xref: /linux/arch/x86/kvm/emulate.c (revision c4ee0af3fa0dc65f690fc908f02b8355f9576ea0)
1 /******************************************************************************
2  * emulate.c
3  *
4  * Generic x86 (32-bit and 64-bit) instruction decoder and emulator.
5  *
6  * Copyright (c) 2005 Keir Fraser
7  *
8  * Linux coding style, mod r/m decoder, segment base fixes, real-mode
9  * privileged instructions:
10  *
11  * Copyright (C) 2006 Qumranet
12  * Copyright 2010 Red Hat, Inc. and/or its affiliates.
13  *
14  *   Avi Kivity <avi@qumranet.com>
15  *   Yaniv Kamay <yaniv@qumranet.com>
16  *
17  * This work is licensed under the terms of the GNU GPL, version 2.  See
18  * the COPYING file in the top-level directory.
19  *
20  * From: xen-unstable 10676:af9809f51f81a3c43f276f00c81a52ef558afda4
21  */
22 
23 #include <linux/kvm_host.h>
24 #include "kvm_cache_regs.h"
25 #include <linux/module.h>
26 #include <asm/kvm_emulate.h>
27 #include <linux/stringify.h>
28 
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 DstImmUByte (OpImmUByte << DstShift)
90 #define DstDX       (OpDX << DstShift)
91 #define DstAccLo    (OpAccLo << DstShift)
92 #define DstMask     (OpMask << DstShift)
93 /* Source operand type. */
94 #define SrcShift    6
95 #define SrcNone     (OpNone << SrcShift)
96 #define SrcReg      (OpReg << SrcShift)
97 #define SrcMem      (OpMem << SrcShift)
98 #define SrcMem16    (OpMem16 << SrcShift)
99 #define SrcMem32    (OpMem32 << SrcShift)
100 #define SrcImm      (OpImm << SrcShift)
101 #define SrcImmByte  (OpImmByte << SrcShift)
102 #define SrcOne      (OpOne << SrcShift)
103 #define SrcImmUByte (OpImmUByte << SrcShift)
104 #define SrcImmU     (OpImmU << SrcShift)
105 #define SrcSI       (OpSI << SrcShift)
106 #define SrcXLat     (OpXLat << SrcShift)
107 #define SrcImmFAddr (OpImmFAddr << SrcShift)
108 #define SrcMemFAddr (OpMemFAddr << SrcShift)
109 #define SrcAcc      (OpAcc << SrcShift)
110 #define SrcImmU16   (OpImmU16 << SrcShift)
111 #define SrcImm64    (OpImm64 << SrcShift)
112 #define SrcDX       (OpDX << SrcShift)
113 #define SrcMem8     (OpMem8 << SrcShift)
114 #define SrcAccHi    (OpAccHi << SrcShift)
115 #define SrcMask     (OpMask << SrcShift)
116 #define BitOp       (1<<11)
117 #define MemAbs      (1<<12)      /* Memory operand is absolute displacement */
118 #define String      (1<<13)     /* String instruction (rep capable) */
119 #define Stack       (1<<14)     /* Stack instruction (push/pop) */
120 #define GroupMask   (7<<15)     /* Opcode uses one of the group mechanisms */
121 #define Group       (1<<15)     /* Bits 3:5 of modrm byte extend opcode */
122 #define GroupDual   (2<<15)     /* Alternate decoding of mod == 3 */
123 #define Prefix      (3<<15)     /* Instruction varies with 66/f2/f3 prefix */
124 #define RMExt       (4<<15)     /* Opcode extension in ModRM r/m if mod == 3 */
125 #define Escape      (5<<15)     /* Escape to coprocessor instruction */
126 #define Sse         (1<<18)     /* SSE Vector instruction */
127 /* Generic ModRM decode. */
128 #define ModRM       (1<<19)
129 /* Destination is only written; never read. */
130 #define Mov         (1<<20)
131 /* Misc flags */
132 #define Prot        (1<<21) /* instruction generates #UD if not in prot-mode */
133 #define EmulateOnUD (1<<22) /* Emulate if unsupported by the host */
134 #define NoAccess    (1<<23) /* Don't access memory (lea/invlpg/verr etc) */
135 #define Op3264      (1<<24) /* Operand is 64b in long mode, 32b otherwise */
136 #define Undefined   (1<<25) /* No Such Instruction */
137 #define Lock        (1<<26) /* lock prefix is allowed for the instruction */
138 #define Priv        (1<<27) /* instruction generates #GP if current CPL != 0 */
139 #define No64	    (1<<28)
140 #define PageTable   (1 << 29)   /* instruction used to write page table */
141 #define NotImpl     (1 << 30)   /* instruction is not implemented */
142 /* Source 2 operand type */
143 #define Src2Shift   (31)
144 #define Src2None    (OpNone << Src2Shift)
145 #define Src2Mem     (OpMem << Src2Shift)
146 #define Src2CL      (OpCL << Src2Shift)
147 #define Src2ImmByte (OpImmByte << Src2Shift)
148 #define Src2One     (OpOne << Src2Shift)
149 #define Src2Imm     (OpImm << Src2Shift)
150 #define Src2ES      (OpES << Src2Shift)
151 #define Src2CS      (OpCS << Src2Shift)
152 #define Src2SS      (OpSS << Src2Shift)
153 #define Src2DS      (OpDS << Src2Shift)
154 #define Src2FS      (OpFS << Src2Shift)
155 #define Src2GS      (OpGS << Src2Shift)
156 #define Src2Mask    (OpMask << Src2Shift)
157 #define Mmx         ((u64)1 << 40)  /* MMX Vector instruction */
158 #define Aligned     ((u64)1 << 41)  /* Explicitly aligned (e.g. MOVDQA) */
159 #define Unaligned   ((u64)1 << 42)  /* Explicitly unaligned (e.g. MOVDQU) */
160 #define Avx         ((u64)1 << 43)  /* Advanced Vector Extensions */
161 #define Fastop      ((u64)1 << 44)  /* Use opcode::u.fastop */
162 #define NoWrite     ((u64)1 << 45)  /* No writeback */
163 #define SrcWrite    ((u64)1 << 46)  /* Write back src operand */
164 
165 #define DstXacc     (DstAccLo | SrcAccHi | SrcWrite)
166 
167 #define X2(x...) x, x
168 #define X3(x...) X2(x), x
169 #define X4(x...) X2(x), X2(x)
170 #define X5(x...) X4(x), x
171 #define X6(x...) X4(x), X2(x)
172 #define X7(x...) X4(x), X3(x)
173 #define X8(x...) X4(x), X4(x)
174 #define X16(x...) X8(x), X8(x)
175 
176 #define NR_FASTOP (ilog2(sizeof(ulong)) + 1)
177 #define FASTOP_SIZE 8
178 
179 /*
180  * fastop functions have a special calling convention:
181  *
182  * dst:    rax        (in/out)
183  * src:    rdx        (in/out)
184  * src2:   rcx        (in)
185  * flags:  rflags     (in/out)
186  * ex:     rsi        (in:fastop pointer, out:zero if exception)
187  *
188  * Moreover, they are all exactly FASTOP_SIZE bytes long, so functions for
189  * different operand sizes can be reached by calculation, rather than a jump
190  * table (which would be bigger than the code).
191  *
192  * fastop functions are declared as taking a never-defined fastop parameter,
193  * so they can't be called from C directly.
194  */
195 
196 struct fastop;
197 
198 struct opcode {
199 	u64 flags : 56;
200 	u64 intercept : 8;
201 	union {
202 		int (*execute)(struct x86_emulate_ctxt *ctxt);
203 		const struct opcode *group;
204 		const struct group_dual *gdual;
205 		const struct gprefix *gprefix;
206 		const struct escape *esc;
207 		void (*fastop)(struct fastop *fake);
208 	} u;
209 	int (*check_perm)(struct x86_emulate_ctxt *ctxt);
210 };
211 
212 struct group_dual {
213 	struct opcode mod012[8];
214 	struct opcode mod3[8];
215 };
216 
217 struct gprefix {
218 	struct opcode pfx_no;
219 	struct opcode pfx_66;
220 	struct opcode pfx_f2;
221 	struct opcode pfx_f3;
222 };
223 
224 struct escape {
225 	struct opcode op[8];
226 	struct opcode high[64];
227 };
228 
229 /* EFLAGS bit definitions. */
230 #define EFLG_ID (1<<21)
231 #define EFLG_VIP (1<<20)
232 #define EFLG_VIF (1<<19)
233 #define EFLG_AC (1<<18)
234 #define EFLG_VM (1<<17)
235 #define EFLG_RF (1<<16)
236 #define EFLG_IOPL (3<<12)
237 #define EFLG_NT (1<<14)
238 #define EFLG_OF (1<<11)
239 #define EFLG_DF (1<<10)
240 #define EFLG_IF (1<<9)
241 #define EFLG_TF (1<<8)
242 #define EFLG_SF (1<<7)
243 #define EFLG_ZF (1<<6)
244 #define EFLG_AF (1<<4)
245 #define EFLG_PF (1<<2)
246 #define EFLG_CF (1<<0)
247 
248 #define EFLG_RESERVED_ZEROS_MASK 0xffc0802a
249 #define EFLG_RESERVED_ONE_MASK 2
250 
251 static ulong reg_read(struct x86_emulate_ctxt *ctxt, unsigned nr)
252 {
253 	if (!(ctxt->regs_valid & (1 << nr))) {
254 		ctxt->regs_valid |= 1 << nr;
255 		ctxt->_regs[nr] = ctxt->ops->read_gpr(ctxt, nr);
256 	}
257 	return ctxt->_regs[nr];
258 }
259 
260 static ulong *reg_write(struct x86_emulate_ctxt *ctxt, unsigned nr)
261 {
262 	ctxt->regs_valid |= 1 << nr;
263 	ctxt->regs_dirty |= 1 << nr;
264 	return &ctxt->_regs[nr];
265 }
266 
267 static ulong *reg_rmw(struct x86_emulate_ctxt *ctxt, unsigned nr)
268 {
269 	reg_read(ctxt, nr);
270 	return reg_write(ctxt, nr);
271 }
272 
273 static void writeback_registers(struct x86_emulate_ctxt *ctxt)
274 {
275 	unsigned reg;
276 
277 	for_each_set_bit(reg, (ulong *)&ctxt->regs_dirty, 16)
278 		ctxt->ops->write_gpr(ctxt, reg, ctxt->_regs[reg]);
279 }
280 
281 static void invalidate_registers(struct x86_emulate_ctxt *ctxt)
282 {
283 	ctxt->regs_dirty = 0;
284 	ctxt->regs_valid = 0;
285 }
286 
287 /*
288  * These EFLAGS bits are restored from saved value during emulation, and
289  * any changes are written back to the saved value after emulation.
290  */
291 #define EFLAGS_MASK (EFLG_OF|EFLG_SF|EFLG_ZF|EFLG_AF|EFLG_PF|EFLG_CF)
292 
293 #ifdef CONFIG_X86_64
294 #define ON64(x) x
295 #else
296 #define ON64(x)
297 #endif
298 
299 static int fastop(struct x86_emulate_ctxt *ctxt, void (*fop)(struct fastop *));
300 
301 #define FOP_ALIGN ".align " __stringify(FASTOP_SIZE) " \n\t"
302 #define FOP_RET   "ret \n\t"
303 
304 #define FOP_START(op) \
305 	extern void em_##op(struct fastop *fake); \
306 	asm(".pushsection .text, \"ax\" \n\t" \
307 	    ".global em_" #op " \n\t" \
308             FOP_ALIGN \
309 	    "em_" #op ": \n\t"
310 
311 #define FOP_END \
312 	    ".popsection")
313 
314 #define FOPNOP() FOP_ALIGN FOP_RET
315 
316 #define FOP1E(op,  dst) \
317 	FOP_ALIGN "10: " #op " %" #dst " \n\t" FOP_RET
318 
319 #define FOP1EEX(op,  dst) \
320 	FOP1E(op, dst) _ASM_EXTABLE(10b, kvm_fastop_exception)
321 
322 #define FASTOP1(op) \
323 	FOP_START(op) \
324 	FOP1E(op##b, al) \
325 	FOP1E(op##w, ax) \
326 	FOP1E(op##l, eax) \
327 	ON64(FOP1E(op##q, rax))	\
328 	FOP_END
329 
330 /* 1-operand, using src2 (for MUL/DIV r/m) */
331 #define FASTOP1SRC2(op, name) \
332 	FOP_START(name) \
333 	FOP1E(op, cl) \
334 	FOP1E(op, cx) \
335 	FOP1E(op, ecx) \
336 	ON64(FOP1E(op, rcx)) \
337 	FOP_END
338 
339 /* 1-operand, using src2 (for MUL/DIV r/m), with exceptions */
340 #define FASTOP1SRC2EX(op, name) \
341 	FOP_START(name) \
342 	FOP1EEX(op, cl) \
343 	FOP1EEX(op, cx) \
344 	FOP1EEX(op, ecx) \
345 	ON64(FOP1EEX(op, rcx)) \
346 	FOP_END
347 
348 #define FOP2E(op,  dst, src)	   \
349 	FOP_ALIGN #op " %" #src ", %" #dst " \n\t" FOP_RET
350 
351 #define FASTOP2(op) \
352 	FOP_START(op) \
353 	FOP2E(op##b, al, dl) \
354 	FOP2E(op##w, ax, dx) \
355 	FOP2E(op##l, eax, edx) \
356 	ON64(FOP2E(op##q, rax, rdx)) \
357 	FOP_END
358 
359 /* 2 operand, word only */
360 #define FASTOP2W(op) \
361 	FOP_START(op) \
362 	FOPNOP() \
363 	FOP2E(op##w, ax, dx) \
364 	FOP2E(op##l, eax, edx) \
365 	ON64(FOP2E(op##q, rax, rdx)) \
366 	FOP_END
367 
368 /* 2 operand, src is CL */
369 #define FASTOP2CL(op) \
370 	FOP_START(op) \
371 	FOP2E(op##b, al, cl) \
372 	FOP2E(op##w, ax, cl) \
373 	FOP2E(op##l, eax, cl) \
374 	ON64(FOP2E(op##q, rax, cl)) \
375 	FOP_END
376 
377 #define FOP3E(op,  dst, src, src2) \
378 	FOP_ALIGN #op " %" #src2 ", %" #src ", %" #dst " \n\t" FOP_RET
379 
380 /* 3-operand, word-only, src2=cl */
381 #define FASTOP3WCL(op) \
382 	FOP_START(op) \
383 	FOPNOP() \
384 	FOP3E(op##w, ax, dx, cl) \
385 	FOP3E(op##l, eax, edx, cl) \
386 	ON64(FOP3E(op##q, rax, rdx, cl)) \
387 	FOP_END
388 
389 /* Special case for SETcc - 1 instruction per cc */
390 #define FOP_SETCC(op) ".align 4; " #op " %al; ret \n\t"
391 
392 asm(".global kvm_fastop_exception \n"
393     "kvm_fastop_exception: xor %esi, %esi; ret");
394 
395 FOP_START(setcc)
396 FOP_SETCC(seto)
397 FOP_SETCC(setno)
398 FOP_SETCC(setc)
399 FOP_SETCC(setnc)
400 FOP_SETCC(setz)
401 FOP_SETCC(setnz)
402 FOP_SETCC(setbe)
403 FOP_SETCC(setnbe)
404 FOP_SETCC(sets)
405 FOP_SETCC(setns)
406 FOP_SETCC(setp)
407 FOP_SETCC(setnp)
408 FOP_SETCC(setl)
409 FOP_SETCC(setnl)
410 FOP_SETCC(setle)
411 FOP_SETCC(setnle)
412 FOP_END;
413 
414 FOP_START(salc) "pushf; sbb %al, %al; popf \n\t" FOP_RET
415 FOP_END;
416 
417 static int emulator_check_intercept(struct x86_emulate_ctxt *ctxt,
418 				    enum x86_intercept intercept,
419 				    enum x86_intercept_stage stage)
420 {
421 	struct x86_instruction_info info = {
422 		.intercept  = intercept,
423 		.rep_prefix = ctxt->rep_prefix,
424 		.modrm_mod  = ctxt->modrm_mod,
425 		.modrm_reg  = ctxt->modrm_reg,
426 		.modrm_rm   = ctxt->modrm_rm,
427 		.src_val    = ctxt->src.val64,
428 		.src_bytes  = ctxt->src.bytes,
429 		.dst_bytes  = ctxt->dst.bytes,
430 		.ad_bytes   = ctxt->ad_bytes,
431 		.next_rip   = ctxt->eip,
432 	};
433 
434 	return ctxt->ops->intercept(ctxt, &info, stage);
435 }
436 
437 static void assign_masked(ulong *dest, ulong src, ulong mask)
438 {
439 	*dest = (*dest & ~mask) | (src & mask);
440 }
441 
442 static inline unsigned long ad_mask(struct x86_emulate_ctxt *ctxt)
443 {
444 	return (1UL << (ctxt->ad_bytes << 3)) - 1;
445 }
446 
447 static ulong stack_mask(struct x86_emulate_ctxt *ctxt)
448 {
449 	u16 sel;
450 	struct desc_struct ss;
451 
452 	if (ctxt->mode == X86EMUL_MODE_PROT64)
453 		return ~0UL;
454 	ctxt->ops->get_segment(ctxt, &sel, &ss, NULL, VCPU_SREG_SS);
455 	return ~0U >> ((ss.d ^ 1) * 16);  /* d=0: 0xffff; d=1: 0xffffffff */
456 }
457 
458 static int stack_size(struct x86_emulate_ctxt *ctxt)
459 {
460 	return (__fls(stack_mask(ctxt)) + 1) >> 3;
461 }
462 
463 /* Access/update address held in a register, based on addressing mode. */
464 static inline unsigned long
465 address_mask(struct x86_emulate_ctxt *ctxt, unsigned long reg)
466 {
467 	if (ctxt->ad_bytes == sizeof(unsigned long))
468 		return reg;
469 	else
470 		return reg & ad_mask(ctxt);
471 }
472 
473 static inline unsigned long
474 register_address(struct x86_emulate_ctxt *ctxt, unsigned long reg)
475 {
476 	return address_mask(ctxt, reg);
477 }
478 
479 static void masked_increment(ulong *reg, ulong mask, int inc)
480 {
481 	assign_masked(reg, *reg + inc, mask);
482 }
483 
484 static inline void
485 register_address_increment(struct x86_emulate_ctxt *ctxt, unsigned long *reg, int inc)
486 {
487 	ulong mask;
488 
489 	if (ctxt->ad_bytes == sizeof(unsigned long))
490 		mask = ~0UL;
491 	else
492 		mask = ad_mask(ctxt);
493 	masked_increment(reg, mask, inc);
494 }
495 
496 static void rsp_increment(struct x86_emulate_ctxt *ctxt, int inc)
497 {
498 	masked_increment(reg_rmw(ctxt, VCPU_REGS_RSP), stack_mask(ctxt), inc);
499 }
500 
501 static inline void jmp_rel(struct x86_emulate_ctxt *ctxt, int rel)
502 {
503 	register_address_increment(ctxt, &ctxt->_eip, rel);
504 }
505 
506 static u32 desc_limit_scaled(struct desc_struct *desc)
507 {
508 	u32 limit = get_desc_limit(desc);
509 
510 	return desc->g ? (limit << 12) | 0xfff : limit;
511 }
512 
513 static void set_seg_override(struct x86_emulate_ctxt *ctxt, int seg)
514 {
515 	ctxt->has_seg_override = true;
516 	ctxt->seg_override = seg;
517 }
518 
519 static unsigned long seg_base(struct x86_emulate_ctxt *ctxt, int seg)
520 {
521 	if (ctxt->mode == X86EMUL_MODE_PROT64 && seg < VCPU_SREG_FS)
522 		return 0;
523 
524 	return ctxt->ops->get_cached_segment_base(ctxt, seg);
525 }
526 
527 static unsigned seg_override(struct x86_emulate_ctxt *ctxt)
528 {
529 	if (!ctxt->has_seg_override)
530 		return 0;
531 
532 	return ctxt->seg_override;
533 }
534 
535 static int emulate_exception(struct x86_emulate_ctxt *ctxt, int vec,
536 			     u32 error, bool valid)
537 {
538 	ctxt->exception.vector = vec;
539 	ctxt->exception.error_code = error;
540 	ctxt->exception.error_code_valid = valid;
541 	return X86EMUL_PROPAGATE_FAULT;
542 }
543 
544 static int emulate_db(struct x86_emulate_ctxt *ctxt)
545 {
546 	return emulate_exception(ctxt, DB_VECTOR, 0, false);
547 }
548 
549 static int emulate_gp(struct x86_emulate_ctxt *ctxt, int err)
550 {
551 	return emulate_exception(ctxt, GP_VECTOR, err, true);
552 }
553 
554 static int emulate_ss(struct x86_emulate_ctxt *ctxt, int err)
555 {
556 	return emulate_exception(ctxt, SS_VECTOR, err, true);
557 }
558 
559 static int emulate_ud(struct x86_emulate_ctxt *ctxt)
560 {
561 	return emulate_exception(ctxt, UD_VECTOR, 0, false);
562 }
563 
564 static int emulate_ts(struct x86_emulate_ctxt *ctxt, int err)
565 {
566 	return emulate_exception(ctxt, TS_VECTOR, err, true);
567 }
568 
569 static int emulate_de(struct x86_emulate_ctxt *ctxt)
570 {
571 	return emulate_exception(ctxt, DE_VECTOR, 0, false);
572 }
573 
574 static int emulate_nm(struct x86_emulate_ctxt *ctxt)
575 {
576 	return emulate_exception(ctxt, NM_VECTOR, 0, false);
577 }
578 
579 static u16 get_segment_selector(struct x86_emulate_ctxt *ctxt, unsigned seg)
580 {
581 	u16 selector;
582 	struct desc_struct desc;
583 
584 	ctxt->ops->get_segment(ctxt, &selector, &desc, NULL, seg);
585 	return selector;
586 }
587 
588 static void set_segment_selector(struct x86_emulate_ctxt *ctxt, u16 selector,
589 				 unsigned seg)
590 {
591 	u16 dummy;
592 	u32 base3;
593 	struct desc_struct desc;
594 
595 	ctxt->ops->get_segment(ctxt, &dummy, &desc, &base3, seg);
596 	ctxt->ops->set_segment(ctxt, selector, &desc, base3, seg);
597 }
598 
599 /*
600  * x86 defines three classes of vector instructions: explicitly
601  * aligned, explicitly unaligned, and the rest, which change behaviour
602  * depending on whether they're AVX encoded or not.
603  *
604  * Also included is CMPXCHG16B which is not a vector instruction, yet it is
605  * subject to the same check.
606  */
607 static bool insn_aligned(struct x86_emulate_ctxt *ctxt, unsigned size)
608 {
609 	if (likely(size < 16))
610 		return false;
611 
612 	if (ctxt->d & Aligned)
613 		return true;
614 	else if (ctxt->d & Unaligned)
615 		return false;
616 	else if (ctxt->d & Avx)
617 		return false;
618 	else
619 		return true;
620 }
621 
622 static int __linearize(struct x86_emulate_ctxt *ctxt,
623 		     struct segmented_address addr,
624 		     unsigned size, bool write, bool fetch,
625 		     ulong *linear)
626 {
627 	struct desc_struct desc;
628 	bool usable;
629 	ulong la;
630 	u32 lim;
631 	u16 sel;
632 	unsigned cpl;
633 
634 	la = seg_base(ctxt, addr.seg) + addr.ea;
635 	switch (ctxt->mode) {
636 	case X86EMUL_MODE_PROT64:
637 		if (((signed long)la << 16) >> 16 != la)
638 			return emulate_gp(ctxt, 0);
639 		break;
640 	default:
641 		usable = ctxt->ops->get_segment(ctxt, &sel, &desc, NULL,
642 						addr.seg);
643 		if (!usable)
644 			goto bad;
645 		/* code segment in protected mode or read-only data segment */
646 		if ((((ctxt->mode != X86EMUL_MODE_REAL) && (desc.type & 8))
647 					|| !(desc.type & 2)) && write)
648 			goto bad;
649 		/* unreadable code segment */
650 		if (!fetch && (desc.type & 8) && !(desc.type & 2))
651 			goto bad;
652 		lim = desc_limit_scaled(&desc);
653 		if ((desc.type & 8) || !(desc.type & 4)) {
654 			/* expand-up segment */
655 			if (addr.ea > lim || (u32)(addr.ea + size - 1) > lim)
656 				goto bad;
657 		} else {
658 			/* expand-down segment */
659 			if (addr.ea <= lim || (u32)(addr.ea + size - 1) <= lim)
660 				goto bad;
661 			lim = desc.d ? 0xffffffff : 0xffff;
662 			if (addr.ea > lim || (u32)(addr.ea + size - 1) > lim)
663 				goto bad;
664 		}
665 		cpl = ctxt->ops->cpl(ctxt);
666 		if (!(desc.type & 8)) {
667 			/* data segment */
668 			if (cpl > desc.dpl)
669 				goto bad;
670 		} else if ((desc.type & 8) && !(desc.type & 4)) {
671 			/* nonconforming code segment */
672 			if (cpl != desc.dpl)
673 				goto bad;
674 		} else if ((desc.type & 8) && (desc.type & 4)) {
675 			/* conforming code segment */
676 			if (cpl < desc.dpl)
677 				goto bad;
678 		}
679 		break;
680 	}
681 	if (fetch ? ctxt->mode != X86EMUL_MODE_PROT64 : ctxt->ad_bytes != 8)
682 		la &= (u32)-1;
683 	if (insn_aligned(ctxt, size) && ((la & (size - 1)) != 0))
684 		return emulate_gp(ctxt, 0);
685 	*linear = la;
686 	return X86EMUL_CONTINUE;
687 bad:
688 	if (addr.seg == VCPU_SREG_SS)
689 		return emulate_ss(ctxt, sel);
690 	else
691 		return emulate_gp(ctxt, sel);
692 }
693 
694 static int linearize(struct x86_emulate_ctxt *ctxt,
695 		     struct segmented_address addr,
696 		     unsigned size, bool write,
697 		     ulong *linear)
698 {
699 	return __linearize(ctxt, addr, size, write, false, linear);
700 }
701 
702 
703 static int segmented_read_std(struct x86_emulate_ctxt *ctxt,
704 			      struct segmented_address addr,
705 			      void *data,
706 			      unsigned size)
707 {
708 	int rc;
709 	ulong linear;
710 
711 	rc = linearize(ctxt, addr, size, false, &linear);
712 	if (rc != X86EMUL_CONTINUE)
713 		return rc;
714 	return ctxt->ops->read_std(ctxt, linear, data, size, &ctxt->exception);
715 }
716 
717 /*
718  * Fetch the next byte of the instruction being emulated which is pointed to
719  * by ctxt->_eip, then increment ctxt->_eip.
720  *
721  * Also prefetch the remaining bytes of the instruction without crossing page
722  * boundary if they are not in fetch_cache yet.
723  */
724 static int do_insn_fetch_byte(struct x86_emulate_ctxt *ctxt, u8 *dest)
725 {
726 	struct fetch_cache *fc = &ctxt->fetch;
727 	int rc;
728 	int size, cur_size;
729 
730 	if (ctxt->_eip == fc->end) {
731 		unsigned long linear;
732 		struct segmented_address addr = { .seg = VCPU_SREG_CS,
733 						  .ea  = ctxt->_eip };
734 		cur_size = fc->end - fc->start;
735 		size = min(15UL - cur_size,
736 			   PAGE_SIZE - offset_in_page(ctxt->_eip));
737 		rc = __linearize(ctxt, addr, size, false, true, &linear);
738 		if (unlikely(rc != X86EMUL_CONTINUE))
739 			return rc;
740 		rc = ctxt->ops->fetch(ctxt, linear, fc->data + cur_size,
741 				      size, &ctxt->exception);
742 		if (unlikely(rc != X86EMUL_CONTINUE))
743 			return rc;
744 		fc->end += size;
745 	}
746 	*dest = fc->data[ctxt->_eip - fc->start];
747 	ctxt->_eip++;
748 	return X86EMUL_CONTINUE;
749 }
750 
751 static int do_insn_fetch(struct x86_emulate_ctxt *ctxt,
752 			 void *dest, unsigned size)
753 {
754 	int rc;
755 
756 	/* x86 instructions are limited to 15 bytes. */
757 	if (unlikely(ctxt->_eip + size - ctxt->eip > 15))
758 		return X86EMUL_UNHANDLEABLE;
759 	while (size--) {
760 		rc = do_insn_fetch_byte(ctxt, dest++);
761 		if (rc != X86EMUL_CONTINUE)
762 			return rc;
763 	}
764 	return X86EMUL_CONTINUE;
765 }
766 
767 /* Fetch next part of the instruction being emulated. */
768 #define insn_fetch(_type, _ctxt)					\
769 ({	unsigned long _x;						\
770 	rc = do_insn_fetch(_ctxt, &_x, sizeof(_type));			\
771 	if (rc != X86EMUL_CONTINUE)					\
772 		goto done;						\
773 	(_type)_x;							\
774 })
775 
776 #define insn_fetch_arr(_arr, _size, _ctxt)				\
777 ({	rc = do_insn_fetch(_ctxt, _arr, (_size));			\
778 	if (rc != X86EMUL_CONTINUE)					\
779 		goto done;						\
780 })
781 
782 /*
783  * Given the 'reg' portion of a ModRM byte, and a register block, return a
784  * pointer into the block that addresses the relevant register.
785  * @highbyte_regs specifies whether to decode AH,CH,DH,BH.
786  */
787 static void *decode_register(struct x86_emulate_ctxt *ctxt, u8 modrm_reg,
788 			     int byteop)
789 {
790 	void *p;
791 	int highbyte_regs = (ctxt->rex_prefix == 0) && byteop;
792 
793 	if (highbyte_regs && modrm_reg >= 4 && modrm_reg < 8)
794 		p = (unsigned char *)reg_rmw(ctxt, modrm_reg & 3) + 1;
795 	else
796 		p = reg_rmw(ctxt, modrm_reg);
797 	return p;
798 }
799 
800 static int read_descriptor(struct x86_emulate_ctxt *ctxt,
801 			   struct segmented_address addr,
802 			   u16 *size, unsigned long *address, int op_bytes)
803 {
804 	int rc;
805 
806 	if (op_bytes == 2)
807 		op_bytes = 3;
808 	*address = 0;
809 	rc = segmented_read_std(ctxt, addr, size, 2);
810 	if (rc != X86EMUL_CONTINUE)
811 		return rc;
812 	addr.ea += 2;
813 	rc = segmented_read_std(ctxt, addr, address, op_bytes);
814 	return rc;
815 }
816 
817 FASTOP2(add);
818 FASTOP2(or);
819 FASTOP2(adc);
820 FASTOP2(sbb);
821 FASTOP2(and);
822 FASTOP2(sub);
823 FASTOP2(xor);
824 FASTOP2(cmp);
825 FASTOP2(test);
826 
827 FASTOP1SRC2(mul, mul_ex);
828 FASTOP1SRC2(imul, imul_ex);
829 FASTOP1SRC2EX(div, div_ex);
830 FASTOP1SRC2EX(idiv, idiv_ex);
831 
832 FASTOP3WCL(shld);
833 FASTOP3WCL(shrd);
834 
835 FASTOP2W(imul);
836 
837 FASTOP1(not);
838 FASTOP1(neg);
839 FASTOP1(inc);
840 FASTOP1(dec);
841 
842 FASTOP2CL(rol);
843 FASTOP2CL(ror);
844 FASTOP2CL(rcl);
845 FASTOP2CL(rcr);
846 FASTOP2CL(shl);
847 FASTOP2CL(shr);
848 FASTOP2CL(sar);
849 
850 FASTOP2W(bsf);
851 FASTOP2W(bsr);
852 FASTOP2W(bt);
853 FASTOP2W(bts);
854 FASTOP2W(btr);
855 FASTOP2W(btc);
856 
857 FASTOP2(xadd);
858 
859 static u8 test_cc(unsigned int condition, unsigned long flags)
860 {
861 	u8 rc;
862 	void (*fop)(void) = (void *)em_setcc + 4 * (condition & 0xf);
863 
864 	flags = (flags & EFLAGS_MASK) | X86_EFLAGS_IF;
865 	asm("push %[flags]; popf; call *%[fastop]"
866 	    : "=a"(rc) : [fastop]"r"(fop), [flags]"r"(flags));
867 	return rc;
868 }
869 
870 static void fetch_register_operand(struct operand *op)
871 {
872 	switch (op->bytes) {
873 	case 1:
874 		op->val = *(u8 *)op->addr.reg;
875 		break;
876 	case 2:
877 		op->val = *(u16 *)op->addr.reg;
878 		break;
879 	case 4:
880 		op->val = *(u32 *)op->addr.reg;
881 		break;
882 	case 8:
883 		op->val = *(u64 *)op->addr.reg;
884 		break;
885 	}
886 }
887 
888 static void read_sse_reg(struct x86_emulate_ctxt *ctxt, sse128_t *data, int reg)
889 {
890 	ctxt->ops->get_fpu(ctxt);
891 	switch (reg) {
892 	case 0: asm("movdqa %%xmm0, %0" : "=m"(*data)); break;
893 	case 1: asm("movdqa %%xmm1, %0" : "=m"(*data)); break;
894 	case 2: asm("movdqa %%xmm2, %0" : "=m"(*data)); break;
895 	case 3: asm("movdqa %%xmm3, %0" : "=m"(*data)); break;
896 	case 4: asm("movdqa %%xmm4, %0" : "=m"(*data)); break;
897 	case 5: asm("movdqa %%xmm5, %0" : "=m"(*data)); break;
898 	case 6: asm("movdqa %%xmm6, %0" : "=m"(*data)); break;
899 	case 7: asm("movdqa %%xmm7, %0" : "=m"(*data)); break;
900 #ifdef CONFIG_X86_64
901 	case 8: asm("movdqa %%xmm8, %0" : "=m"(*data)); break;
902 	case 9: asm("movdqa %%xmm9, %0" : "=m"(*data)); break;
903 	case 10: asm("movdqa %%xmm10, %0" : "=m"(*data)); break;
904 	case 11: asm("movdqa %%xmm11, %0" : "=m"(*data)); break;
905 	case 12: asm("movdqa %%xmm12, %0" : "=m"(*data)); break;
906 	case 13: asm("movdqa %%xmm13, %0" : "=m"(*data)); break;
907 	case 14: asm("movdqa %%xmm14, %0" : "=m"(*data)); break;
908 	case 15: asm("movdqa %%xmm15, %0" : "=m"(*data)); break;
909 #endif
910 	default: BUG();
911 	}
912 	ctxt->ops->put_fpu(ctxt);
913 }
914 
915 static void write_sse_reg(struct x86_emulate_ctxt *ctxt, sse128_t *data,
916 			  int reg)
917 {
918 	ctxt->ops->get_fpu(ctxt);
919 	switch (reg) {
920 	case 0: asm("movdqa %0, %%xmm0" : : "m"(*data)); break;
921 	case 1: asm("movdqa %0, %%xmm1" : : "m"(*data)); break;
922 	case 2: asm("movdqa %0, %%xmm2" : : "m"(*data)); break;
923 	case 3: asm("movdqa %0, %%xmm3" : : "m"(*data)); break;
924 	case 4: asm("movdqa %0, %%xmm4" : : "m"(*data)); break;
925 	case 5: asm("movdqa %0, %%xmm5" : : "m"(*data)); break;
926 	case 6: asm("movdqa %0, %%xmm6" : : "m"(*data)); break;
927 	case 7: asm("movdqa %0, %%xmm7" : : "m"(*data)); break;
928 #ifdef CONFIG_X86_64
929 	case 8: asm("movdqa %0, %%xmm8" : : "m"(*data)); break;
930 	case 9: asm("movdqa %0, %%xmm9" : : "m"(*data)); break;
931 	case 10: asm("movdqa %0, %%xmm10" : : "m"(*data)); break;
932 	case 11: asm("movdqa %0, %%xmm11" : : "m"(*data)); break;
933 	case 12: asm("movdqa %0, %%xmm12" : : "m"(*data)); break;
934 	case 13: asm("movdqa %0, %%xmm13" : : "m"(*data)); break;
935 	case 14: asm("movdqa %0, %%xmm14" : : "m"(*data)); break;
936 	case 15: asm("movdqa %0, %%xmm15" : : "m"(*data)); break;
937 #endif
938 	default: BUG();
939 	}
940 	ctxt->ops->put_fpu(ctxt);
941 }
942 
943 static void read_mmx_reg(struct x86_emulate_ctxt *ctxt, u64 *data, int reg)
944 {
945 	ctxt->ops->get_fpu(ctxt);
946 	switch (reg) {
947 	case 0: asm("movq %%mm0, %0" : "=m"(*data)); break;
948 	case 1: asm("movq %%mm1, %0" : "=m"(*data)); break;
949 	case 2: asm("movq %%mm2, %0" : "=m"(*data)); break;
950 	case 3: asm("movq %%mm3, %0" : "=m"(*data)); break;
951 	case 4: asm("movq %%mm4, %0" : "=m"(*data)); break;
952 	case 5: asm("movq %%mm5, %0" : "=m"(*data)); break;
953 	case 6: asm("movq %%mm6, %0" : "=m"(*data)); break;
954 	case 7: asm("movq %%mm7, %0" : "=m"(*data)); break;
955 	default: BUG();
956 	}
957 	ctxt->ops->put_fpu(ctxt);
958 }
959 
960 static void write_mmx_reg(struct x86_emulate_ctxt *ctxt, u64 *data, int reg)
961 {
962 	ctxt->ops->get_fpu(ctxt);
963 	switch (reg) {
964 	case 0: asm("movq %0, %%mm0" : : "m"(*data)); break;
965 	case 1: asm("movq %0, %%mm1" : : "m"(*data)); break;
966 	case 2: asm("movq %0, %%mm2" : : "m"(*data)); break;
967 	case 3: asm("movq %0, %%mm3" : : "m"(*data)); break;
968 	case 4: asm("movq %0, %%mm4" : : "m"(*data)); break;
969 	case 5: asm("movq %0, %%mm5" : : "m"(*data)); break;
970 	case 6: asm("movq %0, %%mm6" : : "m"(*data)); break;
971 	case 7: asm("movq %0, %%mm7" : : "m"(*data)); break;
972 	default: BUG();
973 	}
974 	ctxt->ops->put_fpu(ctxt);
975 }
976 
977 static int em_fninit(struct x86_emulate_ctxt *ctxt)
978 {
979 	if (ctxt->ops->get_cr(ctxt, 0) & (X86_CR0_TS | X86_CR0_EM))
980 		return emulate_nm(ctxt);
981 
982 	ctxt->ops->get_fpu(ctxt);
983 	asm volatile("fninit");
984 	ctxt->ops->put_fpu(ctxt);
985 	return X86EMUL_CONTINUE;
986 }
987 
988 static int em_fnstcw(struct x86_emulate_ctxt *ctxt)
989 {
990 	u16 fcw;
991 
992 	if (ctxt->ops->get_cr(ctxt, 0) & (X86_CR0_TS | X86_CR0_EM))
993 		return emulate_nm(ctxt);
994 
995 	ctxt->ops->get_fpu(ctxt);
996 	asm volatile("fnstcw %0": "+m"(fcw));
997 	ctxt->ops->put_fpu(ctxt);
998 
999 	/* force 2 byte destination */
1000 	ctxt->dst.bytes = 2;
1001 	ctxt->dst.val = fcw;
1002 
1003 	return X86EMUL_CONTINUE;
1004 }
1005 
1006 static int em_fnstsw(struct x86_emulate_ctxt *ctxt)
1007 {
1008 	u16 fsw;
1009 
1010 	if (ctxt->ops->get_cr(ctxt, 0) & (X86_CR0_TS | X86_CR0_EM))
1011 		return emulate_nm(ctxt);
1012 
1013 	ctxt->ops->get_fpu(ctxt);
1014 	asm volatile("fnstsw %0": "+m"(fsw));
1015 	ctxt->ops->put_fpu(ctxt);
1016 
1017 	/* force 2 byte destination */
1018 	ctxt->dst.bytes = 2;
1019 	ctxt->dst.val = fsw;
1020 
1021 	return X86EMUL_CONTINUE;
1022 }
1023 
1024 static void decode_register_operand(struct x86_emulate_ctxt *ctxt,
1025 				    struct operand *op)
1026 {
1027 	unsigned reg = ctxt->modrm_reg;
1028 
1029 	if (!(ctxt->d & ModRM))
1030 		reg = (ctxt->b & 7) | ((ctxt->rex_prefix & 1) << 3);
1031 
1032 	if (ctxt->d & Sse) {
1033 		op->type = OP_XMM;
1034 		op->bytes = 16;
1035 		op->addr.xmm = reg;
1036 		read_sse_reg(ctxt, &op->vec_val, reg);
1037 		return;
1038 	}
1039 	if (ctxt->d & Mmx) {
1040 		reg &= 7;
1041 		op->type = OP_MM;
1042 		op->bytes = 8;
1043 		op->addr.mm = reg;
1044 		return;
1045 	}
1046 
1047 	op->type = OP_REG;
1048 	op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
1049 	op->addr.reg = decode_register(ctxt, reg, ctxt->d & ByteOp);
1050 
1051 	fetch_register_operand(op);
1052 	op->orig_val = op->val;
1053 }
1054 
1055 static void adjust_modrm_seg(struct x86_emulate_ctxt *ctxt, int base_reg)
1056 {
1057 	if (base_reg == VCPU_REGS_RSP || base_reg == VCPU_REGS_RBP)
1058 		ctxt->modrm_seg = VCPU_SREG_SS;
1059 }
1060 
1061 static int decode_modrm(struct x86_emulate_ctxt *ctxt,
1062 			struct operand *op)
1063 {
1064 	u8 sib;
1065 	int index_reg = 0, base_reg = 0, scale;
1066 	int rc = X86EMUL_CONTINUE;
1067 	ulong modrm_ea = 0;
1068 
1069 	if (ctxt->rex_prefix) {
1070 		ctxt->modrm_reg = (ctxt->rex_prefix & 4) << 1;	/* REX.R */
1071 		index_reg = (ctxt->rex_prefix & 2) << 2; /* REX.X */
1072 		ctxt->modrm_rm = base_reg = (ctxt->rex_prefix & 1) << 3; /* REG.B */
1073 	}
1074 
1075 	ctxt->modrm_mod |= (ctxt->modrm & 0xc0) >> 6;
1076 	ctxt->modrm_reg |= (ctxt->modrm & 0x38) >> 3;
1077 	ctxt->modrm_rm |= (ctxt->modrm & 0x07);
1078 	ctxt->modrm_seg = VCPU_SREG_DS;
1079 
1080 	if (ctxt->modrm_mod == 3) {
1081 		op->type = OP_REG;
1082 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
1083 		op->addr.reg = decode_register(ctxt, ctxt->modrm_rm,
1084 				ctxt->d & ByteOp);
1085 		if (ctxt->d & Sse) {
1086 			op->type = OP_XMM;
1087 			op->bytes = 16;
1088 			op->addr.xmm = ctxt->modrm_rm;
1089 			read_sse_reg(ctxt, &op->vec_val, ctxt->modrm_rm);
1090 			return rc;
1091 		}
1092 		if (ctxt->d & Mmx) {
1093 			op->type = OP_MM;
1094 			op->bytes = 8;
1095 			op->addr.xmm = ctxt->modrm_rm & 7;
1096 			return rc;
1097 		}
1098 		fetch_register_operand(op);
1099 		return rc;
1100 	}
1101 
1102 	op->type = OP_MEM;
1103 
1104 	if (ctxt->ad_bytes == 2) {
1105 		unsigned bx = reg_read(ctxt, VCPU_REGS_RBX);
1106 		unsigned bp = reg_read(ctxt, VCPU_REGS_RBP);
1107 		unsigned si = reg_read(ctxt, VCPU_REGS_RSI);
1108 		unsigned di = reg_read(ctxt, VCPU_REGS_RDI);
1109 
1110 		/* 16-bit ModR/M decode. */
1111 		switch (ctxt->modrm_mod) {
1112 		case 0:
1113 			if (ctxt->modrm_rm == 6)
1114 				modrm_ea += insn_fetch(u16, ctxt);
1115 			break;
1116 		case 1:
1117 			modrm_ea += insn_fetch(s8, ctxt);
1118 			break;
1119 		case 2:
1120 			modrm_ea += insn_fetch(u16, ctxt);
1121 			break;
1122 		}
1123 		switch (ctxt->modrm_rm) {
1124 		case 0:
1125 			modrm_ea += bx + si;
1126 			break;
1127 		case 1:
1128 			modrm_ea += bx + di;
1129 			break;
1130 		case 2:
1131 			modrm_ea += bp + si;
1132 			break;
1133 		case 3:
1134 			modrm_ea += bp + di;
1135 			break;
1136 		case 4:
1137 			modrm_ea += si;
1138 			break;
1139 		case 5:
1140 			modrm_ea += di;
1141 			break;
1142 		case 6:
1143 			if (ctxt->modrm_mod != 0)
1144 				modrm_ea += bp;
1145 			break;
1146 		case 7:
1147 			modrm_ea += bx;
1148 			break;
1149 		}
1150 		if (ctxt->modrm_rm == 2 || ctxt->modrm_rm == 3 ||
1151 		    (ctxt->modrm_rm == 6 && ctxt->modrm_mod != 0))
1152 			ctxt->modrm_seg = VCPU_SREG_SS;
1153 		modrm_ea = (u16)modrm_ea;
1154 	} else {
1155 		/* 32/64-bit ModR/M decode. */
1156 		if ((ctxt->modrm_rm & 7) == 4) {
1157 			sib = insn_fetch(u8, ctxt);
1158 			index_reg |= (sib >> 3) & 7;
1159 			base_reg |= sib & 7;
1160 			scale = sib >> 6;
1161 
1162 			if ((base_reg & 7) == 5 && ctxt->modrm_mod == 0)
1163 				modrm_ea += insn_fetch(s32, ctxt);
1164 			else {
1165 				modrm_ea += reg_read(ctxt, base_reg);
1166 				adjust_modrm_seg(ctxt, base_reg);
1167 			}
1168 			if (index_reg != 4)
1169 				modrm_ea += reg_read(ctxt, index_reg) << scale;
1170 		} else if ((ctxt->modrm_rm & 7) == 5 && ctxt->modrm_mod == 0) {
1171 			if (ctxt->mode == X86EMUL_MODE_PROT64)
1172 				ctxt->rip_relative = 1;
1173 		} else {
1174 			base_reg = ctxt->modrm_rm;
1175 			modrm_ea += reg_read(ctxt, base_reg);
1176 			adjust_modrm_seg(ctxt, base_reg);
1177 		}
1178 		switch (ctxt->modrm_mod) {
1179 		case 0:
1180 			if (ctxt->modrm_rm == 5)
1181 				modrm_ea += insn_fetch(s32, ctxt);
1182 			break;
1183 		case 1:
1184 			modrm_ea += insn_fetch(s8, ctxt);
1185 			break;
1186 		case 2:
1187 			modrm_ea += insn_fetch(s32, ctxt);
1188 			break;
1189 		}
1190 	}
1191 	op->addr.mem.ea = modrm_ea;
1192 done:
1193 	return rc;
1194 }
1195 
1196 static int decode_abs(struct x86_emulate_ctxt *ctxt,
1197 		      struct operand *op)
1198 {
1199 	int rc = X86EMUL_CONTINUE;
1200 
1201 	op->type = OP_MEM;
1202 	switch (ctxt->ad_bytes) {
1203 	case 2:
1204 		op->addr.mem.ea = insn_fetch(u16, ctxt);
1205 		break;
1206 	case 4:
1207 		op->addr.mem.ea = insn_fetch(u32, ctxt);
1208 		break;
1209 	case 8:
1210 		op->addr.mem.ea = insn_fetch(u64, ctxt);
1211 		break;
1212 	}
1213 done:
1214 	return rc;
1215 }
1216 
1217 static void fetch_bit_operand(struct x86_emulate_ctxt *ctxt)
1218 {
1219 	long sv = 0, mask;
1220 
1221 	if (ctxt->dst.type == OP_MEM && ctxt->src.type == OP_REG) {
1222 		mask = ~(ctxt->dst.bytes * 8 - 1);
1223 
1224 		if (ctxt->src.bytes == 2)
1225 			sv = (s16)ctxt->src.val & (s16)mask;
1226 		else if (ctxt->src.bytes == 4)
1227 			sv = (s32)ctxt->src.val & (s32)mask;
1228 
1229 		ctxt->dst.addr.mem.ea += (sv >> 3);
1230 	}
1231 
1232 	/* only subword offset */
1233 	ctxt->src.val &= (ctxt->dst.bytes << 3) - 1;
1234 }
1235 
1236 static int read_emulated(struct x86_emulate_ctxt *ctxt,
1237 			 unsigned long addr, void *dest, unsigned size)
1238 {
1239 	int rc;
1240 	struct read_cache *mc = &ctxt->mem_read;
1241 
1242 	if (mc->pos < mc->end)
1243 		goto read_cached;
1244 
1245 	WARN_ON((mc->end + size) >= sizeof(mc->data));
1246 
1247 	rc = ctxt->ops->read_emulated(ctxt, addr, mc->data + mc->end, size,
1248 				      &ctxt->exception);
1249 	if (rc != X86EMUL_CONTINUE)
1250 		return rc;
1251 
1252 	mc->end += size;
1253 
1254 read_cached:
1255 	memcpy(dest, mc->data + mc->pos, size);
1256 	mc->pos += size;
1257 	return X86EMUL_CONTINUE;
1258 }
1259 
1260 static int segmented_read(struct x86_emulate_ctxt *ctxt,
1261 			  struct segmented_address addr,
1262 			  void *data,
1263 			  unsigned size)
1264 {
1265 	int rc;
1266 	ulong linear;
1267 
1268 	rc = linearize(ctxt, addr, size, false, &linear);
1269 	if (rc != X86EMUL_CONTINUE)
1270 		return rc;
1271 	return read_emulated(ctxt, linear, data, size);
1272 }
1273 
1274 static int segmented_write(struct x86_emulate_ctxt *ctxt,
1275 			   struct segmented_address addr,
1276 			   const void *data,
1277 			   unsigned size)
1278 {
1279 	int rc;
1280 	ulong linear;
1281 
1282 	rc = linearize(ctxt, addr, size, true, &linear);
1283 	if (rc != X86EMUL_CONTINUE)
1284 		return rc;
1285 	return ctxt->ops->write_emulated(ctxt, linear, data, size,
1286 					 &ctxt->exception);
1287 }
1288 
1289 static int segmented_cmpxchg(struct x86_emulate_ctxt *ctxt,
1290 			     struct segmented_address addr,
1291 			     const void *orig_data, const void *data,
1292 			     unsigned size)
1293 {
1294 	int rc;
1295 	ulong linear;
1296 
1297 	rc = linearize(ctxt, addr, size, true, &linear);
1298 	if (rc != X86EMUL_CONTINUE)
1299 		return rc;
1300 	return ctxt->ops->cmpxchg_emulated(ctxt, linear, orig_data, data,
1301 					   size, &ctxt->exception);
1302 }
1303 
1304 static int pio_in_emulated(struct x86_emulate_ctxt *ctxt,
1305 			   unsigned int size, unsigned short port,
1306 			   void *dest)
1307 {
1308 	struct read_cache *rc = &ctxt->io_read;
1309 
1310 	if (rc->pos == rc->end) { /* refill pio read ahead */
1311 		unsigned int in_page, n;
1312 		unsigned int count = ctxt->rep_prefix ?
1313 			address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) : 1;
1314 		in_page = (ctxt->eflags & EFLG_DF) ?
1315 			offset_in_page(reg_read(ctxt, VCPU_REGS_RDI)) :
1316 			PAGE_SIZE - offset_in_page(reg_read(ctxt, VCPU_REGS_RDI));
1317 		n = min(min(in_page, (unsigned int)sizeof(rc->data)) / size,
1318 			count);
1319 		if (n == 0)
1320 			n = 1;
1321 		rc->pos = rc->end = 0;
1322 		if (!ctxt->ops->pio_in_emulated(ctxt, size, port, rc->data, n))
1323 			return 0;
1324 		rc->end = n * size;
1325 	}
1326 
1327 	if (ctxt->rep_prefix && !(ctxt->eflags & EFLG_DF)) {
1328 		ctxt->dst.data = rc->data + rc->pos;
1329 		ctxt->dst.type = OP_MEM_STR;
1330 		ctxt->dst.count = (rc->end - rc->pos) / size;
1331 		rc->pos = rc->end;
1332 	} else {
1333 		memcpy(dest, rc->data + rc->pos, size);
1334 		rc->pos += size;
1335 	}
1336 	return 1;
1337 }
1338 
1339 static int read_interrupt_descriptor(struct x86_emulate_ctxt *ctxt,
1340 				     u16 index, struct desc_struct *desc)
1341 {
1342 	struct desc_ptr dt;
1343 	ulong addr;
1344 
1345 	ctxt->ops->get_idt(ctxt, &dt);
1346 
1347 	if (dt.size < index * 8 + 7)
1348 		return emulate_gp(ctxt, index << 3 | 0x2);
1349 
1350 	addr = dt.address + index * 8;
1351 	return ctxt->ops->read_std(ctxt, addr, desc, sizeof *desc,
1352 				   &ctxt->exception);
1353 }
1354 
1355 static void get_descriptor_table_ptr(struct x86_emulate_ctxt *ctxt,
1356 				     u16 selector, struct desc_ptr *dt)
1357 {
1358 	const struct x86_emulate_ops *ops = ctxt->ops;
1359 
1360 	if (selector & 1 << 2) {
1361 		struct desc_struct desc;
1362 		u16 sel;
1363 
1364 		memset (dt, 0, sizeof *dt);
1365 		if (!ops->get_segment(ctxt, &sel, &desc, NULL, VCPU_SREG_LDTR))
1366 			return;
1367 
1368 		dt->size = desc_limit_scaled(&desc); /* what if limit > 65535? */
1369 		dt->address = get_desc_base(&desc);
1370 	} else
1371 		ops->get_gdt(ctxt, dt);
1372 }
1373 
1374 /* allowed just for 8 bytes segments */
1375 static int read_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1376 				   u16 selector, struct desc_struct *desc,
1377 				   ulong *desc_addr_p)
1378 {
1379 	struct desc_ptr dt;
1380 	u16 index = selector >> 3;
1381 	ulong addr;
1382 
1383 	get_descriptor_table_ptr(ctxt, selector, &dt);
1384 
1385 	if (dt.size < index * 8 + 7)
1386 		return emulate_gp(ctxt, selector & 0xfffc);
1387 
1388 	*desc_addr_p = addr = dt.address + index * 8;
1389 	return ctxt->ops->read_std(ctxt, addr, desc, sizeof *desc,
1390 				   &ctxt->exception);
1391 }
1392 
1393 /* allowed just for 8 bytes segments */
1394 static int write_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1395 				    u16 selector, struct desc_struct *desc)
1396 {
1397 	struct desc_ptr dt;
1398 	u16 index = selector >> 3;
1399 	ulong addr;
1400 
1401 	get_descriptor_table_ptr(ctxt, selector, &dt);
1402 
1403 	if (dt.size < index * 8 + 7)
1404 		return emulate_gp(ctxt, selector & 0xfffc);
1405 
1406 	addr = dt.address + index * 8;
1407 	return ctxt->ops->write_std(ctxt, addr, desc, sizeof *desc,
1408 				    &ctxt->exception);
1409 }
1410 
1411 /* Does not support long mode */
1412 static int load_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1413 				   u16 selector, int seg)
1414 {
1415 	struct desc_struct seg_desc, old_desc;
1416 	u8 dpl, rpl, cpl;
1417 	unsigned err_vec = GP_VECTOR;
1418 	u32 err_code = 0;
1419 	bool null_selector = !(selector & ~0x3); /* 0000-0003 are null */
1420 	ulong desc_addr;
1421 	int ret;
1422 	u16 dummy;
1423 
1424 	memset(&seg_desc, 0, sizeof seg_desc);
1425 
1426 	if (ctxt->mode == X86EMUL_MODE_REAL) {
1427 		/* set real mode segment descriptor (keep limit etc. for
1428 		 * unreal mode) */
1429 		ctxt->ops->get_segment(ctxt, &dummy, &seg_desc, NULL, seg);
1430 		set_desc_base(&seg_desc, selector << 4);
1431 		goto load;
1432 	} else if (seg <= VCPU_SREG_GS && ctxt->mode == X86EMUL_MODE_VM86) {
1433 		/* VM86 needs a clean new segment descriptor */
1434 		set_desc_base(&seg_desc, selector << 4);
1435 		set_desc_limit(&seg_desc, 0xffff);
1436 		seg_desc.type = 3;
1437 		seg_desc.p = 1;
1438 		seg_desc.s = 1;
1439 		seg_desc.dpl = 3;
1440 		goto load;
1441 	}
1442 
1443 	rpl = selector & 3;
1444 	cpl = ctxt->ops->cpl(ctxt);
1445 
1446 	/* NULL selector is not valid for TR, CS and SS (except for long mode) */
1447 	if ((seg == VCPU_SREG_CS
1448 	     || (seg == VCPU_SREG_SS
1449 		 && (ctxt->mode != X86EMUL_MODE_PROT64 || rpl != cpl))
1450 	     || seg == VCPU_SREG_TR)
1451 	    && null_selector)
1452 		goto exception;
1453 
1454 	/* TR should be in GDT only */
1455 	if (seg == VCPU_SREG_TR && (selector & (1 << 2)))
1456 		goto exception;
1457 
1458 	if (null_selector) /* for NULL selector skip all following checks */
1459 		goto load;
1460 
1461 	ret = read_segment_descriptor(ctxt, selector, &seg_desc, &desc_addr);
1462 	if (ret != X86EMUL_CONTINUE)
1463 		return ret;
1464 
1465 	err_code = selector & 0xfffc;
1466 	err_vec = GP_VECTOR;
1467 
1468 	/* can't load system descriptor into segment selector */
1469 	if (seg <= VCPU_SREG_GS && !seg_desc.s)
1470 		goto exception;
1471 
1472 	if (!seg_desc.p) {
1473 		err_vec = (seg == VCPU_SREG_SS) ? SS_VECTOR : NP_VECTOR;
1474 		goto exception;
1475 	}
1476 
1477 	dpl = seg_desc.dpl;
1478 
1479 	switch (seg) {
1480 	case VCPU_SREG_SS:
1481 		/*
1482 		 * segment is not a writable data segment or segment
1483 		 * selector's RPL != CPL or segment selector's RPL != CPL
1484 		 */
1485 		if (rpl != cpl || (seg_desc.type & 0xa) != 0x2 || dpl != cpl)
1486 			goto exception;
1487 		break;
1488 	case VCPU_SREG_CS:
1489 		if (!(seg_desc.type & 8))
1490 			goto exception;
1491 
1492 		if (seg_desc.type & 4) {
1493 			/* conforming */
1494 			if (dpl > cpl)
1495 				goto exception;
1496 		} else {
1497 			/* nonconforming */
1498 			if (rpl > cpl || dpl != cpl)
1499 				goto exception;
1500 		}
1501 		/* CS(RPL) <- CPL */
1502 		selector = (selector & 0xfffc) | cpl;
1503 		break;
1504 	case VCPU_SREG_TR:
1505 		if (seg_desc.s || (seg_desc.type != 1 && seg_desc.type != 9))
1506 			goto exception;
1507 		old_desc = seg_desc;
1508 		seg_desc.type |= 2; /* busy */
1509 		ret = ctxt->ops->cmpxchg_emulated(ctxt, desc_addr, &old_desc, &seg_desc,
1510 						  sizeof(seg_desc), &ctxt->exception);
1511 		if (ret != X86EMUL_CONTINUE)
1512 			return ret;
1513 		break;
1514 	case VCPU_SREG_LDTR:
1515 		if (seg_desc.s || seg_desc.type != 2)
1516 			goto exception;
1517 		break;
1518 	default: /*  DS, ES, FS, or GS */
1519 		/*
1520 		 * segment is not a data or readable code segment or
1521 		 * ((segment is a data or nonconforming code segment)
1522 		 * and (both RPL and CPL > DPL))
1523 		 */
1524 		if ((seg_desc.type & 0xa) == 0x8 ||
1525 		    (((seg_desc.type & 0xc) != 0xc) &&
1526 		     (rpl > dpl && cpl > dpl)))
1527 			goto exception;
1528 		break;
1529 	}
1530 
1531 	if (seg_desc.s) {
1532 		/* mark segment as accessed */
1533 		seg_desc.type |= 1;
1534 		ret = write_segment_descriptor(ctxt, selector, &seg_desc);
1535 		if (ret != X86EMUL_CONTINUE)
1536 			return ret;
1537 	}
1538 load:
1539 	ctxt->ops->set_segment(ctxt, selector, &seg_desc, 0, seg);
1540 	return X86EMUL_CONTINUE;
1541 exception:
1542 	emulate_exception(ctxt, err_vec, err_code, true);
1543 	return X86EMUL_PROPAGATE_FAULT;
1544 }
1545 
1546 static void write_register_operand(struct operand *op)
1547 {
1548 	/* The 4-byte case *is* correct: in 64-bit mode we zero-extend. */
1549 	switch (op->bytes) {
1550 	case 1:
1551 		*(u8 *)op->addr.reg = (u8)op->val;
1552 		break;
1553 	case 2:
1554 		*(u16 *)op->addr.reg = (u16)op->val;
1555 		break;
1556 	case 4:
1557 		*op->addr.reg = (u32)op->val;
1558 		break;	/* 64b: zero-extend */
1559 	case 8:
1560 		*op->addr.reg = op->val;
1561 		break;
1562 	}
1563 }
1564 
1565 static int writeback(struct x86_emulate_ctxt *ctxt, struct operand *op)
1566 {
1567 	int rc;
1568 
1569 	switch (op->type) {
1570 	case OP_REG:
1571 		write_register_operand(op);
1572 		break;
1573 	case OP_MEM:
1574 		if (ctxt->lock_prefix)
1575 			rc = segmented_cmpxchg(ctxt,
1576 					       op->addr.mem,
1577 					       &op->orig_val,
1578 					       &op->val,
1579 					       op->bytes);
1580 		else
1581 			rc = segmented_write(ctxt,
1582 					     op->addr.mem,
1583 					     &op->val,
1584 					     op->bytes);
1585 		if (rc != X86EMUL_CONTINUE)
1586 			return rc;
1587 		break;
1588 	case OP_MEM_STR:
1589 		rc = segmented_write(ctxt,
1590 				op->addr.mem,
1591 				op->data,
1592 				op->bytes * op->count);
1593 		if (rc != X86EMUL_CONTINUE)
1594 			return rc;
1595 		break;
1596 	case OP_XMM:
1597 		write_sse_reg(ctxt, &op->vec_val, op->addr.xmm);
1598 		break;
1599 	case OP_MM:
1600 		write_mmx_reg(ctxt, &op->mm_val, op->addr.mm);
1601 		break;
1602 	case OP_NONE:
1603 		/* no writeback */
1604 		break;
1605 	default:
1606 		break;
1607 	}
1608 	return X86EMUL_CONTINUE;
1609 }
1610 
1611 static int push(struct x86_emulate_ctxt *ctxt, void *data, int bytes)
1612 {
1613 	struct segmented_address addr;
1614 
1615 	rsp_increment(ctxt, -bytes);
1616 	addr.ea = reg_read(ctxt, VCPU_REGS_RSP) & stack_mask(ctxt);
1617 	addr.seg = VCPU_SREG_SS;
1618 
1619 	return segmented_write(ctxt, addr, data, bytes);
1620 }
1621 
1622 static int em_push(struct x86_emulate_ctxt *ctxt)
1623 {
1624 	/* Disable writeback. */
1625 	ctxt->dst.type = OP_NONE;
1626 	return push(ctxt, &ctxt->src.val, ctxt->op_bytes);
1627 }
1628 
1629 static int emulate_pop(struct x86_emulate_ctxt *ctxt,
1630 		       void *dest, int len)
1631 {
1632 	int rc;
1633 	struct segmented_address addr;
1634 
1635 	addr.ea = reg_read(ctxt, VCPU_REGS_RSP) & stack_mask(ctxt);
1636 	addr.seg = VCPU_SREG_SS;
1637 	rc = segmented_read(ctxt, addr, dest, len);
1638 	if (rc != X86EMUL_CONTINUE)
1639 		return rc;
1640 
1641 	rsp_increment(ctxt, len);
1642 	return rc;
1643 }
1644 
1645 static int em_pop(struct x86_emulate_ctxt *ctxt)
1646 {
1647 	return emulate_pop(ctxt, &ctxt->dst.val, ctxt->op_bytes);
1648 }
1649 
1650 static int emulate_popf(struct x86_emulate_ctxt *ctxt,
1651 			void *dest, int len)
1652 {
1653 	int rc;
1654 	unsigned long val, change_mask;
1655 	int iopl = (ctxt->eflags & X86_EFLAGS_IOPL) >> IOPL_SHIFT;
1656 	int cpl = ctxt->ops->cpl(ctxt);
1657 
1658 	rc = emulate_pop(ctxt, &val, len);
1659 	if (rc != X86EMUL_CONTINUE)
1660 		return rc;
1661 
1662 	change_mask = EFLG_CF | EFLG_PF | EFLG_AF | EFLG_ZF | EFLG_SF | EFLG_OF
1663 		| EFLG_TF | EFLG_DF | EFLG_NT | EFLG_RF | EFLG_AC | EFLG_ID;
1664 
1665 	switch(ctxt->mode) {
1666 	case X86EMUL_MODE_PROT64:
1667 	case X86EMUL_MODE_PROT32:
1668 	case X86EMUL_MODE_PROT16:
1669 		if (cpl == 0)
1670 			change_mask |= EFLG_IOPL;
1671 		if (cpl <= iopl)
1672 			change_mask |= EFLG_IF;
1673 		break;
1674 	case X86EMUL_MODE_VM86:
1675 		if (iopl < 3)
1676 			return emulate_gp(ctxt, 0);
1677 		change_mask |= EFLG_IF;
1678 		break;
1679 	default: /* real mode */
1680 		change_mask |= (EFLG_IOPL | EFLG_IF);
1681 		break;
1682 	}
1683 
1684 	*(unsigned long *)dest =
1685 		(ctxt->eflags & ~change_mask) | (val & change_mask);
1686 
1687 	return rc;
1688 }
1689 
1690 static int em_popf(struct x86_emulate_ctxt *ctxt)
1691 {
1692 	ctxt->dst.type = OP_REG;
1693 	ctxt->dst.addr.reg = &ctxt->eflags;
1694 	ctxt->dst.bytes = ctxt->op_bytes;
1695 	return emulate_popf(ctxt, &ctxt->dst.val, ctxt->op_bytes);
1696 }
1697 
1698 static int em_enter(struct x86_emulate_ctxt *ctxt)
1699 {
1700 	int rc;
1701 	unsigned frame_size = ctxt->src.val;
1702 	unsigned nesting_level = ctxt->src2.val & 31;
1703 	ulong rbp;
1704 
1705 	if (nesting_level)
1706 		return X86EMUL_UNHANDLEABLE;
1707 
1708 	rbp = reg_read(ctxt, VCPU_REGS_RBP);
1709 	rc = push(ctxt, &rbp, stack_size(ctxt));
1710 	if (rc != X86EMUL_CONTINUE)
1711 		return rc;
1712 	assign_masked(reg_rmw(ctxt, VCPU_REGS_RBP), reg_read(ctxt, VCPU_REGS_RSP),
1713 		      stack_mask(ctxt));
1714 	assign_masked(reg_rmw(ctxt, VCPU_REGS_RSP),
1715 		      reg_read(ctxt, VCPU_REGS_RSP) - frame_size,
1716 		      stack_mask(ctxt));
1717 	return X86EMUL_CONTINUE;
1718 }
1719 
1720 static int em_leave(struct x86_emulate_ctxt *ctxt)
1721 {
1722 	assign_masked(reg_rmw(ctxt, VCPU_REGS_RSP), reg_read(ctxt, VCPU_REGS_RBP),
1723 		      stack_mask(ctxt));
1724 	return emulate_pop(ctxt, reg_rmw(ctxt, VCPU_REGS_RBP), ctxt->op_bytes);
1725 }
1726 
1727 static int em_push_sreg(struct x86_emulate_ctxt *ctxt)
1728 {
1729 	int seg = ctxt->src2.val;
1730 
1731 	ctxt->src.val = get_segment_selector(ctxt, seg);
1732 
1733 	return em_push(ctxt);
1734 }
1735 
1736 static int em_pop_sreg(struct x86_emulate_ctxt *ctxt)
1737 {
1738 	int seg = ctxt->src2.val;
1739 	unsigned long selector;
1740 	int rc;
1741 
1742 	rc = emulate_pop(ctxt, &selector, ctxt->op_bytes);
1743 	if (rc != X86EMUL_CONTINUE)
1744 		return rc;
1745 
1746 	rc = load_segment_descriptor(ctxt, (u16)selector, seg);
1747 	return rc;
1748 }
1749 
1750 static int em_pusha(struct x86_emulate_ctxt *ctxt)
1751 {
1752 	unsigned long old_esp = reg_read(ctxt, VCPU_REGS_RSP);
1753 	int rc = X86EMUL_CONTINUE;
1754 	int reg = VCPU_REGS_RAX;
1755 
1756 	while (reg <= VCPU_REGS_RDI) {
1757 		(reg == VCPU_REGS_RSP) ?
1758 		(ctxt->src.val = old_esp) : (ctxt->src.val = reg_read(ctxt, reg));
1759 
1760 		rc = em_push(ctxt);
1761 		if (rc != X86EMUL_CONTINUE)
1762 			return rc;
1763 
1764 		++reg;
1765 	}
1766 
1767 	return rc;
1768 }
1769 
1770 static int em_pushf(struct x86_emulate_ctxt *ctxt)
1771 {
1772 	ctxt->src.val =  (unsigned long)ctxt->eflags;
1773 	return em_push(ctxt);
1774 }
1775 
1776 static int em_popa(struct x86_emulate_ctxt *ctxt)
1777 {
1778 	int rc = X86EMUL_CONTINUE;
1779 	int reg = VCPU_REGS_RDI;
1780 
1781 	while (reg >= VCPU_REGS_RAX) {
1782 		if (reg == VCPU_REGS_RSP) {
1783 			rsp_increment(ctxt, ctxt->op_bytes);
1784 			--reg;
1785 		}
1786 
1787 		rc = emulate_pop(ctxt, reg_rmw(ctxt, reg), ctxt->op_bytes);
1788 		if (rc != X86EMUL_CONTINUE)
1789 			break;
1790 		--reg;
1791 	}
1792 	return rc;
1793 }
1794 
1795 static int __emulate_int_real(struct x86_emulate_ctxt *ctxt, int irq)
1796 {
1797 	const struct x86_emulate_ops *ops = ctxt->ops;
1798 	int rc;
1799 	struct desc_ptr dt;
1800 	gva_t cs_addr;
1801 	gva_t eip_addr;
1802 	u16 cs, eip;
1803 
1804 	/* TODO: Add limit checks */
1805 	ctxt->src.val = ctxt->eflags;
1806 	rc = em_push(ctxt);
1807 	if (rc != X86EMUL_CONTINUE)
1808 		return rc;
1809 
1810 	ctxt->eflags &= ~(EFLG_IF | EFLG_TF | EFLG_AC);
1811 
1812 	ctxt->src.val = get_segment_selector(ctxt, VCPU_SREG_CS);
1813 	rc = em_push(ctxt);
1814 	if (rc != X86EMUL_CONTINUE)
1815 		return rc;
1816 
1817 	ctxt->src.val = ctxt->_eip;
1818 	rc = em_push(ctxt);
1819 	if (rc != X86EMUL_CONTINUE)
1820 		return rc;
1821 
1822 	ops->get_idt(ctxt, &dt);
1823 
1824 	eip_addr = dt.address + (irq << 2);
1825 	cs_addr = dt.address + (irq << 2) + 2;
1826 
1827 	rc = ops->read_std(ctxt, cs_addr, &cs, 2, &ctxt->exception);
1828 	if (rc != X86EMUL_CONTINUE)
1829 		return rc;
1830 
1831 	rc = ops->read_std(ctxt, eip_addr, &eip, 2, &ctxt->exception);
1832 	if (rc != X86EMUL_CONTINUE)
1833 		return rc;
1834 
1835 	rc = load_segment_descriptor(ctxt, cs, VCPU_SREG_CS);
1836 	if (rc != X86EMUL_CONTINUE)
1837 		return rc;
1838 
1839 	ctxt->_eip = eip;
1840 
1841 	return rc;
1842 }
1843 
1844 int emulate_int_real(struct x86_emulate_ctxt *ctxt, int irq)
1845 {
1846 	int rc;
1847 
1848 	invalidate_registers(ctxt);
1849 	rc = __emulate_int_real(ctxt, irq);
1850 	if (rc == X86EMUL_CONTINUE)
1851 		writeback_registers(ctxt);
1852 	return rc;
1853 }
1854 
1855 static int emulate_int(struct x86_emulate_ctxt *ctxt, int irq)
1856 {
1857 	switch(ctxt->mode) {
1858 	case X86EMUL_MODE_REAL:
1859 		return __emulate_int_real(ctxt, irq);
1860 	case X86EMUL_MODE_VM86:
1861 	case X86EMUL_MODE_PROT16:
1862 	case X86EMUL_MODE_PROT32:
1863 	case X86EMUL_MODE_PROT64:
1864 	default:
1865 		/* Protected mode interrupts unimplemented yet */
1866 		return X86EMUL_UNHANDLEABLE;
1867 	}
1868 }
1869 
1870 static int emulate_iret_real(struct x86_emulate_ctxt *ctxt)
1871 {
1872 	int rc = X86EMUL_CONTINUE;
1873 	unsigned long temp_eip = 0;
1874 	unsigned long temp_eflags = 0;
1875 	unsigned long cs = 0;
1876 	unsigned long mask = EFLG_CF | EFLG_PF | EFLG_AF | EFLG_ZF | EFLG_SF | EFLG_TF |
1877 			     EFLG_IF | EFLG_DF | EFLG_OF | EFLG_IOPL | EFLG_NT | EFLG_RF |
1878 			     EFLG_AC | EFLG_ID | (1 << 1); /* Last one is the reserved bit */
1879 	unsigned long vm86_mask = EFLG_VM | EFLG_VIF | EFLG_VIP;
1880 
1881 	/* TODO: Add stack limit check */
1882 
1883 	rc = emulate_pop(ctxt, &temp_eip, ctxt->op_bytes);
1884 
1885 	if (rc != X86EMUL_CONTINUE)
1886 		return rc;
1887 
1888 	if (temp_eip & ~0xffff)
1889 		return emulate_gp(ctxt, 0);
1890 
1891 	rc = emulate_pop(ctxt, &cs, ctxt->op_bytes);
1892 
1893 	if (rc != X86EMUL_CONTINUE)
1894 		return rc;
1895 
1896 	rc = emulate_pop(ctxt, &temp_eflags, ctxt->op_bytes);
1897 
1898 	if (rc != X86EMUL_CONTINUE)
1899 		return rc;
1900 
1901 	rc = load_segment_descriptor(ctxt, (u16)cs, VCPU_SREG_CS);
1902 
1903 	if (rc != X86EMUL_CONTINUE)
1904 		return rc;
1905 
1906 	ctxt->_eip = temp_eip;
1907 
1908 
1909 	if (ctxt->op_bytes == 4)
1910 		ctxt->eflags = ((temp_eflags & mask) | (ctxt->eflags & vm86_mask));
1911 	else if (ctxt->op_bytes == 2) {
1912 		ctxt->eflags &= ~0xffff;
1913 		ctxt->eflags |= temp_eflags;
1914 	}
1915 
1916 	ctxt->eflags &= ~EFLG_RESERVED_ZEROS_MASK; /* Clear reserved zeros */
1917 	ctxt->eflags |= EFLG_RESERVED_ONE_MASK;
1918 
1919 	return rc;
1920 }
1921 
1922 static int em_iret(struct x86_emulate_ctxt *ctxt)
1923 {
1924 	switch(ctxt->mode) {
1925 	case X86EMUL_MODE_REAL:
1926 		return emulate_iret_real(ctxt);
1927 	case X86EMUL_MODE_VM86:
1928 	case X86EMUL_MODE_PROT16:
1929 	case X86EMUL_MODE_PROT32:
1930 	case X86EMUL_MODE_PROT64:
1931 	default:
1932 		/* iret from protected mode unimplemented yet */
1933 		return X86EMUL_UNHANDLEABLE;
1934 	}
1935 }
1936 
1937 static int em_jmp_far(struct x86_emulate_ctxt *ctxt)
1938 {
1939 	int rc;
1940 	unsigned short sel;
1941 
1942 	memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
1943 
1944 	rc = load_segment_descriptor(ctxt, sel, VCPU_SREG_CS);
1945 	if (rc != X86EMUL_CONTINUE)
1946 		return rc;
1947 
1948 	ctxt->_eip = 0;
1949 	memcpy(&ctxt->_eip, ctxt->src.valptr, ctxt->op_bytes);
1950 	return X86EMUL_CONTINUE;
1951 }
1952 
1953 static int em_grp45(struct x86_emulate_ctxt *ctxt)
1954 {
1955 	int rc = X86EMUL_CONTINUE;
1956 
1957 	switch (ctxt->modrm_reg) {
1958 	case 2: /* call near abs */ {
1959 		long int old_eip;
1960 		old_eip = ctxt->_eip;
1961 		ctxt->_eip = ctxt->src.val;
1962 		ctxt->src.val = old_eip;
1963 		rc = em_push(ctxt);
1964 		break;
1965 	}
1966 	case 4: /* jmp abs */
1967 		ctxt->_eip = ctxt->src.val;
1968 		break;
1969 	case 5: /* jmp far */
1970 		rc = em_jmp_far(ctxt);
1971 		break;
1972 	case 6:	/* push */
1973 		rc = em_push(ctxt);
1974 		break;
1975 	}
1976 	return rc;
1977 }
1978 
1979 static int em_cmpxchg8b(struct x86_emulate_ctxt *ctxt)
1980 {
1981 	u64 old = ctxt->dst.orig_val64;
1982 
1983 	if (((u32) (old >> 0) != (u32) reg_read(ctxt, VCPU_REGS_RAX)) ||
1984 	    ((u32) (old >> 32) != (u32) reg_read(ctxt, VCPU_REGS_RDX))) {
1985 		*reg_write(ctxt, VCPU_REGS_RAX) = (u32) (old >> 0);
1986 		*reg_write(ctxt, VCPU_REGS_RDX) = (u32) (old >> 32);
1987 		ctxt->eflags &= ~EFLG_ZF;
1988 	} else {
1989 		ctxt->dst.val64 = ((u64)reg_read(ctxt, VCPU_REGS_RCX) << 32) |
1990 			(u32) reg_read(ctxt, VCPU_REGS_RBX);
1991 
1992 		ctxt->eflags |= EFLG_ZF;
1993 	}
1994 	return X86EMUL_CONTINUE;
1995 }
1996 
1997 static int em_ret(struct x86_emulate_ctxt *ctxt)
1998 {
1999 	ctxt->dst.type = OP_REG;
2000 	ctxt->dst.addr.reg = &ctxt->_eip;
2001 	ctxt->dst.bytes = ctxt->op_bytes;
2002 	return em_pop(ctxt);
2003 }
2004 
2005 static int em_ret_far(struct x86_emulate_ctxt *ctxt)
2006 {
2007 	int rc;
2008 	unsigned long cs;
2009 
2010 	rc = emulate_pop(ctxt, &ctxt->_eip, ctxt->op_bytes);
2011 	if (rc != X86EMUL_CONTINUE)
2012 		return rc;
2013 	if (ctxt->op_bytes == 4)
2014 		ctxt->_eip = (u32)ctxt->_eip;
2015 	rc = emulate_pop(ctxt, &cs, ctxt->op_bytes);
2016 	if (rc != X86EMUL_CONTINUE)
2017 		return rc;
2018 	rc = load_segment_descriptor(ctxt, (u16)cs, VCPU_SREG_CS);
2019 	return rc;
2020 }
2021 
2022 static int em_ret_far_imm(struct x86_emulate_ctxt *ctxt)
2023 {
2024         int rc;
2025 
2026         rc = em_ret_far(ctxt);
2027         if (rc != X86EMUL_CONTINUE)
2028                 return rc;
2029         rsp_increment(ctxt, ctxt->src.val);
2030         return X86EMUL_CONTINUE;
2031 }
2032 
2033 static int em_cmpxchg(struct x86_emulate_ctxt *ctxt)
2034 {
2035 	/* Save real source value, then compare EAX against destination. */
2036 	ctxt->src.orig_val = ctxt->src.val;
2037 	ctxt->src.val = reg_read(ctxt, VCPU_REGS_RAX);
2038 	fastop(ctxt, em_cmp);
2039 
2040 	if (ctxt->eflags & EFLG_ZF) {
2041 		/* Success: write back to memory. */
2042 		ctxt->dst.val = ctxt->src.orig_val;
2043 	} else {
2044 		/* Failure: write the value we saw to EAX. */
2045 		ctxt->dst.type = OP_REG;
2046 		ctxt->dst.addr.reg = reg_rmw(ctxt, VCPU_REGS_RAX);
2047 	}
2048 	return X86EMUL_CONTINUE;
2049 }
2050 
2051 static int em_lseg(struct x86_emulate_ctxt *ctxt)
2052 {
2053 	int seg = ctxt->src2.val;
2054 	unsigned short sel;
2055 	int rc;
2056 
2057 	memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
2058 
2059 	rc = load_segment_descriptor(ctxt, sel, seg);
2060 	if (rc != X86EMUL_CONTINUE)
2061 		return rc;
2062 
2063 	ctxt->dst.val = ctxt->src.val;
2064 	return rc;
2065 }
2066 
2067 static void
2068 setup_syscalls_segments(struct x86_emulate_ctxt *ctxt,
2069 			struct desc_struct *cs, struct desc_struct *ss)
2070 {
2071 	cs->l = 0;		/* will be adjusted later */
2072 	set_desc_base(cs, 0);	/* flat segment */
2073 	cs->g = 1;		/* 4kb granularity */
2074 	set_desc_limit(cs, 0xfffff);	/* 4GB limit */
2075 	cs->type = 0x0b;	/* Read, Execute, Accessed */
2076 	cs->s = 1;
2077 	cs->dpl = 0;		/* will be adjusted later */
2078 	cs->p = 1;
2079 	cs->d = 1;
2080 	cs->avl = 0;
2081 
2082 	set_desc_base(ss, 0);	/* flat segment */
2083 	set_desc_limit(ss, 0xfffff);	/* 4GB limit */
2084 	ss->g = 1;		/* 4kb granularity */
2085 	ss->s = 1;
2086 	ss->type = 0x03;	/* Read/Write, Accessed */
2087 	ss->d = 1;		/* 32bit stack segment */
2088 	ss->dpl = 0;
2089 	ss->p = 1;
2090 	ss->l = 0;
2091 	ss->avl = 0;
2092 }
2093 
2094 static bool vendor_intel(struct x86_emulate_ctxt *ctxt)
2095 {
2096 	u32 eax, ebx, ecx, edx;
2097 
2098 	eax = ecx = 0;
2099 	ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
2100 	return ebx == X86EMUL_CPUID_VENDOR_GenuineIntel_ebx
2101 		&& ecx == X86EMUL_CPUID_VENDOR_GenuineIntel_ecx
2102 		&& edx == X86EMUL_CPUID_VENDOR_GenuineIntel_edx;
2103 }
2104 
2105 static bool em_syscall_is_enabled(struct x86_emulate_ctxt *ctxt)
2106 {
2107 	const struct x86_emulate_ops *ops = ctxt->ops;
2108 	u32 eax, ebx, ecx, edx;
2109 
2110 	/*
2111 	 * syscall should always be enabled in longmode - so only become
2112 	 * vendor specific (cpuid) if other modes are active...
2113 	 */
2114 	if (ctxt->mode == X86EMUL_MODE_PROT64)
2115 		return true;
2116 
2117 	eax = 0x00000000;
2118 	ecx = 0x00000000;
2119 	ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
2120 	/*
2121 	 * Intel ("GenuineIntel")
2122 	 * remark: Intel CPUs only support "syscall" in 64bit
2123 	 * longmode. Also an 64bit guest with a
2124 	 * 32bit compat-app running will #UD !! While this
2125 	 * behaviour can be fixed (by emulating) into AMD
2126 	 * response - CPUs of AMD can't behave like Intel.
2127 	 */
2128 	if (ebx == X86EMUL_CPUID_VENDOR_GenuineIntel_ebx &&
2129 	    ecx == X86EMUL_CPUID_VENDOR_GenuineIntel_ecx &&
2130 	    edx == X86EMUL_CPUID_VENDOR_GenuineIntel_edx)
2131 		return false;
2132 
2133 	/* AMD ("AuthenticAMD") */
2134 	if (ebx == X86EMUL_CPUID_VENDOR_AuthenticAMD_ebx &&
2135 	    ecx == X86EMUL_CPUID_VENDOR_AuthenticAMD_ecx &&
2136 	    edx == X86EMUL_CPUID_VENDOR_AuthenticAMD_edx)
2137 		return true;
2138 
2139 	/* AMD ("AMDisbetter!") */
2140 	if (ebx == X86EMUL_CPUID_VENDOR_AMDisbetterI_ebx &&
2141 	    ecx == X86EMUL_CPUID_VENDOR_AMDisbetterI_ecx &&
2142 	    edx == X86EMUL_CPUID_VENDOR_AMDisbetterI_edx)
2143 		return true;
2144 
2145 	/* default: (not Intel, not AMD), apply Intel's stricter rules... */
2146 	return false;
2147 }
2148 
2149 static int em_syscall(struct x86_emulate_ctxt *ctxt)
2150 {
2151 	const struct x86_emulate_ops *ops = ctxt->ops;
2152 	struct desc_struct cs, ss;
2153 	u64 msr_data;
2154 	u16 cs_sel, ss_sel;
2155 	u64 efer = 0;
2156 
2157 	/* syscall is not available in real mode */
2158 	if (ctxt->mode == X86EMUL_MODE_REAL ||
2159 	    ctxt->mode == X86EMUL_MODE_VM86)
2160 		return emulate_ud(ctxt);
2161 
2162 	if (!(em_syscall_is_enabled(ctxt)))
2163 		return emulate_ud(ctxt);
2164 
2165 	ops->get_msr(ctxt, MSR_EFER, &efer);
2166 	setup_syscalls_segments(ctxt, &cs, &ss);
2167 
2168 	if (!(efer & EFER_SCE))
2169 		return emulate_ud(ctxt);
2170 
2171 	ops->get_msr(ctxt, MSR_STAR, &msr_data);
2172 	msr_data >>= 32;
2173 	cs_sel = (u16)(msr_data & 0xfffc);
2174 	ss_sel = (u16)(msr_data + 8);
2175 
2176 	if (efer & EFER_LMA) {
2177 		cs.d = 0;
2178 		cs.l = 1;
2179 	}
2180 	ops->set_segment(ctxt, cs_sel, &cs, 0, VCPU_SREG_CS);
2181 	ops->set_segment(ctxt, ss_sel, &ss, 0, VCPU_SREG_SS);
2182 
2183 	*reg_write(ctxt, VCPU_REGS_RCX) = ctxt->_eip;
2184 	if (efer & EFER_LMA) {
2185 #ifdef CONFIG_X86_64
2186 		*reg_write(ctxt, VCPU_REGS_R11) = ctxt->eflags & ~EFLG_RF;
2187 
2188 		ops->get_msr(ctxt,
2189 			     ctxt->mode == X86EMUL_MODE_PROT64 ?
2190 			     MSR_LSTAR : MSR_CSTAR, &msr_data);
2191 		ctxt->_eip = msr_data;
2192 
2193 		ops->get_msr(ctxt, MSR_SYSCALL_MASK, &msr_data);
2194 		ctxt->eflags &= ~(msr_data | EFLG_RF);
2195 #endif
2196 	} else {
2197 		/* legacy mode */
2198 		ops->get_msr(ctxt, MSR_STAR, &msr_data);
2199 		ctxt->_eip = (u32)msr_data;
2200 
2201 		ctxt->eflags &= ~(EFLG_VM | EFLG_IF | EFLG_RF);
2202 	}
2203 
2204 	return X86EMUL_CONTINUE;
2205 }
2206 
2207 static int em_sysenter(struct x86_emulate_ctxt *ctxt)
2208 {
2209 	const struct x86_emulate_ops *ops = ctxt->ops;
2210 	struct desc_struct cs, ss;
2211 	u64 msr_data;
2212 	u16 cs_sel, ss_sel;
2213 	u64 efer = 0;
2214 
2215 	ops->get_msr(ctxt, MSR_EFER, &efer);
2216 	/* inject #GP if in real mode */
2217 	if (ctxt->mode == X86EMUL_MODE_REAL)
2218 		return emulate_gp(ctxt, 0);
2219 
2220 	/*
2221 	 * Not recognized on AMD in compat mode (but is recognized in legacy
2222 	 * mode).
2223 	 */
2224 	if ((ctxt->mode == X86EMUL_MODE_PROT32) && (efer & EFER_LMA)
2225 	    && !vendor_intel(ctxt))
2226 		return emulate_ud(ctxt);
2227 
2228 	/* XXX sysenter/sysexit have not been tested in 64bit mode.
2229 	* Therefore, we inject an #UD.
2230 	*/
2231 	if (ctxt->mode == X86EMUL_MODE_PROT64)
2232 		return emulate_ud(ctxt);
2233 
2234 	setup_syscalls_segments(ctxt, &cs, &ss);
2235 
2236 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_CS, &msr_data);
2237 	switch (ctxt->mode) {
2238 	case X86EMUL_MODE_PROT32:
2239 		if ((msr_data & 0xfffc) == 0x0)
2240 			return emulate_gp(ctxt, 0);
2241 		break;
2242 	case X86EMUL_MODE_PROT64:
2243 		if (msr_data == 0x0)
2244 			return emulate_gp(ctxt, 0);
2245 		break;
2246 	default:
2247 		break;
2248 	}
2249 
2250 	ctxt->eflags &= ~(EFLG_VM | EFLG_IF | EFLG_RF);
2251 	cs_sel = (u16)msr_data;
2252 	cs_sel &= ~SELECTOR_RPL_MASK;
2253 	ss_sel = cs_sel + 8;
2254 	ss_sel &= ~SELECTOR_RPL_MASK;
2255 	if (ctxt->mode == X86EMUL_MODE_PROT64 || (efer & EFER_LMA)) {
2256 		cs.d = 0;
2257 		cs.l = 1;
2258 	}
2259 
2260 	ops->set_segment(ctxt, cs_sel, &cs, 0, VCPU_SREG_CS);
2261 	ops->set_segment(ctxt, ss_sel, &ss, 0, VCPU_SREG_SS);
2262 
2263 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_EIP, &msr_data);
2264 	ctxt->_eip = msr_data;
2265 
2266 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_ESP, &msr_data);
2267 	*reg_write(ctxt, VCPU_REGS_RSP) = msr_data;
2268 
2269 	return X86EMUL_CONTINUE;
2270 }
2271 
2272 static int em_sysexit(struct x86_emulate_ctxt *ctxt)
2273 {
2274 	const struct x86_emulate_ops *ops = ctxt->ops;
2275 	struct desc_struct cs, ss;
2276 	u64 msr_data;
2277 	int usermode;
2278 	u16 cs_sel = 0, ss_sel = 0;
2279 
2280 	/* inject #GP if in real mode or Virtual 8086 mode */
2281 	if (ctxt->mode == X86EMUL_MODE_REAL ||
2282 	    ctxt->mode == X86EMUL_MODE_VM86)
2283 		return emulate_gp(ctxt, 0);
2284 
2285 	setup_syscalls_segments(ctxt, &cs, &ss);
2286 
2287 	if ((ctxt->rex_prefix & 0x8) != 0x0)
2288 		usermode = X86EMUL_MODE_PROT64;
2289 	else
2290 		usermode = X86EMUL_MODE_PROT32;
2291 
2292 	cs.dpl = 3;
2293 	ss.dpl = 3;
2294 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_CS, &msr_data);
2295 	switch (usermode) {
2296 	case X86EMUL_MODE_PROT32:
2297 		cs_sel = (u16)(msr_data + 16);
2298 		if ((msr_data & 0xfffc) == 0x0)
2299 			return emulate_gp(ctxt, 0);
2300 		ss_sel = (u16)(msr_data + 24);
2301 		break;
2302 	case X86EMUL_MODE_PROT64:
2303 		cs_sel = (u16)(msr_data + 32);
2304 		if (msr_data == 0x0)
2305 			return emulate_gp(ctxt, 0);
2306 		ss_sel = cs_sel + 8;
2307 		cs.d = 0;
2308 		cs.l = 1;
2309 		break;
2310 	}
2311 	cs_sel |= SELECTOR_RPL_MASK;
2312 	ss_sel |= SELECTOR_RPL_MASK;
2313 
2314 	ops->set_segment(ctxt, cs_sel, &cs, 0, VCPU_SREG_CS);
2315 	ops->set_segment(ctxt, ss_sel, &ss, 0, VCPU_SREG_SS);
2316 
2317 	ctxt->_eip = reg_read(ctxt, VCPU_REGS_RDX);
2318 	*reg_write(ctxt, VCPU_REGS_RSP) = reg_read(ctxt, VCPU_REGS_RCX);
2319 
2320 	return X86EMUL_CONTINUE;
2321 }
2322 
2323 static bool emulator_bad_iopl(struct x86_emulate_ctxt *ctxt)
2324 {
2325 	int iopl;
2326 	if (ctxt->mode == X86EMUL_MODE_REAL)
2327 		return false;
2328 	if (ctxt->mode == X86EMUL_MODE_VM86)
2329 		return true;
2330 	iopl = (ctxt->eflags & X86_EFLAGS_IOPL) >> IOPL_SHIFT;
2331 	return ctxt->ops->cpl(ctxt) > iopl;
2332 }
2333 
2334 static bool emulator_io_port_access_allowed(struct x86_emulate_ctxt *ctxt,
2335 					    u16 port, u16 len)
2336 {
2337 	const struct x86_emulate_ops *ops = ctxt->ops;
2338 	struct desc_struct tr_seg;
2339 	u32 base3;
2340 	int r;
2341 	u16 tr, io_bitmap_ptr, perm, bit_idx = port & 0x7;
2342 	unsigned mask = (1 << len) - 1;
2343 	unsigned long base;
2344 
2345 	ops->get_segment(ctxt, &tr, &tr_seg, &base3, VCPU_SREG_TR);
2346 	if (!tr_seg.p)
2347 		return false;
2348 	if (desc_limit_scaled(&tr_seg) < 103)
2349 		return false;
2350 	base = get_desc_base(&tr_seg);
2351 #ifdef CONFIG_X86_64
2352 	base |= ((u64)base3) << 32;
2353 #endif
2354 	r = ops->read_std(ctxt, base + 102, &io_bitmap_ptr, 2, NULL);
2355 	if (r != X86EMUL_CONTINUE)
2356 		return false;
2357 	if (io_bitmap_ptr + port/8 > desc_limit_scaled(&tr_seg))
2358 		return false;
2359 	r = ops->read_std(ctxt, base + io_bitmap_ptr + port/8, &perm, 2, NULL);
2360 	if (r != X86EMUL_CONTINUE)
2361 		return false;
2362 	if ((perm >> bit_idx) & mask)
2363 		return false;
2364 	return true;
2365 }
2366 
2367 static bool emulator_io_permited(struct x86_emulate_ctxt *ctxt,
2368 				 u16 port, u16 len)
2369 {
2370 	if (ctxt->perm_ok)
2371 		return true;
2372 
2373 	if (emulator_bad_iopl(ctxt))
2374 		if (!emulator_io_port_access_allowed(ctxt, port, len))
2375 			return false;
2376 
2377 	ctxt->perm_ok = true;
2378 
2379 	return true;
2380 }
2381 
2382 static void save_state_to_tss16(struct x86_emulate_ctxt *ctxt,
2383 				struct tss_segment_16 *tss)
2384 {
2385 	tss->ip = ctxt->_eip;
2386 	tss->flag = ctxt->eflags;
2387 	tss->ax = reg_read(ctxt, VCPU_REGS_RAX);
2388 	tss->cx = reg_read(ctxt, VCPU_REGS_RCX);
2389 	tss->dx = reg_read(ctxt, VCPU_REGS_RDX);
2390 	tss->bx = reg_read(ctxt, VCPU_REGS_RBX);
2391 	tss->sp = reg_read(ctxt, VCPU_REGS_RSP);
2392 	tss->bp = reg_read(ctxt, VCPU_REGS_RBP);
2393 	tss->si = reg_read(ctxt, VCPU_REGS_RSI);
2394 	tss->di = reg_read(ctxt, VCPU_REGS_RDI);
2395 
2396 	tss->es = get_segment_selector(ctxt, VCPU_SREG_ES);
2397 	tss->cs = get_segment_selector(ctxt, VCPU_SREG_CS);
2398 	tss->ss = get_segment_selector(ctxt, VCPU_SREG_SS);
2399 	tss->ds = get_segment_selector(ctxt, VCPU_SREG_DS);
2400 	tss->ldt = get_segment_selector(ctxt, VCPU_SREG_LDTR);
2401 }
2402 
2403 static int load_state_from_tss16(struct x86_emulate_ctxt *ctxt,
2404 				 struct tss_segment_16 *tss)
2405 {
2406 	int ret;
2407 
2408 	ctxt->_eip = tss->ip;
2409 	ctxt->eflags = tss->flag | 2;
2410 	*reg_write(ctxt, VCPU_REGS_RAX) = tss->ax;
2411 	*reg_write(ctxt, VCPU_REGS_RCX) = tss->cx;
2412 	*reg_write(ctxt, VCPU_REGS_RDX) = tss->dx;
2413 	*reg_write(ctxt, VCPU_REGS_RBX) = tss->bx;
2414 	*reg_write(ctxt, VCPU_REGS_RSP) = tss->sp;
2415 	*reg_write(ctxt, VCPU_REGS_RBP) = tss->bp;
2416 	*reg_write(ctxt, VCPU_REGS_RSI) = tss->si;
2417 	*reg_write(ctxt, VCPU_REGS_RDI) = tss->di;
2418 
2419 	/*
2420 	 * SDM says that segment selectors are loaded before segment
2421 	 * descriptors
2422 	 */
2423 	set_segment_selector(ctxt, tss->ldt, VCPU_SREG_LDTR);
2424 	set_segment_selector(ctxt, tss->es, VCPU_SREG_ES);
2425 	set_segment_selector(ctxt, tss->cs, VCPU_SREG_CS);
2426 	set_segment_selector(ctxt, tss->ss, VCPU_SREG_SS);
2427 	set_segment_selector(ctxt, tss->ds, VCPU_SREG_DS);
2428 
2429 	/*
2430 	 * Now load segment descriptors. If fault happens at this stage
2431 	 * it is handled in a context of new task
2432 	 */
2433 	ret = load_segment_descriptor(ctxt, tss->ldt, VCPU_SREG_LDTR);
2434 	if (ret != X86EMUL_CONTINUE)
2435 		return ret;
2436 	ret = load_segment_descriptor(ctxt, tss->es, VCPU_SREG_ES);
2437 	if (ret != X86EMUL_CONTINUE)
2438 		return ret;
2439 	ret = load_segment_descriptor(ctxt, tss->cs, VCPU_SREG_CS);
2440 	if (ret != X86EMUL_CONTINUE)
2441 		return ret;
2442 	ret = load_segment_descriptor(ctxt, tss->ss, VCPU_SREG_SS);
2443 	if (ret != X86EMUL_CONTINUE)
2444 		return ret;
2445 	ret = load_segment_descriptor(ctxt, tss->ds, VCPU_SREG_DS);
2446 	if (ret != X86EMUL_CONTINUE)
2447 		return ret;
2448 
2449 	return X86EMUL_CONTINUE;
2450 }
2451 
2452 static int task_switch_16(struct x86_emulate_ctxt *ctxt,
2453 			  u16 tss_selector, u16 old_tss_sel,
2454 			  ulong old_tss_base, struct desc_struct *new_desc)
2455 {
2456 	const struct x86_emulate_ops *ops = ctxt->ops;
2457 	struct tss_segment_16 tss_seg;
2458 	int ret;
2459 	u32 new_tss_base = get_desc_base(new_desc);
2460 
2461 	ret = ops->read_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2462 			    &ctxt->exception);
2463 	if (ret != X86EMUL_CONTINUE)
2464 		/* FIXME: need to provide precise fault address */
2465 		return ret;
2466 
2467 	save_state_to_tss16(ctxt, &tss_seg);
2468 
2469 	ret = ops->write_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2470 			     &ctxt->exception);
2471 	if (ret != X86EMUL_CONTINUE)
2472 		/* FIXME: need to provide precise fault address */
2473 		return ret;
2474 
2475 	ret = ops->read_std(ctxt, new_tss_base, &tss_seg, sizeof tss_seg,
2476 			    &ctxt->exception);
2477 	if (ret != X86EMUL_CONTINUE)
2478 		/* FIXME: need to provide precise fault address */
2479 		return ret;
2480 
2481 	if (old_tss_sel != 0xffff) {
2482 		tss_seg.prev_task_link = old_tss_sel;
2483 
2484 		ret = ops->write_std(ctxt, new_tss_base,
2485 				     &tss_seg.prev_task_link,
2486 				     sizeof tss_seg.prev_task_link,
2487 				     &ctxt->exception);
2488 		if (ret != X86EMUL_CONTINUE)
2489 			/* FIXME: need to provide precise fault address */
2490 			return ret;
2491 	}
2492 
2493 	return load_state_from_tss16(ctxt, &tss_seg);
2494 }
2495 
2496 static void save_state_to_tss32(struct x86_emulate_ctxt *ctxt,
2497 				struct tss_segment_32 *tss)
2498 {
2499 	tss->cr3 = ctxt->ops->get_cr(ctxt, 3);
2500 	tss->eip = ctxt->_eip;
2501 	tss->eflags = ctxt->eflags;
2502 	tss->eax = reg_read(ctxt, VCPU_REGS_RAX);
2503 	tss->ecx = reg_read(ctxt, VCPU_REGS_RCX);
2504 	tss->edx = reg_read(ctxt, VCPU_REGS_RDX);
2505 	tss->ebx = reg_read(ctxt, VCPU_REGS_RBX);
2506 	tss->esp = reg_read(ctxt, VCPU_REGS_RSP);
2507 	tss->ebp = reg_read(ctxt, VCPU_REGS_RBP);
2508 	tss->esi = reg_read(ctxt, VCPU_REGS_RSI);
2509 	tss->edi = reg_read(ctxt, VCPU_REGS_RDI);
2510 
2511 	tss->es = get_segment_selector(ctxt, VCPU_SREG_ES);
2512 	tss->cs = get_segment_selector(ctxt, VCPU_SREG_CS);
2513 	tss->ss = get_segment_selector(ctxt, VCPU_SREG_SS);
2514 	tss->ds = get_segment_selector(ctxt, VCPU_SREG_DS);
2515 	tss->fs = get_segment_selector(ctxt, VCPU_SREG_FS);
2516 	tss->gs = get_segment_selector(ctxt, VCPU_SREG_GS);
2517 	tss->ldt_selector = get_segment_selector(ctxt, VCPU_SREG_LDTR);
2518 }
2519 
2520 static int load_state_from_tss32(struct x86_emulate_ctxt *ctxt,
2521 				 struct tss_segment_32 *tss)
2522 {
2523 	int ret;
2524 
2525 	if (ctxt->ops->set_cr(ctxt, 3, tss->cr3))
2526 		return emulate_gp(ctxt, 0);
2527 	ctxt->_eip = tss->eip;
2528 	ctxt->eflags = tss->eflags | 2;
2529 
2530 	/* General purpose registers */
2531 	*reg_write(ctxt, VCPU_REGS_RAX) = tss->eax;
2532 	*reg_write(ctxt, VCPU_REGS_RCX) = tss->ecx;
2533 	*reg_write(ctxt, VCPU_REGS_RDX) = tss->edx;
2534 	*reg_write(ctxt, VCPU_REGS_RBX) = tss->ebx;
2535 	*reg_write(ctxt, VCPU_REGS_RSP) = tss->esp;
2536 	*reg_write(ctxt, VCPU_REGS_RBP) = tss->ebp;
2537 	*reg_write(ctxt, VCPU_REGS_RSI) = tss->esi;
2538 	*reg_write(ctxt, VCPU_REGS_RDI) = tss->edi;
2539 
2540 	/*
2541 	 * SDM says that segment selectors are loaded before segment
2542 	 * descriptors
2543 	 */
2544 	set_segment_selector(ctxt, tss->ldt_selector, VCPU_SREG_LDTR);
2545 	set_segment_selector(ctxt, tss->es, VCPU_SREG_ES);
2546 	set_segment_selector(ctxt, tss->cs, VCPU_SREG_CS);
2547 	set_segment_selector(ctxt, tss->ss, VCPU_SREG_SS);
2548 	set_segment_selector(ctxt, tss->ds, VCPU_SREG_DS);
2549 	set_segment_selector(ctxt, tss->fs, VCPU_SREG_FS);
2550 	set_segment_selector(ctxt, tss->gs, VCPU_SREG_GS);
2551 
2552 	/*
2553 	 * If we're switching between Protected Mode and VM86, we need to make
2554 	 * sure to update the mode before loading the segment descriptors so
2555 	 * that the selectors are interpreted correctly.
2556 	 *
2557 	 * Need to get rflags to the vcpu struct immediately because it
2558 	 * influences the CPL which is checked at least when loading the segment
2559 	 * descriptors and when pushing an error code to the new kernel stack.
2560 	 *
2561 	 * TODO Introduce a separate ctxt->ops->set_cpl callback
2562 	 */
2563 	if (ctxt->eflags & X86_EFLAGS_VM)
2564 		ctxt->mode = X86EMUL_MODE_VM86;
2565 	else
2566 		ctxt->mode = X86EMUL_MODE_PROT32;
2567 
2568 	ctxt->ops->set_rflags(ctxt, ctxt->eflags);
2569 
2570 	/*
2571 	 * Now load segment descriptors. If fault happenes at this stage
2572 	 * it is handled in a context of new task
2573 	 */
2574 	ret = load_segment_descriptor(ctxt, tss->ldt_selector, VCPU_SREG_LDTR);
2575 	if (ret != X86EMUL_CONTINUE)
2576 		return ret;
2577 	ret = load_segment_descriptor(ctxt, tss->es, VCPU_SREG_ES);
2578 	if (ret != X86EMUL_CONTINUE)
2579 		return ret;
2580 	ret = load_segment_descriptor(ctxt, tss->cs, VCPU_SREG_CS);
2581 	if (ret != X86EMUL_CONTINUE)
2582 		return ret;
2583 	ret = load_segment_descriptor(ctxt, tss->ss, VCPU_SREG_SS);
2584 	if (ret != X86EMUL_CONTINUE)
2585 		return ret;
2586 	ret = load_segment_descriptor(ctxt, tss->ds, VCPU_SREG_DS);
2587 	if (ret != X86EMUL_CONTINUE)
2588 		return ret;
2589 	ret = load_segment_descriptor(ctxt, tss->fs, VCPU_SREG_FS);
2590 	if (ret != X86EMUL_CONTINUE)
2591 		return ret;
2592 	ret = load_segment_descriptor(ctxt, tss->gs, VCPU_SREG_GS);
2593 	if (ret != X86EMUL_CONTINUE)
2594 		return ret;
2595 
2596 	return X86EMUL_CONTINUE;
2597 }
2598 
2599 static int task_switch_32(struct x86_emulate_ctxt *ctxt,
2600 			  u16 tss_selector, u16 old_tss_sel,
2601 			  ulong old_tss_base, struct desc_struct *new_desc)
2602 {
2603 	const struct x86_emulate_ops *ops = ctxt->ops;
2604 	struct tss_segment_32 tss_seg;
2605 	int ret;
2606 	u32 new_tss_base = get_desc_base(new_desc);
2607 
2608 	ret = ops->read_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2609 			    &ctxt->exception);
2610 	if (ret != X86EMUL_CONTINUE)
2611 		/* FIXME: need to provide precise fault address */
2612 		return ret;
2613 
2614 	save_state_to_tss32(ctxt, &tss_seg);
2615 
2616 	ret = ops->write_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2617 			     &ctxt->exception);
2618 	if (ret != X86EMUL_CONTINUE)
2619 		/* FIXME: need to provide precise fault address */
2620 		return ret;
2621 
2622 	ret = ops->read_std(ctxt, new_tss_base, &tss_seg, sizeof tss_seg,
2623 			    &ctxt->exception);
2624 	if (ret != X86EMUL_CONTINUE)
2625 		/* FIXME: need to provide precise fault address */
2626 		return ret;
2627 
2628 	if (old_tss_sel != 0xffff) {
2629 		tss_seg.prev_task_link = old_tss_sel;
2630 
2631 		ret = ops->write_std(ctxt, new_tss_base,
2632 				     &tss_seg.prev_task_link,
2633 				     sizeof tss_seg.prev_task_link,
2634 				     &ctxt->exception);
2635 		if (ret != X86EMUL_CONTINUE)
2636 			/* FIXME: need to provide precise fault address */
2637 			return ret;
2638 	}
2639 
2640 	return load_state_from_tss32(ctxt, &tss_seg);
2641 }
2642 
2643 static int emulator_do_task_switch(struct x86_emulate_ctxt *ctxt,
2644 				   u16 tss_selector, int idt_index, int reason,
2645 				   bool has_error_code, u32 error_code)
2646 {
2647 	const struct x86_emulate_ops *ops = ctxt->ops;
2648 	struct desc_struct curr_tss_desc, next_tss_desc;
2649 	int ret;
2650 	u16 old_tss_sel = get_segment_selector(ctxt, VCPU_SREG_TR);
2651 	ulong old_tss_base =
2652 		ops->get_cached_segment_base(ctxt, VCPU_SREG_TR);
2653 	u32 desc_limit;
2654 	ulong desc_addr;
2655 
2656 	/* FIXME: old_tss_base == ~0 ? */
2657 
2658 	ret = read_segment_descriptor(ctxt, tss_selector, &next_tss_desc, &desc_addr);
2659 	if (ret != X86EMUL_CONTINUE)
2660 		return ret;
2661 	ret = read_segment_descriptor(ctxt, old_tss_sel, &curr_tss_desc, &desc_addr);
2662 	if (ret != X86EMUL_CONTINUE)
2663 		return ret;
2664 
2665 	/* FIXME: check that next_tss_desc is tss */
2666 
2667 	/*
2668 	 * Check privileges. The three cases are task switch caused by...
2669 	 *
2670 	 * 1. jmp/call/int to task gate: Check against DPL of the task gate
2671 	 * 2. Exception/IRQ/iret: No check is performed
2672 	 * 3. jmp/call to TSS: Check against DPL of the TSS
2673 	 */
2674 	if (reason == TASK_SWITCH_GATE) {
2675 		if (idt_index != -1) {
2676 			/* Software interrupts */
2677 			struct desc_struct task_gate_desc;
2678 			int dpl;
2679 
2680 			ret = read_interrupt_descriptor(ctxt, idt_index,
2681 							&task_gate_desc);
2682 			if (ret != X86EMUL_CONTINUE)
2683 				return ret;
2684 
2685 			dpl = task_gate_desc.dpl;
2686 			if ((tss_selector & 3) > dpl || ops->cpl(ctxt) > dpl)
2687 				return emulate_gp(ctxt, (idt_index << 3) | 0x2);
2688 		}
2689 	} else if (reason != TASK_SWITCH_IRET) {
2690 		int dpl = next_tss_desc.dpl;
2691 		if ((tss_selector & 3) > dpl || ops->cpl(ctxt) > dpl)
2692 			return emulate_gp(ctxt, tss_selector);
2693 	}
2694 
2695 
2696 	desc_limit = desc_limit_scaled(&next_tss_desc);
2697 	if (!next_tss_desc.p ||
2698 	    ((desc_limit < 0x67 && (next_tss_desc.type & 8)) ||
2699 	     desc_limit < 0x2b)) {
2700 		emulate_ts(ctxt, tss_selector & 0xfffc);
2701 		return X86EMUL_PROPAGATE_FAULT;
2702 	}
2703 
2704 	if (reason == TASK_SWITCH_IRET || reason == TASK_SWITCH_JMP) {
2705 		curr_tss_desc.type &= ~(1 << 1); /* clear busy flag */
2706 		write_segment_descriptor(ctxt, old_tss_sel, &curr_tss_desc);
2707 	}
2708 
2709 	if (reason == TASK_SWITCH_IRET)
2710 		ctxt->eflags = ctxt->eflags & ~X86_EFLAGS_NT;
2711 
2712 	/* set back link to prev task only if NT bit is set in eflags
2713 	   note that old_tss_sel is not used after this point */
2714 	if (reason != TASK_SWITCH_CALL && reason != TASK_SWITCH_GATE)
2715 		old_tss_sel = 0xffff;
2716 
2717 	if (next_tss_desc.type & 8)
2718 		ret = task_switch_32(ctxt, tss_selector, old_tss_sel,
2719 				     old_tss_base, &next_tss_desc);
2720 	else
2721 		ret = task_switch_16(ctxt, tss_selector, old_tss_sel,
2722 				     old_tss_base, &next_tss_desc);
2723 	if (ret != X86EMUL_CONTINUE)
2724 		return ret;
2725 
2726 	if (reason == TASK_SWITCH_CALL || reason == TASK_SWITCH_GATE)
2727 		ctxt->eflags = ctxt->eflags | X86_EFLAGS_NT;
2728 
2729 	if (reason != TASK_SWITCH_IRET) {
2730 		next_tss_desc.type |= (1 << 1); /* set busy flag */
2731 		write_segment_descriptor(ctxt, tss_selector, &next_tss_desc);
2732 	}
2733 
2734 	ops->set_cr(ctxt, 0,  ops->get_cr(ctxt, 0) | X86_CR0_TS);
2735 	ops->set_segment(ctxt, tss_selector, &next_tss_desc, 0, VCPU_SREG_TR);
2736 
2737 	if (has_error_code) {
2738 		ctxt->op_bytes = ctxt->ad_bytes = (next_tss_desc.type & 8) ? 4 : 2;
2739 		ctxt->lock_prefix = 0;
2740 		ctxt->src.val = (unsigned long) error_code;
2741 		ret = em_push(ctxt);
2742 	}
2743 
2744 	return ret;
2745 }
2746 
2747 int emulator_task_switch(struct x86_emulate_ctxt *ctxt,
2748 			 u16 tss_selector, int idt_index, int reason,
2749 			 bool has_error_code, u32 error_code)
2750 {
2751 	int rc;
2752 
2753 	invalidate_registers(ctxt);
2754 	ctxt->_eip = ctxt->eip;
2755 	ctxt->dst.type = OP_NONE;
2756 
2757 	rc = emulator_do_task_switch(ctxt, tss_selector, idt_index, reason,
2758 				     has_error_code, error_code);
2759 
2760 	if (rc == X86EMUL_CONTINUE) {
2761 		ctxt->eip = ctxt->_eip;
2762 		writeback_registers(ctxt);
2763 	}
2764 
2765 	return (rc == X86EMUL_UNHANDLEABLE) ? EMULATION_FAILED : EMULATION_OK;
2766 }
2767 
2768 static void string_addr_inc(struct x86_emulate_ctxt *ctxt, int reg,
2769 		struct operand *op)
2770 {
2771 	int df = (ctxt->eflags & EFLG_DF) ? -op->count : op->count;
2772 
2773 	register_address_increment(ctxt, reg_rmw(ctxt, reg), df * op->bytes);
2774 	op->addr.mem.ea = register_address(ctxt, reg_read(ctxt, reg));
2775 }
2776 
2777 static int em_das(struct x86_emulate_ctxt *ctxt)
2778 {
2779 	u8 al, old_al;
2780 	bool af, cf, old_cf;
2781 
2782 	cf = ctxt->eflags & X86_EFLAGS_CF;
2783 	al = ctxt->dst.val;
2784 
2785 	old_al = al;
2786 	old_cf = cf;
2787 	cf = false;
2788 	af = ctxt->eflags & X86_EFLAGS_AF;
2789 	if ((al & 0x0f) > 9 || af) {
2790 		al -= 6;
2791 		cf = old_cf | (al >= 250);
2792 		af = true;
2793 	} else {
2794 		af = false;
2795 	}
2796 	if (old_al > 0x99 || old_cf) {
2797 		al -= 0x60;
2798 		cf = true;
2799 	}
2800 
2801 	ctxt->dst.val = al;
2802 	/* Set PF, ZF, SF */
2803 	ctxt->src.type = OP_IMM;
2804 	ctxt->src.val = 0;
2805 	ctxt->src.bytes = 1;
2806 	fastop(ctxt, em_or);
2807 	ctxt->eflags &= ~(X86_EFLAGS_AF | X86_EFLAGS_CF);
2808 	if (cf)
2809 		ctxt->eflags |= X86_EFLAGS_CF;
2810 	if (af)
2811 		ctxt->eflags |= X86_EFLAGS_AF;
2812 	return X86EMUL_CONTINUE;
2813 }
2814 
2815 static int em_aam(struct x86_emulate_ctxt *ctxt)
2816 {
2817 	u8 al, ah;
2818 
2819 	if (ctxt->src.val == 0)
2820 		return emulate_de(ctxt);
2821 
2822 	al = ctxt->dst.val & 0xff;
2823 	ah = al / ctxt->src.val;
2824 	al %= ctxt->src.val;
2825 
2826 	ctxt->dst.val = (ctxt->dst.val & 0xffff0000) | al | (ah << 8);
2827 
2828 	/* Set PF, ZF, SF */
2829 	ctxt->src.type = OP_IMM;
2830 	ctxt->src.val = 0;
2831 	ctxt->src.bytes = 1;
2832 	fastop(ctxt, em_or);
2833 
2834 	return X86EMUL_CONTINUE;
2835 }
2836 
2837 static int em_aad(struct x86_emulate_ctxt *ctxt)
2838 {
2839 	u8 al = ctxt->dst.val & 0xff;
2840 	u8 ah = (ctxt->dst.val >> 8) & 0xff;
2841 
2842 	al = (al + (ah * ctxt->src.val)) & 0xff;
2843 
2844 	ctxt->dst.val = (ctxt->dst.val & 0xffff0000) | al;
2845 
2846 	/* Set PF, ZF, SF */
2847 	ctxt->src.type = OP_IMM;
2848 	ctxt->src.val = 0;
2849 	ctxt->src.bytes = 1;
2850 	fastop(ctxt, em_or);
2851 
2852 	return X86EMUL_CONTINUE;
2853 }
2854 
2855 static int em_call(struct x86_emulate_ctxt *ctxt)
2856 {
2857 	long rel = ctxt->src.val;
2858 
2859 	ctxt->src.val = (unsigned long)ctxt->_eip;
2860 	jmp_rel(ctxt, rel);
2861 	return em_push(ctxt);
2862 }
2863 
2864 static int em_call_far(struct x86_emulate_ctxt *ctxt)
2865 {
2866 	u16 sel, old_cs;
2867 	ulong old_eip;
2868 	int rc;
2869 
2870 	old_cs = get_segment_selector(ctxt, VCPU_SREG_CS);
2871 	old_eip = ctxt->_eip;
2872 
2873 	memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
2874 	if (load_segment_descriptor(ctxt, sel, VCPU_SREG_CS))
2875 		return X86EMUL_CONTINUE;
2876 
2877 	ctxt->_eip = 0;
2878 	memcpy(&ctxt->_eip, ctxt->src.valptr, ctxt->op_bytes);
2879 
2880 	ctxt->src.val = old_cs;
2881 	rc = em_push(ctxt);
2882 	if (rc != X86EMUL_CONTINUE)
2883 		return rc;
2884 
2885 	ctxt->src.val = old_eip;
2886 	return em_push(ctxt);
2887 }
2888 
2889 static int em_ret_near_imm(struct x86_emulate_ctxt *ctxt)
2890 {
2891 	int rc;
2892 
2893 	ctxt->dst.type = OP_REG;
2894 	ctxt->dst.addr.reg = &ctxt->_eip;
2895 	ctxt->dst.bytes = ctxt->op_bytes;
2896 	rc = emulate_pop(ctxt, &ctxt->dst.val, ctxt->op_bytes);
2897 	if (rc != X86EMUL_CONTINUE)
2898 		return rc;
2899 	rsp_increment(ctxt, ctxt->src.val);
2900 	return X86EMUL_CONTINUE;
2901 }
2902 
2903 static int em_xchg(struct x86_emulate_ctxt *ctxt)
2904 {
2905 	/* Write back the register source. */
2906 	ctxt->src.val = ctxt->dst.val;
2907 	write_register_operand(&ctxt->src);
2908 
2909 	/* Write back the memory destination with implicit LOCK prefix. */
2910 	ctxt->dst.val = ctxt->src.orig_val;
2911 	ctxt->lock_prefix = 1;
2912 	return X86EMUL_CONTINUE;
2913 }
2914 
2915 static int em_imul_3op(struct x86_emulate_ctxt *ctxt)
2916 {
2917 	ctxt->dst.val = ctxt->src2.val;
2918 	return fastop(ctxt, em_imul);
2919 }
2920 
2921 static int em_cwd(struct x86_emulate_ctxt *ctxt)
2922 {
2923 	ctxt->dst.type = OP_REG;
2924 	ctxt->dst.bytes = ctxt->src.bytes;
2925 	ctxt->dst.addr.reg = reg_rmw(ctxt, VCPU_REGS_RDX);
2926 	ctxt->dst.val = ~((ctxt->src.val >> (ctxt->src.bytes * 8 - 1)) - 1);
2927 
2928 	return X86EMUL_CONTINUE;
2929 }
2930 
2931 static int em_rdtsc(struct x86_emulate_ctxt *ctxt)
2932 {
2933 	u64 tsc = 0;
2934 
2935 	ctxt->ops->get_msr(ctxt, MSR_IA32_TSC, &tsc);
2936 	*reg_write(ctxt, VCPU_REGS_RAX) = (u32)tsc;
2937 	*reg_write(ctxt, VCPU_REGS_RDX) = tsc >> 32;
2938 	return X86EMUL_CONTINUE;
2939 }
2940 
2941 static int em_rdpmc(struct x86_emulate_ctxt *ctxt)
2942 {
2943 	u64 pmc;
2944 
2945 	if (ctxt->ops->read_pmc(ctxt, reg_read(ctxt, VCPU_REGS_RCX), &pmc))
2946 		return emulate_gp(ctxt, 0);
2947 	*reg_write(ctxt, VCPU_REGS_RAX) = (u32)pmc;
2948 	*reg_write(ctxt, VCPU_REGS_RDX) = pmc >> 32;
2949 	return X86EMUL_CONTINUE;
2950 }
2951 
2952 static int em_mov(struct x86_emulate_ctxt *ctxt)
2953 {
2954 	memcpy(ctxt->dst.valptr, ctxt->src.valptr, ctxt->op_bytes);
2955 	return X86EMUL_CONTINUE;
2956 }
2957 
2958 #define FFL(x) bit(X86_FEATURE_##x)
2959 
2960 static int em_movbe(struct x86_emulate_ctxt *ctxt)
2961 {
2962 	u32 ebx, ecx, edx, eax = 1;
2963 	u16 tmp;
2964 
2965 	/*
2966 	 * Check MOVBE is set in the guest-visible CPUID leaf.
2967 	 */
2968 	ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
2969 	if (!(ecx & FFL(MOVBE)))
2970 		return emulate_ud(ctxt);
2971 
2972 	switch (ctxt->op_bytes) {
2973 	case 2:
2974 		/*
2975 		 * From MOVBE definition: "...When the operand size is 16 bits,
2976 		 * the upper word of the destination register remains unchanged
2977 		 * ..."
2978 		 *
2979 		 * Both casting ->valptr and ->val to u16 breaks strict aliasing
2980 		 * rules so we have to do the operation almost per hand.
2981 		 */
2982 		tmp = (u16)ctxt->src.val;
2983 		ctxt->dst.val &= ~0xffffUL;
2984 		ctxt->dst.val |= (unsigned long)swab16(tmp);
2985 		break;
2986 	case 4:
2987 		ctxt->dst.val = swab32((u32)ctxt->src.val);
2988 		break;
2989 	case 8:
2990 		ctxt->dst.val = swab64(ctxt->src.val);
2991 		break;
2992 	default:
2993 		return X86EMUL_PROPAGATE_FAULT;
2994 	}
2995 	return X86EMUL_CONTINUE;
2996 }
2997 
2998 static int em_cr_write(struct x86_emulate_ctxt *ctxt)
2999 {
3000 	if (ctxt->ops->set_cr(ctxt, ctxt->modrm_reg, ctxt->src.val))
3001 		return emulate_gp(ctxt, 0);
3002 
3003 	/* Disable writeback. */
3004 	ctxt->dst.type = OP_NONE;
3005 	return X86EMUL_CONTINUE;
3006 }
3007 
3008 static int em_dr_write(struct x86_emulate_ctxt *ctxt)
3009 {
3010 	unsigned long val;
3011 
3012 	if (ctxt->mode == X86EMUL_MODE_PROT64)
3013 		val = ctxt->src.val & ~0ULL;
3014 	else
3015 		val = ctxt->src.val & ~0U;
3016 
3017 	/* #UD condition is already handled. */
3018 	if (ctxt->ops->set_dr(ctxt, ctxt->modrm_reg, val) < 0)
3019 		return emulate_gp(ctxt, 0);
3020 
3021 	/* Disable writeback. */
3022 	ctxt->dst.type = OP_NONE;
3023 	return X86EMUL_CONTINUE;
3024 }
3025 
3026 static int em_wrmsr(struct x86_emulate_ctxt *ctxt)
3027 {
3028 	u64 msr_data;
3029 
3030 	msr_data = (u32)reg_read(ctxt, VCPU_REGS_RAX)
3031 		| ((u64)reg_read(ctxt, VCPU_REGS_RDX) << 32);
3032 	if (ctxt->ops->set_msr(ctxt, reg_read(ctxt, VCPU_REGS_RCX), msr_data))
3033 		return emulate_gp(ctxt, 0);
3034 
3035 	return X86EMUL_CONTINUE;
3036 }
3037 
3038 static int em_rdmsr(struct x86_emulate_ctxt *ctxt)
3039 {
3040 	u64 msr_data;
3041 
3042 	if (ctxt->ops->get_msr(ctxt, reg_read(ctxt, VCPU_REGS_RCX), &msr_data))
3043 		return emulate_gp(ctxt, 0);
3044 
3045 	*reg_write(ctxt, VCPU_REGS_RAX) = (u32)msr_data;
3046 	*reg_write(ctxt, VCPU_REGS_RDX) = msr_data >> 32;
3047 	return X86EMUL_CONTINUE;
3048 }
3049 
3050 static int em_mov_rm_sreg(struct x86_emulate_ctxt *ctxt)
3051 {
3052 	if (ctxt->modrm_reg > VCPU_SREG_GS)
3053 		return emulate_ud(ctxt);
3054 
3055 	ctxt->dst.val = get_segment_selector(ctxt, ctxt->modrm_reg);
3056 	return X86EMUL_CONTINUE;
3057 }
3058 
3059 static int em_mov_sreg_rm(struct x86_emulate_ctxt *ctxt)
3060 {
3061 	u16 sel = ctxt->src.val;
3062 
3063 	if (ctxt->modrm_reg == VCPU_SREG_CS || ctxt->modrm_reg > VCPU_SREG_GS)
3064 		return emulate_ud(ctxt);
3065 
3066 	if (ctxt->modrm_reg == VCPU_SREG_SS)
3067 		ctxt->interruptibility = KVM_X86_SHADOW_INT_MOV_SS;
3068 
3069 	/* Disable writeback. */
3070 	ctxt->dst.type = OP_NONE;
3071 	return load_segment_descriptor(ctxt, sel, ctxt->modrm_reg);
3072 }
3073 
3074 static int em_lldt(struct x86_emulate_ctxt *ctxt)
3075 {
3076 	u16 sel = ctxt->src.val;
3077 
3078 	/* Disable writeback. */
3079 	ctxt->dst.type = OP_NONE;
3080 	return load_segment_descriptor(ctxt, sel, VCPU_SREG_LDTR);
3081 }
3082 
3083 static int em_ltr(struct x86_emulate_ctxt *ctxt)
3084 {
3085 	u16 sel = ctxt->src.val;
3086 
3087 	/* Disable writeback. */
3088 	ctxt->dst.type = OP_NONE;
3089 	return load_segment_descriptor(ctxt, sel, VCPU_SREG_TR);
3090 }
3091 
3092 static int em_invlpg(struct x86_emulate_ctxt *ctxt)
3093 {
3094 	int rc;
3095 	ulong linear;
3096 
3097 	rc = linearize(ctxt, ctxt->src.addr.mem, 1, false, &linear);
3098 	if (rc == X86EMUL_CONTINUE)
3099 		ctxt->ops->invlpg(ctxt, linear);
3100 	/* Disable writeback. */
3101 	ctxt->dst.type = OP_NONE;
3102 	return X86EMUL_CONTINUE;
3103 }
3104 
3105 static int em_clts(struct x86_emulate_ctxt *ctxt)
3106 {
3107 	ulong cr0;
3108 
3109 	cr0 = ctxt->ops->get_cr(ctxt, 0);
3110 	cr0 &= ~X86_CR0_TS;
3111 	ctxt->ops->set_cr(ctxt, 0, cr0);
3112 	return X86EMUL_CONTINUE;
3113 }
3114 
3115 static int em_vmcall(struct x86_emulate_ctxt *ctxt)
3116 {
3117 	int rc;
3118 
3119 	if (ctxt->modrm_mod != 3 || ctxt->modrm_rm != 1)
3120 		return X86EMUL_UNHANDLEABLE;
3121 
3122 	rc = ctxt->ops->fix_hypercall(ctxt);
3123 	if (rc != X86EMUL_CONTINUE)
3124 		return rc;
3125 
3126 	/* Let the processor re-execute the fixed hypercall */
3127 	ctxt->_eip = ctxt->eip;
3128 	/* Disable writeback. */
3129 	ctxt->dst.type = OP_NONE;
3130 	return X86EMUL_CONTINUE;
3131 }
3132 
3133 static int emulate_store_desc_ptr(struct x86_emulate_ctxt *ctxt,
3134 				  void (*get)(struct x86_emulate_ctxt *ctxt,
3135 					      struct desc_ptr *ptr))
3136 {
3137 	struct desc_ptr desc_ptr;
3138 
3139 	if (ctxt->mode == X86EMUL_MODE_PROT64)
3140 		ctxt->op_bytes = 8;
3141 	get(ctxt, &desc_ptr);
3142 	if (ctxt->op_bytes == 2) {
3143 		ctxt->op_bytes = 4;
3144 		desc_ptr.address &= 0x00ffffff;
3145 	}
3146 	/* Disable writeback. */
3147 	ctxt->dst.type = OP_NONE;
3148 	return segmented_write(ctxt, ctxt->dst.addr.mem,
3149 			       &desc_ptr, 2 + ctxt->op_bytes);
3150 }
3151 
3152 static int em_sgdt(struct x86_emulate_ctxt *ctxt)
3153 {
3154 	return emulate_store_desc_ptr(ctxt, ctxt->ops->get_gdt);
3155 }
3156 
3157 static int em_sidt(struct x86_emulate_ctxt *ctxt)
3158 {
3159 	return emulate_store_desc_ptr(ctxt, ctxt->ops->get_idt);
3160 }
3161 
3162 static int em_lgdt(struct x86_emulate_ctxt *ctxt)
3163 {
3164 	struct desc_ptr desc_ptr;
3165 	int rc;
3166 
3167 	if (ctxt->mode == X86EMUL_MODE_PROT64)
3168 		ctxt->op_bytes = 8;
3169 	rc = read_descriptor(ctxt, ctxt->src.addr.mem,
3170 			     &desc_ptr.size, &desc_ptr.address,
3171 			     ctxt->op_bytes);
3172 	if (rc != X86EMUL_CONTINUE)
3173 		return rc;
3174 	ctxt->ops->set_gdt(ctxt, &desc_ptr);
3175 	/* Disable writeback. */
3176 	ctxt->dst.type = OP_NONE;
3177 	return X86EMUL_CONTINUE;
3178 }
3179 
3180 static int em_vmmcall(struct x86_emulate_ctxt *ctxt)
3181 {
3182 	int rc;
3183 
3184 	rc = ctxt->ops->fix_hypercall(ctxt);
3185 
3186 	/* Disable writeback. */
3187 	ctxt->dst.type = OP_NONE;
3188 	return rc;
3189 }
3190 
3191 static int em_lidt(struct x86_emulate_ctxt *ctxt)
3192 {
3193 	struct desc_ptr desc_ptr;
3194 	int rc;
3195 
3196 	if (ctxt->mode == X86EMUL_MODE_PROT64)
3197 		ctxt->op_bytes = 8;
3198 	rc = read_descriptor(ctxt, ctxt->src.addr.mem,
3199 			     &desc_ptr.size, &desc_ptr.address,
3200 			     ctxt->op_bytes);
3201 	if (rc != X86EMUL_CONTINUE)
3202 		return rc;
3203 	ctxt->ops->set_idt(ctxt, &desc_ptr);
3204 	/* Disable writeback. */
3205 	ctxt->dst.type = OP_NONE;
3206 	return X86EMUL_CONTINUE;
3207 }
3208 
3209 static int em_smsw(struct x86_emulate_ctxt *ctxt)
3210 {
3211 	ctxt->dst.bytes = 2;
3212 	ctxt->dst.val = ctxt->ops->get_cr(ctxt, 0);
3213 	return X86EMUL_CONTINUE;
3214 }
3215 
3216 static int em_lmsw(struct x86_emulate_ctxt *ctxt)
3217 {
3218 	ctxt->ops->set_cr(ctxt, 0, (ctxt->ops->get_cr(ctxt, 0) & ~0x0eul)
3219 			  | (ctxt->src.val & 0x0f));
3220 	ctxt->dst.type = OP_NONE;
3221 	return X86EMUL_CONTINUE;
3222 }
3223 
3224 static int em_loop(struct x86_emulate_ctxt *ctxt)
3225 {
3226 	register_address_increment(ctxt, reg_rmw(ctxt, VCPU_REGS_RCX), -1);
3227 	if ((address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) != 0) &&
3228 	    (ctxt->b == 0xe2 || test_cc(ctxt->b ^ 0x5, ctxt->eflags)))
3229 		jmp_rel(ctxt, ctxt->src.val);
3230 
3231 	return X86EMUL_CONTINUE;
3232 }
3233 
3234 static int em_jcxz(struct x86_emulate_ctxt *ctxt)
3235 {
3236 	if (address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) == 0)
3237 		jmp_rel(ctxt, ctxt->src.val);
3238 
3239 	return X86EMUL_CONTINUE;
3240 }
3241 
3242 static int em_in(struct x86_emulate_ctxt *ctxt)
3243 {
3244 	if (!pio_in_emulated(ctxt, ctxt->dst.bytes, ctxt->src.val,
3245 			     &ctxt->dst.val))
3246 		return X86EMUL_IO_NEEDED;
3247 
3248 	return X86EMUL_CONTINUE;
3249 }
3250 
3251 static int em_out(struct x86_emulate_ctxt *ctxt)
3252 {
3253 	ctxt->ops->pio_out_emulated(ctxt, ctxt->src.bytes, ctxt->dst.val,
3254 				    &ctxt->src.val, 1);
3255 	/* Disable writeback. */
3256 	ctxt->dst.type = OP_NONE;
3257 	return X86EMUL_CONTINUE;
3258 }
3259 
3260 static int em_cli(struct x86_emulate_ctxt *ctxt)
3261 {
3262 	if (emulator_bad_iopl(ctxt))
3263 		return emulate_gp(ctxt, 0);
3264 
3265 	ctxt->eflags &= ~X86_EFLAGS_IF;
3266 	return X86EMUL_CONTINUE;
3267 }
3268 
3269 static int em_sti(struct x86_emulate_ctxt *ctxt)
3270 {
3271 	if (emulator_bad_iopl(ctxt))
3272 		return emulate_gp(ctxt, 0);
3273 
3274 	ctxt->interruptibility = KVM_X86_SHADOW_INT_STI;
3275 	ctxt->eflags |= X86_EFLAGS_IF;
3276 	return X86EMUL_CONTINUE;
3277 }
3278 
3279 static int em_cpuid(struct x86_emulate_ctxt *ctxt)
3280 {
3281 	u32 eax, ebx, ecx, edx;
3282 
3283 	eax = reg_read(ctxt, VCPU_REGS_RAX);
3284 	ecx = reg_read(ctxt, VCPU_REGS_RCX);
3285 	ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
3286 	*reg_write(ctxt, VCPU_REGS_RAX) = eax;
3287 	*reg_write(ctxt, VCPU_REGS_RBX) = ebx;
3288 	*reg_write(ctxt, VCPU_REGS_RCX) = ecx;
3289 	*reg_write(ctxt, VCPU_REGS_RDX) = edx;
3290 	return X86EMUL_CONTINUE;
3291 }
3292 
3293 static int em_sahf(struct x86_emulate_ctxt *ctxt)
3294 {
3295 	u32 flags;
3296 
3297 	flags = EFLG_CF | EFLG_PF | EFLG_AF | EFLG_ZF | EFLG_SF;
3298 	flags &= *reg_rmw(ctxt, VCPU_REGS_RAX) >> 8;
3299 
3300 	ctxt->eflags &= ~0xffUL;
3301 	ctxt->eflags |= flags | X86_EFLAGS_FIXED;
3302 	return X86EMUL_CONTINUE;
3303 }
3304 
3305 static int em_lahf(struct x86_emulate_ctxt *ctxt)
3306 {
3307 	*reg_rmw(ctxt, VCPU_REGS_RAX) &= ~0xff00UL;
3308 	*reg_rmw(ctxt, VCPU_REGS_RAX) |= (ctxt->eflags & 0xff) << 8;
3309 	return X86EMUL_CONTINUE;
3310 }
3311 
3312 static int em_bswap(struct x86_emulate_ctxt *ctxt)
3313 {
3314 	switch (ctxt->op_bytes) {
3315 #ifdef CONFIG_X86_64
3316 	case 8:
3317 		asm("bswap %0" : "+r"(ctxt->dst.val));
3318 		break;
3319 #endif
3320 	default:
3321 		asm("bswap %0" : "+r"(*(u32 *)&ctxt->dst.val));
3322 		break;
3323 	}
3324 	return X86EMUL_CONTINUE;
3325 }
3326 
3327 static bool valid_cr(int nr)
3328 {
3329 	switch (nr) {
3330 	case 0:
3331 	case 2 ... 4:
3332 	case 8:
3333 		return true;
3334 	default:
3335 		return false;
3336 	}
3337 }
3338 
3339 static int check_cr_read(struct x86_emulate_ctxt *ctxt)
3340 {
3341 	if (!valid_cr(ctxt->modrm_reg))
3342 		return emulate_ud(ctxt);
3343 
3344 	return X86EMUL_CONTINUE;
3345 }
3346 
3347 static int check_cr_write(struct x86_emulate_ctxt *ctxt)
3348 {
3349 	u64 new_val = ctxt->src.val64;
3350 	int cr = ctxt->modrm_reg;
3351 	u64 efer = 0;
3352 
3353 	static u64 cr_reserved_bits[] = {
3354 		0xffffffff00000000ULL,
3355 		0, 0, 0, /* CR3 checked later */
3356 		CR4_RESERVED_BITS,
3357 		0, 0, 0,
3358 		CR8_RESERVED_BITS,
3359 	};
3360 
3361 	if (!valid_cr(cr))
3362 		return emulate_ud(ctxt);
3363 
3364 	if (new_val & cr_reserved_bits[cr])
3365 		return emulate_gp(ctxt, 0);
3366 
3367 	switch (cr) {
3368 	case 0: {
3369 		u64 cr4;
3370 		if (((new_val & X86_CR0_PG) && !(new_val & X86_CR0_PE)) ||
3371 		    ((new_val & X86_CR0_NW) && !(new_val & X86_CR0_CD)))
3372 			return emulate_gp(ctxt, 0);
3373 
3374 		cr4 = ctxt->ops->get_cr(ctxt, 4);
3375 		ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3376 
3377 		if ((new_val & X86_CR0_PG) && (efer & EFER_LME) &&
3378 		    !(cr4 & X86_CR4_PAE))
3379 			return emulate_gp(ctxt, 0);
3380 
3381 		break;
3382 		}
3383 	case 3: {
3384 		u64 rsvd = 0;
3385 
3386 		ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3387 		if (efer & EFER_LMA)
3388 			rsvd = CR3_L_MODE_RESERVED_BITS;
3389 		else if (ctxt->ops->get_cr(ctxt, 4) & X86_CR4_PAE)
3390 			rsvd = CR3_PAE_RESERVED_BITS;
3391 		else if (ctxt->ops->get_cr(ctxt, 0) & X86_CR0_PG)
3392 			rsvd = CR3_NONPAE_RESERVED_BITS;
3393 
3394 		if (new_val & rsvd)
3395 			return emulate_gp(ctxt, 0);
3396 
3397 		break;
3398 		}
3399 	case 4: {
3400 		ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3401 
3402 		if ((efer & EFER_LMA) && !(new_val & X86_CR4_PAE))
3403 			return emulate_gp(ctxt, 0);
3404 
3405 		break;
3406 		}
3407 	}
3408 
3409 	return X86EMUL_CONTINUE;
3410 }
3411 
3412 static int check_dr7_gd(struct x86_emulate_ctxt *ctxt)
3413 {
3414 	unsigned long dr7;
3415 
3416 	ctxt->ops->get_dr(ctxt, 7, &dr7);
3417 
3418 	/* Check if DR7.Global_Enable is set */
3419 	return dr7 & (1 << 13);
3420 }
3421 
3422 static int check_dr_read(struct x86_emulate_ctxt *ctxt)
3423 {
3424 	int dr = ctxt->modrm_reg;
3425 	u64 cr4;
3426 
3427 	if (dr > 7)
3428 		return emulate_ud(ctxt);
3429 
3430 	cr4 = ctxt->ops->get_cr(ctxt, 4);
3431 	if ((cr4 & X86_CR4_DE) && (dr == 4 || dr == 5))
3432 		return emulate_ud(ctxt);
3433 
3434 	if (check_dr7_gd(ctxt))
3435 		return emulate_db(ctxt);
3436 
3437 	return X86EMUL_CONTINUE;
3438 }
3439 
3440 static int check_dr_write(struct x86_emulate_ctxt *ctxt)
3441 {
3442 	u64 new_val = ctxt->src.val64;
3443 	int dr = ctxt->modrm_reg;
3444 
3445 	if ((dr == 6 || dr == 7) && (new_val & 0xffffffff00000000ULL))
3446 		return emulate_gp(ctxt, 0);
3447 
3448 	return check_dr_read(ctxt);
3449 }
3450 
3451 static int check_svme(struct x86_emulate_ctxt *ctxt)
3452 {
3453 	u64 efer;
3454 
3455 	ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3456 
3457 	if (!(efer & EFER_SVME))
3458 		return emulate_ud(ctxt);
3459 
3460 	return X86EMUL_CONTINUE;
3461 }
3462 
3463 static int check_svme_pa(struct x86_emulate_ctxt *ctxt)
3464 {
3465 	u64 rax = reg_read(ctxt, VCPU_REGS_RAX);
3466 
3467 	/* Valid physical address? */
3468 	if (rax & 0xffff000000000000ULL)
3469 		return emulate_gp(ctxt, 0);
3470 
3471 	return check_svme(ctxt);
3472 }
3473 
3474 static int check_rdtsc(struct x86_emulate_ctxt *ctxt)
3475 {
3476 	u64 cr4 = ctxt->ops->get_cr(ctxt, 4);
3477 
3478 	if (cr4 & X86_CR4_TSD && ctxt->ops->cpl(ctxt))
3479 		return emulate_ud(ctxt);
3480 
3481 	return X86EMUL_CONTINUE;
3482 }
3483 
3484 static int check_rdpmc(struct x86_emulate_ctxt *ctxt)
3485 {
3486 	u64 cr4 = ctxt->ops->get_cr(ctxt, 4);
3487 	u64 rcx = reg_read(ctxt, VCPU_REGS_RCX);
3488 
3489 	if ((!(cr4 & X86_CR4_PCE) && ctxt->ops->cpl(ctxt)) ||
3490 	    (rcx > 3))
3491 		return emulate_gp(ctxt, 0);
3492 
3493 	return X86EMUL_CONTINUE;
3494 }
3495 
3496 static int check_perm_in(struct x86_emulate_ctxt *ctxt)
3497 {
3498 	ctxt->dst.bytes = min(ctxt->dst.bytes, 4u);
3499 	if (!emulator_io_permited(ctxt, ctxt->src.val, ctxt->dst.bytes))
3500 		return emulate_gp(ctxt, 0);
3501 
3502 	return X86EMUL_CONTINUE;
3503 }
3504 
3505 static int check_perm_out(struct x86_emulate_ctxt *ctxt)
3506 {
3507 	ctxt->src.bytes = min(ctxt->src.bytes, 4u);
3508 	if (!emulator_io_permited(ctxt, ctxt->dst.val, ctxt->src.bytes))
3509 		return emulate_gp(ctxt, 0);
3510 
3511 	return X86EMUL_CONTINUE;
3512 }
3513 
3514 #define D(_y) { .flags = (_y) }
3515 #define DI(_y, _i) { .flags = (_y), .intercept = x86_intercept_##_i }
3516 #define DIP(_y, _i, _p) { .flags = (_y), .intercept = x86_intercept_##_i, \
3517 		      .check_perm = (_p) }
3518 #define N    D(NotImpl)
3519 #define EXT(_f, _e) { .flags = ((_f) | RMExt), .u.group = (_e) }
3520 #define G(_f, _g) { .flags = ((_f) | Group | ModRM), .u.group = (_g) }
3521 #define GD(_f, _g) { .flags = ((_f) | GroupDual | ModRM), .u.gdual = (_g) }
3522 #define E(_f, _e) { .flags = ((_f) | Escape | ModRM), .u.esc = (_e) }
3523 #define I(_f, _e) { .flags = (_f), .u.execute = (_e) }
3524 #define F(_f, _e) { .flags = (_f) | Fastop, .u.fastop = (_e) }
3525 #define II(_f, _e, _i) \
3526 	{ .flags = (_f), .u.execute = (_e), .intercept = x86_intercept_##_i }
3527 #define IIP(_f, _e, _i, _p) \
3528 	{ .flags = (_f), .u.execute = (_e), .intercept = x86_intercept_##_i, \
3529 	  .check_perm = (_p) }
3530 #define GP(_f, _g) { .flags = ((_f) | Prefix), .u.gprefix = (_g) }
3531 
3532 #define D2bv(_f)      D((_f) | ByteOp), D(_f)
3533 #define D2bvIP(_f, _i, _p) DIP((_f) | ByteOp, _i, _p), DIP(_f, _i, _p)
3534 #define I2bv(_f, _e)  I((_f) | ByteOp, _e), I(_f, _e)
3535 #define F2bv(_f, _e)  F((_f) | ByteOp, _e), F(_f, _e)
3536 #define I2bvIP(_f, _e, _i, _p) \
3537 	IIP((_f) | ByteOp, _e, _i, _p), IIP(_f, _e, _i, _p)
3538 
3539 #define F6ALU(_f, _e) F2bv((_f) | DstMem | SrcReg | ModRM, _e),		\
3540 		F2bv(((_f) | DstReg | SrcMem | ModRM) & ~Lock, _e),	\
3541 		F2bv(((_f) & ~Lock) | DstAcc | SrcImm, _e)
3542 
3543 static const struct opcode group7_rm1[] = {
3544 	DI(SrcNone | Priv, monitor),
3545 	DI(SrcNone | Priv, mwait),
3546 	N, N, N, N, N, N,
3547 };
3548 
3549 static const struct opcode group7_rm3[] = {
3550 	DIP(SrcNone | Prot | Priv,		vmrun,		check_svme_pa),
3551 	II(SrcNone  | Prot | EmulateOnUD,	em_vmmcall,	vmmcall),
3552 	DIP(SrcNone | Prot | Priv,		vmload,		check_svme_pa),
3553 	DIP(SrcNone | Prot | Priv,		vmsave,		check_svme_pa),
3554 	DIP(SrcNone | Prot | Priv,		stgi,		check_svme),
3555 	DIP(SrcNone | Prot | Priv,		clgi,		check_svme),
3556 	DIP(SrcNone | Prot | Priv,		skinit,		check_svme),
3557 	DIP(SrcNone | Prot | Priv,		invlpga,	check_svme),
3558 };
3559 
3560 static const struct opcode group7_rm7[] = {
3561 	N,
3562 	DIP(SrcNone, rdtscp, check_rdtsc),
3563 	N, N, N, N, N, N,
3564 };
3565 
3566 static const struct opcode group1[] = {
3567 	F(Lock, em_add),
3568 	F(Lock | PageTable, em_or),
3569 	F(Lock, em_adc),
3570 	F(Lock, em_sbb),
3571 	F(Lock | PageTable, em_and),
3572 	F(Lock, em_sub),
3573 	F(Lock, em_xor),
3574 	F(NoWrite, em_cmp),
3575 };
3576 
3577 static const struct opcode group1A[] = {
3578 	I(DstMem | SrcNone | Mov | Stack, em_pop), N, N, N, N, N, N, N,
3579 };
3580 
3581 static const struct opcode group2[] = {
3582 	F(DstMem | ModRM, em_rol),
3583 	F(DstMem | ModRM, em_ror),
3584 	F(DstMem | ModRM, em_rcl),
3585 	F(DstMem | ModRM, em_rcr),
3586 	F(DstMem | ModRM, em_shl),
3587 	F(DstMem | ModRM, em_shr),
3588 	F(DstMem | ModRM, em_shl),
3589 	F(DstMem | ModRM, em_sar),
3590 };
3591 
3592 static const struct opcode group3[] = {
3593 	F(DstMem | SrcImm | NoWrite, em_test),
3594 	F(DstMem | SrcImm | NoWrite, em_test),
3595 	F(DstMem | SrcNone | Lock, em_not),
3596 	F(DstMem | SrcNone | Lock, em_neg),
3597 	F(DstXacc | Src2Mem, em_mul_ex),
3598 	F(DstXacc | Src2Mem, em_imul_ex),
3599 	F(DstXacc | Src2Mem, em_div_ex),
3600 	F(DstXacc | Src2Mem, em_idiv_ex),
3601 };
3602 
3603 static const struct opcode group4[] = {
3604 	F(ByteOp | DstMem | SrcNone | Lock, em_inc),
3605 	F(ByteOp | DstMem | SrcNone | Lock, em_dec),
3606 	N, N, N, N, N, N,
3607 };
3608 
3609 static const struct opcode group5[] = {
3610 	F(DstMem | SrcNone | Lock,		em_inc),
3611 	F(DstMem | SrcNone | Lock,		em_dec),
3612 	I(SrcMem | Stack,			em_grp45),
3613 	I(SrcMemFAddr | ImplicitOps | Stack,	em_call_far),
3614 	I(SrcMem | Stack,			em_grp45),
3615 	I(SrcMemFAddr | ImplicitOps,		em_grp45),
3616 	I(SrcMem | Stack,			em_grp45), D(Undefined),
3617 };
3618 
3619 static const struct opcode group6[] = {
3620 	DI(Prot,	sldt),
3621 	DI(Prot,	str),
3622 	II(Prot | Priv | SrcMem16, em_lldt, lldt),
3623 	II(Prot | Priv | SrcMem16, em_ltr, ltr),
3624 	N, N, N, N,
3625 };
3626 
3627 static const struct group_dual group7 = { {
3628 	II(Mov | DstMem | Priv,			em_sgdt, sgdt),
3629 	II(Mov | DstMem | Priv,			em_sidt, sidt),
3630 	II(SrcMem | Priv,			em_lgdt, lgdt),
3631 	II(SrcMem | Priv,			em_lidt, lidt),
3632 	II(SrcNone | DstMem | Mov,		em_smsw, smsw), N,
3633 	II(SrcMem16 | Mov | Priv,		em_lmsw, lmsw),
3634 	II(SrcMem | ByteOp | Priv | NoAccess,	em_invlpg, invlpg),
3635 }, {
3636 	I(SrcNone | Priv | EmulateOnUD,	em_vmcall),
3637 	EXT(0, group7_rm1),
3638 	N, EXT(0, group7_rm3),
3639 	II(SrcNone | DstMem | Mov,		em_smsw, smsw), N,
3640 	II(SrcMem16 | Mov | Priv,		em_lmsw, lmsw),
3641 	EXT(0, group7_rm7),
3642 } };
3643 
3644 static const struct opcode group8[] = {
3645 	N, N, N, N,
3646 	F(DstMem | SrcImmByte | NoWrite,		em_bt),
3647 	F(DstMem | SrcImmByte | Lock | PageTable,	em_bts),
3648 	F(DstMem | SrcImmByte | Lock,			em_btr),
3649 	F(DstMem | SrcImmByte | Lock | PageTable,	em_btc),
3650 };
3651 
3652 static const struct group_dual group9 = { {
3653 	N, I(DstMem64 | Lock | PageTable, em_cmpxchg8b), N, N, N, N, N, N,
3654 }, {
3655 	N, N, N, N, N, N, N, N,
3656 } };
3657 
3658 static const struct opcode group11[] = {
3659 	I(DstMem | SrcImm | Mov | PageTable, em_mov),
3660 	X7(D(Undefined)),
3661 };
3662 
3663 static const struct gprefix pfx_0f_6f_0f_7f = {
3664 	I(Mmx, em_mov), I(Sse | Aligned, em_mov), N, I(Sse | Unaligned, em_mov),
3665 };
3666 
3667 static const struct gprefix pfx_vmovntpx = {
3668 	I(0, em_mov), N, N, N,
3669 };
3670 
3671 static const struct escape escape_d9 = { {
3672 	N, N, N, N, N, N, N, I(DstMem, em_fnstcw),
3673 }, {
3674 	/* 0xC0 - 0xC7 */
3675 	N, N, N, N, N, N, N, N,
3676 	/* 0xC8 - 0xCF */
3677 	N, N, N, N, N, N, N, N,
3678 	/* 0xD0 - 0xC7 */
3679 	N, N, N, N, N, N, N, N,
3680 	/* 0xD8 - 0xDF */
3681 	N, N, N, N, N, N, N, N,
3682 	/* 0xE0 - 0xE7 */
3683 	N, N, N, N, N, N, N, N,
3684 	/* 0xE8 - 0xEF */
3685 	N, N, N, N, N, N, N, N,
3686 	/* 0xF0 - 0xF7 */
3687 	N, N, N, N, N, N, N, N,
3688 	/* 0xF8 - 0xFF */
3689 	N, N, N, N, N, N, N, N,
3690 } };
3691 
3692 static const struct escape escape_db = { {
3693 	N, N, N, N, N, N, N, N,
3694 }, {
3695 	/* 0xC0 - 0xC7 */
3696 	N, N, N, N, N, N, N, N,
3697 	/* 0xC8 - 0xCF */
3698 	N, N, N, N, N, N, N, N,
3699 	/* 0xD0 - 0xC7 */
3700 	N, N, N, N, N, N, N, N,
3701 	/* 0xD8 - 0xDF */
3702 	N, N, N, N, N, N, N, N,
3703 	/* 0xE0 - 0xE7 */
3704 	N, N, N, I(ImplicitOps, em_fninit), N, N, N, N,
3705 	/* 0xE8 - 0xEF */
3706 	N, N, N, N, N, N, N, N,
3707 	/* 0xF0 - 0xF7 */
3708 	N, N, N, N, N, N, N, N,
3709 	/* 0xF8 - 0xFF */
3710 	N, N, N, N, N, N, N, N,
3711 } };
3712 
3713 static const struct escape escape_dd = { {
3714 	N, N, N, N, N, N, N, I(DstMem, em_fnstsw),
3715 }, {
3716 	/* 0xC0 - 0xC7 */
3717 	N, N, N, N, N, N, N, N,
3718 	/* 0xC8 - 0xCF */
3719 	N, N, N, N, N, N, N, N,
3720 	/* 0xD0 - 0xC7 */
3721 	N, N, N, N, N, N, N, N,
3722 	/* 0xD8 - 0xDF */
3723 	N, N, N, N, N, N, N, N,
3724 	/* 0xE0 - 0xE7 */
3725 	N, N, N, N, N, N, N, N,
3726 	/* 0xE8 - 0xEF */
3727 	N, N, N, N, N, N, N, N,
3728 	/* 0xF0 - 0xF7 */
3729 	N, N, N, N, N, N, N, N,
3730 	/* 0xF8 - 0xFF */
3731 	N, N, N, N, N, N, N, N,
3732 } };
3733 
3734 static const struct opcode opcode_table[256] = {
3735 	/* 0x00 - 0x07 */
3736 	F6ALU(Lock, em_add),
3737 	I(ImplicitOps | Stack | No64 | Src2ES, em_push_sreg),
3738 	I(ImplicitOps | Stack | No64 | Src2ES, em_pop_sreg),
3739 	/* 0x08 - 0x0F */
3740 	F6ALU(Lock | PageTable, em_or),
3741 	I(ImplicitOps | Stack | No64 | Src2CS, em_push_sreg),
3742 	N,
3743 	/* 0x10 - 0x17 */
3744 	F6ALU(Lock, em_adc),
3745 	I(ImplicitOps | Stack | No64 | Src2SS, em_push_sreg),
3746 	I(ImplicitOps | Stack | No64 | Src2SS, em_pop_sreg),
3747 	/* 0x18 - 0x1F */
3748 	F6ALU(Lock, em_sbb),
3749 	I(ImplicitOps | Stack | No64 | Src2DS, em_push_sreg),
3750 	I(ImplicitOps | Stack | No64 | Src2DS, em_pop_sreg),
3751 	/* 0x20 - 0x27 */
3752 	F6ALU(Lock | PageTable, em_and), N, N,
3753 	/* 0x28 - 0x2F */
3754 	F6ALU(Lock, em_sub), N, I(ByteOp | DstAcc | No64, em_das),
3755 	/* 0x30 - 0x37 */
3756 	F6ALU(Lock, em_xor), N, N,
3757 	/* 0x38 - 0x3F */
3758 	F6ALU(NoWrite, em_cmp), N, N,
3759 	/* 0x40 - 0x4F */
3760 	X8(F(DstReg, em_inc)), X8(F(DstReg, em_dec)),
3761 	/* 0x50 - 0x57 */
3762 	X8(I(SrcReg | Stack, em_push)),
3763 	/* 0x58 - 0x5F */
3764 	X8(I(DstReg | Stack, em_pop)),
3765 	/* 0x60 - 0x67 */
3766 	I(ImplicitOps | Stack | No64, em_pusha),
3767 	I(ImplicitOps | Stack | No64, em_popa),
3768 	N, D(DstReg | SrcMem32 | ModRM | Mov) /* movsxd (x86/64) */ ,
3769 	N, N, N, N,
3770 	/* 0x68 - 0x6F */
3771 	I(SrcImm | Mov | Stack, em_push),
3772 	I(DstReg | SrcMem | ModRM | Src2Imm, em_imul_3op),
3773 	I(SrcImmByte | Mov | Stack, em_push),
3774 	I(DstReg | SrcMem | ModRM | Src2ImmByte, em_imul_3op),
3775 	I2bvIP(DstDI | SrcDX | Mov | String | Unaligned, em_in, ins, check_perm_in), /* insb, insw/insd */
3776 	I2bvIP(SrcSI | DstDX | String, em_out, outs, check_perm_out), /* outsb, outsw/outsd */
3777 	/* 0x70 - 0x7F */
3778 	X16(D(SrcImmByte)),
3779 	/* 0x80 - 0x87 */
3780 	G(ByteOp | DstMem | SrcImm, group1),
3781 	G(DstMem | SrcImm, group1),
3782 	G(ByteOp | DstMem | SrcImm | No64, group1),
3783 	G(DstMem | SrcImmByte, group1),
3784 	F2bv(DstMem | SrcReg | ModRM | NoWrite, em_test),
3785 	I2bv(DstMem | SrcReg | ModRM | Lock | PageTable, em_xchg),
3786 	/* 0x88 - 0x8F */
3787 	I2bv(DstMem | SrcReg | ModRM | Mov | PageTable, em_mov),
3788 	I2bv(DstReg | SrcMem | ModRM | Mov, em_mov),
3789 	I(DstMem | SrcNone | ModRM | Mov | PageTable, em_mov_rm_sreg),
3790 	D(ModRM | SrcMem | NoAccess | DstReg),
3791 	I(ImplicitOps | SrcMem16 | ModRM, em_mov_sreg_rm),
3792 	G(0, group1A),
3793 	/* 0x90 - 0x97 */
3794 	DI(SrcAcc | DstReg, pause), X7(D(SrcAcc | DstReg)),
3795 	/* 0x98 - 0x9F */
3796 	D(DstAcc | SrcNone), I(ImplicitOps | SrcAcc, em_cwd),
3797 	I(SrcImmFAddr | No64, em_call_far), N,
3798 	II(ImplicitOps | Stack, em_pushf, pushf),
3799 	II(ImplicitOps | Stack, em_popf, popf),
3800 	I(ImplicitOps, em_sahf), I(ImplicitOps, em_lahf),
3801 	/* 0xA0 - 0xA7 */
3802 	I2bv(DstAcc | SrcMem | Mov | MemAbs, em_mov),
3803 	I2bv(DstMem | SrcAcc | Mov | MemAbs | PageTable, em_mov),
3804 	I2bv(SrcSI | DstDI | Mov | String, em_mov),
3805 	F2bv(SrcSI | DstDI | String | NoWrite, em_cmp),
3806 	/* 0xA8 - 0xAF */
3807 	F2bv(DstAcc | SrcImm | NoWrite, em_test),
3808 	I2bv(SrcAcc | DstDI | Mov | String, em_mov),
3809 	I2bv(SrcSI | DstAcc | Mov | String, em_mov),
3810 	F2bv(SrcAcc | DstDI | String | NoWrite, em_cmp),
3811 	/* 0xB0 - 0xB7 */
3812 	X8(I(ByteOp | DstReg | SrcImm | Mov, em_mov)),
3813 	/* 0xB8 - 0xBF */
3814 	X8(I(DstReg | SrcImm64 | Mov, em_mov)),
3815 	/* 0xC0 - 0xC7 */
3816 	G(ByteOp | Src2ImmByte, group2), G(Src2ImmByte, group2),
3817 	I(ImplicitOps | Stack | SrcImmU16, em_ret_near_imm),
3818 	I(ImplicitOps | Stack, em_ret),
3819 	I(DstReg | SrcMemFAddr | ModRM | No64 | Src2ES, em_lseg),
3820 	I(DstReg | SrcMemFAddr | ModRM | No64 | Src2DS, em_lseg),
3821 	G(ByteOp, group11), G(0, group11),
3822 	/* 0xC8 - 0xCF */
3823 	I(Stack | SrcImmU16 | Src2ImmByte, em_enter), I(Stack, em_leave),
3824 	I(ImplicitOps | Stack | SrcImmU16, em_ret_far_imm),
3825 	I(ImplicitOps | Stack, em_ret_far),
3826 	D(ImplicitOps), DI(SrcImmByte, intn),
3827 	D(ImplicitOps | No64), II(ImplicitOps, em_iret, iret),
3828 	/* 0xD0 - 0xD7 */
3829 	G(Src2One | ByteOp, group2), G(Src2One, group2),
3830 	G(Src2CL | ByteOp, group2), G(Src2CL, group2),
3831 	I(DstAcc | SrcImmUByte | No64, em_aam),
3832 	I(DstAcc | SrcImmUByte | No64, em_aad),
3833 	F(DstAcc | ByteOp | No64, em_salc),
3834 	I(DstAcc | SrcXLat | ByteOp, em_mov),
3835 	/* 0xD8 - 0xDF */
3836 	N, E(0, &escape_d9), N, E(0, &escape_db), N, E(0, &escape_dd), N, N,
3837 	/* 0xE0 - 0xE7 */
3838 	X3(I(SrcImmByte, em_loop)),
3839 	I(SrcImmByte, em_jcxz),
3840 	I2bvIP(SrcImmUByte | DstAcc, em_in,  in,  check_perm_in),
3841 	I2bvIP(SrcAcc | DstImmUByte, em_out, out, check_perm_out),
3842 	/* 0xE8 - 0xEF */
3843 	I(SrcImm | Stack, em_call), D(SrcImm | ImplicitOps),
3844 	I(SrcImmFAddr | No64, em_jmp_far), D(SrcImmByte | ImplicitOps),
3845 	I2bvIP(SrcDX | DstAcc, em_in,  in,  check_perm_in),
3846 	I2bvIP(SrcAcc | DstDX, em_out, out, check_perm_out),
3847 	/* 0xF0 - 0xF7 */
3848 	N, DI(ImplicitOps, icebp), N, N,
3849 	DI(ImplicitOps | Priv, hlt), D(ImplicitOps),
3850 	G(ByteOp, group3), G(0, group3),
3851 	/* 0xF8 - 0xFF */
3852 	D(ImplicitOps), D(ImplicitOps),
3853 	I(ImplicitOps, em_cli), I(ImplicitOps, em_sti),
3854 	D(ImplicitOps), D(ImplicitOps), G(0, group4), G(0, group5),
3855 };
3856 
3857 static const struct opcode twobyte_table[256] = {
3858 	/* 0x00 - 0x0F */
3859 	G(0, group6), GD(0, &group7), N, N,
3860 	N, I(ImplicitOps | EmulateOnUD, em_syscall),
3861 	II(ImplicitOps | Priv, em_clts, clts), N,
3862 	DI(ImplicitOps | Priv, invd), DI(ImplicitOps | Priv, wbinvd), N, N,
3863 	N, D(ImplicitOps | ModRM), N, N,
3864 	/* 0x10 - 0x1F */
3865 	N, N, N, N, N, N, N, N,
3866 	D(ImplicitOps | ModRM), N, N, N, N, N, N, D(ImplicitOps | ModRM),
3867 	/* 0x20 - 0x2F */
3868 	DIP(ModRM | DstMem | Priv | Op3264, cr_read, check_cr_read),
3869 	DIP(ModRM | DstMem | Priv | Op3264, dr_read, check_dr_read),
3870 	IIP(ModRM | SrcMem | Priv | Op3264, em_cr_write, cr_write, check_cr_write),
3871 	IIP(ModRM | SrcMem | Priv | Op3264, em_dr_write, dr_write, check_dr_write),
3872 	N, N, N, N,
3873 	N, N, N, GP(ModRM | DstMem | SrcReg | Sse | Mov | Aligned, &pfx_vmovntpx),
3874 	N, N, N, N,
3875 	/* 0x30 - 0x3F */
3876 	II(ImplicitOps | Priv, em_wrmsr, wrmsr),
3877 	IIP(ImplicitOps, em_rdtsc, rdtsc, check_rdtsc),
3878 	II(ImplicitOps | Priv, em_rdmsr, rdmsr),
3879 	IIP(ImplicitOps, em_rdpmc, rdpmc, check_rdpmc),
3880 	I(ImplicitOps | EmulateOnUD, em_sysenter),
3881 	I(ImplicitOps | Priv | EmulateOnUD, em_sysexit),
3882 	N, N,
3883 	N, N, N, N, N, N, N, N,
3884 	/* 0x40 - 0x4F */
3885 	X16(D(DstReg | SrcMem | ModRM | Mov)),
3886 	/* 0x50 - 0x5F */
3887 	N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
3888 	/* 0x60 - 0x6F */
3889 	N, N, N, N,
3890 	N, N, N, N,
3891 	N, N, N, N,
3892 	N, N, N, GP(SrcMem | DstReg | ModRM | Mov, &pfx_0f_6f_0f_7f),
3893 	/* 0x70 - 0x7F */
3894 	N, N, N, N,
3895 	N, N, N, N,
3896 	N, N, N, N,
3897 	N, N, N, GP(SrcReg | DstMem | ModRM | Mov, &pfx_0f_6f_0f_7f),
3898 	/* 0x80 - 0x8F */
3899 	X16(D(SrcImm)),
3900 	/* 0x90 - 0x9F */
3901 	X16(D(ByteOp | DstMem | SrcNone | ModRM| Mov)),
3902 	/* 0xA0 - 0xA7 */
3903 	I(Stack | Src2FS, em_push_sreg), I(Stack | Src2FS, em_pop_sreg),
3904 	II(ImplicitOps, em_cpuid, cpuid),
3905 	F(DstMem | SrcReg | ModRM | BitOp | NoWrite, em_bt),
3906 	F(DstMem | SrcReg | Src2ImmByte | ModRM, em_shld),
3907 	F(DstMem | SrcReg | Src2CL | ModRM, em_shld), N, N,
3908 	/* 0xA8 - 0xAF */
3909 	I(Stack | Src2GS, em_push_sreg), I(Stack | Src2GS, em_pop_sreg),
3910 	DI(ImplicitOps, rsm),
3911 	F(DstMem | SrcReg | ModRM | BitOp | Lock | PageTable, em_bts),
3912 	F(DstMem | SrcReg | Src2ImmByte | ModRM, em_shrd),
3913 	F(DstMem | SrcReg | Src2CL | ModRM, em_shrd),
3914 	D(ModRM), F(DstReg | SrcMem | ModRM, em_imul),
3915 	/* 0xB0 - 0xB7 */
3916 	I2bv(DstMem | SrcReg | ModRM | Lock | PageTable, em_cmpxchg),
3917 	I(DstReg | SrcMemFAddr | ModRM | Src2SS, em_lseg),
3918 	F(DstMem | SrcReg | ModRM | BitOp | Lock, em_btr),
3919 	I(DstReg | SrcMemFAddr | ModRM | Src2FS, em_lseg),
3920 	I(DstReg | SrcMemFAddr | ModRM | Src2GS, em_lseg),
3921 	D(DstReg | SrcMem8 | ModRM | Mov), D(DstReg | SrcMem16 | ModRM | Mov),
3922 	/* 0xB8 - 0xBF */
3923 	N, N,
3924 	G(BitOp, group8),
3925 	F(DstMem | SrcReg | ModRM | BitOp | Lock | PageTable, em_btc),
3926 	F(DstReg | SrcMem | ModRM, em_bsf), F(DstReg | SrcMem | ModRM, em_bsr),
3927 	D(DstReg | SrcMem8 | ModRM | Mov), D(DstReg | SrcMem16 | ModRM | Mov),
3928 	/* 0xC0 - 0xC7 */
3929 	F2bv(DstMem | SrcReg | ModRM | SrcWrite | Lock, em_xadd),
3930 	N, D(DstMem | SrcReg | ModRM | Mov),
3931 	N, N, N, GD(0, &group9),
3932 	/* 0xC8 - 0xCF */
3933 	X8(I(DstReg, em_bswap)),
3934 	/* 0xD0 - 0xDF */
3935 	N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
3936 	/* 0xE0 - 0xEF */
3937 	N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
3938 	/* 0xF0 - 0xFF */
3939 	N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N
3940 };
3941 
3942 static const struct gprefix three_byte_0f_38_f0 = {
3943 	I(DstReg | SrcMem | Mov, em_movbe), N, N, N
3944 };
3945 
3946 static const struct gprefix three_byte_0f_38_f1 = {
3947 	I(DstMem | SrcReg | Mov, em_movbe), N, N, N
3948 };
3949 
3950 /*
3951  * Insns below are selected by the prefix which indexed by the third opcode
3952  * byte.
3953  */
3954 static const struct opcode opcode_map_0f_38[256] = {
3955 	/* 0x00 - 0x7f */
3956 	X16(N), X16(N), X16(N), X16(N), X16(N), X16(N), X16(N), X16(N),
3957 	/* 0x80 - 0xef */
3958 	X16(N), X16(N), X16(N), X16(N), X16(N), X16(N), X16(N),
3959 	/* 0xf0 - 0xf1 */
3960 	GP(EmulateOnUD | ModRM | Prefix, &three_byte_0f_38_f0),
3961 	GP(EmulateOnUD | ModRM | Prefix, &three_byte_0f_38_f1),
3962 	/* 0xf2 - 0xff */
3963 	N, N, X4(N), X8(N)
3964 };
3965 
3966 #undef D
3967 #undef N
3968 #undef G
3969 #undef GD
3970 #undef I
3971 #undef GP
3972 #undef EXT
3973 
3974 #undef D2bv
3975 #undef D2bvIP
3976 #undef I2bv
3977 #undef I2bvIP
3978 #undef I6ALU
3979 
3980 static unsigned imm_size(struct x86_emulate_ctxt *ctxt)
3981 {
3982 	unsigned size;
3983 
3984 	size = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
3985 	if (size == 8)
3986 		size = 4;
3987 	return size;
3988 }
3989 
3990 static int decode_imm(struct x86_emulate_ctxt *ctxt, struct operand *op,
3991 		      unsigned size, bool sign_extension)
3992 {
3993 	int rc = X86EMUL_CONTINUE;
3994 
3995 	op->type = OP_IMM;
3996 	op->bytes = size;
3997 	op->addr.mem.ea = ctxt->_eip;
3998 	/* NB. Immediates are sign-extended as necessary. */
3999 	switch (op->bytes) {
4000 	case 1:
4001 		op->val = insn_fetch(s8, ctxt);
4002 		break;
4003 	case 2:
4004 		op->val = insn_fetch(s16, ctxt);
4005 		break;
4006 	case 4:
4007 		op->val = insn_fetch(s32, ctxt);
4008 		break;
4009 	case 8:
4010 		op->val = insn_fetch(s64, ctxt);
4011 		break;
4012 	}
4013 	if (!sign_extension) {
4014 		switch (op->bytes) {
4015 		case 1:
4016 			op->val &= 0xff;
4017 			break;
4018 		case 2:
4019 			op->val &= 0xffff;
4020 			break;
4021 		case 4:
4022 			op->val &= 0xffffffff;
4023 			break;
4024 		}
4025 	}
4026 done:
4027 	return rc;
4028 }
4029 
4030 static int decode_operand(struct x86_emulate_ctxt *ctxt, struct operand *op,
4031 			  unsigned d)
4032 {
4033 	int rc = X86EMUL_CONTINUE;
4034 
4035 	switch (d) {
4036 	case OpReg:
4037 		decode_register_operand(ctxt, op);
4038 		break;
4039 	case OpImmUByte:
4040 		rc = decode_imm(ctxt, op, 1, false);
4041 		break;
4042 	case OpMem:
4043 		ctxt->memop.bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4044 	mem_common:
4045 		*op = ctxt->memop;
4046 		ctxt->memopp = op;
4047 		if ((ctxt->d & BitOp) && op == &ctxt->dst)
4048 			fetch_bit_operand(ctxt);
4049 		op->orig_val = op->val;
4050 		break;
4051 	case OpMem64:
4052 		ctxt->memop.bytes = 8;
4053 		goto mem_common;
4054 	case OpAcc:
4055 		op->type = OP_REG;
4056 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4057 		op->addr.reg = reg_rmw(ctxt, VCPU_REGS_RAX);
4058 		fetch_register_operand(op);
4059 		op->orig_val = op->val;
4060 		break;
4061 	case OpAccLo:
4062 		op->type = OP_REG;
4063 		op->bytes = (ctxt->d & ByteOp) ? 2 : ctxt->op_bytes;
4064 		op->addr.reg = reg_rmw(ctxt, VCPU_REGS_RAX);
4065 		fetch_register_operand(op);
4066 		op->orig_val = op->val;
4067 		break;
4068 	case OpAccHi:
4069 		if (ctxt->d & ByteOp) {
4070 			op->type = OP_NONE;
4071 			break;
4072 		}
4073 		op->type = OP_REG;
4074 		op->bytes = ctxt->op_bytes;
4075 		op->addr.reg = reg_rmw(ctxt, VCPU_REGS_RDX);
4076 		fetch_register_operand(op);
4077 		op->orig_val = op->val;
4078 		break;
4079 	case OpDI:
4080 		op->type = OP_MEM;
4081 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4082 		op->addr.mem.ea =
4083 			register_address(ctxt, reg_read(ctxt, VCPU_REGS_RDI));
4084 		op->addr.mem.seg = VCPU_SREG_ES;
4085 		op->val = 0;
4086 		op->count = 1;
4087 		break;
4088 	case OpDX:
4089 		op->type = OP_REG;
4090 		op->bytes = 2;
4091 		op->addr.reg = reg_rmw(ctxt, VCPU_REGS_RDX);
4092 		fetch_register_operand(op);
4093 		break;
4094 	case OpCL:
4095 		op->bytes = 1;
4096 		op->val = reg_read(ctxt, VCPU_REGS_RCX) & 0xff;
4097 		break;
4098 	case OpImmByte:
4099 		rc = decode_imm(ctxt, op, 1, true);
4100 		break;
4101 	case OpOne:
4102 		op->bytes = 1;
4103 		op->val = 1;
4104 		break;
4105 	case OpImm:
4106 		rc = decode_imm(ctxt, op, imm_size(ctxt), true);
4107 		break;
4108 	case OpImm64:
4109 		rc = decode_imm(ctxt, op, ctxt->op_bytes, true);
4110 		break;
4111 	case OpMem8:
4112 		ctxt->memop.bytes = 1;
4113 		if (ctxt->memop.type == OP_REG) {
4114 			ctxt->memop.addr.reg = decode_register(ctxt,
4115 					ctxt->modrm_rm, true);
4116 			fetch_register_operand(&ctxt->memop);
4117 		}
4118 		goto mem_common;
4119 	case OpMem16:
4120 		ctxt->memop.bytes = 2;
4121 		goto mem_common;
4122 	case OpMem32:
4123 		ctxt->memop.bytes = 4;
4124 		goto mem_common;
4125 	case OpImmU16:
4126 		rc = decode_imm(ctxt, op, 2, false);
4127 		break;
4128 	case OpImmU:
4129 		rc = decode_imm(ctxt, op, imm_size(ctxt), false);
4130 		break;
4131 	case OpSI:
4132 		op->type = OP_MEM;
4133 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4134 		op->addr.mem.ea =
4135 			register_address(ctxt, reg_read(ctxt, VCPU_REGS_RSI));
4136 		op->addr.mem.seg = seg_override(ctxt);
4137 		op->val = 0;
4138 		op->count = 1;
4139 		break;
4140 	case OpXLat:
4141 		op->type = OP_MEM;
4142 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4143 		op->addr.mem.ea =
4144 			register_address(ctxt,
4145 				reg_read(ctxt, VCPU_REGS_RBX) +
4146 				(reg_read(ctxt, VCPU_REGS_RAX) & 0xff));
4147 		op->addr.mem.seg = seg_override(ctxt);
4148 		op->val = 0;
4149 		break;
4150 	case OpImmFAddr:
4151 		op->type = OP_IMM;
4152 		op->addr.mem.ea = ctxt->_eip;
4153 		op->bytes = ctxt->op_bytes + 2;
4154 		insn_fetch_arr(op->valptr, op->bytes, ctxt);
4155 		break;
4156 	case OpMemFAddr:
4157 		ctxt->memop.bytes = ctxt->op_bytes + 2;
4158 		goto mem_common;
4159 	case OpES:
4160 		op->val = VCPU_SREG_ES;
4161 		break;
4162 	case OpCS:
4163 		op->val = VCPU_SREG_CS;
4164 		break;
4165 	case OpSS:
4166 		op->val = VCPU_SREG_SS;
4167 		break;
4168 	case OpDS:
4169 		op->val = VCPU_SREG_DS;
4170 		break;
4171 	case OpFS:
4172 		op->val = VCPU_SREG_FS;
4173 		break;
4174 	case OpGS:
4175 		op->val = VCPU_SREG_GS;
4176 		break;
4177 	case OpImplicit:
4178 		/* Special instructions do their own operand decoding. */
4179 	default:
4180 		op->type = OP_NONE; /* Disable writeback. */
4181 		break;
4182 	}
4183 
4184 done:
4185 	return rc;
4186 }
4187 
4188 int x86_decode_insn(struct x86_emulate_ctxt *ctxt, void *insn, int insn_len)
4189 {
4190 	int rc = X86EMUL_CONTINUE;
4191 	int mode = ctxt->mode;
4192 	int def_op_bytes, def_ad_bytes, goffset, simd_prefix;
4193 	bool op_prefix = false;
4194 	struct opcode opcode;
4195 
4196 	ctxt->memop.type = OP_NONE;
4197 	ctxt->memopp = NULL;
4198 	ctxt->_eip = ctxt->eip;
4199 	ctxt->fetch.start = ctxt->_eip;
4200 	ctxt->fetch.end = ctxt->fetch.start + insn_len;
4201 	ctxt->opcode_len = 1;
4202 	if (insn_len > 0)
4203 		memcpy(ctxt->fetch.data, insn, insn_len);
4204 
4205 	switch (mode) {
4206 	case X86EMUL_MODE_REAL:
4207 	case X86EMUL_MODE_VM86:
4208 	case X86EMUL_MODE_PROT16:
4209 		def_op_bytes = def_ad_bytes = 2;
4210 		break;
4211 	case X86EMUL_MODE_PROT32:
4212 		def_op_bytes = def_ad_bytes = 4;
4213 		break;
4214 #ifdef CONFIG_X86_64
4215 	case X86EMUL_MODE_PROT64:
4216 		def_op_bytes = 4;
4217 		def_ad_bytes = 8;
4218 		break;
4219 #endif
4220 	default:
4221 		return EMULATION_FAILED;
4222 	}
4223 
4224 	ctxt->op_bytes = def_op_bytes;
4225 	ctxt->ad_bytes = def_ad_bytes;
4226 
4227 	/* Legacy prefixes. */
4228 	for (;;) {
4229 		switch (ctxt->b = insn_fetch(u8, ctxt)) {
4230 		case 0x66:	/* operand-size override */
4231 			op_prefix = true;
4232 			/* switch between 2/4 bytes */
4233 			ctxt->op_bytes = def_op_bytes ^ 6;
4234 			break;
4235 		case 0x67:	/* address-size override */
4236 			if (mode == X86EMUL_MODE_PROT64)
4237 				/* switch between 4/8 bytes */
4238 				ctxt->ad_bytes = def_ad_bytes ^ 12;
4239 			else
4240 				/* switch between 2/4 bytes */
4241 				ctxt->ad_bytes = def_ad_bytes ^ 6;
4242 			break;
4243 		case 0x26:	/* ES override */
4244 		case 0x2e:	/* CS override */
4245 		case 0x36:	/* SS override */
4246 		case 0x3e:	/* DS override */
4247 			set_seg_override(ctxt, (ctxt->b >> 3) & 3);
4248 			break;
4249 		case 0x64:	/* FS override */
4250 		case 0x65:	/* GS override */
4251 			set_seg_override(ctxt, ctxt->b & 7);
4252 			break;
4253 		case 0x40 ... 0x4f: /* REX */
4254 			if (mode != X86EMUL_MODE_PROT64)
4255 				goto done_prefixes;
4256 			ctxt->rex_prefix = ctxt->b;
4257 			continue;
4258 		case 0xf0:	/* LOCK */
4259 			ctxt->lock_prefix = 1;
4260 			break;
4261 		case 0xf2:	/* REPNE/REPNZ */
4262 		case 0xf3:	/* REP/REPE/REPZ */
4263 			ctxt->rep_prefix = ctxt->b;
4264 			break;
4265 		default:
4266 			goto done_prefixes;
4267 		}
4268 
4269 		/* Any legacy prefix after a REX prefix nullifies its effect. */
4270 
4271 		ctxt->rex_prefix = 0;
4272 	}
4273 
4274 done_prefixes:
4275 
4276 	/* REX prefix. */
4277 	if (ctxt->rex_prefix & 8)
4278 		ctxt->op_bytes = 8;	/* REX.W */
4279 
4280 	/* Opcode byte(s). */
4281 	opcode = opcode_table[ctxt->b];
4282 	/* Two-byte opcode? */
4283 	if (ctxt->b == 0x0f) {
4284 		ctxt->opcode_len = 2;
4285 		ctxt->b = insn_fetch(u8, ctxt);
4286 		opcode = twobyte_table[ctxt->b];
4287 
4288 		/* 0F_38 opcode map */
4289 		if (ctxt->b == 0x38) {
4290 			ctxt->opcode_len = 3;
4291 			ctxt->b = insn_fetch(u8, ctxt);
4292 			opcode = opcode_map_0f_38[ctxt->b];
4293 		}
4294 	}
4295 	ctxt->d = opcode.flags;
4296 
4297 	if (ctxt->d & ModRM)
4298 		ctxt->modrm = insn_fetch(u8, ctxt);
4299 
4300 	while (ctxt->d & GroupMask) {
4301 		switch (ctxt->d & GroupMask) {
4302 		case Group:
4303 			goffset = (ctxt->modrm >> 3) & 7;
4304 			opcode = opcode.u.group[goffset];
4305 			break;
4306 		case GroupDual:
4307 			goffset = (ctxt->modrm >> 3) & 7;
4308 			if ((ctxt->modrm >> 6) == 3)
4309 				opcode = opcode.u.gdual->mod3[goffset];
4310 			else
4311 				opcode = opcode.u.gdual->mod012[goffset];
4312 			break;
4313 		case RMExt:
4314 			goffset = ctxt->modrm & 7;
4315 			opcode = opcode.u.group[goffset];
4316 			break;
4317 		case Prefix:
4318 			if (ctxt->rep_prefix && op_prefix)
4319 				return EMULATION_FAILED;
4320 			simd_prefix = op_prefix ? 0x66 : ctxt->rep_prefix;
4321 			switch (simd_prefix) {
4322 			case 0x00: opcode = opcode.u.gprefix->pfx_no; break;
4323 			case 0x66: opcode = opcode.u.gprefix->pfx_66; break;
4324 			case 0xf2: opcode = opcode.u.gprefix->pfx_f2; break;
4325 			case 0xf3: opcode = opcode.u.gprefix->pfx_f3; break;
4326 			}
4327 			break;
4328 		case Escape:
4329 			if (ctxt->modrm > 0xbf)
4330 				opcode = opcode.u.esc->high[ctxt->modrm - 0xc0];
4331 			else
4332 				opcode = opcode.u.esc->op[(ctxt->modrm >> 3) & 7];
4333 			break;
4334 		default:
4335 			return EMULATION_FAILED;
4336 		}
4337 
4338 		ctxt->d &= ~(u64)GroupMask;
4339 		ctxt->d |= opcode.flags;
4340 	}
4341 
4342 	ctxt->execute = opcode.u.execute;
4343 	ctxt->check_perm = opcode.check_perm;
4344 	ctxt->intercept = opcode.intercept;
4345 
4346 	/* Unrecognised? */
4347 	if (ctxt->d == 0 || (ctxt->d & NotImpl))
4348 		return EMULATION_FAILED;
4349 
4350 	if (!(ctxt->d & EmulateOnUD) && ctxt->ud)
4351 		return EMULATION_FAILED;
4352 
4353 	if (mode == X86EMUL_MODE_PROT64 && (ctxt->d & Stack))
4354 		ctxt->op_bytes = 8;
4355 
4356 	if (ctxt->d & Op3264) {
4357 		if (mode == X86EMUL_MODE_PROT64)
4358 			ctxt->op_bytes = 8;
4359 		else
4360 			ctxt->op_bytes = 4;
4361 	}
4362 
4363 	if (ctxt->d & Sse)
4364 		ctxt->op_bytes = 16;
4365 	else if (ctxt->d & Mmx)
4366 		ctxt->op_bytes = 8;
4367 
4368 	/* ModRM and SIB bytes. */
4369 	if (ctxt->d & ModRM) {
4370 		rc = decode_modrm(ctxt, &ctxt->memop);
4371 		if (!ctxt->has_seg_override)
4372 			set_seg_override(ctxt, ctxt->modrm_seg);
4373 	} else if (ctxt->d & MemAbs)
4374 		rc = decode_abs(ctxt, &ctxt->memop);
4375 	if (rc != X86EMUL_CONTINUE)
4376 		goto done;
4377 
4378 	if (!ctxt->has_seg_override)
4379 		set_seg_override(ctxt, VCPU_SREG_DS);
4380 
4381 	ctxt->memop.addr.mem.seg = seg_override(ctxt);
4382 
4383 	if (ctxt->memop.type == OP_MEM && ctxt->ad_bytes != 8)
4384 		ctxt->memop.addr.mem.ea = (u32)ctxt->memop.addr.mem.ea;
4385 
4386 	/*
4387 	 * Decode and fetch the source operand: register, memory
4388 	 * or immediate.
4389 	 */
4390 	rc = decode_operand(ctxt, &ctxt->src, (ctxt->d >> SrcShift) & OpMask);
4391 	if (rc != X86EMUL_CONTINUE)
4392 		goto done;
4393 
4394 	/*
4395 	 * Decode and fetch the second source operand: register, memory
4396 	 * or immediate.
4397 	 */
4398 	rc = decode_operand(ctxt, &ctxt->src2, (ctxt->d >> Src2Shift) & OpMask);
4399 	if (rc != X86EMUL_CONTINUE)
4400 		goto done;
4401 
4402 	/* Decode and fetch the destination operand: register or memory. */
4403 	rc = decode_operand(ctxt, &ctxt->dst, (ctxt->d >> DstShift) & OpMask);
4404 
4405 done:
4406 	if (ctxt->memopp && ctxt->memopp->type == OP_MEM && ctxt->rip_relative)
4407 		ctxt->memopp->addr.mem.ea += ctxt->_eip;
4408 
4409 	return (rc != X86EMUL_CONTINUE) ? EMULATION_FAILED : EMULATION_OK;
4410 }
4411 
4412 bool x86_page_table_writing_insn(struct x86_emulate_ctxt *ctxt)
4413 {
4414 	return ctxt->d & PageTable;
4415 }
4416 
4417 static bool string_insn_completed(struct x86_emulate_ctxt *ctxt)
4418 {
4419 	/* The second termination condition only applies for REPE
4420 	 * and REPNE. Test if the repeat string operation prefix is
4421 	 * REPE/REPZ or REPNE/REPNZ and if it's the case it tests the
4422 	 * corresponding termination condition according to:
4423 	 * 	- if REPE/REPZ and ZF = 0 then done
4424 	 * 	- if REPNE/REPNZ and ZF = 1 then done
4425 	 */
4426 	if (((ctxt->b == 0xa6) || (ctxt->b == 0xa7) ||
4427 	     (ctxt->b == 0xae) || (ctxt->b == 0xaf))
4428 	    && (((ctxt->rep_prefix == REPE_PREFIX) &&
4429 		 ((ctxt->eflags & EFLG_ZF) == 0))
4430 		|| ((ctxt->rep_prefix == REPNE_PREFIX) &&
4431 		    ((ctxt->eflags & EFLG_ZF) == EFLG_ZF))))
4432 		return true;
4433 
4434 	return false;
4435 }
4436 
4437 static int flush_pending_x87_faults(struct x86_emulate_ctxt *ctxt)
4438 {
4439 	bool fault = false;
4440 
4441 	ctxt->ops->get_fpu(ctxt);
4442 	asm volatile("1: fwait \n\t"
4443 		     "2: \n\t"
4444 		     ".pushsection .fixup,\"ax\" \n\t"
4445 		     "3: \n\t"
4446 		     "movb $1, %[fault] \n\t"
4447 		     "jmp 2b \n\t"
4448 		     ".popsection \n\t"
4449 		     _ASM_EXTABLE(1b, 3b)
4450 		     : [fault]"+qm"(fault));
4451 	ctxt->ops->put_fpu(ctxt);
4452 
4453 	if (unlikely(fault))
4454 		return emulate_exception(ctxt, MF_VECTOR, 0, false);
4455 
4456 	return X86EMUL_CONTINUE;
4457 }
4458 
4459 static void fetch_possible_mmx_operand(struct x86_emulate_ctxt *ctxt,
4460 				       struct operand *op)
4461 {
4462 	if (op->type == OP_MM)
4463 		read_mmx_reg(ctxt, &op->mm_val, op->addr.mm);
4464 }
4465 
4466 static int fastop(struct x86_emulate_ctxt *ctxt, void (*fop)(struct fastop *))
4467 {
4468 	ulong flags = (ctxt->eflags & EFLAGS_MASK) | X86_EFLAGS_IF;
4469 	if (!(ctxt->d & ByteOp))
4470 		fop += __ffs(ctxt->dst.bytes) * FASTOP_SIZE;
4471 	asm("push %[flags]; popf; call *%[fastop]; pushf; pop %[flags]\n"
4472 	    : "+a"(ctxt->dst.val), "+d"(ctxt->src.val), [flags]"+D"(flags),
4473 	      [fastop]"+S"(fop)
4474 	    : "c"(ctxt->src2.val));
4475 	ctxt->eflags = (ctxt->eflags & ~EFLAGS_MASK) | (flags & EFLAGS_MASK);
4476 	if (!fop) /* exception is returned in fop variable */
4477 		return emulate_de(ctxt);
4478 	return X86EMUL_CONTINUE;
4479 }
4480 
4481 int x86_emulate_insn(struct x86_emulate_ctxt *ctxt)
4482 {
4483 	const struct x86_emulate_ops *ops = ctxt->ops;
4484 	int rc = X86EMUL_CONTINUE;
4485 	int saved_dst_type = ctxt->dst.type;
4486 
4487 	ctxt->mem_read.pos = 0;
4488 
4489 	if ((ctxt->mode == X86EMUL_MODE_PROT64 && (ctxt->d & No64)) ||
4490 			(ctxt->d & Undefined)) {
4491 		rc = emulate_ud(ctxt);
4492 		goto done;
4493 	}
4494 
4495 	/* LOCK prefix is allowed only with some instructions */
4496 	if (ctxt->lock_prefix && (!(ctxt->d & Lock) || ctxt->dst.type != OP_MEM)) {
4497 		rc = emulate_ud(ctxt);
4498 		goto done;
4499 	}
4500 
4501 	if ((ctxt->d & SrcMask) == SrcMemFAddr && ctxt->src.type != OP_MEM) {
4502 		rc = emulate_ud(ctxt);
4503 		goto done;
4504 	}
4505 
4506 	if (((ctxt->d & (Sse|Mmx)) && ((ops->get_cr(ctxt, 0) & X86_CR0_EM)))
4507 	    || ((ctxt->d & Sse) && !(ops->get_cr(ctxt, 4) & X86_CR4_OSFXSR))) {
4508 		rc = emulate_ud(ctxt);
4509 		goto done;
4510 	}
4511 
4512 	if ((ctxt->d & (Sse|Mmx)) && (ops->get_cr(ctxt, 0) & X86_CR0_TS)) {
4513 		rc = emulate_nm(ctxt);
4514 		goto done;
4515 	}
4516 
4517 	if (ctxt->d & Mmx) {
4518 		rc = flush_pending_x87_faults(ctxt);
4519 		if (rc != X86EMUL_CONTINUE)
4520 			goto done;
4521 		/*
4522 		 * Now that we know the fpu is exception safe, we can fetch
4523 		 * operands from it.
4524 		 */
4525 		fetch_possible_mmx_operand(ctxt, &ctxt->src);
4526 		fetch_possible_mmx_operand(ctxt, &ctxt->src2);
4527 		if (!(ctxt->d & Mov))
4528 			fetch_possible_mmx_operand(ctxt, &ctxt->dst);
4529 	}
4530 
4531 	if (unlikely(ctxt->guest_mode) && ctxt->intercept) {
4532 		rc = emulator_check_intercept(ctxt, ctxt->intercept,
4533 					      X86_ICPT_PRE_EXCEPT);
4534 		if (rc != X86EMUL_CONTINUE)
4535 			goto done;
4536 	}
4537 
4538 	/* Privileged instruction can be executed only in CPL=0 */
4539 	if ((ctxt->d & Priv) && ops->cpl(ctxt)) {
4540 		rc = emulate_gp(ctxt, 0);
4541 		goto done;
4542 	}
4543 
4544 	/* Instruction can only be executed in protected mode */
4545 	if ((ctxt->d & Prot) && ctxt->mode < X86EMUL_MODE_PROT16) {
4546 		rc = emulate_ud(ctxt);
4547 		goto done;
4548 	}
4549 
4550 	/* Do instruction specific permission checks */
4551 	if (ctxt->check_perm) {
4552 		rc = ctxt->check_perm(ctxt);
4553 		if (rc != X86EMUL_CONTINUE)
4554 			goto done;
4555 	}
4556 
4557 	if (unlikely(ctxt->guest_mode) && ctxt->intercept) {
4558 		rc = emulator_check_intercept(ctxt, ctxt->intercept,
4559 					      X86_ICPT_POST_EXCEPT);
4560 		if (rc != X86EMUL_CONTINUE)
4561 			goto done;
4562 	}
4563 
4564 	if (ctxt->rep_prefix && (ctxt->d & String)) {
4565 		/* All REP prefixes have the same first termination condition */
4566 		if (address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) == 0) {
4567 			ctxt->eip = ctxt->_eip;
4568 			goto done;
4569 		}
4570 	}
4571 
4572 	if ((ctxt->src.type == OP_MEM) && !(ctxt->d & NoAccess)) {
4573 		rc = segmented_read(ctxt, ctxt->src.addr.mem,
4574 				    ctxt->src.valptr, ctxt->src.bytes);
4575 		if (rc != X86EMUL_CONTINUE)
4576 			goto done;
4577 		ctxt->src.orig_val64 = ctxt->src.val64;
4578 	}
4579 
4580 	if (ctxt->src2.type == OP_MEM) {
4581 		rc = segmented_read(ctxt, ctxt->src2.addr.mem,
4582 				    &ctxt->src2.val, ctxt->src2.bytes);
4583 		if (rc != X86EMUL_CONTINUE)
4584 			goto done;
4585 	}
4586 
4587 	if ((ctxt->d & DstMask) == ImplicitOps)
4588 		goto special_insn;
4589 
4590 
4591 	if ((ctxt->dst.type == OP_MEM) && !(ctxt->d & Mov)) {
4592 		/* optimisation - avoid slow emulated read if Mov */
4593 		rc = segmented_read(ctxt, ctxt->dst.addr.mem,
4594 				   &ctxt->dst.val, ctxt->dst.bytes);
4595 		if (rc != X86EMUL_CONTINUE)
4596 			goto done;
4597 	}
4598 	ctxt->dst.orig_val = ctxt->dst.val;
4599 
4600 special_insn:
4601 
4602 	if (unlikely(ctxt->guest_mode) && ctxt->intercept) {
4603 		rc = emulator_check_intercept(ctxt, ctxt->intercept,
4604 					      X86_ICPT_POST_MEMACCESS);
4605 		if (rc != X86EMUL_CONTINUE)
4606 			goto done;
4607 	}
4608 
4609 	if (ctxt->execute) {
4610 		if (ctxt->d & Fastop) {
4611 			void (*fop)(struct fastop *) = (void *)ctxt->execute;
4612 			rc = fastop(ctxt, fop);
4613 			if (rc != X86EMUL_CONTINUE)
4614 				goto done;
4615 			goto writeback;
4616 		}
4617 		rc = ctxt->execute(ctxt);
4618 		if (rc != X86EMUL_CONTINUE)
4619 			goto done;
4620 		goto writeback;
4621 	}
4622 
4623 	if (ctxt->opcode_len == 2)
4624 		goto twobyte_insn;
4625 	else if (ctxt->opcode_len == 3)
4626 		goto threebyte_insn;
4627 
4628 	switch (ctxt->b) {
4629 	case 0x63:		/* movsxd */
4630 		if (ctxt->mode != X86EMUL_MODE_PROT64)
4631 			goto cannot_emulate;
4632 		ctxt->dst.val = (s32) ctxt->src.val;
4633 		break;
4634 	case 0x70 ... 0x7f: /* jcc (short) */
4635 		if (test_cc(ctxt->b, ctxt->eflags))
4636 			jmp_rel(ctxt, ctxt->src.val);
4637 		break;
4638 	case 0x8d: /* lea r16/r32, m */
4639 		ctxt->dst.val = ctxt->src.addr.mem.ea;
4640 		break;
4641 	case 0x90 ... 0x97: /* nop / xchg reg, rax */
4642 		if (ctxt->dst.addr.reg == reg_rmw(ctxt, VCPU_REGS_RAX))
4643 			break;
4644 		rc = em_xchg(ctxt);
4645 		break;
4646 	case 0x98: /* cbw/cwde/cdqe */
4647 		switch (ctxt->op_bytes) {
4648 		case 2: ctxt->dst.val = (s8)ctxt->dst.val; break;
4649 		case 4: ctxt->dst.val = (s16)ctxt->dst.val; break;
4650 		case 8: ctxt->dst.val = (s32)ctxt->dst.val; break;
4651 		}
4652 		break;
4653 	case 0xcc:		/* int3 */
4654 		rc = emulate_int(ctxt, 3);
4655 		break;
4656 	case 0xcd:		/* int n */
4657 		rc = emulate_int(ctxt, ctxt->src.val);
4658 		break;
4659 	case 0xce:		/* into */
4660 		if (ctxt->eflags & EFLG_OF)
4661 			rc = emulate_int(ctxt, 4);
4662 		break;
4663 	case 0xe9: /* jmp rel */
4664 	case 0xeb: /* jmp rel short */
4665 		jmp_rel(ctxt, ctxt->src.val);
4666 		ctxt->dst.type = OP_NONE; /* Disable writeback. */
4667 		break;
4668 	case 0xf4:              /* hlt */
4669 		ctxt->ops->halt(ctxt);
4670 		break;
4671 	case 0xf5:	/* cmc */
4672 		/* complement carry flag from eflags reg */
4673 		ctxt->eflags ^= EFLG_CF;
4674 		break;
4675 	case 0xf8: /* clc */
4676 		ctxt->eflags &= ~EFLG_CF;
4677 		break;
4678 	case 0xf9: /* stc */
4679 		ctxt->eflags |= EFLG_CF;
4680 		break;
4681 	case 0xfc: /* cld */
4682 		ctxt->eflags &= ~EFLG_DF;
4683 		break;
4684 	case 0xfd: /* std */
4685 		ctxt->eflags |= EFLG_DF;
4686 		break;
4687 	default:
4688 		goto cannot_emulate;
4689 	}
4690 
4691 	if (rc != X86EMUL_CONTINUE)
4692 		goto done;
4693 
4694 writeback:
4695 	if (!(ctxt->d & NoWrite)) {
4696 		rc = writeback(ctxt, &ctxt->dst);
4697 		if (rc != X86EMUL_CONTINUE)
4698 			goto done;
4699 	}
4700 	if (ctxt->d & SrcWrite) {
4701 		BUG_ON(ctxt->src.type == OP_MEM || ctxt->src.type == OP_MEM_STR);
4702 		rc = writeback(ctxt, &ctxt->src);
4703 		if (rc != X86EMUL_CONTINUE)
4704 			goto done;
4705 	}
4706 
4707 	/*
4708 	 * restore dst type in case the decoding will be reused
4709 	 * (happens for string instruction )
4710 	 */
4711 	ctxt->dst.type = saved_dst_type;
4712 
4713 	if ((ctxt->d & SrcMask) == SrcSI)
4714 		string_addr_inc(ctxt, VCPU_REGS_RSI, &ctxt->src);
4715 
4716 	if ((ctxt->d & DstMask) == DstDI)
4717 		string_addr_inc(ctxt, VCPU_REGS_RDI, &ctxt->dst);
4718 
4719 	if (ctxt->rep_prefix && (ctxt->d & String)) {
4720 		unsigned int count;
4721 		struct read_cache *r = &ctxt->io_read;
4722 		if ((ctxt->d & SrcMask) == SrcSI)
4723 			count = ctxt->src.count;
4724 		else
4725 			count = ctxt->dst.count;
4726 		register_address_increment(ctxt, reg_rmw(ctxt, VCPU_REGS_RCX),
4727 				-count);
4728 
4729 		if (!string_insn_completed(ctxt)) {
4730 			/*
4731 			 * Re-enter guest when pio read ahead buffer is empty
4732 			 * or, if it is not used, after each 1024 iteration.
4733 			 */
4734 			if ((r->end != 0 || reg_read(ctxt, VCPU_REGS_RCX) & 0x3ff) &&
4735 			    (r->end == 0 || r->end != r->pos)) {
4736 				/*
4737 				 * Reset read cache. Usually happens before
4738 				 * decode, but since instruction is restarted
4739 				 * we have to do it here.
4740 				 */
4741 				ctxt->mem_read.end = 0;
4742 				writeback_registers(ctxt);
4743 				return EMULATION_RESTART;
4744 			}
4745 			goto done; /* skip rip writeback */
4746 		}
4747 	}
4748 
4749 	ctxt->eip = ctxt->_eip;
4750 
4751 done:
4752 	if (rc == X86EMUL_PROPAGATE_FAULT)
4753 		ctxt->have_exception = true;
4754 	if (rc == X86EMUL_INTERCEPTED)
4755 		return EMULATION_INTERCEPTED;
4756 
4757 	if (rc == X86EMUL_CONTINUE)
4758 		writeback_registers(ctxt);
4759 
4760 	return (rc == X86EMUL_UNHANDLEABLE) ? EMULATION_FAILED : EMULATION_OK;
4761 
4762 twobyte_insn:
4763 	switch (ctxt->b) {
4764 	case 0x09:		/* wbinvd */
4765 		(ctxt->ops->wbinvd)(ctxt);
4766 		break;
4767 	case 0x08:		/* invd */
4768 	case 0x0d:		/* GrpP (prefetch) */
4769 	case 0x18:		/* Grp16 (prefetch/nop) */
4770 	case 0x1f:		/* nop */
4771 		break;
4772 	case 0x20: /* mov cr, reg */
4773 		ctxt->dst.val = ops->get_cr(ctxt, ctxt->modrm_reg);
4774 		break;
4775 	case 0x21: /* mov from dr to reg */
4776 		ops->get_dr(ctxt, ctxt->modrm_reg, &ctxt->dst.val);
4777 		break;
4778 	case 0x40 ... 0x4f:	/* cmov */
4779 		ctxt->dst.val = ctxt->dst.orig_val = ctxt->src.val;
4780 		if (!test_cc(ctxt->b, ctxt->eflags))
4781 			ctxt->dst.type = OP_NONE; /* no writeback */
4782 		break;
4783 	case 0x80 ... 0x8f: /* jnz rel, etc*/
4784 		if (test_cc(ctxt->b, ctxt->eflags))
4785 			jmp_rel(ctxt, ctxt->src.val);
4786 		break;
4787 	case 0x90 ... 0x9f:     /* setcc r/m8 */
4788 		ctxt->dst.val = test_cc(ctxt->b, ctxt->eflags);
4789 		break;
4790 	case 0xae:              /* clflush */
4791 		break;
4792 	case 0xb6 ... 0xb7:	/* movzx */
4793 		ctxt->dst.bytes = ctxt->op_bytes;
4794 		ctxt->dst.val = (ctxt->src.bytes == 1) ? (u8) ctxt->src.val
4795 						       : (u16) ctxt->src.val;
4796 		break;
4797 	case 0xbe ... 0xbf:	/* movsx */
4798 		ctxt->dst.bytes = ctxt->op_bytes;
4799 		ctxt->dst.val = (ctxt->src.bytes == 1) ? (s8) ctxt->src.val :
4800 							(s16) ctxt->src.val;
4801 		break;
4802 	case 0xc3:		/* movnti */
4803 		ctxt->dst.bytes = ctxt->op_bytes;
4804 		ctxt->dst.val = (ctxt->op_bytes == 4) ? (u32) ctxt->src.val :
4805 							(u64) ctxt->src.val;
4806 		break;
4807 	default:
4808 		goto cannot_emulate;
4809 	}
4810 
4811 threebyte_insn:
4812 
4813 	if (rc != X86EMUL_CONTINUE)
4814 		goto done;
4815 
4816 	goto writeback;
4817 
4818 cannot_emulate:
4819 	return EMULATION_FAILED;
4820 }
4821 
4822 void emulator_invalidate_register_cache(struct x86_emulate_ctxt *ctxt)
4823 {
4824 	invalidate_registers(ctxt);
4825 }
4826 
4827 void emulator_writeback_register_cache(struct x86_emulate_ctxt *ctxt)
4828 {
4829 	writeback_registers(ctxt);
4830 }
4831