xref: /openbmc/linux/arch/x86/kvm/emulate.c (revision 80483c3a)
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 <asm/kvm_emulate.h>
26 #include <linux/stringify.h>
27 #include <asm/debugreg.h>
28 
29 #include "x86.h"
30 #include "tss.h"
31 
32 /*
33  * Operand types
34  */
35 #define OpNone             0ull
36 #define OpImplicit         1ull  /* No generic decode */
37 #define OpReg              2ull  /* Register */
38 #define OpMem              3ull  /* Memory */
39 #define OpAcc              4ull  /* Accumulator: AL/AX/EAX/RAX */
40 #define OpDI               5ull  /* ES:DI/EDI/RDI */
41 #define OpMem64            6ull  /* Memory, 64-bit */
42 #define OpImmUByte         7ull  /* Zero-extended 8-bit immediate */
43 #define OpDX               8ull  /* DX register */
44 #define OpCL               9ull  /* CL register (for shifts) */
45 #define OpImmByte         10ull  /* 8-bit sign extended immediate */
46 #define OpOne             11ull  /* Implied 1 */
47 #define OpImm             12ull  /* Sign extended up to 32-bit immediate */
48 #define OpMem16           13ull  /* Memory operand (16-bit). */
49 #define OpMem32           14ull  /* Memory operand (32-bit). */
50 #define OpImmU            15ull  /* Immediate operand, zero extended */
51 #define OpSI              16ull  /* SI/ESI/RSI */
52 #define OpImmFAddr        17ull  /* Immediate far address */
53 #define OpMemFAddr        18ull  /* Far address in memory */
54 #define OpImmU16          19ull  /* Immediate operand, 16 bits, zero extended */
55 #define OpES              20ull  /* ES */
56 #define OpCS              21ull  /* CS */
57 #define OpSS              22ull  /* SS */
58 #define OpDS              23ull  /* DS */
59 #define OpFS              24ull  /* FS */
60 #define OpGS              25ull  /* GS */
61 #define OpMem8            26ull  /* 8-bit zero extended memory operand */
62 #define OpImm64           27ull  /* Sign extended 16/32/64-bit immediate */
63 #define OpXLat            28ull  /* memory at BX/EBX/RBX + zero-extended AL */
64 #define OpAccLo           29ull  /* Low part of extended acc (AX/AX/EAX/RAX) */
65 #define OpAccHi           30ull  /* High part of extended acc (-/DX/EDX/RDX) */
66 
67 #define OpBits             5  /* Width of operand field */
68 #define OpMask             ((1ull << OpBits) - 1)
69 
70 /*
71  * Opcode effective-address decode tables.
72  * Note that we only emulate instructions that have at least one memory
73  * operand (excluding implicit stack references). We assume that stack
74  * references and instruction fetches will never occur in special memory
75  * areas that require emulation. So, for example, 'mov <imm>,<reg>' need
76  * not be handled.
77  */
78 
79 /* Operand sizes: 8-bit operands or specified/overridden size. */
80 #define ByteOp      (1<<0)	/* 8-bit operands. */
81 /* Destination operand type. */
82 #define DstShift    1
83 #define ImplicitOps (OpImplicit << DstShift)
84 #define DstReg      (OpReg << DstShift)
85 #define DstMem      (OpMem << DstShift)
86 #define DstAcc      (OpAcc << DstShift)
87 #define DstDI       (OpDI << DstShift)
88 #define DstMem64    (OpMem64 << DstShift)
89 #define DstMem16    (OpMem16 << DstShift)
90 #define DstImmUByte (OpImmUByte << DstShift)
91 #define DstDX       (OpDX << DstShift)
92 #define DstAccLo    (OpAccLo << DstShift)
93 #define DstMask     (OpMask << DstShift)
94 /* Source operand type. */
95 #define SrcShift    6
96 #define SrcNone     (OpNone << SrcShift)
97 #define SrcReg      (OpReg << SrcShift)
98 #define SrcMem      (OpMem << SrcShift)
99 #define SrcMem16    (OpMem16 << SrcShift)
100 #define SrcMem32    (OpMem32 << SrcShift)
101 #define SrcImm      (OpImm << SrcShift)
102 #define SrcImmByte  (OpImmByte << SrcShift)
103 #define SrcOne      (OpOne << SrcShift)
104 #define SrcImmUByte (OpImmUByte << SrcShift)
105 #define SrcImmU     (OpImmU << SrcShift)
106 #define SrcSI       (OpSI << SrcShift)
107 #define SrcXLat     (OpXLat << SrcShift)
108 #define SrcImmFAddr (OpImmFAddr << SrcShift)
109 #define SrcMemFAddr (OpMemFAddr << SrcShift)
110 #define SrcAcc      (OpAcc << SrcShift)
111 #define SrcImmU16   (OpImmU16 << SrcShift)
112 #define SrcImm64    (OpImm64 << SrcShift)
113 #define SrcDX       (OpDX << SrcShift)
114 #define SrcMem8     (OpMem8 << SrcShift)
115 #define SrcAccHi    (OpAccHi << SrcShift)
116 #define SrcMask     (OpMask << SrcShift)
117 #define BitOp       (1<<11)
118 #define MemAbs      (1<<12)      /* Memory operand is absolute displacement */
119 #define String      (1<<13)     /* String instruction (rep capable) */
120 #define Stack       (1<<14)     /* Stack instruction (push/pop) */
121 #define GroupMask   (7<<15)     /* Opcode uses one of the group mechanisms */
122 #define Group       (1<<15)     /* Bits 3:5 of modrm byte extend opcode */
123 #define GroupDual   (2<<15)     /* Alternate decoding of mod == 3 */
124 #define Prefix      (3<<15)     /* Instruction varies with 66/f2/f3 prefix */
125 #define RMExt       (4<<15)     /* Opcode extension in ModRM r/m if mod == 3 */
126 #define Escape      (5<<15)     /* Escape to coprocessor instruction */
127 #define InstrDual   (6<<15)     /* Alternate instruction decoding of mod == 3 */
128 #define ModeDual    (7<<15)     /* Different instruction for 32/64 bit */
129 #define Sse         (1<<18)     /* SSE Vector instruction */
130 /* Generic ModRM decode. */
131 #define ModRM       (1<<19)
132 /* Destination is only written; never read. */
133 #define Mov         (1<<20)
134 /* Misc flags */
135 #define Prot        (1<<21) /* instruction generates #UD if not in prot-mode */
136 #define EmulateOnUD (1<<22) /* Emulate if unsupported by the host */
137 #define NoAccess    (1<<23) /* Don't access memory (lea/invlpg/verr etc) */
138 #define Op3264      (1<<24) /* Operand is 64b in long mode, 32b otherwise */
139 #define Undefined   (1<<25) /* No Such Instruction */
140 #define Lock        (1<<26) /* lock prefix is allowed for the instruction */
141 #define Priv        (1<<27) /* instruction generates #GP if current CPL != 0 */
142 #define No64	    (1<<28)
143 #define PageTable   (1 << 29)   /* instruction used to write page table */
144 #define NotImpl     (1 << 30)   /* instruction is not implemented */
145 /* Source 2 operand type */
146 #define Src2Shift   (31)
147 #define Src2None    (OpNone << Src2Shift)
148 #define Src2Mem     (OpMem << Src2Shift)
149 #define Src2CL      (OpCL << Src2Shift)
150 #define Src2ImmByte (OpImmByte << Src2Shift)
151 #define Src2One     (OpOne << Src2Shift)
152 #define Src2Imm     (OpImm << Src2Shift)
153 #define Src2ES      (OpES << Src2Shift)
154 #define Src2CS      (OpCS << Src2Shift)
155 #define Src2SS      (OpSS << Src2Shift)
156 #define Src2DS      (OpDS << Src2Shift)
157 #define Src2FS      (OpFS << Src2Shift)
158 #define Src2GS      (OpGS << Src2Shift)
159 #define Src2Mask    (OpMask << Src2Shift)
160 #define Mmx         ((u64)1 << 40)  /* MMX Vector instruction */
161 #define Aligned     ((u64)1 << 41)  /* Explicitly aligned (e.g. MOVDQA) */
162 #define Unaligned   ((u64)1 << 42)  /* Explicitly unaligned (e.g. MOVDQU) */
163 #define Avx         ((u64)1 << 43)  /* Advanced Vector Extensions */
164 #define Fastop      ((u64)1 << 44)  /* Use opcode::u.fastop */
165 #define NoWrite     ((u64)1 << 45)  /* No writeback */
166 #define SrcWrite    ((u64)1 << 46)  /* Write back src operand */
167 #define NoMod	    ((u64)1 << 47)  /* Mod field is ignored */
168 #define Intercept   ((u64)1 << 48)  /* Has valid intercept field */
169 #define CheckPerm   ((u64)1 << 49)  /* Has valid check_perm field */
170 #define PrivUD      ((u64)1 << 51)  /* #UD instead of #GP on CPL > 0 */
171 #define NearBranch  ((u64)1 << 52)  /* Near branches */
172 #define No16	    ((u64)1 << 53)  /* No 16 bit operand */
173 #define IncSP       ((u64)1 << 54)  /* SP is incremented before ModRM calc */
174 
175 #define DstXacc     (DstAccLo | SrcAccHi | SrcWrite)
176 
177 #define X2(x...) x, x
178 #define X3(x...) X2(x), x
179 #define X4(x...) X2(x), X2(x)
180 #define X5(x...) X4(x), x
181 #define X6(x...) X4(x), X2(x)
182 #define X7(x...) X4(x), X3(x)
183 #define X8(x...) X4(x), X4(x)
184 #define X16(x...) X8(x), X8(x)
185 
186 #define NR_FASTOP (ilog2(sizeof(ulong)) + 1)
187 #define FASTOP_SIZE 8
188 
189 /*
190  * fastop functions have a special calling convention:
191  *
192  * dst:    rax        (in/out)
193  * src:    rdx        (in/out)
194  * src2:   rcx        (in)
195  * flags:  rflags     (in/out)
196  * ex:     rsi        (in:fastop pointer, out:zero if exception)
197  *
198  * Moreover, they are all exactly FASTOP_SIZE bytes long, so functions for
199  * different operand sizes can be reached by calculation, rather than a jump
200  * table (which would be bigger than the code).
201  *
202  * fastop functions are declared as taking a never-defined fastop parameter,
203  * so they can't be called from C directly.
204  */
205 
206 struct fastop;
207 
208 struct opcode {
209 	u64 flags : 56;
210 	u64 intercept : 8;
211 	union {
212 		int (*execute)(struct x86_emulate_ctxt *ctxt);
213 		const struct opcode *group;
214 		const struct group_dual *gdual;
215 		const struct gprefix *gprefix;
216 		const struct escape *esc;
217 		const struct instr_dual *idual;
218 		const struct mode_dual *mdual;
219 		void (*fastop)(struct fastop *fake);
220 	} u;
221 	int (*check_perm)(struct x86_emulate_ctxt *ctxt);
222 };
223 
224 struct group_dual {
225 	struct opcode mod012[8];
226 	struct opcode mod3[8];
227 };
228 
229 struct gprefix {
230 	struct opcode pfx_no;
231 	struct opcode pfx_66;
232 	struct opcode pfx_f2;
233 	struct opcode pfx_f3;
234 };
235 
236 struct escape {
237 	struct opcode op[8];
238 	struct opcode high[64];
239 };
240 
241 struct instr_dual {
242 	struct opcode mod012;
243 	struct opcode mod3;
244 };
245 
246 struct mode_dual {
247 	struct opcode mode32;
248 	struct opcode mode64;
249 };
250 
251 #define EFLG_RESERVED_ZEROS_MASK 0xffc0802a
252 
253 enum x86_transfer_type {
254 	X86_TRANSFER_NONE,
255 	X86_TRANSFER_CALL_JMP,
256 	X86_TRANSFER_RET,
257 	X86_TRANSFER_TASK_SWITCH,
258 };
259 
260 static ulong reg_read(struct x86_emulate_ctxt *ctxt, unsigned nr)
261 {
262 	if (!(ctxt->regs_valid & (1 << nr))) {
263 		ctxt->regs_valid |= 1 << nr;
264 		ctxt->_regs[nr] = ctxt->ops->read_gpr(ctxt, nr);
265 	}
266 	return ctxt->_regs[nr];
267 }
268 
269 static ulong *reg_write(struct x86_emulate_ctxt *ctxt, unsigned nr)
270 {
271 	ctxt->regs_valid |= 1 << nr;
272 	ctxt->regs_dirty |= 1 << nr;
273 	return &ctxt->_regs[nr];
274 }
275 
276 static ulong *reg_rmw(struct x86_emulate_ctxt *ctxt, unsigned nr)
277 {
278 	reg_read(ctxt, nr);
279 	return reg_write(ctxt, nr);
280 }
281 
282 static void writeback_registers(struct x86_emulate_ctxt *ctxt)
283 {
284 	unsigned reg;
285 
286 	for_each_set_bit(reg, (ulong *)&ctxt->regs_dirty, 16)
287 		ctxt->ops->write_gpr(ctxt, reg, ctxt->_regs[reg]);
288 }
289 
290 static void invalidate_registers(struct x86_emulate_ctxt *ctxt)
291 {
292 	ctxt->regs_dirty = 0;
293 	ctxt->regs_valid = 0;
294 }
295 
296 /*
297  * These EFLAGS bits are restored from saved value during emulation, and
298  * any changes are written back to the saved value after emulation.
299  */
300 #define EFLAGS_MASK (X86_EFLAGS_OF|X86_EFLAGS_SF|X86_EFLAGS_ZF|X86_EFLAGS_AF|\
301 		     X86_EFLAGS_PF|X86_EFLAGS_CF)
302 
303 #ifdef CONFIG_X86_64
304 #define ON64(x) x
305 #else
306 #define ON64(x)
307 #endif
308 
309 static int fastop(struct x86_emulate_ctxt *ctxt, void (*fop)(struct fastop *));
310 
311 #define FOP_FUNC(name) \
312 	".align " __stringify(FASTOP_SIZE) " \n\t" \
313 	".type " name ", @function \n\t" \
314 	name ":\n\t"
315 
316 #define FOP_RET   "ret \n\t"
317 
318 #define FOP_START(op) \
319 	extern void em_##op(struct fastop *fake); \
320 	asm(".pushsection .text, \"ax\" \n\t" \
321 	    ".global em_" #op " \n\t" \
322 	    FOP_FUNC("em_" #op)
323 
324 #define FOP_END \
325 	    ".popsection")
326 
327 #define FOPNOP() \
328 	FOP_FUNC(__stringify(__UNIQUE_ID(nop))) \
329 	FOP_RET
330 
331 #define FOP1E(op,  dst) \
332 	FOP_FUNC(#op "_" #dst) \
333 	"10: " #op " %" #dst " \n\t" FOP_RET
334 
335 #define FOP1EEX(op,  dst) \
336 	FOP1E(op, dst) _ASM_EXTABLE(10b, kvm_fastop_exception)
337 
338 #define FASTOP1(op) \
339 	FOP_START(op) \
340 	FOP1E(op##b, al) \
341 	FOP1E(op##w, ax) \
342 	FOP1E(op##l, eax) \
343 	ON64(FOP1E(op##q, rax))	\
344 	FOP_END
345 
346 /* 1-operand, using src2 (for MUL/DIV r/m) */
347 #define FASTOP1SRC2(op, name) \
348 	FOP_START(name) \
349 	FOP1E(op, cl) \
350 	FOP1E(op, cx) \
351 	FOP1E(op, ecx) \
352 	ON64(FOP1E(op, rcx)) \
353 	FOP_END
354 
355 /* 1-operand, using src2 (for MUL/DIV r/m), with exceptions */
356 #define FASTOP1SRC2EX(op, name) \
357 	FOP_START(name) \
358 	FOP1EEX(op, cl) \
359 	FOP1EEX(op, cx) \
360 	FOP1EEX(op, ecx) \
361 	ON64(FOP1EEX(op, rcx)) \
362 	FOP_END
363 
364 #define FOP2E(op,  dst, src)	   \
365 	FOP_FUNC(#op "_" #dst "_" #src) \
366 	#op " %" #src ", %" #dst " \n\t" FOP_RET
367 
368 #define FASTOP2(op) \
369 	FOP_START(op) \
370 	FOP2E(op##b, al, dl) \
371 	FOP2E(op##w, ax, dx) \
372 	FOP2E(op##l, eax, edx) \
373 	ON64(FOP2E(op##q, rax, rdx)) \
374 	FOP_END
375 
376 /* 2 operand, word only */
377 #define FASTOP2W(op) \
378 	FOP_START(op) \
379 	FOPNOP() \
380 	FOP2E(op##w, ax, dx) \
381 	FOP2E(op##l, eax, edx) \
382 	ON64(FOP2E(op##q, rax, rdx)) \
383 	FOP_END
384 
385 /* 2 operand, src is CL */
386 #define FASTOP2CL(op) \
387 	FOP_START(op) \
388 	FOP2E(op##b, al, cl) \
389 	FOP2E(op##w, ax, cl) \
390 	FOP2E(op##l, eax, cl) \
391 	ON64(FOP2E(op##q, rax, cl)) \
392 	FOP_END
393 
394 /* 2 operand, src and dest are reversed */
395 #define FASTOP2R(op, name) \
396 	FOP_START(name) \
397 	FOP2E(op##b, dl, al) \
398 	FOP2E(op##w, dx, ax) \
399 	FOP2E(op##l, edx, eax) \
400 	ON64(FOP2E(op##q, rdx, rax)) \
401 	FOP_END
402 
403 #define FOP3E(op,  dst, src, src2) \
404 	FOP_FUNC(#op "_" #dst "_" #src "_" #src2) \
405 	#op " %" #src2 ", %" #src ", %" #dst " \n\t" FOP_RET
406 
407 /* 3-operand, word-only, src2=cl */
408 #define FASTOP3WCL(op) \
409 	FOP_START(op) \
410 	FOPNOP() \
411 	FOP3E(op##w, ax, dx, cl) \
412 	FOP3E(op##l, eax, edx, cl) \
413 	ON64(FOP3E(op##q, rax, rdx, cl)) \
414 	FOP_END
415 
416 /* Special case for SETcc - 1 instruction per cc */
417 #define FOP_SETCC(op) \
418 	".align 4 \n\t" \
419 	".type " #op ", @function \n\t" \
420 	#op ": \n\t" \
421 	#op " %al \n\t" \
422 	FOP_RET
423 
424 asm(".global kvm_fastop_exception \n"
425     "kvm_fastop_exception: xor %esi, %esi; ret");
426 
427 FOP_START(setcc)
428 FOP_SETCC(seto)
429 FOP_SETCC(setno)
430 FOP_SETCC(setc)
431 FOP_SETCC(setnc)
432 FOP_SETCC(setz)
433 FOP_SETCC(setnz)
434 FOP_SETCC(setbe)
435 FOP_SETCC(setnbe)
436 FOP_SETCC(sets)
437 FOP_SETCC(setns)
438 FOP_SETCC(setp)
439 FOP_SETCC(setnp)
440 FOP_SETCC(setl)
441 FOP_SETCC(setnl)
442 FOP_SETCC(setle)
443 FOP_SETCC(setnle)
444 FOP_END;
445 
446 FOP_START(salc) "pushf; sbb %al, %al; popf \n\t" FOP_RET
447 FOP_END;
448 
449 static int emulator_check_intercept(struct x86_emulate_ctxt *ctxt,
450 				    enum x86_intercept intercept,
451 				    enum x86_intercept_stage stage)
452 {
453 	struct x86_instruction_info info = {
454 		.intercept  = intercept,
455 		.rep_prefix = ctxt->rep_prefix,
456 		.modrm_mod  = ctxt->modrm_mod,
457 		.modrm_reg  = ctxt->modrm_reg,
458 		.modrm_rm   = ctxt->modrm_rm,
459 		.src_val    = ctxt->src.val64,
460 		.dst_val    = ctxt->dst.val64,
461 		.src_bytes  = ctxt->src.bytes,
462 		.dst_bytes  = ctxt->dst.bytes,
463 		.ad_bytes   = ctxt->ad_bytes,
464 		.next_rip   = ctxt->eip,
465 	};
466 
467 	return ctxt->ops->intercept(ctxt, &info, stage);
468 }
469 
470 static void assign_masked(ulong *dest, ulong src, ulong mask)
471 {
472 	*dest = (*dest & ~mask) | (src & mask);
473 }
474 
475 static void assign_register(unsigned long *reg, u64 val, int bytes)
476 {
477 	/* The 4-byte case *is* correct: in 64-bit mode we zero-extend. */
478 	switch (bytes) {
479 	case 1:
480 		*(u8 *)reg = (u8)val;
481 		break;
482 	case 2:
483 		*(u16 *)reg = (u16)val;
484 		break;
485 	case 4:
486 		*reg = (u32)val;
487 		break;	/* 64b: zero-extend */
488 	case 8:
489 		*reg = val;
490 		break;
491 	}
492 }
493 
494 static inline unsigned long ad_mask(struct x86_emulate_ctxt *ctxt)
495 {
496 	return (1UL << (ctxt->ad_bytes << 3)) - 1;
497 }
498 
499 static ulong stack_mask(struct x86_emulate_ctxt *ctxt)
500 {
501 	u16 sel;
502 	struct desc_struct ss;
503 
504 	if (ctxt->mode == X86EMUL_MODE_PROT64)
505 		return ~0UL;
506 	ctxt->ops->get_segment(ctxt, &sel, &ss, NULL, VCPU_SREG_SS);
507 	return ~0U >> ((ss.d ^ 1) * 16);  /* d=0: 0xffff; d=1: 0xffffffff */
508 }
509 
510 static int stack_size(struct x86_emulate_ctxt *ctxt)
511 {
512 	return (__fls(stack_mask(ctxt)) + 1) >> 3;
513 }
514 
515 /* Access/update address held in a register, based on addressing mode. */
516 static inline unsigned long
517 address_mask(struct x86_emulate_ctxt *ctxt, unsigned long reg)
518 {
519 	if (ctxt->ad_bytes == sizeof(unsigned long))
520 		return reg;
521 	else
522 		return reg & ad_mask(ctxt);
523 }
524 
525 static inline unsigned long
526 register_address(struct x86_emulate_ctxt *ctxt, int reg)
527 {
528 	return address_mask(ctxt, reg_read(ctxt, reg));
529 }
530 
531 static void masked_increment(ulong *reg, ulong mask, int inc)
532 {
533 	assign_masked(reg, *reg + inc, mask);
534 }
535 
536 static inline void
537 register_address_increment(struct x86_emulate_ctxt *ctxt, int reg, int inc)
538 {
539 	ulong *preg = reg_rmw(ctxt, reg);
540 
541 	assign_register(preg, *preg + inc, ctxt->ad_bytes);
542 }
543 
544 static void rsp_increment(struct x86_emulate_ctxt *ctxt, int inc)
545 {
546 	masked_increment(reg_rmw(ctxt, VCPU_REGS_RSP), stack_mask(ctxt), inc);
547 }
548 
549 static u32 desc_limit_scaled(struct desc_struct *desc)
550 {
551 	u32 limit = get_desc_limit(desc);
552 
553 	return desc->g ? (limit << 12) | 0xfff : limit;
554 }
555 
556 static unsigned long seg_base(struct x86_emulate_ctxt *ctxt, int seg)
557 {
558 	if (ctxt->mode == X86EMUL_MODE_PROT64 && seg < VCPU_SREG_FS)
559 		return 0;
560 
561 	return ctxt->ops->get_cached_segment_base(ctxt, seg);
562 }
563 
564 static int emulate_exception(struct x86_emulate_ctxt *ctxt, int vec,
565 			     u32 error, bool valid)
566 {
567 	WARN_ON(vec > 0x1f);
568 	ctxt->exception.vector = vec;
569 	ctxt->exception.error_code = error;
570 	ctxt->exception.error_code_valid = valid;
571 	return X86EMUL_PROPAGATE_FAULT;
572 }
573 
574 static int emulate_db(struct x86_emulate_ctxt *ctxt)
575 {
576 	return emulate_exception(ctxt, DB_VECTOR, 0, false);
577 }
578 
579 static int emulate_gp(struct x86_emulate_ctxt *ctxt, int err)
580 {
581 	return emulate_exception(ctxt, GP_VECTOR, err, true);
582 }
583 
584 static int emulate_ss(struct x86_emulate_ctxt *ctxt, int err)
585 {
586 	return emulate_exception(ctxt, SS_VECTOR, err, true);
587 }
588 
589 static int emulate_ud(struct x86_emulate_ctxt *ctxt)
590 {
591 	return emulate_exception(ctxt, UD_VECTOR, 0, false);
592 }
593 
594 static int emulate_ts(struct x86_emulate_ctxt *ctxt, int err)
595 {
596 	return emulate_exception(ctxt, TS_VECTOR, err, true);
597 }
598 
599 static int emulate_de(struct x86_emulate_ctxt *ctxt)
600 {
601 	return emulate_exception(ctxt, DE_VECTOR, 0, false);
602 }
603 
604 static int emulate_nm(struct x86_emulate_ctxt *ctxt)
605 {
606 	return emulate_exception(ctxt, NM_VECTOR, 0, false);
607 }
608 
609 static u16 get_segment_selector(struct x86_emulate_ctxt *ctxt, unsigned seg)
610 {
611 	u16 selector;
612 	struct desc_struct desc;
613 
614 	ctxt->ops->get_segment(ctxt, &selector, &desc, NULL, seg);
615 	return selector;
616 }
617 
618 static void set_segment_selector(struct x86_emulate_ctxt *ctxt, u16 selector,
619 				 unsigned seg)
620 {
621 	u16 dummy;
622 	u32 base3;
623 	struct desc_struct desc;
624 
625 	ctxt->ops->get_segment(ctxt, &dummy, &desc, &base3, seg);
626 	ctxt->ops->set_segment(ctxt, selector, &desc, base3, seg);
627 }
628 
629 /*
630  * x86 defines three classes of vector instructions: explicitly
631  * aligned, explicitly unaligned, and the rest, which change behaviour
632  * depending on whether they're AVX encoded or not.
633  *
634  * Also included is CMPXCHG16B which is not a vector instruction, yet it is
635  * subject to the same check.
636  */
637 static bool insn_aligned(struct x86_emulate_ctxt *ctxt, unsigned size)
638 {
639 	if (likely(size < 16))
640 		return false;
641 
642 	if (ctxt->d & Aligned)
643 		return true;
644 	else if (ctxt->d & Unaligned)
645 		return false;
646 	else if (ctxt->d & Avx)
647 		return false;
648 	else
649 		return true;
650 }
651 
652 static __always_inline int __linearize(struct x86_emulate_ctxt *ctxt,
653 				       struct segmented_address addr,
654 				       unsigned *max_size, unsigned size,
655 				       bool write, bool fetch,
656 				       enum x86emul_mode mode, ulong *linear)
657 {
658 	struct desc_struct desc;
659 	bool usable;
660 	ulong la;
661 	u32 lim;
662 	u16 sel;
663 
664 	la = seg_base(ctxt, addr.seg) + addr.ea;
665 	*max_size = 0;
666 	switch (mode) {
667 	case X86EMUL_MODE_PROT64:
668 		*linear = la;
669 		if (is_noncanonical_address(la))
670 			goto bad;
671 
672 		*max_size = min_t(u64, ~0u, (1ull << 48) - la);
673 		if (size > *max_size)
674 			goto bad;
675 		break;
676 	default:
677 		*linear = la = (u32)la;
678 		usable = ctxt->ops->get_segment(ctxt, &sel, &desc, NULL,
679 						addr.seg);
680 		if (!usable)
681 			goto bad;
682 		/* code segment in protected mode or read-only data segment */
683 		if ((((ctxt->mode != X86EMUL_MODE_REAL) && (desc.type & 8))
684 					|| !(desc.type & 2)) && write)
685 			goto bad;
686 		/* unreadable code segment */
687 		if (!fetch && (desc.type & 8) && !(desc.type & 2))
688 			goto bad;
689 		lim = desc_limit_scaled(&desc);
690 		if (!(desc.type & 8) && (desc.type & 4)) {
691 			/* expand-down segment */
692 			if (addr.ea <= lim)
693 				goto bad;
694 			lim = desc.d ? 0xffffffff : 0xffff;
695 		}
696 		if (addr.ea > lim)
697 			goto bad;
698 		if (lim == 0xffffffff)
699 			*max_size = ~0u;
700 		else {
701 			*max_size = (u64)lim + 1 - addr.ea;
702 			if (size > *max_size)
703 				goto bad;
704 		}
705 		break;
706 	}
707 	if (insn_aligned(ctxt, size) && ((la & (size - 1)) != 0))
708 		return emulate_gp(ctxt, 0);
709 	return X86EMUL_CONTINUE;
710 bad:
711 	if (addr.seg == VCPU_SREG_SS)
712 		return emulate_ss(ctxt, 0);
713 	else
714 		return emulate_gp(ctxt, 0);
715 }
716 
717 static int linearize(struct x86_emulate_ctxt *ctxt,
718 		     struct segmented_address addr,
719 		     unsigned size, bool write,
720 		     ulong *linear)
721 {
722 	unsigned max_size;
723 	return __linearize(ctxt, addr, &max_size, size, write, false,
724 			   ctxt->mode, linear);
725 }
726 
727 static inline int assign_eip(struct x86_emulate_ctxt *ctxt, ulong dst,
728 			     enum x86emul_mode mode)
729 {
730 	ulong linear;
731 	int rc;
732 	unsigned max_size;
733 	struct segmented_address addr = { .seg = VCPU_SREG_CS,
734 					   .ea = dst };
735 
736 	if (ctxt->op_bytes != sizeof(unsigned long))
737 		addr.ea = dst & ((1UL << (ctxt->op_bytes << 3)) - 1);
738 	rc = __linearize(ctxt, addr, &max_size, 1, false, true, mode, &linear);
739 	if (rc == X86EMUL_CONTINUE)
740 		ctxt->_eip = addr.ea;
741 	return rc;
742 }
743 
744 static inline int assign_eip_near(struct x86_emulate_ctxt *ctxt, ulong dst)
745 {
746 	return assign_eip(ctxt, dst, ctxt->mode);
747 }
748 
749 static int assign_eip_far(struct x86_emulate_ctxt *ctxt, ulong dst,
750 			  const struct desc_struct *cs_desc)
751 {
752 	enum x86emul_mode mode = ctxt->mode;
753 	int rc;
754 
755 #ifdef CONFIG_X86_64
756 	if (ctxt->mode >= X86EMUL_MODE_PROT16) {
757 		if (cs_desc->l) {
758 			u64 efer = 0;
759 
760 			ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
761 			if (efer & EFER_LMA)
762 				mode = X86EMUL_MODE_PROT64;
763 		} else
764 			mode = X86EMUL_MODE_PROT32; /* temporary value */
765 	}
766 #endif
767 	if (mode == X86EMUL_MODE_PROT16 || mode == X86EMUL_MODE_PROT32)
768 		mode = cs_desc->d ? X86EMUL_MODE_PROT32 : X86EMUL_MODE_PROT16;
769 	rc = assign_eip(ctxt, dst, mode);
770 	if (rc == X86EMUL_CONTINUE)
771 		ctxt->mode = mode;
772 	return rc;
773 }
774 
775 static inline int jmp_rel(struct x86_emulate_ctxt *ctxt, int rel)
776 {
777 	return assign_eip_near(ctxt, ctxt->_eip + rel);
778 }
779 
780 static int segmented_read_std(struct x86_emulate_ctxt *ctxt,
781 			      struct segmented_address addr,
782 			      void *data,
783 			      unsigned size)
784 {
785 	int rc;
786 	ulong linear;
787 
788 	rc = linearize(ctxt, addr, size, false, &linear);
789 	if (rc != X86EMUL_CONTINUE)
790 		return rc;
791 	return ctxt->ops->read_std(ctxt, linear, data, size, &ctxt->exception);
792 }
793 
794 /*
795  * Prefetch the remaining bytes of the instruction without crossing page
796  * boundary if they are not in fetch_cache yet.
797  */
798 static int __do_insn_fetch_bytes(struct x86_emulate_ctxt *ctxt, int op_size)
799 {
800 	int rc;
801 	unsigned size, max_size;
802 	unsigned long linear;
803 	int cur_size = ctxt->fetch.end - ctxt->fetch.data;
804 	struct segmented_address addr = { .seg = VCPU_SREG_CS,
805 					   .ea = ctxt->eip + cur_size };
806 
807 	/*
808 	 * We do not know exactly how many bytes will be needed, and
809 	 * __linearize is expensive, so fetch as much as possible.  We
810 	 * just have to avoid going beyond the 15 byte limit, the end
811 	 * of the segment, or the end of the page.
812 	 *
813 	 * __linearize is called with size 0 so that it does not do any
814 	 * boundary check itself.  Instead, we use max_size to check
815 	 * against op_size.
816 	 */
817 	rc = __linearize(ctxt, addr, &max_size, 0, false, true, ctxt->mode,
818 			 &linear);
819 	if (unlikely(rc != X86EMUL_CONTINUE))
820 		return rc;
821 
822 	size = min_t(unsigned, 15UL ^ cur_size, max_size);
823 	size = min_t(unsigned, size, PAGE_SIZE - offset_in_page(linear));
824 
825 	/*
826 	 * One instruction can only straddle two pages,
827 	 * and one has been loaded at the beginning of
828 	 * x86_decode_insn.  So, if not enough bytes
829 	 * still, we must have hit the 15-byte boundary.
830 	 */
831 	if (unlikely(size < op_size))
832 		return emulate_gp(ctxt, 0);
833 
834 	rc = ctxt->ops->fetch(ctxt, linear, ctxt->fetch.end,
835 			      size, &ctxt->exception);
836 	if (unlikely(rc != X86EMUL_CONTINUE))
837 		return rc;
838 	ctxt->fetch.end += size;
839 	return X86EMUL_CONTINUE;
840 }
841 
842 static __always_inline int do_insn_fetch_bytes(struct x86_emulate_ctxt *ctxt,
843 					       unsigned size)
844 {
845 	unsigned done_size = ctxt->fetch.end - ctxt->fetch.ptr;
846 
847 	if (unlikely(done_size < size))
848 		return __do_insn_fetch_bytes(ctxt, size - done_size);
849 	else
850 		return X86EMUL_CONTINUE;
851 }
852 
853 /* Fetch next part of the instruction being emulated. */
854 #define insn_fetch(_type, _ctxt)					\
855 ({	_type _x;							\
856 									\
857 	rc = do_insn_fetch_bytes(_ctxt, sizeof(_type));			\
858 	if (rc != X86EMUL_CONTINUE)					\
859 		goto done;						\
860 	ctxt->_eip += sizeof(_type);					\
861 	_x = *(_type __aligned(1) *) ctxt->fetch.ptr;			\
862 	ctxt->fetch.ptr += sizeof(_type);				\
863 	_x;								\
864 })
865 
866 #define insn_fetch_arr(_arr, _size, _ctxt)				\
867 ({									\
868 	rc = do_insn_fetch_bytes(_ctxt, _size);				\
869 	if (rc != X86EMUL_CONTINUE)					\
870 		goto done;						\
871 	ctxt->_eip += (_size);						\
872 	memcpy(_arr, ctxt->fetch.ptr, _size);				\
873 	ctxt->fetch.ptr += (_size);					\
874 })
875 
876 /*
877  * Given the 'reg' portion of a ModRM byte, and a register block, return a
878  * pointer into the block that addresses the relevant register.
879  * @highbyte_regs specifies whether to decode AH,CH,DH,BH.
880  */
881 static void *decode_register(struct x86_emulate_ctxt *ctxt, u8 modrm_reg,
882 			     int byteop)
883 {
884 	void *p;
885 	int highbyte_regs = (ctxt->rex_prefix == 0) && byteop;
886 
887 	if (highbyte_regs && modrm_reg >= 4 && modrm_reg < 8)
888 		p = (unsigned char *)reg_rmw(ctxt, modrm_reg & 3) + 1;
889 	else
890 		p = reg_rmw(ctxt, modrm_reg);
891 	return p;
892 }
893 
894 static int read_descriptor(struct x86_emulate_ctxt *ctxt,
895 			   struct segmented_address addr,
896 			   u16 *size, unsigned long *address, int op_bytes)
897 {
898 	int rc;
899 
900 	if (op_bytes == 2)
901 		op_bytes = 3;
902 	*address = 0;
903 	rc = segmented_read_std(ctxt, addr, size, 2);
904 	if (rc != X86EMUL_CONTINUE)
905 		return rc;
906 	addr.ea += 2;
907 	rc = segmented_read_std(ctxt, addr, address, op_bytes);
908 	return rc;
909 }
910 
911 FASTOP2(add);
912 FASTOP2(or);
913 FASTOP2(adc);
914 FASTOP2(sbb);
915 FASTOP2(and);
916 FASTOP2(sub);
917 FASTOP2(xor);
918 FASTOP2(cmp);
919 FASTOP2(test);
920 
921 FASTOP1SRC2(mul, mul_ex);
922 FASTOP1SRC2(imul, imul_ex);
923 FASTOP1SRC2EX(div, div_ex);
924 FASTOP1SRC2EX(idiv, idiv_ex);
925 
926 FASTOP3WCL(shld);
927 FASTOP3WCL(shrd);
928 
929 FASTOP2W(imul);
930 
931 FASTOP1(not);
932 FASTOP1(neg);
933 FASTOP1(inc);
934 FASTOP1(dec);
935 
936 FASTOP2CL(rol);
937 FASTOP2CL(ror);
938 FASTOP2CL(rcl);
939 FASTOP2CL(rcr);
940 FASTOP2CL(shl);
941 FASTOP2CL(shr);
942 FASTOP2CL(sar);
943 
944 FASTOP2W(bsf);
945 FASTOP2W(bsr);
946 FASTOP2W(bt);
947 FASTOP2W(bts);
948 FASTOP2W(btr);
949 FASTOP2W(btc);
950 
951 FASTOP2(xadd);
952 
953 FASTOP2R(cmp, cmp_r);
954 
955 static int em_bsf_c(struct x86_emulate_ctxt *ctxt)
956 {
957 	/* If src is zero, do not writeback, but update flags */
958 	if (ctxt->src.val == 0)
959 		ctxt->dst.type = OP_NONE;
960 	return fastop(ctxt, em_bsf);
961 }
962 
963 static int em_bsr_c(struct x86_emulate_ctxt *ctxt)
964 {
965 	/* If src is zero, do not writeback, but update flags */
966 	if (ctxt->src.val == 0)
967 		ctxt->dst.type = OP_NONE;
968 	return fastop(ctxt, em_bsr);
969 }
970 
971 static __always_inline u8 test_cc(unsigned int condition, unsigned long flags)
972 {
973 	u8 rc;
974 	void (*fop)(void) = (void *)em_setcc + 4 * (condition & 0xf);
975 
976 	flags = (flags & EFLAGS_MASK) | X86_EFLAGS_IF;
977 	asm("push %[flags]; popf; call *%[fastop]"
978 	    : "=a"(rc) : [fastop]"r"(fop), [flags]"r"(flags));
979 	return rc;
980 }
981 
982 static void fetch_register_operand(struct operand *op)
983 {
984 	switch (op->bytes) {
985 	case 1:
986 		op->val = *(u8 *)op->addr.reg;
987 		break;
988 	case 2:
989 		op->val = *(u16 *)op->addr.reg;
990 		break;
991 	case 4:
992 		op->val = *(u32 *)op->addr.reg;
993 		break;
994 	case 8:
995 		op->val = *(u64 *)op->addr.reg;
996 		break;
997 	}
998 }
999 
1000 static void read_sse_reg(struct x86_emulate_ctxt *ctxt, sse128_t *data, int reg)
1001 {
1002 	ctxt->ops->get_fpu(ctxt);
1003 	switch (reg) {
1004 	case 0: asm("movdqa %%xmm0, %0" : "=m"(*data)); break;
1005 	case 1: asm("movdqa %%xmm1, %0" : "=m"(*data)); break;
1006 	case 2: asm("movdqa %%xmm2, %0" : "=m"(*data)); break;
1007 	case 3: asm("movdqa %%xmm3, %0" : "=m"(*data)); break;
1008 	case 4: asm("movdqa %%xmm4, %0" : "=m"(*data)); break;
1009 	case 5: asm("movdqa %%xmm5, %0" : "=m"(*data)); break;
1010 	case 6: asm("movdqa %%xmm6, %0" : "=m"(*data)); break;
1011 	case 7: asm("movdqa %%xmm7, %0" : "=m"(*data)); break;
1012 #ifdef CONFIG_X86_64
1013 	case 8: asm("movdqa %%xmm8, %0" : "=m"(*data)); break;
1014 	case 9: asm("movdqa %%xmm9, %0" : "=m"(*data)); break;
1015 	case 10: asm("movdqa %%xmm10, %0" : "=m"(*data)); break;
1016 	case 11: asm("movdqa %%xmm11, %0" : "=m"(*data)); break;
1017 	case 12: asm("movdqa %%xmm12, %0" : "=m"(*data)); break;
1018 	case 13: asm("movdqa %%xmm13, %0" : "=m"(*data)); break;
1019 	case 14: asm("movdqa %%xmm14, %0" : "=m"(*data)); break;
1020 	case 15: asm("movdqa %%xmm15, %0" : "=m"(*data)); break;
1021 #endif
1022 	default: BUG();
1023 	}
1024 	ctxt->ops->put_fpu(ctxt);
1025 }
1026 
1027 static void write_sse_reg(struct x86_emulate_ctxt *ctxt, sse128_t *data,
1028 			  int reg)
1029 {
1030 	ctxt->ops->get_fpu(ctxt);
1031 	switch (reg) {
1032 	case 0: asm("movdqa %0, %%xmm0" : : "m"(*data)); break;
1033 	case 1: asm("movdqa %0, %%xmm1" : : "m"(*data)); break;
1034 	case 2: asm("movdqa %0, %%xmm2" : : "m"(*data)); break;
1035 	case 3: asm("movdqa %0, %%xmm3" : : "m"(*data)); break;
1036 	case 4: asm("movdqa %0, %%xmm4" : : "m"(*data)); break;
1037 	case 5: asm("movdqa %0, %%xmm5" : : "m"(*data)); break;
1038 	case 6: asm("movdqa %0, %%xmm6" : : "m"(*data)); break;
1039 	case 7: asm("movdqa %0, %%xmm7" : : "m"(*data)); break;
1040 #ifdef CONFIG_X86_64
1041 	case 8: asm("movdqa %0, %%xmm8" : : "m"(*data)); break;
1042 	case 9: asm("movdqa %0, %%xmm9" : : "m"(*data)); break;
1043 	case 10: asm("movdqa %0, %%xmm10" : : "m"(*data)); break;
1044 	case 11: asm("movdqa %0, %%xmm11" : : "m"(*data)); break;
1045 	case 12: asm("movdqa %0, %%xmm12" : : "m"(*data)); break;
1046 	case 13: asm("movdqa %0, %%xmm13" : : "m"(*data)); break;
1047 	case 14: asm("movdqa %0, %%xmm14" : : "m"(*data)); break;
1048 	case 15: asm("movdqa %0, %%xmm15" : : "m"(*data)); break;
1049 #endif
1050 	default: BUG();
1051 	}
1052 	ctxt->ops->put_fpu(ctxt);
1053 }
1054 
1055 static void read_mmx_reg(struct x86_emulate_ctxt *ctxt, u64 *data, int reg)
1056 {
1057 	ctxt->ops->get_fpu(ctxt);
1058 	switch (reg) {
1059 	case 0: asm("movq %%mm0, %0" : "=m"(*data)); break;
1060 	case 1: asm("movq %%mm1, %0" : "=m"(*data)); break;
1061 	case 2: asm("movq %%mm2, %0" : "=m"(*data)); break;
1062 	case 3: asm("movq %%mm3, %0" : "=m"(*data)); break;
1063 	case 4: asm("movq %%mm4, %0" : "=m"(*data)); break;
1064 	case 5: asm("movq %%mm5, %0" : "=m"(*data)); break;
1065 	case 6: asm("movq %%mm6, %0" : "=m"(*data)); break;
1066 	case 7: asm("movq %%mm7, %0" : "=m"(*data)); break;
1067 	default: BUG();
1068 	}
1069 	ctxt->ops->put_fpu(ctxt);
1070 }
1071 
1072 static void write_mmx_reg(struct x86_emulate_ctxt *ctxt, u64 *data, int reg)
1073 {
1074 	ctxt->ops->get_fpu(ctxt);
1075 	switch (reg) {
1076 	case 0: asm("movq %0, %%mm0" : : "m"(*data)); break;
1077 	case 1: asm("movq %0, %%mm1" : : "m"(*data)); break;
1078 	case 2: asm("movq %0, %%mm2" : : "m"(*data)); break;
1079 	case 3: asm("movq %0, %%mm3" : : "m"(*data)); break;
1080 	case 4: asm("movq %0, %%mm4" : : "m"(*data)); break;
1081 	case 5: asm("movq %0, %%mm5" : : "m"(*data)); break;
1082 	case 6: asm("movq %0, %%mm6" : : "m"(*data)); break;
1083 	case 7: asm("movq %0, %%mm7" : : "m"(*data)); break;
1084 	default: BUG();
1085 	}
1086 	ctxt->ops->put_fpu(ctxt);
1087 }
1088 
1089 static int em_fninit(struct x86_emulate_ctxt *ctxt)
1090 {
1091 	if (ctxt->ops->get_cr(ctxt, 0) & (X86_CR0_TS | X86_CR0_EM))
1092 		return emulate_nm(ctxt);
1093 
1094 	ctxt->ops->get_fpu(ctxt);
1095 	asm volatile("fninit");
1096 	ctxt->ops->put_fpu(ctxt);
1097 	return X86EMUL_CONTINUE;
1098 }
1099 
1100 static int em_fnstcw(struct x86_emulate_ctxt *ctxt)
1101 {
1102 	u16 fcw;
1103 
1104 	if (ctxt->ops->get_cr(ctxt, 0) & (X86_CR0_TS | X86_CR0_EM))
1105 		return emulate_nm(ctxt);
1106 
1107 	ctxt->ops->get_fpu(ctxt);
1108 	asm volatile("fnstcw %0": "+m"(fcw));
1109 	ctxt->ops->put_fpu(ctxt);
1110 
1111 	ctxt->dst.val = fcw;
1112 
1113 	return X86EMUL_CONTINUE;
1114 }
1115 
1116 static int em_fnstsw(struct x86_emulate_ctxt *ctxt)
1117 {
1118 	u16 fsw;
1119 
1120 	if (ctxt->ops->get_cr(ctxt, 0) & (X86_CR0_TS | X86_CR0_EM))
1121 		return emulate_nm(ctxt);
1122 
1123 	ctxt->ops->get_fpu(ctxt);
1124 	asm volatile("fnstsw %0": "+m"(fsw));
1125 	ctxt->ops->put_fpu(ctxt);
1126 
1127 	ctxt->dst.val = fsw;
1128 
1129 	return X86EMUL_CONTINUE;
1130 }
1131 
1132 static void decode_register_operand(struct x86_emulate_ctxt *ctxt,
1133 				    struct operand *op)
1134 {
1135 	unsigned reg = ctxt->modrm_reg;
1136 
1137 	if (!(ctxt->d & ModRM))
1138 		reg = (ctxt->b & 7) | ((ctxt->rex_prefix & 1) << 3);
1139 
1140 	if (ctxt->d & Sse) {
1141 		op->type = OP_XMM;
1142 		op->bytes = 16;
1143 		op->addr.xmm = reg;
1144 		read_sse_reg(ctxt, &op->vec_val, reg);
1145 		return;
1146 	}
1147 	if (ctxt->d & Mmx) {
1148 		reg &= 7;
1149 		op->type = OP_MM;
1150 		op->bytes = 8;
1151 		op->addr.mm = reg;
1152 		return;
1153 	}
1154 
1155 	op->type = OP_REG;
1156 	op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
1157 	op->addr.reg = decode_register(ctxt, reg, ctxt->d & ByteOp);
1158 
1159 	fetch_register_operand(op);
1160 	op->orig_val = op->val;
1161 }
1162 
1163 static void adjust_modrm_seg(struct x86_emulate_ctxt *ctxt, int base_reg)
1164 {
1165 	if (base_reg == VCPU_REGS_RSP || base_reg == VCPU_REGS_RBP)
1166 		ctxt->modrm_seg = VCPU_SREG_SS;
1167 }
1168 
1169 static int decode_modrm(struct x86_emulate_ctxt *ctxt,
1170 			struct operand *op)
1171 {
1172 	u8 sib;
1173 	int index_reg, base_reg, scale;
1174 	int rc = X86EMUL_CONTINUE;
1175 	ulong modrm_ea = 0;
1176 
1177 	ctxt->modrm_reg = ((ctxt->rex_prefix << 1) & 8); /* REX.R */
1178 	index_reg = (ctxt->rex_prefix << 2) & 8; /* REX.X */
1179 	base_reg = (ctxt->rex_prefix << 3) & 8; /* REX.B */
1180 
1181 	ctxt->modrm_mod = (ctxt->modrm & 0xc0) >> 6;
1182 	ctxt->modrm_reg |= (ctxt->modrm & 0x38) >> 3;
1183 	ctxt->modrm_rm = base_reg | (ctxt->modrm & 0x07);
1184 	ctxt->modrm_seg = VCPU_SREG_DS;
1185 
1186 	if (ctxt->modrm_mod == 3 || (ctxt->d & NoMod)) {
1187 		op->type = OP_REG;
1188 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
1189 		op->addr.reg = decode_register(ctxt, ctxt->modrm_rm,
1190 				ctxt->d & ByteOp);
1191 		if (ctxt->d & Sse) {
1192 			op->type = OP_XMM;
1193 			op->bytes = 16;
1194 			op->addr.xmm = ctxt->modrm_rm;
1195 			read_sse_reg(ctxt, &op->vec_val, ctxt->modrm_rm);
1196 			return rc;
1197 		}
1198 		if (ctxt->d & Mmx) {
1199 			op->type = OP_MM;
1200 			op->bytes = 8;
1201 			op->addr.mm = ctxt->modrm_rm & 7;
1202 			return rc;
1203 		}
1204 		fetch_register_operand(op);
1205 		return rc;
1206 	}
1207 
1208 	op->type = OP_MEM;
1209 
1210 	if (ctxt->ad_bytes == 2) {
1211 		unsigned bx = reg_read(ctxt, VCPU_REGS_RBX);
1212 		unsigned bp = reg_read(ctxt, VCPU_REGS_RBP);
1213 		unsigned si = reg_read(ctxt, VCPU_REGS_RSI);
1214 		unsigned di = reg_read(ctxt, VCPU_REGS_RDI);
1215 
1216 		/* 16-bit ModR/M decode. */
1217 		switch (ctxt->modrm_mod) {
1218 		case 0:
1219 			if (ctxt->modrm_rm == 6)
1220 				modrm_ea += insn_fetch(u16, ctxt);
1221 			break;
1222 		case 1:
1223 			modrm_ea += insn_fetch(s8, ctxt);
1224 			break;
1225 		case 2:
1226 			modrm_ea += insn_fetch(u16, ctxt);
1227 			break;
1228 		}
1229 		switch (ctxt->modrm_rm) {
1230 		case 0:
1231 			modrm_ea += bx + si;
1232 			break;
1233 		case 1:
1234 			modrm_ea += bx + di;
1235 			break;
1236 		case 2:
1237 			modrm_ea += bp + si;
1238 			break;
1239 		case 3:
1240 			modrm_ea += bp + di;
1241 			break;
1242 		case 4:
1243 			modrm_ea += si;
1244 			break;
1245 		case 5:
1246 			modrm_ea += di;
1247 			break;
1248 		case 6:
1249 			if (ctxt->modrm_mod != 0)
1250 				modrm_ea += bp;
1251 			break;
1252 		case 7:
1253 			modrm_ea += bx;
1254 			break;
1255 		}
1256 		if (ctxt->modrm_rm == 2 || ctxt->modrm_rm == 3 ||
1257 		    (ctxt->modrm_rm == 6 && ctxt->modrm_mod != 0))
1258 			ctxt->modrm_seg = VCPU_SREG_SS;
1259 		modrm_ea = (u16)modrm_ea;
1260 	} else {
1261 		/* 32/64-bit ModR/M decode. */
1262 		if ((ctxt->modrm_rm & 7) == 4) {
1263 			sib = insn_fetch(u8, ctxt);
1264 			index_reg |= (sib >> 3) & 7;
1265 			base_reg |= sib & 7;
1266 			scale = sib >> 6;
1267 
1268 			if ((base_reg & 7) == 5 && ctxt->modrm_mod == 0)
1269 				modrm_ea += insn_fetch(s32, ctxt);
1270 			else {
1271 				modrm_ea += reg_read(ctxt, base_reg);
1272 				adjust_modrm_seg(ctxt, base_reg);
1273 				/* Increment ESP on POP [ESP] */
1274 				if ((ctxt->d & IncSP) &&
1275 				    base_reg == VCPU_REGS_RSP)
1276 					modrm_ea += ctxt->op_bytes;
1277 			}
1278 			if (index_reg != 4)
1279 				modrm_ea += reg_read(ctxt, index_reg) << scale;
1280 		} else if ((ctxt->modrm_rm & 7) == 5 && ctxt->modrm_mod == 0) {
1281 			modrm_ea += insn_fetch(s32, ctxt);
1282 			if (ctxt->mode == X86EMUL_MODE_PROT64)
1283 				ctxt->rip_relative = 1;
1284 		} else {
1285 			base_reg = ctxt->modrm_rm;
1286 			modrm_ea += reg_read(ctxt, base_reg);
1287 			adjust_modrm_seg(ctxt, base_reg);
1288 		}
1289 		switch (ctxt->modrm_mod) {
1290 		case 1:
1291 			modrm_ea += insn_fetch(s8, ctxt);
1292 			break;
1293 		case 2:
1294 			modrm_ea += insn_fetch(s32, ctxt);
1295 			break;
1296 		}
1297 	}
1298 	op->addr.mem.ea = modrm_ea;
1299 	if (ctxt->ad_bytes != 8)
1300 		ctxt->memop.addr.mem.ea = (u32)ctxt->memop.addr.mem.ea;
1301 
1302 done:
1303 	return rc;
1304 }
1305 
1306 static int decode_abs(struct x86_emulate_ctxt *ctxt,
1307 		      struct operand *op)
1308 {
1309 	int rc = X86EMUL_CONTINUE;
1310 
1311 	op->type = OP_MEM;
1312 	switch (ctxt->ad_bytes) {
1313 	case 2:
1314 		op->addr.mem.ea = insn_fetch(u16, ctxt);
1315 		break;
1316 	case 4:
1317 		op->addr.mem.ea = insn_fetch(u32, ctxt);
1318 		break;
1319 	case 8:
1320 		op->addr.mem.ea = insn_fetch(u64, ctxt);
1321 		break;
1322 	}
1323 done:
1324 	return rc;
1325 }
1326 
1327 static void fetch_bit_operand(struct x86_emulate_ctxt *ctxt)
1328 {
1329 	long sv = 0, mask;
1330 
1331 	if (ctxt->dst.type == OP_MEM && ctxt->src.type == OP_REG) {
1332 		mask = ~((long)ctxt->dst.bytes * 8 - 1);
1333 
1334 		if (ctxt->src.bytes == 2)
1335 			sv = (s16)ctxt->src.val & (s16)mask;
1336 		else if (ctxt->src.bytes == 4)
1337 			sv = (s32)ctxt->src.val & (s32)mask;
1338 		else
1339 			sv = (s64)ctxt->src.val & (s64)mask;
1340 
1341 		ctxt->dst.addr.mem.ea = address_mask(ctxt,
1342 					   ctxt->dst.addr.mem.ea + (sv >> 3));
1343 	}
1344 
1345 	/* only subword offset */
1346 	ctxt->src.val &= (ctxt->dst.bytes << 3) - 1;
1347 }
1348 
1349 static int read_emulated(struct x86_emulate_ctxt *ctxt,
1350 			 unsigned long addr, void *dest, unsigned size)
1351 {
1352 	int rc;
1353 	struct read_cache *mc = &ctxt->mem_read;
1354 
1355 	if (mc->pos < mc->end)
1356 		goto read_cached;
1357 
1358 	WARN_ON((mc->end + size) >= sizeof(mc->data));
1359 
1360 	rc = ctxt->ops->read_emulated(ctxt, addr, mc->data + mc->end, size,
1361 				      &ctxt->exception);
1362 	if (rc != X86EMUL_CONTINUE)
1363 		return rc;
1364 
1365 	mc->end += size;
1366 
1367 read_cached:
1368 	memcpy(dest, mc->data + mc->pos, size);
1369 	mc->pos += size;
1370 	return X86EMUL_CONTINUE;
1371 }
1372 
1373 static int segmented_read(struct x86_emulate_ctxt *ctxt,
1374 			  struct segmented_address addr,
1375 			  void *data,
1376 			  unsigned size)
1377 {
1378 	int rc;
1379 	ulong linear;
1380 
1381 	rc = linearize(ctxt, addr, size, false, &linear);
1382 	if (rc != X86EMUL_CONTINUE)
1383 		return rc;
1384 	return read_emulated(ctxt, linear, data, size);
1385 }
1386 
1387 static int segmented_write(struct x86_emulate_ctxt *ctxt,
1388 			   struct segmented_address addr,
1389 			   const void *data,
1390 			   unsigned size)
1391 {
1392 	int rc;
1393 	ulong linear;
1394 
1395 	rc = linearize(ctxt, addr, size, true, &linear);
1396 	if (rc != X86EMUL_CONTINUE)
1397 		return rc;
1398 	return ctxt->ops->write_emulated(ctxt, linear, data, size,
1399 					 &ctxt->exception);
1400 }
1401 
1402 static int segmented_cmpxchg(struct x86_emulate_ctxt *ctxt,
1403 			     struct segmented_address addr,
1404 			     const void *orig_data, const void *data,
1405 			     unsigned size)
1406 {
1407 	int rc;
1408 	ulong linear;
1409 
1410 	rc = linearize(ctxt, addr, size, true, &linear);
1411 	if (rc != X86EMUL_CONTINUE)
1412 		return rc;
1413 	return ctxt->ops->cmpxchg_emulated(ctxt, linear, orig_data, data,
1414 					   size, &ctxt->exception);
1415 }
1416 
1417 static int pio_in_emulated(struct x86_emulate_ctxt *ctxt,
1418 			   unsigned int size, unsigned short port,
1419 			   void *dest)
1420 {
1421 	struct read_cache *rc = &ctxt->io_read;
1422 
1423 	if (rc->pos == rc->end) { /* refill pio read ahead */
1424 		unsigned int in_page, n;
1425 		unsigned int count = ctxt->rep_prefix ?
1426 			address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) : 1;
1427 		in_page = (ctxt->eflags & X86_EFLAGS_DF) ?
1428 			offset_in_page(reg_read(ctxt, VCPU_REGS_RDI)) :
1429 			PAGE_SIZE - offset_in_page(reg_read(ctxt, VCPU_REGS_RDI));
1430 		n = min3(in_page, (unsigned int)sizeof(rc->data) / size, count);
1431 		if (n == 0)
1432 			n = 1;
1433 		rc->pos = rc->end = 0;
1434 		if (!ctxt->ops->pio_in_emulated(ctxt, size, port, rc->data, n))
1435 			return 0;
1436 		rc->end = n * size;
1437 	}
1438 
1439 	if (ctxt->rep_prefix && (ctxt->d & String) &&
1440 	    !(ctxt->eflags & X86_EFLAGS_DF)) {
1441 		ctxt->dst.data = rc->data + rc->pos;
1442 		ctxt->dst.type = OP_MEM_STR;
1443 		ctxt->dst.count = (rc->end - rc->pos) / size;
1444 		rc->pos = rc->end;
1445 	} else {
1446 		memcpy(dest, rc->data + rc->pos, size);
1447 		rc->pos += size;
1448 	}
1449 	return 1;
1450 }
1451 
1452 static int read_interrupt_descriptor(struct x86_emulate_ctxt *ctxt,
1453 				     u16 index, struct desc_struct *desc)
1454 {
1455 	struct desc_ptr dt;
1456 	ulong addr;
1457 
1458 	ctxt->ops->get_idt(ctxt, &dt);
1459 
1460 	if (dt.size < index * 8 + 7)
1461 		return emulate_gp(ctxt, index << 3 | 0x2);
1462 
1463 	addr = dt.address + index * 8;
1464 	return ctxt->ops->read_std(ctxt, addr, desc, sizeof *desc,
1465 				   &ctxt->exception);
1466 }
1467 
1468 static void get_descriptor_table_ptr(struct x86_emulate_ctxt *ctxt,
1469 				     u16 selector, struct desc_ptr *dt)
1470 {
1471 	const struct x86_emulate_ops *ops = ctxt->ops;
1472 	u32 base3 = 0;
1473 
1474 	if (selector & 1 << 2) {
1475 		struct desc_struct desc;
1476 		u16 sel;
1477 
1478 		memset (dt, 0, sizeof *dt);
1479 		if (!ops->get_segment(ctxt, &sel, &desc, &base3,
1480 				      VCPU_SREG_LDTR))
1481 			return;
1482 
1483 		dt->size = desc_limit_scaled(&desc); /* what if limit > 65535? */
1484 		dt->address = get_desc_base(&desc) | ((u64)base3 << 32);
1485 	} else
1486 		ops->get_gdt(ctxt, dt);
1487 }
1488 
1489 static int get_descriptor_ptr(struct x86_emulate_ctxt *ctxt,
1490 			      u16 selector, ulong *desc_addr_p)
1491 {
1492 	struct desc_ptr dt;
1493 	u16 index = selector >> 3;
1494 	ulong addr;
1495 
1496 	get_descriptor_table_ptr(ctxt, selector, &dt);
1497 
1498 	if (dt.size < index * 8 + 7)
1499 		return emulate_gp(ctxt, selector & 0xfffc);
1500 
1501 	addr = dt.address + index * 8;
1502 
1503 #ifdef CONFIG_X86_64
1504 	if (addr >> 32 != 0) {
1505 		u64 efer = 0;
1506 
1507 		ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
1508 		if (!(efer & EFER_LMA))
1509 			addr &= (u32)-1;
1510 	}
1511 #endif
1512 
1513 	*desc_addr_p = addr;
1514 	return X86EMUL_CONTINUE;
1515 }
1516 
1517 /* allowed just for 8 bytes segments */
1518 static int read_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1519 				   u16 selector, struct desc_struct *desc,
1520 				   ulong *desc_addr_p)
1521 {
1522 	int rc;
1523 
1524 	rc = get_descriptor_ptr(ctxt, selector, desc_addr_p);
1525 	if (rc != X86EMUL_CONTINUE)
1526 		return rc;
1527 
1528 	return ctxt->ops->read_std(ctxt, *desc_addr_p, desc, sizeof(*desc),
1529 				   &ctxt->exception);
1530 }
1531 
1532 /* allowed just for 8 bytes segments */
1533 static int write_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1534 				    u16 selector, struct desc_struct *desc)
1535 {
1536 	int rc;
1537 	ulong addr;
1538 
1539 	rc = get_descriptor_ptr(ctxt, selector, &addr);
1540 	if (rc != X86EMUL_CONTINUE)
1541 		return rc;
1542 
1543 	return ctxt->ops->write_std(ctxt, addr, desc, sizeof *desc,
1544 				    &ctxt->exception);
1545 }
1546 
1547 /* Does not support long mode */
1548 static int __load_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1549 				     u16 selector, int seg, u8 cpl,
1550 				     enum x86_transfer_type transfer,
1551 				     struct desc_struct *desc)
1552 {
1553 	struct desc_struct seg_desc, old_desc;
1554 	u8 dpl, rpl;
1555 	unsigned err_vec = GP_VECTOR;
1556 	u32 err_code = 0;
1557 	bool null_selector = !(selector & ~0x3); /* 0000-0003 are null */
1558 	ulong desc_addr;
1559 	int ret;
1560 	u16 dummy;
1561 	u32 base3 = 0;
1562 
1563 	memset(&seg_desc, 0, sizeof seg_desc);
1564 
1565 	if (ctxt->mode == X86EMUL_MODE_REAL) {
1566 		/* set real mode segment descriptor (keep limit etc. for
1567 		 * unreal mode) */
1568 		ctxt->ops->get_segment(ctxt, &dummy, &seg_desc, NULL, seg);
1569 		set_desc_base(&seg_desc, selector << 4);
1570 		goto load;
1571 	} else if (seg <= VCPU_SREG_GS && ctxt->mode == X86EMUL_MODE_VM86) {
1572 		/* VM86 needs a clean new segment descriptor */
1573 		set_desc_base(&seg_desc, selector << 4);
1574 		set_desc_limit(&seg_desc, 0xffff);
1575 		seg_desc.type = 3;
1576 		seg_desc.p = 1;
1577 		seg_desc.s = 1;
1578 		seg_desc.dpl = 3;
1579 		goto load;
1580 	}
1581 
1582 	rpl = selector & 3;
1583 
1584 	/* NULL selector is not valid for TR, CS and SS (except for long mode) */
1585 	if ((seg == VCPU_SREG_CS
1586 	     || (seg == VCPU_SREG_SS
1587 		 && (ctxt->mode != X86EMUL_MODE_PROT64 || rpl != cpl))
1588 	     || seg == VCPU_SREG_TR)
1589 	    && null_selector)
1590 		goto exception;
1591 
1592 	/* TR should be in GDT only */
1593 	if (seg == VCPU_SREG_TR && (selector & (1 << 2)))
1594 		goto exception;
1595 
1596 	if (null_selector) /* for NULL selector skip all following checks */
1597 		goto load;
1598 
1599 	ret = read_segment_descriptor(ctxt, selector, &seg_desc, &desc_addr);
1600 	if (ret != X86EMUL_CONTINUE)
1601 		return ret;
1602 
1603 	err_code = selector & 0xfffc;
1604 	err_vec = (transfer == X86_TRANSFER_TASK_SWITCH) ? TS_VECTOR :
1605 							   GP_VECTOR;
1606 
1607 	/* can't load system descriptor into segment selector */
1608 	if (seg <= VCPU_SREG_GS && !seg_desc.s) {
1609 		if (transfer == X86_TRANSFER_CALL_JMP)
1610 			return X86EMUL_UNHANDLEABLE;
1611 		goto exception;
1612 	}
1613 
1614 	if (!seg_desc.p) {
1615 		err_vec = (seg == VCPU_SREG_SS) ? SS_VECTOR : NP_VECTOR;
1616 		goto exception;
1617 	}
1618 
1619 	dpl = seg_desc.dpl;
1620 
1621 	switch (seg) {
1622 	case VCPU_SREG_SS:
1623 		/*
1624 		 * segment is not a writable data segment or segment
1625 		 * selector's RPL != CPL or segment selector's RPL != CPL
1626 		 */
1627 		if (rpl != cpl || (seg_desc.type & 0xa) != 0x2 || dpl != cpl)
1628 			goto exception;
1629 		break;
1630 	case VCPU_SREG_CS:
1631 		if (!(seg_desc.type & 8))
1632 			goto exception;
1633 
1634 		if (seg_desc.type & 4) {
1635 			/* conforming */
1636 			if (dpl > cpl)
1637 				goto exception;
1638 		} else {
1639 			/* nonconforming */
1640 			if (rpl > cpl || dpl != cpl)
1641 				goto exception;
1642 		}
1643 		/* in long-mode d/b must be clear if l is set */
1644 		if (seg_desc.d && seg_desc.l) {
1645 			u64 efer = 0;
1646 
1647 			ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
1648 			if (efer & EFER_LMA)
1649 				goto exception;
1650 		}
1651 
1652 		/* CS(RPL) <- CPL */
1653 		selector = (selector & 0xfffc) | cpl;
1654 		break;
1655 	case VCPU_SREG_TR:
1656 		if (seg_desc.s || (seg_desc.type != 1 && seg_desc.type != 9))
1657 			goto exception;
1658 		old_desc = seg_desc;
1659 		seg_desc.type |= 2; /* busy */
1660 		ret = ctxt->ops->cmpxchg_emulated(ctxt, desc_addr, &old_desc, &seg_desc,
1661 						  sizeof(seg_desc), &ctxt->exception);
1662 		if (ret != X86EMUL_CONTINUE)
1663 			return ret;
1664 		break;
1665 	case VCPU_SREG_LDTR:
1666 		if (seg_desc.s || seg_desc.type != 2)
1667 			goto exception;
1668 		break;
1669 	default: /*  DS, ES, FS, or GS */
1670 		/*
1671 		 * segment is not a data or readable code segment or
1672 		 * ((segment is a data or nonconforming code segment)
1673 		 * and (both RPL and CPL > DPL))
1674 		 */
1675 		if ((seg_desc.type & 0xa) == 0x8 ||
1676 		    (((seg_desc.type & 0xc) != 0xc) &&
1677 		     (rpl > dpl && cpl > dpl)))
1678 			goto exception;
1679 		break;
1680 	}
1681 
1682 	if (seg_desc.s) {
1683 		/* mark segment as accessed */
1684 		if (!(seg_desc.type & 1)) {
1685 			seg_desc.type |= 1;
1686 			ret = write_segment_descriptor(ctxt, selector,
1687 						       &seg_desc);
1688 			if (ret != X86EMUL_CONTINUE)
1689 				return ret;
1690 		}
1691 	} else if (ctxt->mode == X86EMUL_MODE_PROT64) {
1692 		ret = ctxt->ops->read_std(ctxt, desc_addr+8, &base3,
1693 				sizeof(base3), &ctxt->exception);
1694 		if (ret != X86EMUL_CONTINUE)
1695 			return ret;
1696 		if (is_noncanonical_address(get_desc_base(&seg_desc) |
1697 					     ((u64)base3 << 32)))
1698 			return emulate_gp(ctxt, 0);
1699 	}
1700 load:
1701 	ctxt->ops->set_segment(ctxt, selector, &seg_desc, base3, seg);
1702 	if (desc)
1703 		*desc = seg_desc;
1704 	return X86EMUL_CONTINUE;
1705 exception:
1706 	return emulate_exception(ctxt, err_vec, err_code, true);
1707 }
1708 
1709 static int load_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1710 				   u16 selector, int seg)
1711 {
1712 	u8 cpl = ctxt->ops->cpl(ctxt);
1713 	return __load_segment_descriptor(ctxt, selector, seg, cpl,
1714 					 X86_TRANSFER_NONE, NULL);
1715 }
1716 
1717 static void write_register_operand(struct operand *op)
1718 {
1719 	return assign_register(op->addr.reg, op->val, op->bytes);
1720 }
1721 
1722 static int writeback(struct x86_emulate_ctxt *ctxt, struct operand *op)
1723 {
1724 	switch (op->type) {
1725 	case OP_REG:
1726 		write_register_operand(op);
1727 		break;
1728 	case OP_MEM:
1729 		if (ctxt->lock_prefix)
1730 			return segmented_cmpxchg(ctxt,
1731 						 op->addr.mem,
1732 						 &op->orig_val,
1733 						 &op->val,
1734 						 op->bytes);
1735 		else
1736 			return segmented_write(ctxt,
1737 					       op->addr.mem,
1738 					       &op->val,
1739 					       op->bytes);
1740 		break;
1741 	case OP_MEM_STR:
1742 		return segmented_write(ctxt,
1743 				       op->addr.mem,
1744 				       op->data,
1745 				       op->bytes * op->count);
1746 		break;
1747 	case OP_XMM:
1748 		write_sse_reg(ctxt, &op->vec_val, op->addr.xmm);
1749 		break;
1750 	case OP_MM:
1751 		write_mmx_reg(ctxt, &op->mm_val, op->addr.mm);
1752 		break;
1753 	case OP_NONE:
1754 		/* no writeback */
1755 		break;
1756 	default:
1757 		break;
1758 	}
1759 	return X86EMUL_CONTINUE;
1760 }
1761 
1762 static int push(struct x86_emulate_ctxt *ctxt, void *data, int bytes)
1763 {
1764 	struct segmented_address addr;
1765 
1766 	rsp_increment(ctxt, -bytes);
1767 	addr.ea = reg_read(ctxt, VCPU_REGS_RSP) & stack_mask(ctxt);
1768 	addr.seg = VCPU_SREG_SS;
1769 
1770 	return segmented_write(ctxt, addr, data, bytes);
1771 }
1772 
1773 static int em_push(struct x86_emulate_ctxt *ctxt)
1774 {
1775 	/* Disable writeback. */
1776 	ctxt->dst.type = OP_NONE;
1777 	return push(ctxt, &ctxt->src.val, ctxt->op_bytes);
1778 }
1779 
1780 static int emulate_pop(struct x86_emulate_ctxt *ctxt,
1781 		       void *dest, int len)
1782 {
1783 	int rc;
1784 	struct segmented_address addr;
1785 
1786 	addr.ea = reg_read(ctxt, VCPU_REGS_RSP) & stack_mask(ctxt);
1787 	addr.seg = VCPU_SREG_SS;
1788 	rc = segmented_read(ctxt, addr, dest, len);
1789 	if (rc != X86EMUL_CONTINUE)
1790 		return rc;
1791 
1792 	rsp_increment(ctxt, len);
1793 	return rc;
1794 }
1795 
1796 static int em_pop(struct x86_emulate_ctxt *ctxt)
1797 {
1798 	return emulate_pop(ctxt, &ctxt->dst.val, ctxt->op_bytes);
1799 }
1800 
1801 static int emulate_popf(struct x86_emulate_ctxt *ctxt,
1802 			void *dest, int len)
1803 {
1804 	int rc;
1805 	unsigned long val, change_mask;
1806 	int iopl = (ctxt->eflags & X86_EFLAGS_IOPL) >> X86_EFLAGS_IOPL_BIT;
1807 	int cpl = ctxt->ops->cpl(ctxt);
1808 
1809 	rc = emulate_pop(ctxt, &val, len);
1810 	if (rc != X86EMUL_CONTINUE)
1811 		return rc;
1812 
1813 	change_mask = X86_EFLAGS_CF | X86_EFLAGS_PF | X86_EFLAGS_AF |
1814 		      X86_EFLAGS_ZF | X86_EFLAGS_SF | X86_EFLAGS_OF |
1815 		      X86_EFLAGS_TF | X86_EFLAGS_DF | X86_EFLAGS_NT |
1816 		      X86_EFLAGS_AC | X86_EFLAGS_ID;
1817 
1818 	switch(ctxt->mode) {
1819 	case X86EMUL_MODE_PROT64:
1820 	case X86EMUL_MODE_PROT32:
1821 	case X86EMUL_MODE_PROT16:
1822 		if (cpl == 0)
1823 			change_mask |= X86_EFLAGS_IOPL;
1824 		if (cpl <= iopl)
1825 			change_mask |= X86_EFLAGS_IF;
1826 		break;
1827 	case X86EMUL_MODE_VM86:
1828 		if (iopl < 3)
1829 			return emulate_gp(ctxt, 0);
1830 		change_mask |= X86_EFLAGS_IF;
1831 		break;
1832 	default: /* real mode */
1833 		change_mask |= (X86_EFLAGS_IOPL | X86_EFLAGS_IF);
1834 		break;
1835 	}
1836 
1837 	*(unsigned long *)dest =
1838 		(ctxt->eflags & ~change_mask) | (val & change_mask);
1839 
1840 	return rc;
1841 }
1842 
1843 static int em_popf(struct x86_emulate_ctxt *ctxt)
1844 {
1845 	ctxt->dst.type = OP_REG;
1846 	ctxt->dst.addr.reg = &ctxt->eflags;
1847 	ctxt->dst.bytes = ctxt->op_bytes;
1848 	return emulate_popf(ctxt, &ctxt->dst.val, ctxt->op_bytes);
1849 }
1850 
1851 static int em_enter(struct x86_emulate_ctxt *ctxt)
1852 {
1853 	int rc;
1854 	unsigned frame_size = ctxt->src.val;
1855 	unsigned nesting_level = ctxt->src2.val & 31;
1856 	ulong rbp;
1857 
1858 	if (nesting_level)
1859 		return X86EMUL_UNHANDLEABLE;
1860 
1861 	rbp = reg_read(ctxt, VCPU_REGS_RBP);
1862 	rc = push(ctxt, &rbp, stack_size(ctxt));
1863 	if (rc != X86EMUL_CONTINUE)
1864 		return rc;
1865 	assign_masked(reg_rmw(ctxt, VCPU_REGS_RBP), reg_read(ctxt, VCPU_REGS_RSP),
1866 		      stack_mask(ctxt));
1867 	assign_masked(reg_rmw(ctxt, VCPU_REGS_RSP),
1868 		      reg_read(ctxt, VCPU_REGS_RSP) - frame_size,
1869 		      stack_mask(ctxt));
1870 	return X86EMUL_CONTINUE;
1871 }
1872 
1873 static int em_leave(struct x86_emulate_ctxt *ctxt)
1874 {
1875 	assign_masked(reg_rmw(ctxt, VCPU_REGS_RSP), reg_read(ctxt, VCPU_REGS_RBP),
1876 		      stack_mask(ctxt));
1877 	return emulate_pop(ctxt, reg_rmw(ctxt, VCPU_REGS_RBP), ctxt->op_bytes);
1878 }
1879 
1880 static int em_push_sreg(struct x86_emulate_ctxt *ctxt)
1881 {
1882 	int seg = ctxt->src2.val;
1883 
1884 	ctxt->src.val = get_segment_selector(ctxt, seg);
1885 	if (ctxt->op_bytes == 4) {
1886 		rsp_increment(ctxt, -2);
1887 		ctxt->op_bytes = 2;
1888 	}
1889 
1890 	return em_push(ctxt);
1891 }
1892 
1893 static int em_pop_sreg(struct x86_emulate_ctxt *ctxt)
1894 {
1895 	int seg = ctxt->src2.val;
1896 	unsigned long selector;
1897 	int rc;
1898 
1899 	rc = emulate_pop(ctxt, &selector, 2);
1900 	if (rc != X86EMUL_CONTINUE)
1901 		return rc;
1902 
1903 	if (ctxt->modrm_reg == VCPU_SREG_SS)
1904 		ctxt->interruptibility = KVM_X86_SHADOW_INT_MOV_SS;
1905 	if (ctxt->op_bytes > 2)
1906 		rsp_increment(ctxt, ctxt->op_bytes - 2);
1907 
1908 	rc = load_segment_descriptor(ctxt, (u16)selector, seg);
1909 	return rc;
1910 }
1911 
1912 static int em_pusha(struct x86_emulate_ctxt *ctxt)
1913 {
1914 	unsigned long old_esp = reg_read(ctxt, VCPU_REGS_RSP);
1915 	int rc = X86EMUL_CONTINUE;
1916 	int reg = VCPU_REGS_RAX;
1917 
1918 	while (reg <= VCPU_REGS_RDI) {
1919 		(reg == VCPU_REGS_RSP) ?
1920 		(ctxt->src.val = old_esp) : (ctxt->src.val = reg_read(ctxt, reg));
1921 
1922 		rc = em_push(ctxt);
1923 		if (rc != X86EMUL_CONTINUE)
1924 			return rc;
1925 
1926 		++reg;
1927 	}
1928 
1929 	return rc;
1930 }
1931 
1932 static int em_pushf(struct x86_emulate_ctxt *ctxt)
1933 {
1934 	ctxt->src.val = (unsigned long)ctxt->eflags & ~X86_EFLAGS_VM;
1935 	return em_push(ctxt);
1936 }
1937 
1938 static int em_popa(struct x86_emulate_ctxt *ctxt)
1939 {
1940 	int rc = X86EMUL_CONTINUE;
1941 	int reg = VCPU_REGS_RDI;
1942 	u32 val;
1943 
1944 	while (reg >= VCPU_REGS_RAX) {
1945 		if (reg == VCPU_REGS_RSP) {
1946 			rsp_increment(ctxt, ctxt->op_bytes);
1947 			--reg;
1948 		}
1949 
1950 		rc = emulate_pop(ctxt, &val, ctxt->op_bytes);
1951 		if (rc != X86EMUL_CONTINUE)
1952 			break;
1953 		assign_register(reg_rmw(ctxt, reg), val, ctxt->op_bytes);
1954 		--reg;
1955 	}
1956 	return rc;
1957 }
1958 
1959 static int __emulate_int_real(struct x86_emulate_ctxt *ctxt, int irq)
1960 {
1961 	const struct x86_emulate_ops *ops = ctxt->ops;
1962 	int rc;
1963 	struct desc_ptr dt;
1964 	gva_t cs_addr;
1965 	gva_t eip_addr;
1966 	u16 cs, eip;
1967 
1968 	/* TODO: Add limit checks */
1969 	ctxt->src.val = ctxt->eflags;
1970 	rc = em_push(ctxt);
1971 	if (rc != X86EMUL_CONTINUE)
1972 		return rc;
1973 
1974 	ctxt->eflags &= ~(X86_EFLAGS_IF | X86_EFLAGS_TF | X86_EFLAGS_AC);
1975 
1976 	ctxt->src.val = get_segment_selector(ctxt, VCPU_SREG_CS);
1977 	rc = em_push(ctxt);
1978 	if (rc != X86EMUL_CONTINUE)
1979 		return rc;
1980 
1981 	ctxt->src.val = ctxt->_eip;
1982 	rc = em_push(ctxt);
1983 	if (rc != X86EMUL_CONTINUE)
1984 		return rc;
1985 
1986 	ops->get_idt(ctxt, &dt);
1987 
1988 	eip_addr = dt.address + (irq << 2);
1989 	cs_addr = dt.address + (irq << 2) + 2;
1990 
1991 	rc = ops->read_std(ctxt, cs_addr, &cs, 2, &ctxt->exception);
1992 	if (rc != X86EMUL_CONTINUE)
1993 		return rc;
1994 
1995 	rc = ops->read_std(ctxt, eip_addr, &eip, 2, &ctxt->exception);
1996 	if (rc != X86EMUL_CONTINUE)
1997 		return rc;
1998 
1999 	rc = load_segment_descriptor(ctxt, cs, VCPU_SREG_CS);
2000 	if (rc != X86EMUL_CONTINUE)
2001 		return rc;
2002 
2003 	ctxt->_eip = eip;
2004 
2005 	return rc;
2006 }
2007 
2008 int emulate_int_real(struct x86_emulate_ctxt *ctxt, int irq)
2009 {
2010 	int rc;
2011 
2012 	invalidate_registers(ctxt);
2013 	rc = __emulate_int_real(ctxt, irq);
2014 	if (rc == X86EMUL_CONTINUE)
2015 		writeback_registers(ctxt);
2016 	return rc;
2017 }
2018 
2019 static int emulate_int(struct x86_emulate_ctxt *ctxt, int irq)
2020 {
2021 	switch(ctxt->mode) {
2022 	case X86EMUL_MODE_REAL:
2023 		return __emulate_int_real(ctxt, irq);
2024 	case X86EMUL_MODE_VM86:
2025 	case X86EMUL_MODE_PROT16:
2026 	case X86EMUL_MODE_PROT32:
2027 	case X86EMUL_MODE_PROT64:
2028 	default:
2029 		/* Protected mode interrupts unimplemented yet */
2030 		return X86EMUL_UNHANDLEABLE;
2031 	}
2032 }
2033 
2034 static int emulate_iret_real(struct x86_emulate_ctxt *ctxt)
2035 {
2036 	int rc = X86EMUL_CONTINUE;
2037 	unsigned long temp_eip = 0;
2038 	unsigned long temp_eflags = 0;
2039 	unsigned long cs = 0;
2040 	unsigned long mask = X86_EFLAGS_CF | X86_EFLAGS_PF | X86_EFLAGS_AF |
2041 			     X86_EFLAGS_ZF | X86_EFLAGS_SF | X86_EFLAGS_TF |
2042 			     X86_EFLAGS_IF | X86_EFLAGS_DF | X86_EFLAGS_OF |
2043 			     X86_EFLAGS_IOPL | X86_EFLAGS_NT | X86_EFLAGS_RF |
2044 			     X86_EFLAGS_AC | X86_EFLAGS_ID |
2045 			     X86_EFLAGS_FIXED;
2046 	unsigned long vm86_mask = X86_EFLAGS_VM | X86_EFLAGS_VIF |
2047 				  X86_EFLAGS_VIP;
2048 
2049 	/* TODO: Add stack limit check */
2050 
2051 	rc = emulate_pop(ctxt, &temp_eip, ctxt->op_bytes);
2052 
2053 	if (rc != X86EMUL_CONTINUE)
2054 		return rc;
2055 
2056 	if (temp_eip & ~0xffff)
2057 		return emulate_gp(ctxt, 0);
2058 
2059 	rc = emulate_pop(ctxt, &cs, ctxt->op_bytes);
2060 
2061 	if (rc != X86EMUL_CONTINUE)
2062 		return rc;
2063 
2064 	rc = emulate_pop(ctxt, &temp_eflags, ctxt->op_bytes);
2065 
2066 	if (rc != X86EMUL_CONTINUE)
2067 		return rc;
2068 
2069 	rc = load_segment_descriptor(ctxt, (u16)cs, VCPU_SREG_CS);
2070 
2071 	if (rc != X86EMUL_CONTINUE)
2072 		return rc;
2073 
2074 	ctxt->_eip = temp_eip;
2075 
2076 	if (ctxt->op_bytes == 4)
2077 		ctxt->eflags = ((temp_eflags & mask) | (ctxt->eflags & vm86_mask));
2078 	else if (ctxt->op_bytes == 2) {
2079 		ctxt->eflags &= ~0xffff;
2080 		ctxt->eflags |= temp_eflags;
2081 	}
2082 
2083 	ctxt->eflags &= ~EFLG_RESERVED_ZEROS_MASK; /* Clear reserved zeros */
2084 	ctxt->eflags |= X86_EFLAGS_FIXED;
2085 	ctxt->ops->set_nmi_mask(ctxt, false);
2086 
2087 	return rc;
2088 }
2089 
2090 static int em_iret(struct x86_emulate_ctxt *ctxt)
2091 {
2092 	switch(ctxt->mode) {
2093 	case X86EMUL_MODE_REAL:
2094 		return emulate_iret_real(ctxt);
2095 	case X86EMUL_MODE_VM86:
2096 	case X86EMUL_MODE_PROT16:
2097 	case X86EMUL_MODE_PROT32:
2098 	case X86EMUL_MODE_PROT64:
2099 	default:
2100 		/* iret from protected mode unimplemented yet */
2101 		return X86EMUL_UNHANDLEABLE;
2102 	}
2103 }
2104 
2105 static int em_jmp_far(struct x86_emulate_ctxt *ctxt)
2106 {
2107 	int rc;
2108 	unsigned short sel, old_sel;
2109 	struct desc_struct old_desc, new_desc;
2110 	const struct x86_emulate_ops *ops = ctxt->ops;
2111 	u8 cpl = ctxt->ops->cpl(ctxt);
2112 
2113 	/* Assignment of RIP may only fail in 64-bit mode */
2114 	if (ctxt->mode == X86EMUL_MODE_PROT64)
2115 		ops->get_segment(ctxt, &old_sel, &old_desc, NULL,
2116 				 VCPU_SREG_CS);
2117 
2118 	memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
2119 
2120 	rc = __load_segment_descriptor(ctxt, sel, VCPU_SREG_CS, cpl,
2121 				       X86_TRANSFER_CALL_JMP,
2122 				       &new_desc);
2123 	if (rc != X86EMUL_CONTINUE)
2124 		return rc;
2125 
2126 	rc = assign_eip_far(ctxt, ctxt->src.val, &new_desc);
2127 	if (rc != X86EMUL_CONTINUE) {
2128 		WARN_ON(ctxt->mode != X86EMUL_MODE_PROT64);
2129 		/* assigning eip failed; restore the old cs */
2130 		ops->set_segment(ctxt, old_sel, &old_desc, 0, VCPU_SREG_CS);
2131 		return rc;
2132 	}
2133 	return rc;
2134 }
2135 
2136 static int em_jmp_abs(struct x86_emulate_ctxt *ctxt)
2137 {
2138 	return assign_eip_near(ctxt, ctxt->src.val);
2139 }
2140 
2141 static int em_call_near_abs(struct x86_emulate_ctxt *ctxt)
2142 {
2143 	int rc;
2144 	long int old_eip;
2145 
2146 	old_eip = ctxt->_eip;
2147 	rc = assign_eip_near(ctxt, ctxt->src.val);
2148 	if (rc != X86EMUL_CONTINUE)
2149 		return rc;
2150 	ctxt->src.val = old_eip;
2151 	rc = em_push(ctxt);
2152 	return rc;
2153 }
2154 
2155 static int em_cmpxchg8b(struct x86_emulate_ctxt *ctxt)
2156 {
2157 	u64 old = ctxt->dst.orig_val64;
2158 
2159 	if (ctxt->dst.bytes == 16)
2160 		return X86EMUL_UNHANDLEABLE;
2161 
2162 	if (((u32) (old >> 0) != (u32) reg_read(ctxt, VCPU_REGS_RAX)) ||
2163 	    ((u32) (old >> 32) != (u32) reg_read(ctxt, VCPU_REGS_RDX))) {
2164 		*reg_write(ctxt, VCPU_REGS_RAX) = (u32) (old >> 0);
2165 		*reg_write(ctxt, VCPU_REGS_RDX) = (u32) (old >> 32);
2166 		ctxt->eflags &= ~X86_EFLAGS_ZF;
2167 	} else {
2168 		ctxt->dst.val64 = ((u64)reg_read(ctxt, VCPU_REGS_RCX) << 32) |
2169 			(u32) reg_read(ctxt, VCPU_REGS_RBX);
2170 
2171 		ctxt->eflags |= X86_EFLAGS_ZF;
2172 	}
2173 	return X86EMUL_CONTINUE;
2174 }
2175 
2176 static int em_ret(struct x86_emulate_ctxt *ctxt)
2177 {
2178 	int rc;
2179 	unsigned long eip;
2180 
2181 	rc = emulate_pop(ctxt, &eip, ctxt->op_bytes);
2182 	if (rc != X86EMUL_CONTINUE)
2183 		return rc;
2184 
2185 	return assign_eip_near(ctxt, eip);
2186 }
2187 
2188 static int em_ret_far(struct x86_emulate_ctxt *ctxt)
2189 {
2190 	int rc;
2191 	unsigned long eip, cs;
2192 	u16 old_cs;
2193 	int cpl = ctxt->ops->cpl(ctxt);
2194 	struct desc_struct old_desc, new_desc;
2195 	const struct x86_emulate_ops *ops = ctxt->ops;
2196 
2197 	if (ctxt->mode == X86EMUL_MODE_PROT64)
2198 		ops->get_segment(ctxt, &old_cs, &old_desc, NULL,
2199 				 VCPU_SREG_CS);
2200 
2201 	rc = emulate_pop(ctxt, &eip, ctxt->op_bytes);
2202 	if (rc != X86EMUL_CONTINUE)
2203 		return rc;
2204 	rc = emulate_pop(ctxt, &cs, ctxt->op_bytes);
2205 	if (rc != X86EMUL_CONTINUE)
2206 		return rc;
2207 	/* Outer-privilege level return is not implemented */
2208 	if (ctxt->mode >= X86EMUL_MODE_PROT16 && (cs & 3) > cpl)
2209 		return X86EMUL_UNHANDLEABLE;
2210 	rc = __load_segment_descriptor(ctxt, (u16)cs, VCPU_SREG_CS, cpl,
2211 				       X86_TRANSFER_RET,
2212 				       &new_desc);
2213 	if (rc != X86EMUL_CONTINUE)
2214 		return rc;
2215 	rc = assign_eip_far(ctxt, eip, &new_desc);
2216 	if (rc != X86EMUL_CONTINUE) {
2217 		WARN_ON(ctxt->mode != X86EMUL_MODE_PROT64);
2218 		ops->set_segment(ctxt, old_cs, &old_desc, 0, VCPU_SREG_CS);
2219 	}
2220 	return rc;
2221 }
2222 
2223 static int em_ret_far_imm(struct x86_emulate_ctxt *ctxt)
2224 {
2225         int rc;
2226 
2227         rc = em_ret_far(ctxt);
2228         if (rc != X86EMUL_CONTINUE)
2229                 return rc;
2230         rsp_increment(ctxt, ctxt->src.val);
2231         return X86EMUL_CONTINUE;
2232 }
2233 
2234 static int em_cmpxchg(struct x86_emulate_ctxt *ctxt)
2235 {
2236 	/* Save real source value, then compare EAX against destination. */
2237 	ctxt->dst.orig_val = ctxt->dst.val;
2238 	ctxt->dst.val = reg_read(ctxt, VCPU_REGS_RAX);
2239 	ctxt->src.orig_val = ctxt->src.val;
2240 	ctxt->src.val = ctxt->dst.orig_val;
2241 	fastop(ctxt, em_cmp);
2242 
2243 	if (ctxt->eflags & X86_EFLAGS_ZF) {
2244 		/* Success: write back to memory; no update of EAX */
2245 		ctxt->src.type = OP_NONE;
2246 		ctxt->dst.val = ctxt->src.orig_val;
2247 	} else {
2248 		/* Failure: write the value we saw to EAX. */
2249 		ctxt->src.type = OP_REG;
2250 		ctxt->src.addr.reg = reg_rmw(ctxt, VCPU_REGS_RAX);
2251 		ctxt->src.val = ctxt->dst.orig_val;
2252 		/* Create write-cycle to dest by writing the same value */
2253 		ctxt->dst.val = ctxt->dst.orig_val;
2254 	}
2255 	return X86EMUL_CONTINUE;
2256 }
2257 
2258 static int em_lseg(struct x86_emulate_ctxt *ctxt)
2259 {
2260 	int seg = ctxt->src2.val;
2261 	unsigned short sel;
2262 	int rc;
2263 
2264 	memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
2265 
2266 	rc = load_segment_descriptor(ctxt, sel, seg);
2267 	if (rc != X86EMUL_CONTINUE)
2268 		return rc;
2269 
2270 	ctxt->dst.val = ctxt->src.val;
2271 	return rc;
2272 }
2273 
2274 static int emulator_has_longmode(struct x86_emulate_ctxt *ctxt)
2275 {
2276 	u32 eax, ebx, ecx, edx;
2277 
2278 	eax = 0x80000001;
2279 	ecx = 0;
2280 	ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
2281 	return edx & bit(X86_FEATURE_LM);
2282 }
2283 
2284 #define GET_SMSTATE(type, smbase, offset)				  \
2285 	({								  \
2286 	 type __val;							  \
2287 	 int r = ctxt->ops->read_phys(ctxt, smbase + offset, &__val,      \
2288 				      sizeof(__val));			  \
2289 	 if (r != X86EMUL_CONTINUE)					  \
2290 		 return X86EMUL_UNHANDLEABLE;				  \
2291 	 __val;								  \
2292 	})
2293 
2294 static void rsm_set_desc_flags(struct desc_struct *desc, u32 flags)
2295 {
2296 	desc->g    = (flags >> 23) & 1;
2297 	desc->d    = (flags >> 22) & 1;
2298 	desc->l    = (flags >> 21) & 1;
2299 	desc->avl  = (flags >> 20) & 1;
2300 	desc->p    = (flags >> 15) & 1;
2301 	desc->dpl  = (flags >> 13) & 3;
2302 	desc->s    = (flags >> 12) & 1;
2303 	desc->type = (flags >>  8) & 15;
2304 }
2305 
2306 static int rsm_load_seg_32(struct x86_emulate_ctxt *ctxt, u64 smbase, int n)
2307 {
2308 	struct desc_struct desc;
2309 	int offset;
2310 	u16 selector;
2311 
2312 	selector = GET_SMSTATE(u32, smbase, 0x7fa8 + n * 4);
2313 
2314 	if (n < 3)
2315 		offset = 0x7f84 + n * 12;
2316 	else
2317 		offset = 0x7f2c + (n - 3) * 12;
2318 
2319 	set_desc_base(&desc,      GET_SMSTATE(u32, smbase, offset + 8));
2320 	set_desc_limit(&desc,     GET_SMSTATE(u32, smbase, offset + 4));
2321 	rsm_set_desc_flags(&desc, GET_SMSTATE(u32, smbase, offset));
2322 	ctxt->ops->set_segment(ctxt, selector, &desc, 0, n);
2323 	return X86EMUL_CONTINUE;
2324 }
2325 
2326 static int rsm_load_seg_64(struct x86_emulate_ctxt *ctxt, u64 smbase, int n)
2327 {
2328 	struct desc_struct desc;
2329 	int offset;
2330 	u16 selector;
2331 	u32 base3;
2332 
2333 	offset = 0x7e00 + n * 16;
2334 
2335 	selector =                GET_SMSTATE(u16, smbase, offset);
2336 	rsm_set_desc_flags(&desc, GET_SMSTATE(u16, smbase, offset + 2) << 8);
2337 	set_desc_limit(&desc,     GET_SMSTATE(u32, smbase, offset + 4));
2338 	set_desc_base(&desc,      GET_SMSTATE(u32, smbase, offset + 8));
2339 	base3 =                   GET_SMSTATE(u32, smbase, offset + 12);
2340 
2341 	ctxt->ops->set_segment(ctxt, selector, &desc, base3, n);
2342 	return X86EMUL_CONTINUE;
2343 }
2344 
2345 static int rsm_enter_protected_mode(struct x86_emulate_ctxt *ctxt,
2346 				     u64 cr0, u64 cr4)
2347 {
2348 	int bad;
2349 
2350 	/*
2351 	 * First enable PAE, long mode needs it before CR0.PG = 1 is set.
2352 	 * Then enable protected mode.	However, PCID cannot be enabled
2353 	 * if EFER.LMA=0, so set it separately.
2354 	 */
2355 	bad = ctxt->ops->set_cr(ctxt, 4, cr4 & ~X86_CR4_PCIDE);
2356 	if (bad)
2357 		return X86EMUL_UNHANDLEABLE;
2358 
2359 	bad = ctxt->ops->set_cr(ctxt, 0, cr0);
2360 	if (bad)
2361 		return X86EMUL_UNHANDLEABLE;
2362 
2363 	if (cr4 & X86_CR4_PCIDE) {
2364 		bad = ctxt->ops->set_cr(ctxt, 4, cr4);
2365 		if (bad)
2366 			return X86EMUL_UNHANDLEABLE;
2367 	}
2368 
2369 	return X86EMUL_CONTINUE;
2370 }
2371 
2372 static int rsm_load_state_32(struct x86_emulate_ctxt *ctxt, u64 smbase)
2373 {
2374 	struct desc_struct desc;
2375 	struct desc_ptr dt;
2376 	u16 selector;
2377 	u32 val, cr0, cr4;
2378 	int i;
2379 
2380 	cr0 =                      GET_SMSTATE(u32, smbase, 0x7ffc);
2381 	ctxt->ops->set_cr(ctxt, 3, GET_SMSTATE(u32, smbase, 0x7ff8));
2382 	ctxt->eflags =             GET_SMSTATE(u32, smbase, 0x7ff4) | X86_EFLAGS_FIXED;
2383 	ctxt->_eip =               GET_SMSTATE(u32, smbase, 0x7ff0);
2384 
2385 	for (i = 0; i < 8; i++)
2386 		*reg_write(ctxt, i) = GET_SMSTATE(u32, smbase, 0x7fd0 + i * 4);
2387 
2388 	val = GET_SMSTATE(u32, smbase, 0x7fcc);
2389 	ctxt->ops->set_dr(ctxt, 6, (val & DR6_VOLATILE) | DR6_FIXED_1);
2390 	val = GET_SMSTATE(u32, smbase, 0x7fc8);
2391 	ctxt->ops->set_dr(ctxt, 7, (val & DR7_VOLATILE) | DR7_FIXED_1);
2392 
2393 	selector =                 GET_SMSTATE(u32, smbase, 0x7fc4);
2394 	set_desc_base(&desc,       GET_SMSTATE(u32, smbase, 0x7f64));
2395 	set_desc_limit(&desc,      GET_SMSTATE(u32, smbase, 0x7f60));
2396 	rsm_set_desc_flags(&desc,  GET_SMSTATE(u32, smbase, 0x7f5c));
2397 	ctxt->ops->set_segment(ctxt, selector, &desc, 0, VCPU_SREG_TR);
2398 
2399 	selector =                 GET_SMSTATE(u32, smbase, 0x7fc0);
2400 	set_desc_base(&desc,       GET_SMSTATE(u32, smbase, 0x7f80));
2401 	set_desc_limit(&desc,      GET_SMSTATE(u32, smbase, 0x7f7c));
2402 	rsm_set_desc_flags(&desc,  GET_SMSTATE(u32, smbase, 0x7f78));
2403 	ctxt->ops->set_segment(ctxt, selector, &desc, 0, VCPU_SREG_LDTR);
2404 
2405 	dt.address =               GET_SMSTATE(u32, smbase, 0x7f74);
2406 	dt.size =                  GET_SMSTATE(u32, smbase, 0x7f70);
2407 	ctxt->ops->set_gdt(ctxt, &dt);
2408 
2409 	dt.address =               GET_SMSTATE(u32, smbase, 0x7f58);
2410 	dt.size =                  GET_SMSTATE(u32, smbase, 0x7f54);
2411 	ctxt->ops->set_idt(ctxt, &dt);
2412 
2413 	for (i = 0; i < 6; i++) {
2414 		int r = rsm_load_seg_32(ctxt, smbase, i);
2415 		if (r != X86EMUL_CONTINUE)
2416 			return r;
2417 	}
2418 
2419 	cr4 = GET_SMSTATE(u32, smbase, 0x7f14);
2420 
2421 	ctxt->ops->set_smbase(ctxt, GET_SMSTATE(u32, smbase, 0x7ef8));
2422 
2423 	return rsm_enter_protected_mode(ctxt, cr0, cr4);
2424 }
2425 
2426 static int rsm_load_state_64(struct x86_emulate_ctxt *ctxt, u64 smbase)
2427 {
2428 	struct desc_struct desc;
2429 	struct desc_ptr dt;
2430 	u64 val, cr0, cr4;
2431 	u32 base3;
2432 	u16 selector;
2433 	int i, r;
2434 
2435 	for (i = 0; i < 16; i++)
2436 		*reg_write(ctxt, i) = GET_SMSTATE(u64, smbase, 0x7ff8 - i * 8);
2437 
2438 	ctxt->_eip   = GET_SMSTATE(u64, smbase, 0x7f78);
2439 	ctxt->eflags = GET_SMSTATE(u32, smbase, 0x7f70) | X86_EFLAGS_FIXED;
2440 
2441 	val = GET_SMSTATE(u32, smbase, 0x7f68);
2442 	ctxt->ops->set_dr(ctxt, 6, (val & DR6_VOLATILE) | DR6_FIXED_1);
2443 	val = GET_SMSTATE(u32, smbase, 0x7f60);
2444 	ctxt->ops->set_dr(ctxt, 7, (val & DR7_VOLATILE) | DR7_FIXED_1);
2445 
2446 	cr0 =                       GET_SMSTATE(u64, smbase, 0x7f58);
2447 	ctxt->ops->set_cr(ctxt, 3,  GET_SMSTATE(u64, smbase, 0x7f50));
2448 	cr4 =                       GET_SMSTATE(u64, smbase, 0x7f48);
2449 	ctxt->ops->set_smbase(ctxt, GET_SMSTATE(u32, smbase, 0x7f00));
2450 	val =                       GET_SMSTATE(u64, smbase, 0x7ed0);
2451 	ctxt->ops->set_msr(ctxt, MSR_EFER, val & ~EFER_LMA);
2452 
2453 	selector =                  GET_SMSTATE(u32, smbase, 0x7e90);
2454 	rsm_set_desc_flags(&desc,   GET_SMSTATE(u32, smbase, 0x7e92) << 8);
2455 	set_desc_limit(&desc,       GET_SMSTATE(u32, smbase, 0x7e94));
2456 	set_desc_base(&desc,        GET_SMSTATE(u32, smbase, 0x7e98));
2457 	base3 =                     GET_SMSTATE(u32, smbase, 0x7e9c);
2458 	ctxt->ops->set_segment(ctxt, selector, &desc, base3, VCPU_SREG_TR);
2459 
2460 	dt.size =                   GET_SMSTATE(u32, smbase, 0x7e84);
2461 	dt.address =                GET_SMSTATE(u64, smbase, 0x7e88);
2462 	ctxt->ops->set_idt(ctxt, &dt);
2463 
2464 	selector =                  GET_SMSTATE(u32, smbase, 0x7e70);
2465 	rsm_set_desc_flags(&desc,   GET_SMSTATE(u32, smbase, 0x7e72) << 8);
2466 	set_desc_limit(&desc,       GET_SMSTATE(u32, smbase, 0x7e74));
2467 	set_desc_base(&desc,        GET_SMSTATE(u32, smbase, 0x7e78));
2468 	base3 =                     GET_SMSTATE(u32, smbase, 0x7e7c);
2469 	ctxt->ops->set_segment(ctxt, selector, &desc, base3, VCPU_SREG_LDTR);
2470 
2471 	dt.size =                   GET_SMSTATE(u32, smbase, 0x7e64);
2472 	dt.address =                GET_SMSTATE(u64, smbase, 0x7e68);
2473 	ctxt->ops->set_gdt(ctxt, &dt);
2474 
2475 	r = rsm_enter_protected_mode(ctxt, cr0, cr4);
2476 	if (r != X86EMUL_CONTINUE)
2477 		return r;
2478 
2479 	for (i = 0; i < 6; i++) {
2480 		r = rsm_load_seg_64(ctxt, smbase, i);
2481 		if (r != X86EMUL_CONTINUE)
2482 			return r;
2483 	}
2484 
2485 	return X86EMUL_CONTINUE;
2486 }
2487 
2488 static int em_rsm(struct x86_emulate_ctxt *ctxt)
2489 {
2490 	unsigned long cr0, cr4, efer;
2491 	u64 smbase;
2492 	int ret;
2493 
2494 	if ((ctxt->emul_flags & X86EMUL_SMM_MASK) == 0)
2495 		return emulate_ud(ctxt);
2496 
2497 	/*
2498 	 * Get back to real mode, to prepare a safe state in which to load
2499 	 * CR0/CR3/CR4/EFER.  It's all a bit more complicated if the vCPU
2500 	 * supports long mode.
2501 	 */
2502 	cr4 = ctxt->ops->get_cr(ctxt, 4);
2503 	if (emulator_has_longmode(ctxt)) {
2504 		struct desc_struct cs_desc;
2505 
2506 		/* Zero CR4.PCIDE before CR0.PG.  */
2507 		if (cr4 & X86_CR4_PCIDE) {
2508 			ctxt->ops->set_cr(ctxt, 4, cr4 & ~X86_CR4_PCIDE);
2509 			cr4 &= ~X86_CR4_PCIDE;
2510 		}
2511 
2512 		/* A 32-bit code segment is required to clear EFER.LMA.  */
2513 		memset(&cs_desc, 0, sizeof(cs_desc));
2514 		cs_desc.type = 0xb;
2515 		cs_desc.s = cs_desc.g = cs_desc.p = 1;
2516 		ctxt->ops->set_segment(ctxt, 0, &cs_desc, 0, VCPU_SREG_CS);
2517 	}
2518 
2519 	/* For the 64-bit case, this will clear EFER.LMA.  */
2520 	cr0 = ctxt->ops->get_cr(ctxt, 0);
2521 	if (cr0 & X86_CR0_PE)
2522 		ctxt->ops->set_cr(ctxt, 0, cr0 & ~(X86_CR0_PG | X86_CR0_PE));
2523 
2524 	/* Now clear CR4.PAE (which must be done before clearing EFER.LME).  */
2525 	if (cr4 & X86_CR4_PAE)
2526 		ctxt->ops->set_cr(ctxt, 4, cr4 & ~X86_CR4_PAE);
2527 
2528 	/* And finally go back to 32-bit mode.  */
2529 	efer = 0;
2530 	ctxt->ops->set_msr(ctxt, MSR_EFER, efer);
2531 
2532 	smbase = ctxt->ops->get_smbase(ctxt);
2533 	if (emulator_has_longmode(ctxt))
2534 		ret = rsm_load_state_64(ctxt, smbase + 0x8000);
2535 	else
2536 		ret = rsm_load_state_32(ctxt, smbase + 0x8000);
2537 
2538 	if (ret != X86EMUL_CONTINUE) {
2539 		/* FIXME: should triple fault */
2540 		return X86EMUL_UNHANDLEABLE;
2541 	}
2542 
2543 	if ((ctxt->emul_flags & X86EMUL_SMM_INSIDE_NMI_MASK) == 0)
2544 		ctxt->ops->set_nmi_mask(ctxt, false);
2545 
2546 	ctxt->emul_flags &= ~X86EMUL_SMM_INSIDE_NMI_MASK;
2547 	ctxt->emul_flags &= ~X86EMUL_SMM_MASK;
2548 	return X86EMUL_CONTINUE;
2549 }
2550 
2551 static void
2552 setup_syscalls_segments(struct x86_emulate_ctxt *ctxt,
2553 			struct desc_struct *cs, struct desc_struct *ss)
2554 {
2555 	cs->l = 0;		/* will be adjusted later */
2556 	set_desc_base(cs, 0);	/* flat segment */
2557 	cs->g = 1;		/* 4kb granularity */
2558 	set_desc_limit(cs, 0xfffff);	/* 4GB limit */
2559 	cs->type = 0x0b;	/* Read, Execute, Accessed */
2560 	cs->s = 1;
2561 	cs->dpl = 0;		/* will be adjusted later */
2562 	cs->p = 1;
2563 	cs->d = 1;
2564 	cs->avl = 0;
2565 
2566 	set_desc_base(ss, 0);	/* flat segment */
2567 	set_desc_limit(ss, 0xfffff);	/* 4GB limit */
2568 	ss->g = 1;		/* 4kb granularity */
2569 	ss->s = 1;
2570 	ss->type = 0x03;	/* Read/Write, Accessed */
2571 	ss->d = 1;		/* 32bit stack segment */
2572 	ss->dpl = 0;
2573 	ss->p = 1;
2574 	ss->l = 0;
2575 	ss->avl = 0;
2576 }
2577 
2578 static bool vendor_intel(struct x86_emulate_ctxt *ctxt)
2579 {
2580 	u32 eax, ebx, ecx, edx;
2581 
2582 	eax = ecx = 0;
2583 	ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
2584 	return ebx == X86EMUL_CPUID_VENDOR_GenuineIntel_ebx
2585 		&& ecx == X86EMUL_CPUID_VENDOR_GenuineIntel_ecx
2586 		&& edx == X86EMUL_CPUID_VENDOR_GenuineIntel_edx;
2587 }
2588 
2589 static bool em_syscall_is_enabled(struct x86_emulate_ctxt *ctxt)
2590 {
2591 	const struct x86_emulate_ops *ops = ctxt->ops;
2592 	u32 eax, ebx, ecx, edx;
2593 
2594 	/*
2595 	 * syscall should always be enabled in longmode - so only become
2596 	 * vendor specific (cpuid) if other modes are active...
2597 	 */
2598 	if (ctxt->mode == X86EMUL_MODE_PROT64)
2599 		return true;
2600 
2601 	eax = 0x00000000;
2602 	ecx = 0x00000000;
2603 	ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
2604 	/*
2605 	 * Intel ("GenuineIntel")
2606 	 * remark: Intel CPUs only support "syscall" in 64bit
2607 	 * longmode. Also an 64bit guest with a
2608 	 * 32bit compat-app running will #UD !! While this
2609 	 * behaviour can be fixed (by emulating) into AMD
2610 	 * response - CPUs of AMD can't behave like Intel.
2611 	 */
2612 	if (ebx == X86EMUL_CPUID_VENDOR_GenuineIntel_ebx &&
2613 	    ecx == X86EMUL_CPUID_VENDOR_GenuineIntel_ecx &&
2614 	    edx == X86EMUL_CPUID_VENDOR_GenuineIntel_edx)
2615 		return false;
2616 
2617 	/* AMD ("AuthenticAMD") */
2618 	if (ebx == X86EMUL_CPUID_VENDOR_AuthenticAMD_ebx &&
2619 	    ecx == X86EMUL_CPUID_VENDOR_AuthenticAMD_ecx &&
2620 	    edx == X86EMUL_CPUID_VENDOR_AuthenticAMD_edx)
2621 		return true;
2622 
2623 	/* AMD ("AMDisbetter!") */
2624 	if (ebx == X86EMUL_CPUID_VENDOR_AMDisbetterI_ebx &&
2625 	    ecx == X86EMUL_CPUID_VENDOR_AMDisbetterI_ecx &&
2626 	    edx == X86EMUL_CPUID_VENDOR_AMDisbetterI_edx)
2627 		return true;
2628 
2629 	/* default: (not Intel, not AMD), apply Intel's stricter rules... */
2630 	return false;
2631 }
2632 
2633 static int em_syscall(struct x86_emulate_ctxt *ctxt)
2634 {
2635 	const struct x86_emulate_ops *ops = ctxt->ops;
2636 	struct desc_struct cs, ss;
2637 	u64 msr_data;
2638 	u16 cs_sel, ss_sel;
2639 	u64 efer = 0;
2640 
2641 	/* syscall is not available in real mode */
2642 	if (ctxt->mode == X86EMUL_MODE_REAL ||
2643 	    ctxt->mode == X86EMUL_MODE_VM86)
2644 		return emulate_ud(ctxt);
2645 
2646 	if (!(em_syscall_is_enabled(ctxt)))
2647 		return emulate_ud(ctxt);
2648 
2649 	ops->get_msr(ctxt, MSR_EFER, &efer);
2650 	setup_syscalls_segments(ctxt, &cs, &ss);
2651 
2652 	if (!(efer & EFER_SCE))
2653 		return emulate_ud(ctxt);
2654 
2655 	ops->get_msr(ctxt, MSR_STAR, &msr_data);
2656 	msr_data >>= 32;
2657 	cs_sel = (u16)(msr_data & 0xfffc);
2658 	ss_sel = (u16)(msr_data + 8);
2659 
2660 	if (efer & EFER_LMA) {
2661 		cs.d = 0;
2662 		cs.l = 1;
2663 	}
2664 	ops->set_segment(ctxt, cs_sel, &cs, 0, VCPU_SREG_CS);
2665 	ops->set_segment(ctxt, ss_sel, &ss, 0, VCPU_SREG_SS);
2666 
2667 	*reg_write(ctxt, VCPU_REGS_RCX) = ctxt->_eip;
2668 	if (efer & EFER_LMA) {
2669 #ifdef CONFIG_X86_64
2670 		*reg_write(ctxt, VCPU_REGS_R11) = ctxt->eflags;
2671 
2672 		ops->get_msr(ctxt,
2673 			     ctxt->mode == X86EMUL_MODE_PROT64 ?
2674 			     MSR_LSTAR : MSR_CSTAR, &msr_data);
2675 		ctxt->_eip = msr_data;
2676 
2677 		ops->get_msr(ctxt, MSR_SYSCALL_MASK, &msr_data);
2678 		ctxt->eflags &= ~msr_data;
2679 		ctxt->eflags |= X86_EFLAGS_FIXED;
2680 #endif
2681 	} else {
2682 		/* legacy mode */
2683 		ops->get_msr(ctxt, MSR_STAR, &msr_data);
2684 		ctxt->_eip = (u32)msr_data;
2685 
2686 		ctxt->eflags &= ~(X86_EFLAGS_VM | X86_EFLAGS_IF);
2687 	}
2688 
2689 	return X86EMUL_CONTINUE;
2690 }
2691 
2692 static int em_sysenter(struct x86_emulate_ctxt *ctxt)
2693 {
2694 	const struct x86_emulate_ops *ops = ctxt->ops;
2695 	struct desc_struct cs, ss;
2696 	u64 msr_data;
2697 	u16 cs_sel, ss_sel;
2698 	u64 efer = 0;
2699 
2700 	ops->get_msr(ctxt, MSR_EFER, &efer);
2701 	/* inject #GP if in real mode */
2702 	if (ctxt->mode == X86EMUL_MODE_REAL)
2703 		return emulate_gp(ctxt, 0);
2704 
2705 	/*
2706 	 * Not recognized on AMD in compat mode (but is recognized in legacy
2707 	 * mode).
2708 	 */
2709 	if ((ctxt->mode != X86EMUL_MODE_PROT64) && (efer & EFER_LMA)
2710 	    && !vendor_intel(ctxt))
2711 		return emulate_ud(ctxt);
2712 
2713 	/* sysenter/sysexit have not been tested in 64bit mode. */
2714 	if (ctxt->mode == X86EMUL_MODE_PROT64)
2715 		return X86EMUL_UNHANDLEABLE;
2716 
2717 	setup_syscalls_segments(ctxt, &cs, &ss);
2718 
2719 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_CS, &msr_data);
2720 	if ((msr_data & 0xfffc) == 0x0)
2721 		return emulate_gp(ctxt, 0);
2722 
2723 	ctxt->eflags &= ~(X86_EFLAGS_VM | X86_EFLAGS_IF);
2724 	cs_sel = (u16)msr_data & ~SEGMENT_RPL_MASK;
2725 	ss_sel = cs_sel + 8;
2726 	if (efer & EFER_LMA) {
2727 		cs.d = 0;
2728 		cs.l = 1;
2729 	}
2730 
2731 	ops->set_segment(ctxt, cs_sel, &cs, 0, VCPU_SREG_CS);
2732 	ops->set_segment(ctxt, ss_sel, &ss, 0, VCPU_SREG_SS);
2733 
2734 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_EIP, &msr_data);
2735 	ctxt->_eip = (efer & EFER_LMA) ? msr_data : (u32)msr_data;
2736 
2737 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_ESP, &msr_data);
2738 	*reg_write(ctxt, VCPU_REGS_RSP) = (efer & EFER_LMA) ? msr_data :
2739 							      (u32)msr_data;
2740 
2741 	return X86EMUL_CONTINUE;
2742 }
2743 
2744 static int em_sysexit(struct x86_emulate_ctxt *ctxt)
2745 {
2746 	const struct x86_emulate_ops *ops = ctxt->ops;
2747 	struct desc_struct cs, ss;
2748 	u64 msr_data, rcx, rdx;
2749 	int usermode;
2750 	u16 cs_sel = 0, ss_sel = 0;
2751 
2752 	/* inject #GP if in real mode or Virtual 8086 mode */
2753 	if (ctxt->mode == X86EMUL_MODE_REAL ||
2754 	    ctxt->mode == X86EMUL_MODE_VM86)
2755 		return emulate_gp(ctxt, 0);
2756 
2757 	setup_syscalls_segments(ctxt, &cs, &ss);
2758 
2759 	if ((ctxt->rex_prefix & 0x8) != 0x0)
2760 		usermode = X86EMUL_MODE_PROT64;
2761 	else
2762 		usermode = X86EMUL_MODE_PROT32;
2763 
2764 	rcx = reg_read(ctxt, VCPU_REGS_RCX);
2765 	rdx = reg_read(ctxt, VCPU_REGS_RDX);
2766 
2767 	cs.dpl = 3;
2768 	ss.dpl = 3;
2769 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_CS, &msr_data);
2770 	switch (usermode) {
2771 	case X86EMUL_MODE_PROT32:
2772 		cs_sel = (u16)(msr_data + 16);
2773 		if ((msr_data & 0xfffc) == 0x0)
2774 			return emulate_gp(ctxt, 0);
2775 		ss_sel = (u16)(msr_data + 24);
2776 		rcx = (u32)rcx;
2777 		rdx = (u32)rdx;
2778 		break;
2779 	case X86EMUL_MODE_PROT64:
2780 		cs_sel = (u16)(msr_data + 32);
2781 		if (msr_data == 0x0)
2782 			return emulate_gp(ctxt, 0);
2783 		ss_sel = cs_sel + 8;
2784 		cs.d = 0;
2785 		cs.l = 1;
2786 		if (is_noncanonical_address(rcx) ||
2787 		    is_noncanonical_address(rdx))
2788 			return emulate_gp(ctxt, 0);
2789 		break;
2790 	}
2791 	cs_sel |= SEGMENT_RPL_MASK;
2792 	ss_sel |= SEGMENT_RPL_MASK;
2793 
2794 	ops->set_segment(ctxt, cs_sel, &cs, 0, VCPU_SREG_CS);
2795 	ops->set_segment(ctxt, ss_sel, &ss, 0, VCPU_SREG_SS);
2796 
2797 	ctxt->_eip = rdx;
2798 	*reg_write(ctxt, VCPU_REGS_RSP) = rcx;
2799 
2800 	return X86EMUL_CONTINUE;
2801 }
2802 
2803 static bool emulator_bad_iopl(struct x86_emulate_ctxt *ctxt)
2804 {
2805 	int iopl;
2806 	if (ctxt->mode == X86EMUL_MODE_REAL)
2807 		return false;
2808 	if (ctxt->mode == X86EMUL_MODE_VM86)
2809 		return true;
2810 	iopl = (ctxt->eflags & X86_EFLAGS_IOPL) >> X86_EFLAGS_IOPL_BIT;
2811 	return ctxt->ops->cpl(ctxt) > iopl;
2812 }
2813 
2814 static bool emulator_io_port_access_allowed(struct x86_emulate_ctxt *ctxt,
2815 					    u16 port, u16 len)
2816 {
2817 	const struct x86_emulate_ops *ops = ctxt->ops;
2818 	struct desc_struct tr_seg;
2819 	u32 base3;
2820 	int r;
2821 	u16 tr, io_bitmap_ptr, perm, bit_idx = port & 0x7;
2822 	unsigned mask = (1 << len) - 1;
2823 	unsigned long base;
2824 
2825 	ops->get_segment(ctxt, &tr, &tr_seg, &base3, VCPU_SREG_TR);
2826 	if (!tr_seg.p)
2827 		return false;
2828 	if (desc_limit_scaled(&tr_seg) < 103)
2829 		return false;
2830 	base = get_desc_base(&tr_seg);
2831 #ifdef CONFIG_X86_64
2832 	base |= ((u64)base3) << 32;
2833 #endif
2834 	r = ops->read_std(ctxt, base + 102, &io_bitmap_ptr, 2, NULL);
2835 	if (r != X86EMUL_CONTINUE)
2836 		return false;
2837 	if (io_bitmap_ptr + port/8 > desc_limit_scaled(&tr_seg))
2838 		return false;
2839 	r = ops->read_std(ctxt, base + io_bitmap_ptr + port/8, &perm, 2, NULL);
2840 	if (r != X86EMUL_CONTINUE)
2841 		return false;
2842 	if ((perm >> bit_idx) & mask)
2843 		return false;
2844 	return true;
2845 }
2846 
2847 static bool emulator_io_permited(struct x86_emulate_ctxt *ctxt,
2848 				 u16 port, u16 len)
2849 {
2850 	if (ctxt->perm_ok)
2851 		return true;
2852 
2853 	if (emulator_bad_iopl(ctxt))
2854 		if (!emulator_io_port_access_allowed(ctxt, port, len))
2855 			return false;
2856 
2857 	ctxt->perm_ok = true;
2858 
2859 	return true;
2860 }
2861 
2862 static void string_registers_quirk(struct x86_emulate_ctxt *ctxt)
2863 {
2864 	/*
2865 	 * Intel CPUs mask the counter and pointers in quite strange
2866 	 * manner when ECX is zero due to REP-string optimizations.
2867 	 */
2868 #ifdef CONFIG_X86_64
2869 	if (ctxt->ad_bytes != 4 || !vendor_intel(ctxt))
2870 		return;
2871 
2872 	*reg_write(ctxt, VCPU_REGS_RCX) = 0;
2873 
2874 	switch (ctxt->b) {
2875 	case 0xa4:	/* movsb */
2876 	case 0xa5:	/* movsd/w */
2877 		*reg_rmw(ctxt, VCPU_REGS_RSI) &= (u32)-1;
2878 		/* fall through */
2879 	case 0xaa:	/* stosb */
2880 	case 0xab:	/* stosd/w */
2881 		*reg_rmw(ctxt, VCPU_REGS_RDI) &= (u32)-1;
2882 	}
2883 #endif
2884 }
2885 
2886 static void save_state_to_tss16(struct x86_emulate_ctxt *ctxt,
2887 				struct tss_segment_16 *tss)
2888 {
2889 	tss->ip = ctxt->_eip;
2890 	tss->flag = ctxt->eflags;
2891 	tss->ax = reg_read(ctxt, VCPU_REGS_RAX);
2892 	tss->cx = reg_read(ctxt, VCPU_REGS_RCX);
2893 	tss->dx = reg_read(ctxt, VCPU_REGS_RDX);
2894 	tss->bx = reg_read(ctxt, VCPU_REGS_RBX);
2895 	tss->sp = reg_read(ctxt, VCPU_REGS_RSP);
2896 	tss->bp = reg_read(ctxt, VCPU_REGS_RBP);
2897 	tss->si = reg_read(ctxt, VCPU_REGS_RSI);
2898 	tss->di = reg_read(ctxt, VCPU_REGS_RDI);
2899 
2900 	tss->es = get_segment_selector(ctxt, VCPU_SREG_ES);
2901 	tss->cs = get_segment_selector(ctxt, VCPU_SREG_CS);
2902 	tss->ss = get_segment_selector(ctxt, VCPU_SREG_SS);
2903 	tss->ds = get_segment_selector(ctxt, VCPU_SREG_DS);
2904 	tss->ldt = get_segment_selector(ctxt, VCPU_SREG_LDTR);
2905 }
2906 
2907 static int load_state_from_tss16(struct x86_emulate_ctxt *ctxt,
2908 				 struct tss_segment_16 *tss)
2909 {
2910 	int ret;
2911 	u8 cpl;
2912 
2913 	ctxt->_eip = tss->ip;
2914 	ctxt->eflags = tss->flag | 2;
2915 	*reg_write(ctxt, VCPU_REGS_RAX) = tss->ax;
2916 	*reg_write(ctxt, VCPU_REGS_RCX) = tss->cx;
2917 	*reg_write(ctxt, VCPU_REGS_RDX) = tss->dx;
2918 	*reg_write(ctxt, VCPU_REGS_RBX) = tss->bx;
2919 	*reg_write(ctxt, VCPU_REGS_RSP) = tss->sp;
2920 	*reg_write(ctxt, VCPU_REGS_RBP) = tss->bp;
2921 	*reg_write(ctxt, VCPU_REGS_RSI) = tss->si;
2922 	*reg_write(ctxt, VCPU_REGS_RDI) = tss->di;
2923 
2924 	/*
2925 	 * SDM says that segment selectors are loaded before segment
2926 	 * descriptors
2927 	 */
2928 	set_segment_selector(ctxt, tss->ldt, VCPU_SREG_LDTR);
2929 	set_segment_selector(ctxt, tss->es, VCPU_SREG_ES);
2930 	set_segment_selector(ctxt, tss->cs, VCPU_SREG_CS);
2931 	set_segment_selector(ctxt, tss->ss, VCPU_SREG_SS);
2932 	set_segment_selector(ctxt, tss->ds, VCPU_SREG_DS);
2933 
2934 	cpl = tss->cs & 3;
2935 
2936 	/*
2937 	 * Now load segment descriptors. If fault happens at this stage
2938 	 * it is handled in a context of new task
2939 	 */
2940 	ret = __load_segment_descriptor(ctxt, tss->ldt, VCPU_SREG_LDTR, cpl,
2941 					X86_TRANSFER_TASK_SWITCH, NULL);
2942 	if (ret != X86EMUL_CONTINUE)
2943 		return ret;
2944 	ret = __load_segment_descriptor(ctxt, tss->es, VCPU_SREG_ES, cpl,
2945 					X86_TRANSFER_TASK_SWITCH, NULL);
2946 	if (ret != X86EMUL_CONTINUE)
2947 		return ret;
2948 	ret = __load_segment_descriptor(ctxt, tss->cs, VCPU_SREG_CS, cpl,
2949 					X86_TRANSFER_TASK_SWITCH, NULL);
2950 	if (ret != X86EMUL_CONTINUE)
2951 		return ret;
2952 	ret = __load_segment_descriptor(ctxt, tss->ss, VCPU_SREG_SS, cpl,
2953 					X86_TRANSFER_TASK_SWITCH, NULL);
2954 	if (ret != X86EMUL_CONTINUE)
2955 		return ret;
2956 	ret = __load_segment_descriptor(ctxt, tss->ds, VCPU_SREG_DS, cpl,
2957 					X86_TRANSFER_TASK_SWITCH, NULL);
2958 	if (ret != X86EMUL_CONTINUE)
2959 		return ret;
2960 
2961 	return X86EMUL_CONTINUE;
2962 }
2963 
2964 static int task_switch_16(struct x86_emulate_ctxt *ctxt,
2965 			  u16 tss_selector, u16 old_tss_sel,
2966 			  ulong old_tss_base, struct desc_struct *new_desc)
2967 {
2968 	const struct x86_emulate_ops *ops = ctxt->ops;
2969 	struct tss_segment_16 tss_seg;
2970 	int ret;
2971 	u32 new_tss_base = get_desc_base(new_desc);
2972 
2973 	ret = ops->read_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2974 			    &ctxt->exception);
2975 	if (ret != X86EMUL_CONTINUE)
2976 		return ret;
2977 
2978 	save_state_to_tss16(ctxt, &tss_seg);
2979 
2980 	ret = ops->write_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2981 			     &ctxt->exception);
2982 	if (ret != X86EMUL_CONTINUE)
2983 		return ret;
2984 
2985 	ret = ops->read_std(ctxt, new_tss_base, &tss_seg, sizeof tss_seg,
2986 			    &ctxt->exception);
2987 	if (ret != X86EMUL_CONTINUE)
2988 		return ret;
2989 
2990 	if (old_tss_sel != 0xffff) {
2991 		tss_seg.prev_task_link = old_tss_sel;
2992 
2993 		ret = ops->write_std(ctxt, new_tss_base,
2994 				     &tss_seg.prev_task_link,
2995 				     sizeof tss_seg.prev_task_link,
2996 				     &ctxt->exception);
2997 		if (ret != X86EMUL_CONTINUE)
2998 			return ret;
2999 	}
3000 
3001 	return load_state_from_tss16(ctxt, &tss_seg);
3002 }
3003 
3004 static void save_state_to_tss32(struct x86_emulate_ctxt *ctxt,
3005 				struct tss_segment_32 *tss)
3006 {
3007 	/* CR3 and ldt selector are not saved intentionally */
3008 	tss->eip = ctxt->_eip;
3009 	tss->eflags = ctxt->eflags;
3010 	tss->eax = reg_read(ctxt, VCPU_REGS_RAX);
3011 	tss->ecx = reg_read(ctxt, VCPU_REGS_RCX);
3012 	tss->edx = reg_read(ctxt, VCPU_REGS_RDX);
3013 	tss->ebx = reg_read(ctxt, VCPU_REGS_RBX);
3014 	tss->esp = reg_read(ctxt, VCPU_REGS_RSP);
3015 	tss->ebp = reg_read(ctxt, VCPU_REGS_RBP);
3016 	tss->esi = reg_read(ctxt, VCPU_REGS_RSI);
3017 	tss->edi = reg_read(ctxt, VCPU_REGS_RDI);
3018 
3019 	tss->es = get_segment_selector(ctxt, VCPU_SREG_ES);
3020 	tss->cs = get_segment_selector(ctxt, VCPU_SREG_CS);
3021 	tss->ss = get_segment_selector(ctxt, VCPU_SREG_SS);
3022 	tss->ds = get_segment_selector(ctxt, VCPU_SREG_DS);
3023 	tss->fs = get_segment_selector(ctxt, VCPU_SREG_FS);
3024 	tss->gs = get_segment_selector(ctxt, VCPU_SREG_GS);
3025 }
3026 
3027 static int load_state_from_tss32(struct x86_emulate_ctxt *ctxt,
3028 				 struct tss_segment_32 *tss)
3029 {
3030 	int ret;
3031 	u8 cpl;
3032 
3033 	if (ctxt->ops->set_cr(ctxt, 3, tss->cr3))
3034 		return emulate_gp(ctxt, 0);
3035 	ctxt->_eip = tss->eip;
3036 	ctxt->eflags = tss->eflags | 2;
3037 
3038 	/* General purpose registers */
3039 	*reg_write(ctxt, VCPU_REGS_RAX) = tss->eax;
3040 	*reg_write(ctxt, VCPU_REGS_RCX) = tss->ecx;
3041 	*reg_write(ctxt, VCPU_REGS_RDX) = tss->edx;
3042 	*reg_write(ctxt, VCPU_REGS_RBX) = tss->ebx;
3043 	*reg_write(ctxt, VCPU_REGS_RSP) = tss->esp;
3044 	*reg_write(ctxt, VCPU_REGS_RBP) = tss->ebp;
3045 	*reg_write(ctxt, VCPU_REGS_RSI) = tss->esi;
3046 	*reg_write(ctxt, VCPU_REGS_RDI) = tss->edi;
3047 
3048 	/*
3049 	 * SDM says that segment selectors are loaded before segment
3050 	 * descriptors.  This is important because CPL checks will
3051 	 * use CS.RPL.
3052 	 */
3053 	set_segment_selector(ctxt, tss->ldt_selector, VCPU_SREG_LDTR);
3054 	set_segment_selector(ctxt, tss->es, VCPU_SREG_ES);
3055 	set_segment_selector(ctxt, tss->cs, VCPU_SREG_CS);
3056 	set_segment_selector(ctxt, tss->ss, VCPU_SREG_SS);
3057 	set_segment_selector(ctxt, tss->ds, VCPU_SREG_DS);
3058 	set_segment_selector(ctxt, tss->fs, VCPU_SREG_FS);
3059 	set_segment_selector(ctxt, tss->gs, VCPU_SREG_GS);
3060 
3061 	/*
3062 	 * If we're switching between Protected Mode and VM86, we need to make
3063 	 * sure to update the mode before loading the segment descriptors so
3064 	 * that the selectors are interpreted correctly.
3065 	 */
3066 	if (ctxt->eflags & X86_EFLAGS_VM) {
3067 		ctxt->mode = X86EMUL_MODE_VM86;
3068 		cpl = 3;
3069 	} else {
3070 		ctxt->mode = X86EMUL_MODE_PROT32;
3071 		cpl = tss->cs & 3;
3072 	}
3073 
3074 	/*
3075 	 * Now load segment descriptors. If fault happenes at this stage
3076 	 * it is handled in a context of new task
3077 	 */
3078 	ret = __load_segment_descriptor(ctxt, tss->ldt_selector, VCPU_SREG_LDTR,
3079 					cpl, X86_TRANSFER_TASK_SWITCH, NULL);
3080 	if (ret != X86EMUL_CONTINUE)
3081 		return ret;
3082 	ret = __load_segment_descriptor(ctxt, tss->es, VCPU_SREG_ES, cpl,
3083 					X86_TRANSFER_TASK_SWITCH, NULL);
3084 	if (ret != X86EMUL_CONTINUE)
3085 		return ret;
3086 	ret = __load_segment_descriptor(ctxt, tss->cs, VCPU_SREG_CS, cpl,
3087 					X86_TRANSFER_TASK_SWITCH, NULL);
3088 	if (ret != X86EMUL_CONTINUE)
3089 		return ret;
3090 	ret = __load_segment_descriptor(ctxt, tss->ss, VCPU_SREG_SS, cpl,
3091 					X86_TRANSFER_TASK_SWITCH, NULL);
3092 	if (ret != X86EMUL_CONTINUE)
3093 		return ret;
3094 	ret = __load_segment_descriptor(ctxt, tss->ds, VCPU_SREG_DS, cpl,
3095 					X86_TRANSFER_TASK_SWITCH, NULL);
3096 	if (ret != X86EMUL_CONTINUE)
3097 		return ret;
3098 	ret = __load_segment_descriptor(ctxt, tss->fs, VCPU_SREG_FS, cpl,
3099 					X86_TRANSFER_TASK_SWITCH, NULL);
3100 	if (ret != X86EMUL_CONTINUE)
3101 		return ret;
3102 	ret = __load_segment_descriptor(ctxt, tss->gs, VCPU_SREG_GS, cpl,
3103 					X86_TRANSFER_TASK_SWITCH, NULL);
3104 
3105 	return ret;
3106 }
3107 
3108 static int task_switch_32(struct x86_emulate_ctxt *ctxt,
3109 			  u16 tss_selector, u16 old_tss_sel,
3110 			  ulong old_tss_base, struct desc_struct *new_desc)
3111 {
3112 	const struct x86_emulate_ops *ops = ctxt->ops;
3113 	struct tss_segment_32 tss_seg;
3114 	int ret;
3115 	u32 new_tss_base = get_desc_base(new_desc);
3116 	u32 eip_offset = offsetof(struct tss_segment_32, eip);
3117 	u32 ldt_sel_offset = offsetof(struct tss_segment_32, ldt_selector);
3118 
3119 	ret = ops->read_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
3120 			    &ctxt->exception);
3121 	if (ret != X86EMUL_CONTINUE)
3122 		return ret;
3123 
3124 	save_state_to_tss32(ctxt, &tss_seg);
3125 
3126 	/* Only GP registers and segment selectors are saved */
3127 	ret = ops->write_std(ctxt, old_tss_base + eip_offset, &tss_seg.eip,
3128 			     ldt_sel_offset - eip_offset, &ctxt->exception);
3129 	if (ret != X86EMUL_CONTINUE)
3130 		return ret;
3131 
3132 	ret = ops->read_std(ctxt, new_tss_base, &tss_seg, sizeof tss_seg,
3133 			    &ctxt->exception);
3134 	if (ret != X86EMUL_CONTINUE)
3135 		return ret;
3136 
3137 	if (old_tss_sel != 0xffff) {
3138 		tss_seg.prev_task_link = old_tss_sel;
3139 
3140 		ret = ops->write_std(ctxt, new_tss_base,
3141 				     &tss_seg.prev_task_link,
3142 				     sizeof tss_seg.prev_task_link,
3143 				     &ctxt->exception);
3144 		if (ret != X86EMUL_CONTINUE)
3145 			return ret;
3146 	}
3147 
3148 	return load_state_from_tss32(ctxt, &tss_seg);
3149 }
3150 
3151 static int emulator_do_task_switch(struct x86_emulate_ctxt *ctxt,
3152 				   u16 tss_selector, int idt_index, int reason,
3153 				   bool has_error_code, u32 error_code)
3154 {
3155 	const struct x86_emulate_ops *ops = ctxt->ops;
3156 	struct desc_struct curr_tss_desc, next_tss_desc;
3157 	int ret;
3158 	u16 old_tss_sel = get_segment_selector(ctxt, VCPU_SREG_TR);
3159 	ulong old_tss_base =
3160 		ops->get_cached_segment_base(ctxt, VCPU_SREG_TR);
3161 	u32 desc_limit;
3162 	ulong desc_addr, dr7;
3163 
3164 	/* FIXME: old_tss_base == ~0 ? */
3165 
3166 	ret = read_segment_descriptor(ctxt, tss_selector, &next_tss_desc, &desc_addr);
3167 	if (ret != X86EMUL_CONTINUE)
3168 		return ret;
3169 	ret = read_segment_descriptor(ctxt, old_tss_sel, &curr_tss_desc, &desc_addr);
3170 	if (ret != X86EMUL_CONTINUE)
3171 		return ret;
3172 
3173 	/* FIXME: check that next_tss_desc is tss */
3174 
3175 	/*
3176 	 * Check privileges. The three cases are task switch caused by...
3177 	 *
3178 	 * 1. jmp/call/int to task gate: Check against DPL of the task gate
3179 	 * 2. Exception/IRQ/iret: No check is performed
3180 	 * 3. jmp/call to TSS/task-gate: No check is performed since the
3181 	 *    hardware checks it before exiting.
3182 	 */
3183 	if (reason == TASK_SWITCH_GATE) {
3184 		if (idt_index != -1) {
3185 			/* Software interrupts */
3186 			struct desc_struct task_gate_desc;
3187 			int dpl;
3188 
3189 			ret = read_interrupt_descriptor(ctxt, idt_index,
3190 							&task_gate_desc);
3191 			if (ret != X86EMUL_CONTINUE)
3192 				return ret;
3193 
3194 			dpl = task_gate_desc.dpl;
3195 			if ((tss_selector & 3) > dpl || ops->cpl(ctxt) > dpl)
3196 				return emulate_gp(ctxt, (idt_index << 3) | 0x2);
3197 		}
3198 	}
3199 
3200 	desc_limit = desc_limit_scaled(&next_tss_desc);
3201 	if (!next_tss_desc.p ||
3202 	    ((desc_limit < 0x67 && (next_tss_desc.type & 8)) ||
3203 	     desc_limit < 0x2b)) {
3204 		return emulate_ts(ctxt, tss_selector & 0xfffc);
3205 	}
3206 
3207 	if (reason == TASK_SWITCH_IRET || reason == TASK_SWITCH_JMP) {
3208 		curr_tss_desc.type &= ~(1 << 1); /* clear busy flag */
3209 		write_segment_descriptor(ctxt, old_tss_sel, &curr_tss_desc);
3210 	}
3211 
3212 	if (reason == TASK_SWITCH_IRET)
3213 		ctxt->eflags = ctxt->eflags & ~X86_EFLAGS_NT;
3214 
3215 	/* set back link to prev task only if NT bit is set in eflags
3216 	   note that old_tss_sel is not used after this point */
3217 	if (reason != TASK_SWITCH_CALL && reason != TASK_SWITCH_GATE)
3218 		old_tss_sel = 0xffff;
3219 
3220 	if (next_tss_desc.type & 8)
3221 		ret = task_switch_32(ctxt, tss_selector, old_tss_sel,
3222 				     old_tss_base, &next_tss_desc);
3223 	else
3224 		ret = task_switch_16(ctxt, tss_selector, old_tss_sel,
3225 				     old_tss_base, &next_tss_desc);
3226 	if (ret != X86EMUL_CONTINUE)
3227 		return ret;
3228 
3229 	if (reason == TASK_SWITCH_CALL || reason == TASK_SWITCH_GATE)
3230 		ctxt->eflags = ctxt->eflags | X86_EFLAGS_NT;
3231 
3232 	if (reason != TASK_SWITCH_IRET) {
3233 		next_tss_desc.type |= (1 << 1); /* set busy flag */
3234 		write_segment_descriptor(ctxt, tss_selector, &next_tss_desc);
3235 	}
3236 
3237 	ops->set_cr(ctxt, 0,  ops->get_cr(ctxt, 0) | X86_CR0_TS);
3238 	ops->set_segment(ctxt, tss_selector, &next_tss_desc, 0, VCPU_SREG_TR);
3239 
3240 	if (has_error_code) {
3241 		ctxt->op_bytes = ctxt->ad_bytes = (next_tss_desc.type & 8) ? 4 : 2;
3242 		ctxt->lock_prefix = 0;
3243 		ctxt->src.val = (unsigned long) error_code;
3244 		ret = em_push(ctxt);
3245 	}
3246 
3247 	ops->get_dr(ctxt, 7, &dr7);
3248 	ops->set_dr(ctxt, 7, dr7 & ~(DR_LOCAL_ENABLE_MASK | DR_LOCAL_SLOWDOWN));
3249 
3250 	return ret;
3251 }
3252 
3253 int emulator_task_switch(struct x86_emulate_ctxt *ctxt,
3254 			 u16 tss_selector, int idt_index, int reason,
3255 			 bool has_error_code, u32 error_code)
3256 {
3257 	int rc;
3258 
3259 	invalidate_registers(ctxt);
3260 	ctxt->_eip = ctxt->eip;
3261 	ctxt->dst.type = OP_NONE;
3262 
3263 	rc = emulator_do_task_switch(ctxt, tss_selector, idt_index, reason,
3264 				     has_error_code, error_code);
3265 
3266 	if (rc == X86EMUL_CONTINUE) {
3267 		ctxt->eip = ctxt->_eip;
3268 		writeback_registers(ctxt);
3269 	}
3270 
3271 	return (rc == X86EMUL_UNHANDLEABLE) ? EMULATION_FAILED : EMULATION_OK;
3272 }
3273 
3274 static void string_addr_inc(struct x86_emulate_ctxt *ctxt, int reg,
3275 		struct operand *op)
3276 {
3277 	int df = (ctxt->eflags & X86_EFLAGS_DF) ? -op->count : op->count;
3278 
3279 	register_address_increment(ctxt, reg, df * op->bytes);
3280 	op->addr.mem.ea = register_address(ctxt, reg);
3281 }
3282 
3283 static int em_das(struct x86_emulate_ctxt *ctxt)
3284 {
3285 	u8 al, old_al;
3286 	bool af, cf, old_cf;
3287 
3288 	cf = ctxt->eflags & X86_EFLAGS_CF;
3289 	al = ctxt->dst.val;
3290 
3291 	old_al = al;
3292 	old_cf = cf;
3293 	cf = false;
3294 	af = ctxt->eflags & X86_EFLAGS_AF;
3295 	if ((al & 0x0f) > 9 || af) {
3296 		al -= 6;
3297 		cf = old_cf | (al >= 250);
3298 		af = true;
3299 	} else {
3300 		af = false;
3301 	}
3302 	if (old_al > 0x99 || old_cf) {
3303 		al -= 0x60;
3304 		cf = true;
3305 	}
3306 
3307 	ctxt->dst.val = al;
3308 	/* Set PF, ZF, SF */
3309 	ctxt->src.type = OP_IMM;
3310 	ctxt->src.val = 0;
3311 	ctxt->src.bytes = 1;
3312 	fastop(ctxt, em_or);
3313 	ctxt->eflags &= ~(X86_EFLAGS_AF | X86_EFLAGS_CF);
3314 	if (cf)
3315 		ctxt->eflags |= X86_EFLAGS_CF;
3316 	if (af)
3317 		ctxt->eflags |= X86_EFLAGS_AF;
3318 	return X86EMUL_CONTINUE;
3319 }
3320 
3321 static int em_aam(struct x86_emulate_ctxt *ctxt)
3322 {
3323 	u8 al, ah;
3324 
3325 	if (ctxt->src.val == 0)
3326 		return emulate_de(ctxt);
3327 
3328 	al = ctxt->dst.val & 0xff;
3329 	ah = al / ctxt->src.val;
3330 	al %= ctxt->src.val;
3331 
3332 	ctxt->dst.val = (ctxt->dst.val & 0xffff0000) | al | (ah << 8);
3333 
3334 	/* Set PF, ZF, SF */
3335 	ctxt->src.type = OP_IMM;
3336 	ctxt->src.val = 0;
3337 	ctxt->src.bytes = 1;
3338 	fastop(ctxt, em_or);
3339 
3340 	return X86EMUL_CONTINUE;
3341 }
3342 
3343 static int em_aad(struct x86_emulate_ctxt *ctxt)
3344 {
3345 	u8 al = ctxt->dst.val & 0xff;
3346 	u8 ah = (ctxt->dst.val >> 8) & 0xff;
3347 
3348 	al = (al + (ah * ctxt->src.val)) & 0xff;
3349 
3350 	ctxt->dst.val = (ctxt->dst.val & 0xffff0000) | al;
3351 
3352 	/* Set PF, ZF, SF */
3353 	ctxt->src.type = OP_IMM;
3354 	ctxt->src.val = 0;
3355 	ctxt->src.bytes = 1;
3356 	fastop(ctxt, em_or);
3357 
3358 	return X86EMUL_CONTINUE;
3359 }
3360 
3361 static int em_call(struct x86_emulate_ctxt *ctxt)
3362 {
3363 	int rc;
3364 	long rel = ctxt->src.val;
3365 
3366 	ctxt->src.val = (unsigned long)ctxt->_eip;
3367 	rc = jmp_rel(ctxt, rel);
3368 	if (rc != X86EMUL_CONTINUE)
3369 		return rc;
3370 	return em_push(ctxt);
3371 }
3372 
3373 static int em_call_far(struct x86_emulate_ctxt *ctxt)
3374 {
3375 	u16 sel, old_cs;
3376 	ulong old_eip;
3377 	int rc;
3378 	struct desc_struct old_desc, new_desc;
3379 	const struct x86_emulate_ops *ops = ctxt->ops;
3380 	int cpl = ctxt->ops->cpl(ctxt);
3381 	enum x86emul_mode prev_mode = ctxt->mode;
3382 
3383 	old_eip = ctxt->_eip;
3384 	ops->get_segment(ctxt, &old_cs, &old_desc, NULL, VCPU_SREG_CS);
3385 
3386 	memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
3387 	rc = __load_segment_descriptor(ctxt, sel, VCPU_SREG_CS, cpl,
3388 				       X86_TRANSFER_CALL_JMP, &new_desc);
3389 	if (rc != X86EMUL_CONTINUE)
3390 		return rc;
3391 
3392 	rc = assign_eip_far(ctxt, ctxt->src.val, &new_desc);
3393 	if (rc != X86EMUL_CONTINUE)
3394 		goto fail;
3395 
3396 	ctxt->src.val = old_cs;
3397 	rc = em_push(ctxt);
3398 	if (rc != X86EMUL_CONTINUE)
3399 		goto fail;
3400 
3401 	ctxt->src.val = old_eip;
3402 	rc = em_push(ctxt);
3403 	/* If we failed, we tainted the memory, but the very least we should
3404 	   restore cs */
3405 	if (rc != X86EMUL_CONTINUE) {
3406 		pr_warn_once("faulting far call emulation tainted memory\n");
3407 		goto fail;
3408 	}
3409 	return rc;
3410 fail:
3411 	ops->set_segment(ctxt, old_cs, &old_desc, 0, VCPU_SREG_CS);
3412 	ctxt->mode = prev_mode;
3413 	return rc;
3414 
3415 }
3416 
3417 static int em_ret_near_imm(struct x86_emulate_ctxt *ctxt)
3418 {
3419 	int rc;
3420 	unsigned long eip;
3421 
3422 	rc = emulate_pop(ctxt, &eip, ctxt->op_bytes);
3423 	if (rc != X86EMUL_CONTINUE)
3424 		return rc;
3425 	rc = assign_eip_near(ctxt, eip);
3426 	if (rc != X86EMUL_CONTINUE)
3427 		return rc;
3428 	rsp_increment(ctxt, ctxt->src.val);
3429 	return X86EMUL_CONTINUE;
3430 }
3431 
3432 static int em_xchg(struct x86_emulate_ctxt *ctxt)
3433 {
3434 	/* Write back the register source. */
3435 	ctxt->src.val = ctxt->dst.val;
3436 	write_register_operand(&ctxt->src);
3437 
3438 	/* Write back the memory destination with implicit LOCK prefix. */
3439 	ctxt->dst.val = ctxt->src.orig_val;
3440 	ctxt->lock_prefix = 1;
3441 	return X86EMUL_CONTINUE;
3442 }
3443 
3444 static int em_imul_3op(struct x86_emulate_ctxt *ctxt)
3445 {
3446 	ctxt->dst.val = ctxt->src2.val;
3447 	return fastop(ctxt, em_imul);
3448 }
3449 
3450 static int em_cwd(struct x86_emulate_ctxt *ctxt)
3451 {
3452 	ctxt->dst.type = OP_REG;
3453 	ctxt->dst.bytes = ctxt->src.bytes;
3454 	ctxt->dst.addr.reg = reg_rmw(ctxt, VCPU_REGS_RDX);
3455 	ctxt->dst.val = ~((ctxt->src.val >> (ctxt->src.bytes * 8 - 1)) - 1);
3456 
3457 	return X86EMUL_CONTINUE;
3458 }
3459 
3460 static int em_rdtsc(struct x86_emulate_ctxt *ctxt)
3461 {
3462 	u64 tsc = 0;
3463 
3464 	ctxt->ops->get_msr(ctxt, MSR_IA32_TSC, &tsc);
3465 	*reg_write(ctxt, VCPU_REGS_RAX) = (u32)tsc;
3466 	*reg_write(ctxt, VCPU_REGS_RDX) = tsc >> 32;
3467 	return X86EMUL_CONTINUE;
3468 }
3469 
3470 static int em_rdpmc(struct x86_emulate_ctxt *ctxt)
3471 {
3472 	u64 pmc;
3473 
3474 	if (ctxt->ops->read_pmc(ctxt, reg_read(ctxt, VCPU_REGS_RCX), &pmc))
3475 		return emulate_gp(ctxt, 0);
3476 	*reg_write(ctxt, VCPU_REGS_RAX) = (u32)pmc;
3477 	*reg_write(ctxt, VCPU_REGS_RDX) = pmc >> 32;
3478 	return X86EMUL_CONTINUE;
3479 }
3480 
3481 static int em_mov(struct x86_emulate_ctxt *ctxt)
3482 {
3483 	memcpy(ctxt->dst.valptr, ctxt->src.valptr, sizeof(ctxt->src.valptr));
3484 	return X86EMUL_CONTINUE;
3485 }
3486 
3487 #define FFL(x) bit(X86_FEATURE_##x)
3488 
3489 static int em_movbe(struct x86_emulate_ctxt *ctxt)
3490 {
3491 	u32 ebx, ecx, edx, eax = 1;
3492 	u16 tmp;
3493 
3494 	/*
3495 	 * Check MOVBE is set in the guest-visible CPUID leaf.
3496 	 */
3497 	ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
3498 	if (!(ecx & FFL(MOVBE)))
3499 		return emulate_ud(ctxt);
3500 
3501 	switch (ctxt->op_bytes) {
3502 	case 2:
3503 		/*
3504 		 * From MOVBE definition: "...When the operand size is 16 bits,
3505 		 * the upper word of the destination register remains unchanged
3506 		 * ..."
3507 		 *
3508 		 * Both casting ->valptr and ->val to u16 breaks strict aliasing
3509 		 * rules so we have to do the operation almost per hand.
3510 		 */
3511 		tmp = (u16)ctxt->src.val;
3512 		ctxt->dst.val &= ~0xffffUL;
3513 		ctxt->dst.val |= (unsigned long)swab16(tmp);
3514 		break;
3515 	case 4:
3516 		ctxt->dst.val = swab32((u32)ctxt->src.val);
3517 		break;
3518 	case 8:
3519 		ctxt->dst.val = swab64(ctxt->src.val);
3520 		break;
3521 	default:
3522 		BUG();
3523 	}
3524 	return X86EMUL_CONTINUE;
3525 }
3526 
3527 static int em_cr_write(struct x86_emulate_ctxt *ctxt)
3528 {
3529 	if (ctxt->ops->set_cr(ctxt, ctxt->modrm_reg, ctxt->src.val))
3530 		return emulate_gp(ctxt, 0);
3531 
3532 	/* Disable writeback. */
3533 	ctxt->dst.type = OP_NONE;
3534 	return X86EMUL_CONTINUE;
3535 }
3536 
3537 static int em_dr_write(struct x86_emulate_ctxt *ctxt)
3538 {
3539 	unsigned long val;
3540 
3541 	if (ctxt->mode == X86EMUL_MODE_PROT64)
3542 		val = ctxt->src.val & ~0ULL;
3543 	else
3544 		val = ctxt->src.val & ~0U;
3545 
3546 	/* #UD condition is already handled. */
3547 	if (ctxt->ops->set_dr(ctxt, ctxt->modrm_reg, val) < 0)
3548 		return emulate_gp(ctxt, 0);
3549 
3550 	/* Disable writeback. */
3551 	ctxt->dst.type = OP_NONE;
3552 	return X86EMUL_CONTINUE;
3553 }
3554 
3555 static int em_wrmsr(struct x86_emulate_ctxt *ctxt)
3556 {
3557 	u64 msr_data;
3558 
3559 	msr_data = (u32)reg_read(ctxt, VCPU_REGS_RAX)
3560 		| ((u64)reg_read(ctxt, VCPU_REGS_RDX) << 32);
3561 	if (ctxt->ops->set_msr(ctxt, reg_read(ctxt, VCPU_REGS_RCX), msr_data))
3562 		return emulate_gp(ctxt, 0);
3563 
3564 	return X86EMUL_CONTINUE;
3565 }
3566 
3567 static int em_rdmsr(struct x86_emulate_ctxt *ctxt)
3568 {
3569 	u64 msr_data;
3570 
3571 	if (ctxt->ops->get_msr(ctxt, reg_read(ctxt, VCPU_REGS_RCX), &msr_data))
3572 		return emulate_gp(ctxt, 0);
3573 
3574 	*reg_write(ctxt, VCPU_REGS_RAX) = (u32)msr_data;
3575 	*reg_write(ctxt, VCPU_REGS_RDX) = msr_data >> 32;
3576 	return X86EMUL_CONTINUE;
3577 }
3578 
3579 static int em_mov_rm_sreg(struct x86_emulate_ctxt *ctxt)
3580 {
3581 	if (ctxt->modrm_reg > VCPU_SREG_GS)
3582 		return emulate_ud(ctxt);
3583 
3584 	ctxt->dst.val = get_segment_selector(ctxt, ctxt->modrm_reg);
3585 	if (ctxt->dst.bytes == 4 && ctxt->dst.type == OP_MEM)
3586 		ctxt->dst.bytes = 2;
3587 	return X86EMUL_CONTINUE;
3588 }
3589 
3590 static int em_mov_sreg_rm(struct x86_emulate_ctxt *ctxt)
3591 {
3592 	u16 sel = ctxt->src.val;
3593 
3594 	if (ctxt->modrm_reg == VCPU_SREG_CS || ctxt->modrm_reg > VCPU_SREG_GS)
3595 		return emulate_ud(ctxt);
3596 
3597 	if (ctxt->modrm_reg == VCPU_SREG_SS)
3598 		ctxt->interruptibility = KVM_X86_SHADOW_INT_MOV_SS;
3599 
3600 	/* Disable writeback. */
3601 	ctxt->dst.type = OP_NONE;
3602 	return load_segment_descriptor(ctxt, sel, ctxt->modrm_reg);
3603 }
3604 
3605 static int em_lldt(struct x86_emulate_ctxt *ctxt)
3606 {
3607 	u16 sel = ctxt->src.val;
3608 
3609 	/* Disable writeback. */
3610 	ctxt->dst.type = OP_NONE;
3611 	return load_segment_descriptor(ctxt, sel, VCPU_SREG_LDTR);
3612 }
3613 
3614 static int em_ltr(struct x86_emulate_ctxt *ctxt)
3615 {
3616 	u16 sel = ctxt->src.val;
3617 
3618 	/* Disable writeback. */
3619 	ctxt->dst.type = OP_NONE;
3620 	return load_segment_descriptor(ctxt, sel, VCPU_SREG_TR);
3621 }
3622 
3623 static int em_invlpg(struct x86_emulate_ctxt *ctxt)
3624 {
3625 	int rc;
3626 	ulong linear;
3627 
3628 	rc = linearize(ctxt, ctxt->src.addr.mem, 1, false, &linear);
3629 	if (rc == X86EMUL_CONTINUE)
3630 		ctxt->ops->invlpg(ctxt, linear);
3631 	/* Disable writeback. */
3632 	ctxt->dst.type = OP_NONE;
3633 	return X86EMUL_CONTINUE;
3634 }
3635 
3636 static int em_clts(struct x86_emulate_ctxt *ctxt)
3637 {
3638 	ulong cr0;
3639 
3640 	cr0 = ctxt->ops->get_cr(ctxt, 0);
3641 	cr0 &= ~X86_CR0_TS;
3642 	ctxt->ops->set_cr(ctxt, 0, cr0);
3643 	return X86EMUL_CONTINUE;
3644 }
3645 
3646 static int em_hypercall(struct x86_emulate_ctxt *ctxt)
3647 {
3648 	int rc = ctxt->ops->fix_hypercall(ctxt);
3649 
3650 	if (rc != X86EMUL_CONTINUE)
3651 		return rc;
3652 
3653 	/* Let the processor re-execute the fixed hypercall */
3654 	ctxt->_eip = ctxt->eip;
3655 	/* Disable writeback. */
3656 	ctxt->dst.type = OP_NONE;
3657 	return X86EMUL_CONTINUE;
3658 }
3659 
3660 static int emulate_store_desc_ptr(struct x86_emulate_ctxt *ctxt,
3661 				  void (*get)(struct x86_emulate_ctxt *ctxt,
3662 					      struct desc_ptr *ptr))
3663 {
3664 	struct desc_ptr desc_ptr;
3665 
3666 	if (ctxt->mode == X86EMUL_MODE_PROT64)
3667 		ctxt->op_bytes = 8;
3668 	get(ctxt, &desc_ptr);
3669 	if (ctxt->op_bytes == 2) {
3670 		ctxt->op_bytes = 4;
3671 		desc_ptr.address &= 0x00ffffff;
3672 	}
3673 	/* Disable writeback. */
3674 	ctxt->dst.type = OP_NONE;
3675 	return segmented_write(ctxt, ctxt->dst.addr.mem,
3676 			       &desc_ptr, 2 + ctxt->op_bytes);
3677 }
3678 
3679 static int em_sgdt(struct x86_emulate_ctxt *ctxt)
3680 {
3681 	return emulate_store_desc_ptr(ctxt, ctxt->ops->get_gdt);
3682 }
3683 
3684 static int em_sidt(struct x86_emulate_ctxt *ctxt)
3685 {
3686 	return emulate_store_desc_ptr(ctxt, ctxt->ops->get_idt);
3687 }
3688 
3689 static int em_lgdt_lidt(struct x86_emulate_ctxt *ctxt, bool lgdt)
3690 {
3691 	struct desc_ptr desc_ptr;
3692 	int rc;
3693 
3694 	if (ctxt->mode == X86EMUL_MODE_PROT64)
3695 		ctxt->op_bytes = 8;
3696 	rc = read_descriptor(ctxt, ctxt->src.addr.mem,
3697 			     &desc_ptr.size, &desc_ptr.address,
3698 			     ctxt->op_bytes);
3699 	if (rc != X86EMUL_CONTINUE)
3700 		return rc;
3701 	if (ctxt->mode == X86EMUL_MODE_PROT64 &&
3702 	    is_noncanonical_address(desc_ptr.address))
3703 		return emulate_gp(ctxt, 0);
3704 	if (lgdt)
3705 		ctxt->ops->set_gdt(ctxt, &desc_ptr);
3706 	else
3707 		ctxt->ops->set_idt(ctxt, &desc_ptr);
3708 	/* Disable writeback. */
3709 	ctxt->dst.type = OP_NONE;
3710 	return X86EMUL_CONTINUE;
3711 }
3712 
3713 static int em_lgdt(struct x86_emulate_ctxt *ctxt)
3714 {
3715 	return em_lgdt_lidt(ctxt, true);
3716 }
3717 
3718 static int em_lidt(struct x86_emulate_ctxt *ctxt)
3719 {
3720 	return em_lgdt_lidt(ctxt, false);
3721 }
3722 
3723 static int em_smsw(struct x86_emulate_ctxt *ctxt)
3724 {
3725 	if (ctxt->dst.type == OP_MEM)
3726 		ctxt->dst.bytes = 2;
3727 	ctxt->dst.val = ctxt->ops->get_cr(ctxt, 0);
3728 	return X86EMUL_CONTINUE;
3729 }
3730 
3731 static int em_lmsw(struct x86_emulate_ctxt *ctxt)
3732 {
3733 	ctxt->ops->set_cr(ctxt, 0, (ctxt->ops->get_cr(ctxt, 0) & ~0x0eul)
3734 			  | (ctxt->src.val & 0x0f));
3735 	ctxt->dst.type = OP_NONE;
3736 	return X86EMUL_CONTINUE;
3737 }
3738 
3739 static int em_loop(struct x86_emulate_ctxt *ctxt)
3740 {
3741 	int rc = X86EMUL_CONTINUE;
3742 
3743 	register_address_increment(ctxt, VCPU_REGS_RCX, -1);
3744 	if ((address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) != 0) &&
3745 	    (ctxt->b == 0xe2 || test_cc(ctxt->b ^ 0x5, ctxt->eflags)))
3746 		rc = jmp_rel(ctxt, ctxt->src.val);
3747 
3748 	return rc;
3749 }
3750 
3751 static int em_jcxz(struct x86_emulate_ctxt *ctxt)
3752 {
3753 	int rc = X86EMUL_CONTINUE;
3754 
3755 	if (address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) == 0)
3756 		rc = jmp_rel(ctxt, ctxt->src.val);
3757 
3758 	return rc;
3759 }
3760 
3761 static int em_in(struct x86_emulate_ctxt *ctxt)
3762 {
3763 	if (!pio_in_emulated(ctxt, ctxt->dst.bytes, ctxt->src.val,
3764 			     &ctxt->dst.val))
3765 		return X86EMUL_IO_NEEDED;
3766 
3767 	return X86EMUL_CONTINUE;
3768 }
3769 
3770 static int em_out(struct x86_emulate_ctxt *ctxt)
3771 {
3772 	ctxt->ops->pio_out_emulated(ctxt, ctxt->src.bytes, ctxt->dst.val,
3773 				    &ctxt->src.val, 1);
3774 	/* Disable writeback. */
3775 	ctxt->dst.type = OP_NONE;
3776 	return X86EMUL_CONTINUE;
3777 }
3778 
3779 static int em_cli(struct x86_emulate_ctxt *ctxt)
3780 {
3781 	if (emulator_bad_iopl(ctxt))
3782 		return emulate_gp(ctxt, 0);
3783 
3784 	ctxt->eflags &= ~X86_EFLAGS_IF;
3785 	return X86EMUL_CONTINUE;
3786 }
3787 
3788 static int em_sti(struct x86_emulate_ctxt *ctxt)
3789 {
3790 	if (emulator_bad_iopl(ctxt))
3791 		return emulate_gp(ctxt, 0);
3792 
3793 	ctxt->interruptibility = KVM_X86_SHADOW_INT_STI;
3794 	ctxt->eflags |= X86_EFLAGS_IF;
3795 	return X86EMUL_CONTINUE;
3796 }
3797 
3798 static int em_cpuid(struct x86_emulate_ctxt *ctxt)
3799 {
3800 	u32 eax, ebx, ecx, edx;
3801 
3802 	eax = reg_read(ctxt, VCPU_REGS_RAX);
3803 	ecx = reg_read(ctxt, VCPU_REGS_RCX);
3804 	ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
3805 	*reg_write(ctxt, VCPU_REGS_RAX) = eax;
3806 	*reg_write(ctxt, VCPU_REGS_RBX) = ebx;
3807 	*reg_write(ctxt, VCPU_REGS_RCX) = ecx;
3808 	*reg_write(ctxt, VCPU_REGS_RDX) = edx;
3809 	return X86EMUL_CONTINUE;
3810 }
3811 
3812 static int em_sahf(struct x86_emulate_ctxt *ctxt)
3813 {
3814 	u32 flags;
3815 
3816 	flags = X86_EFLAGS_CF | X86_EFLAGS_PF | X86_EFLAGS_AF | X86_EFLAGS_ZF |
3817 		X86_EFLAGS_SF;
3818 	flags &= *reg_rmw(ctxt, VCPU_REGS_RAX) >> 8;
3819 
3820 	ctxt->eflags &= ~0xffUL;
3821 	ctxt->eflags |= flags | X86_EFLAGS_FIXED;
3822 	return X86EMUL_CONTINUE;
3823 }
3824 
3825 static int em_lahf(struct x86_emulate_ctxt *ctxt)
3826 {
3827 	*reg_rmw(ctxt, VCPU_REGS_RAX) &= ~0xff00UL;
3828 	*reg_rmw(ctxt, VCPU_REGS_RAX) |= (ctxt->eflags & 0xff) << 8;
3829 	return X86EMUL_CONTINUE;
3830 }
3831 
3832 static int em_bswap(struct x86_emulate_ctxt *ctxt)
3833 {
3834 	switch (ctxt->op_bytes) {
3835 #ifdef CONFIG_X86_64
3836 	case 8:
3837 		asm("bswap %0" : "+r"(ctxt->dst.val));
3838 		break;
3839 #endif
3840 	default:
3841 		asm("bswap %0" : "+r"(*(u32 *)&ctxt->dst.val));
3842 		break;
3843 	}
3844 	return X86EMUL_CONTINUE;
3845 }
3846 
3847 static int em_clflush(struct x86_emulate_ctxt *ctxt)
3848 {
3849 	/* emulating clflush regardless of cpuid */
3850 	return X86EMUL_CONTINUE;
3851 }
3852 
3853 static int em_movsxd(struct x86_emulate_ctxt *ctxt)
3854 {
3855 	ctxt->dst.val = (s32) ctxt->src.val;
3856 	return X86EMUL_CONTINUE;
3857 }
3858 
3859 static bool valid_cr(int nr)
3860 {
3861 	switch (nr) {
3862 	case 0:
3863 	case 2 ... 4:
3864 	case 8:
3865 		return true;
3866 	default:
3867 		return false;
3868 	}
3869 }
3870 
3871 static int check_cr_read(struct x86_emulate_ctxt *ctxt)
3872 {
3873 	if (!valid_cr(ctxt->modrm_reg))
3874 		return emulate_ud(ctxt);
3875 
3876 	return X86EMUL_CONTINUE;
3877 }
3878 
3879 static int check_cr_write(struct x86_emulate_ctxt *ctxt)
3880 {
3881 	u64 new_val = ctxt->src.val64;
3882 	int cr = ctxt->modrm_reg;
3883 	u64 efer = 0;
3884 
3885 	static u64 cr_reserved_bits[] = {
3886 		0xffffffff00000000ULL,
3887 		0, 0, 0, /* CR3 checked later */
3888 		CR4_RESERVED_BITS,
3889 		0, 0, 0,
3890 		CR8_RESERVED_BITS,
3891 	};
3892 
3893 	if (!valid_cr(cr))
3894 		return emulate_ud(ctxt);
3895 
3896 	if (new_val & cr_reserved_bits[cr])
3897 		return emulate_gp(ctxt, 0);
3898 
3899 	switch (cr) {
3900 	case 0: {
3901 		u64 cr4;
3902 		if (((new_val & X86_CR0_PG) && !(new_val & X86_CR0_PE)) ||
3903 		    ((new_val & X86_CR0_NW) && !(new_val & X86_CR0_CD)))
3904 			return emulate_gp(ctxt, 0);
3905 
3906 		cr4 = ctxt->ops->get_cr(ctxt, 4);
3907 		ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3908 
3909 		if ((new_val & X86_CR0_PG) && (efer & EFER_LME) &&
3910 		    !(cr4 & X86_CR4_PAE))
3911 			return emulate_gp(ctxt, 0);
3912 
3913 		break;
3914 		}
3915 	case 3: {
3916 		u64 rsvd = 0;
3917 
3918 		ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3919 		if (efer & EFER_LMA)
3920 			rsvd = CR3_L_MODE_RESERVED_BITS & ~CR3_PCID_INVD;
3921 
3922 		if (new_val & rsvd)
3923 			return emulate_gp(ctxt, 0);
3924 
3925 		break;
3926 		}
3927 	case 4: {
3928 		ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3929 
3930 		if ((efer & EFER_LMA) && !(new_val & X86_CR4_PAE))
3931 			return emulate_gp(ctxt, 0);
3932 
3933 		break;
3934 		}
3935 	}
3936 
3937 	return X86EMUL_CONTINUE;
3938 }
3939 
3940 static int check_dr7_gd(struct x86_emulate_ctxt *ctxt)
3941 {
3942 	unsigned long dr7;
3943 
3944 	ctxt->ops->get_dr(ctxt, 7, &dr7);
3945 
3946 	/* Check if DR7.Global_Enable is set */
3947 	return dr7 & (1 << 13);
3948 }
3949 
3950 static int check_dr_read(struct x86_emulate_ctxt *ctxt)
3951 {
3952 	int dr = ctxt->modrm_reg;
3953 	u64 cr4;
3954 
3955 	if (dr > 7)
3956 		return emulate_ud(ctxt);
3957 
3958 	cr4 = ctxt->ops->get_cr(ctxt, 4);
3959 	if ((cr4 & X86_CR4_DE) && (dr == 4 || dr == 5))
3960 		return emulate_ud(ctxt);
3961 
3962 	if (check_dr7_gd(ctxt)) {
3963 		ulong dr6;
3964 
3965 		ctxt->ops->get_dr(ctxt, 6, &dr6);
3966 		dr6 &= ~15;
3967 		dr6 |= DR6_BD | DR6_RTM;
3968 		ctxt->ops->set_dr(ctxt, 6, dr6);
3969 		return emulate_db(ctxt);
3970 	}
3971 
3972 	return X86EMUL_CONTINUE;
3973 }
3974 
3975 static int check_dr_write(struct x86_emulate_ctxt *ctxt)
3976 {
3977 	u64 new_val = ctxt->src.val64;
3978 	int dr = ctxt->modrm_reg;
3979 
3980 	if ((dr == 6 || dr == 7) && (new_val & 0xffffffff00000000ULL))
3981 		return emulate_gp(ctxt, 0);
3982 
3983 	return check_dr_read(ctxt);
3984 }
3985 
3986 static int check_svme(struct x86_emulate_ctxt *ctxt)
3987 {
3988 	u64 efer;
3989 
3990 	ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3991 
3992 	if (!(efer & EFER_SVME))
3993 		return emulate_ud(ctxt);
3994 
3995 	return X86EMUL_CONTINUE;
3996 }
3997 
3998 static int check_svme_pa(struct x86_emulate_ctxt *ctxt)
3999 {
4000 	u64 rax = reg_read(ctxt, VCPU_REGS_RAX);
4001 
4002 	/* Valid physical address? */
4003 	if (rax & 0xffff000000000000ULL)
4004 		return emulate_gp(ctxt, 0);
4005 
4006 	return check_svme(ctxt);
4007 }
4008 
4009 static int check_rdtsc(struct x86_emulate_ctxt *ctxt)
4010 {
4011 	u64 cr4 = ctxt->ops->get_cr(ctxt, 4);
4012 
4013 	if (cr4 & X86_CR4_TSD && ctxt->ops->cpl(ctxt))
4014 		return emulate_ud(ctxt);
4015 
4016 	return X86EMUL_CONTINUE;
4017 }
4018 
4019 static int check_rdpmc(struct x86_emulate_ctxt *ctxt)
4020 {
4021 	u64 cr4 = ctxt->ops->get_cr(ctxt, 4);
4022 	u64 rcx = reg_read(ctxt, VCPU_REGS_RCX);
4023 
4024 	if ((!(cr4 & X86_CR4_PCE) && ctxt->ops->cpl(ctxt)) ||
4025 	    ctxt->ops->check_pmc(ctxt, rcx))
4026 		return emulate_gp(ctxt, 0);
4027 
4028 	return X86EMUL_CONTINUE;
4029 }
4030 
4031 static int check_perm_in(struct x86_emulate_ctxt *ctxt)
4032 {
4033 	ctxt->dst.bytes = min(ctxt->dst.bytes, 4u);
4034 	if (!emulator_io_permited(ctxt, ctxt->src.val, ctxt->dst.bytes))
4035 		return emulate_gp(ctxt, 0);
4036 
4037 	return X86EMUL_CONTINUE;
4038 }
4039 
4040 static int check_perm_out(struct x86_emulate_ctxt *ctxt)
4041 {
4042 	ctxt->src.bytes = min(ctxt->src.bytes, 4u);
4043 	if (!emulator_io_permited(ctxt, ctxt->dst.val, ctxt->src.bytes))
4044 		return emulate_gp(ctxt, 0);
4045 
4046 	return X86EMUL_CONTINUE;
4047 }
4048 
4049 #define D(_y) { .flags = (_y) }
4050 #define DI(_y, _i) { .flags = (_y)|Intercept, .intercept = x86_intercept_##_i }
4051 #define DIP(_y, _i, _p) { .flags = (_y)|Intercept|CheckPerm, \
4052 		      .intercept = x86_intercept_##_i, .check_perm = (_p) }
4053 #define N    D(NotImpl)
4054 #define EXT(_f, _e) { .flags = ((_f) | RMExt), .u.group = (_e) }
4055 #define G(_f, _g) { .flags = ((_f) | Group | ModRM), .u.group = (_g) }
4056 #define GD(_f, _g) { .flags = ((_f) | GroupDual | ModRM), .u.gdual = (_g) }
4057 #define ID(_f, _i) { .flags = ((_f) | InstrDual | ModRM), .u.idual = (_i) }
4058 #define MD(_f, _m) { .flags = ((_f) | ModeDual), .u.mdual = (_m) }
4059 #define E(_f, _e) { .flags = ((_f) | Escape | ModRM), .u.esc = (_e) }
4060 #define I(_f, _e) { .flags = (_f), .u.execute = (_e) }
4061 #define F(_f, _e) { .flags = (_f) | Fastop, .u.fastop = (_e) }
4062 #define II(_f, _e, _i) \
4063 	{ .flags = (_f)|Intercept, .u.execute = (_e), .intercept = x86_intercept_##_i }
4064 #define IIP(_f, _e, _i, _p) \
4065 	{ .flags = (_f)|Intercept|CheckPerm, .u.execute = (_e), \
4066 	  .intercept = x86_intercept_##_i, .check_perm = (_p) }
4067 #define GP(_f, _g) { .flags = ((_f) | Prefix), .u.gprefix = (_g) }
4068 
4069 #define D2bv(_f)      D((_f) | ByteOp), D(_f)
4070 #define D2bvIP(_f, _i, _p) DIP((_f) | ByteOp, _i, _p), DIP(_f, _i, _p)
4071 #define I2bv(_f, _e)  I((_f) | ByteOp, _e), I(_f, _e)
4072 #define F2bv(_f, _e)  F((_f) | ByteOp, _e), F(_f, _e)
4073 #define I2bvIP(_f, _e, _i, _p) \
4074 	IIP((_f) | ByteOp, _e, _i, _p), IIP(_f, _e, _i, _p)
4075 
4076 #define F6ALU(_f, _e) F2bv((_f) | DstMem | SrcReg | ModRM, _e),		\
4077 		F2bv(((_f) | DstReg | SrcMem | ModRM) & ~Lock, _e),	\
4078 		F2bv(((_f) & ~Lock) | DstAcc | SrcImm, _e)
4079 
4080 static const struct opcode group7_rm0[] = {
4081 	N,
4082 	I(SrcNone | Priv | EmulateOnUD,	em_hypercall),
4083 	N, N, N, N, N, N,
4084 };
4085 
4086 static const struct opcode group7_rm1[] = {
4087 	DI(SrcNone | Priv, monitor),
4088 	DI(SrcNone | Priv, mwait),
4089 	N, N, N, N, N, N,
4090 };
4091 
4092 static const struct opcode group7_rm3[] = {
4093 	DIP(SrcNone | Prot | Priv,		vmrun,		check_svme_pa),
4094 	II(SrcNone  | Prot | EmulateOnUD,	em_hypercall,	vmmcall),
4095 	DIP(SrcNone | Prot | Priv,		vmload,		check_svme_pa),
4096 	DIP(SrcNone | Prot | Priv,		vmsave,		check_svme_pa),
4097 	DIP(SrcNone | Prot | Priv,		stgi,		check_svme),
4098 	DIP(SrcNone | Prot | Priv,		clgi,		check_svme),
4099 	DIP(SrcNone | Prot | Priv,		skinit,		check_svme),
4100 	DIP(SrcNone | Prot | Priv,		invlpga,	check_svme),
4101 };
4102 
4103 static const struct opcode group7_rm7[] = {
4104 	N,
4105 	DIP(SrcNone, rdtscp, check_rdtsc),
4106 	N, N, N, N, N, N,
4107 };
4108 
4109 static const struct opcode group1[] = {
4110 	F(Lock, em_add),
4111 	F(Lock | PageTable, em_or),
4112 	F(Lock, em_adc),
4113 	F(Lock, em_sbb),
4114 	F(Lock | PageTable, em_and),
4115 	F(Lock, em_sub),
4116 	F(Lock, em_xor),
4117 	F(NoWrite, em_cmp),
4118 };
4119 
4120 static const struct opcode group1A[] = {
4121 	I(DstMem | SrcNone | Mov | Stack | IncSP, em_pop), N, N, N, N, N, N, N,
4122 };
4123 
4124 static const struct opcode group2[] = {
4125 	F(DstMem | ModRM, em_rol),
4126 	F(DstMem | ModRM, em_ror),
4127 	F(DstMem | ModRM, em_rcl),
4128 	F(DstMem | ModRM, em_rcr),
4129 	F(DstMem | ModRM, em_shl),
4130 	F(DstMem | ModRM, em_shr),
4131 	F(DstMem | ModRM, em_shl),
4132 	F(DstMem | ModRM, em_sar),
4133 };
4134 
4135 static const struct opcode group3[] = {
4136 	F(DstMem | SrcImm | NoWrite, em_test),
4137 	F(DstMem | SrcImm | NoWrite, em_test),
4138 	F(DstMem | SrcNone | Lock, em_not),
4139 	F(DstMem | SrcNone | Lock, em_neg),
4140 	F(DstXacc | Src2Mem, em_mul_ex),
4141 	F(DstXacc | Src2Mem, em_imul_ex),
4142 	F(DstXacc | Src2Mem, em_div_ex),
4143 	F(DstXacc | Src2Mem, em_idiv_ex),
4144 };
4145 
4146 static const struct opcode group4[] = {
4147 	F(ByteOp | DstMem | SrcNone | Lock, em_inc),
4148 	F(ByteOp | DstMem | SrcNone | Lock, em_dec),
4149 	N, N, N, N, N, N,
4150 };
4151 
4152 static const struct opcode group5[] = {
4153 	F(DstMem | SrcNone | Lock,		em_inc),
4154 	F(DstMem | SrcNone | Lock,		em_dec),
4155 	I(SrcMem | NearBranch,			em_call_near_abs),
4156 	I(SrcMemFAddr | ImplicitOps,		em_call_far),
4157 	I(SrcMem | NearBranch,			em_jmp_abs),
4158 	I(SrcMemFAddr | ImplicitOps,		em_jmp_far),
4159 	I(SrcMem | Stack,			em_push), D(Undefined),
4160 };
4161 
4162 static const struct opcode group6[] = {
4163 	DI(Prot | DstMem,	sldt),
4164 	DI(Prot | DstMem,	str),
4165 	II(Prot | Priv | SrcMem16, em_lldt, lldt),
4166 	II(Prot | Priv | SrcMem16, em_ltr, ltr),
4167 	N, N, N, N,
4168 };
4169 
4170 static const struct group_dual group7 = { {
4171 	II(Mov | DstMem,			em_sgdt, sgdt),
4172 	II(Mov | DstMem,			em_sidt, sidt),
4173 	II(SrcMem | Priv,			em_lgdt, lgdt),
4174 	II(SrcMem | Priv,			em_lidt, lidt),
4175 	II(SrcNone | DstMem | Mov,		em_smsw, smsw), N,
4176 	II(SrcMem16 | Mov | Priv,		em_lmsw, lmsw),
4177 	II(SrcMem | ByteOp | Priv | NoAccess,	em_invlpg, invlpg),
4178 }, {
4179 	EXT(0, group7_rm0),
4180 	EXT(0, group7_rm1),
4181 	N, EXT(0, group7_rm3),
4182 	II(SrcNone | DstMem | Mov,		em_smsw, smsw), N,
4183 	II(SrcMem16 | Mov | Priv,		em_lmsw, lmsw),
4184 	EXT(0, group7_rm7),
4185 } };
4186 
4187 static const struct opcode group8[] = {
4188 	N, N, N, N,
4189 	F(DstMem | SrcImmByte | NoWrite,		em_bt),
4190 	F(DstMem | SrcImmByte | Lock | PageTable,	em_bts),
4191 	F(DstMem | SrcImmByte | Lock,			em_btr),
4192 	F(DstMem | SrcImmByte | Lock | PageTable,	em_btc),
4193 };
4194 
4195 static const struct group_dual group9 = { {
4196 	N, I(DstMem64 | Lock | PageTable, em_cmpxchg8b), N, N, N, N, N, N,
4197 }, {
4198 	N, N, N, N, N, N, N, N,
4199 } };
4200 
4201 static const struct opcode group11[] = {
4202 	I(DstMem | SrcImm | Mov | PageTable, em_mov),
4203 	X7(D(Undefined)),
4204 };
4205 
4206 static const struct gprefix pfx_0f_ae_7 = {
4207 	I(SrcMem | ByteOp, em_clflush), N, N, N,
4208 };
4209 
4210 static const struct group_dual group15 = { {
4211 	N, N, N, N, N, N, N, GP(0, &pfx_0f_ae_7),
4212 }, {
4213 	N, N, N, N, N, N, N, N,
4214 } };
4215 
4216 static const struct gprefix pfx_0f_6f_0f_7f = {
4217 	I(Mmx, em_mov), I(Sse | Aligned, em_mov), N, I(Sse | Unaligned, em_mov),
4218 };
4219 
4220 static const struct instr_dual instr_dual_0f_2b = {
4221 	I(0, em_mov), N
4222 };
4223 
4224 static const struct gprefix pfx_0f_2b = {
4225 	ID(0, &instr_dual_0f_2b), ID(0, &instr_dual_0f_2b), N, N,
4226 };
4227 
4228 static const struct gprefix pfx_0f_28_0f_29 = {
4229 	I(Aligned, em_mov), I(Aligned, em_mov), N, N,
4230 };
4231 
4232 static const struct gprefix pfx_0f_e7 = {
4233 	N, I(Sse, em_mov), N, N,
4234 };
4235 
4236 static const struct escape escape_d9 = { {
4237 	N, N, N, N, N, N, N, I(DstMem16 | Mov, em_fnstcw),
4238 }, {
4239 	/* 0xC0 - 0xC7 */
4240 	N, N, N, N, N, N, N, N,
4241 	/* 0xC8 - 0xCF */
4242 	N, N, N, N, N, N, N, N,
4243 	/* 0xD0 - 0xC7 */
4244 	N, N, N, N, N, N, N, N,
4245 	/* 0xD8 - 0xDF */
4246 	N, N, N, N, N, N, N, N,
4247 	/* 0xE0 - 0xE7 */
4248 	N, N, N, N, N, N, N, N,
4249 	/* 0xE8 - 0xEF */
4250 	N, N, N, N, N, N, N, N,
4251 	/* 0xF0 - 0xF7 */
4252 	N, N, N, N, N, N, N, N,
4253 	/* 0xF8 - 0xFF */
4254 	N, N, N, N, N, N, N, N,
4255 } };
4256 
4257 static const struct escape escape_db = { {
4258 	N, N, N, N, N, N, N, N,
4259 }, {
4260 	/* 0xC0 - 0xC7 */
4261 	N, N, N, N, N, N, N, N,
4262 	/* 0xC8 - 0xCF */
4263 	N, N, N, N, N, N, N, N,
4264 	/* 0xD0 - 0xC7 */
4265 	N, N, N, N, N, N, N, N,
4266 	/* 0xD8 - 0xDF */
4267 	N, N, N, N, N, N, N, N,
4268 	/* 0xE0 - 0xE7 */
4269 	N, N, N, I(ImplicitOps, em_fninit), N, N, N, N,
4270 	/* 0xE8 - 0xEF */
4271 	N, N, N, N, N, N, N, N,
4272 	/* 0xF0 - 0xF7 */
4273 	N, N, N, N, N, N, N, N,
4274 	/* 0xF8 - 0xFF */
4275 	N, N, N, N, N, N, N, N,
4276 } };
4277 
4278 static const struct escape escape_dd = { {
4279 	N, N, N, N, N, N, N, I(DstMem16 | Mov, em_fnstsw),
4280 }, {
4281 	/* 0xC0 - 0xC7 */
4282 	N, N, N, N, N, N, N, N,
4283 	/* 0xC8 - 0xCF */
4284 	N, N, N, N, N, N, N, N,
4285 	/* 0xD0 - 0xC7 */
4286 	N, N, N, N, N, N, N, N,
4287 	/* 0xD8 - 0xDF */
4288 	N, N, N, N, N, N, N, N,
4289 	/* 0xE0 - 0xE7 */
4290 	N, N, N, N, N, N, N, N,
4291 	/* 0xE8 - 0xEF */
4292 	N, N, N, N, N, N, N, N,
4293 	/* 0xF0 - 0xF7 */
4294 	N, N, N, N, N, N, N, N,
4295 	/* 0xF8 - 0xFF */
4296 	N, N, N, N, N, N, N, N,
4297 } };
4298 
4299 static const struct instr_dual instr_dual_0f_c3 = {
4300 	I(DstMem | SrcReg | ModRM | No16 | Mov, em_mov), N
4301 };
4302 
4303 static const struct mode_dual mode_dual_63 = {
4304 	N, I(DstReg | SrcMem32 | ModRM | Mov, em_movsxd)
4305 };
4306 
4307 static const struct opcode opcode_table[256] = {
4308 	/* 0x00 - 0x07 */
4309 	F6ALU(Lock, em_add),
4310 	I(ImplicitOps | Stack | No64 | Src2ES, em_push_sreg),
4311 	I(ImplicitOps | Stack | No64 | Src2ES, em_pop_sreg),
4312 	/* 0x08 - 0x0F */
4313 	F6ALU(Lock | PageTable, em_or),
4314 	I(ImplicitOps | Stack | No64 | Src2CS, em_push_sreg),
4315 	N,
4316 	/* 0x10 - 0x17 */
4317 	F6ALU(Lock, em_adc),
4318 	I(ImplicitOps | Stack | No64 | Src2SS, em_push_sreg),
4319 	I(ImplicitOps | Stack | No64 | Src2SS, em_pop_sreg),
4320 	/* 0x18 - 0x1F */
4321 	F6ALU(Lock, em_sbb),
4322 	I(ImplicitOps | Stack | No64 | Src2DS, em_push_sreg),
4323 	I(ImplicitOps | Stack | No64 | Src2DS, em_pop_sreg),
4324 	/* 0x20 - 0x27 */
4325 	F6ALU(Lock | PageTable, em_and), N, N,
4326 	/* 0x28 - 0x2F */
4327 	F6ALU(Lock, em_sub), N, I(ByteOp | DstAcc | No64, em_das),
4328 	/* 0x30 - 0x37 */
4329 	F6ALU(Lock, em_xor), N, N,
4330 	/* 0x38 - 0x3F */
4331 	F6ALU(NoWrite, em_cmp), N, N,
4332 	/* 0x40 - 0x4F */
4333 	X8(F(DstReg, em_inc)), X8(F(DstReg, em_dec)),
4334 	/* 0x50 - 0x57 */
4335 	X8(I(SrcReg | Stack, em_push)),
4336 	/* 0x58 - 0x5F */
4337 	X8(I(DstReg | Stack, em_pop)),
4338 	/* 0x60 - 0x67 */
4339 	I(ImplicitOps | Stack | No64, em_pusha),
4340 	I(ImplicitOps | Stack | No64, em_popa),
4341 	N, MD(ModRM, &mode_dual_63),
4342 	N, N, N, N,
4343 	/* 0x68 - 0x6F */
4344 	I(SrcImm | Mov | Stack, em_push),
4345 	I(DstReg | SrcMem | ModRM | Src2Imm, em_imul_3op),
4346 	I(SrcImmByte | Mov | Stack, em_push),
4347 	I(DstReg | SrcMem | ModRM | Src2ImmByte, em_imul_3op),
4348 	I2bvIP(DstDI | SrcDX | Mov | String | Unaligned, em_in, ins, check_perm_in), /* insb, insw/insd */
4349 	I2bvIP(SrcSI | DstDX | String, em_out, outs, check_perm_out), /* outsb, outsw/outsd */
4350 	/* 0x70 - 0x7F */
4351 	X16(D(SrcImmByte | NearBranch)),
4352 	/* 0x80 - 0x87 */
4353 	G(ByteOp | DstMem | SrcImm, group1),
4354 	G(DstMem | SrcImm, group1),
4355 	G(ByteOp | DstMem | SrcImm | No64, group1),
4356 	G(DstMem | SrcImmByte, group1),
4357 	F2bv(DstMem | SrcReg | ModRM | NoWrite, em_test),
4358 	I2bv(DstMem | SrcReg | ModRM | Lock | PageTable, em_xchg),
4359 	/* 0x88 - 0x8F */
4360 	I2bv(DstMem | SrcReg | ModRM | Mov | PageTable, em_mov),
4361 	I2bv(DstReg | SrcMem | ModRM | Mov, em_mov),
4362 	I(DstMem | SrcNone | ModRM | Mov | PageTable, em_mov_rm_sreg),
4363 	D(ModRM | SrcMem | NoAccess | DstReg),
4364 	I(ImplicitOps | SrcMem16 | ModRM, em_mov_sreg_rm),
4365 	G(0, group1A),
4366 	/* 0x90 - 0x97 */
4367 	DI(SrcAcc | DstReg, pause), X7(D(SrcAcc | DstReg)),
4368 	/* 0x98 - 0x9F */
4369 	D(DstAcc | SrcNone), I(ImplicitOps | SrcAcc, em_cwd),
4370 	I(SrcImmFAddr | No64, em_call_far), N,
4371 	II(ImplicitOps | Stack, em_pushf, pushf),
4372 	II(ImplicitOps | Stack, em_popf, popf),
4373 	I(ImplicitOps, em_sahf), I(ImplicitOps, em_lahf),
4374 	/* 0xA0 - 0xA7 */
4375 	I2bv(DstAcc | SrcMem | Mov | MemAbs, em_mov),
4376 	I2bv(DstMem | SrcAcc | Mov | MemAbs | PageTable, em_mov),
4377 	I2bv(SrcSI | DstDI | Mov | String, em_mov),
4378 	F2bv(SrcSI | DstDI | String | NoWrite, em_cmp_r),
4379 	/* 0xA8 - 0xAF */
4380 	F2bv(DstAcc | SrcImm | NoWrite, em_test),
4381 	I2bv(SrcAcc | DstDI | Mov | String, em_mov),
4382 	I2bv(SrcSI | DstAcc | Mov | String, em_mov),
4383 	F2bv(SrcAcc | DstDI | String | NoWrite, em_cmp_r),
4384 	/* 0xB0 - 0xB7 */
4385 	X8(I(ByteOp | DstReg | SrcImm | Mov, em_mov)),
4386 	/* 0xB8 - 0xBF */
4387 	X8(I(DstReg | SrcImm64 | Mov, em_mov)),
4388 	/* 0xC0 - 0xC7 */
4389 	G(ByteOp | Src2ImmByte, group2), G(Src2ImmByte, group2),
4390 	I(ImplicitOps | NearBranch | SrcImmU16, em_ret_near_imm),
4391 	I(ImplicitOps | NearBranch, em_ret),
4392 	I(DstReg | SrcMemFAddr | ModRM | No64 | Src2ES, em_lseg),
4393 	I(DstReg | SrcMemFAddr | ModRM | No64 | Src2DS, em_lseg),
4394 	G(ByteOp, group11), G(0, group11),
4395 	/* 0xC8 - 0xCF */
4396 	I(Stack | SrcImmU16 | Src2ImmByte, em_enter), I(Stack, em_leave),
4397 	I(ImplicitOps | SrcImmU16, em_ret_far_imm),
4398 	I(ImplicitOps, em_ret_far),
4399 	D(ImplicitOps), DI(SrcImmByte, intn),
4400 	D(ImplicitOps | No64), II(ImplicitOps, em_iret, iret),
4401 	/* 0xD0 - 0xD7 */
4402 	G(Src2One | ByteOp, group2), G(Src2One, group2),
4403 	G(Src2CL | ByteOp, group2), G(Src2CL, group2),
4404 	I(DstAcc | SrcImmUByte | No64, em_aam),
4405 	I(DstAcc | SrcImmUByte | No64, em_aad),
4406 	F(DstAcc | ByteOp | No64, em_salc),
4407 	I(DstAcc | SrcXLat | ByteOp, em_mov),
4408 	/* 0xD8 - 0xDF */
4409 	N, E(0, &escape_d9), N, E(0, &escape_db), N, E(0, &escape_dd), N, N,
4410 	/* 0xE0 - 0xE7 */
4411 	X3(I(SrcImmByte | NearBranch, em_loop)),
4412 	I(SrcImmByte | NearBranch, em_jcxz),
4413 	I2bvIP(SrcImmUByte | DstAcc, em_in,  in,  check_perm_in),
4414 	I2bvIP(SrcAcc | DstImmUByte, em_out, out, check_perm_out),
4415 	/* 0xE8 - 0xEF */
4416 	I(SrcImm | NearBranch, em_call), D(SrcImm | ImplicitOps | NearBranch),
4417 	I(SrcImmFAddr | No64, em_jmp_far),
4418 	D(SrcImmByte | ImplicitOps | NearBranch),
4419 	I2bvIP(SrcDX | DstAcc, em_in,  in,  check_perm_in),
4420 	I2bvIP(SrcAcc | DstDX, em_out, out, check_perm_out),
4421 	/* 0xF0 - 0xF7 */
4422 	N, DI(ImplicitOps, icebp), N, N,
4423 	DI(ImplicitOps | Priv, hlt), D(ImplicitOps),
4424 	G(ByteOp, group3), G(0, group3),
4425 	/* 0xF8 - 0xFF */
4426 	D(ImplicitOps), D(ImplicitOps),
4427 	I(ImplicitOps, em_cli), I(ImplicitOps, em_sti),
4428 	D(ImplicitOps), D(ImplicitOps), G(0, group4), G(0, group5),
4429 };
4430 
4431 static const struct opcode twobyte_table[256] = {
4432 	/* 0x00 - 0x0F */
4433 	G(0, group6), GD(0, &group7), N, N,
4434 	N, I(ImplicitOps | EmulateOnUD, em_syscall),
4435 	II(ImplicitOps | Priv, em_clts, clts), N,
4436 	DI(ImplicitOps | Priv, invd), DI(ImplicitOps | Priv, wbinvd), N, N,
4437 	N, D(ImplicitOps | ModRM | SrcMem | NoAccess), N, N,
4438 	/* 0x10 - 0x1F */
4439 	N, N, N, N, N, N, N, N,
4440 	D(ImplicitOps | ModRM | SrcMem | NoAccess),
4441 	N, N, N, N, N, N, D(ImplicitOps | ModRM | SrcMem | NoAccess),
4442 	/* 0x20 - 0x2F */
4443 	DIP(ModRM | DstMem | Priv | Op3264 | NoMod, cr_read, check_cr_read),
4444 	DIP(ModRM | DstMem | Priv | Op3264 | NoMod, dr_read, check_dr_read),
4445 	IIP(ModRM | SrcMem | Priv | Op3264 | NoMod, em_cr_write, cr_write,
4446 						check_cr_write),
4447 	IIP(ModRM | SrcMem | Priv | Op3264 | NoMod, em_dr_write, dr_write,
4448 						check_dr_write),
4449 	N, N, N, N,
4450 	GP(ModRM | DstReg | SrcMem | Mov | Sse, &pfx_0f_28_0f_29),
4451 	GP(ModRM | DstMem | SrcReg | Mov | Sse, &pfx_0f_28_0f_29),
4452 	N, GP(ModRM | DstMem | SrcReg | Mov | Sse, &pfx_0f_2b),
4453 	N, N, N, N,
4454 	/* 0x30 - 0x3F */
4455 	II(ImplicitOps | Priv, em_wrmsr, wrmsr),
4456 	IIP(ImplicitOps, em_rdtsc, rdtsc, check_rdtsc),
4457 	II(ImplicitOps | Priv, em_rdmsr, rdmsr),
4458 	IIP(ImplicitOps, em_rdpmc, rdpmc, check_rdpmc),
4459 	I(ImplicitOps | EmulateOnUD, em_sysenter),
4460 	I(ImplicitOps | Priv | EmulateOnUD, em_sysexit),
4461 	N, N,
4462 	N, N, N, N, N, N, N, N,
4463 	/* 0x40 - 0x4F */
4464 	X16(D(DstReg | SrcMem | ModRM)),
4465 	/* 0x50 - 0x5F */
4466 	N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
4467 	/* 0x60 - 0x6F */
4468 	N, N, N, N,
4469 	N, N, N, N,
4470 	N, N, N, N,
4471 	N, N, N, GP(SrcMem | DstReg | ModRM | Mov, &pfx_0f_6f_0f_7f),
4472 	/* 0x70 - 0x7F */
4473 	N, N, N, N,
4474 	N, N, N, N,
4475 	N, N, N, N,
4476 	N, N, N, GP(SrcReg | DstMem | ModRM | Mov, &pfx_0f_6f_0f_7f),
4477 	/* 0x80 - 0x8F */
4478 	X16(D(SrcImm | NearBranch)),
4479 	/* 0x90 - 0x9F */
4480 	X16(D(ByteOp | DstMem | SrcNone | ModRM| Mov)),
4481 	/* 0xA0 - 0xA7 */
4482 	I(Stack | Src2FS, em_push_sreg), I(Stack | Src2FS, em_pop_sreg),
4483 	II(ImplicitOps, em_cpuid, cpuid),
4484 	F(DstMem | SrcReg | ModRM | BitOp | NoWrite, em_bt),
4485 	F(DstMem | SrcReg | Src2ImmByte | ModRM, em_shld),
4486 	F(DstMem | SrcReg | Src2CL | ModRM, em_shld), N, N,
4487 	/* 0xA8 - 0xAF */
4488 	I(Stack | Src2GS, em_push_sreg), I(Stack | Src2GS, em_pop_sreg),
4489 	II(EmulateOnUD | ImplicitOps, em_rsm, rsm),
4490 	F(DstMem | SrcReg | ModRM | BitOp | Lock | PageTable, em_bts),
4491 	F(DstMem | SrcReg | Src2ImmByte | ModRM, em_shrd),
4492 	F(DstMem | SrcReg | Src2CL | ModRM, em_shrd),
4493 	GD(0, &group15), F(DstReg | SrcMem | ModRM, em_imul),
4494 	/* 0xB0 - 0xB7 */
4495 	I2bv(DstMem | SrcReg | ModRM | Lock | PageTable | SrcWrite, em_cmpxchg),
4496 	I(DstReg | SrcMemFAddr | ModRM | Src2SS, em_lseg),
4497 	F(DstMem | SrcReg | ModRM | BitOp | Lock, em_btr),
4498 	I(DstReg | SrcMemFAddr | ModRM | Src2FS, em_lseg),
4499 	I(DstReg | SrcMemFAddr | ModRM | Src2GS, em_lseg),
4500 	D(DstReg | SrcMem8 | ModRM | Mov), D(DstReg | SrcMem16 | ModRM | Mov),
4501 	/* 0xB8 - 0xBF */
4502 	N, N,
4503 	G(BitOp, group8),
4504 	F(DstMem | SrcReg | ModRM | BitOp | Lock | PageTable, em_btc),
4505 	I(DstReg | SrcMem | ModRM, em_bsf_c),
4506 	I(DstReg | SrcMem | ModRM, em_bsr_c),
4507 	D(DstReg | SrcMem8 | ModRM | Mov), D(DstReg | SrcMem16 | ModRM | Mov),
4508 	/* 0xC0 - 0xC7 */
4509 	F2bv(DstMem | SrcReg | ModRM | SrcWrite | Lock, em_xadd),
4510 	N, ID(0, &instr_dual_0f_c3),
4511 	N, N, N, GD(0, &group9),
4512 	/* 0xC8 - 0xCF */
4513 	X8(I(DstReg, em_bswap)),
4514 	/* 0xD0 - 0xDF */
4515 	N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
4516 	/* 0xE0 - 0xEF */
4517 	N, N, N, N, N, N, N, GP(SrcReg | DstMem | ModRM | Mov, &pfx_0f_e7),
4518 	N, N, N, N, N, N, N, N,
4519 	/* 0xF0 - 0xFF */
4520 	N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N
4521 };
4522 
4523 static const struct instr_dual instr_dual_0f_38_f0 = {
4524 	I(DstReg | SrcMem | Mov, em_movbe), N
4525 };
4526 
4527 static const struct instr_dual instr_dual_0f_38_f1 = {
4528 	I(DstMem | SrcReg | Mov, em_movbe), N
4529 };
4530 
4531 static const struct gprefix three_byte_0f_38_f0 = {
4532 	ID(0, &instr_dual_0f_38_f0), N, N, N
4533 };
4534 
4535 static const struct gprefix three_byte_0f_38_f1 = {
4536 	ID(0, &instr_dual_0f_38_f1), N, N, N
4537 };
4538 
4539 /*
4540  * Insns below are selected by the prefix which indexed by the third opcode
4541  * byte.
4542  */
4543 static const struct opcode opcode_map_0f_38[256] = {
4544 	/* 0x00 - 0x7f */
4545 	X16(N), X16(N), X16(N), X16(N), X16(N), X16(N), X16(N), X16(N),
4546 	/* 0x80 - 0xef */
4547 	X16(N), X16(N), X16(N), X16(N), X16(N), X16(N), X16(N),
4548 	/* 0xf0 - 0xf1 */
4549 	GP(EmulateOnUD | ModRM, &three_byte_0f_38_f0),
4550 	GP(EmulateOnUD | ModRM, &three_byte_0f_38_f1),
4551 	/* 0xf2 - 0xff */
4552 	N, N, X4(N), X8(N)
4553 };
4554 
4555 #undef D
4556 #undef N
4557 #undef G
4558 #undef GD
4559 #undef I
4560 #undef GP
4561 #undef EXT
4562 #undef MD
4563 #undef ID
4564 
4565 #undef D2bv
4566 #undef D2bvIP
4567 #undef I2bv
4568 #undef I2bvIP
4569 #undef I6ALU
4570 
4571 static unsigned imm_size(struct x86_emulate_ctxt *ctxt)
4572 {
4573 	unsigned size;
4574 
4575 	size = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4576 	if (size == 8)
4577 		size = 4;
4578 	return size;
4579 }
4580 
4581 static int decode_imm(struct x86_emulate_ctxt *ctxt, struct operand *op,
4582 		      unsigned size, bool sign_extension)
4583 {
4584 	int rc = X86EMUL_CONTINUE;
4585 
4586 	op->type = OP_IMM;
4587 	op->bytes = size;
4588 	op->addr.mem.ea = ctxt->_eip;
4589 	/* NB. Immediates are sign-extended as necessary. */
4590 	switch (op->bytes) {
4591 	case 1:
4592 		op->val = insn_fetch(s8, ctxt);
4593 		break;
4594 	case 2:
4595 		op->val = insn_fetch(s16, ctxt);
4596 		break;
4597 	case 4:
4598 		op->val = insn_fetch(s32, ctxt);
4599 		break;
4600 	case 8:
4601 		op->val = insn_fetch(s64, ctxt);
4602 		break;
4603 	}
4604 	if (!sign_extension) {
4605 		switch (op->bytes) {
4606 		case 1:
4607 			op->val &= 0xff;
4608 			break;
4609 		case 2:
4610 			op->val &= 0xffff;
4611 			break;
4612 		case 4:
4613 			op->val &= 0xffffffff;
4614 			break;
4615 		}
4616 	}
4617 done:
4618 	return rc;
4619 }
4620 
4621 static int decode_operand(struct x86_emulate_ctxt *ctxt, struct operand *op,
4622 			  unsigned d)
4623 {
4624 	int rc = X86EMUL_CONTINUE;
4625 
4626 	switch (d) {
4627 	case OpReg:
4628 		decode_register_operand(ctxt, op);
4629 		break;
4630 	case OpImmUByte:
4631 		rc = decode_imm(ctxt, op, 1, false);
4632 		break;
4633 	case OpMem:
4634 		ctxt->memop.bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4635 	mem_common:
4636 		*op = ctxt->memop;
4637 		ctxt->memopp = op;
4638 		if (ctxt->d & BitOp)
4639 			fetch_bit_operand(ctxt);
4640 		op->orig_val = op->val;
4641 		break;
4642 	case OpMem64:
4643 		ctxt->memop.bytes = (ctxt->op_bytes == 8) ? 16 : 8;
4644 		goto mem_common;
4645 	case OpAcc:
4646 		op->type = OP_REG;
4647 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4648 		op->addr.reg = reg_rmw(ctxt, VCPU_REGS_RAX);
4649 		fetch_register_operand(op);
4650 		op->orig_val = op->val;
4651 		break;
4652 	case OpAccLo:
4653 		op->type = OP_REG;
4654 		op->bytes = (ctxt->d & ByteOp) ? 2 : ctxt->op_bytes;
4655 		op->addr.reg = reg_rmw(ctxt, VCPU_REGS_RAX);
4656 		fetch_register_operand(op);
4657 		op->orig_val = op->val;
4658 		break;
4659 	case OpAccHi:
4660 		if (ctxt->d & ByteOp) {
4661 			op->type = OP_NONE;
4662 			break;
4663 		}
4664 		op->type = OP_REG;
4665 		op->bytes = ctxt->op_bytes;
4666 		op->addr.reg = reg_rmw(ctxt, VCPU_REGS_RDX);
4667 		fetch_register_operand(op);
4668 		op->orig_val = op->val;
4669 		break;
4670 	case OpDI:
4671 		op->type = OP_MEM;
4672 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4673 		op->addr.mem.ea =
4674 			register_address(ctxt, VCPU_REGS_RDI);
4675 		op->addr.mem.seg = VCPU_SREG_ES;
4676 		op->val = 0;
4677 		op->count = 1;
4678 		break;
4679 	case OpDX:
4680 		op->type = OP_REG;
4681 		op->bytes = 2;
4682 		op->addr.reg = reg_rmw(ctxt, VCPU_REGS_RDX);
4683 		fetch_register_operand(op);
4684 		break;
4685 	case OpCL:
4686 		op->type = OP_IMM;
4687 		op->bytes = 1;
4688 		op->val = reg_read(ctxt, VCPU_REGS_RCX) & 0xff;
4689 		break;
4690 	case OpImmByte:
4691 		rc = decode_imm(ctxt, op, 1, true);
4692 		break;
4693 	case OpOne:
4694 		op->type = OP_IMM;
4695 		op->bytes = 1;
4696 		op->val = 1;
4697 		break;
4698 	case OpImm:
4699 		rc = decode_imm(ctxt, op, imm_size(ctxt), true);
4700 		break;
4701 	case OpImm64:
4702 		rc = decode_imm(ctxt, op, ctxt->op_bytes, true);
4703 		break;
4704 	case OpMem8:
4705 		ctxt->memop.bytes = 1;
4706 		if (ctxt->memop.type == OP_REG) {
4707 			ctxt->memop.addr.reg = decode_register(ctxt,
4708 					ctxt->modrm_rm, true);
4709 			fetch_register_operand(&ctxt->memop);
4710 		}
4711 		goto mem_common;
4712 	case OpMem16:
4713 		ctxt->memop.bytes = 2;
4714 		goto mem_common;
4715 	case OpMem32:
4716 		ctxt->memop.bytes = 4;
4717 		goto mem_common;
4718 	case OpImmU16:
4719 		rc = decode_imm(ctxt, op, 2, false);
4720 		break;
4721 	case OpImmU:
4722 		rc = decode_imm(ctxt, op, imm_size(ctxt), false);
4723 		break;
4724 	case OpSI:
4725 		op->type = OP_MEM;
4726 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4727 		op->addr.mem.ea =
4728 			register_address(ctxt, VCPU_REGS_RSI);
4729 		op->addr.mem.seg = ctxt->seg_override;
4730 		op->val = 0;
4731 		op->count = 1;
4732 		break;
4733 	case OpXLat:
4734 		op->type = OP_MEM;
4735 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4736 		op->addr.mem.ea =
4737 			address_mask(ctxt,
4738 				reg_read(ctxt, VCPU_REGS_RBX) +
4739 				(reg_read(ctxt, VCPU_REGS_RAX) & 0xff));
4740 		op->addr.mem.seg = ctxt->seg_override;
4741 		op->val = 0;
4742 		break;
4743 	case OpImmFAddr:
4744 		op->type = OP_IMM;
4745 		op->addr.mem.ea = ctxt->_eip;
4746 		op->bytes = ctxt->op_bytes + 2;
4747 		insn_fetch_arr(op->valptr, op->bytes, ctxt);
4748 		break;
4749 	case OpMemFAddr:
4750 		ctxt->memop.bytes = ctxt->op_bytes + 2;
4751 		goto mem_common;
4752 	case OpES:
4753 		op->type = OP_IMM;
4754 		op->val = VCPU_SREG_ES;
4755 		break;
4756 	case OpCS:
4757 		op->type = OP_IMM;
4758 		op->val = VCPU_SREG_CS;
4759 		break;
4760 	case OpSS:
4761 		op->type = OP_IMM;
4762 		op->val = VCPU_SREG_SS;
4763 		break;
4764 	case OpDS:
4765 		op->type = OP_IMM;
4766 		op->val = VCPU_SREG_DS;
4767 		break;
4768 	case OpFS:
4769 		op->type = OP_IMM;
4770 		op->val = VCPU_SREG_FS;
4771 		break;
4772 	case OpGS:
4773 		op->type = OP_IMM;
4774 		op->val = VCPU_SREG_GS;
4775 		break;
4776 	case OpImplicit:
4777 		/* Special instructions do their own operand decoding. */
4778 	default:
4779 		op->type = OP_NONE; /* Disable writeback. */
4780 		break;
4781 	}
4782 
4783 done:
4784 	return rc;
4785 }
4786 
4787 int x86_decode_insn(struct x86_emulate_ctxt *ctxt, void *insn, int insn_len)
4788 {
4789 	int rc = X86EMUL_CONTINUE;
4790 	int mode = ctxt->mode;
4791 	int def_op_bytes, def_ad_bytes, goffset, simd_prefix;
4792 	bool op_prefix = false;
4793 	bool has_seg_override = false;
4794 	struct opcode opcode;
4795 
4796 	ctxt->memop.type = OP_NONE;
4797 	ctxt->memopp = NULL;
4798 	ctxt->_eip = ctxt->eip;
4799 	ctxt->fetch.ptr = ctxt->fetch.data;
4800 	ctxt->fetch.end = ctxt->fetch.data + insn_len;
4801 	ctxt->opcode_len = 1;
4802 	if (insn_len > 0)
4803 		memcpy(ctxt->fetch.data, insn, insn_len);
4804 	else {
4805 		rc = __do_insn_fetch_bytes(ctxt, 1);
4806 		if (rc != X86EMUL_CONTINUE)
4807 			return rc;
4808 	}
4809 
4810 	switch (mode) {
4811 	case X86EMUL_MODE_REAL:
4812 	case X86EMUL_MODE_VM86:
4813 	case X86EMUL_MODE_PROT16:
4814 		def_op_bytes = def_ad_bytes = 2;
4815 		break;
4816 	case X86EMUL_MODE_PROT32:
4817 		def_op_bytes = def_ad_bytes = 4;
4818 		break;
4819 #ifdef CONFIG_X86_64
4820 	case X86EMUL_MODE_PROT64:
4821 		def_op_bytes = 4;
4822 		def_ad_bytes = 8;
4823 		break;
4824 #endif
4825 	default:
4826 		return EMULATION_FAILED;
4827 	}
4828 
4829 	ctxt->op_bytes = def_op_bytes;
4830 	ctxt->ad_bytes = def_ad_bytes;
4831 
4832 	/* Legacy prefixes. */
4833 	for (;;) {
4834 		switch (ctxt->b = insn_fetch(u8, ctxt)) {
4835 		case 0x66:	/* operand-size override */
4836 			op_prefix = true;
4837 			/* switch between 2/4 bytes */
4838 			ctxt->op_bytes = def_op_bytes ^ 6;
4839 			break;
4840 		case 0x67:	/* address-size override */
4841 			if (mode == X86EMUL_MODE_PROT64)
4842 				/* switch between 4/8 bytes */
4843 				ctxt->ad_bytes = def_ad_bytes ^ 12;
4844 			else
4845 				/* switch between 2/4 bytes */
4846 				ctxt->ad_bytes = def_ad_bytes ^ 6;
4847 			break;
4848 		case 0x26:	/* ES override */
4849 		case 0x2e:	/* CS override */
4850 		case 0x36:	/* SS override */
4851 		case 0x3e:	/* DS override */
4852 			has_seg_override = true;
4853 			ctxt->seg_override = (ctxt->b >> 3) & 3;
4854 			break;
4855 		case 0x64:	/* FS override */
4856 		case 0x65:	/* GS override */
4857 			has_seg_override = true;
4858 			ctxt->seg_override = ctxt->b & 7;
4859 			break;
4860 		case 0x40 ... 0x4f: /* REX */
4861 			if (mode != X86EMUL_MODE_PROT64)
4862 				goto done_prefixes;
4863 			ctxt->rex_prefix = ctxt->b;
4864 			continue;
4865 		case 0xf0:	/* LOCK */
4866 			ctxt->lock_prefix = 1;
4867 			break;
4868 		case 0xf2:	/* REPNE/REPNZ */
4869 		case 0xf3:	/* REP/REPE/REPZ */
4870 			ctxt->rep_prefix = ctxt->b;
4871 			break;
4872 		default:
4873 			goto done_prefixes;
4874 		}
4875 
4876 		/* Any legacy prefix after a REX prefix nullifies its effect. */
4877 
4878 		ctxt->rex_prefix = 0;
4879 	}
4880 
4881 done_prefixes:
4882 
4883 	/* REX prefix. */
4884 	if (ctxt->rex_prefix & 8)
4885 		ctxt->op_bytes = 8;	/* REX.W */
4886 
4887 	/* Opcode byte(s). */
4888 	opcode = opcode_table[ctxt->b];
4889 	/* Two-byte opcode? */
4890 	if (ctxt->b == 0x0f) {
4891 		ctxt->opcode_len = 2;
4892 		ctxt->b = insn_fetch(u8, ctxt);
4893 		opcode = twobyte_table[ctxt->b];
4894 
4895 		/* 0F_38 opcode map */
4896 		if (ctxt->b == 0x38) {
4897 			ctxt->opcode_len = 3;
4898 			ctxt->b = insn_fetch(u8, ctxt);
4899 			opcode = opcode_map_0f_38[ctxt->b];
4900 		}
4901 	}
4902 	ctxt->d = opcode.flags;
4903 
4904 	if (ctxt->d & ModRM)
4905 		ctxt->modrm = insn_fetch(u8, ctxt);
4906 
4907 	/* vex-prefix instructions are not implemented */
4908 	if (ctxt->opcode_len == 1 && (ctxt->b == 0xc5 || ctxt->b == 0xc4) &&
4909 	    (mode == X86EMUL_MODE_PROT64 || (ctxt->modrm & 0xc0) == 0xc0)) {
4910 		ctxt->d = NotImpl;
4911 	}
4912 
4913 	while (ctxt->d & GroupMask) {
4914 		switch (ctxt->d & GroupMask) {
4915 		case Group:
4916 			goffset = (ctxt->modrm >> 3) & 7;
4917 			opcode = opcode.u.group[goffset];
4918 			break;
4919 		case GroupDual:
4920 			goffset = (ctxt->modrm >> 3) & 7;
4921 			if ((ctxt->modrm >> 6) == 3)
4922 				opcode = opcode.u.gdual->mod3[goffset];
4923 			else
4924 				opcode = opcode.u.gdual->mod012[goffset];
4925 			break;
4926 		case RMExt:
4927 			goffset = ctxt->modrm & 7;
4928 			opcode = opcode.u.group[goffset];
4929 			break;
4930 		case Prefix:
4931 			if (ctxt->rep_prefix && op_prefix)
4932 				return EMULATION_FAILED;
4933 			simd_prefix = op_prefix ? 0x66 : ctxt->rep_prefix;
4934 			switch (simd_prefix) {
4935 			case 0x00: opcode = opcode.u.gprefix->pfx_no; break;
4936 			case 0x66: opcode = opcode.u.gprefix->pfx_66; break;
4937 			case 0xf2: opcode = opcode.u.gprefix->pfx_f2; break;
4938 			case 0xf3: opcode = opcode.u.gprefix->pfx_f3; break;
4939 			}
4940 			break;
4941 		case Escape:
4942 			if (ctxt->modrm > 0xbf)
4943 				opcode = opcode.u.esc->high[ctxt->modrm - 0xc0];
4944 			else
4945 				opcode = opcode.u.esc->op[(ctxt->modrm >> 3) & 7];
4946 			break;
4947 		case InstrDual:
4948 			if ((ctxt->modrm >> 6) == 3)
4949 				opcode = opcode.u.idual->mod3;
4950 			else
4951 				opcode = opcode.u.idual->mod012;
4952 			break;
4953 		case ModeDual:
4954 			if (ctxt->mode == X86EMUL_MODE_PROT64)
4955 				opcode = opcode.u.mdual->mode64;
4956 			else
4957 				opcode = opcode.u.mdual->mode32;
4958 			break;
4959 		default:
4960 			return EMULATION_FAILED;
4961 		}
4962 
4963 		ctxt->d &= ~(u64)GroupMask;
4964 		ctxt->d |= opcode.flags;
4965 	}
4966 
4967 	/* Unrecognised? */
4968 	if (ctxt->d == 0)
4969 		return EMULATION_FAILED;
4970 
4971 	ctxt->execute = opcode.u.execute;
4972 
4973 	if (unlikely(ctxt->ud) && likely(!(ctxt->d & EmulateOnUD)))
4974 		return EMULATION_FAILED;
4975 
4976 	if (unlikely(ctxt->d &
4977 	    (NotImpl|Stack|Op3264|Sse|Mmx|Intercept|CheckPerm|NearBranch|
4978 	     No16))) {
4979 		/*
4980 		 * These are copied unconditionally here, and checked unconditionally
4981 		 * in x86_emulate_insn.
4982 		 */
4983 		ctxt->check_perm = opcode.check_perm;
4984 		ctxt->intercept = opcode.intercept;
4985 
4986 		if (ctxt->d & NotImpl)
4987 			return EMULATION_FAILED;
4988 
4989 		if (mode == X86EMUL_MODE_PROT64) {
4990 			if (ctxt->op_bytes == 4 && (ctxt->d & Stack))
4991 				ctxt->op_bytes = 8;
4992 			else if (ctxt->d & NearBranch)
4993 				ctxt->op_bytes = 8;
4994 		}
4995 
4996 		if (ctxt->d & Op3264) {
4997 			if (mode == X86EMUL_MODE_PROT64)
4998 				ctxt->op_bytes = 8;
4999 			else
5000 				ctxt->op_bytes = 4;
5001 		}
5002 
5003 		if ((ctxt->d & No16) && ctxt->op_bytes == 2)
5004 			ctxt->op_bytes = 4;
5005 
5006 		if (ctxt->d & Sse)
5007 			ctxt->op_bytes = 16;
5008 		else if (ctxt->d & Mmx)
5009 			ctxt->op_bytes = 8;
5010 	}
5011 
5012 	/* ModRM and SIB bytes. */
5013 	if (ctxt->d & ModRM) {
5014 		rc = decode_modrm(ctxt, &ctxt->memop);
5015 		if (!has_seg_override) {
5016 			has_seg_override = true;
5017 			ctxt->seg_override = ctxt->modrm_seg;
5018 		}
5019 	} else if (ctxt->d & MemAbs)
5020 		rc = decode_abs(ctxt, &ctxt->memop);
5021 	if (rc != X86EMUL_CONTINUE)
5022 		goto done;
5023 
5024 	if (!has_seg_override)
5025 		ctxt->seg_override = VCPU_SREG_DS;
5026 
5027 	ctxt->memop.addr.mem.seg = ctxt->seg_override;
5028 
5029 	/*
5030 	 * Decode and fetch the source operand: register, memory
5031 	 * or immediate.
5032 	 */
5033 	rc = decode_operand(ctxt, &ctxt->src, (ctxt->d >> SrcShift) & OpMask);
5034 	if (rc != X86EMUL_CONTINUE)
5035 		goto done;
5036 
5037 	/*
5038 	 * Decode and fetch the second source operand: register, memory
5039 	 * or immediate.
5040 	 */
5041 	rc = decode_operand(ctxt, &ctxt->src2, (ctxt->d >> Src2Shift) & OpMask);
5042 	if (rc != X86EMUL_CONTINUE)
5043 		goto done;
5044 
5045 	/* Decode and fetch the destination operand: register or memory. */
5046 	rc = decode_operand(ctxt, &ctxt->dst, (ctxt->d >> DstShift) & OpMask);
5047 
5048 	if (ctxt->rip_relative)
5049 		ctxt->memopp->addr.mem.ea = address_mask(ctxt,
5050 					ctxt->memopp->addr.mem.ea + ctxt->_eip);
5051 
5052 done:
5053 	return (rc != X86EMUL_CONTINUE) ? EMULATION_FAILED : EMULATION_OK;
5054 }
5055 
5056 bool x86_page_table_writing_insn(struct x86_emulate_ctxt *ctxt)
5057 {
5058 	return ctxt->d & PageTable;
5059 }
5060 
5061 static bool string_insn_completed(struct x86_emulate_ctxt *ctxt)
5062 {
5063 	/* The second termination condition only applies for REPE
5064 	 * and REPNE. Test if the repeat string operation prefix is
5065 	 * REPE/REPZ or REPNE/REPNZ and if it's the case it tests the
5066 	 * corresponding termination condition according to:
5067 	 * 	- if REPE/REPZ and ZF = 0 then done
5068 	 * 	- if REPNE/REPNZ and ZF = 1 then done
5069 	 */
5070 	if (((ctxt->b == 0xa6) || (ctxt->b == 0xa7) ||
5071 	     (ctxt->b == 0xae) || (ctxt->b == 0xaf))
5072 	    && (((ctxt->rep_prefix == REPE_PREFIX) &&
5073 		 ((ctxt->eflags & X86_EFLAGS_ZF) == 0))
5074 		|| ((ctxt->rep_prefix == REPNE_PREFIX) &&
5075 		    ((ctxt->eflags & X86_EFLAGS_ZF) == X86_EFLAGS_ZF))))
5076 		return true;
5077 
5078 	return false;
5079 }
5080 
5081 static int flush_pending_x87_faults(struct x86_emulate_ctxt *ctxt)
5082 {
5083 	bool fault = false;
5084 
5085 	ctxt->ops->get_fpu(ctxt);
5086 	asm volatile("1: fwait \n\t"
5087 		     "2: \n\t"
5088 		     ".pushsection .fixup,\"ax\" \n\t"
5089 		     "3: \n\t"
5090 		     "movb $1, %[fault] \n\t"
5091 		     "jmp 2b \n\t"
5092 		     ".popsection \n\t"
5093 		     _ASM_EXTABLE(1b, 3b)
5094 		     : [fault]"+qm"(fault));
5095 	ctxt->ops->put_fpu(ctxt);
5096 
5097 	if (unlikely(fault))
5098 		return emulate_exception(ctxt, MF_VECTOR, 0, false);
5099 
5100 	return X86EMUL_CONTINUE;
5101 }
5102 
5103 static void fetch_possible_mmx_operand(struct x86_emulate_ctxt *ctxt,
5104 				       struct operand *op)
5105 {
5106 	if (op->type == OP_MM)
5107 		read_mmx_reg(ctxt, &op->mm_val, op->addr.mm);
5108 }
5109 
5110 static int fastop(struct x86_emulate_ctxt *ctxt, void (*fop)(struct fastop *))
5111 {
5112 	register void *__sp asm(_ASM_SP);
5113 	ulong flags = (ctxt->eflags & EFLAGS_MASK) | X86_EFLAGS_IF;
5114 
5115 	if (!(ctxt->d & ByteOp))
5116 		fop += __ffs(ctxt->dst.bytes) * FASTOP_SIZE;
5117 
5118 	asm("push %[flags]; popf; call *%[fastop]; pushf; pop %[flags]\n"
5119 	    : "+a"(ctxt->dst.val), "+d"(ctxt->src.val), [flags]"+D"(flags),
5120 	      [fastop]"+S"(fop), "+r"(__sp)
5121 	    : "c"(ctxt->src2.val));
5122 
5123 	ctxt->eflags = (ctxt->eflags & ~EFLAGS_MASK) | (flags & EFLAGS_MASK);
5124 	if (!fop) /* exception is returned in fop variable */
5125 		return emulate_de(ctxt);
5126 	return X86EMUL_CONTINUE;
5127 }
5128 
5129 void init_decode_cache(struct x86_emulate_ctxt *ctxt)
5130 {
5131 	memset(&ctxt->rip_relative, 0,
5132 	       (void *)&ctxt->modrm - (void *)&ctxt->rip_relative);
5133 
5134 	ctxt->io_read.pos = 0;
5135 	ctxt->io_read.end = 0;
5136 	ctxt->mem_read.end = 0;
5137 }
5138 
5139 int x86_emulate_insn(struct x86_emulate_ctxt *ctxt)
5140 {
5141 	const struct x86_emulate_ops *ops = ctxt->ops;
5142 	int rc = X86EMUL_CONTINUE;
5143 	int saved_dst_type = ctxt->dst.type;
5144 
5145 	ctxt->mem_read.pos = 0;
5146 
5147 	/* LOCK prefix is allowed only with some instructions */
5148 	if (ctxt->lock_prefix && (!(ctxt->d & Lock) || ctxt->dst.type != OP_MEM)) {
5149 		rc = emulate_ud(ctxt);
5150 		goto done;
5151 	}
5152 
5153 	if ((ctxt->d & SrcMask) == SrcMemFAddr && ctxt->src.type != OP_MEM) {
5154 		rc = emulate_ud(ctxt);
5155 		goto done;
5156 	}
5157 
5158 	if (unlikely(ctxt->d &
5159 		     (No64|Undefined|Sse|Mmx|Intercept|CheckPerm|Priv|Prot|String))) {
5160 		if ((ctxt->mode == X86EMUL_MODE_PROT64 && (ctxt->d & No64)) ||
5161 				(ctxt->d & Undefined)) {
5162 			rc = emulate_ud(ctxt);
5163 			goto done;
5164 		}
5165 
5166 		if (((ctxt->d & (Sse|Mmx)) && ((ops->get_cr(ctxt, 0) & X86_CR0_EM)))
5167 		    || ((ctxt->d & Sse) && !(ops->get_cr(ctxt, 4) & X86_CR4_OSFXSR))) {
5168 			rc = emulate_ud(ctxt);
5169 			goto done;
5170 		}
5171 
5172 		if ((ctxt->d & (Sse|Mmx)) && (ops->get_cr(ctxt, 0) & X86_CR0_TS)) {
5173 			rc = emulate_nm(ctxt);
5174 			goto done;
5175 		}
5176 
5177 		if (ctxt->d & Mmx) {
5178 			rc = flush_pending_x87_faults(ctxt);
5179 			if (rc != X86EMUL_CONTINUE)
5180 				goto done;
5181 			/*
5182 			 * Now that we know the fpu is exception safe, we can fetch
5183 			 * operands from it.
5184 			 */
5185 			fetch_possible_mmx_operand(ctxt, &ctxt->src);
5186 			fetch_possible_mmx_operand(ctxt, &ctxt->src2);
5187 			if (!(ctxt->d & Mov))
5188 				fetch_possible_mmx_operand(ctxt, &ctxt->dst);
5189 		}
5190 
5191 		if (unlikely(ctxt->emul_flags & X86EMUL_GUEST_MASK) && ctxt->intercept) {
5192 			rc = emulator_check_intercept(ctxt, ctxt->intercept,
5193 						      X86_ICPT_PRE_EXCEPT);
5194 			if (rc != X86EMUL_CONTINUE)
5195 				goto done;
5196 		}
5197 
5198 		/* Instruction can only be executed in protected mode */
5199 		if ((ctxt->d & Prot) && ctxt->mode < X86EMUL_MODE_PROT16) {
5200 			rc = emulate_ud(ctxt);
5201 			goto done;
5202 		}
5203 
5204 		/* Privileged instruction can be executed only in CPL=0 */
5205 		if ((ctxt->d & Priv) && ops->cpl(ctxt)) {
5206 			if (ctxt->d & PrivUD)
5207 				rc = emulate_ud(ctxt);
5208 			else
5209 				rc = emulate_gp(ctxt, 0);
5210 			goto done;
5211 		}
5212 
5213 		/* Do instruction specific permission checks */
5214 		if (ctxt->d & CheckPerm) {
5215 			rc = ctxt->check_perm(ctxt);
5216 			if (rc != X86EMUL_CONTINUE)
5217 				goto done;
5218 		}
5219 
5220 		if (unlikely(ctxt->emul_flags & X86EMUL_GUEST_MASK) && (ctxt->d & Intercept)) {
5221 			rc = emulator_check_intercept(ctxt, ctxt->intercept,
5222 						      X86_ICPT_POST_EXCEPT);
5223 			if (rc != X86EMUL_CONTINUE)
5224 				goto done;
5225 		}
5226 
5227 		if (ctxt->rep_prefix && (ctxt->d & String)) {
5228 			/* All REP prefixes have the same first termination condition */
5229 			if (address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) == 0) {
5230 				string_registers_quirk(ctxt);
5231 				ctxt->eip = ctxt->_eip;
5232 				ctxt->eflags &= ~X86_EFLAGS_RF;
5233 				goto done;
5234 			}
5235 		}
5236 	}
5237 
5238 	if ((ctxt->src.type == OP_MEM) && !(ctxt->d & NoAccess)) {
5239 		rc = segmented_read(ctxt, ctxt->src.addr.mem,
5240 				    ctxt->src.valptr, ctxt->src.bytes);
5241 		if (rc != X86EMUL_CONTINUE)
5242 			goto done;
5243 		ctxt->src.orig_val64 = ctxt->src.val64;
5244 	}
5245 
5246 	if (ctxt->src2.type == OP_MEM) {
5247 		rc = segmented_read(ctxt, ctxt->src2.addr.mem,
5248 				    &ctxt->src2.val, ctxt->src2.bytes);
5249 		if (rc != X86EMUL_CONTINUE)
5250 			goto done;
5251 	}
5252 
5253 	if ((ctxt->d & DstMask) == ImplicitOps)
5254 		goto special_insn;
5255 
5256 
5257 	if ((ctxt->dst.type == OP_MEM) && !(ctxt->d & Mov)) {
5258 		/* optimisation - avoid slow emulated read if Mov */
5259 		rc = segmented_read(ctxt, ctxt->dst.addr.mem,
5260 				   &ctxt->dst.val, ctxt->dst.bytes);
5261 		if (rc != X86EMUL_CONTINUE) {
5262 			if (!(ctxt->d & NoWrite) &&
5263 			    rc == X86EMUL_PROPAGATE_FAULT &&
5264 			    ctxt->exception.vector == PF_VECTOR)
5265 				ctxt->exception.error_code |= PFERR_WRITE_MASK;
5266 			goto done;
5267 		}
5268 	}
5269 	/* Copy full 64-bit value for CMPXCHG8B.  */
5270 	ctxt->dst.orig_val64 = ctxt->dst.val64;
5271 
5272 special_insn:
5273 
5274 	if (unlikely(ctxt->emul_flags & X86EMUL_GUEST_MASK) && (ctxt->d & Intercept)) {
5275 		rc = emulator_check_intercept(ctxt, ctxt->intercept,
5276 					      X86_ICPT_POST_MEMACCESS);
5277 		if (rc != X86EMUL_CONTINUE)
5278 			goto done;
5279 	}
5280 
5281 	if (ctxt->rep_prefix && (ctxt->d & String))
5282 		ctxt->eflags |= X86_EFLAGS_RF;
5283 	else
5284 		ctxt->eflags &= ~X86_EFLAGS_RF;
5285 
5286 	if (ctxt->execute) {
5287 		if (ctxt->d & Fastop) {
5288 			void (*fop)(struct fastop *) = (void *)ctxt->execute;
5289 			rc = fastop(ctxt, fop);
5290 			if (rc != X86EMUL_CONTINUE)
5291 				goto done;
5292 			goto writeback;
5293 		}
5294 		rc = ctxt->execute(ctxt);
5295 		if (rc != X86EMUL_CONTINUE)
5296 			goto done;
5297 		goto writeback;
5298 	}
5299 
5300 	if (ctxt->opcode_len == 2)
5301 		goto twobyte_insn;
5302 	else if (ctxt->opcode_len == 3)
5303 		goto threebyte_insn;
5304 
5305 	switch (ctxt->b) {
5306 	case 0x70 ... 0x7f: /* jcc (short) */
5307 		if (test_cc(ctxt->b, ctxt->eflags))
5308 			rc = jmp_rel(ctxt, ctxt->src.val);
5309 		break;
5310 	case 0x8d: /* lea r16/r32, m */
5311 		ctxt->dst.val = ctxt->src.addr.mem.ea;
5312 		break;
5313 	case 0x90 ... 0x97: /* nop / xchg reg, rax */
5314 		if (ctxt->dst.addr.reg == reg_rmw(ctxt, VCPU_REGS_RAX))
5315 			ctxt->dst.type = OP_NONE;
5316 		else
5317 			rc = em_xchg(ctxt);
5318 		break;
5319 	case 0x98: /* cbw/cwde/cdqe */
5320 		switch (ctxt->op_bytes) {
5321 		case 2: ctxt->dst.val = (s8)ctxt->dst.val; break;
5322 		case 4: ctxt->dst.val = (s16)ctxt->dst.val; break;
5323 		case 8: ctxt->dst.val = (s32)ctxt->dst.val; break;
5324 		}
5325 		break;
5326 	case 0xcc:		/* int3 */
5327 		rc = emulate_int(ctxt, 3);
5328 		break;
5329 	case 0xcd:		/* int n */
5330 		rc = emulate_int(ctxt, ctxt->src.val);
5331 		break;
5332 	case 0xce:		/* into */
5333 		if (ctxt->eflags & X86_EFLAGS_OF)
5334 			rc = emulate_int(ctxt, 4);
5335 		break;
5336 	case 0xe9: /* jmp rel */
5337 	case 0xeb: /* jmp rel short */
5338 		rc = jmp_rel(ctxt, ctxt->src.val);
5339 		ctxt->dst.type = OP_NONE; /* Disable writeback. */
5340 		break;
5341 	case 0xf4:              /* hlt */
5342 		ctxt->ops->halt(ctxt);
5343 		break;
5344 	case 0xf5:	/* cmc */
5345 		/* complement carry flag from eflags reg */
5346 		ctxt->eflags ^= X86_EFLAGS_CF;
5347 		break;
5348 	case 0xf8: /* clc */
5349 		ctxt->eflags &= ~X86_EFLAGS_CF;
5350 		break;
5351 	case 0xf9: /* stc */
5352 		ctxt->eflags |= X86_EFLAGS_CF;
5353 		break;
5354 	case 0xfc: /* cld */
5355 		ctxt->eflags &= ~X86_EFLAGS_DF;
5356 		break;
5357 	case 0xfd: /* std */
5358 		ctxt->eflags |= X86_EFLAGS_DF;
5359 		break;
5360 	default:
5361 		goto cannot_emulate;
5362 	}
5363 
5364 	if (rc != X86EMUL_CONTINUE)
5365 		goto done;
5366 
5367 writeback:
5368 	if (ctxt->d & SrcWrite) {
5369 		BUG_ON(ctxt->src.type == OP_MEM || ctxt->src.type == OP_MEM_STR);
5370 		rc = writeback(ctxt, &ctxt->src);
5371 		if (rc != X86EMUL_CONTINUE)
5372 			goto done;
5373 	}
5374 	if (!(ctxt->d & NoWrite)) {
5375 		rc = writeback(ctxt, &ctxt->dst);
5376 		if (rc != X86EMUL_CONTINUE)
5377 			goto done;
5378 	}
5379 
5380 	/*
5381 	 * restore dst type in case the decoding will be reused
5382 	 * (happens for string instruction )
5383 	 */
5384 	ctxt->dst.type = saved_dst_type;
5385 
5386 	if ((ctxt->d & SrcMask) == SrcSI)
5387 		string_addr_inc(ctxt, VCPU_REGS_RSI, &ctxt->src);
5388 
5389 	if ((ctxt->d & DstMask) == DstDI)
5390 		string_addr_inc(ctxt, VCPU_REGS_RDI, &ctxt->dst);
5391 
5392 	if (ctxt->rep_prefix && (ctxt->d & String)) {
5393 		unsigned int count;
5394 		struct read_cache *r = &ctxt->io_read;
5395 		if ((ctxt->d & SrcMask) == SrcSI)
5396 			count = ctxt->src.count;
5397 		else
5398 			count = ctxt->dst.count;
5399 		register_address_increment(ctxt, VCPU_REGS_RCX, -count);
5400 
5401 		if (!string_insn_completed(ctxt)) {
5402 			/*
5403 			 * Re-enter guest when pio read ahead buffer is empty
5404 			 * or, if it is not used, after each 1024 iteration.
5405 			 */
5406 			if ((r->end != 0 || reg_read(ctxt, VCPU_REGS_RCX) & 0x3ff) &&
5407 			    (r->end == 0 || r->end != r->pos)) {
5408 				/*
5409 				 * Reset read cache. Usually happens before
5410 				 * decode, but since instruction is restarted
5411 				 * we have to do it here.
5412 				 */
5413 				ctxt->mem_read.end = 0;
5414 				writeback_registers(ctxt);
5415 				return EMULATION_RESTART;
5416 			}
5417 			goto done; /* skip rip writeback */
5418 		}
5419 		ctxt->eflags &= ~X86_EFLAGS_RF;
5420 	}
5421 
5422 	ctxt->eip = ctxt->_eip;
5423 
5424 done:
5425 	if (rc == X86EMUL_PROPAGATE_FAULT) {
5426 		WARN_ON(ctxt->exception.vector > 0x1f);
5427 		ctxt->have_exception = true;
5428 	}
5429 	if (rc == X86EMUL_INTERCEPTED)
5430 		return EMULATION_INTERCEPTED;
5431 
5432 	if (rc == X86EMUL_CONTINUE)
5433 		writeback_registers(ctxt);
5434 
5435 	return (rc == X86EMUL_UNHANDLEABLE) ? EMULATION_FAILED : EMULATION_OK;
5436 
5437 twobyte_insn:
5438 	switch (ctxt->b) {
5439 	case 0x09:		/* wbinvd */
5440 		(ctxt->ops->wbinvd)(ctxt);
5441 		break;
5442 	case 0x08:		/* invd */
5443 	case 0x0d:		/* GrpP (prefetch) */
5444 	case 0x18:		/* Grp16 (prefetch/nop) */
5445 	case 0x1f:		/* nop */
5446 		break;
5447 	case 0x20: /* mov cr, reg */
5448 		ctxt->dst.val = ops->get_cr(ctxt, ctxt->modrm_reg);
5449 		break;
5450 	case 0x21: /* mov from dr to reg */
5451 		ops->get_dr(ctxt, ctxt->modrm_reg, &ctxt->dst.val);
5452 		break;
5453 	case 0x40 ... 0x4f:	/* cmov */
5454 		if (test_cc(ctxt->b, ctxt->eflags))
5455 			ctxt->dst.val = ctxt->src.val;
5456 		else if (ctxt->op_bytes != 4)
5457 			ctxt->dst.type = OP_NONE; /* no writeback */
5458 		break;
5459 	case 0x80 ... 0x8f: /* jnz rel, etc*/
5460 		if (test_cc(ctxt->b, ctxt->eflags))
5461 			rc = jmp_rel(ctxt, ctxt->src.val);
5462 		break;
5463 	case 0x90 ... 0x9f:     /* setcc r/m8 */
5464 		ctxt->dst.val = test_cc(ctxt->b, ctxt->eflags);
5465 		break;
5466 	case 0xb6 ... 0xb7:	/* movzx */
5467 		ctxt->dst.bytes = ctxt->op_bytes;
5468 		ctxt->dst.val = (ctxt->src.bytes == 1) ? (u8) ctxt->src.val
5469 						       : (u16) ctxt->src.val;
5470 		break;
5471 	case 0xbe ... 0xbf:	/* movsx */
5472 		ctxt->dst.bytes = ctxt->op_bytes;
5473 		ctxt->dst.val = (ctxt->src.bytes == 1) ? (s8) ctxt->src.val :
5474 							(s16) ctxt->src.val;
5475 		break;
5476 	default:
5477 		goto cannot_emulate;
5478 	}
5479 
5480 threebyte_insn:
5481 
5482 	if (rc != X86EMUL_CONTINUE)
5483 		goto done;
5484 
5485 	goto writeback;
5486 
5487 cannot_emulate:
5488 	return EMULATION_FAILED;
5489 }
5490 
5491 void emulator_invalidate_register_cache(struct x86_emulate_ctxt *ctxt)
5492 {
5493 	invalidate_registers(ctxt);
5494 }
5495 
5496 void emulator_writeback_register_cache(struct x86_emulate_ctxt *ctxt)
5497 {
5498 	writeback_registers(ctxt);
5499 }
5500