, group6), GD(0, &group7), N, N,
3829  N, I(ImplicitOps | VendorSpecific, em_syscall),
3830  II(ImplicitOps | Priv, em_clts, clts), N,
3831  DI(ImplicitOps | Priv, invd), DI(ImplicitOps | Priv, wbinvd), N, N,
3832  N, D(ImplicitOps | ModRM), N, N,
3833  /* 0x10 - 0x1F */
3834  N, N, N, N, N, N, N, N, D(ImplicitOps | ModRM), N, N, N, N, N, N, N,
3835  /* 0x20 - 0x2F */
3836  DIP(ModRM | DstMem | Priv | Op3264, cr_read, check_cr_read),
3837  DIP(ModRM | DstMem | Priv | Op3264, dr_read, check_dr_read),
3838  IIP(ModRM | SrcMem | Priv | Op3264, em_cr_write, cr_write, check_cr_write),
3839  IIP(ModRM | SrcMem | Priv | Op3264, em_dr_write, dr_write, check_dr_write),
3840  N, N, N, N,
3841  N, N, N, GP(ModRM | DstMem | SrcReg | Sse | Mov | Aligned, &pfx_vmovntpx),
3842  N, N, N, N,
3843  /* 0x30 - 0x3F */
3844  II(ImplicitOps | Priv, em_wrmsr, wrmsr),
3845  IIP(ImplicitOps, em_rdtsc, rdtsc, check_rdtsc),
3846  II(ImplicitOps | Priv, em_rdmsr, rdmsr),
3847  IIP(ImplicitOps, em_rdpmc, rdpmc, check_rdpmc),
3848  I(ImplicitOps | VendorSpecific, em_sysenter),
3849  I(ImplicitOps | Priv | VendorSpecific, em_sysexit),
3850  N, N,
3851  N, N, N, N, N, N, N, N,
3852  /* 0x40 - 0x4F */
3853  X16(D(DstReg | SrcMem | ModRM | Mov)),
3854  /* 0x50 - 0x5F */
3855  N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
3856  /* 0x60 - 0x6F */
3857  N, N, N, N,
3858  N, N, N, N,
3859  N, N, N, N,
3860  N, N, N, GP(SrcMem | DstReg | ModRM | Mov, &pfx_0f_6f_0f_7f),
3861  /* 0x70 - 0x7F */
3862  N, N, N, N,
3863  N, N, N, N,
3864  N, N, N, N,
3865  N, N, N, GP(SrcReg | DstMem | ModRM | Mov, &pfx_0f_6f_0f_7f),
3866  /* 0x80 - 0x8F */
3867  X16(D(SrcImm)),
3868  /* 0x90 - 0x9F */
3869  X16(D(ByteOp | DstMem | SrcNone | ModRM| Mov)),
3870  /* 0xA0 - 0xA7 */
3871  I(Stack | Src2FS, em_push_sreg), I(Stack | Src2FS, em_pop_sreg),
3872  II(ImplicitOps, em_cpuid, cpuid), I(DstMem | SrcReg | ModRM | BitOp, em_bt),
3873  D(DstMem | SrcReg | Src2ImmByte | ModRM),
3874  D(DstMem | SrcReg | Src2CL | ModRM), N, N,
3875  /* 0xA8 - 0xAF */
3876  I(Stack | Src2GS, em_push_sreg), I(Stack | Src2GS, em_pop_sreg),
3877  DI(ImplicitOps, rsm),
3878  I(DstMem | SrcReg | ModRM | BitOp | Lock | PageTable, em_bts),
3879  D(DstMem | SrcReg | Src2ImmByte | ModRM),
3880  D(DstMem | SrcReg | Src2CL | ModRM),
3881  D(ModRM), I(DstReg | SrcMem | ModRM, em_imul),
3882  /* 0xB0 - 0xB7 */
3883  I2bv(DstMem | SrcReg | ModRM | Lock | PageTable, em_cmpxchg),
3884  I(DstReg | SrcMemFAddr | ModRM | Src2SS, em_lseg),
3885  I(DstMem | SrcReg | ModRM | BitOp | Lock, em_btr),
3886  I(DstReg | SrcMemFAddr | ModRM | Src2FS, em_lseg),
3887  I(DstReg | SrcMemFAddr | ModRM | Src2GS, em_lseg),
3888  D(DstReg | SrcMem8 | ModRM | Mov), D(DstReg | SrcMem16 | ModRM | Mov),
3889  /* 0xB8 - 0xBF */
3890  N, N,
3891  G(BitOp, group8),
3892  I(DstMem | SrcReg | ModRM | BitOp | Lock | PageTable, em_btc),
3893  I(DstReg | SrcMem | ModRM, em_bsf), I(DstReg | SrcMem | ModRM, em_bsr),
3894  D(DstReg | SrcMem8 | ModRM | Mov), D(DstReg | SrcMem16 | ModRM | Mov),
3895  /* 0xC0 - 0xC7 */
3896  D2bv(DstMem | SrcReg | ModRM | Lock),
3897  N, D(DstMem | SrcReg | ModRM | Mov),
3898  N, N, N, GD(0, &group9),
3899  /* 0xC8 - 0xCF */
3900  X8(I(DstReg, em_bswap)),
3901  /* 0xD0 - 0xDF */
3902  N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
3903  /* 0xE0 - 0xEF */
3904  N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
3905  /* 0xF0 - 0xFF */
3906  N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N
3907 };
3908 
3909 #undef D
3910 #undef N
3911 #undef G
3912 #undef GD
3913 #undef I
3914 #undef GP
3915 #undef EXT
3916 
3917 #undef D2bv
3918 #undef D2bvIP
3919 #undef I2bv
3920 #undef I2bvIP
3921 #undef I6ALU
3922 
3923 static unsigned imm_size(struct x86_emulate_ctxt *ctxt)
3924 {
3925  unsigned size;
3926 
3927  size = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
3928  if (size == 8)
3929  size = 4;
3930  return size;
3931 }
3932 
3933 static int decode_imm(struct x86_emulate_ctxt *ctxt, struct operand *op,
3934  unsigned size, bool sign_extension)
3935 {
3936  int rc = X86EMUL_CONTINUE;
3937 
3938  op->type = OP_IMM;
3939  op->bytes = size;
3940  op->addr.mem.ea = ctxt->_eip;
3941  /* NB. Immediates are sign-extended as necessary. */
3942  switch (op->bytes) {
3943  case 1:
3944  op->val = insn_fetch(s8, ctxt);
3945  break;
3946  case 2:
3947  op->val = insn_fetch(s16, ctxt);
3948  break;
3949  case 4:
3950  op->val = insn_fetch(s32, ctxt);
3951  break;
3952  }
3953  if (!sign_extension) {
3954  switch (op->bytes) {
3955  case 1:
3956  op->val &= 0xff;
3957  break;
3958  case 2:
3959  op->val &= 0xffff;
3960  break;
3961  case 4:
3962  op->val &= 0xffffffff;
3963  break;
3964  }
3965  }
3966 done:
3967  return rc;
3968 }
3969 
3970 static int decode_operand(struct x86_emulate_ctxt *ctxt, struct operand *op,
3971  unsigned d)
3972 {
3973  int rc = X86EMUL_CONTINUE;
3974 
3975  switch (d) {
3976  case OpReg:
3977  decode_register_operand(ctxt, op);
3978  break;
3979  case OpImmUByte:
3980  rc = decode_imm(ctxt, op, 1, false);
3981  break;
3982  case OpMem:
3983  ctxt->memop.bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
3984  mem_common:
3985  *op = ctxt->memop;
3986  ctxt->memopp = op;
3987  if ((ctxt->d & BitOp) && op == &ctxt->dst)
3988  fetch_bit_operand(ctxt);
3989  op->orig_val = op->val;
3990  break;
3991  case OpMem64:
3992  ctxt->memop.bytes = 8;
3993  goto mem_common;
3994  case OpAcc:
3995  op->type = OP_REG;
3996  op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
3997  op->addr.reg = reg_rmw(ctxt, VCPU_REGS_RAX);
3998  fetch_register_operand(op);
3999  op->orig_val = op->val;
4000  break;
4001  case OpDI:
4002  op->type = OP_MEM;
4003  op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4004  op->addr.mem.ea =
4005  register_address(ctxt, reg_read(ctxt, VCPU_REGS_RDI));
4006  op->addr.mem.seg = VCPU_SREG_ES;
4007  op->val = 0;
4008  op->count = 1;
4009  break;
4010  case OpDX:
4011  op->type = OP_REG;
4012  op->bytes = 2;
4013  op->addr.reg = reg_rmw(ctxt, VCPU_REGS_RDX);
4014  fetch_register_operand(op);
4015  break;
4016  case OpCL:
4017  op->bytes = 1;
4018  op->val = reg_read(ctxt, VCPU_REGS_RCX) & 0xff;
4019  break;
4020  case OpImmByte:
4021  rc = decode_imm(ctxt, op, 1, true);
4022  break;
4023  case OpOne:
4024  op->bytes = 1;
4025  op->val = 1;
4026  break;
4027  case OpImm:
4028  rc = decode_imm(ctxt, op, imm_size(ctxt), true);
4029  break;
4030  case OpMem8:
4031  ctxt->memop.bytes = 1;
4032  goto mem_common;
4033  case OpMem16:
4034  ctxt->memop.bytes = 2;
4035  goto mem_common;
4036  case OpMem32:
4037  ctxt->memop.bytes = 4;
4038  goto mem_common;
4039  case OpImmU16:
4040  rc = decode_imm(ctxt, op, 2, false);
4041  break;
4042  case OpImmU:
4043  rc = decode_imm(ctxt, op, imm_size(ctxt), false);
4044  break;
4045  case OpSI:
4046  op->type = OP_MEM;
4047  op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4048  op->addr.mem.ea =
4049  register_address(ctxt, reg_read(ctxt, VCPU_REGS_RSI));
4050  op->addr.mem.seg = seg_override(ctxt);
4051  op->val = 0;
4052  op->count = 1;
4053  break;
4054  case OpImmFAddr:
4055  op->type = OP_IMM;
4056  op->addr.mem.ea = ctxt->_eip;
4057  op->bytes = ctxt->op_bytes + 2;
4058  insn_fetch_arr(op->valptr, op->bytes, ctxt);
4059  break;
4060  case OpMemFAddr:
4061  ctxt->memop.bytes = ctxt->op_bytes + 2;
4062  goto mem_common;
4063  case OpES:
4064  op->val = VCPU_SREG_ES;
4065  break;
4066  case OpCS:
4067  op->val = VCPU_SREG_CS;
4068  break;
4069  case OpSS:
4070  op->val = VCPU_SREG_SS;
4071  break;
4072  case OpDS:
4073  op->val = VCPU_SREG_DS;
4074  break;
4075  case OpFS:
4076  op->val = VCPU_SREG_FS;
4077  break;
4078  case OpGS:
4079  op->val = VCPU_SREG_GS;
4080  break;
4081  case OpImplicit:
4082  /* Special instructions do their own operand decoding. */
4083  default:
4084  op->type = OP_NONE; /* Disable writeback. */
4085  break;
4086  }
4087 
4088 done:
4089  return rc;
4090 }
4091 
4092 int x86_decode_insn(struct x86_emulate_ctxt *ctxt, void *insn, int insn_len)
4093 {
4094  int rc = X86EMUL_CONTINUE;
4095  int mode = ctxt->mode;
4096  int def_op_bytes, def_ad_bytes, goffset, simd_prefix;
4097  bool op_prefix = false;
4098  struct opcode opcode;
4099 
4100  ctxt->memop.type = OP_NONE;
4101  ctxt->memopp = NULL;
4102  ctxt->_eip = ctxt->eip;
4103  ctxt->fetch.start = ctxt->_eip;
4104  ctxt->fetch.end = ctxt->fetch.start + insn_len;
4105  if (insn_len > 0)
4106  memcpy(ctxt->fetch.data, insn, insn_len);
4107 
4108  switch (mode) {
4109  case X86EMUL_MODE_REAL:
4110  case X86EMUL_MODE_VM86:
4111  case X86EMUL_MODE_PROT16:
4112  def_op_bytes = def_ad_bytes = 2;
4113  break;
4114  case X86EMUL_MODE_PROT32:
4115  def_op_bytes = def_ad_bytes = 4;
4116  break;
4117 #ifdef CONFIG_X86_64
4118  case X86EMUL_MODE_PROT64:
4119  def_op_bytes = 4;
4120  def_ad_bytes = 8;
4121  break;
4122 #endif
4123  default:
4124  return EMULATION_FAILED;
4125  }
4126 
4127  ctxt->op_bytes = def_op_bytes;
4128  ctxt->ad_bytes = def_ad_bytes;
4129 
4130  /* Legacy prefixes. */
4131  for (;;) {
4132  switch (ctxt->b = insn_fetch(u8, ctxt)) {
4133  case 0x66: /* operand-size override */
4134  op_prefix = true;
4135  /* switch between 2/4 bytes */
4136  ctxt->op_bytes = def_op_bytes ^ 6;
4137  break;
4138  case 0x67: /* address-size override */
4139  if (mode == X86EMUL_MODE_PROT64)
4140  /* switch between 4/8 bytes */
4141  ctxt->ad_bytes = def_ad_bytes ^ 12;
4142  else
4143  /* switch between 2/4 bytes */
4144  ctxt->ad_bytes = def_ad_bytes ^ 6;
4145  break;
4146  case 0x26: /* ES override */
4147  case 0x2e: /* CS override */
4148  case 0x36: /* SS override */
4149  case 0x3e: /* DS override */
4150  set_seg_override(ctxt, (ctxt->b >> 3) & 3);
4151  break;
4152  case 0x64: /* FS override */
4153  case 0x65: /* GS override */
4154  set_seg_override(ctxt, ctxt->b & 7);
4155  break;
4156  case 0x40 ... 0x4f: /* REX */
4157  if (mode != X86EMUL_MODE_PROT64)
4158  goto done_prefixes;
4159  ctxt->rex_prefix = ctxt->b;
4160  continue;
4161  case 0xf0: /* LOCK */
4162  ctxt->lock_prefix = 1;
4163  break;
4164  case 0xf2: /* REPNE/REPNZ */
4165  case 0xf3: /* REP/REPE/REPZ */
4166  ctxt->rep_prefix = ctxt->b;
4167  break;
4168  default:
4169  goto done_prefixes;
4170  }
4171 
4172  /* Any legacy prefix after a REX prefix nullifies its effect. */
4173 
4174  ctxt->rex_prefix = 0;
4175  }
4176 
4177 done_prefixes:
4178 
4179  /* REX prefix. */
4180  if (ctxt->rex_prefix & 8)
4181  ctxt->op_bytes = 8; /* REX.W */
4182 
4183  /* Opcode byte(s). */
4184  opcode = opcode_table[ctxt->b];
4185  /* Two-byte opcode? */
4186  if (ctxt->b == 0x0f) {
4187  ctxt->twobyte = 1;
4188  ctxt->b = insn_fetch(u8, ctxt);
4189  opcode = twobyte_table[ctxt->b];
4190  }
4191  ctxt->d = opcode.flags;
4192 
4193  if (ctxt->d & ModRM)
4194  ctxt->modrm = insn_fetch(u8, ctxt);
4195 
4196  while (ctxt->d & GroupMask) {
4197  switch (ctxt->d & GroupMask) {
4198  case Group:
4199  goffset = (ctxt->modrm >> 3) & 7;
4200  opcode = opcode.u.group[goffset];
4201  break;
4202  case GroupDual:
4203  goffset = (ctxt->modrm >> 3) & 7;
4204  if ((ctxt->modrm >> 6) == 3)
4205  opcode = opcode.u.gdual->mod3[goffset];
4206  else
4207  opcode = opcode.u.gdual->mod012[goffset];
4208  break;
4209  case RMExt:
4210  goffset = ctxt->modrm & 7;
4211  opcode = opcode.u.group[goffset];
4212  break;
4213  case Prefix:
4214  if (ctxt->rep_prefix && op_prefix)
4215  return EMULATION_FAILED;
4216  simd_prefix = op_prefix ? 0x66 : ctxt->rep_prefix;
4217  switch (simd_prefix) {
4218  case 0x00: opcode = opcode.u.gprefix->pfx_no; break;
4219  case 0x66: opcode = opcode.u.gprefix->pfx_66; break;
4220  case 0xf2: opcode = opcode.u.gprefix->pfx_f2; break;
4221  case 0xf3: opcode = opcode.u.gprefix->pfx_f3; break;
4222  }
4223  break;
4224  default:
4225  return EMULATION_FAILED;
4226  }
4227 
4228  ctxt->d &= ~(u64)GroupMask;
4229  ctxt->d |= opcode.flags;
4230  }
4231 
4232  ctxt->execute = opcode.u.execute;
4233  ctxt->check_perm = opcode.check_perm;
4234  ctxt->intercept = opcode.intercept;
4235 
4236  /* Unrecognised? */
4237  if (ctxt->d == 0 || (ctxt->d & Undefined))
4238  return EMULATION_FAILED;
4239 
4240  if (!(ctxt->d & VendorSpecific) && ctxt->only_vendor_specific_insn)
4241  return EMULATION_FAILED;
4242 
4243  if (mode == X86EMUL_MODE_PROT64 && (ctxt->d & Stack))
4244  ctxt->op_bytes = 8;
4245 
4246  if (ctxt->d & Op3264) {
4247  if (mode == X86EMUL_MODE_PROT64)
4248  ctxt->op_bytes = 8;
4249  else
4250  ctxt->op_bytes = 4;
4251  }
4252 
4253  if (ctxt->d & Sse)
4254  ctxt->op_bytes = 16;
4255  else if (ctxt->d & Mmx)
4256  ctxt->op_bytes = 8;
4257 
4258  /* ModRM and SIB bytes. */
4259  if (ctxt->d & ModRM) {
4260  rc = decode_modrm(ctxt, &ctxt->memop);
4261  if (!ctxt->has_seg_override)
4262  set_seg_override(ctxt, ctxt->modrm_seg);
4263  } else if (ctxt->d & MemAbs)
4264  rc = decode_abs(ctxt, &ctxt->memop);
4265  if (rc != X86EMUL_CONTINUE)
4266  goto done;
4267 
4268  if (!ctxt->has_seg_override)
4269  set_seg_override(ctxt, VCPU_SREG_DS);
4270 
4271  ctxt->memop.addr.mem.seg = seg_override(ctxt);
4272 
4273  if (ctxt->memop.type == OP_MEM && ctxt->ad_bytes != 8)
4274  ctxt->memop.addr.mem.ea = (u32)ctxt->memop.addr.mem.ea;
4275 
4276  /*
4277  * Decode and fetch the source operand: register, memory
4278  * or immediate.
4279  */
4280  rc = decode_operand(ctxt, &ctxt->src, (ctxt->d >> SrcShift) & OpMask);
4281  if (rc != X86EMUL_CONTINUE)
4282  goto done;
4283 
4284  /*
4285  * Decode and fetch the second source operand: register, memory
4286  * or immediate.
4287  */
4288  rc = decode_operand(ctxt, &ctxt->src2, (ctxt->d >> Src2Shift) & OpMask);
4289  if (rc != X86EMUL_CONTINUE)
4290  goto done;
4291 
4292  /* Decode and fetch the destination operand: register or memory. */
4293  rc = decode_operand(ctxt, &ctxt->dst, (ctxt->d >> DstShift) & OpMask);
4294 
4295 done:
4296  if (ctxt->memopp && ctxt->memopp->type == OP_MEM && ctxt->rip_relative)
4297  ctxt->memopp->addr.mem.ea += ctxt->_eip;
4298 
4299  return (rc != X86EMUL_CONTINUE) ? EMULATION_FAILED : EMULATION_OK;
4300 }
4301 
4303 {
4304  return ctxt->d & PageTable;
4305 }
4306 
4307 static bool string_insn_completed(struct x86_emulate_ctxt *ctxt)
4308 {
4309  /* The second termination condition only applies for REPE
4310  * and REPNE. Test if the repeat string operation prefix is
4311  * REPE/REPZ or REPNE/REPNZ and if it's the case it tests the
4312  * corresponding termination condition according to:
4313  * - if REPE/REPZ and ZF = 0 then done
4314  * - if REPNE/REPNZ and ZF = 1 then done
4315  */
4316  if (((ctxt->b == 0xa6) || (ctxt->b == 0xa7) ||
4317  (ctxt->b == 0xae) || (ctxt->b == 0xaf))
4318  && (((ctxt->rep_prefix == REPE_PREFIX) &&
4319  ((ctxt->eflags & EFLG_ZF) == 0))
4320  || ((ctxt->rep_prefix == REPNE_PREFIX) &&
4321  ((ctxt->eflags & EFLG_ZF) == EFLG_ZF))))
4322  return true;
4323 
4324  return false;
4325 }
4326 
4327 static int flush_pending_x87_faults(struct x86_emulate_ctxt *ctxt)
4328 {
4329  bool fault = false;
4330 
4331  ctxt->ops->get_fpu(ctxt);
4332  asm volatile("1: fwait \n\t"
4333  "2: \n\t"
4334  ".pushsection .fixup,\"ax\" \n\t"
4335  "3: \n\t"
4336  "movb $1, %[fault] \n\t"
4337  "jmp 2b \n\t"
4338  ".popsection \n\t"
4339  _ASM_EXTABLE(1b, 3b)
4340  : [fault]"+qm"(fault));
4341  ctxt->ops->put_fpu(ctxt);
4342 
4343  if (unlikely(fault))
4344  return emulate_exception(ctxt, MF_VECTOR, 0, false);
4345 
4346  return X86EMUL_CONTINUE;
4347 }
4348 
4349 static void fetch_possible_mmx_operand(struct x86_emulate_ctxt *ctxt,
4350  struct operand *op)
4351 {
4352  if (op->type == OP_MM)
4353  read_mmx_reg(ctxt, &op->mm_val, op->addr.mm);
4354 }
4355 
4356 
4358 {
4359  const struct x86_emulate_ops *ops = ctxt->ops;
4360  int rc = X86EMUL_CONTINUE;
4361  int saved_dst_type = ctxt->dst.type;
4362 
4363  ctxt->mem_read.pos = 0;
4364 
4365  if (ctxt->mode == X86EMUL_MODE_PROT64 && (ctxt->d & No64)) {
4366  rc = emulate_ud(ctxt);
4367  goto done;
4368  }
4369 
4370  /* LOCK prefix is allowed only with some instructions */
4371  if (ctxt->lock_prefix && (!(ctxt->d & Lock) || ctxt->dst.type != OP_MEM)) {
4372  rc = emulate_ud(ctxt);
4373  goto done;
4374  }
4375 
4376  if ((ctxt->d & SrcMask) == SrcMemFAddr && ctxt->src.type != OP_MEM) {
4377  rc = emulate_ud(ctxt);
4378  goto done;
4379  }
4380 
4381  if (((ctxt->d & (Sse|Mmx)) && ((ops->get_cr(ctxt, 0) & X86_CR0_EM)))
4382  || ((ctxt->d & Sse) && !(ops->get_cr(ctxt, 4) & X86_CR4_OSFXSR))) {
4383  rc = emulate_ud(ctxt);
4384  goto done;
4385  }
4386 
4387  if ((ctxt->d & (Sse|Mmx)) && (ops->get_cr(ctxt, 0) & X86_CR0_TS)) {
4388  rc = emulate_nm(ctxt);
4389  goto done;
4390  }
4391 
4392  if (ctxt->d & Mmx) {
4393  rc = flush_pending_x87_faults(ctxt);
4394  if (rc != X86EMUL_CONTINUE)
4395  goto done;
4396  /*
4397  * Now that we know the fpu is exception safe, we can fetch
4398  * operands from it.
4399  */
4400  fetch_possible_mmx_operand(ctxt, &ctxt->src);
4401  fetch_possible_mmx_operand(ctxt, &ctxt->src2);
4402  if (!(ctxt->d & Mov))
4403  fetch_possible_mmx_operand(ctxt, &ctxt->dst);
4404  }
4405 
4406  if (unlikely(ctxt->guest_mode) && ctxt->intercept) {
4407  rc = emulator_check_intercept(ctxt, ctxt->intercept,
4409  if (rc != X86EMUL_CONTINUE)
4410  goto done;
4411  }
4412 
4413  /* Privileged instruction can be executed only in CPL=0 */
4414  if ((ctxt->d & Priv) && ops->cpl(ctxt)) {
4415  rc = emulate_gp(ctxt, 0);
4416  goto done;
4417  }
4418 
4419  /* Instruction can only be executed in protected mode */
4420  if ((ctxt->d & Prot) && ctxt->mode < X86EMUL_MODE_PROT16) {
4421  rc = emulate_ud(ctxt);
4422  goto done;
4423  }
4424 
4425  /* Do instruction specific permission checks */
4426  if (ctxt->check_perm) {
4427  rc = ctxt->check_perm(ctxt);
4428  if (rc != X86EMUL_CONTINUE)
4429  goto done;
4430  }
4431 
4432  if (unlikely(ctxt->guest_mode) && ctxt->intercept) {
4433  rc = emulator_check_intercept(ctxt, ctxt->intercept,
4435  if (rc != X86EMUL_CONTINUE)
4436  goto done;
4437  }
4438 
4439  if (ctxt->rep_prefix && (ctxt->d & String)) {
4440  /* All REP prefixes have the same first termination condition */
4441  if (address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) == 0) {
4442  ctxt->eip = ctxt->_eip;
4443  goto done;
4444  }
4445  }
4446 
4447  if ((ctxt->src.type == OP_MEM) && !(ctxt->d & NoAccess)) {
4448  rc = segmented_read(ctxt, ctxt->src.addr.mem,
4449  ctxt->src.valptr, ctxt->src.bytes);
4450  if (rc != X86EMUL_CONTINUE)
4451  goto done;
4452  ctxt->src.orig_val64 = ctxt->src.val64;
4453  }
4454 
4455  if (ctxt->src2.type == OP_MEM) {
4456  rc = segmented_read(ctxt, ctxt->src2.addr.mem,
4457  &ctxt->src2.val, ctxt->src2.bytes);
4458  if (rc != X86EMUL_CONTINUE)
4459  goto done;
4460  }
4461 
4462  if ((ctxt->d & DstMask) == ImplicitOps)
4463  goto special_insn;
4464 
4465 
4466  if ((ctxt->dst.type == OP_MEM) && !(ctxt->d & Mov)) {
4467  /* optimisation - avoid slow emulated read if Mov */
4468  rc = segmented_read(ctxt, ctxt->dst.addr.mem,
4469  &ctxt->dst.val, ctxt->dst.bytes);
4470  if (rc != X86EMUL_CONTINUE)
4471  goto done;
4472  }
4473  ctxt->dst.orig_val = ctxt->dst.val;
4474 
4475 special_insn:
4476 
4477  if (unlikely(ctxt->guest_mode) && ctxt->intercept) {
4478  rc = emulator_check_intercept(ctxt, ctxt->intercept,
4480  if (rc != X86EMUL_CONTINUE)
4481  goto done;
4482  }
4483 
4484  if (ctxt->execute) {
4485  rc = ctxt->execute(ctxt);
4486  if (rc != X86EMUL_CONTINUE)
4487  goto done;
4488  goto writeback;
4489  }
4490 
4491  if (ctxt->twobyte)
4492  goto twobyte_insn;
4493 
4494  switch (ctxt->b) {
4495  case 0x40 ... 0x47: /* inc r16/r32 */
4496  emulate_1op(ctxt, "inc");
4497  break;
4498  case 0x48 ... 0x4f: /* dec r16/r32 */
4499  emulate_1op(ctxt, "dec");
4500  break;
4501  case 0x63: /* movsxd */
4502  if (ctxt->mode != X86EMUL_MODE_PROT64)
4503  goto cannot_emulate;
4504  ctxt->dst.val = (s32) ctxt->src.val;
4505  break;
4506  case 0x70 ... 0x7f: /* jcc (short) */
4507  if (test_cc(ctxt->b, ctxt->eflags))
4508  jmp_rel(ctxt, ctxt->src.val);
4509  break;
4510  case 0x8d: /* lea r16/r32, m */
4511  ctxt->dst.val = ctxt->src.addr.mem.ea;
4512  break;
4513  case 0x90 ... 0x97: /* nop / xchg reg, rax */
4514  if (ctxt->dst.addr.reg == reg_rmw(ctxt, VCPU_REGS_RAX))
4515  break;
4516  rc = em_xchg(ctxt);
4517  break;
4518  case 0x98: /* cbw/cwde/cdqe */
4519  switch (ctxt->op_bytes) {
4520  case 2: ctxt->dst.val = (s8)ctxt->dst.val; break;
4521  case 4: ctxt->dst.val = (s16)ctxt->dst.val; break;
4522  case 8: ctxt->dst.val = (s32)ctxt->dst.val; break;
4523  }
4524  break;
4525  case 0xc0 ... 0xc1:
4526  rc = em_grp2(ctxt);
4527  break;
4528  case 0xcc: /* int3 */
4529  rc = emulate_int(ctxt, 3);
4530  break;
4531  case 0xcd: /* int n */
4532  rc = emulate_int(ctxt, ctxt->src.val);
4533  break;
4534  case 0xce: /* into */
4535  if (ctxt->eflags & EFLG_OF)
4536  rc = emulate_int(ctxt, 4);
4537  break;
4538  case 0xd0 ... 0xd1: /* Grp2 */
4539  rc = em_grp2(ctxt);
4540  break;
4541  case 0xd2 ... 0xd3: /* Grp2 */
4542  ctxt->src.val = reg_read(ctxt, VCPU_REGS_RCX);
4543  rc = em_grp2(ctxt);
4544  break;
4545  case 0xe9: /* jmp rel */
4546  case 0xeb: /* jmp rel short */
4547  jmp_rel(ctxt, ctxt->src.val);
4548  ctxt->dst.type = OP_NONE; /* Disable writeback. */
4549  break;
4550  case 0xf4: /* hlt */
4551  ctxt->ops->halt(ctxt);
4552  break;
4553  case 0xf5: /* cmc */
4554  /* complement carry flag from eflags reg */
4555  ctxt->eflags ^= EFLG_CF;
4556  break;
4557  case 0xf8: /* clc */
4558  ctxt->eflags &= ~EFLG_CF;
4559  break;
4560  case 0xf9: /* stc */
4561  ctxt->eflags |= EFLG_CF;
4562  break;
4563  case 0xfc: /* cld */
4564  ctxt->eflags &= ~EFLG_DF;
4565  break;
4566  case 0xfd: /* std */
4567  ctxt->eflags |= EFLG_DF;
4568  break;
4569  default:
4570  goto cannot_emulate;
4571  }
4572 
4573  if (rc != X86EMUL_CONTINUE)
4574  goto done;
4575 
4576 writeback:
4577  rc = writeback(ctxt);
4578  if (rc != X86EMUL_CONTINUE)
4579  goto done;
4580 
4581  /*
4582  * restore dst type in case the decoding will be reused
4583  * (happens for string instruction )
4584  */
4585  ctxt->dst.type = saved_dst_type;
4586 
4587  if ((ctxt->d & SrcMask) == SrcSI)
4588  string_addr_inc(ctxt, VCPU_REGS_RSI, &ctxt->src);
4589 
4590  if ((ctxt->d & DstMask) == DstDI)
4591  string_addr_inc(ctxt, VCPU_REGS_RDI, &ctxt->dst);
4592 
4593  if (ctxt->rep_prefix && (ctxt->d & String)) {
4594  unsigned int count;
4595  struct read_cache *r = &ctxt->io_read;
4596  if ((ctxt->d & SrcMask) == SrcSI)
4597  count = ctxt->src.count;
4598  else
4599  count = ctxt->dst.count;
4600  register_address_increment(ctxt, reg_rmw(ctxt, VCPU_REGS_RCX),
4601  -count);
4602 
4603  if (!string_insn_completed(ctxt)) {
4604  /*
4605  * Re-enter guest when pio read ahead buffer is empty
4606  * or, if it is not used, after each 1024 iteration.
4607  */
4608  if ((r->end != 0 || reg_read(ctxt, VCPU_REGS_RCX) & 0x3ff) &&
4609  (r->end == 0 || r->end != r->pos)) {
4610  /*
4611  * Reset read cache. Usually happens before
4612  * decode, but since instruction is restarted
4613  * we have to do it here.
4614  */
4615  ctxt->mem_read.end = 0;
4616  writeback_registers(ctxt);
4617  return EMULATION_RESTART;
4618  }
4619  goto done; /* skip rip writeback */
4620  }
4621  }
4622 
4623  ctxt->eip = ctxt->_eip;
4624 
4625 done:
4626  if (rc == X86EMUL_PROPAGATE_FAULT)
4627  ctxt->have_exception = true;
4628  if (rc == X86EMUL_INTERCEPTED)
4629  return EMULATION_INTERCEPTED;
4630 
4631  if (rc == X86EMUL_CONTINUE)
4632  writeback_registers(ctxt);
4633 
4635 
4636 twobyte_insn:
4637  switch (ctxt->b) {
4638  case 0x09: /* wbinvd */
4639  (ctxt->ops->wbinvd)(ctxt);
4640  break;
4641  case 0x08: /* invd */
4642  case 0x0d: /* GrpP (prefetch) */
4643  case 0x18: /* Grp16 (prefetch/nop) */
4644  break;
4645  case 0x20: /* mov cr, reg */
4646  ctxt->dst.val = ops->get_cr(ctxt, ctxt->modrm_reg);
4647  break;
4648  case 0x21: /* mov from dr to reg */
4649  ops->get_dr(ctxt, ctxt->modrm_reg, &ctxt->dst.val);
4650  break;
4651  case 0x40 ... 0x4f: /* cmov */
4652  ctxt->dst.val = ctxt->dst.orig_val = ctxt->src.val;
4653  if (!test_cc(ctxt->b, ctxt->eflags))
4654  ctxt->dst.type = OP_NONE; /* no writeback */
4655  break;
4656  case 0x80 ... 0x8f: /* jnz rel, etc*/
4657  if (test_cc(ctxt->b, ctxt->eflags))
4658  jmp_rel(ctxt, ctxt->src.val);
4659  break;
4660  case 0x90 ... 0x9f: /* setcc r/m8 */
4661  ctxt->dst.val = test_cc(ctxt->b, ctxt->eflags);
4662  break;
4663  case 0xa4: /* shld imm8, r, r/m */
4664  case 0xa5: /* shld cl, r, r/m */
4665  emulate_2op_cl(ctxt, "shld");
4666  break;
4667  case 0xac: /* shrd imm8, r, r/m */
4668  case 0xad: /* shrd cl, r, r/m */
4669  emulate_2op_cl(ctxt, "shrd");
4670  break;
4671  case 0xae: /* clflush */
4672  break;
4673  case 0xb6 ... 0xb7: /* movzx */
4674  ctxt->dst.bytes = ctxt->op_bytes;
4675  ctxt->dst.val = (ctxt->src.bytes == 1) ? (u8) ctxt->src.val
4676  : (u16) ctxt->src.val;
4677  break;
4678  case 0xbe ... 0xbf: /* movsx */
4679  ctxt->dst.bytes = ctxt->op_bytes;
4680  ctxt->dst.val = (ctxt->src.bytes == 1) ? (s8) ctxt->src.val :
4681  (s16) ctxt->src.val;
4682  break;
4683  case 0xc0 ... 0xc1: /* xadd */
4684  emulate_2op_SrcV(ctxt, "add");
4685  /* Write back the register source. */
4686  ctxt->src.val = ctxt->dst.orig_val;
4687  write_register_operand(&ctxt->src);
4688  break;
4689  case 0xc3: /* movnti */
4690  ctxt->dst.bytes = ctxt->op_bytes;
4691  ctxt->dst.val = (ctxt->op_bytes == 4) ? (u32) ctxt->src.val :
4692  (u64) ctxt->src.val;
4693  break;
4694  default:
4695  goto cannot_emulate;
4696  }
4697 
4698  if (rc != X86EMUL_CONTINUE)
4699  goto done;
4700 
4701  goto writeback;
4702 
4703 cannot_emulate:
4704  return EMULATION_FAILED;
4705 }
4706 
4708 {
4709  invalidate_registers(ctxt);
4710 }
4711 
4713 {
4714  writeback_registers(ctxt);
4715 }