xref: /openbmc/linux/arch/x86/kvm/emulate.c (revision 78700c0a)
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 <avi@qumranet.com>
15  *   Yaniv Kamay <yaniv@qumranet.com>
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 #include <linux/stringify.h>
28 #include <asm/debugreg.h>
29 
30 #include "x86.h"
31 #include "tss.h"
32 
33 /*
34  * Operand types
35  */
36 #define OpNone             0ull
37 #define OpImplicit         1ull  /* No generic decode */
38 #define OpReg              2ull  /* Register */
39 #define OpMem              3ull  /* Memory */
40 #define OpAcc              4ull  /* Accumulator: AL/AX/EAX/RAX */
41 #define OpDI               5ull  /* ES:DI/EDI/RDI */
42 #define OpMem64            6ull  /* Memory, 64-bit */
43 #define OpImmUByte         7ull  /* Zero-extended 8-bit immediate */
44 #define OpDX               8ull  /* DX register */
45 #define OpCL               9ull  /* CL register (for shifts) */
46 #define OpImmByte         10ull  /* 8-bit sign extended immediate */
47 #define OpOne             11ull  /* Implied 1 */
48 #define OpImm             12ull  /* Sign extended up to 32-bit immediate */
49 #define OpMem16           13ull  /* Memory operand (16-bit). */
50 #define OpMem32           14ull  /* Memory operand (32-bit). */
51 #define OpImmU            15ull  /* Immediate operand, zero extended */
52 #define OpSI              16ull  /* SI/ESI/RSI */
53 #define OpImmFAddr        17ull  /* Immediate far address */
54 #define OpMemFAddr        18ull  /* Far address in memory */
55 #define OpImmU16          19ull  /* Immediate operand, 16 bits, zero extended */
56 #define OpES              20ull  /* ES */
57 #define OpCS              21ull  /* CS */
58 #define OpSS              22ull  /* SS */
59 #define OpDS              23ull  /* DS */
60 #define OpFS              24ull  /* FS */
61 #define OpGS              25ull  /* GS */
62 #define OpMem8            26ull  /* 8-bit zero extended memory operand */
63 #define OpImm64           27ull  /* Sign extended 16/32/64-bit immediate */
64 #define OpXLat            28ull  /* memory at BX/EBX/RBX + zero-extended AL */
65 #define OpAccLo           29ull  /* Low part of extended acc (AX/AX/EAX/RAX) */
66 #define OpAccHi           30ull  /* High part of extended acc (-/DX/EDX/RDX) */
67 
68 #define OpBits             5  /* Width of operand field */
69 #define OpMask             ((1ull << OpBits) - 1)
70 
71 /*
72  * Opcode effective-address decode tables.
73  * Note that we only emulate instructions that have at least one memory
74  * operand (excluding implicit stack references). We assume that stack
75  * references and instruction fetches will never occur in special memory
76  * areas that require emulation. So, for example, 'mov <imm>,<reg>' need
77  * not be handled.
78  */
79 
80 /* Operand sizes: 8-bit operands or specified/overridden size. */
81 #define ByteOp      (1<<0)	/* 8-bit operands. */
82 /* Destination operand type. */
83 #define DstShift    1
84 #define ImplicitOps (OpImplicit << DstShift)
85 #define DstReg      (OpReg << DstShift)
86 #define DstMem      (OpMem << DstShift)
87 #define DstAcc      (OpAcc << DstShift)
88 #define DstDI       (OpDI << DstShift)
89 #define DstMem64    (OpMem64 << DstShift)
90 #define DstMem16    (OpMem16 << DstShift)
91 #define DstImmUByte (OpImmUByte << DstShift)
92 #define DstDX       (OpDX << DstShift)
93 #define DstAccLo    (OpAccLo << DstShift)
94 #define DstMask     (OpMask << DstShift)
95 /* Source operand type. */
96 #define SrcShift    6
97 #define SrcNone     (OpNone << SrcShift)
98 #define SrcReg      (OpReg << SrcShift)
99 #define SrcMem      (OpMem << SrcShift)
100 #define SrcMem16    (OpMem16 << SrcShift)
101 #define SrcMem32    (OpMem32 << SrcShift)
102 #define SrcImm      (OpImm << SrcShift)
103 #define SrcImmByte  (OpImmByte << SrcShift)
104 #define SrcOne      (OpOne << SrcShift)
105 #define SrcImmUByte (OpImmUByte << SrcShift)
106 #define SrcImmU     (OpImmU << SrcShift)
107 #define SrcSI       (OpSI << SrcShift)
108 #define SrcXLat     (OpXLat << SrcShift)
109 #define SrcImmFAddr (OpImmFAddr << SrcShift)
110 #define SrcMemFAddr (OpMemFAddr << SrcShift)
111 #define SrcAcc      (OpAcc << SrcShift)
112 #define SrcImmU16   (OpImmU16 << SrcShift)
113 #define SrcImm64    (OpImm64 << SrcShift)
114 #define SrcDX       (OpDX << SrcShift)
115 #define SrcMem8     (OpMem8 << SrcShift)
116 #define SrcAccHi    (OpAccHi << SrcShift)
117 #define SrcMask     (OpMask << SrcShift)
118 #define BitOp       (1<<11)
119 #define MemAbs      (1<<12)      /* Memory operand is absolute displacement */
120 #define String      (1<<13)     /* String instruction (rep capable) */
121 #define Stack       (1<<14)     /* Stack instruction (push/pop) */
122 #define GroupMask   (7<<15)     /* Opcode uses one of the group mechanisms */
123 #define Group       (1<<15)     /* Bits 3:5 of modrm byte extend opcode */
124 #define GroupDual   (2<<15)     /* Alternate decoding of mod == 3 */
125 #define Prefix      (3<<15)     /* Instruction varies with 66/f2/f3 prefix */
126 #define RMExt       (4<<15)     /* Opcode extension in ModRM r/m if mod == 3 */
127 #define Escape      (5<<15)     /* Escape to coprocessor instruction */
128 #define InstrDual   (6<<15)     /* Alternate instruction decoding of mod == 3 */
129 #define ModeDual    (7<<15)     /* Different instruction for 32/64 bit */
130 #define Sse         (1<<18)     /* SSE Vector instruction */
131 /* Generic ModRM decode. */
132 #define ModRM       (1<<19)
133 /* Destination is only written; never read. */
134 #define Mov         (1<<20)
135 /* Misc flags */
136 #define Prot        (1<<21) /* instruction generates #UD if not in prot-mode */
137 #define EmulateOnUD (1<<22) /* Emulate if unsupported by the host */
138 #define NoAccess    (1<<23) /* Don't access memory (lea/invlpg/verr etc) */
139 #define Op3264      (1<<24) /* Operand is 64b in long mode, 32b otherwise */
140 #define Undefined   (1<<25) /* No Such Instruction */
141 #define Lock        (1<<26) /* lock prefix is allowed for the instruction */
142 #define Priv        (1<<27) /* instruction generates #GP if current CPL != 0 */
143 #define No64	    (1<<28)
144 #define PageTable   (1 << 29)   /* instruction used to write page table */
145 #define NotImpl     (1 << 30)   /* instruction is not implemented */
146 /* Source 2 operand type */
147 #define Src2Shift   (31)
148 #define Src2None    (OpNone << Src2Shift)
149 #define Src2Mem     (OpMem << Src2Shift)
150 #define Src2CL      (OpCL << Src2Shift)
151 #define Src2ImmByte (OpImmByte << Src2Shift)
152 #define Src2One     (OpOne << Src2Shift)
153 #define Src2Imm     (OpImm << Src2Shift)
154 #define Src2ES      (OpES << Src2Shift)
155 #define Src2CS      (OpCS << Src2Shift)
156 #define Src2SS      (OpSS << Src2Shift)
157 #define Src2DS      (OpDS << Src2Shift)
158 #define Src2FS      (OpFS << Src2Shift)
159 #define Src2GS      (OpGS << Src2Shift)
160 #define Src2Mask    (OpMask << Src2Shift)
161 #define Mmx         ((u64)1 << 40)  /* MMX Vector instruction */
162 #define Aligned     ((u64)1 << 41)  /* Explicitly aligned (e.g. MOVDQA) */
163 #define Unaligned   ((u64)1 << 42)  /* Explicitly unaligned (e.g. MOVDQU) */
164 #define Avx         ((u64)1 << 43)  /* Advanced Vector Extensions */
165 #define Fastop      ((u64)1 << 44)  /* Use opcode::u.fastop */
166 #define NoWrite     ((u64)1 << 45)  /* No writeback */
167 #define SrcWrite    ((u64)1 << 46)  /* Write back src operand */
168 #define NoMod	    ((u64)1 << 47)  /* Mod field is ignored */
169 #define Intercept   ((u64)1 << 48)  /* Has valid intercept field */
170 #define CheckPerm   ((u64)1 << 49)  /* Has valid check_perm field */
171 #define PrivUD      ((u64)1 << 51)  /* #UD instead of #GP on CPL > 0 */
172 #define NearBranch  ((u64)1 << 52)  /* Near branches */
173 #define No16	    ((u64)1 << 53)  /* No 16 bit operand */
174 #define IncSP       ((u64)1 << 54)  /* SP is incremented before ModRM calc */
175 
176 #define DstXacc     (DstAccLo | SrcAccHi | SrcWrite)
177 
178 #define X2(x...) x, x
179 #define X3(x...) X2(x), x
180 #define X4(x...) X2(x), X2(x)
181 #define X5(x...) X4(x), x
182 #define X6(x...) X4(x), X2(x)
183 #define X7(x...) X4(x), X3(x)
184 #define X8(x...) X4(x), X4(x)
185 #define X16(x...) X8(x), X8(x)
186 
187 #define NR_FASTOP (ilog2(sizeof(ulong)) + 1)
188 #define FASTOP_SIZE 8
189 
190 /*
191  * fastop functions have a special calling convention:
192  *
193  * dst:    rax        (in/out)
194  * src:    rdx        (in/out)
195  * src2:   rcx        (in)
196  * flags:  rflags     (in/out)
197  * ex:     rsi        (in:fastop pointer, out:zero if exception)
198  *
199  * Moreover, they are all exactly FASTOP_SIZE bytes long, so functions for
200  * different operand sizes can be reached by calculation, rather than a jump
201  * table (which would be bigger than the code).
202  *
203  * fastop functions are declared as taking a never-defined fastop parameter,
204  * so they can't be called from C directly.
205  */
206 
207 struct fastop;
208 
209 struct opcode {
210 	u64 flags : 56;
211 	u64 intercept : 8;
212 	union {
213 		int (*execute)(struct x86_emulate_ctxt *ctxt);
214 		const struct opcode *group;
215 		const struct group_dual *gdual;
216 		const struct gprefix *gprefix;
217 		const struct escape *esc;
218 		const struct instr_dual *idual;
219 		const struct mode_dual *mdual;
220 		void (*fastop)(struct fastop *fake);
221 	} u;
222 	int (*check_perm)(struct x86_emulate_ctxt *ctxt);
223 };
224 
225 struct group_dual {
226 	struct opcode mod012[8];
227 	struct opcode mod3[8];
228 };
229 
230 struct gprefix {
231 	struct opcode pfx_no;
232 	struct opcode pfx_66;
233 	struct opcode pfx_f2;
234 	struct opcode pfx_f3;
235 };
236 
237 struct escape {
238 	struct opcode op[8];
239 	struct opcode high[64];
240 };
241 
242 struct instr_dual {
243 	struct opcode mod012;
244 	struct opcode mod3;
245 };
246 
247 struct mode_dual {
248 	struct opcode mode32;
249 	struct opcode mode64;
250 };
251 
252 #define EFLG_RESERVED_ZEROS_MASK 0xffc0802a
253 
254 enum x86_transfer_type {
255 	X86_TRANSFER_NONE,
256 	X86_TRANSFER_CALL_JMP,
257 	X86_TRANSFER_RET,
258 	X86_TRANSFER_TASK_SWITCH,
259 };
260 
261 static ulong reg_read(struct x86_emulate_ctxt *ctxt, unsigned nr)
262 {
263 	if (!(ctxt->regs_valid & (1 << nr))) {
264 		ctxt->regs_valid |= 1 << nr;
265 		ctxt->_regs[nr] = ctxt->ops->read_gpr(ctxt, nr);
266 	}
267 	return ctxt->_regs[nr];
268 }
269 
270 static ulong *reg_write(struct x86_emulate_ctxt *ctxt, unsigned nr)
271 {
272 	ctxt->regs_valid |= 1 << nr;
273 	ctxt->regs_dirty |= 1 << nr;
274 	return &ctxt->_regs[nr];
275 }
276 
277 static ulong *reg_rmw(struct x86_emulate_ctxt *ctxt, unsigned nr)
278 {
279 	reg_read(ctxt, nr);
280 	return reg_write(ctxt, nr);
281 }
282 
283 static void writeback_registers(struct x86_emulate_ctxt *ctxt)
284 {
285 	unsigned reg;
286 
287 	for_each_set_bit(reg, (ulong *)&ctxt->regs_dirty, 16)
288 		ctxt->ops->write_gpr(ctxt, reg, ctxt->_regs[reg]);
289 }
290 
291 static void invalidate_registers(struct x86_emulate_ctxt *ctxt)
292 {
293 	ctxt->regs_dirty = 0;
294 	ctxt->regs_valid = 0;
295 }
296 
297 /*
298  * These EFLAGS bits are restored from saved value during emulation, and
299  * any changes are written back to the saved value after emulation.
300  */
301 #define EFLAGS_MASK (X86_EFLAGS_OF|X86_EFLAGS_SF|X86_EFLAGS_ZF|X86_EFLAGS_AF|\
302 		     X86_EFLAGS_PF|X86_EFLAGS_CF)
303 
304 #ifdef CONFIG_X86_64
305 #define ON64(x) x
306 #else
307 #define ON64(x)
308 #endif
309 
310 static int fastop(struct x86_emulate_ctxt *ctxt, void (*fop)(struct fastop *));
311 
312 #define FOP_FUNC(name) \
313 	".align " __stringify(FASTOP_SIZE) " \n\t" \
314 	".type " name ", @function \n\t" \
315 	name ":\n\t"
316 
317 #define FOP_RET   "ret \n\t"
318 
319 #define FOP_START(op) \
320 	extern void em_##op(struct fastop *fake); \
321 	asm(".pushsection .text, \"ax\" \n\t" \
322 	    ".global em_" #op " \n\t" \
323 	    FOP_FUNC("em_" #op)
324 
325 #define FOP_END \
326 	    ".popsection")
327 
328 #define FOPNOP() \
329 	FOP_FUNC(__stringify(__UNIQUE_ID(nop))) \
330 	FOP_RET
331 
332 #define FOP1E(op,  dst) \
333 	FOP_FUNC(#op "_" #dst) \
334 	"10: " #op " %" #dst " \n\t" FOP_RET
335 
336 #define FOP1EEX(op,  dst) \
337 	FOP1E(op, dst) _ASM_EXTABLE(10b, kvm_fastop_exception)
338 
339 #define FASTOP1(op) \
340 	FOP_START(op) \
341 	FOP1E(op##b, al) \
342 	FOP1E(op##w, ax) \
343 	FOP1E(op##l, eax) \
344 	ON64(FOP1E(op##q, rax))	\
345 	FOP_END
346 
347 /* 1-operand, using src2 (for MUL/DIV r/m) */
348 #define FASTOP1SRC2(op, name) \
349 	FOP_START(name) \
350 	FOP1E(op, cl) \
351 	FOP1E(op, cx) \
352 	FOP1E(op, ecx) \
353 	ON64(FOP1E(op, rcx)) \
354 	FOP_END
355 
356 /* 1-operand, using src2 (for MUL/DIV r/m), with exceptions */
357 #define FASTOP1SRC2EX(op, name) \
358 	FOP_START(name) \
359 	FOP1EEX(op, cl) \
360 	FOP1EEX(op, cx) \
361 	FOP1EEX(op, ecx) \
362 	ON64(FOP1EEX(op, rcx)) \
363 	FOP_END
364 
365 #define FOP2E(op,  dst, src)	   \
366 	FOP_FUNC(#op "_" #dst "_" #src) \
367 	#op " %" #src ", %" #dst " \n\t" FOP_RET
368 
369 #define FASTOP2(op) \
370 	FOP_START(op) \
371 	FOP2E(op##b, al, dl) \
372 	FOP2E(op##w, ax, dx) \
373 	FOP2E(op##l, eax, edx) \
374 	ON64(FOP2E(op##q, rax, rdx)) \
375 	FOP_END
376 
377 /* 2 operand, word only */
378 #define FASTOP2W(op) \
379 	FOP_START(op) \
380 	FOPNOP() \
381 	FOP2E(op##w, ax, dx) \
382 	FOP2E(op##l, eax, edx) \
383 	ON64(FOP2E(op##q, rax, rdx)) \
384 	FOP_END
385 
386 /* 2 operand, src is CL */
387 #define FASTOP2CL(op) \
388 	FOP_START(op) \
389 	FOP2E(op##b, al, cl) \
390 	FOP2E(op##w, ax, cl) \
391 	FOP2E(op##l, eax, cl) \
392 	ON64(FOP2E(op##q, rax, cl)) \
393 	FOP_END
394 
395 /* 2 operand, src and dest are reversed */
396 #define FASTOP2R(op, name) \
397 	FOP_START(name) \
398 	FOP2E(op##b, dl, al) \
399 	FOP2E(op##w, dx, ax) \
400 	FOP2E(op##l, edx, eax) \
401 	ON64(FOP2E(op##q, rdx, rax)) \
402 	FOP_END
403 
404 #define FOP3E(op,  dst, src, src2) \
405 	FOP_FUNC(#op "_" #dst "_" #src "_" #src2) \
406 	#op " %" #src2 ", %" #src ", %" #dst " \n\t" FOP_RET
407 
408 /* 3-operand, word-only, src2=cl */
409 #define FASTOP3WCL(op) \
410 	FOP_START(op) \
411 	FOPNOP() \
412 	FOP3E(op##w, ax, dx, cl) \
413 	FOP3E(op##l, eax, edx, cl) \
414 	ON64(FOP3E(op##q, rax, rdx, cl)) \
415 	FOP_END
416 
417 /* Special case for SETcc - 1 instruction per cc */
418 #define FOP_SETCC(op) \
419 	".align 4 \n\t" \
420 	".type " #op ", @function \n\t" \
421 	#op ": \n\t" \
422 	#op " %al \n\t" \
423 	FOP_RET
424 
425 asm(".global kvm_fastop_exception \n"
426     "kvm_fastop_exception: xor %esi, %esi; ret");
427 
428 FOP_START(setcc)
429 FOP_SETCC(seto)
430 FOP_SETCC(setno)
431 FOP_SETCC(setc)
432 FOP_SETCC(setnc)
433 FOP_SETCC(setz)
434 FOP_SETCC(setnz)
435 FOP_SETCC(setbe)
436 FOP_SETCC(setnbe)
437 FOP_SETCC(sets)
438 FOP_SETCC(setns)
439 FOP_SETCC(setp)
440 FOP_SETCC(setnp)
441 FOP_SETCC(setl)
442 FOP_SETCC(setnl)
443 FOP_SETCC(setle)
444 FOP_SETCC(setnle)
445 FOP_END;
446 
447 FOP_START(salc) "pushf; sbb %al, %al; popf \n\t" FOP_RET
448 FOP_END;
449 
450 static int emulator_check_intercept(struct x86_emulate_ctxt *ctxt,
451 				    enum x86_intercept intercept,
452 				    enum x86_intercept_stage stage)
453 {
454 	struct x86_instruction_info info = {
455 		.intercept  = intercept,
456 		.rep_prefix = ctxt->rep_prefix,
457 		.modrm_mod  = ctxt->modrm_mod,
458 		.modrm_reg  = ctxt->modrm_reg,
459 		.modrm_rm   = ctxt->modrm_rm,
460 		.src_val    = ctxt->src.val64,
461 		.dst_val    = ctxt->dst.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 void assign_register(unsigned long *reg, u64 val, int bytes)
477 {
478 	/* The 4-byte case *is* correct: in 64-bit mode we zero-extend. */
479 	switch (bytes) {
480 	case 1:
481 		*(u8 *)reg = (u8)val;
482 		break;
483 	case 2:
484 		*(u16 *)reg = (u16)val;
485 		break;
486 	case 4:
487 		*reg = (u32)val;
488 		break;	/* 64b: zero-extend */
489 	case 8:
490 		*reg = val;
491 		break;
492 	}
493 }
494 
495 static inline unsigned long ad_mask(struct x86_emulate_ctxt *ctxt)
496 {
497 	return (1UL << (ctxt->ad_bytes << 3)) - 1;
498 }
499 
500 static ulong stack_mask(struct x86_emulate_ctxt *ctxt)
501 {
502 	u16 sel;
503 	struct desc_struct ss;
504 
505 	if (ctxt->mode == X86EMUL_MODE_PROT64)
506 		return ~0UL;
507 	ctxt->ops->get_segment(ctxt, &sel, &ss, NULL, VCPU_SREG_SS);
508 	return ~0U >> ((ss.d ^ 1) * 16);  /* d=0: 0xffff; d=1: 0xffffffff */
509 }
510 
511 static int stack_size(struct x86_emulate_ctxt *ctxt)
512 {
513 	return (__fls(stack_mask(ctxt)) + 1) >> 3;
514 }
515 
516 /* Access/update address held in a register, based on addressing mode. */
517 static inline unsigned long
518 address_mask(struct x86_emulate_ctxt *ctxt, unsigned long reg)
519 {
520 	if (ctxt->ad_bytes == sizeof(unsigned long))
521 		return reg;
522 	else
523 		return reg & ad_mask(ctxt);
524 }
525 
526 static inline unsigned long
527 register_address(struct x86_emulate_ctxt *ctxt, int reg)
528 {
529 	return address_mask(ctxt, reg_read(ctxt, reg));
530 }
531 
532 static void masked_increment(ulong *reg, ulong mask, int inc)
533 {
534 	assign_masked(reg, *reg + inc, mask);
535 }
536 
537 static inline void
538 register_address_increment(struct x86_emulate_ctxt *ctxt, int reg, int inc)
539 {
540 	ulong *preg = reg_rmw(ctxt, reg);
541 
542 	assign_register(preg, *preg + inc, ctxt->ad_bytes);
543 }
544 
545 static void rsp_increment(struct x86_emulate_ctxt *ctxt, int inc)
546 {
547 	masked_increment(reg_rmw(ctxt, VCPU_REGS_RSP), stack_mask(ctxt), inc);
548 }
549 
550 static u32 desc_limit_scaled(struct desc_struct *desc)
551 {
552 	u32 limit = get_desc_limit(desc);
553 
554 	return desc->g ? (limit << 12) | 0xfff : limit;
555 }
556 
557 static unsigned long seg_base(struct x86_emulate_ctxt *ctxt, int seg)
558 {
559 	if (ctxt->mode == X86EMUL_MODE_PROT64 && seg < VCPU_SREG_FS)
560 		return 0;
561 
562 	return ctxt->ops->get_cached_segment_base(ctxt, seg);
563 }
564 
565 static int emulate_exception(struct x86_emulate_ctxt *ctxt, int vec,
566 			     u32 error, bool valid)
567 {
568 	WARN_ON(vec > 0x1f);
569 	ctxt->exception.vector = vec;
570 	ctxt->exception.error_code = error;
571 	ctxt->exception.error_code_valid = valid;
572 	return X86EMUL_PROPAGATE_FAULT;
573 }
574 
575 static int emulate_db(struct x86_emulate_ctxt *ctxt)
576 {
577 	return emulate_exception(ctxt, DB_VECTOR, 0, false);
578 }
579 
580 static int emulate_gp(struct x86_emulate_ctxt *ctxt, int err)
581 {
582 	return emulate_exception(ctxt, GP_VECTOR, err, true);
583 }
584 
585 static int emulate_ss(struct x86_emulate_ctxt *ctxt, int err)
586 {
587 	return emulate_exception(ctxt, SS_VECTOR, err, true);
588 }
589 
590 static int emulate_ud(struct x86_emulate_ctxt *ctxt)
591 {
592 	return emulate_exception(ctxt, UD_VECTOR, 0, false);
593 }
594 
595 static int emulate_ts(struct x86_emulate_ctxt *ctxt, int err)
596 {
597 	return emulate_exception(ctxt, TS_VECTOR, err, true);
598 }
599 
600 static int emulate_de(struct x86_emulate_ctxt *ctxt)
601 {
602 	return emulate_exception(ctxt, DE_VECTOR, 0, false);
603 }
604 
605 static int emulate_nm(struct x86_emulate_ctxt *ctxt)
606 {
607 	return emulate_exception(ctxt, NM_VECTOR, 0, false);
608 }
609 
610 static u16 get_segment_selector(struct x86_emulate_ctxt *ctxt, unsigned seg)
611 {
612 	u16 selector;
613 	struct desc_struct desc;
614 
615 	ctxt->ops->get_segment(ctxt, &selector, &desc, NULL, seg);
616 	return selector;
617 }
618 
619 static void set_segment_selector(struct x86_emulate_ctxt *ctxt, u16 selector,
620 				 unsigned seg)
621 {
622 	u16 dummy;
623 	u32 base3;
624 	struct desc_struct desc;
625 
626 	ctxt->ops->get_segment(ctxt, &dummy, &desc, &base3, seg);
627 	ctxt->ops->set_segment(ctxt, selector, &desc, base3, seg);
628 }
629 
630 /*
631  * x86 defines three classes of vector instructions: explicitly
632  * aligned, explicitly unaligned, and the rest, which change behaviour
633  * depending on whether they're AVX encoded or not.
634  *
635  * Also included is CMPXCHG16B which is not a vector instruction, yet it is
636  * subject to the same check.
637  */
638 static bool insn_aligned(struct x86_emulate_ctxt *ctxt, unsigned size)
639 {
640 	if (likely(size < 16))
641 		return false;
642 
643 	if (ctxt->d & Aligned)
644 		return true;
645 	else if (ctxt->d & Unaligned)
646 		return false;
647 	else if (ctxt->d & Avx)
648 		return false;
649 	else
650 		return true;
651 }
652 
653 static __always_inline int __linearize(struct x86_emulate_ctxt *ctxt,
654 				       struct segmented_address addr,
655 				       unsigned *max_size, unsigned size,
656 				       bool write, bool fetch,
657 				       enum x86emul_mode mode, ulong *linear)
658 {
659 	struct desc_struct desc;
660 	bool usable;
661 	ulong la;
662 	u32 lim;
663 	u16 sel;
664 
665 	la = seg_base(ctxt, addr.seg) + addr.ea;
666 	*max_size = 0;
667 	switch (mode) {
668 	case X86EMUL_MODE_PROT64:
669 		*linear = la;
670 		if (is_noncanonical_address(la))
671 			goto bad;
672 
673 		*max_size = min_t(u64, ~0u, (1ull << 48) - la);
674 		if (size > *max_size)
675 			goto bad;
676 		break;
677 	default:
678 		*linear = la = (u32)la;
679 		usable = ctxt->ops->get_segment(ctxt, &sel, &desc, NULL,
680 						addr.seg);
681 		if (!usable)
682 			goto bad;
683 		/* code segment in protected mode or read-only data segment */
684 		if ((((ctxt->mode != X86EMUL_MODE_REAL) && (desc.type & 8))
685 					|| !(desc.type & 2)) && write)
686 			goto bad;
687 		/* unreadable code segment */
688 		if (!fetch && (desc.type & 8) && !(desc.type & 2))
689 			goto bad;
690 		lim = desc_limit_scaled(&desc);
691 		if (!(desc.type & 8) && (desc.type & 4)) {
692 			/* expand-down segment */
693 			if (addr.ea <= lim)
694 				goto bad;
695 			lim = desc.d ? 0xffffffff : 0xffff;
696 		}
697 		if (addr.ea > lim)
698 			goto bad;
699 		if (lim == 0xffffffff)
700 			*max_size = ~0u;
701 		else {
702 			*max_size = (u64)lim + 1 - addr.ea;
703 			if (size > *max_size)
704 				goto bad;
705 		}
706 		break;
707 	}
708 	if (insn_aligned(ctxt, size) && ((la & (size - 1)) != 0))
709 		return emulate_gp(ctxt, 0);
710 	return X86EMUL_CONTINUE;
711 bad:
712 	if (addr.seg == VCPU_SREG_SS)
713 		return emulate_ss(ctxt, 0);
714 	else
715 		return emulate_gp(ctxt, 0);
716 }
717 
718 static int linearize(struct x86_emulate_ctxt *ctxt,
719 		     struct segmented_address addr,
720 		     unsigned size, bool write,
721 		     ulong *linear)
722 {
723 	unsigned max_size;
724 	return __linearize(ctxt, addr, &max_size, size, write, false,
725 			   ctxt->mode, linear);
726 }
727 
728 static inline int assign_eip(struct x86_emulate_ctxt *ctxt, ulong dst,
729 			     enum x86emul_mode mode)
730 {
731 	ulong linear;
732 	int rc;
733 	unsigned max_size;
734 	struct segmented_address addr = { .seg = VCPU_SREG_CS,
735 					   .ea = dst };
736 
737 	if (ctxt->op_bytes != sizeof(unsigned long))
738 		addr.ea = dst & ((1UL << (ctxt->op_bytes << 3)) - 1);
739 	rc = __linearize(ctxt, addr, &max_size, 1, false, true, mode, &linear);
740 	if (rc == X86EMUL_CONTINUE)
741 		ctxt->_eip = addr.ea;
742 	return rc;
743 }
744 
745 static inline int assign_eip_near(struct x86_emulate_ctxt *ctxt, ulong dst)
746 {
747 	return assign_eip(ctxt, dst, ctxt->mode);
748 }
749 
750 static int assign_eip_far(struct x86_emulate_ctxt *ctxt, ulong dst,
751 			  const struct desc_struct *cs_desc)
752 {
753 	enum x86emul_mode mode = ctxt->mode;
754 	int rc;
755 
756 #ifdef CONFIG_X86_64
757 	if (ctxt->mode >= X86EMUL_MODE_PROT16) {
758 		if (cs_desc->l) {
759 			u64 efer = 0;
760 
761 			ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
762 			if (efer & EFER_LMA)
763 				mode = X86EMUL_MODE_PROT64;
764 		} else
765 			mode = X86EMUL_MODE_PROT32; /* temporary value */
766 	}
767 #endif
768 	if (mode == X86EMUL_MODE_PROT16 || mode == X86EMUL_MODE_PROT32)
769 		mode = cs_desc->d ? X86EMUL_MODE_PROT32 : X86EMUL_MODE_PROT16;
770 	rc = assign_eip(ctxt, dst, mode);
771 	if (rc == X86EMUL_CONTINUE)
772 		ctxt->mode = mode;
773 	return rc;
774 }
775 
776 static inline int jmp_rel(struct x86_emulate_ctxt *ctxt, int rel)
777 {
778 	return assign_eip_near(ctxt, ctxt->_eip + rel);
779 }
780 
781 static int segmented_read_std(struct x86_emulate_ctxt *ctxt,
782 			      struct segmented_address addr,
783 			      void *data,
784 			      unsigned size)
785 {
786 	int rc;
787 	ulong linear;
788 
789 	rc = linearize(ctxt, addr, size, false, &linear);
790 	if (rc != X86EMUL_CONTINUE)
791 		return rc;
792 	return ctxt->ops->read_std(ctxt, linear, data, size, &ctxt->exception);
793 }
794 
795 /*
796  * Prefetch the remaining bytes of the instruction without crossing page
797  * boundary if they are not in fetch_cache yet.
798  */
799 static int __do_insn_fetch_bytes(struct x86_emulate_ctxt *ctxt, int op_size)
800 {
801 	int rc;
802 	unsigned size, max_size;
803 	unsigned long linear;
804 	int cur_size = ctxt->fetch.end - ctxt->fetch.data;
805 	struct segmented_address addr = { .seg = VCPU_SREG_CS,
806 					   .ea = ctxt->eip + cur_size };
807 
808 	/*
809 	 * We do not know exactly how many bytes will be needed, and
810 	 * __linearize is expensive, so fetch as much as possible.  We
811 	 * just have to avoid going beyond the 15 byte limit, the end
812 	 * of the segment, or the end of the page.
813 	 *
814 	 * __linearize is called with size 0 so that it does not do any
815 	 * boundary check itself.  Instead, we use max_size to check
816 	 * against op_size.
817 	 */
818 	rc = __linearize(ctxt, addr, &max_size, 0, false, true, ctxt->mode,
819 			 &linear);
820 	if (unlikely(rc != X86EMUL_CONTINUE))
821 		return rc;
822 
823 	size = min_t(unsigned, 15UL ^ cur_size, max_size);
824 	size = min_t(unsigned, size, PAGE_SIZE - offset_in_page(linear));
825 
826 	/*
827 	 * One instruction can only straddle two pages,
828 	 * and one has been loaded at the beginning of
829 	 * x86_decode_insn.  So, if not enough bytes
830 	 * still, we must have hit the 15-byte boundary.
831 	 */
832 	if (unlikely(size < op_size))
833 		return emulate_gp(ctxt, 0);
834 
835 	rc = ctxt->ops->fetch(ctxt, linear, ctxt->fetch.end,
836 			      size, &ctxt->exception);
837 	if (unlikely(rc != X86EMUL_CONTINUE))
838 		return rc;
839 	ctxt->fetch.end += size;
840 	return X86EMUL_CONTINUE;
841 }
842 
843 static __always_inline int do_insn_fetch_bytes(struct x86_emulate_ctxt *ctxt,
844 					       unsigned size)
845 {
846 	unsigned done_size = ctxt->fetch.end - ctxt->fetch.ptr;
847 
848 	if (unlikely(done_size < size))
849 		return __do_insn_fetch_bytes(ctxt, size - done_size);
850 	else
851 		return X86EMUL_CONTINUE;
852 }
853 
854 /* Fetch next part of the instruction being emulated. */
855 #define insn_fetch(_type, _ctxt)					\
856 ({	_type _x;							\
857 									\
858 	rc = do_insn_fetch_bytes(_ctxt, sizeof(_type));			\
859 	if (rc != X86EMUL_CONTINUE)					\
860 		goto done;						\
861 	ctxt->_eip += sizeof(_type);					\
862 	_x = *(_type __aligned(1) *) ctxt->fetch.ptr;			\
863 	ctxt->fetch.ptr += sizeof(_type);				\
864 	_x;								\
865 })
866 
867 #define insn_fetch_arr(_arr, _size, _ctxt)				\
868 ({									\
869 	rc = do_insn_fetch_bytes(_ctxt, _size);				\
870 	if (rc != X86EMUL_CONTINUE)					\
871 		goto done;						\
872 	ctxt->_eip += (_size);						\
873 	memcpy(_arr, ctxt->fetch.ptr, _size);				\
874 	ctxt->fetch.ptr += (_size);					\
875 })
876 
877 /*
878  * Given the 'reg' portion of a ModRM byte, and a register block, return a
879  * pointer into the block that addresses the relevant register.
880  * @highbyte_regs specifies whether to decode AH,CH,DH,BH.
881  */
882 static void *decode_register(struct x86_emulate_ctxt *ctxt, u8 modrm_reg,
883 			     int byteop)
884 {
885 	void *p;
886 	int highbyte_regs = (ctxt->rex_prefix == 0) && byteop;
887 
888 	if (highbyte_regs && modrm_reg >= 4 && modrm_reg < 8)
889 		p = (unsigned char *)reg_rmw(ctxt, modrm_reg & 3) + 1;
890 	else
891 		p = reg_rmw(ctxt, modrm_reg);
892 	return p;
893 }
894 
895 static int read_descriptor(struct x86_emulate_ctxt *ctxt,
896 			   struct segmented_address addr,
897 			   u16 *size, unsigned long *address, int op_bytes)
898 {
899 	int rc;
900 
901 	if (op_bytes == 2)
902 		op_bytes = 3;
903 	*address = 0;
904 	rc = segmented_read_std(ctxt, addr, size, 2);
905 	if (rc != X86EMUL_CONTINUE)
906 		return rc;
907 	addr.ea += 2;
908 	rc = segmented_read_std(ctxt, addr, address, op_bytes);
909 	return rc;
910 }
911 
912 FASTOP2(add);
913 FASTOP2(or);
914 FASTOP2(adc);
915 FASTOP2(sbb);
916 FASTOP2(and);
917 FASTOP2(sub);
918 FASTOP2(xor);
919 FASTOP2(cmp);
920 FASTOP2(test);
921 
922 FASTOP1SRC2(mul, mul_ex);
923 FASTOP1SRC2(imul, imul_ex);
924 FASTOP1SRC2EX(div, div_ex);
925 FASTOP1SRC2EX(idiv, idiv_ex);
926 
927 FASTOP3WCL(shld);
928 FASTOP3WCL(shrd);
929 
930 FASTOP2W(imul);
931 
932 FASTOP1(not);
933 FASTOP1(neg);
934 FASTOP1(inc);
935 FASTOP1(dec);
936 
937 FASTOP2CL(rol);
938 FASTOP2CL(ror);
939 FASTOP2CL(rcl);
940 FASTOP2CL(rcr);
941 FASTOP2CL(shl);
942 FASTOP2CL(shr);
943 FASTOP2CL(sar);
944 
945 FASTOP2W(bsf);
946 FASTOP2W(bsr);
947 FASTOP2W(bt);
948 FASTOP2W(bts);
949 FASTOP2W(btr);
950 FASTOP2W(btc);
951 
952 FASTOP2(xadd);
953 
954 FASTOP2R(cmp, cmp_r);
955 
956 static int em_bsf_c(struct x86_emulate_ctxt *ctxt)
957 {
958 	/* If src is zero, do not writeback, but update flags */
959 	if (ctxt->src.val == 0)
960 		ctxt->dst.type = OP_NONE;
961 	return fastop(ctxt, em_bsf);
962 }
963 
964 static int em_bsr_c(struct x86_emulate_ctxt *ctxt)
965 {
966 	/* If src is zero, do not writeback, but update flags */
967 	if (ctxt->src.val == 0)
968 		ctxt->dst.type = OP_NONE;
969 	return fastop(ctxt, em_bsr);
970 }
971 
972 static __always_inline u8 test_cc(unsigned int condition, unsigned long flags)
973 {
974 	u8 rc;
975 	void (*fop)(void) = (void *)em_setcc + 4 * (condition & 0xf);
976 
977 	flags = (flags & EFLAGS_MASK) | X86_EFLAGS_IF;
978 	asm("push %[flags]; popf; call *%[fastop]"
979 	    : "=a"(rc) : [fastop]"r"(fop), [flags]"r"(flags));
980 	return rc;
981 }
982 
983 static void fetch_register_operand(struct operand *op)
984 {
985 	switch (op->bytes) {
986 	case 1:
987 		op->val = *(u8 *)op->addr.reg;
988 		break;
989 	case 2:
990 		op->val = *(u16 *)op->addr.reg;
991 		break;
992 	case 4:
993 		op->val = *(u32 *)op->addr.reg;
994 		break;
995 	case 8:
996 		op->val = *(u64 *)op->addr.reg;
997 		break;
998 	}
999 }
1000 
1001 static void read_sse_reg(struct x86_emulate_ctxt *ctxt, sse128_t *data, int reg)
1002 {
1003 	ctxt->ops->get_fpu(ctxt);
1004 	switch (reg) {
1005 	case 0: asm("movdqa %%xmm0, %0" : "=m"(*data)); break;
1006 	case 1: asm("movdqa %%xmm1, %0" : "=m"(*data)); break;
1007 	case 2: asm("movdqa %%xmm2, %0" : "=m"(*data)); break;
1008 	case 3: asm("movdqa %%xmm3, %0" : "=m"(*data)); break;
1009 	case 4: asm("movdqa %%xmm4, %0" : "=m"(*data)); break;
1010 	case 5: asm("movdqa %%xmm5, %0" : "=m"(*data)); break;
1011 	case 6: asm("movdqa %%xmm6, %0" : "=m"(*data)); break;
1012 	case 7: asm("movdqa %%xmm7, %0" : "=m"(*data)); break;
1013 #ifdef CONFIG_X86_64
1014 	case 8: asm("movdqa %%xmm8, %0" : "=m"(*data)); break;
1015 	case 9: asm("movdqa %%xmm9, %0" : "=m"(*data)); break;
1016 	case 10: asm("movdqa %%xmm10, %0" : "=m"(*data)); break;
1017 	case 11: asm("movdqa %%xmm11, %0" : "=m"(*data)); break;
1018 	case 12: asm("movdqa %%xmm12, %0" : "=m"(*data)); break;
1019 	case 13: asm("movdqa %%xmm13, %0" : "=m"(*data)); break;
1020 	case 14: asm("movdqa %%xmm14, %0" : "=m"(*data)); break;
1021 	case 15: asm("movdqa %%xmm15, %0" : "=m"(*data)); break;
1022 #endif
1023 	default: BUG();
1024 	}
1025 	ctxt->ops->put_fpu(ctxt);
1026 }
1027 
1028 static void write_sse_reg(struct x86_emulate_ctxt *ctxt, sse128_t *data,
1029 			  int reg)
1030 {
1031 	ctxt->ops->get_fpu(ctxt);
1032 	switch (reg) {
1033 	case 0: asm("movdqa %0, %%xmm0" : : "m"(*data)); break;
1034 	case 1: asm("movdqa %0, %%xmm1" : : "m"(*data)); break;
1035 	case 2: asm("movdqa %0, %%xmm2" : : "m"(*data)); break;
1036 	case 3: asm("movdqa %0, %%xmm3" : : "m"(*data)); break;
1037 	case 4: asm("movdqa %0, %%xmm4" : : "m"(*data)); break;
1038 	case 5: asm("movdqa %0, %%xmm5" : : "m"(*data)); break;
1039 	case 6: asm("movdqa %0, %%xmm6" : : "m"(*data)); break;
1040 	case 7: asm("movdqa %0, %%xmm7" : : "m"(*data)); break;
1041 #ifdef CONFIG_X86_64
1042 	case 8: asm("movdqa %0, %%xmm8" : : "m"(*data)); break;
1043 	case 9: asm("movdqa %0, %%xmm9" : : "m"(*data)); break;
1044 	case 10: asm("movdqa %0, %%xmm10" : : "m"(*data)); break;
1045 	case 11: asm("movdqa %0, %%xmm11" : : "m"(*data)); break;
1046 	case 12: asm("movdqa %0, %%xmm12" : : "m"(*data)); break;
1047 	case 13: asm("movdqa %0, %%xmm13" : : "m"(*data)); break;
1048 	case 14: asm("movdqa %0, %%xmm14" : : "m"(*data)); break;
1049 	case 15: asm("movdqa %0, %%xmm15" : : "m"(*data)); break;
1050 #endif
1051 	default: BUG();
1052 	}
1053 	ctxt->ops->put_fpu(ctxt);
1054 }
1055 
1056 static void read_mmx_reg(struct x86_emulate_ctxt *ctxt, u64 *data, int reg)
1057 {
1058 	ctxt->ops->get_fpu(ctxt);
1059 	switch (reg) {
1060 	case 0: asm("movq %%mm0, %0" : "=m"(*data)); break;
1061 	case 1: asm("movq %%mm1, %0" : "=m"(*data)); break;
1062 	case 2: asm("movq %%mm2, %0" : "=m"(*data)); break;
1063 	case 3: asm("movq %%mm3, %0" : "=m"(*data)); break;
1064 	case 4: asm("movq %%mm4, %0" : "=m"(*data)); break;
1065 	case 5: asm("movq %%mm5, %0" : "=m"(*data)); break;
1066 	case 6: asm("movq %%mm6, %0" : "=m"(*data)); break;
1067 	case 7: asm("movq %%mm7, %0" : "=m"(*data)); break;
1068 	default: BUG();
1069 	}
1070 	ctxt->ops->put_fpu(ctxt);
1071 }
1072 
1073 static void write_mmx_reg(struct x86_emulate_ctxt *ctxt, u64 *data, int reg)
1074 {
1075 	ctxt->ops->get_fpu(ctxt);
1076 	switch (reg) {
1077 	case 0: asm("movq %0, %%mm0" : : "m"(*data)); break;
1078 	case 1: asm("movq %0, %%mm1" : : "m"(*data)); break;
1079 	case 2: asm("movq %0, %%mm2" : : "m"(*data)); break;
1080 	case 3: asm("movq %0, %%mm3" : : "m"(*data)); break;
1081 	case 4: asm("movq %0, %%mm4" : : "m"(*data)); break;
1082 	case 5: asm("movq %0, %%mm5" : : "m"(*data)); break;
1083 	case 6: asm("movq %0, %%mm6" : : "m"(*data)); break;
1084 	case 7: asm("movq %0, %%mm7" : : "m"(*data)); break;
1085 	default: BUG();
1086 	}
1087 	ctxt->ops->put_fpu(ctxt);
1088 }
1089 
1090 static int em_fninit(struct x86_emulate_ctxt *ctxt)
1091 {
1092 	if (ctxt->ops->get_cr(ctxt, 0) & (X86_CR0_TS | X86_CR0_EM))
1093 		return emulate_nm(ctxt);
1094 
1095 	ctxt->ops->get_fpu(ctxt);
1096 	asm volatile("fninit");
1097 	ctxt->ops->put_fpu(ctxt);
1098 	return X86EMUL_CONTINUE;
1099 }
1100 
1101 static int em_fnstcw(struct x86_emulate_ctxt *ctxt)
1102 {
1103 	u16 fcw;
1104 
1105 	if (ctxt->ops->get_cr(ctxt, 0) & (X86_CR0_TS | X86_CR0_EM))
1106 		return emulate_nm(ctxt);
1107 
1108 	ctxt->ops->get_fpu(ctxt);
1109 	asm volatile("fnstcw %0": "+m"(fcw));
1110 	ctxt->ops->put_fpu(ctxt);
1111 
1112 	ctxt->dst.val = fcw;
1113 
1114 	return X86EMUL_CONTINUE;
1115 }
1116 
1117 static int em_fnstsw(struct x86_emulate_ctxt *ctxt)
1118 {
1119 	u16 fsw;
1120 
1121 	if (ctxt->ops->get_cr(ctxt, 0) & (X86_CR0_TS | X86_CR0_EM))
1122 		return emulate_nm(ctxt);
1123 
1124 	ctxt->ops->get_fpu(ctxt);
1125 	asm volatile("fnstsw %0": "+m"(fsw));
1126 	ctxt->ops->put_fpu(ctxt);
1127 
1128 	ctxt->dst.val = fsw;
1129 
1130 	return X86EMUL_CONTINUE;
1131 }
1132 
1133 static void decode_register_operand(struct x86_emulate_ctxt *ctxt,
1134 				    struct operand *op)
1135 {
1136 	unsigned reg = ctxt->modrm_reg;
1137 
1138 	if (!(ctxt->d & ModRM))
1139 		reg = (ctxt->b & 7) | ((ctxt->rex_prefix & 1) << 3);
1140 
1141 	if (ctxt->d & Sse) {
1142 		op->type = OP_XMM;
1143 		op->bytes = 16;
1144 		op->addr.xmm = reg;
1145 		read_sse_reg(ctxt, &op->vec_val, reg);
1146 		return;
1147 	}
1148 	if (ctxt->d & Mmx) {
1149 		reg &= 7;
1150 		op->type = OP_MM;
1151 		op->bytes = 8;
1152 		op->addr.mm = reg;
1153 		return;
1154 	}
1155 
1156 	op->type = OP_REG;
1157 	op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
1158 	op->addr.reg = decode_register(ctxt, reg, ctxt->d & ByteOp);
1159 
1160 	fetch_register_operand(op);
1161 	op->orig_val = op->val;
1162 }
1163 
1164 static void adjust_modrm_seg(struct x86_emulate_ctxt *ctxt, int base_reg)
1165 {
1166 	if (base_reg == VCPU_REGS_RSP || base_reg == VCPU_REGS_RBP)
1167 		ctxt->modrm_seg = VCPU_SREG_SS;
1168 }
1169 
1170 static int decode_modrm(struct x86_emulate_ctxt *ctxt,
1171 			struct operand *op)
1172 {
1173 	u8 sib;
1174 	int index_reg, base_reg, scale;
1175 	int rc = X86EMUL_CONTINUE;
1176 	ulong modrm_ea = 0;
1177 
1178 	ctxt->modrm_reg = ((ctxt->rex_prefix << 1) & 8); /* REX.R */
1179 	index_reg = (ctxt->rex_prefix << 2) & 8; /* REX.X */
1180 	base_reg = (ctxt->rex_prefix << 3) & 8; /* REX.B */
1181 
1182 	ctxt->modrm_mod = (ctxt->modrm & 0xc0) >> 6;
1183 	ctxt->modrm_reg |= (ctxt->modrm & 0x38) >> 3;
1184 	ctxt->modrm_rm = base_reg | (ctxt->modrm & 0x07);
1185 	ctxt->modrm_seg = VCPU_SREG_DS;
1186 
1187 	if (ctxt->modrm_mod == 3 || (ctxt->d & NoMod)) {
1188 		op->type = OP_REG;
1189 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
1190 		op->addr.reg = decode_register(ctxt, ctxt->modrm_rm,
1191 				ctxt->d & ByteOp);
1192 		if (ctxt->d & Sse) {
1193 			op->type = OP_XMM;
1194 			op->bytes = 16;
1195 			op->addr.xmm = ctxt->modrm_rm;
1196 			read_sse_reg(ctxt, &op->vec_val, ctxt->modrm_rm);
1197 			return rc;
1198 		}
1199 		if (ctxt->d & Mmx) {
1200 			op->type = OP_MM;
1201 			op->bytes = 8;
1202 			op->addr.mm = ctxt->modrm_rm & 7;
1203 			return rc;
1204 		}
1205 		fetch_register_operand(op);
1206 		return rc;
1207 	}
1208 
1209 	op->type = OP_MEM;
1210 
1211 	if (ctxt->ad_bytes == 2) {
1212 		unsigned bx = reg_read(ctxt, VCPU_REGS_RBX);
1213 		unsigned bp = reg_read(ctxt, VCPU_REGS_RBP);
1214 		unsigned si = reg_read(ctxt, VCPU_REGS_RSI);
1215 		unsigned di = reg_read(ctxt, VCPU_REGS_RDI);
1216 
1217 		/* 16-bit ModR/M decode. */
1218 		switch (ctxt->modrm_mod) {
1219 		case 0:
1220 			if (ctxt->modrm_rm == 6)
1221 				modrm_ea += insn_fetch(u16, ctxt);
1222 			break;
1223 		case 1:
1224 			modrm_ea += insn_fetch(s8, ctxt);
1225 			break;
1226 		case 2:
1227 			modrm_ea += insn_fetch(u16, ctxt);
1228 			break;
1229 		}
1230 		switch (ctxt->modrm_rm) {
1231 		case 0:
1232 			modrm_ea += bx + si;
1233 			break;
1234 		case 1:
1235 			modrm_ea += bx + di;
1236 			break;
1237 		case 2:
1238 			modrm_ea += bp + si;
1239 			break;
1240 		case 3:
1241 			modrm_ea += bp + di;
1242 			break;
1243 		case 4:
1244 			modrm_ea += si;
1245 			break;
1246 		case 5:
1247 			modrm_ea += di;
1248 			break;
1249 		case 6:
1250 			if (ctxt->modrm_mod != 0)
1251 				modrm_ea += bp;
1252 			break;
1253 		case 7:
1254 			modrm_ea += bx;
1255 			break;
1256 		}
1257 		if (ctxt->modrm_rm == 2 || ctxt->modrm_rm == 3 ||
1258 		    (ctxt->modrm_rm == 6 && ctxt->modrm_mod != 0))
1259 			ctxt->modrm_seg = VCPU_SREG_SS;
1260 		modrm_ea = (u16)modrm_ea;
1261 	} else {
1262 		/* 32/64-bit ModR/M decode. */
1263 		if ((ctxt->modrm_rm & 7) == 4) {
1264 			sib = insn_fetch(u8, ctxt);
1265 			index_reg |= (sib >> 3) & 7;
1266 			base_reg |= sib & 7;
1267 			scale = sib >> 6;
1268 
1269 			if ((base_reg & 7) == 5 && ctxt->modrm_mod == 0)
1270 				modrm_ea += insn_fetch(s32, ctxt);
1271 			else {
1272 				modrm_ea += reg_read(ctxt, base_reg);
1273 				adjust_modrm_seg(ctxt, base_reg);
1274 				/* Increment ESP on POP [ESP] */
1275 				if ((ctxt->d & IncSP) &&
1276 				    base_reg == VCPU_REGS_RSP)
1277 					modrm_ea += ctxt->op_bytes;
1278 			}
1279 			if (index_reg != 4)
1280 				modrm_ea += reg_read(ctxt, index_reg) << scale;
1281 		} else if ((ctxt->modrm_rm & 7) == 5 && ctxt->modrm_mod == 0) {
1282 			modrm_ea += insn_fetch(s32, ctxt);
1283 			if (ctxt->mode == X86EMUL_MODE_PROT64)
1284 				ctxt->rip_relative = 1;
1285 		} else {
1286 			base_reg = ctxt->modrm_rm;
1287 			modrm_ea += reg_read(ctxt, base_reg);
1288 			adjust_modrm_seg(ctxt, base_reg);
1289 		}
1290 		switch (ctxt->modrm_mod) {
1291 		case 1:
1292 			modrm_ea += insn_fetch(s8, ctxt);
1293 			break;
1294 		case 2:
1295 			modrm_ea += insn_fetch(s32, ctxt);
1296 			break;
1297 		}
1298 	}
1299 	op->addr.mem.ea = modrm_ea;
1300 	if (ctxt->ad_bytes != 8)
1301 		ctxt->memop.addr.mem.ea = (u32)ctxt->memop.addr.mem.ea;
1302 
1303 done:
1304 	return rc;
1305 }
1306 
1307 static int decode_abs(struct x86_emulate_ctxt *ctxt,
1308 		      struct operand *op)
1309 {
1310 	int rc = X86EMUL_CONTINUE;
1311 
1312 	op->type = OP_MEM;
1313 	switch (ctxt->ad_bytes) {
1314 	case 2:
1315 		op->addr.mem.ea = insn_fetch(u16, ctxt);
1316 		break;
1317 	case 4:
1318 		op->addr.mem.ea = insn_fetch(u32, ctxt);
1319 		break;
1320 	case 8:
1321 		op->addr.mem.ea = insn_fetch(u64, ctxt);
1322 		break;
1323 	}
1324 done:
1325 	return rc;
1326 }
1327 
1328 static void fetch_bit_operand(struct x86_emulate_ctxt *ctxt)
1329 {
1330 	long sv = 0, mask;
1331 
1332 	if (ctxt->dst.type == OP_MEM && ctxt->src.type == OP_REG) {
1333 		mask = ~((long)ctxt->dst.bytes * 8 - 1);
1334 
1335 		if (ctxt->src.bytes == 2)
1336 			sv = (s16)ctxt->src.val & (s16)mask;
1337 		else if (ctxt->src.bytes == 4)
1338 			sv = (s32)ctxt->src.val & (s32)mask;
1339 		else
1340 			sv = (s64)ctxt->src.val & (s64)mask;
1341 
1342 		ctxt->dst.addr.mem.ea = address_mask(ctxt,
1343 					   ctxt->dst.addr.mem.ea + (sv >> 3));
1344 	}
1345 
1346 	/* only subword offset */
1347 	ctxt->src.val &= (ctxt->dst.bytes << 3) - 1;
1348 }
1349 
1350 static int read_emulated(struct x86_emulate_ctxt *ctxt,
1351 			 unsigned long addr, void *dest, unsigned size)
1352 {
1353 	int rc;
1354 	struct read_cache *mc = &ctxt->mem_read;
1355 
1356 	if (mc->pos < mc->end)
1357 		goto read_cached;
1358 
1359 	WARN_ON((mc->end + size) >= sizeof(mc->data));
1360 
1361 	rc = ctxt->ops->read_emulated(ctxt, addr, mc->data + mc->end, size,
1362 				      &ctxt->exception);
1363 	if (rc != X86EMUL_CONTINUE)
1364 		return rc;
1365 
1366 	mc->end += size;
1367 
1368 read_cached:
1369 	memcpy(dest, mc->data + mc->pos, size);
1370 	mc->pos += size;
1371 	return X86EMUL_CONTINUE;
1372 }
1373 
1374 static int segmented_read(struct x86_emulate_ctxt *ctxt,
1375 			  struct segmented_address addr,
1376 			  void *data,
1377 			  unsigned size)
1378 {
1379 	int rc;
1380 	ulong linear;
1381 
1382 	rc = linearize(ctxt, addr, size, false, &linear);
1383 	if (rc != X86EMUL_CONTINUE)
1384 		return rc;
1385 	return read_emulated(ctxt, linear, data, size);
1386 }
1387 
1388 static int segmented_write(struct x86_emulate_ctxt *ctxt,
1389 			   struct segmented_address addr,
1390 			   const void *data,
1391 			   unsigned size)
1392 {
1393 	int rc;
1394 	ulong linear;
1395 
1396 	rc = linearize(ctxt, addr, size, true, &linear);
1397 	if (rc != X86EMUL_CONTINUE)
1398 		return rc;
1399 	return ctxt->ops->write_emulated(ctxt, linear, data, size,
1400 					 &ctxt->exception);
1401 }
1402 
1403 static int segmented_cmpxchg(struct x86_emulate_ctxt *ctxt,
1404 			     struct segmented_address addr,
1405 			     const void *orig_data, const void *data,
1406 			     unsigned size)
1407 {
1408 	int rc;
1409 	ulong linear;
1410 
1411 	rc = linearize(ctxt, addr, size, true, &linear);
1412 	if (rc != X86EMUL_CONTINUE)
1413 		return rc;
1414 	return ctxt->ops->cmpxchg_emulated(ctxt, linear, orig_data, data,
1415 					   size, &ctxt->exception);
1416 }
1417 
1418 static int pio_in_emulated(struct x86_emulate_ctxt *ctxt,
1419 			   unsigned int size, unsigned short port,
1420 			   void *dest)
1421 {
1422 	struct read_cache *rc = &ctxt->io_read;
1423 
1424 	if (rc->pos == rc->end) { /* refill pio read ahead */
1425 		unsigned int in_page, n;
1426 		unsigned int count = ctxt->rep_prefix ?
1427 			address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) : 1;
1428 		in_page = (ctxt->eflags & X86_EFLAGS_DF) ?
1429 			offset_in_page(reg_read(ctxt, VCPU_REGS_RDI)) :
1430 			PAGE_SIZE - offset_in_page(reg_read(ctxt, VCPU_REGS_RDI));
1431 		n = min3(in_page, (unsigned int)sizeof(rc->data) / size, count);
1432 		if (n == 0)
1433 			n = 1;
1434 		rc->pos = rc->end = 0;
1435 		if (!ctxt->ops->pio_in_emulated(ctxt, size, port, rc->data, n))
1436 			return 0;
1437 		rc->end = n * size;
1438 	}
1439 
1440 	if (ctxt->rep_prefix && (ctxt->d & String) &&
1441 	    !(ctxt->eflags & X86_EFLAGS_DF)) {
1442 		ctxt->dst.data = rc->data + rc->pos;
1443 		ctxt->dst.type = OP_MEM_STR;
1444 		ctxt->dst.count = (rc->end - rc->pos) / size;
1445 		rc->pos = rc->end;
1446 	} else {
1447 		memcpy(dest, rc->data + rc->pos, size);
1448 		rc->pos += size;
1449 	}
1450 	return 1;
1451 }
1452 
1453 static int read_interrupt_descriptor(struct x86_emulate_ctxt *ctxt,
1454 				     u16 index, struct desc_struct *desc)
1455 {
1456 	struct desc_ptr dt;
1457 	ulong addr;
1458 
1459 	ctxt->ops->get_idt(ctxt, &dt);
1460 
1461 	if (dt.size < index * 8 + 7)
1462 		return emulate_gp(ctxt, index << 3 | 0x2);
1463 
1464 	addr = dt.address + index * 8;
1465 	return ctxt->ops->read_std(ctxt, addr, desc, sizeof *desc,
1466 				   &ctxt->exception);
1467 }
1468 
1469 static void get_descriptor_table_ptr(struct x86_emulate_ctxt *ctxt,
1470 				     u16 selector, struct desc_ptr *dt)
1471 {
1472 	const struct x86_emulate_ops *ops = ctxt->ops;
1473 	u32 base3 = 0;
1474 
1475 	if (selector & 1 << 2) {
1476 		struct desc_struct desc;
1477 		u16 sel;
1478 
1479 		memset (dt, 0, sizeof *dt);
1480 		if (!ops->get_segment(ctxt, &sel, &desc, &base3,
1481 				      VCPU_SREG_LDTR))
1482 			return;
1483 
1484 		dt->size = desc_limit_scaled(&desc); /* what if limit > 65535? */
1485 		dt->address = get_desc_base(&desc) | ((u64)base3 << 32);
1486 	} else
1487 		ops->get_gdt(ctxt, dt);
1488 }
1489 
1490 static int get_descriptor_ptr(struct x86_emulate_ctxt *ctxt,
1491 			      u16 selector, ulong *desc_addr_p)
1492 {
1493 	struct desc_ptr dt;
1494 	u16 index = selector >> 3;
1495 	ulong addr;
1496 
1497 	get_descriptor_table_ptr(ctxt, selector, &dt);
1498 
1499 	if (dt.size < index * 8 + 7)
1500 		return emulate_gp(ctxt, selector & 0xfffc);
1501 
1502 	addr = dt.address + index * 8;
1503 
1504 #ifdef CONFIG_X86_64
1505 	if (addr >> 32 != 0) {
1506 		u64 efer = 0;
1507 
1508 		ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
1509 		if (!(efer & EFER_LMA))
1510 			addr &= (u32)-1;
1511 	}
1512 #endif
1513 
1514 	*desc_addr_p = addr;
1515 	return X86EMUL_CONTINUE;
1516 }
1517 
1518 /* allowed just for 8 bytes segments */
1519 static int read_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1520 				   u16 selector, struct desc_struct *desc,
1521 				   ulong *desc_addr_p)
1522 {
1523 	int rc;
1524 
1525 	rc = get_descriptor_ptr(ctxt, selector, desc_addr_p);
1526 	if (rc != X86EMUL_CONTINUE)
1527 		return rc;
1528 
1529 	return ctxt->ops->read_std(ctxt, *desc_addr_p, desc, sizeof(*desc),
1530 				   &ctxt->exception);
1531 }
1532 
1533 /* allowed just for 8 bytes segments */
1534 static int write_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1535 				    u16 selector, struct desc_struct *desc)
1536 {
1537 	int rc;
1538 	ulong addr;
1539 
1540 	rc = get_descriptor_ptr(ctxt, selector, &addr);
1541 	if (rc != X86EMUL_CONTINUE)
1542 		return rc;
1543 
1544 	return ctxt->ops->write_std(ctxt, addr, desc, sizeof *desc,
1545 				    &ctxt->exception);
1546 }
1547 
1548 /* Does not support long mode */
1549 static int __load_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1550 				     u16 selector, int seg, u8 cpl,
1551 				     enum x86_transfer_type transfer,
1552 				     struct desc_struct *desc)
1553 {
1554 	struct desc_struct seg_desc, old_desc;
1555 	u8 dpl, rpl;
1556 	unsigned err_vec = GP_VECTOR;
1557 	u32 err_code = 0;
1558 	bool null_selector = !(selector & ~0x3); /* 0000-0003 are null */
1559 	ulong desc_addr;
1560 	int ret;
1561 	u16 dummy;
1562 	u32 base3 = 0;
1563 
1564 	memset(&seg_desc, 0, sizeof seg_desc);
1565 
1566 	if (ctxt->mode == X86EMUL_MODE_REAL) {
1567 		/* set real mode segment descriptor (keep limit etc. for
1568 		 * unreal mode) */
1569 		ctxt->ops->get_segment(ctxt, &dummy, &seg_desc, NULL, seg);
1570 		set_desc_base(&seg_desc, selector << 4);
1571 		goto load;
1572 	} else if (seg <= VCPU_SREG_GS && ctxt->mode == X86EMUL_MODE_VM86) {
1573 		/* VM86 needs a clean new segment descriptor */
1574 		set_desc_base(&seg_desc, selector << 4);
1575 		set_desc_limit(&seg_desc, 0xffff);
1576 		seg_desc.type = 3;
1577 		seg_desc.p = 1;
1578 		seg_desc.s = 1;
1579 		seg_desc.dpl = 3;
1580 		goto load;
1581 	}
1582 
1583 	rpl = selector & 3;
1584 
1585 	/* NULL selector is not valid for TR, CS and SS (except for long mode) */
1586 	if ((seg == VCPU_SREG_CS
1587 	     || (seg == VCPU_SREG_SS
1588 		 && (ctxt->mode != X86EMUL_MODE_PROT64 || rpl != cpl))
1589 	     || seg == VCPU_SREG_TR)
1590 	    && null_selector)
1591 		goto exception;
1592 
1593 	/* TR should be in GDT only */
1594 	if (seg == VCPU_SREG_TR && (selector & (1 << 2)))
1595 		goto exception;
1596 
1597 	if (null_selector) /* for NULL selector skip all following checks */
1598 		goto load;
1599 
1600 	ret = read_segment_descriptor(ctxt, selector, &seg_desc, &desc_addr);
1601 	if (ret != X86EMUL_CONTINUE)
1602 		return ret;
1603 
1604 	err_code = selector & 0xfffc;
1605 	err_vec = (transfer == X86_TRANSFER_TASK_SWITCH) ? TS_VECTOR :
1606 							   GP_VECTOR;
1607 
1608 	/* can't load system descriptor into segment selector */
1609 	if (seg <= VCPU_SREG_GS && !seg_desc.s) {
1610 		if (transfer == X86_TRANSFER_CALL_JMP)
1611 			return X86EMUL_UNHANDLEABLE;
1612 		goto exception;
1613 	}
1614 
1615 	if (!seg_desc.p) {
1616 		err_vec = (seg == VCPU_SREG_SS) ? SS_VECTOR : NP_VECTOR;
1617 		goto exception;
1618 	}
1619 
1620 	dpl = seg_desc.dpl;
1621 
1622 	switch (seg) {
1623 	case VCPU_SREG_SS:
1624 		/*
1625 		 * segment is not a writable data segment or segment
1626 		 * selector's RPL != CPL or segment selector's RPL != CPL
1627 		 */
1628 		if (rpl != cpl || (seg_desc.type & 0xa) != 0x2 || dpl != cpl)
1629 			goto exception;
1630 		break;
1631 	case VCPU_SREG_CS:
1632 		if (!(seg_desc.type & 8))
1633 			goto exception;
1634 
1635 		if (seg_desc.type & 4) {
1636 			/* conforming */
1637 			if (dpl > cpl)
1638 				goto exception;
1639 		} else {
1640 			/* nonconforming */
1641 			if (rpl > cpl || dpl != cpl)
1642 				goto exception;
1643 		}
1644 		/* in long-mode d/b must be clear if l is set */
1645 		if (seg_desc.d && seg_desc.l) {
1646 			u64 efer = 0;
1647 
1648 			ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
1649 			if (efer & EFER_LMA)
1650 				goto exception;
1651 		}
1652 
1653 		/* CS(RPL) <- CPL */
1654 		selector = (selector & 0xfffc) | cpl;
1655 		break;
1656 	case VCPU_SREG_TR:
1657 		if (seg_desc.s || (seg_desc.type != 1 && seg_desc.type != 9))
1658 			goto exception;
1659 		old_desc = seg_desc;
1660 		seg_desc.type |= 2; /* busy */
1661 		ret = ctxt->ops->cmpxchg_emulated(ctxt, desc_addr, &old_desc, &seg_desc,
1662 						  sizeof(seg_desc), &ctxt->exception);
1663 		if (ret != X86EMUL_CONTINUE)
1664 			return ret;
1665 		break;
1666 	case VCPU_SREG_LDTR:
1667 		if (seg_desc.s || seg_desc.type != 2)
1668 			goto exception;
1669 		break;
1670 	default: /*  DS, ES, FS, or GS */
1671 		/*
1672 		 * segment is not a data or readable code segment or
1673 		 * ((segment is a data or nonconforming code segment)
1674 		 * and (both RPL and CPL > DPL))
1675 		 */
1676 		if ((seg_desc.type & 0xa) == 0x8 ||
1677 		    (((seg_desc.type & 0xc) != 0xc) &&
1678 		     (rpl > dpl && cpl > dpl)))
1679 			goto exception;
1680 		break;
1681 	}
1682 
1683 	if (seg_desc.s) {
1684 		/* mark segment as accessed */
1685 		if (!(seg_desc.type & 1)) {
1686 			seg_desc.type |= 1;
1687 			ret = write_segment_descriptor(ctxt, selector,
1688 						       &seg_desc);
1689 			if (ret != X86EMUL_CONTINUE)
1690 				return ret;
1691 		}
1692 	} else if (ctxt->mode == X86EMUL_MODE_PROT64) {
1693 		ret = ctxt->ops->read_std(ctxt, desc_addr+8, &base3,
1694 				sizeof(base3), &ctxt->exception);
1695 		if (ret != X86EMUL_CONTINUE)
1696 			return ret;
1697 		if (is_noncanonical_address(get_desc_base(&seg_desc) |
1698 					     ((u64)base3 << 32)))
1699 			return emulate_gp(ctxt, 0);
1700 	}
1701 load:
1702 	ctxt->ops->set_segment(ctxt, selector, &seg_desc, base3, seg);
1703 	if (desc)
1704 		*desc = seg_desc;
1705 	return X86EMUL_CONTINUE;
1706 exception:
1707 	return emulate_exception(ctxt, err_vec, err_code, true);
1708 }
1709 
1710 static int load_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1711 				   u16 selector, int seg)
1712 {
1713 	u8 cpl = ctxt->ops->cpl(ctxt);
1714 	return __load_segment_descriptor(ctxt, selector, seg, cpl,
1715 					 X86_TRANSFER_NONE, NULL);
1716 }
1717 
1718 static void write_register_operand(struct operand *op)
1719 {
1720 	return assign_register(op->addr.reg, op->val, op->bytes);
1721 }
1722 
1723 static int writeback(struct x86_emulate_ctxt *ctxt, struct operand *op)
1724 {
1725 	switch (op->type) {
1726 	case OP_REG:
1727 		write_register_operand(op);
1728 		break;
1729 	case OP_MEM:
1730 		if (ctxt->lock_prefix)
1731 			return segmented_cmpxchg(ctxt,
1732 						 op->addr.mem,
1733 						 &op->orig_val,
1734 						 &op->val,
1735 						 op->bytes);
1736 		else
1737 			return segmented_write(ctxt,
1738 					       op->addr.mem,
1739 					       &op->val,
1740 					       op->bytes);
1741 		break;
1742 	case OP_MEM_STR:
1743 		return segmented_write(ctxt,
1744 				       op->addr.mem,
1745 				       op->data,
1746 				       op->bytes * op->count);
1747 		break;
1748 	case OP_XMM:
1749 		write_sse_reg(ctxt, &op->vec_val, op->addr.xmm);
1750 		break;
1751 	case OP_MM:
1752 		write_mmx_reg(ctxt, &op->mm_val, op->addr.mm);
1753 		break;
1754 	case OP_NONE:
1755 		/* no writeback */
1756 		break;
1757 	default:
1758 		break;
1759 	}
1760 	return X86EMUL_CONTINUE;
1761 }
1762 
1763 static int push(struct x86_emulate_ctxt *ctxt, void *data, int bytes)
1764 {
1765 	struct segmented_address addr;
1766 
1767 	rsp_increment(ctxt, -bytes);
1768 	addr.ea = reg_read(ctxt, VCPU_REGS_RSP) & stack_mask(ctxt);
1769 	addr.seg = VCPU_SREG_SS;
1770 
1771 	return segmented_write(ctxt, addr, data, bytes);
1772 }
1773 
1774 static int em_push(struct x86_emulate_ctxt *ctxt)
1775 {
1776 	/* Disable writeback. */
1777 	ctxt->dst.type = OP_NONE;
1778 	return push(ctxt, &ctxt->src.val, ctxt->op_bytes);
1779 }
1780 
1781 static int emulate_pop(struct x86_emulate_ctxt *ctxt,
1782 		       void *dest, int len)
1783 {
1784 	int rc;
1785 	struct segmented_address addr;
1786 
1787 	addr.ea = reg_read(ctxt, VCPU_REGS_RSP) & stack_mask(ctxt);
1788 	addr.seg = VCPU_SREG_SS;
1789 	rc = segmented_read(ctxt, addr, dest, len);
1790 	if (rc != X86EMUL_CONTINUE)
1791 		return rc;
1792 
1793 	rsp_increment(ctxt, len);
1794 	return rc;
1795 }
1796 
1797 static int em_pop(struct x86_emulate_ctxt *ctxt)
1798 {
1799 	return emulate_pop(ctxt, &ctxt->dst.val, ctxt->op_bytes);
1800 }
1801 
1802 static int emulate_popf(struct x86_emulate_ctxt *ctxt,
1803 			void *dest, int len)
1804 {
1805 	int rc;
1806 	unsigned long val, change_mask;
1807 	int iopl = (ctxt->eflags & X86_EFLAGS_IOPL) >> X86_EFLAGS_IOPL_BIT;
1808 	int cpl = ctxt->ops->cpl(ctxt);
1809 
1810 	rc = emulate_pop(ctxt, &val, len);
1811 	if (rc != X86EMUL_CONTINUE)
1812 		return rc;
1813 
1814 	change_mask = X86_EFLAGS_CF | X86_EFLAGS_PF | X86_EFLAGS_AF |
1815 		      X86_EFLAGS_ZF | X86_EFLAGS_SF | X86_EFLAGS_OF |
1816 		      X86_EFLAGS_TF | X86_EFLAGS_DF | X86_EFLAGS_NT |
1817 		      X86_EFLAGS_AC | X86_EFLAGS_ID;
1818 
1819 	switch(ctxt->mode) {
1820 	case X86EMUL_MODE_PROT64:
1821 	case X86EMUL_MODE_PROT32:
1822 	case X86EMUL_MODE_PROT16:
1823 		if (cpl == 0)
1824 			change_mask |= X86_EFLAGS_IOPL;
1825 		if (cpl <= iopl)
1826 			change_mask |= X86_EFLAGS_IF;
1827 		break;
1828 	case X86EMUL_MODE_VM86:
1829 		if (iopl < 3)
1830 			return emulate_gp(ctxt, 0);
1831 		change_mask |= X86_EFLAGS_IF;
1832 		break;
1833 	default: /* real mode */
1834 		change_mask |= (X86_EFLAGS_IOPL | X86_EFLAGS_IF);
1835 		break;
1836 	}
1837 
1838 	*(unsigned long *)dest =
1839 		(ctxt->eflags & ~change_mask) | (val & change_mask);
1840 
1841 	return rc;
1842 }
1843 
1844 static int em_popf(struct x86_emulate_ctxt *ctxt)
1845 {
1846 	ctxt->dst.type = OP_REG;
1847 	ctxt->dst.addr.reg = &ctxt->eflags;
1848 	ctxt->dst.bytes = ctxt->op_bytes;
1849 	return emulate_popf(ctxt, &ctxt->dst.val, ctxt->op_bytes);
1850 }
1851 
1852 static int em_enter(struct x86_emulate_ctxt *ctxt)
1853 {
1854 	int rc;
1855 	unsigned frame_size = ctxt->src.val;
1856 	unsigned nesting_level = ctxt->src2.val & 31;
1857 	ulong rbp;
1858 
1859 	if (nesting_level)
1860 		return X86EMUL_UNHANDLEABLE;
1861 
1862 	rbp = reg_read(ctxt, VCPU_REGS_RBP);
1863 	rc = push(ctxt, &rbp, stack_size(ctxt));
1864 	if (rc != X86EMUL_CONTINUE)
1865 		return rc;
1866 	assign_masked(reg_rmw(ctxt, VCPU_REGS_RBP), reg_read(ctxt, VCPU_REGS_RSP),
1867 		      stack_mask(ctxt));
1868 	assign_masked(reg_rmw(ctxt, VCPU_REGS_RSP),
1869 		      reg_read(ctxt, VCPU_REGS_RSP) - frame_size,
1870 		      stack_mask(ctxt));
1871 	return X86EMUL_CONTINUE;
1872 }
1873 
1874 static int em_leave(struct x86_emulate_ctxt *ctxt)
1875 {
1876 	assign_masked(reg_rmw(ctxt, VCPU_REGS_RSP), reg_read(ctxt, VCPU_REGS_RBP),
1877 		      stack_mask(ctxt));
1878 	return emulate_pop(ctxt, reg_rmw(ctxt, VCPU_REGS_RBP), ctxt->op_bytes);
1879 }
1880 
1881 static int em_push_sreg(struct x86_emulate_ctxt *ctxt)
1882 {
1883 	int seg = ctxt->src2.val;
1884 
1885 	ctxt->src.val = get_segment_selector(ctxt, seg);
1886 	if (ctxt->op_bytes == 4) {
1887 		rsp_increment(ctxt, -2);
1888 		ctxt->op_bytes = 2;
1889 	}
1890 
1891 	return em_push(ctxt);
1892 }
1893 
1894 static int em_pop_sreg(struct x86_emulate_ctxt *ctxt)
1895 {
1896 	int seg = ctxt->src2.val;
1897 	unsigned long selector;
1898 	int rc;
1899 
1900 	rc = emulate_pop(ctxt, &selector, 2);
1901 	if (rc != X86EMUL_CONTINUE)
1902 		return rc;
1903 
1904 	if (ctxt->modrm_reg == VCPU_SREG_SS)
1905 		ctxt->interruptibility = KVM_X86_SHADOW_INT_MOV_SS;
1906 	if (ctxt->op_bytes > 2)
1907 		rsp_increment(ctxt, ctxt->op_bytes - 2);
1908 
1909 	rc = load_segment_descriptor(ctxt, (u16)selector, seg);
1910 	return rc;
1911 }
1912 
1913 static int em_pusha(struct x86_emulate_ctxt *ctxt)
1914 {
1915 	unsigned long old_esp = reg_read(ctxt, VCPU_REGS_RSP);
1916 	int rc = X86EMUL_CONTINUE;
1917 	int reg = VCPU_REGS_RAX;
1918 
1919 	while (reg <= VCPU_REGS_RDI) {
1920 		(reg == VCPU_REGS_RSP) ?
1921 		(ctxt->src.val = old_esp) : (ctxt->src.val = reg_read(ctxt, reg));
1922 
1923 		rc = em_push(ctxt);
1924 		if (rc != X86EMUL_CONTINUE)
1925 			return rc;
1926 
1927 		++reg;
1928 	}
1929 
1930 	return rc;
1931 }
1932 
1933 static int em_pushf(struct x86_emulate_ctxt *ctxt)
1934 {
1935 	ctxt->src.val = (unsigned long)ctxt->eflags & ~X86_EFLAGS_VM;
1936 	return em_push(ctxt);
1937 }
1938 
1939 static int em_popa(struct x86_emulate_ctxt *ctxt)
1940 {
1941 	int rc = X86EMUL_CONTINUE;
1942 	int reg = VCPU_REGS_RDI;
1943 	u32 val;
1944 
1945 	while (reg >= VCPU_REGS_RAX) {
1946 		if (reg == VCPU_REGS_RSP) {
1947 			rsp_increment(ctxt, ctxt->op_bytes);
1948 			--reg;
1949 		}
1950 
1951 		rc = emulate_pop(ctxt, &val, ctxt->op_bytes);
1952 		if (rc != X86EMUL_CONTINUE)
1953 			break;
1954 		assign_register(reg_rmw(ctxt, reg), val, ctxt->op_bytes);
1955 		--reg;
1956 	}
1957 	return rc;
1958 }
1959 
1960 static int __emulate_int_real(struct x86_emulate_ctxt *ctxt, int irq)
1961 {
1962 	const struct x86_emulate_ops *ops = ctxt->ops;
1963 	int rc;
1964 	struct desc_ptr dt;
1965 	gva_t cs_addr;
1966 	gva_t eip_addr;
1967 	u16 cs, eip;
1968 
1969 	/* TODO: Add limit checks */
1970 	ctxt->src.val = ctxt->eflags;
1971 	rc = em_push(ctxt);
1972 	if (rc != X86EMUL_CONTINUE)
1973 		return rc;
1974 
1975 	ctxt->eflags &= ~(X86_EFLAGS_IF | X86_EFLAGS_TF | X86_EFLAGS_AC);
1976 
1977 	ctxt->src.val = get_segment_selector(ctxt, VCPU_SREG_CS);
1978 	rc = em_push(ctxt);
1979 	if (rc != X86EMUL_CONTINUE)
1980 		return rc;
1981 
1982 	ctxt->src.val = ctxt->_eip;
1983 	rc = em_push(ctxt);
1984 	if (rc != X86EMUL_CONTINUE)
1985 		return rc;
1986 
1987 	ops->get_idt(ctxt, &dt);
1988 
1989 	eip_addr = dt.address + (irq << 2);
1990 	cs_addr = dt.address + (irq << 2) + 2;
1991 
1992 	rc = ops->read_std(ctxt, cs_addr, &cs, 2, &ctxt->exception);
1993 	if (rc != X86EMUL_CONTINUE)
1994 		return rc;
1995 
1996 	rc = ops->read_std(ctxt, eip_addr, &eip, 2, &ctxt->exception);
1997 	if (rc != X86EMUL_CONTINUE)
1998 		return rc;
1999 
2000 	rc = load_segment_descriptor(ctxt, cs, VCPU_SREG_CS);
2001 	if (rc != X86EMUL_CONTINUE)
2002 		return rc;
2003 
2004 	ctxt->_eip = eip;
2005 
2006 	return rc;
2007 }
2008 
2009 int emulate_int_real(struct x86_emulate_ctxt *ctxt, int irq)
2010 {
2011 	int rc;
2012 
2013 	invalidate_registers(ctxt);
2014 	rc = __emulate_int_real(ctxt, irq);
2015 	if (rc == X86EMUL_CONTINUE)
2016 		writeback_registers(ctxt);
2017 	return rc;
2018 }
2019 
2020 static int emulate_int(struct x86_emulate_ctxt *ctxt, int irq)
2021 {
2022 	switch(ctxt->mode) {
2023 	case X86EMUL_MODE_REAL:
2024 		return __emulate_int_real(ctxt, irq);
2025 	case X86EMUL_MODE_VM86:
2026 	case X86EMUL_MODE_PROT16:
2027 	case X86EMUL_MODE_PROT32:
2028 	case X86EMUL_MODE_PROT64:
2029 	default:
2030 		/* Protected mode interrupts unimplemented yet */
2031 		return X86EMUL_UNHANDLEABLE;
2032 	}
2033 }
2034 
2035 static int emulate_iret_real(struct x86_emulate_ctxt *ctxt)
2036 {
2037 	int rc = X86EMUL_CONTINUE;
2038 	unsigned long temp_eip = 0;
2039 	unsigned long temp_eflags = 0;
2040 	unsigned long cs = 0;
2041 	unsigned long mask = X86_EFLAGS_CF | X86_EFLAGS_PF | X86_EFLAGS_AF |
2042 			     X86_EFLAGS_ZF | X86_EFLAGS_SF | X86_EFLAGS_TF |
2043 			     X86_EFLAGS_IF | X86_EFLAGS_DF | X86_EFLAGS_OF |
2044 			     X86_EFLAGS_IOPL | X86_EFLAGS_NT | X86_EFLAGS_RF |
2045 			     X86_EFLAGS_AC | X86_EFLAGS_ID |
2046 			     X86_EFLAGS_FIXED;
2047 	unsigned long vm86_mask = X86_EFLAGS_VM | X86_EFLAGS_VIF |
2048 				  X86_EFLAGS_VIP;
2049 
2050 	/* TODO: Add stack limit check */
2051 
2052 	rc = emulate_pop(ctxt, &temp_eip, ctxt->op_bytes);
2053 
2054 	if (rc != X86EMUL_CONTINUE)
2055 		return rc;
2056 
2057 	if (temp_eip & ~0xffff)
2058 		return emulate_gp(ctxt, 0);
2059 
2060 	rc = emulate_pop(ctxt, &cs, ctxt->op_bytes);
2061 
2062 	if (rc != X86EMUL_CONTINUE)
2063 		return rc;
2064 
2065 	rc = emulate_pop(ctxt, &temp_eflags, ctxt->op_bytes);
2066 
2067 	if (rc != X86EMUL_CONTINUE)
2068 		return rc;
2069 
2070 	rc = load_segment_descriptor(ctxt, (u16)cs, VCPU_SREG_CS);
2071 
2072 	if (rc != X86EMUL_CONTINUE)
2073 		return rc;
2074 
2075 	ctxt->_eip = temp_eip;
2076 
2077 	if (ctxt->op_bytes == 4)
2078 		ctxt->eflags = ((temp_eflags & mask) | (ctxt->eflags & vm86_mask));
2079 	else if (ctxt->op_bytes == 2) {
2080 		ctxt->eflags &= ~0xffff;
2081 		ctxt->eflags |= temp_eflags;
2082 	}
2083 
2084 	ctxt->eflags &= ~EFLG_RESERVED_ZEROS_MASK; /* Clear reserved zeros */
2085 	ctxt->eflags |= X86_EFLAGS_FIXED;
2086 	ctxt->ops->set_nmi_mask(ctxt, false);
2087 
2088 	return rc;
2089 }
2090 
2091 static int em_iret(struct x86_emulate_ctxt *ctxt)
2092 {
2093 	switch(ctxt->mode) {
2094 	case X86EMUL_MODE_REAL:
2095 		return emulate_iret_real(ctxt);
2096 	case X86EMUL_MODE_VM86:
2097 	case X86EMUL_MODE_PROT16:
2098 	case X86EMUL_MODE_PROT32:
2099 	case X86EMUL_MODE_PROT64:
2100 	default:
2101 		/* iret from protected mode unimplemented yet */
2102 		return X86EMUL_UNHANDLEABLE;
2103 	}
2104 }
2105 
2106 static int em_jmp_far(struct x86_emulate_ctxt *ctxt)
2107 {
2108 	int rc;
2109 	unsigned short sel, old_sel;
2110 	struct desc_struct old_desc, new_desc;
2111 	const struct x86_emulate_ops *ops = ctxt->ops;
2112 	u8 cpl = ctxt->ops->cpl(ctxt);
2113 
2114 	/* Assignment of RIP may only fail in 64-bit mode */
2115 	if (ctxt->mode == X86EMUL_MODE_PROT64)
2116 		ops->get_segment(ctxt, &old_sel, &old_desc, NULL,
2117 				 VCPU_SREG_CS);
2118 
2119 	memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
2120 
2121 	rc = __load_segment_descriptor(ctxt, sel, VCPU_SREG_CS, cpl,
2122 				       X86_TRANSFER_CALL_JMP,
2123 				       &new_desc);
2124 	if (rc != X86EMUL_CONTINUE)
2125 		return rc;
2126 
2127 	rc = assign_eip_far(ctxt, ctxt->src.val, &new_desc);
2128 	if (rc != X86EMUL_CONTINUE) {
2129 		WARN_ON(ctxt->mode != X86EMUL_MODE_PROT64);
2130 		/* assigning eip failed; restore the old cs */
2131 		ops->set_segment(ctxt, old_sel, &old_desc, 0, VCPU_SREG_CS);
2132 		return rc;
2133 	}
2134 	return rc;
2135 }
2136 
2137 static int em_jmp_abs(struct x86_emulate_ctxt *ctxt)
2138 {
2139 	return assign_eip_near(ctxt, ctxt->src.val);
2140 }
2141 
2142 static int em_call_near_abs(struct x86_emulate_ctxt *ctxt)
2143 {
2144 	int rc;
2145 	long int old_eip;
2146 
2147 	old_eip = ctxt->_eip;
2148 	rc = assign_eip_near(ctxt, ctxt->src.val);
2149 	if (rc != X86EMUL_CONTINUE)
2150 		return rc;
2151 	ctxt->src.val = old_eip;
2152 	rc = em_push(ctxt);
2153 	return rc;
2154 }
2155 
2156 static int em_cmpxchg8b(struct x86_emulate_ctxt *ctxt)
2157 {
2158 	u64 old = ctxt->dst.orig_val64;
2159 
2160 	if (ctxt->dst.bytes == 16)
2161 		return X86EMUL_UNHANDLEABLE;
2162 
2163 	if (((u32) (old >> 0) != (u32) reg_read(ctxt, VCPU_REGS_RAX)) ||
2164 	    ((u32) (old >> 32) != (u32) reg_read(ctxt, VCPU_REGS_RDX))) {
2165 		*reg_write(ctxt, VCPU_REGS_RAX) = (u32) (old >> 0);
2166 		*reg_write(ctxt, VCPU_REGS_RDX) = (u32) (old >> 32);
2167 		ctxt->eflags &= ~X86_EFLAGS_ZF;
2168 	} else {
2169 		ctxt->dst.val64 = ((u64)reg_read(ctxt, VCPU_REGS_RCX) << 32) |
2170 			(u32) reg_read(ctxt, VCPU_REGS_RBX);
2171 
2172 		ctxt->eflags |= X86_EFLAGS_ZF;
2173 	}
2174 	return X86EMUL_CONTINUE;
2175 }
2176 
2177 static int em_ret(struct x86_emulate_ctxt *ctxt)
2178 {
2179 	int rc;
2180 	unsigned long eip;
2181 
2182 	rc = emulate_pop(ctxt, &eip, ctxt->op_bytes);
2183 	if (rc != X86EMUL_CONTINUE)
2184 		return rc;
2185 
2186 	return assign_eip_near(ctxt, eip);
2187 }
2188 
2189 static int em_ret_far(struct x86_emulate_ctxt *ctxt)
2190 {
2191 	int rc;
2192 	unsigned long eip, cs;
2193 	u16 old_cs;
2194 	int cpl = ctxt->ops->cpl(ctxt);
2195 	struct desc_struct old_desc, new_desc;
2196 	const struct x86_emulate_ops *ops = ctxt->ops;
2197 
2198 	if (ctxt->mode == X86EMUL_MODE_PROT64)
2199 		ops->get_segment(ctxt, &old_cs, &old_desc, NULL,
2200 				 VCPU_SREG_CS);
2201 
2202 	rc = emulate_pop(ctxt, &eip, ctxt->op_bytes);
2203 	if (rc != X86EMUL_CONTINUE)
2204 		return rc;
2205 	rc = emulate_pop(ctxt, &cs, ctxt->op_bytes);
2206 	if (rc != X86EMUL_CONTINUE)
2207 		return rc;
2208 	/* Outer-privilege level return is not implemented */
2209 	if (ctxt->mode >= X86EMUL_MODE_PROT16 && (cs & 3) > cpl)
2210 		return X86EMUL_UNHANDLEABLE;
2211 	rc = __load_segment_descriptor(ctxt, (u16)cs, VCPU_SREG_CS, cpl,
2212 				       X86_TRANSFER_RET,
2213 				       &new_desc);
2214 	if (rc != X86EMUL_CONTINUE)
2215 		return rc;
2216 	rc = assign_eip_far(ctxt, eip, &new_desc);
2217 	if (rc != X86EMUL_CONTINUE) {
2218 		WARN_ON(ctxt->mode != X86EMUL_MODE_PROT64);
2219 		ops->set_segment(ctxt, old_cs, &old_desc, 0, VCPU_SREG_CS);
2220 	}
2221 	return rc;
2222 }
2223 
2224 static int em_ret_far_imm(struct x86_emulate_ctxt *ctxt)
2225 {
2226         int rc;
2227 
2228         rc = em_ret_far(ctxt);
2229         if (rc != X86EMUL_CONTINUE)
2230                 return rc;
2231         rsp_increment(ctxt, ctxt->src.val);
2232         return X86EMUL_CONTINUE;
2233 }
2234 
2235 static int em_cmpxchg(struct x86_emulate_ctxt *ctxt)
2236 {
2237 	/* Save real source value, then compare EAX against destination. */
2238 	ctxt->dst.orig_val = ctxt->dst.val;
2239 	ctxt->dst.val = reg_read(ctxt, VCPU_REGS_RAX);
2240 	ctxt->src.orig_val = ctxt->src.val;
2241 	ctxt->src.val = ctxt->dst.orig_val;
2242 	fastop(ctxt, em_cmp);
2243 
2244 	if (ctxt->eflags & X86_EFLAGS_ZF) {
2245 		/* Success: write back to memory; no update of EAX */
2246 		ctxt->src.type = OP_NONE;
2247 		ctxt->dst.val = ctxt->src.orig_val;
2248 	} else {
2249 		/* Failure: write the value we saw to EAX. */
2250 		ctxt->src.type = OP_REG;
2251 		ctxt->src.addr.reg = reg_rmw(ctxt, VCPU_REGS_RAX);
2252 		ctxt->src.val = ctxt->dst.orig_val;
2253 		/* Create write-cycle to dest by writing the same value */
2254 		ctxt->dst.val = ctxt->dst.orig_val;
2255 	}
2256 	return X86EMUL_CONTINUE;
2257 }
2258 
2259 static int em_lseg(struct x86_emulate_ctxt *ctxt)
2260 {
2261 	int seg = ctxt->src2.val;
2262 	unsigned short sel;
2263 	int rc;
2264 
2265 	memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
2266 
2267 	rc = load_segment_descriptor(ctxt, sel, seg);
2268 	if (rc != X86EMUL_CONTINUE)
2269 		return rc;
2270 
2271 	ctxt->dst.val = ctxt->src.val;
2272 	return rc;
2273 }
2274 
2275 static int emulator_has_longmode(struct x86_emulate_ctxt *ctxt)
2276 {
2277 	u32 eax, ebx, ecx, edx;
2278 
2279 	eax = 0x80000001;
2280 	ecx = 0;
2281 	ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
2282 	return edx & bit(X86_FEATURE_LM);
2283 }
2284 
2285 #define GET_SMSTATE(type, smbase, offset)				  \
2286 	({								  \
2287 	 type __val;							  \
2288 	 int r = ctxt->ops->read_phys(ctxt, smbase + offset, &__val,      \
2289 				      sizeof(__val));			  \
2290 	 if (r != X86EMUL_CONTINUE)					  \
2291 		 return X86EMUL_UNHANDLEABLE;				  \
2292 	 __val;								  \
2293 	})
2294 
2295 static void rsm_set_desc_flags(struct desc_struct *desc, u32 flags)
2296 {
2297 	desc->g    = (flags >> 23) & 1;
2298 	desc->d    = (flags >> 22) & 1;
2299 	desc->l    = (flags >> 21) & 1;
2300 	desc->avl  = (flags >> 20) & 1;
2301 	desc->p    = (flags >> 15) & 1;
2302 	desc->dpl  = (flags >> 13) & 3;
2303 	desc->s    = (flags >> 12) & 1;
2304 	desc->type = (flags >>  8) & 15;
2305 }
2306 
2307 static int rsm_load_seg_32(struct x86_emulate_ctxt *ctxt, u64 smbase, int n)
2308 {
2309 	struct desc_struct desc;
2310 	int offset;
2311 	u16 selector;
2312 
2313 	selector = GET_SMSTATE(u32, smbase, 0x7fa8 + n * 4);
2314 
2315 	if (n < 3)
2316 		offset = 0x7f84 + n * 12;
2317 	else
2318 		offset = 0x7f2c + (n - 3) * 12;
2319 
2320 	set_desc_base(&desc,      GET_SMSTATE(u32, smbase, offset + 8));
2321 	set_desc_limit(&desc,     GET_SMSTATE(u32, smbase, offset + 4));
2322 	rsm_set_desc_flags(&desc, GET_SMSTATE(u32, smbase, offset));
2323 	ctxt->ops->set_segment(ctxt, selector, &desc, 0, n);
2324 	return X86EMUL_CONTINUE;
2325 }
2326 
2327 static int rsm_load_seg_64(struct x86_emulate_ctxt *ctxt, u64 smbase, int n)
2328 {
2329 	struct desc_struct desc;
2330 	int offset;
2331 	u16 selector;
2332 	u32 base3;
2333 
2334 	offset = 0x7e00 + n * 16;
2335 
2336 	selector =                GET_SMSTATE(u16, smbase, offset);
2337 	rsm_set_desc_flags(&desc, GET_SMSTATE(u16, smbase, offset + 2) << 8);
2338 	set_desc_limit(&desc,     GET_SMSTATE(u32, smbase, offset + 4));
2339 	set_desc_base(&desc,      GET_SMSTATE(u32, smbase, offset + 8));
2340 	base3 =                   GET_SMSTATE(u32, smbase, offset + 12);
2341 
2342 	ctxt->ops->set_segment(ctxt, selector, &desc, base3, n);
2343 	return X86EMUL_CONTINUE;
2344 }
2345 
2346 static int rsm_enter_protected_mode(struct x86_emulate_ctxt *ctxt,
2347 				     u64 cr0, u64 cr4)
2348 {
2349 	int bad;
2350 
2351 	/*
2352 	 * First enable PAE, long mode needs it before CR0.PG = 1 is set.
2353 	 * Then enable protected mode.	However, PCID cannot be enabled
2354 	 * if EFER.LMA=0, so set it separately.
2355 	 */
2356 	bad = ctxt->ops->set_cr(ctxt, 4, cr4 & ~X86_CR4_PCIDE);
2357 	if (bad)
2358 		return X86EMUL_UNHANDLEABLE;
2359 
2360 	bad = ctxt->ops->set_cr(ctxt, 0, cr0);
2361 	if (bad)
2362 		return X86EMUL_UNHANDLEABLE;
2363 
2364 	if (cr4 & X86_CR4_PCIDE) {
2365 		bad = ctxt->ops->set_cr(ctxt, 4, cr4);
2366 		if (bad)
2367 			return X86EMUL_UNHANDLEABLE;
2368 	}
2369 
2370 	return X86EMUL_CONTINUE;
2371 }
2372 
2373 static int rsm_load_state_32(struct x86_emulate_ctxt *ctxt, u64 smbase)
2374 {
2375 	struct desc_struct desc;
2376 	struct desc_ptr dt;
2377 	u16 selector;
2378 	u32 val, cr0, cr4;
2379 	int i;
2380 
2381 	cr0 =                      GET_SMSTATE(u32, smbase, 0x7ffc);
2382 	ctxt->ops->set_cr(ctxt, 3, GET_SMSTATE(u32, smbase, 0x7ff8));
2383 	ctxt->eflags =             GET_SMSTATE(u32, smbase, 0x7ff4) | X86_EFLAGS_FIXED;
2384 	ctxt->_eip =               GET_SMSTATE(u32, smbase, 0x7ff0);
2385 
2386 	for (i = 0; i < 8; i++)
2387 		*reg_write(ctxt, i) = GET_SMSTATE(u32, smbase, 0x7fd0 + i * 4);
2388 
2389 	val = GET_SMSTATE(u32, smbase, 0x7fcc);
2390 	ctxt->ops->set_dr(ctxt, 6, (val & DR6_VOLATILE) | DR6_FIXED_1);
2391 	val = GET_SMSTATE(u32, smbase, 0x7fc8);
2392 	ctxt->ops->set_dr(ctxt, 7, (val & DR7_VOLATILE) | DR7_FIXED_1);
2393 
2394 	selector =                 GET_SMSTATE(u32, smbase, 0x7fc4);
2395 	set_desc_base(&desc,       GET_SMSTATE(u32, smbase, 0x7f64));
2396 	set_desc_limit(&desc,      GET_SMSTATE(u32, smbase, 0x7f60));
2397 	rsm_set_desc_flags(&desc,  GET_SMSTATE(u32, smbase, 0x7f5c));
2398 	ctxt->ops->set_segment(ctxt, selector, &desc, 0, VCPU_SREG_TR);
2399 
2400 	selector =                 GET_SMSTATE(u32, smbase, 0x7fc0);
2401 	set_desc_base(&desc,       GET_SMSTATE(u32, smbase, 0x7f80));
2402 	set_desc_limit(&desc,      GET_SMSTATE(u32, smbase, 0x7f7c));
2403 	rsm_set_desc_flags(&desc,  GET_SMSTATE(u32, smbase, 0x7f78));
2404 	ctxt->ops->set_segment(ctxt, selector, &desc, 0, VCPU_SREG_LDTR);
2405 
2406 	dt.address =               GET_SMSTATE(u32, smbase, 0x7f74);
2407 	dt.size =                  GET_SMSTATE(u32, smbase, 0x7f70);
2408 	ctxt->ops->set_gdt(ctxt, &dt);
2409 
2410 	dt.address =               GET_SMSTATE(u32, smbase, 0x7f58);
2411 	dt.size =                  GET_SMSTATE(u32, smbase, 0x7f54);
2412 	ctxt->ops->set_idt(ctxt, &dt);
2413 
2414 	for (i = 0; i < 6; i++) {
2415 		int r = rsm_load_seg_32(ctxt, smbase, i);
2416 		if (r != X86EMUL_CONTINUE)
2417 			return r;
2418 	}
2419 
2420 	cr4 = GET_SMSTATE(u32, smbase, 0x7f14);
2421 
2422 	ctxt->ops->set_smbase(ctxt, GET_SMSTATE(u32, smbase, 0x7ef8));
2423 
2424 	return rsm_enter_protected_mode(ctxt, cr0, cr4);
2425 }
2426 
2427 static int rsm_load_state_64(struct x86_emulate_ctxt *ctxt, u64 smbase)
2428 {
2429 	struct desc_struct desc;
2430 	struct desc_ptr dt;
2431 	u64 val, cr0, cr4;
2432 	u32 base3;
2433 	u16 selector;
2434 	int i, r;
2435 
2436 	for (i = 0; i < 16; i++)
2437 		*reg_write(ctxt, i) = GET_SMSTATE(u64, smbase, 0x7ff8 - i * 8);
2438 
2439 	ctxt->_eip   = GET_SMSTATE(u64, smbase, 0x7f78);
2440 	ctxt->eflags = GET_SMSTATE(u32, smbase, 0x7f70) | X86_EFLAGS_FIXED;
2441 
2442 	val = GET_SMSTATE(u32, smbase, 0x7f68);
2443 	ctxt->ops->set_dr(ctxt, 6, (val & DR6_VOLATILE) | DR6_FIXED_1);
2444 	val = GET_SMSTATE(u32, smbase, 0x7f60);
2445 	ctxt->ops->set_dr(ctxt, 7, (val & DR7_VOLATILE) | DR7_FIXED_1);
2446 
2447 	cr0 =                       GET_SMSTATE(u64, smbase, 0x7f58);
2448 	ctxt->ops->set_cr(ctxt, 3,  GET_SMSTATE(u64, smbase, 0x7f50));
2449 	cr4 =                       GET_SMSTATE(u64, smbase, 0x7f48);
2450 	ctxt->ops->set_smbase(ctxt, GET_SMSTATE(u32, smbase, 0x7f00));
2451 	val =                       GET_SMSTATE(u64, smbase, 0x7ed0);
2452 	ctxt->ops->set_msr(ctxt, MSR_EFER, val & ~EFER_LMA);
2453 
2454 	selector =                  GET_SMSTATE(u32, smbase, 0x7e90);
2455 	rsm_set_desc_flags(&desc,   GET_SMSTATE(u32, smbase, 0x7e92) << 8);
2456 	set_desc_limit(&desc,       GET_SMSTATE(u32, smbase, 0x7e94));
2457 	set_desc_base(&desc,        GET_SMSTATE(u32, smbase, 0x7e98));
2458 	base3 =                     GET_SMSTATE(u32, smbase, 0x7e9c);
2459 	ctxt->ops->set_segment(ctxt, selector, &desc, base3, VCPU_SREG_TR);
2460 
2461 	dt.size =                   GET_SMSTATE(u32, smbase, 0x7e84);
2462 	dt.address =                GET_SMSTATE(u64, smbase, 0x7e88);
2463 	ctxt->ops->set_idt(ctxt, &dt);
2464 
2465 	selector =                  GET_SMSTATE(u32, smbase, 0x7e70);
2466 	rsm_set_desc_flags(&desc,   GET_SMSTATE(u32, smbase, 0x7e72) << 8);
2467 	set_desc_limit(&desc,       GET_SMSTATE(u32, smbase, 0x7e74));
2468 	set_desc_base(&desc,        GET_SMSTATE(u32, smbase, 0x7e78));
2469 	base3 =                     GET_SMSTATE(u32, smbase, 0x7e7c);
2470 	ctxt->ops->set_segment(ctxt, selector, &desc, base3, VCPU_SREG_LDTR);
2471 
2472 	dt.size =                   GET_SMSTATE(u32, smbase, 0x7e64);
2473 	dt.address =                GET_SMSTATE(u64, smbase, 0x7e68);
2474 	ctxt->ops->set_gdt(ctxt, &dt);
2475 
2476 	r = rsm_enter_protected_mode(ctxt, cr0, cr4);
2477 	if (r != X86EMUL_CONTINUE)
2478 		return r;
2479 
2480 	for (i = 0; i < 6; i++) {
2481 		r = rsm_load_seg_64(ctxt, smbase, i);
2482 		if (r != X86EMUL_CONTINUE)
2483 			return r;
2484 	}
2485 
2486 	return X86EMUL_CONTINUE;
2487 }
2488 
2489 static int em_rsm(struct x86_emulate_ctxt *ctxt)
2490 {
2491 	unsigned long cr0, cr4, efer;
2492 	u64 smbase;
2493 	int ret;
2494 
2495 	if ((ctxt->emul_flags & X86EMUL_SMM_MASK) == 0)
2496 		return emulate_ud(ctxt);
2497 
2498 	/*
2499 	 * Get back to real mode, to prepare a safe state in which to load
2500 	 * CR0/CR3/CR4/EFER.  It's all a bit more complicated if the vCPU
2501 	 * supports long mode.
2502 	 */
2503 	cr4 = ctxt->ops->get_cr(ctxt, 4);
2504 	if (emulator_has_longmode(ctxt)) {
2505 		struct desc_struct cs_desc;
2506 
2507 		/* Zero CR4.PCIDE before CR0.PG.  */
2508 		if (cr4 & X86_CR4_PCIDE) {
2509 			ctxt->ops->set_cr(ctxt, 4, cr4 & ~X86_CR4_PCIDE);
2510 			cr4 &= ~X86_CR4_PCIDE;
2511 		}
2512 
2513 		/* A 32-bit code segment is required to clear EFER.LMA.  */
2514 		memset(&cs_desc, 0, sizeof(cs_desc));
2515 		cs_desc.type = 0xb;
2516 		cs_desc.s = cs_desc.g = cs_desc.p = 1;
2517 		ctxt->ops->set_segment(ctxt, 0, &cs_desc, 0, VCPU_SREG_CS);
2518 	}
2519 
2520 	/* For the 64-bit case, this will clear EFER.LMA.  */
2521 	cr0 = ctxt->ops->get_cr(ctxt, 0);
2522 	if (cr0 & X86_CR0_PE)
2523 		ctxt->ops->set_cr(ctxt, 0, cr0 & ~(X86_CR0_PG | X86_CR0_PE));
2524 
2525 	/* Now clear CR4.PAE (which must be done before clearing EFER.LME).  */
2526 	if (cr4 & X86_CR4_PAE)
2527 		ctxt->ops->set_cr(ctxt, 4, cr4 & ~X86_CR4_PAE);
2528 
2529 	/* And finally go back to 32-bit mode.  */
2530 	efer = 0;
2531 	ctxt->ops->set_msr(ctxt, MSR_EFER, efer);
2532 
2533 	smbase = ctxt->ops->get_smbase(ctxt);
2534 	if (emulator_has_longmode(ctxt))
2535 		ret = rsm_load_state_64(ctxt, smbase + 0x8000);
2536 	else
2537 		ret = rsm_load_state_32(ctxt, smbase + 0x8000);
2538 
2539 	if (ret != X86EMUL_CONTINUE) {
2540 		/* FIXME: should triple fault */
2541 		return X86EMUL_UNHANDLEABLE;
2542 	}
2543 
2544 	if ((ctxt->emul_flags & X86EMUL_SMM_INSIDE_NMI_MASK) == 0)
2545 		ctxt->ops->set_nmi_mask(ctxt, false);
2546 
2547 	ctxt->emul_flags &= ~X86EMUL_SMM_INSIDE_NMI_MASK;
2548 	ctxt->emul_flags &= ~X86EMUL_SMM_MASK;
2549 	return X86EMUL_CONTINUE;
2550 }
2551 
2552 static void
2553 setup_syscalls_segments(struct x86_emulate_ctxt *ctxt,
2554 			struct desc_struct *cs, struct desc_struct *ss)
2555 {
2556 	cs->l = 0;		/* will be adjusted later */
2557 	set_desc_base(cs, 0);	/* flat segment */
2558 	cs->g = 1;		/* 4kb granularity */
2559 	set_desc_limit(cs, 0xfffff);	/* 4GB limit */
2560 	cs->type = 0x0b;	/* Read, Execute, Accessed */
2561 	cs->s = 1;
2562 	cs->dpl = 0;		/* will be adjusted later */
2563 	cs->p = 1;
2564 	cs->d = 1;
2565 	cs->avl = 0;
2566 
2567 	set_desc_base(ss, 0);	/* flat segment */
2568 	set_desc_limit(ss, 0xfffff);	/* 4GB limit */
2569 	ss->g = 1;		/* 4kb granularity */
2570 	ss->s = 1;
2571 	ss->type = 0x03;	/* Read/Write, Accessed */
2572 	ss->d = 1;		/* 32bit stack segment */
2573 	ss->dpl = 0;
2574 	ss->p = 1;
2575 	ss->l = 0;
2576 	ss->avl = 0;
2577 }
2578 
2579 static bool vendor_intel(struct x86_emulate_ctxt *ctxt)
2580 {
2581 	u32 eax, ebx, ecx, edx;
2582 
2583 	eax = ecx = 0;
2584 	ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
2585 	return ebx == X86EMUL_CPUID_VENDOR_GenuineIntel_ebx
2586 		&& ecx == X86EMUL_CPUID_VENDOR_GenuineIntel_ecx
2587 		&& edx == X86EMUL_CPUID_VENDOR_GenuineIntel_edx;
2588 }
2589 
2590 static bool em_syscall_is_enabled(struct x86_emulate_ctxt *ctxt)
2591 {
2592 	const struct x86_emulate_ops *ops = ctxt->ops;
2593 	u32 eax, ebx, ecx, edx;
2594 
2595 	/*
2596 	 * syscall should always be enabled in longmode - so only become
2597 	 * vendor specific (cpuid) if other modes are active...
2598 	 */
2599 	if (ctxt->mode == X86EMUL_MODE_PROT64)
2600 		return true;
2601 
2602 	eax = 0x00000000;
2603 	ecx = 0x00000000;
2604 	ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
2605 	/*
2606 	 * Intel ("GenuineIntel")
2607 	 * remark: Intel CPUs only support "syscall" in 64bit
2608 	 * longmode. Also an 64bit guest with a
2609 	 * 32bit compat-app running will #UD !! While this
2610 	 * behaviour can be fixed (by emulating) into AMD
2611 	 * response - CPUs of AMD can't behave like Intel.
2612 	 */
2613 	if (ebx == X86EMUL_CPUID_VENDOR_GenuineIntel_ebx &&
2614 	    ecx == X86EMUL_CPUID_VENDOR_GenuineIntel_ecx &&
2615 	    edx == X86EMUL_CPUID_VENDOR_GenuineIntel_edx)
2616 		return false;
2617 
2618 	/* AMD ("AuthenticAMD") */
2619 	if (ebx == X86EMUL_CPUID_VENDOR_AuthenticAMD_ebx &&
2620 	    ecx == X86EMUL_CPUID_VENDOR_AuthenticAMD_ecx &&
2621 	    edx == X86EMUL_CPUID_VENDOR_AuthenticAMD_edx)
2622 		return true;
2623 
2624 	/* AMD ("AMDisbetter!") */
2625 	if (ebx == X86EMUL_CPUID_VENDOR_AMDisbetterI_ebx &&
2626 	    ecx == X86EMUL_CPUID_VENDOR_AMDisbetterI_ecx &&
2627 	    edx == X86EMUL_CPUID_VENDOR_AMDisbetterI_edx)
2628 		return true;
2629 
2630 	/* default: (not Intel, not AMD), apply Intel's stricter rules... */
2631 	return false;
2632 }
2633 
2634 static int em_syscall(struct x86_emulate_ctxt *ctxt)
2635 {
2636 	const struct x86_emulate_ops *ops = ctxt->ops;
2637 	struct desc_struct cs, ss;
2638 	u64 msr_data;
2639 	u16 cs_sel, ss_sel;
2640 	u64 efer = 0;
2641 
2642 	/* syscall is not available in real mode */
2643 	if (ctxt->mode == X86EMUL_MODE_REAL ||
2644 	    ctxt->mode == X86EMUL_MODE_VM86)
2645 		return emulate_ud(ctxt);
2646 
2647 	if (!(em_syscall_is_enabled(ctxt)))
2648 		return emulate_ud(ctxt);
2649 
2650 	ops->get_msr(ctxt, MSR_EFER, &efer);
2651 	setup_syscalls_segments(ctxt, &cs, &ss);
2652 
2653 	if (!(efer & EFER_SCE))
2654 		return emulate_ud(ctxt);
2655 
2656 	ops->get_msr(ctxt, MSR_STAR, &msr_data);
2657 	msr_data >>= 32;
2658 	cs_sel = (u16)(msr_data & 0xfffc);
2659 	ss_sel = (u16)(msr_data + 8);
2660 
2661 	if (efer & EFER_LMA) {
2662 		cs.d = 0;
2663 		cs.l = 1;
2664 	}
2665 	ops->set_segment(ctxt, cs_sel, &cs, 0, VCPU_SREG_CS);
2666 	ops->set_segment(ctxt, ss_sel, &ss, 0, VCPU_SREG_SS);
2667 
2668 	*reg_write(ctxt, VCPU_REGS_RCX) = ctxt->_eip;
2669 	if (efer & EFER_LMA) {
2670 #ifdef CONFIG_X86_64
2671 		*reg_write(ctxt, VCPU_REGS_R11) = ctxt->eflags;
2672 
2673 		ops->get_msr(ctxt,
2674 			     ctxt->mode == X86EMUL_MODE_PROT64 ?
2675 			     MSR_LSTAR : MSR_CSTAR, &msr_data);
2676 		ctxt->_eip = msr_data;
2677 
2678 		ops->get_msr(ctxt, MSR_SYSCALL_MASK, &msr_data);
2679 		ctxt->eflags &= ~msr_data;
2680 		ctxt->eflags |= X86_EFLAGS_FIXED;
2681 #endif
2682 	} else {
2683 		/* legacy mode */
2684 		ops->get_msr(ctxt, MSR_STAR, &msr_data);
2685 		ctxt->_eip = (u32)msr_data;
2686 
2687 		ctxt->eflags &= ~(X86_EFLAGS_VM | X86_EFLAGS_IF);
2688 	}
2689 
2690 	return X86EMUL_CONTINUE;
2691 }
2692 
2693 static int em_sysenter(struct x86_emulate_ctxt *ctxt)
2694 {
2695 	const struct x86_emulate_ops *ops = ctxt->ops;
2696 	struct desc_struct cs, ss;
2697 	u64 msr_data;
2698 	u16 cs_sel, ss_sel;
2699 	u64 efer = 0;
2700 
2701 	ops->get_msr(ctxt, MSR_EFER, &efer);
2702 	/* inject #GP if in real mode */
2703 	if (ctxt->mode == X86EMUL_MODE_REAL)
2704 		return emulate_gp(ctxt, 0);
2705 
2706 	/*
2707 	 * Not recognized on AMD in compat mode (but is recognized in legacy
2708 	 * mode).
2709 	 */
2710 	if ((ctxt->mode != X86EMUL_MODE_PROT64) && (efer & EFER_LMA)
2711 	    && !vendor_intel(ctxt))
2712 		return emulate_ud(ctxt);
2713 
2714 	/* sysenter/sysexit have not been tested in 64bit mode. */
2715 	if (ctxt->mode == X86EMUL_MODE_PROT64)
2716 		return X86EMUL_UNHANDLEABLE;
2717 
2718 	setup_syscalls_segments(ctxt, &cs, &ss);
2719 
2720 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_CS, &msr_data);
2721 	if ((msr_data & 0xfffc) == 0x0)
2722 		return emulate_gp(ctxt, 0);
2723 
2724 	ctxt->eflags &= ~(X86_EFLAGS_VM | X86_EFLAGS_IF);
2725 	cs_sel = (u16)msr_data & ~SEGMENT_RPL_MASK;
2726 	ss_sel = cs_sel + 8;
2727 	if (efer & EFER_LMA) {
2728 		cs.d = 0;
2729 		cs.l = 1;
2730 	}
2731 
2732 	ops->set_segment(ctxt, cs_sel, &cs, 0, VCPU_SREG_CS);
2733 	ops->set_segment(ctxt, ss_sel, &ss, 0, VCPU_SREG_SS);
2734 
2735 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_EIP, &msr_data);
2736 	ctxt->_eip = (efer & EFER_LMA) ? msr_data : (u32)msr_data;
2737 
2738 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_ESP, &msr_data);
2739 	*reg_write(ctxt, VCPU_REGS_RSP) = (efer & EFER_LMA) ? msr_data :
2740 							      (u32)msr_data;
2741 
2742 	return X86EMUL_CONTINUE;
2743 }
2744 
2745 static int em_sysexit(struct x86_emulate_ctxt *ctxt)
2746 {
2747 	const struct x86_emulate_ops *ops = ctxt->ops;
2748 	struct desc_struct cs, ss;
2749 	u64 msr_data, rcx, rdx;
2750 	int usermode;
2751 	u16 cs_sel = 0, ss_sel = 0;
2752 
2753 	/* inject #GP if in real mode or Virtual 8086 mode */
2754 	if (ctxt->mode == X86EMUL_MODE_REAL ||
2755 	    ctxt->mode == X86EMUL_MODE_VM86)
2756 		return emulate_gp(ctxt, 0);
2757 
2758 	setup_syscalls_segments(ctxt, &cs, &ss);
2759 
2760 	if ((ctxt->rex_prefix & 0x8) != 0x0)
2761 		usermode = X86EMUL_MODE_PROT64;
2762 	else
2763 		usermode = X86EMUL_MODE_PROT32;
2764 
2765 	rcx = reg_read(ctxt, VCPU_REGS_RCX);
2766 	rdx = reg_read(ctxt, VCPU_REGS_RDX);
2767 
2768 	cs.dpl = 3;
2769 	ss.dpl = 3;
2770 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_CS, &msr_data);
2771 	switch (usermode) {
2772 	case X86EMUL_MODE_PROT32:
2773 		cs_sel = (u16)(msr_data + 16);
2774 		if ((msr_data & 0xfffc) == 0x0)
2775 			return emulate_gp(ctxt, 0);
2776 		ss_sel = (u16)(msr_data + 24);
2777 		rcx = (u32)rcx;
2778 		rdx = (u32)rdx;
2779 		break;
2780 	case X86EMUL_MODE_PROT64:
2781 		cs_sel = (u16)(msr_data + 32);
2782 		if (msr_data == 0x0)
2783 			return emulate_gp(ctxt, 0);
2784 		ss_sel = cs_sel + 8;
2785 		cs.d = 0;
2786 		cs.l = 1;
2787 		if (is_noncanonical_address(rcx) ||
2788 		    is_noncanonical_address(rdx))
2789 			return emulate_gp(ctxt, 0);
2790 		break;
2791 	}
2792 	cs_sel |= SEGMENT_RPL_MASK;
2793 	ss_sel |= SEGMENT_RPL_MASK;
2794 
2795 	ops->set_segment(ctxt, cs_sel, &cs, 0, VCPU_SREG_CS);
2796 	ops->set_segment(ctxt, ss_sel, &ss, 0, VCPU_SREG_SS);
2797 
2798 	ctxt->_eip = rdx;
2799 	*reg_write(ctxt, VCPU_REGS_RSP) = rcx;
2800 
2801 	return X86EMUL_CONTINUE;
2802 }
2803 
2804 static bool emulator_bad_iopl(struct x86_emulate_ctxt *ctxt)
2805 {
2806 	int iopl;
2807 	if (ctxt->mode == X86EMUL_MODE_REAL)
2808 		return false;
2809 	if (ctxt->mode == X86EMUL_MODE_VM86)
2810 		return true;
2811 	iopl = (ctxt->eflags & X86_EFLAGS_IOPL) >> X86_EFLAGS_IOPL_BIT;
2812 	return ctxt->ops->cpl(ctxt) > iopl;
2813 }
2814 
2815 static bool emulator_io_port_access_allowed(struct x86_emulate_ctxt *ctxt,
2816 					    u16 port, u16 len)
2817 {
2818 	const struct x86_emulate_ops *ops = ctxt->ops;
2819 	struct desc_struct tr_seg;
2820 	u32 base3;
2821 	int r;
2822 	u16 tr, io_bitmap_ptr, perm, bit_idx = port & 0x7;
2823 	unsigned mask = (1 << len) - 1;
2824 	unsigned long base;
2825 
2826 	ops->get_segment(ctxt, &tr, &tr_seg, &base3, VCPU_SREG_TR);
2827 	if (!tr_seg.p)
2828 		return false;
2829 	if (desc_limit_scaled(&tr_seg) < 103)
2830 		return false;
2831 	base = get_desc_base(&tr_seg);
2832 #ifdef CONFIG_X86_64
2833 	base |= ((u64)base3) << 32;
2834 #endif
2835 	r = ops->read_std(ctxt, base + 102, &io_bitmap_ptr, 2, NULL);
2836 	if (r != X86EMUL_CONTINUE)
2837 		return false;
2838 	if (io_bitmap_ptr + port/8 > desc_limit_scaled(&tr_seg))
2839 		return false;
2840 	r = ops->read_std(ctxt, base + io_bitmap_ptr + port/8, &perm, 2, NULL);
2841 	if (r != X86EMUL_CONTINUE)
2842 		return false;
2843 	if ((perm >> bit_idx) & mask)
2844 		return false;
2845 	return true;
2846 }
2847 
2848 static bool emulator_io_permited(struct x86_emulate_ctxt *ctxt,
2849 				 u16 port, u16 len)
2850 {
2851 	if (ctxt->perm_ok)
2852 		return true;
2853 
2854 	if (emulator_bad_iopl(ctxt))
2855 		if (!emulator_io_port_access_allowed(ctxt, port, len))
2856 			return false;
2857 
2858 	ctxt->perm_ok = true;
2859 
2860 	return true;
2861 }
2862 
2863 static void string_registers_quirk(struct x86_emulate_ctxt *ctxt)
2864 {
2865 	/*
2866 	 * Intel CPUs mask the counter and pointers in quite strange
2867 	 * manner when ECX is zero due to REP-string optimizations.
2868 	 */
2869 #ifdef CONFIG_X86_64
2870 	if (ctxt->ad_bytes != 4 || !vendor_intel(ctxt))
2871 		return;
2872 
2873 	*reg_write(ctxt, VCPU_REGS_RCX) = 0;
2874 
2875 	switch (ctxt->b) {
2876 	case 0xa4:	/* movsb */
2877 	case 0xa5:	/* movsd/w */
2878 		*reg_rmw(ctxt, VCPU_REGS_RSI) &= (u32)-1;
2879 		/* fall through */
2880 	case 0xaa:	/* stosb */
2881 	case 0xab:	/* stosd/w */
2882 		*reg_rmw(ctxt, VCPU_REGS_RDI) &= (u32)-1;
2883 	}
2884 #endif
2885 }
2886 
2887 static void save_state_to_tss16(struct x86_emulate_ctxt *ctxt,
2888 				struct tss_segment_16 *tss)
2889 {
2890 	tss->ip = ctxt->_eip;
2891 	tss->flag = ctxt->eflags;
2892 	tss->ax = reg_read(ctxt, VCPU_REGS_RAX);
2893 	tss->cx = reg_read(ctxt, VCPU_REGS_RCX);
2894 	tss->dx = reg_read(ctxt, VCPU_REGS_RDX);
2895 	tss->bx = reg_read(ctxt, VCPU_REGS_RBX);
2896 	tss->sp = reg_read(ctxt, VCPU_REGS_RSP);
2897 	tss->bp = reg_read(ctxt, VCPU_REGS_RBP);
2898 	tss->si = reg_read(ctxt, VCPU_REGS_RSI);
2899 	tss->di = reg_read(ctxt, VCPU_REGS_RDI);
2900 
2901 	tss->es = get_segment_selector(ctxt, VCPU_SREG_ES);
2902 	tss->cs = get_segment_selector(ctxt, VCPU_SREG_CS);
2903 	tss->ss = get_segment_selector(ctxt, VCPU_SREG_SS);
2904 	tss->ds = get_segment_selector(ctxt, VCPU_SREG_DS);
2905 	tss->ldt = get_segment_selector(ctxt, VCPU_SREG_LDTR);
2906 }
2907 
2908 static int load_state_from_tss16(struct x86_emulate_ctxt *ctxt,
2909 				 struct tss_segment_16 *tss)
2910 {
2911 	int ret;
2912 	u8 cpl;
2913 
2914 	ctxt->_eip = tss->ip;
2915 	ctxt->eflags = tss->flag | 2;
2916 	*reg_write(ctxt, VCPU_REGS_RAX) = tss->ax;
2917 	*reg_write(ctxt, VCPU_REGS_RCX) = tss->cx;
2918 	*reg_write(ctxt, VCPU_REGS_RDX) = tss->dx;
2919 	*reg_write(ctxt, VCPU_REGS_RBX) = tss->bx;
2920 	*reg_write(ctxt, VCPU_REGS_RSP) = tss->sp;
2921 	*reg_write(ctxt, VCPU_REGS_RBP) = tss->bp;
2922 	*reg_write(ctxt, VCPU_REGS_RSI) = tss->si;
2923 	*reg_write(ctxt, VCPU_REGS_RDI) = tss->di;
2924 
2925 	/*
2926 	 * SDM says that segment selectors are loaded before segment
2927 	 * descriptors
2928 	 */
2929 	set_segment_selector(ctxt, tss->ldt, VCPU_SREG_LDTR);
2930 	set_segment_selector(ctxt, tss->es, VCPU_SREG_ES);
2931 	set_segment_selector(ctxt, tss->cs, VCPU_SREG_CS);
2932 	set_segment_selector(ctxt, tss->ss, VCPU_SREG_SS);
2933 	set_segment_selector(ctxt, tss->ds, VCPU_SREG_DS);
2934 
2935 	cpl = tss->cs & 3;
2936 
2937 	/*
2938 	 * Now load segment descriptors. If fault happens at this stage
2939 	 * it is handled in a context of new task
2940 	 */
2941 	ret = __load_segment_descriptor(ctxt, tss->ldt, VCPU_SREG_LDTR, cpl,
2942 					X86_TRANSFER_TASK_SWITCH, NULL);
2943 	if (ret != X86EMUL_CONTINUE)
2944 		return ret;
2945 	ret = __load_segment_descriptor(ctxt, tss->es, VCPU_SREG_ES, cpl,
2946 					X86_TRANSFER_TASK_SWITCH, NULL);
2947 	if (ret != X86EMUL_CONTINUE)
2948 		return ret;
2949 	ret = __load_segment_descriptor(ctxt, tss->cs, VCPU_SREG_CS, cpl,
2950 					X86_TRANSFER_TASK_SWITCH, NULL);
2951 	if (ret != X86EMUL_CONTINUE)
2952 		return ret;
2953 	ret = __load_segment_descriptor(ctxt, tss->ss, VCPU_SREG_SS, cpl,
2954 					X86_TRANSFER_TASK_SWITCH, NULL);
2955 	if (ret != X86EMUL_CONTINUE)
2956 		return ret;
2957 	ret = __load_segment_descriptor(ctxt, tss->ds, VCPU_SREG_DS, cpl,
2958 					X86_TRANSFER_TASK_SWITCH, NULL);
2959 	if (ret != X86EMUL_CONTINUE)
2960 		return ret;
2961 
2962 	return X86EMUL_CONTINUE;
2963 }
2964 
2965 static int task_switch_16(struct x86_emulate_ctxt *ctxt,
2966 			  u16 tss_selector, u16 old_tss_sel,
2967 			  ulong old_tss_base, struct desc_struct *new_desc)
2968 {
2969 	const struct x86_emulate_ops *ops = ctxt->ops;
2970 	struct tss_segment_16 tss_seg;
2971 	int ret;
2972 	u32 new_tss_base = get_desc_base(new_desc);
2973 
2974 	ret = ops->read_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2975 			    &ctxt->exception);
2976 	if (ret != X86EMUL_CONTINUE)
2977 		return ret;
2978 
2979 	save_state_to_tss16(ctxt, &tss_seg);
2980 
2981 	ret = ops->write_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2982 			     &ctxt->exception);
2983 	if (ret != X86EMUL_CONTINUE)
2984 		return ret;
2985 
2986 	ret = ops->read_std(ctxt, new_tss_base, &tss_seg, sizeof tss_seg,
2987 			    &ctxt->exception);
2988 	if (ret != X86EMUL_CONTINUE)
2989 		return ret;
2990 
2991 	if (old_tss_sel != 0xffff) {
2992 		tss_seg.prev_task_link = old_tss_sel;
2993 
2994 		ret = ops->write_std(ctxt, new_tss_base,
2995 				     &tss_seg.prev_task_link,
2996 				     sizeof tss_seg.prev_task_link,
2997 				     &ctxt->exception);
2998 		if (ret != X86EMUL_CONTINUE)
2999 			return ret;
3000 	}
3001 
3002 	return load_state_from_tss16(ctxt, &tss_seg);
3003 }
3004 
3005 static void save_state_to_tss32(struct x86_emulate_ctxt *ctxt,
3006 				struct tss_segment_32 *tss)
3007 {
3008 	/* CR3 and ldt selector are not saved intentionally */
3009 	tss->eip = ctxt->_eip;
3010 	tss->eflags = ctxt->eflags;
3011 	tss->eax = reg_read(ctxt, VCPU_REGS_RAX);
3012 	tss->ecx = reg_read(ctxt, VCPU_REGS_RCX);
3013 	tss->edx = reg_read(ctxt, VCPU_REGS_RDX);
3014 	tss->ebx = reg_read(ctxt, VCPU_REGS_RBX);
3015 	tss->esp = reg_read(ctxt, VCPU_REGS_RSP);
3016 	tss->ebp = reg_read(ctxt, VCPU_REGS_RBP);
3017 	tss->esi = reg_read(ctxt, VCPU_REGS_RSI);
3018 	tss->edi = reg_read(ctxt, VCPU_REGS_RDI);
3019 
3020 	tss->es = get_segment_selector(ctxt, VCPU_SREG_ES);
3021 	tss->cs = get_segment_selector(ctxt, VCPU_SREG_CS);
3022 	tss->ss = get_segment_selector(ctxt, VCPU_SREG_SS);
3023 	tss->ds = get_segment_selector(ctxt, VCPU_SREG_DS);
3024 	tss->fs = get_segment_selector(ctxt, VCPU_SREG_FS);
3025 	tss->gs = get_segment_selector(ctxt, VCPU_SREG_GS);
3026 }
3027 
3028 static int load_state_from_tss32(struct x86_emulate_ctxt *ctxt,
3029 				 struct tss_segment_32 *tss)
3030 {
3031 	int ret;
3032 	u8 cpl;
3033 
3034 	if (ctxt->ops->set_cr(ctxt, 3, tss->cr3))
3035 		return emulate_gp(ctxt, 0);
3036 	ctxt->_eip = tss->eip;
3037 	ctxt->eflags = tss->eflags | 2;
3038 
3039 	/* General purpose registers */
3040 	*reg_write(ctxt, VCPU_REGS_RAX) = tss->eax;
3041 	*reg_write(ctxt, VCPU_REGS_RCX) = tss->ecx;
3042 	*reg_write(ctxt, VCPU_REGS_RDX) = tss->edx;
3043 	*reg_write(ctxt, VCPU_REGS_RBX) = tss->ebx;
3044 	*reg_write(ctxt, VCPU_REGS_RSP) = tss->esp;
3045 	*reg_write(ctxt, VCPU_REGS_RBP) = tss->ebp;
3046 	*reg_write(ctxt, VCPU_REGS_RSI) = tss->esi;
3047 	*reg_write(ctxt, VCPU_REGS_RDI) = tss->edi;
3048 
3049 	/*
3050 	 * SDM says that segment selectors are loaded before segment
3051 	 * descriptors.  This is important because CPL checks will
3052 	 * use CS.RPL.
3053 	 */
3054 	set_segment_selector(ctxt, tss->ldt_selector, VCPU_SREG_LDTR);
3055 	set_segment_selector(ctxt, tss->es, VCPU_SREG_ES);
3056 	set_segment_selector(ctxt, tss->cs, VCPU_SREG_CS);
3057 	set_segment_selector(ctxt, tss->ss, VCPU_SREG_SS);
3058 	set_segment_selector(ctxt, tss->ds, VCPU_SREG_DS);
3059 	set_segment_selector(ctxt, tss->fs, VCPU_SREG_FS);
3060 	set_segment_selector(ctxt, tss->gs, VCPU_SREG_GS);
3061 
3062 	/*
3063 	 * If we're switching between Protected Mode and VM86, we need to make
3064 	 * sure to update the mode before loading the segment descriptors so
3065 	 * that the selectors are interpreted correctly.
3066 	 */
3067 	if (ctxt->eflags & X86_EFLAGS_VM) {
3068 		ctxt->mode = X86EMUL_MODE_VM86;
3069 		cpl = 3;
3070 	} else {
3071 		ctxt->mode = X86EMUL_MODE_PROT32;
3072 		cpl = tss->cs & 3;
3073 	}
3074 
3075 	/*
3076 	 * Now load segment descriptors. If fault happenes at this stage
3077 	 * it is handled in a context of new task
3078 	 */
3079 	ret = __load_segment_descriptor(ctxt, tss->ldt_selector, VCPU_SREG_LDTR,
3080 					cpl, X86_TRANSFER_TASK_SWITCH, NULL);
3081 	if (ret != X86EMUL_CONTINUE)
3082 		return ret;
3083 	ret = __load_segment_descriptor(ctxt, tss->es, VCPU_SREG_ES, cpl,
3084 					X86_TRANSFER_TASK_SWITCH, NULL);
3085 	if (ret != X86EMUL_CONTINUE)
3086 		return ret;
3087 	ret = __load_segment_descriptor(ctxt, tss->cs, VCPU_SREG_CS, cpl,
3088 					X86_TRANSFER_TASK_SWITCH, NULL);
3089 	if (ret != X86EMUL_CONTINUE)
3090 		return ret;
3091 	ret = __load_segment_descriptor(ctxt, tss->ss, VCPU_SREG_SS, cpl,
3092 					X86_TRANSFER_TASK_SWITCH, NULL);
3093 	if (ret != X86EMUL_CONTINUE)
3094 		return ret;
3095 	ret = __load_segment_descriptor(ctxt, tss->ds, VCPU_SREG_DS, cpl,
3096 					X86_TRANSFER_TASK_SWITCH, NULL);
3097 	if (ret != X86EMUL_CONTINUE)
3098 		return ret;
3099 	ret = __load_segment_descriptor(ctxt, tss->fs, VCPU_SREG_FS, cpl,
3100 					X86_TRANSFER_TASK_SWITCH, NULL);
3101 	if (ret != X86EMUL_CONTINUE)
3102 		return ret;
3103 	ret = __load_segment_descriptor(ctxt, tss->gs, VCPU_SREG_GS, cpl,
3104 					X86_TRANSFER_TASK_SWITCH, NULL);
3105 
3106 	return ret;
3107 }
3108 
3109 static int task_switch_32(struct x86_emulate_ctxt *ctxt,
3110 			  u16 tss_selector, u16 old_tss_sel,
3111 			  ulong old_tss_base, struct desc_struct *new_desc)
3112 {
3113 	const struct x86_emulate_ops *ops = ctxt->ops;
3114 	struct tss_segment_32 tss_seg;
3115 	int ret;
3116 	u32 new_tss_base = get_desc_base(new_desc);
3117 	u32 eip_offset = offsetof(struct tss_segment_32, eip);
3118 	u32 ldt_sel_offset = offsetof(struct tss_segment_32, ldt_selector);
3119 
3120 	ret = ops->read_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
3121 			    &ctxt->exception);
3122 	if (ret != X86EMUL_CONTINUE)
3123 		return ret;
3124 
3125 	save_state_to_tss32(ctxt, &tss_seg);
3126 
3127 	/* Only GP registers and segment selectors are saved */
3128 	ret = ops->write_std(ctxt, old_tss_base + eip_offset, &tss_seg.eip,
3129 			     ldt_sel_offset - eip_offset, &ctxt->exception);
3130 	if (ret != X86EMUL_CONTINUE)
3131 		return ret;
3132 
3133 	ret = ops->read_std(ctxt, new_tss_base, &tss_seg, sizeof tss_seg,
3134 			    &ctxt->exception);
3135 	if (ret != X86EMUL_CONTINUE)
3136 		return ret;
3137 
3138 	if (old_tss_sel != 0xffff) {
3139 		tss_seg.prev_task_link = old_tss_sel;
3140 
3141 		ret = ops->write_std(ctxt, new_tss_base,
3142 				     &tss_seg.prev_task_link,
3143 				     sizeof tss_seg.prev_task_link,
3144 				     &ctxt->exception);
3145 		if (ret != X86EMUL_CONTINUE)
3146 			return ret;
3147 	}
3148 
3149 	return load_state_from_tss32(ctxt, &tss_seg);
3150 }
3151 
3152 static int emulator_do_task_switch(struct x86_emulate_ctxt *ctxt,
3153 				   u16 tss_selector, int idt_index, int reason,
3154 				   bool has_error_code, u32 error_code)
3155 {
3156 	const struct x86_emulate_ops *ops = ctxt->ops;
3157 	struct desc_struct curr_tss_desc, next_tss_desc;
3158 	int ret;
3159 	u16 old_tss_sel = get_segment_selector(ctxt, VCPU_SREG_TR);
3160 	ulong old_tss_base =
3161 		ops->get_cached_segment_base(ctxt, VCPU_SREG_TR);
3162 	u32 desc_limit;
3163 	ulong desc_addr, dr7;
3164 
3165 	/* FIXME: old_tss_base == ~0 ? */
3166 
3167 	ret = read_segment_descriptor(ctxt, tss_selector, &next_tss_desc, &desc_addr);
3168 	if (ret != X86EMUL_CONTINUE)
3169 		return ret;
3170 	ret = read_segment_descriptor(ctxt, old_tss_sel, &curr_tss_desc, &desc_addr);
3171 	if (ret != X86EMUL_CONTINUE)
3172 		return ret;
3173 
3174 	/* FIXME: check that next_tss_desc is tss */
3175 
3176 	/*
3177 	 * Check privileges. The three cases are task switch caused by...
3178 	 *
3179 	 * 1. jmp/call/int to task gate: Check against DPL of the task gate
3180 	 * 2. Exception/IRQ/iret: No check is performed
3181 	 * 3. jmp/call to TSS/task-gate: No check is performed since the
3182 	 *    hardware checks it before exiting.
3183 	 */
3184 	if (reason == TASK_SWITCH_GATE) {
3185 		if (idt_index != -1) {
3186 			/* Software interrupts */
3187 			struct desc_struct task_gate_desc;
3188 			int dpl;
3189 
3190 			ret = read_interrupt_descriptor(ctxt, idt_index,
3191 							&task_gate_desc);
3192 			if (ret != X86EMUL_CONTINUE)
3193 				return ret;
3194 
3195 			dpl = task_gate_desc.dpl;
3196 			if ((tss_selector & 3) > dpl || ops->cpl(ctxt) > dpl)
3197 				return emulate_gp(ctxt, (idt_index << 3) | 0x2);
3198 		}
3199 	}
3200 
3201 	desc_limit = desc_limit_scaled(&next_tss_desc);
3202 	if (!next_tss_desc.p ||
3203 	    ((desc_limit < 0x67 && (next_tss_desc.type & 8)) ||
3204 	     desc_limit < 0x2b)) {
3205 		return emulate_ts(ctxt, tss_selector & 0xfffc);
3206 	}
3207 
3208 	if (reason == TASK_SWITCH_IRET || reason == TASK_SWITCH_JMP) {
3209 		curr_tss_desc.type &= ~(1 << 1); /* clear busy flag */
3210 		write_segment_descriptor(ctxt, old_tss_sel, &curr_tss_desc);
3211 	}
3212 
3213 	if (reason == TASK_SWITCH_IRET)
3214 		ctxt->eflags = ctxt->eflags & ~X86_EFLAGS_NT;
3215 
3216 	/* set back link to prev task only if NT bit is set in eflags
3217 	   note that old_tss_sel is not used after this point */
3218 	if (reason != TASK_SWITCH_CALL && reason != TASK_SWITCH_GATE)
3219 		old_tss_sel = 0xffff;
3220 
3221 	if (next_tss_desc.type & 8)
3222 		ret = task_switch_32(ctxt, tss_selector, old_tss_sel,
3223 				     old_tss_base, &next_tss_desc);
3224 	else
3225 		ret = task_switch_16(ctxt, tss_selector, old_tss_sel,
3226 				     old_tss_base, &next_tss_desc);
3227 	if (ret != X86EMUL_CONTINUE)
3228 		return ret;
3229 
3230 	if (reason == TASK_SWITCH_CALL || reason == TASK_SWITCH_GATE)
3231 		ctxt->eflags = ctxt->eflags | X86_EFLAGS_NT;
3232 
3233 	if (reason != TASK_SWITCH_IRET) {
3234 		next_tss_desc.type |= (1 << 1); /* set busy flag */
3235 		write_segment_descriptor(ctxt, tss_selector, &next_tss_desc);
3236 	}
3237 
3238 	ops->set_cr(ctxt, 0,  ops->get_cr(ctxt, 0) | X86_CR0_TS);
3239 	ops->set_segment(ctxt, tss_selector, &next_tss_desc, 0, VCPU_SREG_TR);
3240 
3241 	if (has_error_code) {
3242 		ctxt->op_bytes = ctxt->ad_bytes = (next_tss_desc.type & 8) ? 4 : 2;
3243 		ctxt->lock_prefix = 0;
3244 		ctxt->src.val = (unsigned long) error_code;
3245 		ret = em_push(ctxt);
3246 	}
3247 
3248 	ops->get_dr(ctxt, 7, &dr7);
3249 	ops->set_dr(ctxt, 7, dr7 & ~(DR_LOCAL_ENABLE_MASK | DR_LOCAL_SLOWDOWN));
3250 
3251 	return ret;
3252 }
3253 
3254 int emulator_task_switch(struct x86_emulate_ctxt *ctxt,
3255 			 u16 tss_selector, int idt_index, int reason,
3256 			 bool has_error_code, u32 error_code)
3257 {
3258 	int rc;
3259 
3260 	invalidate_registers(ctxt);
3261 	ctxt->_eip = ctxt->eip;
3262 	ctxt->dst.type = OP_NONE;
3263 
3264 	rc = emulator_do_task_switch(ctxt, tss_selector, idt_index, reason,
3265 				     has_error_code, error_code);
3266 
3267 	if (rc == X86EMUL_CONTINUE) {
3268 		ctxt->eip = ctxt->_eip;
3269 		writeback_registers(ctxt);
3270 	}
3271 
3272 	return (rc == X86EMUL_UNHANDLEABLE) ? EMULATION_FAILED : EMULATION_OK;
3273 }
3274 
3275 static void string_addr_inc(struct x86_emulate_ctxt *ctxt, int reg,
3276 		struct operand *op)
3277 {
3278 	int df = (ctxt->eflags & X86_EFLAGS_DF) ? -op->count : op->count;
3279 
3280 	register_address_increment(ctxt, reg, df * op->bytes);
3281 	op->addr.mem.ea = register_address(ctxt, reg);
3282 }
3283 
3284 static int em_das(struct x86_emulate_ctxt *ctxt)
3285 {
3286 	u8 al, old_al;
3287 	bool af, cf, old_cf;
3288 
3289 	cf = ctxt->eflags & X86_EFLAGS_CF;
3290 	al = ctxt->dst.val;
3291 
3292 	old_al = al;
3293 	old_cf = cf;
3294 	cf = false;
3295 	af = ctxt->eflags & X86_EFLAGS_AF;
3296 	if ((al & 0x0f) > 9 || af) {
3297 		al -= 6;
3298 		cf = old_cf | (al >= 250);
3299 		af = true;
3300 	} else {
3301 		af = false;
3302 	}
3303 	if (old_al > 0x99 || old_cf) {
3304 		al -= 0x60;
3305 		cf = true;
3306 	}
3307 
3308 	ctxt->dst.val = al;
3309 	/* Set PF, ZF, SF */
3310 	ctxt->src.type = OP_IMM;
3311 	ctxt->src.val = 0;
3312 	ctxt->src.bytes = 1;
3313 	fastop(ctxt, em_or);
3314 	ctxt->eflags &= ~(X86_EFLAGS_AF | X86_EFLAGS_CF);
3315 	if (cf)
3316 		ctxt->eflags |= X86_EFLAGS_CF;
3317 	if (af)
3318 		ctxt->eflags |= X86_EFLAGS_AF;
3319 	return X86EMUL_CONTINUE;
3320 }
3321 
3322 static int em_aam(struct x86_emulate_ctxt *ctxt)
3323 {
3324 	u8 al, ah;
3325 
3326 	if (ctxt->src.val == 0)
3327 		return emulate_de(ctxt);
3328 
3329 	al = ctxt->dst.val & 0xff;
3330 	ah = al / ctxt->src.val;
3331 	al %= ctxt->src.val;
3332 
3333 	ctxt->dst.val = (ctxt->dst.val & 0xffff0000) | al | (ah << 8);
3334 
3335 	/* Set PF, ZF, SF */
3336 	ctxt->src.type = OP_IMM;
3337 	ctxt->src.val = 0;
3338 	ctxt->src.bytes = 1;
3339 	fastop(ctxt, em_or);
3340 
3341 	return X86EMUL_CONTINUE;
3342 }
3343 
3344 static int em_aad(struct x86_emulate_ctxt *ctxt)
3345 {
3346 	u8 al = ctxt->dst.val & 0xff;
3347 	u8 ah = (ctxt->dst.val >> 8) & 0xff;
3348 
3349 	al = (al + (ah * ctxt->src.val)) & 0xff;
3350 
3351 	ctxt->dst.val = (ctxt->dst.val & 0xffff0000) | al;
3352 
3353 	/* Set PF, ZF, SF */
3354 	ctxt->src.type = OP_IMM;
3355 	ctxt->src.val = 0;
3356 	ctxt->src.bytes = 1;
3357 	fastop(ctxt, em_or);
3358 
3359 	return X86EMUL_CONTINUE;
3360 }
3361 
3362 static int em_call(struct x86_emulate_ctxt *ctxt)
3363 {
3364 	int rc;
3365 	long rel = ctxt->src.val;
3366 
3367 	ctxt->src.val = (unsigned long)ctxt->_eip;
3368 	rc = jmp_rel(ctxt, rel);
3369 	if (rc != X86EMUL_CONTINUE)
3370 		return rc;
3371 	return em_push(ctxt);
3372 }
3373 
3374 static int em_call_far(struct x86_emulate_ctxt *ctxt)
3375 {
3376 	u16 sel, old_cs;
3377 	ulong old_eip;
3378 	int rc;
3379 	struct desc_struct old_desc, new_desc;
3380 	const struct x86_emulate_ops *ops = ctxt->ops;
3381 	int cpl = ctxt->ops->cpl(ctxt);
3382 	enum x86emul_mode prev_mode = ctxt->mode;
3383 
3384 	old_eip = ctxt->_eip;
3385 	ops->get_segment(ctxt, &old_cs, &old_desc, NULL, VCPU_SREG_CS);
3386 
3387 	memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
3388 	rc = __load_segment_descriptor(ctxt, sel, VCPU_SREG_CS, cpl,
3389 				       X86_TRANSFER_CALL_JMP, &new_desc);
3390 	if (rc != X86EMUL_CONTINUE)
3391 		return rc;
3392 
3393 	rc = assign_eip_far(ctxt, ctxt->src.val, &new_desc);
3394 	if (rc != X86EMUL_CONTINUE)
3395 		goto fail;
3396 
3397 	ctxt->src.val = old_cs;
3398 	rc = em_push(ctxt);
3399 	if (rc != X86EMUL_CONTINUE)
3400 		goto fail;
3401 
3402 	ctxt->src.val = old_eip;
3403 	rc = em_push(ctxt);
3404 	/* If we failed, we tainted the memory, but the very least we should
3405 	   restore cs */
3406 	if (rc != X86EMUL_CONTINUE) {
3407 		pr_warn_once("faulting far call emulation tainted memory\n");
3408 		goto fail;
3409 	}
3410 	return rc;
3411 fail:
3412 	ops->set_segment(ctxt, old_cs, &old_desc, 0, VCPU_SREG_CS);
3413 	ctxt->mode = prev_mode;
3414 	return rc;
3415 
3416 }
3417 
3418 static int em_ret_near_imm(struct x86_emulate_ctxt *ctxt)
3419 {
3420 	int rc;
3421 	unsigned long eip;
3422 
3423 	rc = emulate_pop(ctxt, &eip, ctxt->op_bytes);
3424 	if (rc != X86EMUL_CONTINUE)
3425 		return rc;
3426 	rc = assign_eip_near(ctxt, eip);
3427 	if (rc != X86EMUL_CONTINUE)
3428 		return rc;
3429 	rsp_increment(ctxt, ctxt->src.val);
3430 	return X86EMUL_CONTINUE;
3431 }
3432 
3433 static int em_xchg(struct x86_emulate_ctxt *ctxt)
3434 {
3435 	/* Write back the register source. */
3436 	ctxt->src.val = ctxt->dst.val;
3437 	write_register_operand(&ctxt->src);
3438 
3439 	/* Write back the memory destination with implicit LOCK prefix. */
3440 	ctxt->dst.val = ctxt->src.orig_val;
3441 	ctxt->lock_prefix = 1;
3442 	return X86EMUL_CONTINUE;
3443 }
3444 
3445 static int em_imul_3op(struct x86_emulate_ctxt *ctxt)
3446 {
3447 	ctxt->dst.val = ctxt->src2.val;
3448 	return fastop(ctxt, em_imul);
3449 }
3450 
3451 static int em_cwd(struct x86_emulate_ctxt *ctxt)
3452 {
3453 	ctxt->dst.type = OP_REG;
3454 	ctxt->dst.bytes = ctxt->src.bytes;
3455 	ctxt->dst.addr.reg = reg_rmw(ctxt, VCPU_REGS_RDX);
3456 	ctxt->dst.val = ~((ctxt->src.val >> (ctxt->src.bytes * 8 - 1)) - 1);
3457 
3458 	return X86EMUL_CONTINUE;
3459 }
3460 
3461 static int em_rdtsc(struct x86_emulate_ctxt *ctxt)
3462 {
3463 	u64 tsc = 0;
3464 
3465 	ctxt->ops->get_msr(ctxt, MSR_IA32_TSC, &tsc);
3466 	*reg_write(ctxt, VCPU_REGS_RAX) = (u32)tsc;
3467 	*reg_write(ctxt, VCPU_REGS_RDX) = tsc >> 32;
3468 	return X86EMUL_CONTINUE;
3469 }
3470 
3471 static int em_rdpmc(struct x86_emulate_ctxt *ctxt)
3472 {
3473 	u64 pmc;
3474 
3475 	if (ctxt->ops->read_pmc(ctxt, reg_read(ctxt, VCPU_REGS_RCX), &pmc))
3476 		return emulate_gp(ctxt, 0);
3477 	*reg_write(ctxt, VCPU_REGS_RAX) = (u32)pmc;
3478 	*reg_write(ctxt, VCPU_REGS_RDX) = pmc >> 32;
3479 	return X86EMUL_CONTINUE;
3480 }
3481 
3482 static int em_mov(struct x86_emulate_ctxt *ctxt)
3483 {
3484 	memcpy(ctxt->dst.valptr, ctxt->src.valptr, sizeof(ctxt->src.valptr));
3485 	return X86EMUL_CONTINUE;
3486 }
3487 
3488 #define FFL(x) bit(X86_FEATURE_##x)
3489 
3490 static int em_movbe(struct x86_emulate_ctxt *ctxt)
3491 {
3492 	u32 ebx, ecx, edx, eax = 1;
3493 	u16 tmp;
3494 
3495 	/*
3496 	 * Check MOVBE is set in the guest-visible CPUID leaf.
3497 	 */
3498 	ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
3499 	if (!(ecx & FFL(MOVBE)))
3500 		return emulate_ud(ctxt);
3501 
3502 	switch (ctxt->op_bytes) {
3503 	case 2:
3504 		/*
3505 		 * From MOVBE definition: "...When the operand size is 16 bits,
3506 		 * the upper word of the destination register remains unchanged
3507 		 * ..."
3508 		 *
3509 		 * Both casting ->valptr and ->val to u16 breaks strict aliasing
3510 		 * rules so we have to do the operation almost per hand.
3511 		 */
3512 		tmp = (u16)ctxt->src.val;
3513 		ctxt->dst.val &= ~0xffffUL;
3514 		ctxt->dst.val |= (unsigned long)swab16(tmp);
3515 		break;
3516 	case 4:
3517 		ctxt->dst.val = swab32((u32)ctxt->src.val);
3518 		break;
3519 	case 8:
3520 		ctxt->dst.val = swab64(ctxt->src.val);
3521 		break;
3522 	default:
3523 		BUG();
3524 	}
3525 	return X86EMUL_CONTINUE;
3526 }
3527 
3528 static int em_cr_write(struct x86_emulate_ctxt *ctxt)
3529 {
3530 	if (ctxt->ops->set_cr(ctxt, ctxt->modrm_reg, ctxt->src.val))
3531 		return emulate_gp(ctxt, 0);
3532 
3533 	/* Disable writeback. */
3534 	ctxt->dst.type = OP_NONE;
3535 	return X86EMUL_CONTINUE;
3536 }
3537 
3538 static int em_dr_write(struct x86_emulate_ctxt *ctxt)
3539 {
3540 	unsigned long val;
3541 
3542 	if (ctxt->mode == X86EMUL_MODE_PROT64)
3543 		val = ctxt->src.val & ~0ULL;
3544 	else
3545 		val = ctxt->src.val & ~0U;
3546 
3547 	/* #UD condition is already handled. */
3548 	if (ctxt->ops->set_dr(ctxt, ctxt->modrm_reg, val) < 0)
3549 		return emulate_gp(ctxt, 0);
3550 
3551 	/* Disable writeback. */
3552 	ctxt->dst.type = OP_NONE;
3553 	return X86EMUL_CONTINUE;
3554 }
3555 
3556 static int em_wrmsr(struct x86_emulate_ctxt *ctxt)
3557 {
3558 	u64 msr_data;
3559 
3560 	msr_data = (u32)reg_read(ctxt, VCPU_REGS_RAX)
3561 		| ((u64)reg_read(ctxt, VCPU_REGS_RDX) << 32);
3562 	if (ctxt->ops->set_msr(ctxt, reg_read(ctxt, VCPU_REGS_RCX), msr_data))
3563 		return emulate_gp(ctxt, 0);
3564 
3565 	return X86EMUL_CONTINUE;
3566 }
3567 
3568 static int em_rdmsr(struct x86_emulate_ctxt *ctxt)
3569 {
3570 	u64 msr_data;
3571 
3572 	if (ctxt->ops->get_msr(ctxt, reg_read(ctxt, VCPU_REGS_RCX), &msr_data))
3573 		return emulate_gp(ctxt, 0);
3574 
3575 	*reg_write(ctxt, VCPU_REGS_RAX) = (u32)msr_data;
3576 	*reg_write(ctxt, VCPU_REGS_RDX) = msr_data >> 32;
3577 	return X86EMUL_CONTINUE;
3578 }
3579 
3580 static int em_mov_rm_sreg(struct x86_emulate_ctxt *ctxt)
3581 {
3582 	if (ctxt->modrm_reg > VCPU_SREG_GS)
3583 		return emulate_ud(ctxt);
3584 
3585 	ctxt->dst.val = get_segment_selector(ctxt, ctxt->modrm_reg);
3586 	if (ctxt->dst.bytes == 4 && ctxt->dst.type == OP_MEM)
3587 		ctxt->dst.bytes = 2;
3588 	return X86EMUL_CONTINUE;
3589 }
3590 
3591 static int em_mov_sreg_rm(struct x86_emulate_ctxt *ctxt)
3592 {
3593 	u16 sel = ctxt->src.val;
3594 
3595 	if (ctxt->modrm_reg == VCPU_SREG_CS || ctxt->modrm_reg > VCPU_SREG_GS)
3596 		return emulate_ud(ctxt);
3597 
3598 	if (ctxt->modrm_reg == VCPU_SREG_SS)
3599 		ctxt->interruptibility = KVM_X86_SHADOW_INT_MOV_SS;
3600 
3601 	/* Disable writeback. */
3602 	ctxt->dst.type = OP_NONE;
3603 	return load_segment_descriptor(ctxt, sel, ctxt->modrm_reg);
3604 }
3605 
3606 static int em_lldt(struct x86_emulate_ctxt *ctxt)
3607 {
3608 	u16 sel = ctxt->src.val;
3609 
3610 	/* Disable writeback. */
3611 	ctxt->dst.type = OP_NONE;
3612 	return load_segment_descriptor(ctxt, sel, VCPU_SREG_LDTR);
3613 }
3614 
3615 static int em_ltr(struct x86_emulate_ctxt *ctxt)
3616 {
3617 	u16 sel = ctxt->src.val;
3618 
3619 	/* Disable writeback. */
3620 	ctxt->dst.type = OP_NONE;
3621 	return load_segment_descriptor(ctxt, sel, VCPU_SREG_TR);
3622 }
3623 
3624 static int em_invlpg(struct x86_emulate_ctxt *ctxt)
3625 {
3626 	int rc;
3627 	ulong linear;
3628 
3629 	rc = linearize(ctxt, ctxt->src.addr.mem, 1, false, &linear);
3630 	if (rc == X86EMUL_CONTINUE)
3631 		ctxt->ops->invlpg(ctxt, linear);
3632 	/* Disable writeback. */
3633 	ctxt->dst.type = OP_NONE;
3634 	return X86EMUL_CONTINUE;
3635 }
3636 
3637 static int em_clts(struct x86_emulate_ctxt *ctxt)
3638 {
3639 	ulong cr0;
3640 
3641 	cr0 = ctxt->ops->get_cr(ctxt, 0);
3642 	cr0 &= ~X86_CR0_TS;
3643 	ctxt->ops->set_cr(ctxt, 0, cr0);
3644 	return X86EMUL_CONTINUE;
3645 }
3646 
3647 static int em_hypercall(struct x86_emulate_ctxt *ctxt)
3648 {
3649 	int rc = ctxt->ops->fix_hypercall(ctxt);
3650 
3651 	if (rc != X86EMUL_CONTINUE)
3652 		return rc;
3653 
3654 	/* Let the processor re-execute the fixed hypercall */
3655 	ctxt->_eip = ctxt->eip;
3656 	/* Disable writeback. */
3657 	ctxt->dst.type = OP_NONE;
3658 	return X86EMUL_CONTINUE;
3659 }
3660 
3661 static int emulate_store_desc_ptr(struct x86_emulate_ctxt *ctxt,
3662 				  void (*get)(struct x86_emulate_ctxt *ctxt,
3663 					      struct desc_ptr *ptr))
3664 {
3665 	struct desc_ptr desc_ptr;
3666 
3667 	if (ctxt->mode == X86EMUL_MODE_PROT64)
3668 		ctxt->op_bytes = 8;
3669 	get(ctxt, &desc_ptr);
3670 	if (ctxt->op_bytes == 2) {
3671 		ctxt->op_bytes = 4;
3672 		desc_ptr.address &= 0x00ffffff;
3673 	}
3674 	/* Disable writeback. */
3675 	ctxt->dst.type = OP_NONE;
3676 	return segmented_write(ctxt, ctxt->dst.addr.mem,
3677 			       &desc_ptr, 2 + ctxt->op_bytes);
3678 }
3679 
3680 static int em_sgdt(struct x86_emulate_ctxt *ctxt)
3681 {
3682 	return emulate_store_desc_ptr(ctxt, ctxt->ops->get_gdt);
3683 }
3684 
3685 static int em_sidt(struct x86_emulate_ctxt *ctxt)
3686 {
3687 	return emulate_store_desc_ptr(ctxt, ctxt->ops->get_idt);
3688 }
3689 
3690 static int em_lgdt_lidt(struct x86_emulate_ctxt *ctxt, bool lgdt)
3691 {
3692 	struct desc_ptr desc_ptr;
3693 	int rc;
3694 
3695 	if (ctxt->mode == X86EMUL_MODE_PROT64)
3696 		ctxt->op_bytes = 8;
3697 	rc = read_descriptor(ctxt, ctxt->src.addr.mem,
3698 			     &desc_ptr.size, &desc_ptr.address,
3699 			     ctxt->op_bytes);
3700 	if (rc != X86EMUL_CONTINUE)
3701 		return rc;
3702 	if (ctxt->mode == X86EMUL_MODE_PROT64 &&
3703 	    is_noncanonical_address(desc_ptr.address))
3704 		return emulate_gp(ctxt, 0);
3705 	if (lgdt)
3706 		ctxt->ops->set_gdt(ctxt, &desc_ptr);
3707 	else
3708 		ctxt->ops->set_idt(ctxt, &desc_ptr);
3709 	/* Disable writeback. */
3710 	ctxt->dst.type = OP_NONE;
3711 	return X86EMUL_CONTINUE;
3712 }
3713 
3714 static int em_lgdt(struct x86_emulate_ctxt *ctxt)
3715 {
3716 	return em_lgdt_lidt(ctxt, true);
3717 }
3718 
3719 static int em_lidt(struct x86_emulate_ctxt *ctxt)
3720 {
3721 	return em_lgdt_lidt(ctxt, false);
3722 }
3723 
3724 static int em_smsw(struct x86_emulate_ctxt *ctxt)
3725 {
3726 	if (ctxt->dst.type == OP_MEM)
3727 		ctxt->dst.bytes = 2;
3728 	ctxt->dst.val = ctxt->ops->get_cr(ctxt, 0);
3729 	return X86EMUL_CONTINUE;
3730 }
3731 
3732 static int em_lmsw(struct x86_emulate_ctxt *ctxt)
3733 {
3734 	ctxt->ops->set_cr(ctxt, 0, (ctxt->ops->get_cr(ctxt, 0) & ~0x0eul)
3735 			  | (ctxt->src.val & 0x0f));
3736 	ctxt->dst.type = OP_NONE;
3737 	return X86EMUL_CONTINUE;
3738 }
3739 
3740 static int em_loop(struct x86_emulate_ctxt *ctxt)
3741 {
3742 	int rc = X86EMUL_CONTINUE;
3743 
3744 	register_address_increment(ctxt, VCPU_REGS_RCX, -1);
3745 	if ((address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) != 0) &&
3746 	    (ctxt->b == 0xe2 || test_cc(ctxt->b ^ 0x5, ctxt->eflags)))
3747 		rc = jmp_rel(ctxt, ctxt->src.val);
3748 
3749 	return rc;
3750 }
3751 
3752 static int em_jcxz(struct x86_emulate_ctxt *ctxt)
3753 {
3754 	int rc = X86EMUL_CONTINUE;
3755 
3756 	if (address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) == 0)
3757 		rc = jmp_rel(ctxt, ctxt->src.val);
3758 
3759 	return rc;
3760 }
3761 
3762 static int em_in(struct x86_emulate_ctxt *ctxt)
3763 {
3764 	if (!pio_in_emulated(ctxt, ctxt->dst.bytes, ctxt->src.val,
3765 			     &ctxt->dst.val))
3766 		return X86EMUL_IO_NEEDED;
3767 
3768 	return X86EMUL_CONTINUE;
3769 }
3770 
3771 static int em_out(struct x86_emulate_ctxt *ctxt)
3772 {
3773 	ctxt->ops->pio_out_emulated(ctxt, ctxt->src.bytes, ctxt->dst.val,
3774 				    &ctxt->src.val, 1);
3775 	/* Disable writeback. */
3776 	ctxt->dst.type = OP_NONE;
3777 	return X86EMUL_CONTINUE;
3778 }
3779 
3780 static int em_cli(struct x86_emulate_ctxt *ctxt)
3781 {
3782 	if (emulator_bad_iopl(ctxt))
3783 		return emulate_gp(ctxt, 0);
3784 
3785 	ctxt->eflags &= ~X86_EFLAGS_IF;
3786 	return X86EMUL_CONTINUE;
3787 }
3788 
3789 static int em_sti(struct x86_emulate_ctxt *ctxt)
3790 {
3791 	if (emulator_bad_iopl(ctxt))
3792 		return emulate_gp(ctxt, 0);
3793 
3794 	ctxt->interruptibility = KVM_X86_SHADOW_INT_STI;
3795 	ctxt->eflags |= X86_EFLAGS_IF;
3796 	return X86EMUL_CONTINUE;
3797 }
3798 
3799 static int em_cpuid(struct x86_emulate_ctxt *ctxt)
3800 {
3801 	u32 eax, ebx, ecx, edx;
3802 
3803 	eax = reg_read(ctxt, VCPU_REGS_RAX);
3804 	ecx = reg_read(ctxt, VCPU_REGS_RCX);
3805 	ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
3806 	*reg_write(ctxt, VCPU_REGS_RAX) = eax;
3807 	*reg_write(ctxt, VCPU_REGS_RBX) = ebx;
3808 	*reg_write(ctxt, VCPU_REGS_RCX) = ecx;
3809 	*reg_write(ctxt, VCPU_REGS_RDX) = edx;
3810 	return X86EMUL_CONTINUE;
3811 }
3812 
3813 static int em_sahf(struct x86_emulate_ctxt *ctxt)
3814 {
3815 	u32 flags;
3816 
3817 	flags = X86_EFLAGS_CF | X86_EFLAGS_PF | X86_EFLAGS_AF | X86_EFLAGS_ZF |
3818 		X86_EFLAGS_SF;
3819 	flags &= *reg_rmw(ctxt, VCPU_REGS_RAX) >> 8;
3820 
3821 	ctxt->eflags &= ~0xffUL;
3822 	ctxt->eflags |= flags | X86_EFLAGS_FIXED;
3823 	return X86EMUL_CONTINUE;
3824 }
3825 
3826 static int em_lahf(struct x86_emulate_ctxt *ctxt)
3827 {
3828 	*reg_rmw(ctxt, VCPU_REGS_RAX) &= ~0xff00UL;
3829 	*reg_rmw(ctxt, VCPU_REGS_RAX) |= (ctxt->eflags & 0xff) << 8;
3830 	return X86EMUL_CONTINUE;
3831 }
3832 
3833 static int em_bswap(struct x86_emulate_ctxt *ctxt)
3834 {
3835 	switch (ctxt->op_bytes) {
3836 #ifdef CONFIG_X86_64
3837 	case 8:
3838 		asm("bswap %0" : "+r"(ctxt->dst.val));
3839 		break;
3840 #endif
3841 	default:
3842 		asm("bswap %0" : "+r"(*(u32 *)&ctxt->dst.val));
3843 		break;
3844 	}
3845 	return X86EMUL_CONTINUE;
3846 }
3847 
3848 static int em_clflush(struct x86_emulate_ctxt *ctxt)
3849 {
3850 	/* emulating clflush regardless of cpuid */
3851 	return X86EMUL_CONTINUE;
3852 }
3853 
3854 static int em_movsxd(struct x86_emulate_ctxt *ctxt)
3855 {
3856 	ctxt->dst.val = (s32) ctxt->src.val;
3857 	return X86EMUL_CONTINUE;
3858 }
3859 
3860 static bool valid_cr(int nr)
3861 {
3862 	switch (nr) {
3863 	case 0:
3864 	case 2 ... 4:
3865 	case 8:
3866 		return true;
3867 	default:
3868 		return false;
3869 	}
3870 }
3871 
3872 static int check_cr_read(struct x86_emulate_ctxt *ctxt)
3873 {
3874 	if (!valid_cr(ctxt->modrm_reg))
3875 		return emulate_ud(ctxt);
3876 
3877 	return X86EMUL_CONTINUE;
3878 }
3879 
3880 static int check_cr_write(struct x86_emulate_ctxt *ctxt)
3881 {
3882 	u64 new_val = ctxt->src.val64;
3883 	int cr = ctxt->modrm_reg;
3884 	u64 efer = 0;
3885 
3886 	static u64 cr_reserved_bits[] = {
3887 		0xffffffff00000000ULL,
3888 		0, 0, 0, /* CR3 checked later */
3889 		CR4_RESERVED_BITS,
3890 		0, 0, 0,
3891 		CR8_RESERVED_BITS,
3892 	};
3893 
3894 	if (!valid_cr(cr))
3895 		return emulate_ud(ctxt);
3896 
3897 	if (new_val & cr_reserved_bits[cr])
3898 		return emulate_gp(ctxt, 0);
3899 
3900 	switch (cr) {
3901 	case 0: {
3902 		u64 cr4;
3903 		if (((new_val & X86_CR0_PG) && !(new_val & X86_CR0_PE)) ||
3904 		    ((new_val & X86_CR0_NW) && !(new_val & X86_CR0_CD)))
3905 			return emulate_gp(ctxt, 0);
3906 
3907 		cr4 = ctxt->ops->get_cr(ctxt, 4);
3908 		ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3909 
3910 		if ((new_val & X86_CR0_PG) && (efer & EFER_LME) &&
3911 		    !(cr4 & X86_CR4_PAE))
3912 			return emulate_gp(ctxt, 0);
3913 
3914 		break;
3915 		}
3916 	case 3: {
3917 		u64 rsvd = 0;
3918 
3919 		ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3920 		if (efer & EFER_LMA)
3921 			rsvd = CR3_L_MODE_RESERVED_BITS & ~CR3_PCID_INVD;
3922 
3923 		if (new_val & rsvd)
3924 			return emulate_gp(ctxt, 0);
3925 
3926 		break;
3927 		}
3928 	case 4: {
3929 		ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3930 
3931 		if ((efer & EFER_LMA) && !(new_val & X86_CR4_PAE))
3932 			return emulate_gp(ctxt, 0);
3933 
3934 		break;
3935 		}
3936 	}
3937 
3938 	return X86EMUL_CONTINUE;
3939 }
3940 
3941 static int check_dr7_gd(struct x86_emulate_ctxt *ctxt)
3942 {
3943 	unsigned long dr7;
3944 
3945 	ctxt->ops->get_dr(ctxt, 7, &dr7);
3946 
3947 	/* Check if DR7.Global_Enable is set */
3948 	return dr7 & (1 << 13);
3949 }
3950 
3951 static int check_dr_read(struct x86_emulate_ctxt *ctxt)
3952 {
3953 	int dr = ctxt->modrm_reg;
3954 	u64 cr4;
3955 
3956 	if (dr > 7)
3957 		return emulate_ud(ctxt);
3958 
3959 	cr4 = ctxt->ops->get_cr(ctxt, 4);
3960 	if ((cr4 & X86_CR4_DE) && (dr == 4 || dr == 5))
3961 		return emulate_ud(ctxt);
3962 
3963 	if (check_dr7_gd(ctxt)) {
3964 		ulong dr6;
3965 
3966 		ctxt->ops->get_dr(ctxt, 6, &dr6);
3967 		dr6 &= ~15;
3968 		dr6 |= DR6_BD | DR6_RTM;
3969 		ctxt->ops->set_dr(ctxt, 6, dr6);
3970 		return emulate_db(ctxt);
3971 	}
3972 
3973 	return X86EMUL_CONTINUE;
3974 }
3975 
3976 static int check_dr_write(struct x86_emulate_ctxt *ctxt)
3977 {
3978 	u64 new_val = ctxt->src.val64;
3979 	int dr = ctxt->modrm_reg;
3980 
3981 	if ((dr == 6 || dr == 7) && (new_val & 0xffffffff00000000ULL))
3982 		return emulate_gp(ctxt, 0);
3983 
3984 	return check_dr_read(ctxt);
3985 }
3986 
3987 static int check_svme(struct x86_emulate_ctxt *ctxt)
3988 {
3989 	u64 efer;
3990 
3991 	ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3992 
3993 	if (!(efer & EFER_SVME))
3994 		return emulate_ud(ctxt);
3995 
3996 	return X86EMUL_CONTINUE;
3997 }
3998 
3999 static int check_svme_pa(struct x86_emulate_ctxt *ctxt)
4000 {
4001 	u64 rax = reg_read(ctxt, VCPU_REGS_RAX);
4002 
4003 	/* Valid physical address? */
4004 	if (rax & 0xffff000000000000ULL)
4005 		return emulate_gp(ctxt, 0);
4006 
4007 	return check_svme(ctxt);
4008 }
4009 
4010 static int check_rdtsc(struct x86_emulate_ctxt *ctxt)
4011 {
4012 	u64 cr4 = ctxt->ops->get_cr(ctxt, 4);
4013 
4014 	if (cr4 & X86_CR4_TSD && ctxt->ops->cpl(ctxt))
4015 		return emulate_ud(ctxt);
4016 
4017 	return X86EMUL_CONTINUE;
4018 }
4019 
4020 static int check_rdpmc(struct x86_emulate_ctxt *ctxt)
4021 {
4022 	u64 cr4 = ctxt->ops->get_cr(ctxt, 4);
4023 	u64 rcx = reg_read(ctxt, VCPU_REGS_RCX);
4024 
4025 	if ((!(cr4 & X86_CR4_PCE) && ctxt->ops->cpl(ctxt)) ||
4026 	    ctxt->ops->check_pmc(ctxt, rcx))
4027 		return emulate_gp(ctxt, 0);
4028 
4029 	return X86EMUL_CONTINUE;
4030 }
4031 
4032 static int check_perm_in(struct x86_emulate_ctxt *ctxt)
4033 {
4034 	ctxt->dst.bytes = min(ctxt->dst.bytes, 4u);
4035 	if (!emulator_io_permited(ctxt, ctxt->src.val, ctxt->dst.bytes))
4036 		return emulate_gp(ctxt, 0);
4037 
4038 	return X86EMUL_CONTINUE;
4039 }
4040 
4041 static int check_perm_out(struct x86_emulate_ctxt *ctxt)
4042 {
4043 	ctxt->src.bytes = min(ctxt->src.bytes, 4u);
4044 	if (!emulator_io_permited(ctxt, ctxt->dst.val, ctxt->src.bytes))
4045 		return emulate_gp(ctxt, 0);
4046 
4047 	return X86EMUL_CONTINUE;
4048 }
4049 
4050 #define D(_y) { .flags = (_y) }
4051 #define DI(_y, _i) { .flags = (_y)|Intercept, .intercept = x86_intercept_##_i }
4052 #define DIP(_y, _i, _p) { .flags = (_y)|Intercept|CheckPerm, \
4053 		      .intercept = x86_intercept_##_i, .check_perm = (_p) }
4054 #define N    D(NotImpl)
4055 #define EXT(_f, _e) { .flags = ((_f) | RMExt), .u.group = (_e) }
4056 #define G(_f, _g) { .flags = ((_f) | Group | ModRM), .u.group = (_g) }
4057 #define GD(_f, _g) { .flags = ((_f) | GroupDual | ModRM), .u.gdual = (_g) }
4058 #define ID(_f, _i) { .flags = ((_f) | InstrDual | ModRM), .u.idual = (_i) }
4059 #define MD(_f, _m) { .flags = ((_f) | ModeDual), .u.mdual = (_m) }
4060 #define E(_f, _e) { .flags = ((_f) | Escape | ModRM), .u.esc = (_e) }
4061 #define I(_f, _e) { .flags = (_f), .u.execute = (_e) }
4062 #define F(_f, _e) { .flags = (_f) | Fastop, .u.fastop = (_e) }
4063 #define II(_f, _e, _i) \
4064 	{ .flags = (_f)|Intercept, .u.execute = (_e), .intercept = x86_intercept_##_i }
4065 #define IIP(_f, _e, _i, _p) \
4066 	{ .flags = (_f)|Intercept|CheckPerm, .u.execute = (_e), \
4067 	  .intercept = x86_intercept_##_i, .check_perm = (_p) }
4068 #define GP(_f, _g) { .flags = ((_f) | Prefix), .u.gprefix = (_g) }
4069 
4070 #define D2bv(_f)      D((_f) | ByteOp), D(_f)
4071 #define D2bvIP(_f, _i, _p) DIP((_f) | ByteOp, _i, _p), DIP(_f, _i, _p)
4072 #define I2bv(_f, _e)  I((_f) | ByteOp, _e), I(_f, _e)
4073 #define F2bv(_f, _e)  F((_f) | ByteOp, _e), F(_f, _e)
4074 #define I2bvIP(_f, _e, _i, _p) \
4075 	IIP((_f) | ByteOp, _e, _i, _p), IIP(_f, _e, _i, _p)
4076 
4077 #define F6ALU(_f, _e) F2bv((_f) | DstMem | SrcReg | ModRM, _e),		\
4078 		F2bv(((_f) | DstReg | SrcMem | ModRM) & ~Lock, _e),	\
4079 		F2bv(((_f) & ~Lock) | DstAcc | SrcImm, _e)
4080 
4081 static const struct opcode group7_rm0[] = {
4082 	N,
4083 	I(SrcNone | Priv | EmulateOnUD,	em_hypercall),
4084 	N, N, N, N, N, N,
4085 };
4086 
4087 static const struct opcode group7_rm1[] = {
4088 	DI(SrcNone | Priv, monitor),
4089 	DI(SrcNone | Priv, mwait),
4090 	N, N, N, N, N, N,
4091 };
4092 
4093 static const struct opcode group7_rm3[] = {
4094 	DIP(SrcNone | Prot | Priv,		vmrun,		check_svme_pa),
4095 	II(SrcNone  | Prot | EmulateOnUD,	em_hypercall,	vmmcall),
4096 	DIP(SrcNone | Prot | Priv,		vmload,		check_svme_pa),
4097 	DIP(SrcNone | Prot | Priv,		vmsave,		check_svme_pa),
4098 	DIP(SrcNone | Prot | Priv,		stgi,		check_svme),
4099 	DIP(SrcNone | Prot | Priv,		clgi,		check_svme),
4100 	DIP(SrcNone | Prot | Priv,		skinit,		check_svme),
4101 	DIP(SrcNone | Prot | Priv,		invlpga,	check_svme),
4102 };
4103 
4104 static const struct opcode group7_rm7[] = {
4105 	N,
4106 	DIP(SrcNone, rdtscp, check_rdtsc),
4107 	N, N, N, N, N, N,
4108 };
4109 
4110 static const struct opcode group1[] = {
4111 	F(Lock, em_add),
4112 	F(Lock | PageTable, em_or),
4113 	F(Lock, em_adc),
4114 	F(Lock, em_sbb),
4115 	F(Lock | PageTable, em_and),
4116 	F(Lock, em_sub),
4117 	F(Lock, em_xor),
4118 	F(NoWrite, em_cmp),
4119 };
4120 
4121 static const struct opcode group1A[] = {
4122 	I(DstMem | SrcNone | Mov | Stack | IncSP, em_pop), N, N, N, N, N, N, N,
4123 };
4124 
4125 static const struct opcode group2[] = {
4126 	F(DstMem | ModRM, em_rol),
4127 	F(DstMem | ModRM, em_ror),
4128 	F(DstMem | ModRM, em_rcl),
4129 	F(DstMem | ModRM, em_rcr),
4130 	F(DstMem | ModRM, em_shl),
4131 	F(DstMem | ModRM, em_shr),
4132 	F(DstMem | ModRM, em_shl),
4133 	F(DstMem | ModRM, em_sar),
4134 };
4135 
4136 static const struct opcode group3[] = {
4137 	F(DstMem | SrcImm | NoWrite, em_test),
4138 	F(DstMem | SrcImm | NoWrite, em_test),
4139 	F(DstMem | SrcNone | Lock, em_not),
4140 	F(DstMem | SrcNone | Lock, em_neg),
4141 	F(DstXacc | Src2Mem, em_mul_ex),
4142 	F(DstXacc | Src2Mem, em_imul_ex),
4143 	F(DstXacc | Src2Mem, em_div_ex),
4144 	F(DstXacc | Src2Mem, em_idiv_ex),
4145 };
4146 
4147 static const struct opcode group4[] = {
4148 	F(ByteOp | DstMem | SrcNone | Lock, em_inc),
4149 	F(ByteOp | DstMem | SrcNone | Lock, em_dec),
4150 	N, N, N, N, N, N,
4151 };
4152 
4153 static const struct opcode group5[] = {
4154 	F(DstMem | SrcNone | Lock,		em_inc),
4155 	F(DstMem | SrcNone | Lock,		em_dec),
4156 	I(SrcMem | NearBranch,			em_call_near_abs),
4157 	I(SrcMemFAddr | ImplicitOps,		em_call_far),
4158 	I(SrcMem | NearBranch,			em_jmp_abs),
4159 	I(SrcMemFAddr | ImplicitOps,		em_jmp_far),
4160 	I(SrcMem | Stack,			em_push), D(Undefined),
4161 };
4162 
4163 static const struct opcode group6[] = {
4164 	DI(Prot | DstMem,	sldt),
4165 	DI(Prot | DstMem,	str),
4166 	II(Prot | Priv | SrcMem16, em_lldt, lldt),
4167 	II(Prot | Priv | SrcMem16, em_ltr, ltr),
4168 	N, N, N, N,
4169 };
4170 
4171 static const struct group_dual group7 = { {
4172 	II(Mov | DstMem,			em_sgdt, sgdt),
4173 	II(Mov | DstMem,			em_sidt, sidt),
4174 	II(SrcMem | Priv,			em_lgdt, lgdt),
4175 	II(SrcMem | Priv,			em_lidt, lidt),
4176 	II(SrcNone | DstMem | Mov,		em_smsw, smsw), N,
4177 	II(SrcMem16 | Mov | Priv,		em_lmsw, lmsw),
4178 	II(SrcMem | ByteOp | Priv | NoAccess,	em_invlpg, invlpg),
4179 }, {
4180 	EXT(0, group7_rm0),
4181 	EXT(0, group7_rm1),
4182 	N, EXT(0, group7_rm3),
4183 	II(SrcNone | DstMem | Mov,		em_smsw, smsw), N,
4184 	II(SrcMem16 | Mov | Priv,		em_lmsw, lmsw),
4185 	EXT(0, group7_rm7),
4186 } };
4187 
4188 static const struct opcode group8[] = {
4189 	N, N, N, N,
4190 	F(DstMem | SrcImmByte | NoWrite,		em_bt),
4191 	F(DstMem | SrcImmByte | Lock | PageTable,	em_bts),
4192 	F(DstMem | SrcImmByte | Lock,			em_btr),
4193 	F(DstMem | SrcImmByte | Lock | PageTable,	em_btc),
4194 };
4195 
4196 static const struct group_dual group9 = { {
4197 	N, I(DstMem64 | Lock | PageTable, em_cmpxchg8b), N, N, N, N, N, N,
4198 }, {
4199 	N, N, N, N, N, N, N, N,
4200 } };
4201 
4202 static const struct opcode group11[] = {
4203 	I(DstMem | SrcImm | Mov | PageTable, em_mov),
4204 	X7(D(Undefined)),
4205 };
4206 
4207 static const struct gprefix pfx_0f_ae_7 = {
4208 	I(SrcMem | ByteOp, em_clflush), N, N, N,
4209 };
4210 
4211 static const struct group_dual group15 = { {
4212 	N, N, N, N, N, N, N, GP(0, &pfx_0f_ae_7),
4213 }, {
4214 	N, N, N, N, N, N, N, N,
4215 } };
4216 
4217 static const struct gprefix pfx_0f_6f_0f_7f = {
4218 	I(Mmx, em_mov), I(Sse | Aligned, em_mov), N, I(Sse | Unaligned, em_mov),
4219 };
4220 
4221 static const struct instr_dual instr_dual_0f_2b = {
4222 	I(0, em_mov), N
4223 };
4224 
4225 static const struct gprefix pfx_0f_2b = {
4226 	ID(0, &instr_dual_0f_2b), ID(0, &instr_dual_0f_2b), N, N,
4227 };
4228 
4229 static const struct gprefix pfx_0f_28_0f_29 = {
4230 	I(Aligned, em_mov), I(Aligned, em_mov), N, N,
4231 };
4232 
4233 static const struct gprefix pfx_0f_e7 = {
4234 	N, I(Sse, em_mov), N, N,
4235 };
4236 
4237 static const struct escape escape_d9 = { {
4238 	N, N, N, N, N, N, N, I(DstMem16 | Mov, em_fnstcw),
4239 }, {
4240 	/* 0xC0 - 0xC7 */
4241 	N, N, N, N, N, N, N, N,
4242 	/* 0xC8 - 0xCF */
4243 	N, N, N, N, N, N, N, N,
4244 	/* 0xD0 - 0xC7 */
4245 	N, N, N, N, N, N, N, N,
4246 	/* 0xD8 - 0xDF */
4247 	N, N, N, N, N, N, N, N,
4248 	/* 0xE0 - 0xE7 */
4249 	N, N, N, N, N, N, N, N,
4250 	/* 0xE8 - 0xEF */
4251 	N, N, N, N, N, N, N, N,
4252 	/* 0xF0 - 0xF7 */
4253 	N, N, N, N, N, N, N, N,
4254 	/* 0xF8 - 0xFF */
4255 	N, N, N, N, N, N, N, N,
4256 } };
4257 
4258 static const struct escape escape_db = { {
4259 	N, N, N, N, N, N, N, N,
4260 }, {
4261 	/* 0xC0 - 0xC7 */
4262 	N, N, N, N, N, N, N, N,
4263 	/* 0xC8 - 0xCF */
4264 	N, N, N, N, N, N, N, N,
4265 	/* 0xD0 - 0xC7 */
4266 	N, N, N, N, N, N, N, N,
4267 	/* 0xD8 - 0xDF */
4268 	N, N, N, N, N, N, N, N,
4269 	/* 0xE0 - 0xE7 */
4270 	N, N, N, I(ImplicitOps, em_fninit), N, N, N, N,
4271 	/* 0xE8 - 0xEF */
4272 	N, N, N, N, N, N, N, N,
4273 	/* 0xF0 - 0xF7 */
4274 	N, N, N, N, N, N, N, N,
4275 	/* 0xF8 - 0xFF */
4276 	N, N, N, N, N, N, N, N,
4277 } };
4278 
4279 static const struct escape escape_dd = { {
4280 	N, N, N, N, N, N, N, I(DstMem16 | Mov, em_fnstsw),
4281 }, {
4282 	/* 0xC0 - 0xC7 */
4283 	N, N, N, N, N, N, N, N,
4284 	/* 0xC8 - 0xCF */
4285 	N, N, N, N, N, N, N, N,
4286 	/* 0xD0 - 0xC7 */
4287 	N, N, N, N, N, N, N, N,
4288 	/* 0xD8 - 0xDF */
4289 	N, N, N, N, N, N, N, N,
4290 	/* 0xE0 - 0xE7 */
4291 	N, N, N, N, N, N, N, N,
4292 	/* 0xE8 - 0xEF */
4293 	N, N, N, N, N, N, N, N,
4294 	/* 0xF0 - 0xF7 */
4295 	N, N, N, N, N, N, N, N,
4296 	/* 0xF8 - 0xFF */
4297 	N, N, N, N, N, N, N, N,
4298 } };
4299 
4300 static const struct instr_dual instr_dual_0f_c3 = {
4301 	I(DstMem | SrcReg | ModRM | No16 | Mov, em_mov), N
4302 };
4303 
4304 static const struct mode_dual mode_dual_63 = {
4305 	N, I(DstReg | SrcMem32 | ModRM | Mov, em_movsxd)
4306 };
4307 
4308 static const struct opcode opcode_table[256] = {
4309 	/* 0x00 - 0x07 */
4310 	F6ALU(Lock, em_add),
4311 	I(ImplicitOps | Stack | No64 | Src2ES, em_push_sreg),
4312 	I(ImplicitOps | Stack | No64 | Src2ES, em_pop_sreg),
4313 	/* 0x08 - 0x0F */
4314 	F6ALU(Lock | PageTable, em_or),
4315 	I(ImplicitOps | Stack | No64 | Src2CS, em_push_sreg),
4316 	N,
4317 	/* 0x10 - 0x17 */
4318 	F6ALU(Lock, em_adc),
4319 	I(ImplicitOps | Stack | No64 | Src2SS, em_push_sreg),
4320 	I(ImplicitOps | Stack | No64 | Src2SS, em_pop_sreg),
4321 	/* 0x18 - 0x1F */
4322 	F6ALU(Lock, em_sbb),
4323 	I(ImplicitOps | Stack | No64 | Src2DS, em_push_sreg),
4324 	I(ImplicitOps | Stack | No64 | Src2DS, em_pop_sreg),
4325 	/* 0x20 - 0x27 */
4326 	F6ALU(Lock | PageTable, em_and), N, N,
4327 	/* 0x28 - 0x2F */
4328 	F6ALU(Lock, em_sub), N, I(ByteOp | DstAcc | No64, em_das),
4329 	/* 0x30 - 0x37 */
4330 	F6ALU(Lock, em_xor), N, N,
4331 	/* 0x38 - 0x3F */
4332 	F6ALU(NoWrite, em_cmp), N, N,
4333 	/* 0x40 - 0x4F */
4334 	X8(F(DstReg, em_inc)), X8(F(DstReg, em_dec)),
4335 	/* 0x50 - 0x57 */
4336 	X8(I(SrcReg | Stack, em_push)),
4337 	/* 0x58 - 0x5F */
4338 	X8(I(DstReg | Stack, em_pop)),
4339 	/* 0x60 - 0x67 */
4340 	I(ImplicitOps | Stack | No64, em_pusha),
4341 	I(ImplicitOps | Stack | No64, em_popa),
4342 	N, MD(ModRM, &mode_dual_63),
4343 	N, N, N, N,
4344 	/* 0x68 - 0x6F */
4345 	I(SrcImm | Mov | Stack, em_push),
4346 	I(DstReg | SrcMem | ModRM | Src2Imm, em_imul_3op),
4347 	I(SrcImmByte | Mov | Stack, em_push),
4348 	I(DstReg | SrcMem | ModRM | Src2ImmByte, em_imul_3op),
4349 	I2bvIP(DstDI | SrcDX | Mov | String | Unaligned, em_in, ins, check_perm_in), /* insb, insw/insd */
4350 	I2bvIP(SrcSI | DstDX | String, em_out, outs, check_perm_out), /* outsb, outsw/outsd */
4351 	/* 0x70 - 0x7F */
4352 	X16(D(SrcImmByte | NearBranch)),
4353 	/* 0x80 - 0x87 */
4354 	G(ByteOp | DstMem | SrcImm, group1),
4355 	G(DstMem | SrcImm, group1),
4356 	G(ByteOp | DstMem | SrcImm | No64, group1),
4357 	G(DstMem | SrcImmByte, group1),
4358 	F2bv(DstMem | SrcReg | ModRM | NoWrite, em_test),
4359 	I2bv(DstMem | SrcReg | ModRM | Lock | PageTable, em_xchg),
4360 	/* 0x88 - 0x8F */
4361 	I2bv(DstMem | SrcReg | ModRM | Mov | PageTable, em_mov),
4362 	I2bv(DstReg | SrcMem | ModRM | Mov, em_mov),
4363 	I(DstMem | SrcNone | ModRM | Mov | PageTable, em_mov_rm_sreg),
4364 	D(ModRM | SrcMem | NoAccess | DstReg),
4365 	I(ImplicitOps | SrcMem16 | ModRM, em_mov_sreg_rm),
4366 	G(0, group1A),
4367 	/* 0x90 - 0x97 */
4368 	DI(SrcAcc | DstReg, pause), X7(D(SrcAcc | DstReg)),
4369 	/* 0x98 - 0x9F */
4370 	D(DstAcc | SrcNone), I(ImplicitOps | SrcAcc, em_cwd),
4371 	I(SrcImmFAddr | No64, em_call_far), N,
4372 	II(ImplicitOps | Stack, em_pushf, pushf),
4373 	II(ImplicitOps | Stack, em_popf, popf),
4374 	I(ImplicitOps, em_sahf), I(ImplicitOps, em_lahf),
4375 	/* 0xA0 - 0xA7 */
4376 	I2bv(DstAcc | SrcMem | Mov | MemAbs, em_mov),
4377 	I2bv(DstMem | SrcAcc | Mov | MemAbs | PageTable, em_mov),
4378 	I2bv(SrcSI | DstDI | Mov | String, em_mov),
4379 	F2bv(SrcSI | DstDI | String | NoWrite, em_cmp_r),
4380 	/* 0xA8 - 0xAF */
4381 	F2bv(DstAcc | SrcImm | NoWrite, em_test),
4382 	I2bv(SrcAcc | DstDI | Mov | String, em_mov),
4383 	I2bv(SrcSI | DstAcc | Mov | String, em_mov),
4384 	F2bv(SrcAcc | DstDI | String | NoWrite, em_cmp_r),
4385 	/* 0xB0 - 0xB7 */
4386 	X8(I(ByteOp | DstReg | SrcImm | Mov, em_mov)),
4387 	/* 0xB8 - 0xBF */
4388 	X8(I(DstReg | SrcImm64 | Mov, em_mov)),
4389 	/* 0xC0 - 0xC7 */
4390 	G(ByteOp | Src2ImmByte, group2), G(Src2ImmByte, group2),
4391 	I(ImplicitOps | NearBranch | SrcImmU16, em_ret_near_imm),
4392 	I(ImplicitOps | NearBranch, em_ret),
4393 	I(DstReg | SrcMemFAddr | ModRM | No64 | Src2ES, em_lseg),
4394 	I(DstReg | SrcMemFAddr | ModRM | No64 | Src2DS, em_lseg),
4395 	G(ByteOp, group11), G(0, group11),
4396 	/* 0xC8 - 0xCF */
4397 	I(Stack | SrcImmU16 | Src2ImmByte, em_enter), I(Stack, em_leave),
4398 	I(ImplicitOps | SrcImmU16, em_ret_far_imm),
4399 	I(ImplicitOps, em_ret_far),
4400 	D(ImplicitOps), DI(SrcImmByte, intn),
4401 	D(ImplicitOps | No64), II(ImplicitOps, em_iret, iret),
4402 	/* 0xD0 - 0xD7 */
4403 	G(Src2One | ByteOp, group2), G(Src2One, group2),
4404 	G(Src2CL | ByteOp, group2), G(Src2CL, group2),
4405 	I(DstAcc | SrcImmUByte | No64, em_aam),
4406 	I(DstAcc | SrcImmUByte | No64, em_aad),
4407 	F(DstAcc | ByteOp | No64, em_salc),
4408 	I(DstAcc | SrcXLat | ByteOp, em_mov),
4409 	/* 0xD8 - 0xDF */
4410 	N, E(0, &escape_d9), N, E(0, &escape_db), N, E(0, &escape_dd), N, N,
4411 	/* 0xE0 - 0xE7 */
4412 	X3(I(SrcImmByte | NearBranch, em_loop)),
4413 	I(SrcImmByte | NearBranch, em_jcxz),
4414 	I2bvIP(SrcImmUByte | DstAcc, em_in,  in,  check_perm_in),
4415 	I2bvIP(SrcAcc | DstImmUByte, em_out, out, check_perm_out),
4416 	/* 0xE8 - 0xEF */
4417 	I(SrcImm | NearBranch, em_call), D(SrcImm | ImplicitOps | NearBranch),
4418 	I(SrcImmFAddr | No64, em_jmp_far),
4419 	D(SrcImmByte | ImplicitOps | NearBranch),
4420 	I2bvIP(SrcDX | DstAcc, em_in,  in,  check_perm_in),
4421 	I2bvIP(SrcAcc | DstDX, em_out, out, check_perm_out),
4422 	/* 0xF0 - 0xF7 */
4423 	N, DI(ImplicitOps, icebp), N, N,
4424 	DI(ImplicitOps | Priv, hlt), D(ImplicitOps),
4425 	G(ByteOp, group3), G(0, group3),
4426 	/* 0xF8 - 0xFF */
4427 	D(ImplicitOps), D(ImplicitOps),
4428 	I(ImplicitOps, em_cli), I(ImplicitOps, em_sti),
4429 	D(ImplicitOps), D(ImplicitOps), G(0, group4), G(0, group5),
4430 };
4431 
4432 static const struct opcode twobyte_table[256] = {
4433 	/* 0x00 - 0x0F */
4434 	G(0, group6), GD(0, &group7), N, N,
4435 	N, I(ImplicitOps | EmulateOnUD, em_syscall),
4436 	II(ImplicitOps | Priv, em_clts, clts), N,
4437 	DI(ImplicitOps | Priv, invd), DI(ImplicitOps | Priv, wbinvd), N, N,
4438 	N, D(ImplicitOps | ModRM | SrcMem | NoAccess), N, N,
4439 	/* 0x10 - 0x1F */
4440 	N, N, N, N, N, N, N, N,
4441 	D(ImplicitOps | ModRM | SrcMem | NoAccess),
4442 	N, N, N, N, N, N, D(ImplicitOps | ModRM | SrcMem | NoAccess),
4443 	/* 0x20 - 0x2F */
4444 	DIP(ModRM | DstMem | Priv | Op3264 | NoMod, cr_read, check_cr_read),
4445 	DIP(ModRM | DstMem | Priv | Op3264 | NoMod, dr_read, check_dr_read),
4446 	IIP(ModRM | SrcMem | Priv | Op3264 | NoMod, em_cr_write, cr_write,
4447 						check_cr_write),
4448 	IIP(ModRM | SrcMem | Priv | Op3264 | NoMod, em_dr_write, dr_write,
4449 						check_dr_write),
4450 	N, N, N, N,
4451 	GP(ModRM | DstReg | SrcMem | Mov | Sse, &pfx_0f_28_0f_29),
4452 	GP(ModRM | DstMem | SrcReg | Mov | Sse, &pfx_0f_28_0f_29),
4453 	N, GP(ModRM | DstMem | SrcReg | Mov | Sse, &pfx_0f_2b),
4454 	N, N, N, N,
4455 	/* 0x30 - 0x3F */
4456 	II(ImplicitOps | Priv, em_wrmsr, wrmsr),
4457 	IIP(ImplicitOps, em_rdtsc, rdtsc, check_rdtsc),
4458 	II(ImplicitOps | Priv, em_rdmsr, rdmsr),
4459 	IIP(ImplicitOps, em_rdpmc, rdpmc, check_rdpmc),
4460 	I(ImplicitOps | EmulateOnUD, em_sysenter),
4461 	I(ImplicitOps | Priv | EmulateOnUD, em_sysexit),
4462 	N, N,
4463 	N, N, N, N, N, N, N, N,
4464 	/* 0x40 - 0x4F */
4465 	X16(D(DstReg | SrcMem | ModRM)),
4466 	/* 0x50 - 0x5F */
4467 	N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
4468 	/* 0x60 - 0x6F */
4469 	N, N, N, N,
4470 	N, N, N, N,
4471 	N, N, N, N,
4472 	N, N, N, GP(SrcMem | DstReg | ModRM | Mov, &pfx_0f_6f_0f_7f),
4473 	/* 0x70 - 0x7F */
4474 	N, N, N, N,
4475 	N, N, N, N,
4476 	N, N, N, N,
4477 	N, N, N, GP(SrcReg | DstMem | ModRM | Mov, &pfx_0f_6f_0f_7f),
4478 	/* 0x80 - 0x8F */
4479 	X16(D(SrcImm | NearBranch)),
4480 	/* 0x90 - 0x9F */
4481 	X16(D(ByteOp | DstMem | SrcNone | ModRM| Mov)),
4482 	/* 0xA0 - 0xA7 */
4483 	I(Stack | Src2FS, em_push_sreg), I(Stack | Src2FS, em_pop_sreg),
4484 	II(ImplicitOps, em_cpuid, cpuid),
4485 	F(DstMem | SrcReg | ModRM | BitOp | NoWrite, em_bt),
4486 	F(DstMem | SrcReg | Src2ImmByte | ModRM, em_shld),
4487 	F(DstMem | SrcReg | Src2CL | ModRM, em_shld), N, N,
4488 	/* 0xA8 - 0xAF */
4489 	I(Stack | Src2GS, em_push_sreg), I(Stack | Src2GS, em_pop_sreg),
4490 	II(EmulateOnUD | ImplicitOps, em_rsm, rsm),
4491 	F(DstMem | SrcReg | ModRM | BitOp | Lock | PageTable, em_bts),
4492 	F(DstMem | SrcReg | Src2ImmByte | ModRM, em_shrd),
4493 	F(DstMem | SrcReg | Src2CL | ModRM, em_shrd),
4494 	GD(0, &group15), F(DstReg | SrcMem | ModRM, em_imul),
4495 	/* 0xB0 - 0xB7 */
4496 	I2bv(DstMem | SrcReg | ModRM | Lock | PageTable | SrcWrite, em_cmpxchg),
4497 	I(DstReg | SrcMemFAddr | ModRM | Src2SS, em_lseg),
4498 	F(DstMem | SrcReg | ModRM | BitOp | Lock, em_btr),
4499 	I(DstReg | SrcMemFAddr | ModRM | Src2FS, em_lseg),
4500 	I(DstReg | SrcMemFAddr | ModRM | Src2GS, em_lseg),
4501 	D(DstReg | SrcMem8 | ModRM | Mov), D(DstReg | SrcMem16 | ModRM | Mov),
4502 	/* 0xB8 - 0xBF */
4503 	N, N,
4504 	G(BitOp, group8),
4505 	F(DstMem | SrcReg | ModRM | BitOp | Lock | PageTable, em_btc),
4506 	I(DstReg | SrcMem | ModRM, em_bsf_c),
4507 	I(DstReg | SrcMem | ModRM, em_bsr_c),
4508 	D(DstReg | SrcMem8 | ModRM | Mov), D(DstReg | SrcMem16 | ModRM | Mov),
4509 	/* 0xC0 - 0xC7 */
4510 	F2bv(DstMem | SrcReg | ModRM | SrcWrite | Lock, em_xadd),
4511 	N, ID(0, &instr_dual_0f_c3),
4512 	N, N, N, GD(0, &group9),
4513 	/* 0xC8 - 0xCF */
4514 	X8(I(DstReg, em_bswap)),
4515 	/* 0xD0 - 0xDF */
4516 	N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
4517 	/* 0xE0 - 0xEF */
4518 	N, N, N, N, N, N, N, GP(SrcReg | DstMem | ModRM | Mov, &pfx_0f_e7),
4519 	N, N, N, N, N, N, N, N,
4520 	/* 0xF0 - 0xFF */
4521 	N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N
4522 };
4523 
4524 static const struct instr_dual instr_dual_0f_38_f0 = {
4525 	I(DstReg | SrcMem | Mov, em_movbe), N
4526 };
4527 
4528 static const struct instr_dual instr_dual_0f_38_f1 = {
4529 	I(DstMem | SrcReg | Mov, em_movbe), N
4530 };
4531 
4532 static const struct gprefix three_byte_0f_38_f0 = {
4533 	ID(0, &instr_dual_0f_38_f0), N, N, N
4534 };
4535 
4536 static const struct gprefix three_byte_0f_38_f1 = {
4537 	ID(0, &instr_dual_0f_38_f1), N, N, N
4538 };
4539 
4540 /*
4541  * Insns below are selected by the prefix which indexed by the third opcode
4542  * byte.
4543  */
4544 static const struct opcode opcode_map_0f_38[256] = {
4545 	/* 0x00 - 0x7f */
4546 	X16(N), X16(N), X16(N), X16(N), X16(N), X16(N), X16(N), X16(N),
4547 	/* 0x80 - 0xef */
4548 	X16(N), X16(N), X16(N), X16(N), X16(N), X16(N), X16(N),
4549 	/* 0xf0 - 0xf1 */
4550 	GP(EmulateOnUD | ModRM, &three_byte_0f_38_f0),
4551 	GP(EmulateOnUD | ModRM, &three_byte_0f_38_f1),
4552 	/* 0xf2 - 0xff */
4553 	N, N, X4(N), X8(N)
4554 };
4555 
4556 #undef D
4557 #undef N
4558 #undef G
4559 #undef GD
4560 #undef I
4561 #undef GP
4562 #undef EXT
4563 #undef MD
4564 #undef ID
4565 
4566 #undef D2bv
4567 #undef D2bvIP
4568 #undef I2bv
4569 #undef I2bvIP
4570 #undef I6ALU
4571 
4572 static unsigned imm_size(struct x86_emulate_ctxt *ctxt)
4573 {
4574 	unsigned size;
4575 
4576 	size = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4577 	if (size == 8)
4578 		size = 4;
4579 	return size;
4580 }
4581 
4582 static int decode_imm(struct x86_emulate_ctxt *ctxt, struct operand *op,
4583 		      unsigned size, bool sign_extension)
4584 {
4585 	int rc = X86EMUL_CONTINUE;
4586 
4587 	op->type = OP_IMM;
4588 	op->bytes = size;
4589 	op->addr.mem.ea = ctxt->_eip;
4590 	/* NB. Immediates are sign-extended as necessary. */
4591 	switch (op->bytes) {
4592 	case 1:
4593 		op->val = insn_fetch(s8, ctxt);
4594 		break;
4595 	case 2:
4596 		op->val = insn_fetch(s16, ctxt);
4597 		break;
4598 	case 4:
4599 		op->val = insn_fetch(s32, ctxt);
4600 		break;
4601 	case 8:
4602 		op->val = insn_fetch(s64, ctxt);
4603 		break;
4604 	}
4605 	if (!sign_extension) {
4606 		switch (op->bytes) {
4607 		case 1:
4608 			op->val &= 0xff;
4609 			break;
4610 		case 2:
4611 			op->val &= 0xffff;
4612 			break;
4613 		case 4:
4614 			op->val &= 0xffffffff;
4615 			break;
4616 		}
4617 	}
4618 done:
4619 	return rc;
4620 }
4621 
4622 static int decode_operand(struct x86_emulate_ctxt *ctxt, struct operand *op,
4623 			  unsigned d)
4624 {
4625 	int rc = X86EMUL_CONTINUE;
4626 
4627 	switch (d) {
4628 	case OpReg:
4629 		decode_register_operand(ctxt, op);
4630 		break;
4631 	case OpImmUByte:
4632 		rc = decode_imm(ctxt, op, 1, false);
4633 		break;
4634 	case OpMem:
4635 		ctxt->memop.bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4636 	mem_common:
4637 		*op = ctxt->memop;
4638 		ctxt->memopp = op;
4639 		if (ctxt->d & BitOp)
4640 			fetch_bit_operand(ctxt);
4641 		op->orig_val = op->val;
4642 		break;
4643 	case OpMem64:
4644 		ctxt->memop.bytes = (ctxt->op_bytes == 8) ? 16 : 8;
4645 		goto mem_common;
4646 	case OpAcc:
4647 		op->type = OP_REG;
4648 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4649 		op->addr.reg = reg_rmw(ctxt, VCPU_REGS_RAX);
4650 		fetch_register_operand(op);
4651 		op->orig_val = op->val;
4652 		break;
4653 	case OpAccLo:
4654 		op->type = OP_REG;
4655 		op->bytes = (ctxt->d & ByteOp) ? 2 : ctxt->op_bytes;
4656 		op->addr.reg = reg_rmw(ctxt, VCPU_REGS_RAX);
4657 		fetch_register_operand(op);
4658 		op->orig_val = op->val;
4659 		break;
4660 	case OpAccHi:
4661 		if (ctxt->d & ByteOp) {
4662 			op->type = OP_NONE;
4663 			break;
4664 		}
4665 		op->type = OP_REG;
4666 		op->bytes = ctxt->op_bytes;
4667 		op->addr.reg = reg_rmw(ctxt, VCPU_REGS_RDX);
4668 		fetch_register_operand(op);
4669 		op->orig_val = op->val;
4670 		break;
4671 	case OpDI:
4672 		op->type = OP_MEM;
4673 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4674 		op->addr.mem.ea =
4675 			register_address(ctxt, VCPU_REGS_RDI);
4676 		op->addr.mem.seg = VCPU_SREG_ES;
4677 		op->val = 0;
4678 		op->count = 1;
4679 		break;
4680 	case OpDX:
4681 		op->type = OP_REG;
4682 		op->bytes = 2;
4683 		op->addr.reg = reg_rmw(ctxt, VCPU_REGS_RDX);
4684 		fetch_register_operand(op);
4685 		break;
4686 	case OpCL:
4687 		op->type = OP_IMM;
4688 		op->bytes = 1;
4689 		op->val = reg_read(ctxt, VCPU_REGS_RCX) & 0xff;
4690 		break;
4691 	case OpImmByte:
4692 		rc = decode_imm(ctxt, op, 1, true);
4693 		break;
4694 	case OpOne:
4695 		op->type = OP_IMM;
4696 		op->bytes = 1;
4697 		op->val = 1;
4698 		break;
4699 	case OpImm:
4700 		rc = decode_imm(ctxt, op, imm_size(ctxt), true);
4701 		break;
4702 	case OpImm64:
4703 		rc = decode_imm(ctxt, op, ctxt->op_bytes, true);
4704 		break;
4705 	case OpMem8:
4706 		ctxt->memop.bytes = 1;
4707 		if (ctxt->memop.type == OP_REG) {
4708 			ctxt->memop.addr.reg = decode_register(ctxt,
4709 					ctxt->modrm_rm, true);
4710 			fetch_register_operand(&ctxt->memop);
4711 		}
4712 		goto mem_common;
4713 	case OpMem16:
4714 		ctxt->memop.bytes = 2;
4715 		goto mem_common;
4716 	case OpMem32:
4717 		ctxt->memop.bytes = 4;
4718 		goto mem_common;
4719 	case OpImmU16:
4720 		rc = decode_imm(ctxt, op, 2, false);
4721 		break;
4722 	case OpImmU:
4723 		rc = decode_imm(ctxt, op, imm_size(ctxt), false);
4724 		break;
4725 	case OpSI:
4726 		op->type = OP_MEM;
4727 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4728 		op->addr.mem.ea =
4729 			register_address(ctxt, VCPU_REGS_RSI);
4730 		op->addr.mem.seg = ctxt->seg_override;
4731 		op->val = 0;
4732 		op->count = 1;
4733 		break;
4734 	case OpXLat:
4735 		op->type = OP_MEM;
4736 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4737 		op->addr.mem.ea =
4738 			address_mask(ctxt,
4739 				reg_read(ctxt, VCPU_REGS_RBX) +
4740 				(reg_read(ctxt, VCPU_REGS_RAX) & 0xff));
4741 		op->addr.mem.seg = ctxt->seg_override;
4742 		op->val = 0;
4743 		break;
4744 	case OpImmFAddr:
4745 		op->type = OP_IMM;
4746 		op->addr.mem.ea = ctxt->_eip;
4747 		op->bytes = ctxt->op_bytes + 2;
4748 		insn_fetch_arr(op->valptr, op->bytes, ctxt);
4749 		break;
4750 	case OpMemFAddr:
4751 		ctxt->memop.bytes = ctxt->op_bytes + 2;
4752 		goto mem_common;
4753 	case OpES:
4754 		op->type = OP_IMM;
4755 		op->val = VCPU_SREG_ES;
4756 		break;
4757 	case OpCS:
4758 		op->type = OP_IMM;
4759 		op->val = VCPU_SREG_CS;
4760 		break;
4761 	case OpSS:
4762 		op->type = OP_IMM;
4763 		op->val = VCPU_SREG_SS;
4764 		break;
4765 	case OpDS:
4766 		op->type = OP_IMM;
4767 		op->val = VCPU_SREG_DS;
4768 		break;
4769 	case OpFS:
4770 		op->type = OP_IMM;
4771 		op->val = VCPU_SREG_FS;
4772 		break;
4773 	case OpGS:
4774 		op->type = OP_IMM;
4775 		op->val = VCPU_SREG_GS;
4776 		break;
4777 	case OpImplicit:
4778 		/* Special instructions do their own operand decoding. */
4779 	default:
4780 		op->type = OP_NONE; /* Disable writeback. */
4781 		break;
4782 	}
4783 
4784 done:
4785 	return rc;
4786 }
4787 
4788 int x86_decode_insn(struct x86_emulate_ctxt *ctxt, void *insn, int insn_len)
4789 {
4790 	int rc = X86EMUL_CONTINUE;
4791 	int mode = ctxt->mode;
4792 	int def_op_bytes, def_ad_bytes, goffset, simd_prefix;
4793 	bool op_prefix = false;
4794 	bool has_seg_override = false;
4795 	struct opcode opcode;
4796 
4797 	ctxt->memop.type = OP_NONE;
4798 	ctxt->memopp = NULL;
4799 	ctxt->_eip = ctxt->eip;
4800 	ctxt->fetch.ptr = ctxt->fetch.data;
4801 	ctxt->fetch.end = ctxt->fetch.data + insn_len;
4802 	ctxt->opcode_len = 1;
4803 	if (insn_len > 0)
4804 		memcpy(ctxt->fetch.data, insn, insn_len);
4805 	else {
4806 		rc = __do_insn_fetch_bytes(ctxt, 1);
4807 		if (rc != X86EMUL_CONTINUE)
4808 			return rc;
4809 	}
4810 
4811 	switch (mode) {
4812 	case X86EMUL_MODE_REAL:
4813 	case X86EMUL_MODE_VM86:
4814 	case X86EMUL_MODE_PROT16:
4815 		def_op_bytes = def_ad_bytes = 2;
4816 		break;
4817 	case X86EMUL_MODE_PROT32:
4818 		def_op_bytes = def_ad_bytes = 4;
4819 		break;
4820 #ifdef CONFIG_X86_64
4821 	case X86EMUL_MODE_PROT64:
4822 		def_op_bytes = 4;
4823 		def_ad_bytes = 8;
4824 		break;
4825 #endif
4826 	default:
4827 		return EMULATION_FAILED;
4828 	}
4829 
4830 	ctxt->op_bytes = def_op_bytes;
4831 	ctxt->ad_bytes = def_ad_bytes;
4832 
4833 	/* Legacy prefixes. */
4834 	for (;;) {
4835 		switch (ctxt->b = insn_fetch(u8, ctxt)) {
4836 		case 0x66:	/* operand-size override */
4837 			op_prefix = true;
4838 			/* switch between 2/4 bytes */
4839 			ctxt->op_bytes = def_op_bytes ^ 6;
4840 			break;
4841 		case 0x67:	/* address-size override */
4842 			if (mode == X86EMUL_MODE_PROT64)
4843 				/* switch between 4/8 bytes */
4844 				ctxt->ad_bytes = def_ad_bytes ^ 12;
4845 			else
4846 				/* switch between 2/4 bytes */
4847 				ctxt->ad_bytes = def_ad_bytes ^ 6;
4848 			break;
4849 		case 0x26:	/* ES override */
4850 		case 0x2e:	/* CS override */
4851 		case 0x36:	/* SS override */
4852 		case 0x3e:	/* DS override */
4853 			has_seg_override = true;
4854 			ctxt->seg_override = (ctxt->b >> 3) & 3;
4855 			break;
4856 		case 0x64:	/* FS override */
4857 		case 0x65:	/* GS override */
4858 			has_seg_override = true;
4859 			ctxt->seg_override = ctxt->b & 7;
4860 			break;
4861 		case 0x40 ... 0x4f: /* REX */
4862 			if (mode != X86EMUL_MODE_PROT64)
4863 				goto done_prefixes;
4864 			ctxt->rex_prefix = ctxt->b;
4865 			continue;
4866 		case 0xf0:	/* LOCK */
4867 			ctxt->lock_prefix = 1;
4868 			break;
4869 		case 0xf2:	/* REPNE/REPNZ */
4870 		case 0xf3:	/* REP/REPE/REPZ */
4871 			ctxt->rep_prefix = ctxt->b;
4872 			break;
4873 		default:
4874 			goto done_prefixes;
4875 		}
4876 
4877 		/* Any legacy prefix after a REX prefix nullifies its effect. */
4878 
4879 		ctxt->rex_prefix = 0;
4880 	}
4881 
4882 done_prefixes:
4883 
4884 	/* REX prefix. */
4885 	if (ctxt->rex_prefix & 8)
4886 		ctxt->op_bytes = 8;	/* REX.W */
4887 
4888 	/* Opcode byte(s). */
4889 	opcode = opcode_table[ctxt->b];
4890 	/* Two-byte opcode? */
4891 	if (ctxt->b == 0x0f) {
4892 		ctxt->opcode_len = 2;
4893 		ctxt->b = insn_fetch(u8, ctxt);
4894 		opcode = twobyte_table[ctxt->b];
4895 
4896 		/* 0F_38 opcode map */
4897 		if (ctxt->b == 0x38) {
4898 			ctxt->opcode_len = 3;
4899 			ctxt->b = insn_fetch(u8, ctxt);
4900 			opcode = opcode_map_0f_38[ctxt->b];
4901 		}
4902 	}
4903 	ctxt->d = opcode.flags;
4904 
4905 	if (ctxt->d & ModRM)
4906 		ctxt->modrm = insn_fetch(u8, ctxt);
4907 
4908 	/* vex-prefix instructions are not implemented */
4909 	if (ctxt->opcode_len == 1 && (ctxt->b == 0xc5 || ctxt->b == 0xc4) &&
4910 	    (mode == X86EMUL_MODE_PROT64 || (ctxt->modrm & 0xc0) == 0xc0)) {
4911 		ctxt->d = NotImpl;
4912 	}
4913 
4914 	while (ctxt->d & GroupMask) {
4915 		switch (ctxt->d & GroupMask) {
4916 		case Group:
4917 			goffset = (ctxt->modrm >> 3) & 7;
4918 			opcode = opcode.u.group[goffset];
4919 			break;
4920 		case GroupDual:
4921 			goffset = (ctxt->modrm >> 3) & 7;
4922 			if ((ctxt->modrm >> 6) == 3)
4923 				opcode = opcode.u.gdual->mod3[goffset];
4924 			else
4925 				opcode = opcode.u.gdual->mod012[goffset];
4926 			break;
4927 		case RMExt:
4928 			goffset = ctxt->modrm & 7;
4929 			opcode = opcode.u.group[goffset];
4930 			break;
4931 		case Prefix:
4932 			if (ctxt->rep_prefix && op_prefix)
4933 				return EMULATION_FAILED;
4934 			simd_prefix = op_prefix ? 0x66 : ctxt->rep_prefix;
4935 			switch (simd_prefix) {
4936 			case 0x00: opcode = opcode.u.gprefix->pfx_no; break;
4937 			case 0x66: opcode = opcode.u.gprefix->pfx_66; break;
4938 			case 0xf2: opcode = opcode.u.gprefix->pfx_f2; break;
4939 			case 0xf3: opcode = opcode.u.gprefix->pfx_f3; break;
4940 			}
4941 			break;
4942 		case Escape:
4943 			if (ctxt->modrm > 0xbf)
4944 				opcode = opcode.u.esc->high[ctxt->modrm - 0xc0];
4945 			else
4946 				opcode = opcode.u.esc->op[(ctxt->modrm >> 3) & 7];
4947 			break;
4948 		case InstrDual:
4949 			if ((ctxt->modrm >> 6) == 3)
4950 				opcode = opcode.u.idual->mod3;
4951 			else
4952 				opcode = opcode.u.idual->mod012;
4953 			break;
4954 		case ModeDual:
4955 			if (ctxt->mode == X86EMUL_MODE_PROT64)
4956 				opcode = opcode.u.mdual->mode64;
4957 			else
4958 				opcode = opcode.u.mdual->mode32;
4959 			break;
4960 		default:
4961 			return EMULATION_FAILED;
4962 		}
4963 
4964 		ctxt->d &= ~(u64)GroupMask;
4965 		ctxt->d |= opcode.flags;
4966 	}
4967 
4968 	/* Unrecognised? */
4969 	if (ctxt->d == 0)
4970 		return EMULATION_FAILED;
4971 
4972 	ctxt->execute = opcode.u.execute;
4973 
4974 	if (unlikely(ctxt->ud) && likely(!(ctxt->d & EmulateOnUD)))
4975 		return EMULATION_FAILED;
4976 
4977 	if (unlikely(ctxt->d &
4978 	    (NotImpl|Stack|Op3264|Sse|Mmx|Intercept|CheckPerm|NearBranch|
4979 	     No16))) {
4980 		/*
4981 		 * These are copied unconditionally here, and checked unconditionally
4982 		 * in x86_emulate_insn.
4983 		 */
4984 		ctxt->check_perm = opcode.check_perm;
4985 		ctxt->intercept = opcode.intercept;
4986 
4987 		if (ctxt->d & NotImpl)
4988 			return EMULATION_FAILED;
4989 
4990 		if (mode == X86EMUL_MODE_PROT64) {
4991 			if (ctxt->op_bytes == 4 && (ctxt->d & Stack))
4992 				ctxt->op_bytes = 8;
4993 			else if (ctxt->d & NearBranch)
4994 				ctxt->op_bytes = 8;
4995 		}
4996 
4997 		if (ctxt->d & Op3264) {
4998 			if (mode == X86EMUL_MODE_PROT64)
4999 				ctxt->op_bytes = 8;
5000 			else
5001 				ctxt->op_bytes = 4;
5002 		}
5003 
5004 		if ((ctxt->d & No16) && ctxt->op_bytes == 2)
5005 			ctxt->op_bytes = 4;
5006 
5007 		if (ctxt->d & Sse)
5008 			ctxt->op_bytes = 16;
5009 		else if (ctxt->d & Mmx)
5010 			ctxt->op_bytes = 8;
5011 	}
5012 
5013 	/* ModRM and SIB bytes. */
5014 	if (ctxt->d & ModRM) {
5015 		rc = decode_modrm(ctxt, &ctxt->memop);
5016 		if (!has_seg_override) {
5017 			has_seg_override = true;
5018 			ctxt->seg_override = ctxt->modrm_seg;
5019 		}
5020 	} else if (ctxt->d & MemAbs)
5021 		rc = decode_abs(ctxt, &ctxt->memop);
5022 	if (rc != X86EMUL_CONTINUE)
5023 		goto done;
5024 
5025 	if (!has_seg_override)
5026 		ctxt->seg_override = VCPU_SREG_DS;
5027 
5028 	ctxt->memop.addr.mem.seg = ctxt->seg_override;
5029 
5030 	/*
5031 	 * Decode and fetch the source operand: register, memory
5032 	 * or immediate.
5033 	 */
5034 	rc = decode_operand(ctxt, &ctxt->src, (ctxt->d >> SrcShift) & OpMask);
5035 	if (rc != X86EMUL_CONTINUE)
5036 		goto done;
5037 
5038 	/*
5039 	 * Decode and fetch the second source operand: register, memory
5040 	 * or immediate.
5041 	 */
5042 	rc = decode_operand(ctxt, &ctxt->src2, (ctxt->d >> Src2Shift) & OpMask);
5043 	if (rc != X86EMUL_CONTINUE)
5044 		goto done;
5045 
5046 	/* Decode and fetch the destination operand: register or memory. */
5047 	rc = decode_operand(ctxt, &ctxt->dst, (ctxt->d >> DstShift) & OpMask);
5048 
5049 	if (ctxt->rip_relative)
5050 		ctxt->memopp->addr.mem.ea = address_mask(ctxt,
5051 					ctxt->memopp->addr.mem.ea + ctxt->_eip);
5052 
5053 done:
5054 	return (rc != X86EMUL_CONTINUE) ? EMULATION_FAILED : EMULATION_OK;
5055 }
5056 
5057 bool x86_page_table_writing_insn(struct x86_emulate_ctxt *ctxt)
5058 {
5059 	return ctxt->d & PageTable;
5060 }
5061 
5062 static bool string_insn_completed(struct x86_emulate_ctxt *ctxt)
5063 {
5064 	/* The second termination condition only applies for REPE
5065 	 * and REPNE. Test if the repeat string operation prefix is
5066 	 * REPE/REPZ or REPNE/REPNZ and if it's the case it tests the
5067 	 * corresponding termination condition according to:
5068 	 * 	- if REPE/REPZ and ZF = 0 then done
5069 	 * 	- if REPNE/REPNZ and ZF = 1 then done
5070 	 */
5071 	if (((ctxt->b == 0xa6) || (ctxt->b == 0xa7) ||
5072 	     (ctxt->b == 0xae) || (ctxt->b == 0xaf))
5073 	    && (((ctxt->rep_prefix == REPE_PREFIX) &&
5074 		 ((ctxt->eflags & X86_EFLAGS_ZF) == 0))
5075 		|| ((ctxt->rep_prefix == REPNE_PREFIX) &&
5076 		    ((ctxt->eflags & X86_EFLAGS_ZF) == X86_EFLAGS_ZF))))
5077 		return true;
5078 
5079 	return false;
5080 }
5081 
5082 static int flush_pending_x87_faults(struct x86_emulate_ctxt *ctxt)
5083 {
5084 	bool fault = false;
5085 
5086 	ctxt->ops->get_fpu(ctxt);
5087 	asm volatile("1: fwait \n\t"
5088 		     "2: \n\t"
5089 		     ".pushsection .fixup,\"ax\" \n\t"
5090 		     "3: \n\t"
5091 		     "movb $1, %[fault] \n\t"
5092 		     "jmp 2b \n\t"
5093 		     ".popsection \n\t"
5094 		     _ASM_EXTABLE(1b, 3b)
5095 		     : [fault]"+qm"(fault));
5096 	ctxt->ops->put_fpu(ctxt);
5097 
5098 	if (unlikely(fault))
5099 		return emulate_exception(ctxt, MF_VECTOR, 0, false);
5100 
5101 	return X86EMUL_CONTINUE;
5102 }
5103 
5104 static void fetch_possible_mmx_operand(struct x86_emulate_ctxt *ctxt,
5105 				       struct operand *op)
5106 {
5107 	if (op->type == OP_MM)
5108 		read_mmx_reg(ctxt, &op->mm_val, op->addr.mm);
5109 }
5110 
5111 static int fastop(struct x86_emulate_ctxt *ctxt, void (*fop)(struct fastop *))
5112 {
5113 	register void *__sp asm(_ASM_SP);
5114 	ulong flags = (ctxt->eflags & EFLAGS_MASK) | X86_EFLAGS_IF;
5115 
5116 	if (!(ctxt->d & ByteOp))
5117 		fop += __ffs(ctxt->dst.bytes) * FASTOP_SIZE;
5118 
5119 	asm("push %[flags]; popf; call *%[fastop]; pushf; pop %[flags]\n"
5120 	    : "+a"(ctxt->dst.val), "+d"(ctxt->src.val), [flags]"+D"(flags),
5121 	      [fastop]"+S"(fop), "+r"(__sp)
5122 	    : "c"(ctxt->src2.val));
5123 
5124 	ctxt->eflags = (ctxt->eflags & ~EFLAGS_MASK) | (flags & EFLAGS_MASK);
5125 	if (!fop) /* exception is returned in fop variable */
5126 		return emulate_de(ctxt);
5127 	return X86EMUL_CONTINUE;
5128 }
5129 
5130 void init_decode_cache(struct x86_emulate_ctxt *ctxt)
5131 {
5132 	memset(&ctxt->rip_relative, 0,
5133 	       (void *)&ctxt->modrm - (void *)&ctxt->rip_relative);
5134 
5135 	ctxt->io_read.pos = 0;
5136 	ctxt->io_read.end = 0;
5137 	ctxt->mem_read.end = 0;
5138 }
5139 
5140 int x86_emulate_insn(struct x86_emulate_ctxt *ctxt)
5141 {
5142 	const struct x86_emulate_ops *ops = ctxt->ops;
5143 	int rc = X86EMUL_CONTINUE;
5144 	int saved_dst_type = ctxt->dst.type;
5145 
5146 	ctxt->mem_read.pos = 0;
5147 
5148 	/* LOCK prefix is allowed only with some instructions */
5149 	if (ctxt->lock_prefix && (!(ctxt->d & Lock) || ctxt->dst.type != OP_MEM)) {
5150 		rc = emulate_ud(ctxt);
5151 		goto done;
5152 	}
5153 
5154 	if ((ctxt->d & SrcMask) == SrcMemFAddr && ctxt->src.type != OP_MEM) {
5155 		rc = emulate_ud(ctxt);
5156 		goto done;
5157 	}
5158 
5159 	if (unlikely(ctxt->d &
5160 		     (No64|Undefined|Sse|Mmx|Intercept|CheckPerm|Priv|Prot|String))) {
5161 		if ((ctxt->mode == X86EMUL_MODE_PROT64 && (ctxt->d & No64)) ||
5162 				(ctxt->d & Undefined)) {
5163 			rc = emulate_ud(ctxt);
5164 			goto done;
5165 		}
5166 
5167 		if (((ctxt->d & (Sse|Mmx)) && ((ops->get_cr(ctxt, 0) & X86_CR0_EM)))
5168 		    || ((ctxt->d & Sse) && !(ops->get_cr(ctxt, 4) & X86_CR4_OSFXSR))) {
5169 			rc = emulate_ud(ctxt);
5170 			goto done;
5171 		}
5172 
5173 		if ((ctxt->d & (Sse|Mmx)) && (ops->get_cr(ctxt, 0) & X86_CR0_TS)) {
5174 			rc = emulate_nm(ctxt);
5175 			goto done;
5176 		}
5177 
5178 		if (ctxt->d & Mmx) {
5179 			rc = flush_pending_x87_faults(ctxt);
5180 			if (rc != X86EMUL_CONTINUE)
5181 				goto done;
5182 			/*
5183 			 * Now that we know the fpu is exception safe, we can fetch
5184 			 * operands from it.
5185 			 */
5186 			fetch_possible_mmx_operand(ctxt, &ctxt->src);
5187 			fetch_possible_mmx_operand(ctxt, &ctxt->src2);
5188 			if (!(ctxt->d & Mov))
5189 				fetch_possible_mmx_operand(ctxt, &ctxt->dst);
5190 		}
5191 
5192 		if (unlikely(ctxt->emul_flags & X86EMUL_GUEST_MASK) && ctxt->intercept) {
5193 			rc = emulator_check_intercept(ctxt, ctxt->intercept,
5194 						      X86_ICPT_PRE_EXCEPT);
5195 			if (rc != X86EMUL_CONTINUE)
5196 				goto done;
5197 		}
5198 
5199 		/* Instruction can only be executed in protected mode */
5200 		if ((ctxt->d & Prot) && ctxt->mode < X86EMUL_MODE_PROT16) {
5201 			rc = emulate_ud(ctxt);
5202 			goto done;
5203 		}
5204 
5205 		/* Privileged instruction can be executed only in CPL=0 */
5206 		if ((ctxt->d & Priv) && ops->cpl(ctxt)) {
5207 			if (ctxt->d & PrivUD)
5208 				rc = emulate_ud(ctxt);
5209 			else
5210 				rc = emulate_gp(ctxt, 0);
5211 			goto done;
5212 		}
5213 
5214 		/* Do instruction specific permission checks */
5215 		if (ctxt->d & CheckPerm) {
5216 			rc = ctxt->check_perm(ctxt);
5217 			if (rc != X86EMUL_CONTINUE)
5218 				goto done;
5219 		}
5220 
5221 		if (unlikely(ctxt->emul_flags & X86EMUL_GUEST_MASK) && (ctxt->d & Intercept)) {
5222 			rc = emulator_check_intercept(ctxt, ctxt->intercept,
5223 						      X86_ICPT_POST_EXCEPT);
5224 			if (rc != X86EMUL_CONTINUE)
5225 				goto done;
5226 		}
5227 
5228 		if (ctxt->rep_prefix && (ctxt->d & String)) {
5229 			/* All REP prefixes have the same first termination condition */
5230 			if (address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) == 0) {
5231 				string_registers_quirk(ctxt);
5232 				ctxt->eip = ctxt->_eip;
5233 				ctxt->eflags &= ~X86_EFLAGS_RF;
5234 				goto done;
5235 			}
5236 		}
5237 	}
5238 
5239 	if ((ctxt->src.type == OP_MEM) && !(ctxt->d & NoAccess)) {
5240 		rc = segmented_read(ctxt, ctxt->src.addr.mem,
5241 				    ctxt->src.valptr, ctxt->src.bytes);
5242 		if (rc != X86EMUL_CONTINUE)
5243 			goto done;
5244 		ctxt->src.orig_val64 = ctxt->src.val64;
5245 	}
5246 
5247 	if (ctxt->src2.type == OP_MEM) {
5248 		rc = segmented_read(ctxt, ctxt->src2.addr.mem,
5249 				    &ctxt->src2.val, ctxt->src2.bytes);
5250 		if (rc != X86EMUL_CONTINUE)
5251 			goto done;
5252 	}
5253 
5254 	if ((ctxt->d & DstMask) == ImplicitOps)
5255 		goto special_insn;
5256 
5257 
5258 	if ((ctxt->dst.type == OP_MEM) && !(ctxt->d & Mov)) {
5259 		/* optimisation - avoid slow emulated read if Mov */
5260 		rc = segmented_read(ctxt, ctxt->dst.addr.mem,
5261 				   &ctxt->dst.val, ctxt->dst.bytes);
5262 		if (rc != X86EMUL_CONTINUE) {
5263 			if (!(ctxt->d & NoWrite) &&
5264 			    rc == X86EMUL_PROPAGATE_FAULT &&
5265 			    ctxt->exception.vector == PF_VECTOR)
5266 				ctxt->exception.error_code |= PFERR_WRITE_MASK;
5267 			goto done;
5268 		}
5269 	}
5270 	/* Copy full 64-bit value for CMPXCHG8B.  */
5271 	ctxt->dst.orig_val64 = ctxt->dst.val64;
5272 
5273 special_insn:
5274 
5275 	if (unlikely(ctxt->emul_flags & X86EMUL_GUEST_MASK) && (ctxt->d & Intercept)) {
5276 		rc = emulator_check_intercept(ctxt, ctxt->intercept,
5277 					      X86_ICPT_POST_MEMACCESS);
5278 		if (rc != X86EMUL_CONTINUE)
5279 			goto done;
5280 	}
5281 
5282 	if (ctxt->rep_prefix && (ctxt->d & String))
5283 		ctxt->eflags |= X86_EFLAGS_RF;
5284 	else
5285 		ctxt->eflags &= ~X86_EFLAGS_RF;
5286 
5287 	if (ctxt->execute) {
5288 		if (ctxt->d & Fastop) {
5289 			void (*fop)(struct fastop *) = (void *)ctxt->execute;
5290 			rc = fastop(ctxt, fop);
5291 			if (rc != X86EMUL_CONTINUE)
5292 				goto done;
5293 			goto writeback;
5294 		}
5295 		rc = ctxt->execute(ctxt);
5296 		if (rc != X86EMUL_CONTINUE)
5297 			goto done;
5298 		goto writeback;
5299 	}
5300 
5301 	if (ctxt->opcode_len == 2)
5302 		goto twobyte_insn;
5303 	else if (ctxt->opcode_len == 3)
5304 		goto threebyte_insn;
5305 
5306 	switch (ctxt->b) {
5307 	case 0x70 ... 0x7f: /* jcc (short) */
5308 		if (test_cc(ctxt->b, ctxt->eflags))
5309 			rc = jmp_rel(ctxt, ctxt->src.val);
5310 		break;
5311 	case 0x8d: /* lea r16/r32, m */
5312 		ctxt->dst.val = ctxt->src.addr.mem.ea;
5313 		break;
5314 	case 0x90 ... 0x97: /* nop / xchg reg, rax */
5315 		if (ctxt->dst.addr.reg == reg_rmw(ctxt, VCPU_REGS_RAX))
5316 			ctxt->dst.type = OP_NONE;
5317 		else
5318 			rc = em_xchg(ctxt);
5319 		break;
5320 	case 0x98: /* cbw/cwde/cdqe */
5321 		switch (ctxt->op_bytes) {
5322 		case 2: ctxt->dst.val = (s8)ctxt->dst.val; break;
5323 		case 4: ctxt->dst.val = (s16)ctxt->dst.val; break;
5324 		case 8: ctxt->dst.val = (s32)ctxt->dst.val; break;
5325 		}
5326 		break;
5327 	case 0xcc:		/* int3 */
5328 		rc = emulate_int(ctxt, 3);
5329 		break;
5330 	case 0xcd:		/* int n */
5331 		rc = emulate_int(ctxt, ctxt->src.val);
5332 		break;
5333 	case 0xce:		/* into */
5334 		if (ctxt->eflags & X86_EFLAGS_OF)
5335 			rc = emulate_int(ctxt, 4);
5336 		break;
5337 	case 0xe9: /* jmp rel */
5338 	case 0xeb: /* jmp rel short */
5339 		rc = jmp_rel(ctxt, ctxt->src.val);
5340 		ctxt->dst.type = OP_NONE; /* Disable writeback. */
5341 		break;
5342 	case 0xf4:              /* hlt */
5343 		ctxt->ops->halt(ctxt);
5344 		break;
5345 	case 0xf5:	/* cmc */
5346 		/* complement carry flag from eflags reg */
5347 		ctxt->eflags ^= X86_EFLAGS_CF;
5348 		break;
5349 	case 0xf8: /* clc */
5350 		ctxt->eflags &= ~X86_EFLAGS_CF;
5351 		break;
5352 	case 0xf9: /* stc */
5353 		ctxt->eflags |= X86_EFLAGS_CF;
5354 		break;
5355 	case 0xfc: /* cld */
5356 		ctxt->eflags &= ~X86_EFLAGS_DF;
5357 		break;
5358 	case 0xfd: /* std */
5359 		ctxt->eflags |= X86_EFLAGS_DF;
5360 		break;
5361 	default:
5362 		goto cannot_emulate;
5363 	}
5364 
5365 	if (rc != X86EMUL_CONTINUE)
5366 		goto done;
5367 
5368 writeback:
5369 	if (ctxt->d & SrcWrite) {
5370 		BUG_ON(ctxt->src.type == OP_MEM || ctxt->src.type == OP_MEM_STR);
5371 		rc = writeback(ctxt, &ctxt->src);
5372 		if (rc != X86EMUL_CONTINUE)
5373 			goto done;
5374 	}
5375 	if (!(ctxt->d & NoWrite)) {
5376 		rc = writeback(ctxt, &ctxt->dst);
5377 		if (rc != X86EMUL_CONTINUE)
5378 			goto done;
5379 	}
5380 
5381 	/*
5382 	 * restore dst type in case the decoding will be reused
5383 	 * (happens for string instruction )
5384 	 */
5385 	ctxt->dst.type = saved_dst_type;
5386 
5387 	if ((ctxt->d & SrcMask) == SrcSI)
5388 		string_addr_inc(ctxt, VCPU_REGS_RSI, &ctxt->src);
5389 
5390 	if ((ctxt->d & DstMask) == DstDI)
5391 		string_addr_inc(ctxt, VCPU_REGS_RDI, &ctxt->dst);
5392 
5393 	if (ctxt->rep_prefix && (ctxt->d & String)) {
5394 		unsigned int count;
5395 		struct read_cache *r = &ctxt->io_read;
5396 		if ((ctxt->d & SrcMask) == SrcSI)
5397 			count = ctxt->src.count;
5398 		else
5399 			count = ctxt->dst.count;
5400 		register_address_increment(ctxt, VCPU_REGS_RCX, -count);
5401 
5402 		if (!string_insn_completed(ctxt)) {
5403 			/*
5404 			 * Re-enter guest when pio read ahead buffer is empty
5405 			 * or, if it is not used, after each 1024 iteration.
5406 			 */
5407 			if ((r->end != 0 || reg_read(ctxt, VCPU_REGS_RCX) & 0x3ff) &&
5408 			    (r->end == 0 || r->end != r->pos)) {
5409 				/*
5410 				 * Reset read cache. Usually happens before
5411 				 * decode, but since instruction is restarted
5412 				 * we have to do it here.
5413 				 */
5414 				ctxt->mem_read.end = 0;
5415 				writeback_registers(ctxt);
5416 				return EMULATION_RESTART;
5417 			}
5418 			goto done; /* skip rip writeback */
5419 		}
5420 		ctxt->eflags &= ~X86_EFLAGS_RF;
5421 	}
5422 
5423 	ctxt->eip = ctxt->_eip;
5424 
5425 done:
5426 	if (rc == X86EMUL_PROPAGATE_FAULT) {
5427 		WARN_ON(ctxt->exception.vector > 0x1f);
5428 		ctxt->have_exception = true;
5429 	}
5430 	if (rc == X86EMUL_INTERCEPTED)
5431 		return EMULATION_INTERCEPTED;
5432 
5433 	if (rc == X86EMUL_CONTINUE)
5434 		writeback_registers(ctxt);
5435 
5436 	return (rc == X86EMUL_UNHANDLEABLE) ? EMULATION_FAILED : EMULATION_OK;
5437 
5438 twobyte_insn:
5439 	switch (ctxt->b) {
5440 	case 0x09:		/* wbinvd */
5441 		(ctxt->ops->wbinvd)(ctxt);
5442 		break;
5443 	case 0x08:		/* invd */
5444 	case 0x0d:		/* GrpP (prefetch) */
5445 	case 0x18:		/* Grp16 (prefetch/nop) */
5446 	case 0x1f:		/* nop */
5447 		break;
5448 	case 0x20: /* mov cr, reg */
5449 		ctxt->dst.val = ops->get_cr(ctxt, ctxt->modrm_reg);
5450 		break;
5451 	case 0x21: /* mov from dr to reg */
5452 		ops->get_dr(ctxt, ctxt->modrm_reg, &ctxt->dst.val);
5453 		break;
5454 	case 0x40 ... 0x4f:	/* cmov */
5455 		if (test_cc(ctxt->b, ctxt->eflags))
5456 			ctxt->dst.val = ctxt->src.val;
5457 		else if (ctxt->op_bytes != 4)
5458 			ctxt->dst.type = OP_NONE; /* no writeback */
5459 		break;
5460 	case 0x80 ... 0x8f: /* jnz rel, etc*/
5461 		if (test_cc(ctxt->b, ctxt->eflags))
5462 			rc = jmp_rel(ctxt, ctxt->src.val);
5463 		break;
5464 	case 0x90 ... 0x9f:     /* setcc r/m8 */
5465 		ctxt->dst.val = test_cc(ctxt->b, ctxt->eflags);
5466 		break;
5467 	case 0xb6 ... 0xb7:	/* movzx */
5468 		ctxt->dst.bytes = ctxt->op_bytes;
5469 		ctxt->dst.val = (ctxt->src.bytes == 1) ? (u8) ctxt->src.val
5470 						       : (u16) ctxt->src.val;
5471 		break;
5472 	case 0xbe ... 0xbf:	/* movsx */
5473 		ctxt->dst.bytes = ctxt->op_bytes;
5474 		ctxt->dst.val = (ctxt->src.bytes == 1) ? (s8) ctxt->src.val :
5475 							(s16) ctxt->src.val;
5476 		break;
5477 	default:
5478 		goto cannot_emulate;
5479 	}
5480 
5481 threebyte_insn:
5482 
5483 	if (rc != X86EMUL_CONTINUE)
5484 		goto done;
5485 
5486 	goto writeback;
5487 
5488 cannot_emulate:
5489 	return EMULATION_FAILED;
5490 }
5491 
5492 void emulator_invalidate_register_cache(struct x86_emulate_ctxt *ctxt)
5493 {
5494 	invalidate_registers(ctxt);
5495 }
5496 
5497 void emulator_writeback_register_cache(struct x86_emulate_ctxt *ctxt)
5498 {
5499 	writeback_registers(ctxt);
5500 }
5501