xref: /openbmc/linux/arch/x86/kvm/emulate.c (revision fea88a0c)
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 
28 #include "x86.h"
29 #include "tss.h"
30 
31 /*
32  * Operand types
33  */
34 #define OpNone             0ull
35 #define OpImplicit         1ull  /* No generic decode */
36 #define OpReg              2ull  /* Register */
37 #define OpMem              3ull  /* Memory */
38 #define OpAcc              4ull  /* Accumulator: AL/AX/EAX/RAX */
39 #define OpDI               5ull  /* ES:DI/EDI/RDI */
40 #define OpMem64            6ull  /* Memory, 64-bit */
41 #define OpImmUByte         7ull  /* Zero-extended 8-bit immediate */
42 #define OpDX               8ull  /* DX register */
43 #define OpCL               9ull  /* CL register (for shifts) */
44 #define OpImmByte         10ull  /* 8-bit sign extended immediate */
45 #define OpOne             11ull  /* Implied 1 */
46 #define OpImm             12ull  /* Sign extended immediate */
47 #define OpMem16           13ull  /* Memory operand (16-bit). */
48 #define OpMem32           14ull  /* Memory operand (32-bit). */
49 #define OpImmU            15ull  /* Immediate operand, zero extended */
50 #define OpSI              16ull  /* SI/ESI/RSI */
51 #define OpImmFAddr        17ull  /* Immediate far address */
52 #define OpMemFAddr        18ull  /* Far address in memory */
53 #define OpImmU16          19ull  /* Immediate operand, 16 bits, zero extended */
54 #define OpES              20ull  /* ES */
55 #define OpCS              21ull  /* CS */
56 #define OpSS              22ull  /* SS */
57 #define OpDS              23ull  /* DS */
58 #define OpFS              24ull  /* FS */
59 #define OpGS              25ull  /* GS */
60 #define OpMem8            26ull  /* 8-bit zero extended memory operand */
61 
62 #define OpBits             5  /* Width of operand field */
63 #define OpMask             ((1ull << OpBits) - 1)
64 
65 /*
66  * Opcode effective-address decode tables.
67  * Note that we only emulate instructions that have at least one memory
68  * operand (excluding implicit stack references). We assume that stack
69  * references and instruction fetches will never occur in special memory
70  * areas that require emulation. So, for example, 'mov <imm>,<reg>' need
71  * not be handled.
72  */
73 
74 /* Operand sizes: 8-bit operands or specified/overridden size. */
75 #define ByteOp      (1<<0)	/* 8-bit operands. */
76 /* Destination operand type. */
77 #define DstShift    1
78 #define ImplicitOps (OpImplicit << DstShift)
79 #define DstReg      (OpReg << DstShift)
80 #define DstMem      (OpMem << DstShift)
81 #define DstAcc      (OpAcc << DstShift)
82 #define DstDI       (OpDI << DstShift)
83 #define DstMem64    (OpMem64 << DstShift)
84 #define DstImmUByte (OpImmUByte << DstShift)
85 #define DstDX       (OpDX << DstShift)
86 #define DstMask     (OpMask << DstShift)
87 /* Source operand type. */
88 #define SrcShift    6
89 #define SrcNone     (OpNone << SrcShift)
90 #define SrcReg      (OpReg << SrcShift)
91 #define SrcMem      (OpMem << SrcShift)
92 #define SrcMem16    (OpMem16 << SrcShift)
93 #define SrcMem32    (OpMem32 << SrcShift)
94 #define SrcImm      (OpImm << SrcShift)
95 #define SrcImmByte  (OpImmByte << SrcShift)
96 #define SrcOne      (OpOne << SrcShift)
97 #define SrcImmUByte (OpImmUByte << SrcShift)
98 #define SrcImmU     (OpImmU << SrcShift)
99 #define SrcSI       (OpSI << SrcShift)
100 #define SrcImmFAddr (OpImmFAddr << SrcShift)
101 #define SrcMemFAddr (OpMemFAddr << SrcShift)
102 #define SrcAcc      (OpAcc << SrcShift)
103 #define SrcImmU16   (OpImmU16 << SrcShift)
104 #define SrcDX       (OpDX << SrcShift)
105 #define SrcMem8     (OpMem8 << SrcShift)
106 #define SrcMask     (OpMask << SrcShift)
107 #define BitOp       (1<<11)
108 #define MemAbs      (1<<12)      /* Memory operand is absolute displacement */
109 #define String      (1<<13)     /* String instruction (rep capable) */
110 #define Stack       (1<<14)     /* Stack instruction (push/pop) */
111 #define GroupMask   (7<<15)     /* Opcode uses one of the group mechanisms */
112 #define Group       (1<<15)     /* Bits 3:5 of modrm byte extend opcode */
113 #define GroupDual   (2<<15)     /* Alternate decoding of mod == 3 */
114 #define Prefix      (3<<15)     /* Instruction varies with 66/f2/f3 prefix */
115 #define RMExt       (4<<15)     /* Opcode extension in ModRM r/m if mod == 3 */
116 #define Sse         (1<<18)     /* SSE Vector instruction */
117 /* Generic ModRM decode. */
118 #define ModRM       (1<<19)
119 /* Destination is only written; never read. */
120 #define Mov         (1<<20)
121 /* Misc flags */
122 #define Prot        (1<<21) /* instruction generates #UD if not in prot-mode */
123 #define VendorSpecific (1<<22) /* Vendor specific instruction */
124 #define NoAccess    (1<<23) /* Don't access memory (lea/invlpg/verr etc) */
125 #define Op3264      (1<<24) /* Operand is 64b in long mode, 32b otherwise */
126 #define Undefined   (1<<25) /* No Such Instruction */
127 #define Lock        (1<<26) /* lock prefix is allowed for the instruction */
128 #define Priv        (1<<27) /* instruction generates #GP if current CPL != 0 */
129 #define No64	    (1<<28)
130 #define PageTable   (1 << 29)   /* instruction used to write page table */
131 /* Source 2 operand type */
132 #define Src2Shift   (30)
133 #define Src2None    (OpNone << Src2Shift)
134 #define Src2CL      (OpCL << Src2Shift)
135 #define Src2ImmByte (OpImmByte << Src2Shift)
136 #define Src2One     (OpOne << Src2Shift)
137 #define Src2Imm     (OpImm << Src2Shift)
138 #define Src2ES      (OpES << Src2Shift)
139 #define Src2CS      (OpCS << Src2Shift)
140 #define Src2SS      (OpSS << Src2Shift)
141 #define Src2DS      (OpDS << Src2Shift)
142 #define Src2FS      (OpFS << Src2Shift)
143 #define Src2GS      (OpGS << Src2Shift)
144 #define Src2Mask    (OpMask << Src2Shift)
145 
146 #define X2(x...) x, x
147 #define X3(x...) X2(x), x
148 #define X4(x...) X2(x), X2(x)
149 #define X5(x...) X4(x), x
150 #define X6(x...) X4(x), X2(x)
151 #define X7(x...) X4(x), X3(x)
152 #define X8(x...) X4(x), X4(x)
153 #define X16(x...) X8(x), X8(x)
154 
155 struct opcode {
156 	u64 flags : 56;
157 	u64 intercept : 8;
158 	union {
159 		int (*execute)(struct x86_emulate_ctxt *ctxt);
160 		struct opcode *group;
161 		struct group_dual *gdual;
162 		struct gprefix *gprefix;
163 	} u;
164 	int (*check_perm)(struct x86_emulate_ctxt *ctxt);
165 };
166 
167 struct group_dual {
168 	struct opcode mod012[8];
169 	struct opcode mod3[8];
170 };
171 
172 struct gprefix {
173 	struct opcode pfx_no;
174 	struct opcode pfx_66;
175 	struct opcode pfx_f2;
176 	struct opcode pfx_f3;
177 };
178 
179 /* EFLAGS bit definitions. */
180 #define EFLG_ID (1<<21)
181 #define EFLG_VIP (1<<20)
182 #define EFLG_VIF (1<<19)
183 #define EFLG_AC (1<<18)
184 #define EFLG_VM (1<<17)
185 #define EFLG_RF (1<<16)
186 #define EFLG_IOPL (3<<12)
187 #define EFLG_NT (1<<14)
188 #define EFLG_OF (1<<11)
189 #define EFLG_DF (1<<10)
190 #define EFLG_IF (1<<9)
191 #define EFLG_TF (1<<8)
192 #define EFLG_SF (1<<7)
193 #define EFLG_ZF (1<<6)
194 #define EFLG_AF (1<<4)
195 #define EFLG_PF (1<<2)
196 #define EFLG_CF (1<<0)
197 
198 #define EFLG_RESERVED_ZEROS_MASK 0xffc0802a
199 #define EFLG_RESERVED_ONE_MASK 2
200 
201 /*
202  * Instruction emulation:
203  * Most instructions are emulated directly via a fragment of inline assembly
204  * code. This allows us to save/restore EFLAGS and thus very easily pick up
205  * any modified flags.
206  */
207 
208 #if defined(CONFIG_X86_64)
209 #define _LO32 "k"		/* force 32-bit operand */
210 #define _STK  "%%rsp"		/* stack pointer */
211 #elif defined(__i386__)
212 #define _LO32 ""		/* force 32-bit operand */
213 #define _STK  "%%esp"		/* stack pointer */
214 #endif
215 
216 /*
217  * These EFLAGS bits are restored from saved value during emulation, and
218  * any changes are written back to the saved value after emulation.
219  */
220 #define EFLAGS_MASK (EFLG_OF|EFLG_SF|EFLG_ZF|EFLG_AF|EFLG_PF|EFLG_CF)
221 
222 /* Before executing instruction: restore necessary bits in EFLAGS. */
223 #define _PRE_EFLAGS(_sav, _msk, _tmp)					\
224 	/* EFLAGS = (_sav & _msk) | (EFLAGS & ~_msk); _sav &= ~_msk; */ \
225 	"movl %"_sav",%"_LO32 _tmp"; "                                  \
226 	"push %"_tmp"; "                                                \
227 	"push %"_tmp"; "                                                \
228 	"movl %"_msk",%"_LO32 _tmp"; "                                  \
229 	"andl %"_LO32 _tmp",("_STK"); "                                 \
230 	"pushf; "                                                       \
231 	"notl %"_LO32 _tmp"; "                                          \
232 	"andl %"_LO32 _tmp",("_STK"); "                                 \
233 	"andl %"_LO32 _tmp","__stringify(BITS_PER_LONG/4)"("_STK"); "	\
234 	"pop  %"_tmp"; "                                                \
235 	"orl  %"_LO32 _tmp",("_STK"); "                                 \
236 	"popf; "                                                        \
237 	"pop  %"_sav"; "
238 
239 /* After executing instruction: write-back necessary bits in EFLAGS. */
240 #define _POST_EFLAGS(_sav, _msk, _tmp) \
241 	/* _sav |= EFLAGS & _msk; */		\
242 	"pushf; "				\
243 	"pop  %"_tmp"; "			\
244 	"andl %"_msk",%"_LO32 _tmp"; "		\
245 	"orl  %"_LO32 _tmp",%"_sav"; "
246 
247 #ifdef CONFIG_X86_64
248 #define ON64(x) x
249 #else
250 #define ON64(x)
251 #endif
252 
253 #define ____emulate_2op(ctxt, _op, _x, _y, _suffix, _dsttype)	\
254 	do {								\
255 		__asm__ __volatile__ (					\
256 			_PRE_EFLAGS("0", "4", "2")			\
257 			_op _suffix " %"_x"3,%1; "			\
258 			_POST_EFLAGS("0", "4", "2")			\
259 			: "=m" ((ctxt)->eflags),			\
260 			  "+q" (*(_dsttype*)&(ctxt)->dst.val),		\
261 			  "=&r" (_tmp)					\
262 			: _y ((ctxt)->src.val), "i" (EFLAGS_MASK));	\
263 	} while (0)
264 
265 
266 /* Raw emulation: instruction has two explicit operands. */
267 #define __emulate_2op_nobyte(ctxt,_op,_wx,_wy,_lx,_ly,_qx,_qy)		\
268 	do {								\
269 		unsigned long _tmp;					\
270 									\
271 		switch ((ctxt)->dst.bytes) {				\
272 		case 2:							\
273 			____emulate_2op(ctxt,_op,_wx,_wy,"w",u16);	\
274 			break;						\
275 		case 4:							\
276 			____emulate_2op(ctxt,_op,_lx,_ly,"l",u32);	\
277 			break;						\
278 		case 8:							\
279 			ON64(____emulate_2op(ctxt,_op,_qx,_qy,"q",u64)); \
280 			break;						\
281 		}							\
282 	} while (0)
283 
284 #define __emulate_2op(ctxt,_op,_bx,_by,_wx,_wy,_lx,_ly,_qx,_qy)		     \
285 	do {								     \
286 		unsigned long _tmp;					     \
287 		switch ((ctxt)->dst.bytes) {				     \
288 		case 1:							     \
289 			____emulate_2op(ctxt,_op,_bx,_by,"b",u8);	     \
290 			break;						     \
291 		default:						     \
292 			__emulate_2op_nobyte(ctxt, _op,			     \
293 					     _wx, _wy, _lx, _ly, _qx, _qy);  \
294 			break;						     \
295 		}							     \
296 	} while (0)
297 
298 /* Source operand is byte-sized and may be restricted to just %cl. */
299 #define emulate_2op_SrcB(ctxt, _op)					\
300 	__emulate_2op(ctxt, _op, "b", "c", "b", "c", "b", "c", "b", "c")
301 
302 /* Source operand is byte, word, long or quad sized. */
303 #define emulate_2op_SrcV(ctxt, _op)					\
304 	__emulate_2op(ctxt, _op, "b", "q", "w", "r", _LO32, "r", "", "r")
305 
306 /* Source operand is word, long or quad sized. */
307 #define emulate_2op_SrcV_nobyte(ctxt, _op)				\
308 	__emulate_2op_nobyte(ctxt, _op, "w", "r", _LO32, "r", "", "r")
309 
310 /* Instruction has three operands and one operand is stored in ECX register */
311 #define __emulate_2op_cl(ctxt, _op, _suffix, _type)		\
312 	do {								\
313 		unsigned long _tmp;					\
314 		_type _clv  = (ctxt)->src2.val;				\
315 		_type _srcv = (ctxt)->src.val;				\
316 		_type _dstv = (ctxt)->dst.val;				\
317 									\
318 		__asm__ __volatile__ (					\
319 			_PRE_EFLAGS("0", "5", "2")			\
320 			_op _suffix " %4,%1 \n"				\
321 			_POST_EFLAGS("0", "5", "2")			\
322 			: "=m" ((ctxt)->eflags), "+r" (_dstv), "=&r" (_tmp) \
323 			: "c" (_clv) , "r" (_srcv), "i" (EFLAGS_MASK)	\
324 			);						\
325 									\
326 		(ctxt)->src2.val  = (unsigned long) _clv;		\
327 		(ctxt)->src2.val = (unsigned long) _srcv;		\
328 		(ctxt)->dst.val = (unsigned long) _dstv;		\
329 	} while (0)
330 
331 #define emulate_2op_cl(ctxt, _op)					\
332 	do {								\
333 		switch ((ctxt)->dst.bytes) {				\
334 		case 2:							\
335 			__emulate_2op_cl(ctxt, _op, "w", u16);		\
336 			break;						\
337 		case 4:							\
338 			__emulate_2op_cl(ctxt, _op, "l", u32);		\
339 			break;						\
340 		case 8:							\
341 			ON64(__emulate_2op_cl(ctxt, _op, "q", ulong));	\
342 			break;						\
343 		}							\
344 	} while (0)
345 
346 #define __emulate_1op(ctxt, _op, _suffix)				\
347 	do {								\
348 		unsigned long _tmp;					\
349 									\
350 		__asm__ __volatile__ (					\
351 			_PRE_EFLAGS("0", "3", "2")			\
352 			_op _suffix " %1; "				\
353 			_POST_EFLAGS("0", "3", "2")			\
354 			: "=m" ((ctxt)->eflags), "+m" ((ctxt)->dst.val), \
355 			  "=&r" (_tmp)					\
356 			: "i" (EFLAGS_MASK));				\
357 	} while (0)
358 
359 /* Instruction has only one explicit operand (no source operand). */
360 #define emulate_1op(ctxt, _op)						\
361 	do {								\
362 		switch ((ctxt)->dst.bytes) {				\
363 		case 1:	__emulate_1op(ctxt, _op, "b"); break;		\
364 		case 2:	__emulate_1op(ctxt, _op, "w"); break;		\
365 		case 4:	__emulate_1op(ctxt, _op, "l"); break;		\
366 		case 8:	ON64(__emulate_1op(ctxt, _op, "q")); break;	\
367 		}							\
368 	} while (0)
369 
370 #define __emulate_1op_rax_rdx(ctxt, _op, _suffix, _ex)			\
371 	do {								\
372 		unsigned long _tmp;					\
373 		ulong *rax = &(ctxt)->regs[VCPU_REGS_RAX];		\
374 		ulong *rdx = &(ctxt)->regs[VCPU_REGS_RDX];		\
375 									\
376 		__asm__ __volatile__ (					\
377 			_PRE_EFLAGS("0", "5", "1")			\
378 			"1: \n\t"					\
379 			_op _suffix " %6; "				\
380 			"2: \n\t"					\
381 			_POST_EFLAGS("0", "5", "1")			\
382 			".pushsection .fixup,\"ax\" \n\t"		\
383 			"3: movb $1, %4 \n\t"				\
384 			"jmp 2b \n\t"					\
385 			".popsection \n\t"				\
386 			_ASM_EXTABLE(1b, 3b)				\
387 			: "=m" ((ctxt)->eflags), "=&r" (_tmp),		\
388 			  "+a" (*rax), "+d" (*rdx), "+qm"(_ex)		\
389 			: "i" (EFLAGS_MASK), "m" ((ctxt)->src.val),	\
390 			  "a" (*rax), "d" (*rdx));			\
391 	} while (0)
392 
393 /* instruction has only one source operand, destination is implicit (e.g. mul, div, imul, idiv) */
394 #define emulate_1op_rax_rdx(ctxt, _op, _ex)	\
395 	do {								\
396 		switch((ctxt)->src.bytes) {				\
397 		case 1:							\
398 			__emulate_1op_rax_rdx(ctxt, _op, "b", _ex);	\
399 			break;						\
400 		case 2:							\
401 			__emulate_1op_rax_rdx(ctxt, _op, "w", _ex);	\
402 			break;						\
403 		case 4:							\
404 			__emulate_1op_rax_rdx(ctxt, _op, "l", _ex);	\
405 			break;						\
406 		case 8: ON64(						\
407 			__emulate_1op_rax_rdx(ctxt, _op, "q", _ex));	\
408 			break;						\
409 		}							\
410 	} while (0)
411 
412 static int emulator_check_intercept(struct x86_emulate_ctxt *ctxt,
413 				    enum x86_intercept intercept,
414 				    enum x86_intercept_stage stage)
415 {
416 	struct x86_instruction_info info = {
417 		.intercept  = intercept,
418 		.rep_prefix = ctxt->rep_prefix,
419 		.modrm_mod  = ctxt->modrm_mod,
420 		.modrm_reg  = ctxt->modrm_reg,
421 		.modrm_rm   = ctxt->modrm_rm,
422 		.src_val    = ctxt->src.val64,
423 		.src_bytes  = ctxt->src.bytes,
424 		.dst_bytes  = ctxt->dst.bytes,
425 		.ad_bytes   = ctxt->ad_bytes,
426 		.next_rip   = ctxt->eip,
427 	};
428 
429 	return ctxt->ops->intercept(ctxt, &info, stage);
430 }
431 
432 static inline unsigned long ad_mask(struct x86_emulate_ctxt *ctxt)
433 {
434 	return (1UL << (ctxt->ad_bytes << 3)) - 1;
435 }
436 
437 /* Access/update address held in a register, based on addressing mode. */
438 static inline unsigned long
439 address_mask(struct x86_emulate_ctxt *ctxt, unsigned long reg)
440 {
441 	if (ctxt->ad_bytes == sizeof(unsigned long))
442 		return reg;
443 	else
444 		return reg & ad_mask(ctxt);
445 }
446 
447 static inline unsigned long
448 register_address(struct x86_emulate_ctxt *ctxt, unsigned long reg)
449 {
450 	return address_mask(ctxt, reg);
451 }
452 
453 static inline void
454 register_address_increment(struct x86_emulate_ctxt *ctxt, unsigned long *reg, int inc)
455 {
456 	if (ctxt->ad_bytes == sizeof(unsigned long))
457 		*reg += inc;
458 	else
459 		*reg = (*reg & ~ad_mask(ctxt)) | ((*reg + inc) & ad_mask(ctxt));
460 }
461 
462 static inline void jmp_rel(struct x86_emulate_ctxt *ctxt, int rel)
463 {
464 	register_address_increment(ctxt, &ctxt->_eip, rel);
465 }
466 
467 static u32 desc_limit_scaled(struct desc_struct *desc)
468 {
469 	u32 limit = get_desc_limit(desc);
470 
471 	return desc->g ? (limit << 12) | 0xfff : limit;
472 }
473 
474 static void set_seg_override(struct x86_emulate_ctxt *ctxt, int seg)
475 {
476 	ctxt->has_seg_override = true;
477 	ctxt->seg_override = seg;
478 }
479 
480 static unsigned long seg_base(struct x86_emulate_ctxt *ctxt, int seg)
481 {
482 	if (ctxt->mode == X86EMUL_MODE_PROT64 && seg < VCPU_SREG_FS)
483 		return 0;
484 
485 	return ctxt->ops->get_cached_segment_base(ctxt, seg);
486 }
487 
488 static unsigned seg_override(struct x86_emulate_ctxt *ctxt)
489 {
490 	if (!ctxt->has_seg_override)
491 		return 0;
492 
493 	return ctxt->seg_override;
494 }
495 
496 static int emulate_exception(struct x86_emulate_ctxt *ctxt, int vec,
497 			     u32 error, bool valid)
498 {
499 	ctxt->exception.vector = vec;
500 	ctxt->exception.error_code = error;
501 	ctxt->exception.error_code_valid = valid;
502 	return X86EMUL_PROPAGATE_FAULT;
503 }
504 
505 static int emulate_db(struct x86_emulate_ctxt *ctxt)
506 {
507 	return emulate_exception(ctxt, DB_VECTOR, 0, false);
508 }
509 
510 static int emulate_gp(struct x86_emulate_ctxt *ctxt, int err)
511 {
512 	return emulate_exception(ctxt, GP_VECTOR, err, true);
513 }
514 
515 static int emulate_ss(struct x86_emulate_ctxt *ctxt, int err)
516 {
517 	return emulate_exception(ctxt, SS_VECTOR, err, true);
518 }
519 
520 static int emulate_ud(struct x86_emulate_ctxt *ctxt)
521 {
522 	return emulate_exception(ctxt, UD_VECTOR, 0, false);
523 }
524 
525 static int emulate_ts(struct x86_emulate_ctxt *ctxt, int err)
526 {
527 	return emulate_exception(ctxt, TS_VECTOR, err, true);
528 }
529 
530 static int emulate_de(struct x86_emulate_ctxt *ctxt)
531 {
532 	return emulate_exception(ctxt, DE_VECTOR, 0, false);
533 }
534 
535 static int emulate_nm(struct x86_emulate_ctxt *ctxt)
536 {
537 	return emulate_exception(ctxt, NM_VECTOR, 0, false);
538 }
539 
540 static u16 get_segment_selector(struct x86_emulate_ctxt *ctxt, unsigned seg)
541 {
542 	u16 selector;
543 	struct desc_struct desc;
544 
545 	ctxt->ops->get_segment(ctxt, &selector, &desc, NULL, seg);
546 	return selector;
547 }
548 
549 static void set_segment_selector(struct x86_emulate_ctxt *ctxt, u16 selector,
550 				 unsigned seg)
551 {
552 	u16 dummy;
553 	u32 base3;
554 	struct desc_struct desc;
555 
556 	ctxt->ops->get_segment(ctxt, &dummy, &desc, &base3, seg);
557 	ctxt->ops->set_segment(ctxt, selector, &desc, base3, seg);
558 }
559 
560 static int __linearize(struct x86_emulate_ctxt *ctxt,
561 		     struct segmented_address addr,
562 		     unsigned size, bool write, bool fetch,
563 		     ulong *linear)
564 {
565 	struct desc_struct desc;
566 	bool usable;
567 	ulong la;
568 	u32 lim;
569 	u16 sel;
570 	unsigned cpl, rpl;
571 
572 	la = seg_base(ctxt, addr.seg) + addr.ea;
573 	switch (ctxt->mode) {
574 	case X86EMUL_MODE_REAL:
575 		break;
576 	case X86EMUL_MODE_PROT64:
577 		if (((signed long)la << 16) >> 16 != la)
578 			return emulate_gp(ctxt, 0);
579 		break;
580 	default:
581 		usable = ctxt->ops->get_segment(ctxt, &sel, &desc, NULL,
582 						addr.seg);
583 		if (!usable)
584 			goto bad;
585 		/* code segment or read-only data segment */
586 		if (((desc.type & 8) || !(desc.type & 2)) && write)
587 			goto bad;
588 		/* unreadable code segment */
589 		if (!fetch && (desc.type & 8) && !(desc.type & 2))
590 			goto bad;
591 		lim = desc_limit_scaled(&desc);
592 		if ((desc.type & 8) || !(desc.type & 4)) {
593 			/* expand-up segment */
594 			if (addr.ea > lim || (u32)(addr.ea + size - 1) > lim)
595 				goto bad;
596 		} else {
597 			/* exapand-down segment */
598 			if (addr.ea <= lim || (u32)(addr.ea + size - 1) <= lim)
599 				goto bad;
600 			lim = desc.d ? 0xffffffff : 0xffff;
601 			if (addr.ea > lim || (u32)(addr.ea + size - 1) > lim)
602 				goto bad;
603 		}
604 		cpl = ctxt->ops->cpl(ctxt);
605 		rpl = sel & 3;
606 		cpl = max(cpl, rpl);
607 		if (!(desc.type & 8)) {
608 			/* data segment */
609 			if (cpl > desc.dpl)
610 				goto bad;
611 		} else if ((desc.type & 8) && !(desc.type & 4)) {
612 			/* nonconforming code segment */
613 			if (cpl != desc.dpl)
614 				goto bad;
615 		} else if ((desc.type & 8) && (desc.type & 4)) {
616 			/* conforming code segment */
617 			if (cpl < desc.dpl)
618 				goto bad;
619 		}
620 		break;
621 	}
622 	if (fetch ? ctxt->mode != X86EMUL_MODE_PROT64 : ctxt->ad_bytes != 8)
623 		la &= (u32)-1;
624 	*linear = la;
625 	return X86EMUL_CONTINUE;
626 bad:
627 	if (addr.seg == VCPU_SREG_SS)
628 		return emulate_ss(ctxt, addr.seg);
629 	else
630 		return emulate_gp(ctxt, addr.seg);
631 }
632 
633 static int linearize(struct x86_emulate_ctxt *ctxt,
634 		     struct segmented_address addr,
635 		     unsigned size, bool write,
636 		     ulong *linear)
637 {
638 	return __linearize(ctxt, addr, size, write, false, linear);
639 }
640 
641 
642 static int segmented_read_std(struct x86_emulate_ctxt *ctxt,
643 			      struct segmented_address addr,
644 			      void *data,
645 			      unsigned size)
646 {
647 	int rc;
648 	ulong linear;
649 
650 	rc = linearize(ctxt, addr, size, false, &linear);
651 	if (rc != X86EMUL_CONTINUE)
652 		return rc;
653 	return ctxt->ops->read_std(ctxt, linear, data, size, &ctxt->exception);
654 }
655 
656 /*
657  * Fetch the next byte of the instruction being emulated which is pointed to
658  * by ctxt->_eip, then increment ctxt->_eip.
659  *
660  * Also prefetch the remaining bytes of the instruction without crossing page
661  * boundary if they are not in fetch_cache yet.
662  */
663 static int do_insn_fetch_byte(struct x86_emulate_ctxt *ctxt, u8 *dest)
664 {
665 	struct fetch_cache *fc = &ctxt->fetch;
666 	int rc;
667 	int size, cur_size;
668 
669 	if (ctxt->_eip == fc->end) {
670 		unsigned long linear;
671 		struct segmented_address addr = { .seg = VCPU_SREG_CS,
672 						  .ea  = ctxt->_eip };
673 		cur_size = fc->end - fc->start;
674 		size = min(15UL - cur_size,
675 			   PAGE_SIZE - offset_in_page(ctxt->_eip));
676 		rc = __linearize(ctxt, addr, size, false, true, &linear);
677 		if (unlikely(rc != X86EMUL_CONTINUE))
678 			return rc;
679 		rc = ctxt->ops->fetch(ctxt, linear, fc->data + cur_size,
680 				      size, &ctxt->exception);
681 		if (unlikely(rc != X86EMUL_CONTINUE))
682 			return rc;
683 		fc->end += size;
684 	}
685 	*dest = fc->data[ctxt->_eip - fc->start];
686 	ctxt->_eip++;
687 	return X86EMUL_CONTINUE;
688 }
689 
690 static int do_insn_fetch(struct x86_emulate_ctxt *ctxt,
691 			 void *dest, unsigned size)
692 {
693 	int rc;
694 
695 	/* x86 instructions are limited to 15 bytes. */
696 	if (unlikely(ctxt->_eip + size - ctxt->eip > 15))
697 		return X86EMUL_UNHANDLEABLE;
698 	while (size--) {
699 		rc = do_insn_fetch_byte(ctxt, dest++);
700 		if (rc != X86EMUL_CONTINUE)
701 			return rc;
702 	}
703 	return X86EMUL_CONTINUE;
704 }
705 
706 /* Fetch next part of the instruction being emulated. */
707 #define insn_fetch(_type, _ctxt)					\
708 ({	unsigned long _x;						\
709 	rc = do_insn_fetch(_ctxt, &_x, sizeof(_type));			\
710 	if (rc != X86EMUL_CONTINUE)					\
711 		goto done;						\
712 	(_type)_x;							\
713 })
714 
715 #define insn_fetch_arr(_arr, _size, _ctxt)				\
716 ({	rc = do_insn_fetch(_ctxt, _arr, (_size));			\
717 	if (rc != X86EMUL_CONTINUE)					\
718 		goto done;						\
719 })
720 
721 /*
722  * Given the 'reg' portion of a ModRM byte, and a register block, return a
723  * pointer into the block that addresses the relevant register.
724  * @highbyte_regs specifies whether to decode AH,CH,DH,BH.
725  */
726 static void *decode_register(u8 modrm_reg, unsigned long *regs,
727 			     int highbyte_regs)
728 {
729 	void *p;
730 
731 	p = &regs[modrm_reg];
732 	if (highbyte_regs && modrm_reg >= 4 && modrm_reg < 8)
733 		p = (unsigned char *)&regs[modrm_reg & 3] + 1;
734 	return p;
735 }
736 
737 static int read_descriptor(struct x86_emulate_ctxt *ctxt,
738 			   struct segmented_address addr,
739 			   u16 *size, unsigned long *address, int op_bytes)
740 {
741 	int rc;
742 
743 	if (op_bytes == 2)
744 		op_bytes = 3;
745 	*address = 0;
746 	rc = segmented_read_std(ctxt, addr, size, 2);
747 	if (rc != X86EMUL_CONTINUE)
748 		return rc;
749 	addr.ea += 2;
750 	rc = segmented_read_std(ctxt, addr, address, op_bytes);
751 	return rc;
752 }
753 
754 static int test_cc(unsigned int condition, unsigned int flags)
755 {
756 	int rc = 0;
757 
758 	switch ((condition & 15) >> 1) {
759 	case 0: /* o */
760 		rc |= (flags & EFLG_OF);
761 		break;
762 	case 1: /* b/c/nae */
763 		rc |= (flags & EFLG_CF);
764 		break;
765 	case 2: /* z/e */
766 		rc |= (flags & EFLG_ZF);
767 		break;
768 	case 3: /* be/na */
769 		rc |= (flags & (EFLG_CF|EFLG_ZF));
770 		break;
771 	case 4: /* s */
772 		rc |= (flags & EFLG_SF);
773 		break;
774 	case 5: /* p/pe */
775 		rc |= (flags & EFLG_PF);
776 		break;
777 	case 7: /* le/ng */
778 		rc |= (flags & EFLG_ZF);
779 		/* fall through */
780 	case 6: /* l/nge */
781 		rc |= (!(flags & EFLG_SF) != !(flags & EFLG_OF));
782 		break;
783 	}
784 
785 	/* Odd condition identifiers (lsb == 1) have inverted sense. */
786 	return (!!rc ^ (condition & 1));
787 }
788 
789 static void fetch_register_operand(struct operand *op)
790 {
791 	switch (op->bytes) {
792 	case 1:
793 		op->val = *(u8 *)op->addr.reg;
794 		break;
795 	case 2:
796 		op->val = *(u16 *)op->addr.reg;
797 		break;
798 	case 4:
799 		op->val = *(u32 *)op->addr.reg;
800 		break;
801 	case 8:
802 		op->val = *(u64 *)op->addr.reg;
803 		break;
804 	}
805 }
806 
807 static void read_sse_reg(struct x86_emulate_ctxt *ctxt, sse128_t *data, int reg)
808 {
809 	ctxt->ops->get_fpu(ctxt);
810 	switch (reg) {
811 	case 0: asm("movdqu %%xmm0, %0" : "=m"(*data)); break;
812 	case 1: asm("movdqu %%xmm1, %0" : "=m"(*data)); break;
813 	case 2: asm("movdqu %%xmm2, %0" : "=m"(*data)); break;
814 	case 3: asm("movdqu %%xmm3, %0" : "=m"(*data)); break;
815 	case 4: asm("movdqu %%xmm4, %0" : "=m"(*data)); break;
816 	case 5: asm("movdqu %%xmm5, %0" : "=m"(*data)); break;
817 	case 6: asm("movdqu %%xmm6, %0" : "=m"(*data)); break;
818 	case 7: asm("movdqu %%xmm7, %0" : "=m"(*data)); break;
819 #ifdef CONFIG_X86_64
820 	case 8: asm("movdqu %%xmm8, %0" : "=m"(*data)); break;
821 	case 9: asm("movdqu %%xmm9, %0" : "=m"(*data)); break;
822 	case 10: asm("movdqu %%xmm10, %0" : "=m"(*data)); break;
823 	case 11: asm("movdqu %%xmm11, %0" : "=m"(*data)); break;
824 	case 12: asm("movdqu %%xmm12, %0" : "=m"(*data)); break;
825 	case 13: asm("movdqu %%xmm13, %0" : "=m"(*data)); break;
826 	case 14: asm("movdqu %%xmm14, %0" : "=m"(*data)); break;
827 	case 15: asm("movdqu %%xmm15, %0" : "=m"(*data)); break;
828 #endif
829 	default: BUG();
830 	}
831 	ctxt->ops->put_fpu(ctxt);
832 }
833 
834 static void write_sse_reg(struct x86_emulate_ctxt *ctxt, sse128_t *data,
835 			  int reg)
836 {
837 	ctxt->ops->get_fpu(ctxt);
838 	switch (reg) {
839 	case 0: asm("movdqu %0, %%xmm0" : : "m"(*data)); break;
840 	case 1: asm("movdqu %0, %%xmm1" : : "m"(*data)); break;
841 	case 2: asm("movdqu %0, %%xmm2" : : "m"(*data)); break;
842 	case 3: asm("movdqu %0, %%xmm3" : : "m"(*data)); break;
843 	case 4: asm("movdqu %0, %%xmm4" : : "m"(*data)); break;
844 	case 5: asm("movdqu %0, %%xmm5" : : "m"(*data)); break;
845 	case 6: asm("movdqu %0, %%xmm6" : : "m"(*data)); break;
846 	case 7: asm("movdqu %0, %%xmm7" : : "m"(*data)); break;
847 #ifdef CONFIG_X86_64
848 	case 8: asm("movdqu %0, %%xmm8" : : "m"(*data)); break;
849 	case 9: asm("movdqu %0, %%xmm9" : : "m"(*data)); break;
850 	case 10: asm("movdqu %0, %%xmm10" : : "m"(*data)); break;
851 	case 11: asm("movdqu %0, %%xmm11" : : "m"(*data)); break;
852 	case 12: asm("movdqu %0, %%xmm12" : : "m"(*data)); break;
853 	case 13: asm("movdqu %0, %%xmm13" : : "m"(*data)); break;
854 	case 14: asm("movdqu %0, %%xmm14" : : "m"(*data)); break;
855 	case 15: asm("movdqu %0, %%xmm15" : : "m"(*data)); break;
856 #endif
857 	default: BUG();
858 	}
859 	ctxt->ops->put_fpu(ctxt);
860 }
861 
862 static void decode_register_operand(struct x86_emulate_ctxt *ctxt,
863 				    struct operand *op)
864 {
865 	unsigned reg = ctxt->modrm_reg;
866 	int highbyte_regs = ctxt->rex_prefix == 0;
867 
868 	if (!(ctxt->d & ModRM))
869 		reg = (ctxt->b & 7) | ((ctxt->rex_prefix & 1) << 3);
870 
871 	if (ctxt->d & Sse) {
872 		op->type = OP_XMM;
873 		op->bytes = 16;
874 		op->addr.xmm = reg;
875 		read_sse_reg(ctxt, &op->vec_val, reg);
876 		return;
877 	}
878 
879 	op->type = OP_REG;
880 	if (ctxt->d & ByteOp) {
881 		op->addr.reg = decode_register(reg, ctxt->regs, highbyte_regs);
882 		op->bytes = 1;
883 	} else {
884 		op->addr.reg = decode_register(reg, ctxt->regs, 0);
885 		op->bytes = ctxt->op_bytes;
886 	}
887 	fetch_register_operand(op);
888 	op->orig_val = op->val;
889 }
890 
891 static int decode_modrm(struct x86_emulate_ctxt *ctxt,
892 			struct operand *op)
893 {
894 	u8 sib;
895 	int index_reg = 0, base_reg = 0, scale;
896 	int rc = X86EMUL_CONTINUE;
897 	ulong modrm_ea = 0;
898 
899 	if (ctxt->rex_prefix) {
900 		ctxt->modrm_reg = (ctxt->rex_prefix & 4) << 1;	/* REX.R */
901 		index_reg = (ctxt->rex_prefix & 2) << 2; /* REX.X */
902 		ctxt->modrm_rm = base_reg = (ctxt->rex_prefix & 1) << 3; /* REG.B */
903 	}
904 
905 	ctxt->modrm = insn_fetch(u8, ctxt);
906 	ctxt->modrm_mod |= (ctxt->modrm & 0xc0) >> 6;
907 	ctxt->modrm_reg |= (ctxt->modrm & 0x38) >> 3;
908 	ctxt->modrm_rm |= (ctxt->modrm & 0x07);
909 	ctxt->modrm_seg = VCPU_SREG_DS;
910 
911 	if (ctxt->modrm_mod == 3) {
912 		op->type = OP_REG;
913 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
914 		op->addr.reg = decode_register(ctxt->modrm_rm,
915 					       ctxt->regs, ctxt->d & ByteOp);
916 		if (ctxt->d & Sse) {
917 			op->type = OP_XMM;
918 			op->bytes = 16;
919 			op->addr.xmm = ctxt->modrm_rm;
920 			read_sse_reg(ctxt, &op->vec_val, ctxt->modrm_rm);
921 			return rc;
922 		}
923 		fetch_register_operand(op);
924 		return rc;
925 	}
926 
927 	op->type = OP_MEM;
928 
929 	if (ctxt->ad_bytes == 2) {
930 		unsigned bx = ctxt->regs[VCPU_REGS_RBX];
931 		unsigned bp = ctxt->regs[VCPU_REGS_RBP];
932 		unsigned si = ctxt->regs[VCPU_REGS_RSI];
933 		unsigned di = ctxt->regs[VCPU_REGS_RDI];
934 
935 		/* 16-bit ModR/M decode. */
936 		switch (ctxt->modrm_mod) {
937 		case 0:
938 			if (ctxt->modrm_rm == 6)
939 				modrm_ea += insn_fetch(u16, ctxt);
940 			break;
941 		case 1:
942 			modrm_ea += insn_fetch(s8, ctxt);
943 			break;
944 		case 2:
945 			modrm_ea += insn_fetch(u16, ctxt);
946 			break;
947 		}
948 		switch (ctxt->modrm_rm) {
949 		case 0:
950 			modrm_ea += bx + si;
951 			break;
952 		case 1:
953 			modrm_ea += bx + di;
954 			break;
955 		case 2:
956 			modrm_ea += bp + si;
957 			break;
958 		case 3:
959 			modrm_ea += bp + di;
960 			break;
961 		case 4:
962 			modrm_ea += si;
963 			break;
964 		case 5:
965 			modrm_ea += di;
966 			break;
967 		case 6:
968 			if (ctxt->modrm_mod != 0)
969 				modrm_ea += bp;
970 			break;
971 		case 7:
972 			modrm_ea += bx;
973 			break;
974 		}
975 		if (ctxt->modrm_rm == 2 || ctxt->modrm_rm == 3 ||
976 		    (ctxt->modrm_rm == 6 && ctxt->modrm_mod != 0))
977 			ctxt->modrm_seg = VCPU_SREG_SS;
978 		modrm_ea = (u16)modrm_ea;
979 	} else {
980 		/* 32/64-bit ModR/M decode. */
981 		if ((ctxt->modrm_rm & 7) == 4) {
982 			sib = insn_fetch(u8, ctxt);
983 			index_reg |= (sib >> 3) & 7;
984 			base_reg |= sib & 7;
985 			scale = sib >> 6;
986 
987 			if ((base_reg & 7) == 5 && ctxt->modrm_mod == 0)
988 				modrm_ea += insn_fetch(s32, ctxt);
989 			else
990 				modrm_ea += ctxt->regs[base_reg];
991 			if (index_reg != 4)
992 				modrm_ea += ctxt->regs[index_reg] << scale;
993 		} else if ((ctxt->modrm_rm & 7) == 5 && ctxt->modrm_mod == 0) {
994 			if (ctxt->mode == X86EMUL_MODE_PROT64)
995 				ctxt->rip_relative = 1;
996 		} else
997 			modrm_ea += ctxt->regs[ctxt->modrm_rm];
998 		switch (ctxt->modrm_mod) {
999 		case 0:
1000 			if (ctxt->modrm_rm == 5)
1001 				modrm_ea += insn_fetch(s32, ctxt);
1002 			break;
1003 		case 1:
1004 			modrm_ea += insn_fetch(s8, ctxt);
1005 			break;
1006 		case 2:
1007 			modrm_ea += insn_fetch(s32, ctxt);
1008 			break;
1009 		}
1010 	}
1011 	op->addr.mem.ea = modrm_ea;
1012 done:
1013 	return rc;
1014 }
1015 
1016 static int decode_abs(struct x86_emulate_ctxt *ctxt,
1017 		      struct operand *op)
1018 {
1019 	int rc = X86EMUL_CONTINUE;
1020 
1021 	op->type = OP_MEM;
1022 	switch (ctxt->ad_bytes) {
1023 	case 2:
1024 		op->addr.mem.ea = insn_fetch(u16, ctxt);
1025 		break;
1026 	case 4:
1027 		op->addr.mem.ea = insn_fetch(u32, ctxt);
1028 		break;
1029 	case 8:
1030 		op->addr.mem.ea = insn_fetch(u64, ctxt);
1031 		break;
1032 	}
1033 done:
1034 	return rc;
1035 }
1036 
1037 static void fetch_bit_operand(struct x86_emulate_ctxt *ctxt)
1038 {
1039 	long sv = 0, mask;
1040 
1041 	if (ctxt->dst.type == OP_MEM && ctxt->src.type == OP_REG) {
1042 		mask = ~(ctxt->dst.bytes * 8 - 1);
1043 
1044 		if (ctxt->src.bytes == 2)
1045 			sv = (s16)ctxt->src.val & (s16)mask;
1046 		else if (ctxt->src.bytes == 4)
1047 			sv = (s32)ctxt->src.val & (s32)mask;
1048 
1049 		ctxt->dst.addr.mem.ea += (sv >> 3);
1050 	}
1051 
1052 	/* only subword offset */
1053 	ctxt->src.val &= (ctxt->dst.bytes << 3) - 1;
1054 }
1055 
1056 static int read_emulated(struct x86_emulate_ctxt *ctxt,
1057 			 unsigned long addr, void *dest, unsigned size)
1058 {
1059 	int rc;
1060 	struct read_cache *mc = &ctxt->mem_read;
1061 
1062 	while (size) {
1063 		int n = min(size, 8u);
1064 		size -= n;
1065 		if (mc->pos < mc->end)
1066 			goto read_cached;
1067 
1068 		rc = ctxt->ops->read_emulated(ctxt, addr, mc->data + mc->end, n,
1069 					      &ctxt->exception);
1070 		if (rc != X86EMUL_CONTINUE)
1071 			return rc;
1072 		mc->end += n;
1073 
1074 	read_cached:
1075 		memcpy(dest, mc->data + mc->pos, n);
1076 		mc->pos += n;
1077 		dest += n;
1078 		addr += n;
1079 	}
1080 	return X86EMUL_CONTINUE;
1081 }
1082 
1083 static int segmented_read(struct x86_emulate_ctxt *ctxt,
1084 			  struct segmented_address addr,
1085 			  void *data,
1086 			  unsigned size)
1087 {
1088 	int rc;
1089 	ulong linear;
1090 
1091 	rc = linearize(ctxt, addr, size, false, &linear);
1092 	if (rc != X86EMUL_CONTINUE)
1093 		return rc;
1094 	return read_emulated(ctxt, linear, data, size);
1095 }
1096 
1097 static int segmented_write(struct x86_emulate_ctxt *ctxt,
1098 			   struct segmented_address addr,
1099 			   const void *data,
1100 			   unsigned size)
1101 {
1102 	int rc;
1103 	ulong linear;
1104 
1105 	rc = linearize(ctxt, addr, size, true, &linear);
1106 	if (rc != X86EMUL_CONTINUE)
1107 		return rc;
1108 	return ctxt->ops->write_emulated(ctxt, linear, data, size,
1109 					 &ctxt->exception);
1110 }
1111 
1112 static int segmented_cmpxchg(struct x86_emulate_ctxt *ctxt,
1113 			     struct segmented_address addr,
1114 			     const void *orig_data, const void *data,
1115 			     unsigned size)
1116 {
1117 	int rc;
1118 	ulong linear;
1119 
1120 	rc = linearize(ctxt, addr, size, true, &linear);
1121 	if (rc != X86EMUL_CONTINUE)
1122 		return rc;
1123 	return ctxt->ops->cmpxchg_emulated(ctxt, linear, orig_data, data,
1124 					   size, &ctxt->exception);
1125 }
1126 
1127 static int pio_in_emulated(struct x86_emulate_ctxt *ctxt,
1128 			   unsigned int size, unsigned short port,
1129 			   void *dest)
1130 {
1131 	struct read_cache *rc = &ctxt->io_read;
1132 
1133 	if (rc->pos == rc->end) { /* refill pio read ahead */
1134 		unsigned int in_page, n;
1135 		unsigned int count = ctxt->rep_prefix ?
1136 			address_mask(ctxt, ctxt->regs[VCPU_REGS_RCX]) : 1;
1137 		in_page = (ctxt->eflags & EFLG_DF) ?
1138 			offset_in_page(ctxt->regs[VCPU_REGS_RDI]) :
1139 			PAGE_SIZE - offset_in_page(ctxt->regs[VCPU_REGS_RDI]);
1140 		n = min(min(in_page, (unsigned int)sizeof(rc->data)) / size,
1141 			count);
1142 		if (n == 0)
1143 			n = 1;
1144 		rc->pos = rc->end = 0;
1145 		if (!ctxt->ops->pio_in_emulated(ctxt, size, port, rc->data, n))
1146 			return 0;
1147 		rc->end = n * size;
1148 	}
1149 
1150 	memcpy(dest, rc->data + rc->pos, size);
1151 	rc->pos += size;
1152 	return 1;
1153 }
1154 
1155 static int read_interrupt_descriptor(struct x86_emulate_ctxt *ctxt,
1156 				     u16 index, struct desc_struct *desc)
1157 {
1158 	struct desc_ptr dt;
1159 	ulong addr;
1160 
1161 	ctxt->ops->get_idt(ctxt, &dt);
1162 
1163 	if (dt.size < index * 8 + 7)
1164 		return emulate_gp(ctxt, index << 3 | 0x2);
1165 
1166 	addr = dt.address + index * 8;
1167 	return ctxt->ops->read_std(ctxt, addr, desc, sizeof *desc,
1168 				   &ctxt->exception);
1169 }
1170 
1171 static void get_descriptor_table_ptr(struct x86_emulate_ctxt *ctxt,
1172 				     u16 selector, struct desc_ptr *dt)
1173 {
1174 	struct x86_emulate_ops *ops = ctxt->ops;
1175 
1176 	if (selector & 1 << 2) {
1177 		struct desc_struct desc;
1178 		u16 sel;
1179 
1180 		memset (dt, 0, sizeof *dt);
1181 		if (!ops->get_segment(ctxt, &sel, &desc, NULL, VCPU_SREG_LDTR))
1182 			return;
1183 
1184 		dt->size = desc_limit_scaled(&desc); /* what if limit > 65535? */
1185 		dt->address = get_desc_base(&desc);
1186 	} else
1187 		ops->get_gdt(ctxt, dt);
1188 }
1189 
1190 /* allowed just for 8 bytes segments */
1191 static int read_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1192 				   u16 selector, struct desc_struct *desc)
1193 {
1194 	struct desc_ptr dt;
1195 	u16 index = selector >> 3;
1196 	ulong addr;
1197 
1198 	get_descriptor_table_ptr(ctxt, selector, &dt);
1199 
1200 	if (dt.size < index * 8 + 7)
1201 		return emulate_gp(ctxt, selector & 0xfffc);
1202 
1203 	addr = dt.address + index * 8;
1204 	return ctxt->ops->read_std(ctxt, addr, desc, sizeof *desc,
1205 				   &ctxt->exception);
1206 }
1207 
1208 /* allowed just for 8 bytes segments */
1209 static int write_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1210 				    u16 selector, struct desc_struct *desc)
1211 {
1212 	struct desc_ptr dt;
1213 	u16 index = selector >> 3;
1214 	ulong addr;
1215 
1216 	get_descriptor_table_ptr(ctxt, selector, &dt);
1217 
1218 	if (dt.size < index * 8 + 7)
1219 		return emulate_gp(ctxt, selector & 0xfffc);
1220 
1221 	addr = dt.address + index * 8;
1222 	return ctxt->ops->write_std(ctxt, addr, desc, sizeof *desc,
1223 				    &ctxt->exception);
1224 }
1225 
1226 /* Does not support long mode */
1227 static int load_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1228 				   u16 selector, int seg)
1229 {
1230 	struct desc_struct seg_desc;
1231 	u8 dpl, rpl, cpl;
1232 	unsigned err_vec = GP_VECTOR;
1233 	u32 err_code = 0;
1234 	bool null_selector = !(selector & ~0x3); /* 0000-0003 are null */
1235 	int ret;
1236 
1237 	memset(&seg_desc, 0, sizeof seg_desc);
1238 
1239 	if ((seg <= VCPU_SREG_GS && ctxt->mode == X86EMUL_MODE_VM86)
1240 	    || ctxt->mode == X86EMUL_MODE_REAL) {
1241 		/* set real mode segment descriptor */
1242 		set_desc_base(&seg_desc, selector << 4);
1243 		set_desc_limit(&seg_desc, 0xffff);
1244 		seg_desc.type = 3;
1245 		seg_desc.p = 1;
1246 		seg_desc.s = 1;
1247 		if (ctxt->mode == X86EMUL_MODE_VM86)
1248 			seg_desc.dpl = 3;
1249 		goto load;
1250 	}
1251 
1252 	/* NULL selector is not valid for TR, CS and SS */
1253 	if ((seg == VCPU_SREG_CS || seg == VCPU_SREG_SS || seg == VCPU_SREG_TR)
1254 	    && null_selector)
1255 		goto exception;
1256 
1257 	/* TR should be in GDT only */
1258 	if (seg == VCPU_SREG_TR && (selector & (1 << 2)))
1259 		goto exception;
1260 
1261 	if (null_selector) /* for NULL selector skip all following checks */
1262 		goto load;
1263 
1264 	ret = read_segment_descriptor(ctxt, selector, &seg_desc);
1265 	if (ret != X86EMUL_CONTINUE)
1266 		return ret;
1267 
1268 	err_code = selector & 0xfffc;
1269 	err_vec = GP_VECTOR;
1270 
1271 	/* can't load system descriptor into segment selecor */
1272 	if (seg <= VCPU_SREG_GS && !seg_desc.s)
1273 		goto exception;
1274 
1275 	if (!seg_desc.p) {
1276 		err_vec = (seg == VCPU_SREG_SS) ? SS_VECTOR : NP_VECTOR;
1277 		goto exception;
1278 	}
1279 
1280 	rpl = selector & 3;
1281 	dpl = seg_desc.dpl;
1282 	cpl = ctxt->ops->cpl(ctxt);
1283 
1284 	switch (seg) {
1285 	case VCPU_SREG_SS:
1286 		/*
1287 		 * segment is not a writable data segment or segment
1288 		 * selector's RPL != CPL or segment selector's RPL != CPL
1289 		 */
1290 		if (rpl != cpl || (seg_desc.type & 0xa) != 0x2 || dpl != cpl)
1291 			goto exception;
1292 		break;
1293 	case VCPU_SREG_CS:
1294 		if (!(seg_desc.type & 8))
1295 			goto exception;
1296 
1297 		if (seg_desc.type & 4) {
1298 			/* conforming */
1299 			if (dpl > cpl)
1300 				goto exception;
1301 		} else {
1302 			/* nonconforming */
1303 			if (rpl > cpl || dpl != cpl)
1304 				goto exception;
1305 		}
1306 		/* CS(RPL) <- CPL */
1307 		selector = (selector & 0xfffc) | cpl;
1308 		break;
1309 	case VCPU_SREG_TR:
1310 		if (seg_desc.s || (seg_desc.type != 1 && seg_desc.type != 9))
1311 			goto exception;
1312 		break;
1313 	case VCPU_SREG_LDTR:
1314 		if (seg_desc.s || seg_desc.type != 2)
1315 			goto exception;
1316 		break;
1317 	default: /*  DS, ES, FS, or GS */
1318 		/*
1319 		 * segment is not a data or readable code segment or
1320 		 * ((segment is a data or nonconforming code segment)
1321 		 * and (both RPL and CPL > DPL))
1322 		 */
1323 		if ((seg_desc.type & 0xa) == 0x8 ||
1324 		    (((seg_desc.type & 0xc) != 0xc) &&
1325 		     (rpl > dpl && cpl > dpl)))
1326 			goto exception;
1327 		break;
1328 	}
1329 
1330 	if (seg_desc.s) {
1331 		/* mark segment as accessed */
1332 		seg_desc.type |= 1;
1333 		ret = write_segment_descriptor(ctxt, selector, &seg_desc);
1334 		if (ret != X86EMUL_CONTINUE)
1335 			return ret;
1336 	}
1337 load:
1338 	ctxt->ops->set_segment(ctxt, selector, &seg_desc, 0, seg);
1339 	return X86EMUL_CONTINUE;
1340 exception:
1341 	emulate_exception(ctxt, err_vec, err_code, true);
1342 	return X86EMUL_PROPAGATE_FAULT;
1343 }
1344 
1345 static void write_register_operand(struct operand *op)
1346 {
1347 	/* The 4-byte case *is* correct: in 64-bit mode we zero-extend. */
1348 	switch (op->bytes) {
1349 	case 1:
1350 		*(u8 *)op->addr.reg = (u8)op->val;
1351 		break;
1352 	case 2:
1353 		*(u16 *)op->addr.reg = (u16)op->val;
1354 		break;
1355 	case 4:
1356 		*op->addr.reg = (u32)op->val;
1357 		break;	/* 64b: zero-extend */
1358 	case 8:
1359 		*op->addr.reg = op->val;
1360 		break;
1361 	}
1362 }
1363 
1364 static int writeback(struct x86_emulate_ctxt *ctxt)
1365 {
1366 	int rc;
1367 
1368 	switch (ctxt->dst.type) {
1369 	case OP_REG:
1370 		write_register_operand(&ctxt->dst);
1371 		break;
1372 	case OP_MEM:
1373 		if (ctxt->lock_prefix)
1374 			rc = segmented_cmpxchg(ctxt,
1375 					       ctxt->dst.addr.mem,
1376 					       &ctxt->dst.orig_val,
1377 					       &ctxt->dst.val,
1378 					       ctxt->dst.bytes);
1379 		else
1380 			rc = segmented_write(ctxt,
1381 					     ctxt->dst.addr.mem,
1382 					     &ctxt->dst.val,
1383 					     ctxt->dst.bytes);
1384 		if (rc != X86EMUL_CONTINUE)
1385 			return rc;
1386 		break;
1387 	case OP_XMM:
1388 		write_sse_reg(ctxt, &ctxt->dst.vec_val, ctxt->dst.addr.xmm);
1389 		break;
1390 	case OP_NONE:
1391 		/* no writeback */
1392 		break;
1393 	default:
1394 		break;
1395 	}
1396 	return X86EMUL_CONTINUE;
1397 }
1398 
1399 static int em_push(struct x86_emulate_ctxt *ctxt)
1400 {
1401 	struct segmented_address addr;
1402 
1403 	register_address_increment(ctxt, &ctxt->regs[VCPU_REGS_RSP], -ctxt->op_bytes);
1404 	addr.ea = register_address(ctxt, ctxt->regs[VCPU_REGS_RSP]);
1405 	addr.seg = VCPU_SREG_SS;
1406 
1407 	/* Disable writeback. */
1408 	ctxt->dst.type = OP_NONE;
1409 	return segmented_write(ctxt, addr, &ctxt->src.val, ctxt->op_bytes);
1410 }
1411 
1412 static int emulate_pop(struct x86_emulate_ctxt *ctxt,
1413 		       void *dest, int len)
1414 {
1415 	int rc;
1416 	struct segmented_address addr;
1417 
1418 	addr.ea = register_address(ctxt, ctxt->regs[VCPU_REGS_RSP]);
1419 	addr.seg = VCPU_SREG_SS;
1420 	rc = segmented_read(ctxt, addr, dest, len);
1421 	if (rc != X86EMUL_CONTINUE)
1422 		return rc;
1423 
1424 	register_address_increment(ctxt, &ctxt->regs[VCPU_REGS_RSP], len);
1425 	return rc;
1426 }
1427 
1428 static int em_pop(struct x86_emulate_ctxt *ctxt)
1429 {
1430 	return emulate_pop(ctxt, &ctxt->dst.val, ctxt->op_bytes);
1431 }
1432 
1433 static int emulate_popf(struct x86_emulate_ctxt *ctxt,
1434 			void *dest, int len)
1435 {
1436 	int rc;
1437 	unsigned long val, change_mask;
1438 	int iopl = (ctxt->eflags & X86_EFLAGS_IOPL) >> IOPL_SHIFT;
1439 	int cpl = ctxt->ops->cpl(ctxt);
1440 
1441 	rc = emulate_pop(ctxt, &val, len);
1442 	if (rc != X86EMUL_CONTINUE)
1443 		return rc;
1444 
1445 	change_mask = EFLG_CF | EFLG_PF | EFLG_AF | EFLG_ZF | EFLG_SF | EFLG_OF
1446 		| EFLG_TF | EFLG_DF | EFLG_NT | EFLG_RF | EFLG_AC | EFLG_ID;
1447 
1448 	switch(ctxt->mode) {
1449 	case X86EMUL_MODE_PROT64:
1450 	case X86EMUL_MODE_PROT32:
1451 	case X86EMUL_MODE_PROT16:
1452 		if (cpl == 0)
1453 			change_mask |= EFLG_IOPL;
1454 		if (cpl <= iopl)
1455 			change_mask |= EFLG_IF;
1456 		break;
1457 	case X86EMUL_MODE_VM86:
1458 		if (iopl < 3)
1459 			return emulate_gp(ctxt, 0);
1460 		change_mask |= EFLG_IF;
1461 		break;
1462 	default: /* real mode */
1463 		change_mask |= (EFLG_IOPL | EFLG_IF);
1464 		break;
1465 	}
1466 
1467 	*(unsigned long *)dest =
1468 		(ctxt->eflags & ~change_mask) | (val & change_mask);
1469 
1470 	return rc;
1471 }
1472 
1473 static int em_popf(struct x86_emulate_ctxt *ctxt)
1474 {
1475 	ctxt->dst.type = OP_REG;
1476 	ctxt->dst.addr.reg = &ctxt->eflags;
1477 	ctxt->dst.bytes = ctxt->op_bytes;
1478 	return emulate_popf(ctxt, &ctxt->dst.val, ctxt->op_bytes);
1479 }
1480 
1481 static int em_push_sreg(struct x86_emulate_ctxt *ctxt)
1482 {
1483 	int seg = ctxt->src2.val;
1484 
1485 	ctxt->src.val = get_segment_selector(ctxt, seg);
1486 
1487 	return em_push(ctxt);
1488 }
1489 
1490 static int em_pop_sreg(struct x86_emulate_ctxt *ctxt)
1491 {
1492 	int seg = ctxt->src2.val;
1493 	unsigned long selector;
1494 	int rc;
1495 
1496 	rc = emulate_pop(ctxt, &selector, ctxt->op_bytes);
1497 	if (rc != X86EMUL_CONTINUE)
1498 		return rc;
1499 
1500 	rc = load_segment_descriptor(ctxt, (u16)selector, seg);
1501 	return rc;
1502 }
1503 
1504 static int em_pusha(struct x86_emulate_ctxt *ctxt)
1505 {
1506 	unsigned long old_esp = ctxt->regs[VCPU_REGS_RSP];
1507 	int rc = X86EMUL_CONTINUE;
1508 	int reg = VCPU_REGS_RAX;
1509 
1510 	while (reg <= VCPU_REGS_RDI) {
1511 		(reg == VCPU_REGS_RSP) ?
1512 		(ctxt->src.val = old_esp) : (ctxt->src.val = ctxt->regs[reg]);
1513 
1514 		rc = em_push(ctxt);
1515 		if (rc != X86EMUL_CONTINUE)
1516 			return rc;
1517 
1518 		++reg;
1519 	}
1520 
1521 	return rc;
1522 }
1523 
1524 static int em_pushf(struct x86_emulate_ctxt *ctxt)
1525 {
1526 	ctxt->src.val =  (unsigned long)ctxt->eflags;
1527 	return em_push(ctxt);
1528 }
1529 
1530 static int em_popa(struct x86_emulate_ctxt *ctxt)
1531 {
1532 	int rc = X86EMUL_CONTINUE;
1533 	int reg = VCPU_REGS_RDI;
1534 
1535 	while (reg >= VCPU_REGS_RAX) {
1536 		if (reg == VCPU_REGS_RSP) {
1537 			register_address_increment(ctxt, &ctxt->regs[VCPU_REGS_RSP],
1538 							ctxt->op_bytes);
1539 			--reg;
1540 		}
1541 
1542 		rc = emulate_pop(ctxt, &ctxt->regs[reg], ctxt->op_bytes);
1543 		if (rc != X86EMUL_CONTINUE)
1544 			break;
1545 		--reg;
1546 	}
1547 	return rc;
1548 }
1549 
1550 int emulate_int_real(struct x86_emulate_ctxt *ctxt, int irq)
1551 {
1552 	struct x86_emulate_ops *ops = ctxt->ops;
1553 	int rc;
1554 	struct desc_ptr dt;
1555 	gva_t cs_addr;
1556 	gva_t eip_addr;
1557 	u16 cs, eip;
1558 
1559 	/* TODO: Add limit checks */
1560 	ctxt->src.val = ctxt->eflags;
1561 	rc = em_push(ctxt);
1562 	if (rc != X86EMUL_CONTINUE)
1563 		return rc;
1564 
1565 	ctxt->eflags &= ~(EFLG_IF | EFLG_TF | EFLG_AC);
1566 
1567 	ctxt->src.val = get_segment_selector(ctxt, VCPU_SREG_CS);
1568 	rc = em_push(ctxt);
1569 	if (rc != X86EMUL_CONTINUE)
1570 		return rc;
1571 
1572 	ctxt->src.val = ctxt->_eip;
1573 	rc = em_push(ctxt);
1574 	if (rc != X86EMUL_CONTINUE)
1575 		return rc;
1576 
1577 	ops->get_idt(ctxt, &dt);
1578 
1579 	eip_addr = dt.address + (irq << 2);
1580 	cs_addr = dt.address + (irq << 2) + 2;
1581 
1582 	rc = ops->read_std(ctxt, cs_addr, &cs, 2, &ctxt->exception);
1583 	if (rc != X86EMUL_CONTINUE)
1584 		return rc;
1585 
1586 	rc = ops->read_std(ctxt, eip_addr, &eip, 2, &ctxt->exception);
1587 	if (rc != X86EMUL_CONTINUE)
1588 		return rc;
1589 
1590 	rc = load_segment_descriptor(ctxt, cs, VCPU_SREG_CS);
1591 	if (rc != X86EMUL_CONTINUE)
1592 		return rc;
1593 
1594 	ctxt->_eip = eip;
1595 
1596 	return rc;
1597 }
1598 
1599 static int emulate_int(struct x86_emulate_ctxt *ctxt, int irq)
1600 {
1601 	switch(ctxt->mode) {
1602 	case X86EMUL_MODE_REAL:
1603 		return emulate_int_real(ctxt, irq);
1604 	case X86EMUL_MODE_VM86:
1605 	case X86EMUL_MODE_PROT16:
1606 	case X86EMUL_MODE_PROT32:
1607 	case X86EMUL_MODE_PROT64:
1608 	default:
1609 		/* Protected mode interrupts unimplemented yet */
1610 		return X86EMUL_UNHANDLEABLE;
1611 	}
1612 }
1613 
1614 static int emulate_iret_real(struct x86_emulate_ctxt *ctxt)
1615 {
1616 	int rc = X86EMUL_CONTINUE;
1617 	unsigned long temp_eip = 0;
1618 	unsigned long temp_eflags = 0;
1619 	unsigned long cs = 0;
1620 	unsigned long mask = EFLG_CF | EFLG_PF | EFLG_AF | EFLG_ZF | EFLG_SF | EFLG_TF |
1621 			     EFLG_IF | EFLG_DF | EFLG_OF | EFLG_IOPL | EFLG_NT | EFLG_RF |
1622 			     EFLG_AC | EFLG_ID | (1 << 1); /* Last one is the reserved bit */
1623 	unsigned long vm86_mask = EFLG_VM | EFLG_VIF | EFLG_VIP;
1624 
1625 	/* TODO: Add stack limit check */
1626 
1627 	rc = emulate_pop(ctxt, &temp_eip, ctxt->op_bytes);
1628 
1629 	if (rc != X86EMUL_CONTINUE)
1630 		return rc;
1631 
1632 	if (temp_eip & ~0xffff)
1633 		return emulate_gp(ctxt, 0);
1634 
1635 	rc = emulate_pop(ctxt, &cs, ctxt->op_bytes);
1636 
1637 	if (rc != X86EMUL_CONTINUE)
1638 		return rc;
1639 
1640 	rc = emulate_pop(ctxt, &temp_eflags, ctxt->op_bytes);
1641 
1642 	if (rc != X86EMUL_CONTINUE)
1643 		return rc;
1644 
1645 	rc = load_segment_descriptor(ctxt, (u16)cs, VCPU_SREG_CS);
1646 
1647 	if (rc != X86EMUL_CONTINUE)
1648 		return rc;
1649 
1650 	ctxt->_eip = temp_eip;
1651 
1652 
1653 	if (ctxt->op_bytes == 4)
1654 		ctxt->eflags = ((temp_eflags & mask) | (ctxt->eflags & vm86_mask));
1655 	else if (ctxt->op_bytes == 2) {
1656 		ctxt->eflags &= ~0xffff;
1657 		ctxt->eflags |= temp_eflags;
1658 	}
1659 
1660 	ctxt->eflags &= ~EFLG_RESERVED_ZEROS_MASK; /* Clear reserved zeros */
1661 	ctxt->eflags |= EFLG_RESERVED_ONE_MASK;
1662 
1663 	return rc;
1664 }
1665 
1666 static int em_iret(struct x86_emulate_ctxt *ctxt)
1667 {
1668 	switch(ctxt->mode) {
1669 	case X86EMUL_MODE_REAL:
1670 		return emulate_iret_real(ctxt);
1671 	case X86EMUL_MODE_VM86:
1672 	case X86EMUL_MODE_PROT16:
1673 	case X86EMUL_MODE_PROT32:
1674 	case X86EMUL_MODE_PROT64:
1675 	default:
1676 		/* iret from protected mode unimplemented yet */
1677 		return X86EMUL_UNHANDLEABLE;
1678 	}
1679 }
1680 
1681 static int em_jmp_far(struct x86_emulate_ctxt *ctxt)
1682 {
1683 	int rc;
1684 	unsigned short sel;
1685 
1686 	memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
1687 
1688 	rc = load_segment_descriptor(ctxt, sel, VCPU_SREG_CS);
1689 	if (rc != X86EMUL_CONTINUE)
1690 		return rc;
1691 
1692 	ctxt->_eip = 0;
1693 	memcpy(&ctxt->_eip, ctxt->src.valptr, ctxt->op_bytes);
1694 	return X86EMUL_CONTINUE;
1695 }
1696 
1697 static int em_grp2(struct x86_emulate_ctxt *ctxt)
1698 {
1699 	switch (ctxt->modrm_reg) {
1700 	case 0:	/* rol */
1701 		emulate_2op_SrcB(ctxt, "rol");
1702 		break;
1703 	case 1:	/* ror */
1704 		emulate_2op_SrcB(ctxt, "ror");
1705 		break;
1706 	case 2:	/* rcl */
1707 		emulate_2op_SrcB(ctxt, "rcl");
1708 		break;
1709 	case 3:	/* rcr */
1710 		emulate_2op_SrcB(ctxt, "rcr");
1711 		break;
1712 	case 4:	/* sal/shl */
1713 	case 6:	/* sal/shl */
1714 		emulate_2op_SrcB(ctxt, "sal");
1715 		break;
1716 	case 5:	/* shr */
1717 		emulate_2op_SrcB(ctxt, "shr");
1718 		break;
1719 	case 7:	/* sar */
1720 		emulate_2op_SrcB(ctxt, "sar");
1721 		break;
1722 	}
1723 	return X86EMUL_CONTINUE;
1724 }
1725 
1726 static int em_not(struct x86_emulate_ctxt *ctxt)
1727 {
1728 	ctxt->dst.val = ~ctxt->dst.val;
1729 	return X86EMUL_CONTINUE;
1730 }
1731 
1732 static int em_neg(struct x86_emulate_ctxt *ctxt)
1733 {
1734 	emulate_1op(ctxt, "neg");
1735 	return X86EMUL_CONTINUE;
1736 }
1737 
1738 static int em_mul_ex(struct x86_emulate_ctxt *ctxt)
1739 {
1740 	u8 ex = 0;
1741 
1742 	emulate_1op_rax_rdx(ctxt, "mul", ex);
1743 	return X86EMUL_CONTINUE;
1744 }
1745 
1746 static int em_imul_ex(struct x86_emulate_ctxt *ctxt)
1747 {
1748 	u8 ex = 0;
1749 
1750 	emulate_1op_rax_rdx(ctxt, "imul", ex);
1751 	return X86EMUL_CONTINUE;
1752 }
1753 
1754 static int em_div_ex(struct x86_emulate_ctxt *ctxt)
1755 {
1756 	u8 de = 0;
1757 
1758 	emulate_1op_rax_rdx(ctxt, "div", de);
1759 	if (de)
1760 		return emulate_de(ctxt);
1761 	return X86EMUL_CONTINUE;
1762 }
1763 
1764 static int em_idiv_ex(struct x86_emulate_ctxt *ctxt)
1765 {
1766 	u8 de = 0;
1767 
1768 	emulate_1op_rax_rdx(ctxt, "idiv", de);
1769 	if (de)
1770 		return emulate_de(ctxt);
1771 	return X86EMUL_CONTINUE;
1772 }
1773 
1774 static int em_grp45(struct x86_emulate_ctxt *ctxt)
1775 {
1776 	int rc = X86EMUL_CONTINUE;
1777 
1778 	switch (ctxt->modrm_reg) {
1779 	case 0:	/* inc */
1780 		emulate_1op(ctxt, "inc");
1781 		break;
1782 	case 1:	/* dec */
1783 		emulate_1op(ctxt, "dec");
1784 		break;
1785 	case 2: /* call near abs */ {
1786 		long int old_eip;
1787 		old_eip = ctxt->_eip;
1788 		ctxt->_eip = ctxt->src.val;
1789 		ctxt->src.val = old_eip;
1790 		rc = em_push(ctxt);
1791 		break;
1792 	}
1793 	case 4: /* jmp abs */
1794 		ctxt->_eip = ctxt->src.val;
1795 		break;
1796 	case 5: /* jmp far */
1797 		rc = em_jmp_far(ctxt);
1798 		break;
1799 	case 6:	/* push */
1800 		rc = em_push(ctxt);
1801 		break;
1802 	}
1803 	return rc;
1804 }
1805 
1806 static int em_cmpxchg8b(struct x86_emulate_ctxt *ctxt)
1807 {
1808 	u64 old = ctxt->dst.orig_val64;
1809 
1810 	if (((u32) (old >> 0) != (u32) ctxt->regs[VCPU_REGS_RAX]) ||
1811 	    ((u32) (old >> 32) != (u32) ctxt->regs[VCPU_REGS_RDX])) {
1812 		ctxt->regs[VCPU_REGS_RAX] = (u32) (old >> 0);
1813 		ctxt->regs[VCPU_REGS_RDX] = (u32) (old >> 32);
1814 		ctxt->eflags &= ~EFLG_ZF;
1815 	} else {
1816 		ctxt->dst.val64 = ((u64)ctxt->regs[VCPU_REGS_RCX] << 32) |
1817 			(u32) ctxt->regs[VCPU_REGS_RBX];
1818 
1819 		ctxt->eflags |= EFLG_ZF;
1820 	}
1821 	return X86EMUL_CONTINUE;
1822 }
1823 
1824 static int em_ret(struct x86_emulate_ctxt *ctxt)
1825 {
1826 	ctxt->dst.type = OP_REG;
1827 	ctxt->dst.addr.reg = &ctxt->_eip;
1828 	ctxt->dst.bytes = ctxt->op_bytes;
1829 	return em_pop(ctxt);
1830 }
1831 
1832 static int em_ret_far(struct x86_emulate_ctxt *ctxt)
1833 {
1834 	int rc;
1835 	unsigned long cs;
1836 
1837 	rc = emulate_pop(ctxt, &ctxt->_eip, ctxt->op_bytes);
1838 	if (rc != X86EMUL_CONTINUE)
1839 		return rc;
1840 	if (ctxt->op_bytes == 4)
1841 		ctxt->_eip = (u32)ctxt->_eip;
1842 	rc = emulate_pop(ctxt, &cs, ctxt->op_bytes);
1843 	if (rc != X86EMUL_CONTINUE)
1844 		return rc;
1845 	rc = load_segment_descriptor(ctxt, (u16)cs, VCPU_SREG_CS);
1846 	return rc;
1847 }
1848 
1849 static int em_cmpxchg(struct x86_emulate_ctxt *ctxt)
1850 {
1851 	/* Save real source value, then compare EAX against destination. */
1852 	ctxt->src.orig_val = ctxt->src.val;
1853 	ctxt->src.val = ctxt->regs[VCPU_REGS_RAX];
1854 	emulate_2op_SrcV(ctxt, "cmp");
1855 
1856 	if (ctxt->eflags & EFLG_ZF) {
1857 		/* Success: write back to memory. */
1858 		ctxt->dst.val = ctxt->src.orig_val;
1859 	} else {
1860 		/* Failure: write the value we saw to EAX. */
1861 		ctxt->dst.type = OP_REG;
1862 		ctxt->dst.addr.reg = (unsigned long *)&ctxt->regs[VCPU_REGS_RAX];
1863 	}
1864 	return X86EMUL_CONTINUE;
1865 }
1866 
1867 static int em_lseg(struct x86_emulate_ctxt *ctxt)
1868 {
1869 	int seg = ctxt->src2.val;
1870 	unsigned short sel;
1871 	int rc;
1872 
1873 	memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
1874 
1875 	rc = load_segment_descriptor(ctxt, sel, seg);
1876 	if (rc != X86EMUL_CONTINUE)
1877 		return rc;
1878 
1879 	ctxt->dst.val = ctxt->src.val;
1880 	return rc;
1881 }
1882 
1883 static void
1884 setup_syscalls_segments(struct x86_emulate_ctxt *ctxt,
1885 			struct desc_struct *cs, struct desc_struct *ss)
1886 {
1887 	u16 selector;
1888 
1889 	memset(cs, 0, sizeof(struct desc_struct));
1890 	ctxt->ops->get_segment(ctxt, &selector, cs, NULL, VCPU_SREG_CS);
1891 	memset(ss, 0, sizeof(struct desc_struct));
1892 
1893 	cs->l = 0;		/* will be adjusted later */
1894 	set_desc_base(cs, 0);	/* flat segment */
1895 	cs->g = 1;		/* 4kb granularity */
1896 	set_desc_limit(cs, 0xfffff);	/* 4GB limit */
1897 	cs->type = 0x0b;	/* Read, Execute, Accessed */
1898 	cs->s = 1;
1899 	cs->dpl = 0;		/* will be adjusted later */
1900 	cs->p = 1;
1901 	cs->d = 1;
1902 
1903 	set_desc_base(ss, 0);	/* flat segment */
1904 	set_desc_limit(ss, 0xfffff);	/* 4GB limit */
1905 	ss->g = 1;		/* 4kb granularity */
1906 	ss->s = 1;
1907 	ss->type = 0x03;	/* Read/Write, Accessed */
1908 	ss->d = 1;		/* 32bit stack segment */
1909 	ss->dpl = 0;
1910 	ss->p = 1;
1911 }
1912 
1913 static bool vendor_intel(struct x86_emulate_ctxt *ctxt)
1914 {
1915 	u32 eax, ebx, ecx, edx;
1916 
1917 	eax = ecx = 0;
1918 	return ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx)
1919 		&& ebx == X86EMUL_CPUID_VENDOR_GenuineIntel_ebx
1920 		&& ecx == X86EMUL_CPUID_VENDOR_GenuineIntel_ecx
1921 		&& edx == X86EMUL_CPUID_VENDOR_GenuineIntel_edx;
1922 }
1923 
1924 static bool em_syscall_is_enabled(struct x86_emulate_ctxt *ctxt)
1925 {
1926 	struct x86_emulate_ops *ops = ctxt->ops;
1927 	u32 eax, ebx, ecx, edx;
1928 
1929 	/*
1930 	 * syscall should always be enabled in longmode - so only become
1931 	 * vendor specific (cpuid) if other modes are active...
1932 	 */
1933 	if (ctxt->mode == X86EMUL_MODE_PROT64)
1934 		return true;
1935 
1936 	eax = 0x00000000;
1937 	ecx = 0x00000000;
1938 	if (ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx)) {
1939 		/*
1940 		 * Intel ("GenuineIntel")
1941 		 * remark: Intel CPUs only support "syscall" in 64bit
1942 		 * longmode. Also an 64bit guest with a
1943 		 * 32bit compat-app running will #UD !! While this
1944 		 * behaviour can be fixed (by emulating) into AMD
1945 		 * response - CPUs of AMD can't behave like Intel.
1946 		 */
1947 		if (ebx == X86EMUL_CPUID_VENDOR_GenuineIntel_ebx &&
1948 		    ecx == X86EMUL_CPUID_VENDOR_GenuineIntel_ecx &&
1949 		    edx == X86EMUL_CPUID_VENDOR_GenuineIntel_edx)
1950 			return false;
1951 
1952 		/* AMD ("AuthenticAMD") */
1953 		if (ebx == X86EMUL_CPUID_VENDOR_AuthenticAMD_ebx &&
1954 		    ecx == X86EMUL_CPUID_VENDOR_AuthenticAMD_ecx &&
1955 		    edx == X86EMUL_CPUID_VENDOR_AuthenticAMD_edx)
1956 			return true;
1957 
1958 		/* AMD ("AMDisbetter!") */
1959 		if (ebx == X86EMUL_CPUID_VENDOR_AMDisbetterI_ebx &&
1960 		    ecx == X86EMUL_CPUID_VENDOR_AMDisbetterI_ecx &&
1961 		    edx == X86EMUL_CPUID_VENDOR_AMDisbetterI_edx)
1962 			return true;
1963 	}
1964 
1965 	/* default: (not Intel, not AMD), apply Intel's stricter rules... */
1966 	return false;
1967 }
1968 
1969 static int em_syscall(struct x86_emulate_ctxt *ctxt)
1970 {
1971 	struct x86_emulate_ops *ops = ctxt->ops;
1972 	struct desc_struct cs, ss;
1973 	u64 msr_data;
1974 	u16 cs_sel, ss_sel;
1975 	u64 efer = 0;
1976 
1977 	/* syscall is not available in real mode */
1978 	if (ctxt->mode == X86EMUL_MODE_REAL ||
1979 	    ctxt->mode == X86EMUL_MODE_VM86)
1980 		return emulate_ud(ctxt);
1981 
1982 	if (!(em_syscall_is_enabled(ctxt)))
1983 		return emulate_ud(ctxt);
1984 
1985 	ops->get_msr(ctxt, MSR_EFER, &efer);
1986 	setup_syscalls_segments(ctxt, &cs, &ss);
1987 
1988 	if (!(efer & EFER_SCE))
1989 		return emulate_ud(ctxt);
1990 
1991 	ops->get_msr(ctxt, MSR_STAR, &msr_data);
1992 	msr_data >>= 32;
1993 	cs_sel = (u16)(msr_data & 0xfffc);
1994 	ss_sel = (u16)(msr_data + 8);
1995 
1996 	if (efer & EFER_LMA) {
1997 		cs.d = 0;
1998 		cs.l = 1;
1999 	}
2000 	ops->set_segment(ctxt, cs_sel, &cs, 0, VCPU_SREG_CS);
2001 	ops->set_segment(ctxt, ss_sel, &ss, 0, VCPU_SREG_SS);
2002 
2003 	ctxt->regs[VCPU_REGS_RCX] = ctxt->_eip;
2004 	if (efer & EFER_LMA) {
2005 #ifdef CONFIG_X86_64
2006 		ctxt->regs[VCPU_REGS_R11] = ctxt->eflags & ~EFLG_RF;
2007 
2008 		ops->get_msr(ctxt,
2009 			     ctxt->mode == X86EMUL_MODE_PROT64 ?
2010 			     MSR_LSTAR : MSR_CSTAR, &msr_data);
2011 		ctxt->_eip = msr_data;
2012 
2013 		ops->get_msr(ctxt, MSR_SYSCALL_MASK, &msr_data);
2014 		ctxt->eflags &= ~(msr_data | EFLG_RF);
2015 #endif
2016 	} else {
2017 		/* legacy mode */
2018 		ops->get_msr(ctxt, MSR_STAR, &msr_data);
2019 		ctxt->_eip = (u32)msr_data;
2020 
2021 		ctxt->eflags &= ~(EFLG_VM | EFLG_IF | EFLG_RF);
2022 	}
2023 
2024 	return X86EMUL_CONTINUE;
2025 }
2026 
2027 static int em_sysenter(struct x86_emulate_ctxt *ctxt)
2028 {
2029 	struct x86_emulate_ops *ops = ctxt->ops;
2030 	struct desc_struct cs, ss;
2031 	u64 msr_data;
2032 	u16 cs_sel, ss_sel;
2033 	u64 efer = 0;
2034 
2035 	ops->get_msr(ctxt, MSR_EFER, &efer);
2036 	/* inject #GP if in real mode */
2037 	if (ctxt->mode == X86EMUL_MODE_REAL)
2038 		return emulate_gp(ctxt, 0);
2039 
2040 	/*
2041 	 * Not recognized on AMD in compat mode (but is recognized in legacy
2042 	 * mode).
2043 	 */
2044 	if ((ctxt->mode == X86EMUL_MODE_PROT32) && (efer & EFER_LMA)
2045 	    && !vendor_intel(ctxt))
2046 		return emulate_ud(ctxt);
2047 
2048 	/* XXX sysenter/sysexit have not been tested in 64bit mode.
2049 	* Therefore, we inject an #UD.
2050 	*/
2051 	if (ctxt->mode == X86EMUL_MODE_PROT64)
2052 		return emulate_ud(ctxt);
2053 
2054 	setup_syscalls_segments(ctxt, &cs, &ss);
2055 
2056 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_CS, &msr_data);
2057 	switch (ctxt->mode) {
2058 	case X86EMUL_MODE_PROT32:
2059 		if ((msr_data & 0xfffc) == 0x0)
2060 			return emulate_gp(ctxt, 0);
2061 		break;
2062 	case X86EMUL_MODE_PROT64:
2063 		if (msr_data == 0x0)
2064 			return emulate_gp(ctxt, 0);
2065 		break;
2066 	}
2067 
2068 	ctxt->eflags &= ~(EFLG_VM | EFLG_IF | EFLG_RF);
2069 	cs_sel = (u16)msr_data;
2070 	cs_sel &= ~SELECTOR_RPL_MASK;
2071 	ss_sel = cs_sel + 8;
2072 	ss_sel &= ~SELECTOR_RPL_MASK;
2073 	if (ctxt->mode == X86EMUL_MODE_PROT64 || (efer & EFER_LMA)) {
2074 		cs.d = 0;
2075 		cs.l = 1;
2076 	}
2077 
2078 	ops->set_segment(ctxt, cs_sel, &cs, 0, VCPU_SREG_CS);
2079 	ops->set_segment(ctxt, ss_sel, &ss, 0, VCPU_SREG_SS);
2080 
2081 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_EIP, &msr_data);
2082 	ctxt->_eip = msr_data;
2083 
2084 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_ESP, &msr_data);
2085 	ctxt->regs[VCPU_REGS_RSP] = msr_data;
2086 
2087 	return X86EMUL_CONTINUE;
2088 }
2089 
2090 static int em_sysexit(struct x86_emulate_ctxt *ctxt)
2091 {
2092 	struct x86_emulate_ops *ops = ctxt->ops;
2093 	struct desc_struct cs, ss;
2094 	u64 msr_data;
2095 	int usermode;
2096 	u16 cs_sel = 0, ss_sel = 0;
2097 
2098 	/* inject #GP if in real mode or Virtual 8086 mode */
2099 	if (ctxt->mode == X86EMUL_MODE_REAL ||
2100 	    ctxt->mode == X86EMUL_MODE_VM86)
2101 		return emulate_gp(ctxt, 0);
2102 
2103 	setup_syscalls_segments(ctxt, &cs, &ss);
2104 
2105 	if ((ctxt->rex_prefix & 0x8) != 0x0)
2106 		usermode = X86EMUL_MODE_PROT64;
2107 	else
2108 		usermode = X86EMUL_MODE_PROT32;
2109 
2110 	cs.dpl = 3;
2111 	ss.dpl = 3;
2112 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_CS, &msr_data);
2113 	switch (usermode) {
2114 	case X86EMUL_MODE_PROT32:
2115 		cs_sel = (u16)(msr_data + 16);
2116 		if ((msr_data & 0xfffc) == 0x0)
2117 			return emulate_gp(ctxt, 0);
2118 		ss_sel = (u16)(msr_data + 24);
2119 		break;
2120 	case X86EMUL_MODE_PROT64:
2121 		cs_sel = (u16)(msr_data + 32);
2122 		if (msr_data == 0x0)
2123 			return emulate_gp(ctxt, 0);
2124 		ss_sel = cs_sel + 8;
2125 		cs.d = 0;
2126 		cs.l = 1;
2127 		break;
2128 	}
2129 	cs_sel |= SELECTOR_RPL_MASK;
2130 	ss_sel |= SELECTOR_RPL_MASK;
2131 
2132 	ops->set_segment(ctxt, cs_sel, &cs, 0, VCPU_SREG_CS);
2133 	ops->set_segment(ctxt, ss_sel, &ss, 0, VCPU_SREG_SS);
2134 
2135 	ctxt->_eip = ctxt->regs[VCPU_REGS_RDX];
2136 	ctxt->regs[VCPU_REGS_RSP] = ctxt->regs[VCPU_REGS_RCX];
2137 
2138 	return X86EMUL_CONTINUE;
2139 }
2140 
2141 static bool emulator_bad_iopl(struct x86_emulate_ctxt *ctxt)
2142 {
2143 	int iopl;
2144 	if (ctxt->mode == X86EMUL_MODE_REAL)
2145 		return false;
2146 	if (ctxt->mode == X86EMUL_MODE_VM86)
2147 		return true;
2148 	iopl = (ctxt->eflags & X86_EFLAGS_IOPL) >> IOPL_SHIFT;
2149 	return ctxt->ops->cpl(ctxt) > iopl;
2150 }
2151 
2152 static bool emulator_io_port_access_allowed(struct x86_emulate_ctxt *ctxt,
2153 					    u16 port, u16 len)
2154 {
2155 	struct x86_emulate_ops *ops = ctxt->ops;
2156 	struct desc_struct tr_seg;
2157 	u32 base3;
2158 	int r;
2159 	u16 tr, io_bitmap_ptr, perm, bit_idx = port & 0x7;
2160 	unsigned mask = (1 << len) - 1;
2161 	unsigned long base;
2162 
2163 	ops->get_segment(ctxt, &tr, &tr_seg, &base3, VCPU_SREG_TR);
2164 	if (!tr_seg.p)
2165 		return false;
2166 	if (desc_limit_scaled(&tr_seg) < 103)
2167 		return false;
2168 	base = get_desc_base(&tr_seg);
2169 #ifdef CONFIG_X86_64
2170 	base |= ((u64)base3) << 32;
2171 #endif
2172 	r = ops->read_std(ctxt, base + 102, &io_bitmap_ptr, 2, NULL);
2173 	if (r != X86EMUL_CONTINUE)
2174 		return false;
2175 	if (io_bitmap_ptr + port/8 > desc_limit_scaled(&tr_seg))
2176 		return false;
2177 	r = ops->read_std(ctxt, base + io_bitmap_ptr + port/8, &perm, 2, NULL);
2178 	if (r != X86EMUL_CONTINUE)
2179 		return false;
2180 	if ((perm >> bit_idx) & mask)
2181 		return false;
2182 	return true;
2183 }
2184 
2185 static bool emulator_io_permited(struct x86_emulate_ctxt *ctxt,
2186 				 u16 port, u16 len)
2187 {
2188 	if (ctxt->perm_ok)
2189 		return true;
2190 
2191 	if (emulator_bad_iopl(ctxt))
2192 		if (!emulator_io_port_access_allowed(ctxt, port, len))
2193 			return false;
2194 
2195 	ctxt->perm_ok = true;
2196 
2197 	return true;
2198 }
2199 
2200 static void save_state_to_tss16(struct x86_emulate_ctxt *ctxt,
2201 				struct tss_segment_16 *tss)
2202 {
2203 	tss->ip = ctxt->_eip;
2204 	tss->flag = ctxt->eflags;
2205 	tss->ax = ctxt->regs[VCPU_REGS_RAX];
2206 	tss->cx = ctxt->regs[VCPU_REGS_RCX];
2207 	tss->dx = ctxt->regs[VCPU_REGS_RDX];
2208 	tss->bx = ctxt->regs[VCPU_REGS_RBX];
2209 	tss->sp = ctxt->regs[VCPU_REGS_RSP];
2210 	tss->bp = ctxt->regs[VCPU_REGS_RBP];
2211 	tss->si = ctxt->regs[VCPU_REGS_RSI];
2212 	tss->di = ctxt->regs[VCPU_REGS_RDI];
2213 
2214 	tss->es = get_segment_selector(ctxt, VCPU_SREG_ES);
2215 	tss->cs = get_segment_selector(ctxt, VCPU_SREG_CS);
2216 	tss->ss = get_segment_selector(ctxt, VCPU_SREG_SS);
2217 	tss->ds = get_segment_selector(ctxt, VCPU_SREG_DS);
2218 	tss->ldt = get_segment_selector(ctxt, VCPU_SREG_LDTR);
2219 }
2220 
2221 static int load_state_from_tss16(struct x86_emulate_ctxt *ctxt,
2222 				 struct tss_segment_16 *tss)
2223 {
2224 	int ret;
2225 
2226 	ctxt->_eip = tss->ip;
2227 	ctxt->eflags = tss->flag | 2;
2228 	ctxt->regs[VCPU_REGS_RAX] = tss->ax;
2229 	ctxt->regs[VCPU_REGS_RCX] = tss->cx;
2230 	ctxt->regs[VCPU_REGS_RDX] = tss->dx;
2231 	ctxt->regs[VCPU_REGS_RBX] = tss->bx;
2232 	ctxt->regs[VCPU_REGS_RSP] = tss->sp;
2233 	ctxt->regs[VCPU_REGS_RBP] = tss->bp;
2234 	ctxt->regs[VCPU_REGS_RSI] = tss->si;
2235 	ctxt->regs[VCPU_REGS_RDI] = tss->di;
2236 
2237 	/*
2238 	 * SDM says that segment selectors are loaded before segment
2239 	 * descriptors
2240 	 */
2241 	set_segment_selector(ctxt, tss->ldt, VCPU_SREG_LDTR);
2242 	set_segment_selector(ctxt, tss->es, VCPU_SREG_ES);
2243 	set_segment_selector(ctxt, tss->cs, VCPU_SREG_CS);
2244 	set_segment_selector(ctxt, tss->ss, VCPU_SREG_SS);
2245 	set_segment_selector(ctxt, tss->ds, VCPU_SREG_DS);
2246 
2247 	/*
2248 	 * Now load segment descriptors. If fault happenes at this stage
2249 	 * it is handled in a context of new task
2250 	 */
2251 	ret = load_segment_descriptor(ctxt, tss->ldt, VCPU_SREG_LDTR);
2252 	if (ret != X86EMUL_CONTINUE)
2253 		return ret;
2254 	ret = load_segment_descriptor(ctxt, tss->es, VCPU_SREG_ES);
2255 	if (ret != X86EMUL_CONTINUE)
2256 		return ret;
2257 	ret = load_segment_descriptor(ctxt, tss->cs, VCPU_SREG_CS);
2258 	if (ret != X86EMUL_CONTINUE)
2259 		return ret;
2260 	ret = load_segment_descriptor(ctxt, tss->ss, VCPU_SREG_SS);
2261 	if (ret != X86EMUL_CONTINUE)
2262 		return ret;
2263 	ret = load_segment_descriptor(ctxt, tss->ds, VCPU_SREG_DS);
2264 	if (ret != X86EMUL_CONTINUE)
2265 		return ret;
2266 
2267 	return X86EMUL_CONTINUE;
2268 }
2269 
2270 static int task_switch_16(struct x86_emulate_ctxt *ctxt,
2271 			  u16 tss_selector, u16 old_tss_sel,
2272 			  ulong old_tss_base, struct desc_struct *new_desc)
2273 {
2274 	struct x86_emulate_ops *ops = ctxt->ops;
2275 	struct tss_segment_16 tss_seg;
2276 	int ret;
2277 	u32 new_tss_base = get_desc_base(new_desc);
2278 
2279 	ret = ops->read_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2280 			    &ctxt->exception);
2281 	if (ret != X86EMUL_CONTINUE)
2282 		/* FIXME: need to provide precise fault address */
2283 		return ret;
2284 
2285 	save_state_to_tss16(ctxt, &tss_seg);
2286 
2287 	ret = ops->write_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2288 			     &ctxt->exception);
2289 	if (ret != X86EMUL_CONTINUE)
2290 		/* FIXME: need to provide precise fault address */
2291 		return ret;
2292 
2293 	ret = ops->read_std(ctxt, new_tss_base, &tss_seg, sizeof tss_seg,
2294 			    &ctxt->exception);
2295 	if (ret != X86EMUL_CONTINUE)
2296 		/* FIXME: need to provide precise fault address */
2297 		return ret;
2298 
2299 	if (old_tss_sel != 0xffff) {
2300 		tss_seg.prev_task_link = old_tss_sel;
2301 
2302 		ret = ops->write_std(ctxt, new_tss_base,
2303 				     &tss_seg.prev_task_link,
2304 				     sizeof tss_seg.prev_task_link,
2305 				     &ctxt->exception);
2306 		if (ret != X86EMUL_CONTINUE)
2307 			/* FIXME: need to provide precise fault address */
2308 			return ret;
2309 	}
2310 
2311 	return load_state_from_tss16(ctxt, &tss_seg);
2312 }
2313 
2314 static void save_state_to_tss32(struct x86_emulate_ctxt *ctxt,
2315 				struct tss_segment_32 *tss)
2316 {
2317 	tss->cr3 = ctxt->ops->get_cr(ctxt, 3);
2318 	tss->eip = ctxt->_eip;
2319 	tss->eflags = ctxt->eflags;
2320 	tss->eax = ctxt->regs[VCPU_REGS_RAX];
2321 	tss->ecx = ctxt->regs[VCPU_REGS_RCX];
2322 	tss->edx = ctxt->regs[VCPU_REGS_RDX];
2323 	tss->ebx = ctxt->regs[VCPU_REGS_RBX];
2324 	tss->esp = ctxt->regs[VCPU_REGS_RSP];
2325 	tss->ebp = ctxt->regs[VCPU_REGS_RBP];
2326 	tss->esi = ctxt->regs[VCPU_REGS_RSI];
2327 	tss->edi = ctxt->regs[VCPU_REGS_RDI];
2328 
2329 	tss->es = get_segment_selector(ctxt, VCPU_SREG_ES);
2330 	tss->cs = get_segment_selector(ctxt, VCPU_SREG_CS);
2331 	tss->ss = get_segment_selector(ctxt, VCPU_SREG_SS);
2332 	tss->ds = get_segment_selector(ctxt, VCPU_SREG_DS);
2333 	tss->fs = get_segment_selector(ctxt, VCPU_SREG_FS);
2334 	tss->gs = get_segment_selector(ctxt, VCPU_SREG_GS);
2335 	tss->ldt_selector = get_segment_selector(ctxt, VCPU_SREG_LDTR);
2336 }
2337 
2338 static int load_state_from_tss32(struct x86_emulate_ctxt *ctxt,
2339 				 struct tss_segment_32 *tss)
2340 {
2341 	int ret;
2342 
2343 	if (ctxt->ops->set_cr(ctxt, 3, tss->cr3))
2344 		return emulate_gp(ctxt, 0);
2345 	ctxt->_eip = tss->eip;
2346 	ctxt->eflags = tss->eflags | 2;
2347 
2348 	/* General purpose registers */
2349 	ctxt->regs[VCPU_REGS_RAX] = tss->eax;
2350 	ctxt->regs[VCPU_REGS_RCX] = tss->ecx;
2351 	ctxt->regs[VCPU_REGS_RDX] = tss->edx;
2352 	ctxt->regs[VCPU_REGS_RBX] = tss->ebx;
2353 	ctxt->regs[VCPU_REGS_RSP] = tss->esp;
2354 	ctxt->regs[VCPU_REGS_RBP] = tss->ebp;
2355 	ctxt->regs[VCPU_REGS_RSI] = tss->esi;
2356 	ctxt->regs[VCPU_REGS_RDI] = tss->edi;
2357 
2358 	/*
2359 	 * SDM says that segment selectors are loaded before segment
2360 	 * descriptors
2361 	 */
2362 	set_segment_selector(ctxt, tss->ldt_selector, VCPU_SREG_LDTR);
2363 	set_segment_selector(ctxt, tss->es, VCPU_SREG_ES);
2364 	set_segment_selector(ctxt, tss->cs, VCPU_SREG_CS);
2365 	set_segment_selector(ctxt, tss->ss, VCPU_SREG_SS);
2366 	set_segment_selector(ctxt, tss->ds, VCPU_SREG_DS);
2367 	set_segment_selector(ctxt, tss->fs, VCPU_SREG_FS);
2368 	set_segment_selector(ctxt, tss->gs, VCPU_SREG_GS);
2369 
2370 	/*
2371 	 * If we're switching between Protected Mode and VM86, we need to make
2372 	 * sure to update the mode before loading the segment descriptors so
2373 	 * that the selectors are interpreted correctly.
2374 	 *
2375 	 * Need to get rflags to the vcpu struct immediately because it
2376 	 * influences the CPL which is checked at least when loading the segment
2377 	 * descriptors and when pushing an error code to the new kernel stack.
2378 	 *
2379 	 * TODO Introduce a separate ctxt->ops->set_cpl callback
2380 	 */
2381 	if (ctxt->eflags & X86_EFLAGS_VM)
2382 		ctxt->mode = X86EMUL_MODE_VM86;
2383 	else
2384 		ctxt->mode = X86EMUL_MODE_PROT32;
2385 
2386 	ctxt->ops->set_rflags(ctxt, ctxt->eflags);
2387 
2388 	/*
2389 	 * Now load segment descriptors. If fault happenes at this stage
2390 	 * it is handled in a context of new task
2391 	 */
2392 	ret = load_segment_descriptor(ctxt, tss->ldt_selector, VCPU_SREG_LDTR);
2393 	if (ret != X86EMUL_CONTINUE)
2394 		return ret;
2395 	ret = load_segment_descriptor(ctxt, tss->es, VCPU_SREG_ES);
2396 	if (ret != X86EMUL_CONTINUE)
2397 		return ret;
2398 	ret = load_segment_descriptor(ctxt, tss->cs, VCPU_SREG_CS);
2399 	if (ret != X86EMUL_CONTINUE)
2400 		return ret;
2401 	ret = load_segment_descriptor(ctxt, tss->ss, VCPU_SREG_SS);
2402 	if (ret != X86EMUL_CONTINUE)
2403 		return ret;
2404 	ret = load_segment_descriptor(ctxt, tss->ds, VCPU_SREG_DS);
2405 	if (ret != X86EMUL_CONTINUE)
2406 		return ret;
2407 	ret = load_segment_descriptor(ctxt, tss->fs, VCPU_SREG_FS);
2408 	if (ret != X86EMUL_CONTINUE)
2409 		return ret;
2410 	ret = load_segment_descriptor(ctxt, tss->gs, VCPU_SREG_GS);
2411 	if (ret != X86EMUL_CONTINUE)
2412 		return ret;
2413 
2414 	return X86EMUL_CONTINUE;
2415 }
2416 
2417 static int task_switch_32(struct x86_emulate_ctxt *ctxt,
2418 			  u16 tss_selector, u16 old_tss_sel,
2419 			  ulong old_tss_base, struct desc_struct *new_desc)
2420 {
2421 	struct x86_emulate_ops *ops = ctxt->ops;
2422 	struct tss_segment_32 tss_seg;
2423 	int ret;
2424 	u32 new_tss_base = get_desc_base(new_desc);
2425 
2426 	ret = ops->read_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2427 			    &ctxt->exception);
2428 	if (ret != X86EMUL_CONTINUE)
2429 		/* FIXME: need to provide precise fault address */
2430 		return ret;
2431 
2432 	save_state_to_tss32(ctxt, &tss_seg);
2433 
2434 	ret = ops->write_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2435 			     &ctxt->exception);
2436 	if (ret != X86EMUL_CONTINUE)
2437 		/* FIXME: need to provide precise fault address */
2438 		return ret;
2439 
2440 	ret = ops->read_std(ctxt, new_tss_base, &tss_seg, sizeof tss_seg,
2441 			    &ctxt->exception);
2442 	if (ret != X86EMUL_CONTINUE)
2443 		/* FIXME: need to provide precise fault address */
2444 		return ret;
2445 
2446 	if (old_tss_sel != 0xffff) {
2447 		tss_seg.prev_task_link = old_tss_sel;
2448 
2449 		ret = ops->write_std(ctxt, new_tss_base,
2450 				     &tss_seg.prev_task_link,
2451 				     sizeof tss_seg.prev_task_link,
2452 				     &ctxt->exception);
2453 		if (ret != X86EMUL_CONTINUE)
2454 			/* FIXME: need to provide precise fault address */
2455 			return ret;
2456 	}
2457 
2458 	return load_state_from_tss32(ctxt, &tss_seg);
2459 }
2460 
2461 static int emulator_do_task_switch(struct x86_emulate_ctxt *ctxt,
2462 				   u16 tss_selector, int idt_index, int reason,
2463 				   bool has_error_code, u32 error_code)
2464 {
2465 	struct x86_emulate_ops *ops = ctxt->ops;
2466 	struct desc_struct curr_tss_desc, next_tss_desc;
2467 	int ret;
2468 	u16 old_tss_sel = get_segment_selector(ctxt, VCPU_SREG_TR);
2469 	ulong old_tss_base =
2470 		ops->get_cached_segment_base(ctxt, VCPU_SREG_TR);
2471 	u32 desc_limit;
2472 
2473 	/* FIXME: old_tss_base == ~0 ? */
2474 
2475 	ret = read_segment_descriptor(ctxt, tss_selector, &next_tss_desc);
2476 	if (ret != X86EMUL_CONTINUE)
2477 		return ret;
2478 	ret = read_segment_descriptor(ctxt, old_tss_sel, &curr_tss_desc);
2479 	if (ret != X86EMUL_CONTINUE)
2480 		return ret;
2481 
2482 	/* FIXME: check that next_tss_desc is tss */
2483 
2484 	/*
2485 	 * Check privileges. The three cases are task switch caused by...
2486 	 *
2487 	 * 1. jmp/call/int to task gate: Check against DPL of the task gate
2488 	 * 2. Exception/IRQ/iret: No check is performed
2489 	 * 3. jmp/call to TSS: Check agains DPL of the TSS
2490 	 */
2491 	if (reason == TASK_SWITCH_GATE) {
2492 		if (idt_index != -1) {
2493 			/* Software interrupts */
2494 			struct desc_struct task_gate_desc;
2495 			int dpl;
2496 
2497 			ret = read_interrupt_descriptor(ctxt, idt_index,
2498 							&task_gate_desc);
2499 			if (ret != X86EMUL_CONTINUE)
2500 				return ret;
2501 
2502 			dpl = task_gate_desc.dpl;
2503 			if ((tss_selector & 3) > dpl || ops->cpl(ctxt) > dpl)
2504 				return emulate_gp(ctxt, (idt_index << 3) | 0x2);
2505 		}
2506 	} else if (reason != TASK_SWITCH_IRET) {
2507 		int dpl = next_tss_desc.dpl;
2508 		if ((tss_selector & 3) > dpl || ops->cpl(ctxt) > dpl)
2509 			return emulate_gp(ctxt, tss_selector);
2510 	}
2511 
2512 
2513 	desc_limit = desc_limit_scaled(&next_tss_desc);
2514 	if (!next_tss_desc.p ||
2515 	    ((desc_limit < 0x67 && (next_tss_desc.type & 8)) ||
2516 	     desc_limit < 0x2b)) {
2517 		emulate_ts(ctxt, tss_selector & 0xfffc);
2518 		return X86EMUL_PROPAGATE_FAULT;
2519 	}
2520 
2521 	if (reason == TASK_SWITCH_IRET || reason == TASK_SWITCH_JMP) {
2522 		curr_tss_desc.type &= ~(1 << 1); /* clear busy flag */
2523 		write_segment_descriptor(ctxt, old_tss_sel, &curr_tss_desc);
2524 	}
2525 
2526 	if (reason == TASK_SWITCH_IRET)
2527 		ctxt->eflags = ctxt->eflags & ~X86_EFLAGS_NT;
2528 
2529 	/* set back link to prev task only if NT bit is set in eflags
2530 	   note that old_tss_sel is not used afetr this point */
2531 	if (reason != TASK_SWITCH_CALL && reason != TASK_SWITCH_GATE)
2532 		old_tss_sel = 0xffff;
2533 
2534 	if (next_tss_desc.type & 8)
2535 		ret = task_switch_32(ctxt, tss_selector, old_tss_sel,
2536 				     old_tss_base, &next_tss_desc);
2537 	else
2538 		ret = task_switch_16(ctxt, tss_selector, old_tss_sel,
2539 				     old_tss_base, &next_tss_desc);
2540 	if (ret != X86EMUL_CONTINUE)
2541 		return ret;
2542 
2543 	if (reason == TASK_SWITCH_CALL || reason == TASK_SWITCH_GATE)
2544 		ctxt->eflags = ctxt->eflags | X86_EFLAGS_NT;
2545 
2546 	if (reason != TASK_SWITCH_IRET) {
2547 		next_tss_desc.type |= (1 << 1); /* set busy flag */
2548 		write_segment_descriptor(ctxt, tss_selector, &next_tss_desc);
2549 	}
2550 
2551 	ops->set_cr(ctxt, 0,  ops->get_cr(ctxt, 0) | X86_CR0_TS);
2552 	ops->set_segment(ctxt, tss_selector, &next_tss_desc, 0, VCPU_SREG_TR);
2553 
2554 	if (has_error_code) {
2555 		ctxt->op_bytes = ctxt->ad_bytes = (next_tss_desc.type & 8) ? 4 : 2;
2556 		ctxt->lock_prefix = 0;
2557 		ctxt->src.val = (unsigned long) error_code;
2558 		ret = em_push(ctxt);
2559 	}
2560 
2561 	return ret;
2562 }
2563 
2564 int emulator_task_switch(struct x86_emulate_ctxt *ctxt,
2565 			 u16 tss_selector, int idt_index, int reason,
2566 			 bool has_error_code, u32 error_code)
2567 {
2568 	int rc;
2569 
2570 	ctxt->_eip = ctxt->eip;
2571 	ctxt->dst.type = OP_NONE;
2572 
2573 	rc = emulator_do_task_switch(ctxt, tss_selector, idt_index, reason,
2574 				     has_error_code, error_code);
2575 
2576 	if (rc == X86EMUL_CONTINUE)
2577 		ctxt->eip = ctxt->_eip;
2578 
2579 	return (rc == X86EMUL_UNHANDLEABLE) ? EMULATION_FAILED : EMULATION_OK;
2580 }
2581 
2582 static void string_addr_inc(struct x86_emulate_ctxt *ctxt, unsigned seg,
2583 			    int reg, struct operand *op)
2584 {
2585 	int df = (ctxt->eflags & EFLG_DF) ? -1 : 1;
2586 
2587 	register_address_increment(ctxt, &ctxt->regs[reg], df * op->bytes);
2588 	op->addr.mem.ea = register_address(ctxt, ctxt->regs[reg]);
2589 	op->addr.mem.seg = seg;
2590 }
2591 
2592 static int em_das(struct x86_emulate_ctxt *ctxt)
2593 {
2594 	u8 al, old_al;
2595 	bool af, cf, old_cf;
2596 
2597 	cf = ctxt->eflags & X86_EFLAGS_CF;
2598 	al = ctxt->dst.val;
2599 
2600 	old_al = al;
2601 	old_cf = cf;
2602 	cf = false;
2603 	af = ctxt->eflags & X86_EFLAGS_AF;
2604 	if ((al & 0x0f) > 9 || af) {
2605 		al -= 6;
2606 		cf = old_cf | (al >= 250);
2607 		af = true;
2608 	} else {
2609 		af = false;
2610 	}
2611 	if (old_al > 0x99 || old_cf) {
2612 		al -= 0x60;
2613 		cf = true;
2614 	}
2615 
2616 	ctxt->dst.val = al;
2617 	/* Set PF, ZF, SF */
2618 	ctxt->src.type = OP_IMM;
2619 	ctxt->src.val = 0;
2620 	ctxt->src.bytes = 1;
2621 	emulate_2op_SrcV(ctxt, "or");
2622 	ctxt->eflags &= ~(X86_EFLAGS_AF | X86_EFLAGS_CF);
2623 	if (cf)
2624 		ctxt->eflags |= X86_EFLAGS_CF;
2625 	if (af)
2626 		ctxt->eflags |= X86_EFLAGS_AF;
2627 	return X86EMUL_CONTINUE;
2628 }
2629 
2630 static int em_call(struct x86_emulate_ctxt *ctxt)
2631 {
2632 	long rel = ctxt->src.val;
2633 
2634 	ctxt->src.val = (unsigned long)ctxt->_eip;
2635 	jmp_rel(ctxt, rel);
2636 	return em_push(ctxt);
2637 }
2638 
2639 static int em_call_far(struct x86_emulate_ctxt *ctxt)
2640 {
2641 	u16 sel, old_cs;
2642 	ulong old_eip;
2643 	int rc;
2644 
2645 	old_cs = get_segment_selector(ctxt, VCPU_SREG_CS);
2646 	old_eip = ctxt->_eip;
2647 
2648 	memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
2649 	if (load_segment_descriptor(ctxt, sel, VCPU_SREG_CS))
2650 		return X86EMUL_CONTINUE;
2651 
2652 	ctxt->_eip = 0;
2653 	memcpy(&ctxt->_eip, ctxt->src.valptr, ctxt->op_bytes);
2654 
2655 	ctxt->src.val = old_cs;
2656 	rc = em_push(ctxt);
2657 	if (rc != X86EMUL_CONTINUE)
2658 		return rc;
2659 
2660 	ctxt->src.val = old_eip;
2661 	return em_push(ctxt);
2662 }
2663 
2664 static int em_ret_near_imm(struct x86_emulate_ctxt *ctxt)
2665 {
2666 	int rc;
2667 
2668 	ctxt->dst.type = OP_REG;
2669 	ctxt->dst.addr.reg = &ctxt->_eip;
2670 	ctxt->dst.bytes = ctxt->op_bytes;
2671 	rc = emulate_pop(ctxt, &ctxt->dst.val, ctxt->op_bytes);
2672 	if (rc != X86EMUL_CONTINUE)
2673 		return rc;
2674 	register_address_increment(ctxt, &ctxt->regs[VCPU_REGS_RSP], ctxt->src.val);
2675 	return X86EMUL_CONTINUE;
2676 }
2677 
2678 static int em_add(struct x86_emulate_ctxt *ctxt)
2679 {
2680 	emulate_2op_SrcV(ctxt, "add");
2681 	return X86EMUL_CONTINUE;
2682 }
2683 
2684 static int em_or(struct x86_emulate_ctxt *ctxt)
2685 {
2686 	emulate_2op_SrcV(ctxt, "or");
2687 	return X86EMUL_CONTINUE;
2688 }
2689 
2690 static int em_adc(struct x86_emulate_ctxt *ctxt)
2691 {
2692 	emulate_2op_SrcV(ctxt, "adc");
2693 	return X86EMUL_CONTINUE;
2694 }
2695 
2696 static int em_sbb(struct x86_emulate_ctxt *ctxt)
2697 {
2698 	emulate_2op_SrcV(ctxt, "sbb");
2699 	return X86EMUL_CONTINUE;
2700 }
2701 
2702 static int em_and(struct x86_emulate_ctxt *ctxt)
2703 {
2704 	emulate_2op_SrcV(ctxt, "and");
2705 	return X86EMUL_CONTINUE;
2706 }
2707 
2708 static int em_sub(struct x86_emulate_ctxt *ctxt)
2709 {
2710 	emulate_2op_SrcV(ctxt, "sub");
2711 	return X86EMUL_CONTINUE;
2712 }
2713 
2714 static int em_xor(struct x86_emulate_ctxt *ctxt)
2715 {
2716 	emulate_2op_SrcV(ctxt, "xor");
2717 	return X86EMUL_CONTINUE;
2718 }
2719 
2720 static int em_cmp(struct x86_emulate_ctxt *ctxt)
2721 {
2722 	emulate_2op_SrcV(ctxt, "cmp");
2723 	/* Disable writeback. */
2724 	ctxt->dst.type = OP_NONE;
2725 	return X86EMUL_CONTINUE;
2726 }
2727 
2728 static int em_test(struct x86_emulate_ctxt *ctxt)
2729 {
2730 	emulate_2op_SrcV(ctxt, "test");
2731 	/* Disable writeback. */
2732 	ctxt->dst.type = OP_NONE;
2733 	return X86EMUL_CONTINUE;
2734 }
2735 
2736 static int em_xchg(struct x86_emulate_ctxt *ctxt)
2737 {
2738 	/* Write back the register source. */
2739 	ctxt->src.val = ctxt->dst.val;
2740 	write_register_operand(&ctxt->src);
2741 
2742 	/* Write back the memory destination with implicit LOCK prefix. */
2743 	ctxt->dst.val = ctxt->src.orig_val;
2744 	ctxt->lock_prefix = 1;
2745 	return X86EMUL_CONTINUE;
2746 }
2747 
2748 static int em_imul(struct x86_emulate_ctxt *ctxt)
2749 {
2750 	emulate_2op_SrcV_nobyte(ctxt, "imul");
2751 	return X86EMUL_CONTINUE;
2752 }
2753 
2754 static int em_imul_3op(struct x86_emulate_ctxt *ctxt)
2755 {
2756 	ctxt->dst.val = ctxt->src2.val;
2757 	return em_imul(ctxt);
2758 }
2759 
2760 static int em_cwd(struct x86_emulate_ctxt *ctxt)
2761 {
2762 	ctxt->dst.type = OP_REG;
2763 	ctxt->dst.bytes = ctxt->src.bytes;
2764 	ctxt->dst.addr.reg = &ctxt->regs[VCPU_REGS_RDX];
2765 	ctxt->dst.val = ~((ctxt->src.val >> (ctxt->src.bytes * 8 - 1)) - 1);
2766 
2767 	return X86EMUL_CONTINUE;
2768 }
2769 
2770 static int em_rdtsc(struct x86_emulate_ctxt *ctxt)
2771 {
2772 	u64 tsc = 0;
2773 
2774 	ctxt->ops->get_msr(ctxt, MSR_IA32_TSC, &tsc);
2775 	ctxt->regs[VCPU_REGS_RAX] = (u32)tsc;
2776 	ctxt->regs[VCPU_REGS_RDX] = tsc >> 32;
2777 	return X86EMUL_CONTINUE;
2778 }
2779 
2780 static int em_rdpmc(struct x86_emulate_ctxt *ctxt)
2781 {
2782 	u64 pmc;
2783 
2784 	if (ctxt->ops->read_pmc(ctxt, ctxt->regs[VCPU_REGS_RCX], &pmc))
2785 		return emulate_gp(ctxt, 0);
2786 	ctxt->regs[VCPU_REGS_RAX] = (u32)pmc;
2787 	ctxt->regs[VCPU_REGS_RDX] = pmc >> 32;
2788 	return X86EMUL_CONTINUE;
2789 }
2790 
2791 static int em_mov(struct x86_emulate_ctxt *ctxt)
2792 {
2793 	ctxt->dst.val = ctxt->src.val;
2794 	return X86EMUL_CONTINUE;
2795 }
2796 
2797 static int em_cr_write(struct x86_emulate_ctxt *ctxt)
2798 {
2799 	if (ctxt->ops->set_cr(ctxt, ctxt->modrm_reg, ctxt->src.val))
2800 		return emulate_gp(ctxt, 0);
2801 
2802 	/* Disable writeback. */
2803 	ctxt->dst.type = OP_NONE;
2804 	return X86EMUL_CONTINUE;
2805 }
2806 
2807 static int em_dr_write(struct x86_emulate_ctxt *ctxt)
2808 {
2809 	unsigned long val;
2810 
2811 	if (ctxt->mode == X86EMUL_MODE_PROT64)
2812 		val = ctxt->src.val & ~0ULL;
2813 	else
2814 		val = ctxt->src.val & ~0U;
2815 
2816 	/* #UD condition is already handled. */
2817 	if (ctxt->ops->set_dr(ctxt, ctxt->modrm_reg, val) < 0)
2818 		return emulate_gp(ctxt, 0);
2819 
2820 	/* Disable writeback. */
2821 	ctxt->dst.type = OP_NONE;
2822 	return X86EMUL_CONTINUE;
2823 }
2824 
2825 static int em_wrmsr(struct x86_emulate_ctxt *ctxt)
2826 {
2827 	u64 msr_data;
2828 
2829 	msr_data = (u32)ctxt->regs[VCPU_REGS_RAX]
2830 		| ((u64)ctxt->regs[VCPU_REGS_RDX] << 32);
2831 	if (ctxt->ops->set_msr(ctxt, ctxt->regs[VCPU_REGS_RCX], msr_data))
2832 		return emulate_gp(ctxt, 0);
2833 
2834 	return X86EMUL_CONTINUE;
2835 }
2836 
2837 static int em_rdmsr(struct x86_emulate_ctxt *ctxt)
2838 {
2839 	u64 msr_data;
2840 
2841 	if (ctxt->ops->get_msr(ctxt, ctxt->regs[VCPU_REGS_RCX], &msr_data))
2842 		return emulate_gp(ctxt, 0);
2843 
2844 	ctxt->regs[VCPU_REGS_RAX] = (u32)msr_data;
2845 	ctxt->regs[VCPU_REGS_RDX] = msr_data >> 32;
2846 	return X86EMUL_CONTINUE;
2847 }
2848 
2849 static int em_mov_rm_sreg(struct x86_emulate_ctxt *ctxt)
2850 {
2851 	if (ctxt->modrm_reg > VCPU_SREG_GS)
2852 		return emulate_ud(ctxt);
2853 
2854 	ctxt->dst.val = get_segment_selector(ctxt, ctxt->modrm_reg);
2855 	return X86EMUL_CONTINUE;
2856 }
2857 
2858 static int em_mov_sreg_rm(struct x86_emulate_ctxt *ctxt)
2859 {
2860 	u16 sel = ctxt->src.val;
2861 
2862 	if (ctxt->modrm_reg == VCPU_SREG_CS || ctxt->modrm_reg > VCPU_SREG_GS)
2863 		return emulate_ud(ctxt);
2864 
2865 	if (ctxt->modrm_reg == VCPU_SREG_SS)
2866 		ctxt->interruptibility = KVM_X86_SHADOW_INT_MOV_SS;
2867 
2868 	/* Disable writeback. */
2869 	ctxt->dst.type = OP_NONE;
2870 	return load_segment_descriptor(ctxt, sel, ctxt->modrm_reg);
2871 }
2872 
2873 static int em_movdqu(struct x86_emulate_ctxt *ctxt)
2874 {
2875 	memcpy(&ctxt->dst.vec_val, &ctxt->src.vec_val, ctxt->op_bytes);
2876 	return X86EMUL_CONTINUE;
2877 }
2878 
2879 static int em_invlpg(struct x86_emulate_ctxt *ctxt)
2880 {
2881 	int rc;
2882 	ulong linear;
2883 
2884 	rc = linearize(ctxt, ctxt->src.addr.mem, 1, false, &linear);
2885 	if (rc == X86EMUL_CONTINUE)
2886 		ctxt->ops->invlpg(ctxt, linear);
2887 	/* Disable writeback. */
2888 	ctxt->dst.type = OP_NONE;
2889 	return X86EMUL_CONTINUE;
2890 }
2891 
2892 static int em_clts(struct x86_emulate_ctxt *ctxt)
2893 {
2894 	ulong cr0;
2895 
2896 	cr0 = ctxt->ops->get_cr(ctxt, 0);
2897 	cr0 &= ~X86_CR0_TS;
2898 	ctxt->ops->set_cr(ctxt, 0, cr0);
2899 	return X86EMUL_CONTINUE;
2900 }
2901 
2902 static int em_vmcall(struct x86_emulate_ctxt *ctxt)
2903 {
2904 	int rc;
2905 
2906 	if (ctxt->modrm_mod != 3 || ctxt->modrm_rm != 1)
2907 		return X86EMUL_UNHANDLEABLE;
2908 
2909 	rc = ctxt->ops->fix_hypercall(ctxt);
2910 	if (rc != X86EMUL_CONTINUE)
2911 		return rc;
2912 
2913 	/* Let the processor re-execute the fixed hypercall */
2914 	ctxt->_eip = ctxt->eip;
2915 	/* Disable writeback. */
2916 	ctxt->dst.type = OP_NONE;
2917 	return X86EMUL_CONTINUE;
2918 }
2919 
2920 static int em_lgdt(struct x86_emulate_ctxt *ctxt)
2921 {
2922 	struct desc_ptr desc_ptr;
2923 	int rc;
2924 
2925 	rc = read_descriptor(ctxt, ctxt->src.addr.mem,
2926 			     &desc_ptr.size, &desc_ptr.address,
2927 			     ctxt->op_bytes);
2928 	if (rc != X86EMUL_CONTINUE)
2929 		return rc;
2930 	ctxt->ops->set_gdt(ctxt, &desc_ptr);
2931 	/* Disable writeback. */
2932 	ctxt->dst.type = OP_NONE;
2933 	return X86EMUL_CONTINUE;
2934 }
2935 
2936 static int em_vmmcall(struct x86_emulate_ctxt *ctxt)
2937 {
2938 	int rc;
2939 
2940 	rc = ctxt->ops->fix_hypercall(ctxt);
2941 
2942 	/* Disable writeback. */
2943 	ctxt->dst.type = OP_NONE;
2944 	return rc;
2945 }
2946 
2947 static int em_lidt(struct x86_emulate_ctxt *ctxt)
2948 {
2949 	struct desc_ptr desc_ptr;
2950 	int rc;
2951 
2952 	rc = read_descriptor(ctxt, ctxt->src.addr.mem,
2953 			     &desc_ptr.size, &desc_ptr.address,
2954 			     ctxt->op_bytes);
2955 	if (rc != X86EMUL_CONTINUE)
2956 		return rc;
2957 	ctxt->ops->set_idt(ctxt, &desc_ptr);
2958 	/* Disable writeback. */
2959 	ctxt->dst.type = OP_NONE;
2960 	return X86EMUL_CONTINUE;
2961 }
2962 
2963 static int em_smsw(struct x86_emulate_ctxt *ctxt)
2964 {
2965 	ctxt->dst.bytes = 2;
2966 	ctxt->dst.val = ctxt->ops->get_cr(ctxt, 0);
2967 	return X86EMUL_CONTINUE;
2968 }
2969 
2970 static int em_lmsw(struct x86_emulate_ctxt *ctxt)
2971 {
2972 	ctxt->ops->set_cr(ctxt, 0, (ctxt->ops->get_cr(ctxt, 0) & ~0x0eul)
2973 			  | (ctxt->src.val & 0x0f));
2974 	ctxt->dst.type = OP_NONE;
2975 	return X86EMUL_CONTINUE;
2976 }
2977 
2978 static int em_loop(struct x86_emulate_ctxt *ctxt)
2979 {
2980 	register_address_increment(ctxt, &ctxt->regs[VCPU_REGS_RCX], -1);
2981 	if ((address_mask(ctxt, ctxt->regs[VCPU_REGS_RCX]) != 0) &&
2982 	    (ctxt->b == 0xe2 || test_cc(ctxt->b ^ 0x5, ctxt->eflags)))
2983 		jmp_rel(ctxt, ctxt->src.val);
2984 
2985 	return X86EMUL_CONTINUE;
2986 }
2987 
2988 static int em_jcxz(struct x86_emulate_ctxt *ctxt)
2989 {
2990 	if (address_mask(ctxt, ctxt->regs[VCPU_REGS_RCX]) == 0)
2991 		jmp_rel(ctxt, ctxt->src.val);
2992 
2993 	return X86EMUL_CONTINUE;
2994 }
2995 
2996 static int em_in(struct x86_emulate_ctxt *ctxt)
2997 {
2998 	if (!pio_in_emulated(ctxt, ctxt->dst.bytes, ctxt->src.val,
2999 			     &ctxt->dst.val))
3000 		return X86EMUL_IO_NEEDED;
3001 
3002 	return X86EMUL_CONTINUE;
3003 }
3004 
3005 static int em_out(struct x86_emulate_ctxt *ctxt)
3006 {
3007 	ctxt->ops->pio_out_emulated(ctxt, ctxt->src.bytes, ctxt->dst.val,
3008 				    &ctxt->src.val, 1);
3009 	/* Disable writeback. */
3010 	ctxt->dst.type = OP_NONE;
3011 	return X86EMUL_CONTINUE;
3012 }
3013 
3014 static int em_cli(struct x86_emulate_ctxt *ctxt)
3015 {
3016 	if (emulator_bad_iopl(ctxt))
3017 		return emulate_gp(ctxt, 0);
3018 
3019 	ctxt->eflags &= ~X86_EFLAGS_IF;
3020 	return X86EMUL_CONTINUE;
3021 }
3022 
3023 static int em_sti(struct x86_emulate_ctxt *ctxt)
3024 {
3025 	if (emulator_bad_iopl(ctxt))
3026 		return emulate_gp(ctxt, 0);
3027 
3028 	ctxt->interruptibility = KVM_X86_SHADOW_INT_STI;
3029 	ctxt->eflags |= X86_EFLAGS_IF;
3030 	return X86EMUL_CONTINUE;
3031 }
3032 
3033 static int em_bt(struct x86_emulate_ctxt *ctxt)
3034 {
3035 	/* Disable writeback. */
3036 	ctxt->dst.type = OP_NONE;
3037 	/* only subword offset */
3038 	ctxt->src.val &= (ctxt->dst.bytes << 3) - 1;
3039 
3040 	emulate_2op_SrcV_nobyte(ctxt, "bt");
3041 	return X86EMUL_CONTINUE;
3042 }
3043 
3044 static int em_bts(struct x86_emulate_ctxt *ctxt)
3045 {
3046 	emulate_2op_SrcV_nobyte(ctxt, "bts");
3047 	return X86EMUL_CONTINUE;
3048 }
3049 
3050 static int em_btr(struct x86_emulate_ctxt *ctxt)
3051 {
3052 	emulate_2op_SrcV_nobyte(ctxt, "btr");
3053 	return X86EMUL_CONTINUE;
3054 }
3055 
3056 static int em_btc(struct x86_emulate_ctxt *ctxt)
3057 {
3058 	emulate_2op_SrcV_nobyte(ctxt, "btc");
3059 	return X86EMUL_CONTINUE;
3060 }
3061 
3062 static int em_bsf(struct x86_emulate_ctxt *ctxt)
3063 {
3064 	u8 zf;
3065 
3066 	__asm__ ("bsf %2, %0; setz %1"
3067 		 : "=r"(ctxt->dst.val), "=q"(zf)
3068 		 : "r"(ctxt->src.val));
3069 
3070 	ctxt->eflags &= ~X86_EFLAGS_ZF;
3071 	if (zf) {
3072 		ctxt->eflags |= X86_EFLAGS_ZF;
3073 		/* Disable writeback. */
3074 		ctxt->dst.type = OP_NONE;
3075 	}
3076 	return X86EMUL_CONTINUE;
3077 }
3078 
3079 static int em_bsr(struct x86_emulate_ctxt *ctxt)
3080 {
3081 	u8 zf;
3082 
3083 	__asm__ ("bsr %2, %0; setz %1"
3084 		 : "=r"(ctxt->dst.val), "=q"(zf)
3085 		 : "r"(ctxt->src.val));
3086 
3087 	ctxt->eflags &= ~X86_EFLAGS_ZF;
3088 	if (zf) {
3089 		ctxt->eflags |= X86_EFLAGS_ZF;
3090 		/* Disable writeback. */
3091 		ctxt->dst.type = OP_NONE;
3092 	}
3093 	return X86EMUL_CONTINUE;
3094 }
3095 
3096 static bool valid_cr(int nr)
3097 {
3098 	switch (nr) {
3099 	case 0:
3100 	case 2 ... 4:
3101 	case 8:
3102 		return true;
3103 	default:
3104 		return false;
3105 	}
3106 }
3107 
3108 static int check_cr_read(struct x86_emulate_ctxt *ctxt)
3109 {
3110 	if (!valid_cr(ctxt->modrm_reg))
3111 		return emulate_ud(ctxt);
3112 
3113 	return X86EMUL_CONTINUE;
3114 }
3115 
3116 static int check_cr_write(struct x86_emulate_ctxt *ctxt)
3117 {
3118 	u64 new_val = ctxt->src.val64;
3119 	int cr = ctxt->modrm_reg;
3120 	u64 efer = 0;
3121 
3122 	static u64 cr_reserved_bits[] = {
3123 		0xffffffff00000000ULL,
3124 		0, 0, 0, /* CR3 checked later */
3125 		CR4_RESERVED_BITS,
3126 		0, 0, 0,
3127 		CR8_RESERVED_BITS,
3128 	};
3129 
3130 	if (!valid_cr(cr))
3131 		return emulate_ud(ctxt);
3132 
3133 	if (new_val & cr_reserved_bits[cr])
3134 		return emulate_gp(ctxt, 0);
3135 
3136 	switch (cr) {
3137 	case 0: {
3138 		u64 cr4;
3139 		if (((new_val & X86_CR0_PG) && !(new_val & X86_CR0_PE)) ||
3140 		    ((new_val & X86_CR0_NW) && !(new_val & X86_CR0_CD)))
3141 			return emulate_gp(ctxt, 0);
3142 
3143 		cr4 = ctxt->ops->get_cr(ctxt, 4);
3144 		ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3145 
3146 		if ((new_val & X86_CR0_PG) && (efer & EFER_LME) &&
3147 		    !(cr4 & X86_CR4_PAE))
3148 			return emulate_gp(ctxt, 0);
3149 
3150 		break;
3151 		}
3152 	case 3: {
3153 		u64 rsvd = 0;
3154 
3155 		ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3156 		if (efer & EFER_LMA)
3157 			rsvd = CR3_L_MODE_RESERVED_BITS;
3158 		else if (ctxt->ops->get_cr(ctxt, 4) & X86_CR4_PAE)
3159 			rsvd = CR3_PAE_RESERVED_BITS;
3160 		else if (ctxt->ops->get_cr(ctxt, 0) & X86_CR0_PG)
3161 			rsvd = CR3_NONPAE_RESERVED_BITS;
3162 
3163 		if (new_val & rsvd)
3164 			return emulate_gp(ctxt, 0);
3165 
3166 		break;
3167 		}
3168 	case 4: {
3169 		ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3170 
3171 		if ((efer & EFER_LMA) && !(new_val & X86_CR4_PAE))
3172 			return emulate_gp(ctxt, 0);
3173 
3174 		break;
3175 		}
3176 	}
3177 
3178 	return X86EMUL_CONTINUE;
3179 }
3180 
3181 static int check_dr7_gd(struct x86_emulate_ctxt *ctxt)
3182 {
3183 	unsigned long dr7;
3184 
3185 	ctxt->ops->get_dr(ctxt, 7, &dr7);
3186 
3187 	/* Check if DR7.Global_Enable is set */
3188 	return dr7 & (1 << 13);
3189 }
3190 
3191 static int check_dr_read(struct x86_emulate_ctxt *ctxt)
3192 {
3193 	int dr = ctxt->modrm_reg;
3194 	u64 cr4;
3195 
3196 	if (dr > 7)
3197 		return emulate_ud(ctxt);
3198 
3199 	cr4 = ctxt->ops->get_cr(ctxt, 4);
3200 	if ((cr4 & X86_CR4_DE) && (dr == 4 || dr == 5))
3201 		return emulate_ud(ctxt);
3202 
3203 	if (check_dr7_gd(ctxt))
3204 		return emulate_db(ctxt);
3205 
3206 	return X86EMUL_CONTINUE;
3207 }
3208 
3209 static int check_dr_write(struct x86_emulate_ctxt *ctxt)
3210 {
3211 	u64 new_val = ctxt->src.val64;
3212 	int dr = ctxt->modrm_reg;
3213 
3214 	if ((dr == 6 || dr == 7) && (new_val & 0xffffffff00000000ULL))
3215 		return emulate_gp(ctxt, 0);
3216 
3217 	return check_dr_read(ctxt);
3218 }
3219 
3220 static int check_svme(struct x86_emulate_ctxt *ctxt)
3221 {
3222 	u64 efer;
3223 
3224 	ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3225 
3226 	if (!(efer & EFER_SVME))
3227 		return emulate_ud(ctxt);
3228 
3229 	return X86EMUL_CONTINUE;
3230 }
3231 
3232 static int check_svme_pa(struct x86_emulate_ctxt *ctxt)
3233 {
3234 	u64 rax = ctxt->regs[VCPU_REGS_RAX];
3235 
3236 	/* Valid physical address? */
3237 	if (rax & 0xffff000000000000ULL)
3238 		return emulate_gp(ctxt, 0);
3239 
3240 	return check_svme(ctxt);
3241 }
3242 
3243 static int check_rdtsc(struct x86_emulate_ctxt *ctxt)
3244 {
3245 	u64 cr4 = ctxt->ops->get_cr(ctxt, 4);
3246 
3247 	if (cr4 & X86_CR4_TSD && ctxt->ops->cpl(ctxt))
3248 		return emulate_ud(ctxt);
3249 
3250 	return X86EMUL_CONTINUE;
3251 }
3252 
3253 static int check_rdpmc(struct x86_emulate_ctxt *ctxt)
3254 {
3255 	u64 cr4 = ctxt->ops->get_cr(ctxt, 4);
3256 	u64 rcx = ctxt->regs[VCPU_REGS_RCX];
3257 
3258 	if ((!(cr4 & X86_CR4_PCE) && ctxt->ops->cpl(ctxt)) ||
3259 	    (rcx > 3))
3260 		return emulate_gp(ctxt, 0);
3261 
3262 	return X86EMUL_CONTINUE;
3263 }
3264 
3265 static int check_perm_in(struct x86_emulate_ctxt *ctxt)
3266 {
3267 	ctxt->dst.bytes = min(ctxt->dst.bytes, 4u);
3268 	if (!emulator_io_permited(ctxt, ctxt->src.val, ctxt->dst.bytes))
3269 		return emulate_gp(ctxt, 0);
3270 
3271 	return X86EMUL_CONTINUE;
3272 }
3273 
3274 static int check_perm_out(struct x86_emulate_ctxt *ctxt)
3275 {
3276 	ctxt->src.bytes = min(ctxt->src.bytes, 4u);
3277 	if (!emulator_io_permited(ctxt, ctxt->dst.val, ctxt->src.bytes))
3278 		return emulate_gp(ctxt, 0);
3279 
3280 	return X86EMUL_CONTINUE;
3281 }
3282 
3283 #define D(_y) { .flags = (_y) }
3284 #define DI(_y, _i) { .flags = (_y), .intercept = x86_intercept_##_i }
3285 #define DIP(_y, _i, _p) { .flags = (_y), .intercept = x86_intercept_##_i, \
3286 		      .check_perm = (_p) }
3287 #define N    D(0)
3288 #define EXT(_f, _e) { .flags = ((_f) | RMExt), .u.group = (_e) }
3289 #define G(_f, _g) { .flags = ((_f) | Group), .u.group = (_g) }
3290 #define GD(_f, _g) { .flags = ((_f) | GroupDual), .u.gdual = (_g) }
3291 #define I(_f, _e) { .flags = (_f), .u.execute = (_e) }
3292 #define II(_f, _e, _i) \
3293 	{ .flags = (_f), .u.execute = (_e), .intercept = x86_intercept_##_i }
3294 #define IIP(_f, _e, _i, _p) \
3295 	{ .flags = (_f), .u.execute = (_e), .intercept = x86_intercept_##_i, \
3296 	  .check_perm = (_p) }
3297 #define GP(_f, _g) { .flags = ((_f) | Prefix), .u.gprefix = (_g) }
3298 
3299 #define D2bv(_f)      D((_f) | ByteOp), D(_f)
3300 #define D2bvIP(_f, _i, _p) DIP((_f) | ByteOp, _i, _p), DIP(_f, _i, _p)
3301 #define I2bv(_f, _e)  I((_f) | ByteOp, _e), I(_f, _e)
3302 #define I2bvIP(_f, _e, _i, _p) \
3303 	IIP((_f) | ByteOp, _e, _i, _p), IIP(_f, _e, _i, _p)
3304 
3305 #define I6ALU(_f, _e) I2bv((_f) | DstMem | SrcReg | ModRM, _e),		\
3306 		I2bv(((_f) | DstReg | SrcMem | ModRM) & ~Lock, _e),	\
3307 		I2bv(((_f) & ~Lock) | DstAcc | SrcImm, _e)
3308 
3309 static struct opcode group7_rm1[] = {
3310 	DI(SrcNone | ModRM | Priv, monitor),
3311 	DI(SrcNone | ModRM | Priv, mwait),
3312 	N, N, N, N, N, N,
3313 };
3314 
3315 static struct opcode group7_rm3[] = {
3316 	DIP(SrcNone | ModRM | Prot | Priv, vmrun,   check_svme_pa),
3317 	II(SrcNone | ModRM | Prot | VendorSpecific, em_vmmcall, vmmcall),
3318 	DIP(SrcNone | ModRM | Prot | Priv, vmload,  check_svme_pa),
3319 	DIP(SrcNone | ModRM | Prot | Priv, vmsave,  check_svme_pa),
3320 	DIP(SrcNone | ModRM | Prot | Priv, stgi,    check_svme),
3321 	DIP(SrcNone | ModRM | Prot | Priv, clgi,    check_svme),
3322 	DIP(SrcNone | ModRM | Prot | Priv, skinit,  check_svme),
3323 	DIP(SrcNone | ModRM | Prot | Priv, invlpga, check_svme),
3324 };
3325 
3326 static struct opcode group7_rm7[] = {
3327 	N,
3328 	DIP(SrcNone | ModRM, rdtscp, check_rdtsc),
3329 	N, N, N, N, N, N,
3330 };
3331 
3332 static struct opcode group1[] = {
3333 	I(Lock, em_add),
3334 	I(Lock | PageTable, em_or),
3335 	I(Lock, em_adc),
3336 	I(Lock, em_sbb),
3337 	I(Lock | PageTable, em_and),
3338 	I(Lock, em_sub),
3339 	I(Lock, em_xor),
3340 	I(0, em_cmp),
3341 };
3342 
3343 static struct opcode group1A[] = {
3344 	I(DstMem | SrcNone | ModRM | Mov | Stack, em_pop), N, N, N, N, N, N, N,
3345 };
3346 
3347 static struct opcode group3[] = {
3348 	I(DstMem | SrcImm | ModRM, em_test),
3349 	I(DstMem | SrcImm | ModRM, em_test),
3350 	I(DstMem | SrcNone | ModRM | Lock, em_not),
3351 	I(DstMem | SrcNone | ModRM | Lock, em_neg),
3352 	I(SrcMem | ModRM, em_mul_ex),
3353 	I(SrcMem | ModRM, em_imul_ex),
3354 	I(SrcMem | ModRM, em_div_ex),
3355 	I(SrcMem | ModRM, em_idiv_ex),
3356 };
3357 
3358 static struct opcode group4[] = {
3359 	I(ByteOp | DstMem | SrcNone | ModRM | Lock, em_grp45),
3360 	I(ByteOp | DstMem | SrcNone | ModRM | Lock, em_grp45),
3361 	N, N, N, N, N, N,
3362 };
3363 
3364 static struct opcode group5[] = {
3365 	I(DstMem | SrcNone | ModRM | Lock, em_grp45),
3366 	I(DstMem | SrcNone | ModRM | Lock, em_grp45),
3367 	I(SrcMem | ModRM | Stack, em_grp45),
3368 	I(SrcMemFAddr | ModRM | ImplicitOps | Stack, em_call_far),
3369 	I(SrcMem | ModRM | Stack, em_grp45),
3370 	I(SrcMemFAddr | ModRM | ImplicitOps, em_grp45),
3371 	I(SrcMem | ModRM | Stack, em_grp45), N,
3372 };
3373 
3374 static struct opcode group6[] = {
3375 	DI(ModRM | Prot,        sldt),
3376 	DI(ModRM | Prot,        str),
3377 	DI(ModRM | Prot | Priv, lldt),
3378 	DI(ModRM | Prot | Priv, ltr),
3379 	N, N, N, N,
3380 };
3381 
3382 static struct group_dual group7 = { {
3383 	DI(ModRM | Mov | DstMem | Priv, sgdt),
3384 	DI(ModRM | Mov | DstMem | Priv, sidt),
3385 	II(ModRM | SrcMem | Priv, em_lgdt, lgdt),
3386 	II(ModRM | SrcMem | Priv, em_lidt, lidt),
3387 	II(SrcNone | ModRM | DstMem | Mov, em_smsw, smsw), N,
3388 	II(SrcMem16 | ModRM | Mov | Priv, em_lmsw, lmsw),
3389 	II(SrcMem | ModRM | ByteOp | Priv | NoAccess, em_invlpg, invlpg),
3390 }, {
3391 	I(SrcNone | ModRM | Priv | VendorSpecific, em_vmcall),
3392 	EXT(0, group7_rm1),
3393 	N, EXT(0, group7_rm3),
3394 	II(SrcNone | ModRM | DstMem | Mov, em_smsw, smsw), N,
3395 	II(SrcMem16 | ModRM | Mov | Priv, em_lmsw, lmsw), EXT(0, group7_rm7),
3396 } };
3397 
3398 static struct opcode group8[] = {
3399 	N, N, N, N,
3400 	I(DstMem | SrcImmByte | ModRM, em_bt),
3401 	I(DstMem | SrcImmByte | ModRM | Lock | PageTable, em_bts),
3402 	I(DstMem | SrcImmByte | ModRM | Lock, em_btr),
3403 	I(DstMem | SrcImmByte | ModRM | Lock | PageTable, em_btc),
3404 };
3405 
3406 static struct group_dual group9 = { {
3407 	N, I(DstMem64 | ModRM | Lock | PageTable, em_cmpxchg8b), N, N, N, N, N, N,
3408 }, {
3409 	N, N, N, N, N, N, N, N,
3410 } };
3411 
3412 static struct opcode group11[] = {
3413 	I(DstMem | SrcImm | ModRM | Mov | PageTable, em_mov),
3414 	X7(D(Undefined)),
3415 };
3416 
3417 static struct gprefix pfx_0f_6f_0f_7f = {
3418 	N, N, N, I(Sse, em_movdqu),
3419 };
3420 
3421 static struct opcode opcode_table[256] = {
3422 	/* 0x00 - 0x07 */
3423 	I6ALU(Lock, em_add),
3424 	I(ImplicitOps | Stack | No64 | Src2ES, em_push_sreg),
3425 	I(ImplicitOps | Stack | No64 | Src2ES, em_pop_sreg),
3426 	/* 0x08 - 0x0F */
3427 	I6ALU(Lock | PageTable, em_or),
3428 	I(ImplicitOps | Stack | No64 | Src2CS, em_push_sreg),
3429 	N,
3430 	/* 0x10 - 0x17 */
3431 	I6ALU(Lock, em_adc),
3432 	I(ImplicitOps | Stack | No64 | Src2SS, em_push_sreg),
3433 	I(ImplicitOps | Stack | No64 | Src2SS, em_pop_sreg),
3434 	/* 0x18 - 0x1F */
3435 	I6ALU(Lock, em_sbb),
3436 	I(ImplicitOps | Stack | No64 | Src2DS, em_push_sreg),
3437 	I(ImplicitOps | Stack | No64 | Src2DS, em_pop_sreg),
3438 	/* 0x20 - 0x27 */
3439 	I6ALU(Lock | PageTable, em_and), N, N,
3440 	/* 0x28 - 0x2F */
3441 	I6ALU(Lock, em_sub), N, I(ByteOp | DstAcc | No64, em_das),
3442 	/* 0x30 - 0x37 */
3443 	I6ALU(Lock, em_xor), N, N,
3444 	/* 0x38 - 0x3F */
3445 	I6ALU(0, em_cmp), N, N,
3446 	/* 0x40 - 0x4F */
3447 	X16(D(DstReg)),
3448 	/* 0x50 - 0x57 */
3449 	X8(I(SrcReg | Stack, em_push)),
3450 	/* 0x58 - 0x5F */
3451 	X8(I(DstReg | Stack, em_pop)),
3452 	/* 0x60 - 0x67 */
3453 	I(ImplicitOps | Stack | No64, em_pusha),
3454 	I(ImplicitOps | Stack | No64, em_popa),
3455 	N, D(DstReg | SrcMem32 | ModRM | Mov) /* movsxd (x86/64) */ ,
3456 	N, N, N, N,
3457 	/* 0x68 - 0x6F */
3458 	I(SrcImm | Mov | Stack, em_push),
3459 	I(DstReg | SrcMem | ModRM | Src2Imm, em_imul_3op),
3460 	I(SrcImmByte | Mov | Stack, em_push),
3461 	I(DstReg | SrcMem | ModRM | Src2ImmByte, em_imul_3op),
3462 	I2bvIP(DstDI | SrcDX | Mov | String, em_in, ins, check_perm_in), /* insb, insw/insd */
3463 	I2bvIP(SrcSI | DstDX | String, em_out, outs, check_perm_out), /* outsb, outsw/outsd */
3464 	/* 0x70 - 0x7F */
3465 	X16(D(SrcImmByte)),
3466 	/* 0x80 - 0x87 */
3467 	G(ByteOp | DstMem | SrcImm | ModRM | Group, group1),
3468 	G(DstMem | SrcImm | ModRM | Group, group1),
3469 	G(ByteOp | DstMem | SrcImm | ModRM | No64 | Group, group1),
3470 	G(DstMem | SrcImmByte | ModRM | Group, group1),
3471 	I2bv(DstMem | SrcReg | ModRM, em_test),
3472 	I2bv(DstMem | SrcReg | ModRM | Lock | PageTable, em_xchg),
3473 	/* 0x88 - 0x8F */
3474 	I2bv(DstMem | SrcReg | ModRM | Mov | PageTable, em_mov),
3475 	I2bv(DstReg | SrcMem | ModRM | Mov, em_mov),
3476 	I(DstMem | SrcNone | ModRM | Mov | PageTable, em_mov_rm_sreg),
3477 	D(ModRM | SrcMem | NoAccess | DstReg),
3478 	I(ImplicitOps | SrcMem16 | ModRM, em_mov_sreg_rm),
3479 	G(0, group1A),
3480 	/* 0x90 - 0x97 */
3481 	DI(SrcAcc | DstReg, pause), X7(D(SrcAcc | DstReg)),
3482 	/* 0x98 - 0x9F */
3483 	D(DstAcc | SrcNone), I(ImplicitOps | SrcAcc, em_cwd),
3484 	I(SrcImmFAddr | No64, em_call_far), N,
3485 	II(ImplicitOps | Stack, em_pushf, pushf),
3486 	II(ImplicitOps | Stack, em_popf, popf), N, N,
3487 	/* 0xA0 - 0xA7 */
3488 	I2bv(DstAcc | SrcMem | Mov | MemAbs, em_mov),
3489 	I2bv(DstMem | SrcAcc | Mov | MemAbs | PageTable, em_mov),
3490 	I2bv(SrcSI | DstDI | Mov | String, em_mov),
3491 	I2bv(SrcSI | DstDI | String, em_cmp),
3492 	/* 0xA8 - 0xAF */
3493 	I2bv(DstAcc | SrcImm, em_test),
3494 	I2bv(SrcAcc | DstDI | Mov | String, em_mov),
3495 	I2bv(SrcSI | DstAcc | Mov | String, em_mov),
3496 	I2bv(SrcAcc | DstDI | String, em_cmp),
3497 	/* 0xB0 - 0xB7 */
3498 	X8(I(ByteOp | DstReg | SrcImm | Mov, em_mov)),
3499 	/* 0xB8 - 0xBF */
3500 	X8(I(DstReg | SrcImm | Mov, em_mov)),
3501 	/* 0xC0 - 0xC7 */
3502 	D2bv(DstMem | SrcImmByte | ModRM),
3503 	I(ImplicitOps | Stack | SrcImmU16, em_ret_near_imm),
3504 	I(ImplicitOps | Stack, em_ret),
3505 	I(DstReg | SrcMemFAddr | ModRM | No64 | Src2ES, em_lseg),
3506 	I(DstReg | SrcMemFAddr | ModRM | No64 | Src2DS, em_lseg),
3507 	G(ByteOp, group11), G(0, group11),
3508 	/* 0xC8 - 0xCF */
3509 	N, N, N, I(ImplicitOps | Stack, em_ret_far),
3510 	D(ImplicitOps), DI(SrcImmByte, intn),
3511 	D(ImplicitOps | No64), II(ImplicitOps, em_iret, iret),
3512 	/* 0xD0 - 0xD7 */
3513 	D2bv(DstMem | SrcOne | ModRM), D2bv(DstMem | ModRM),
3514 	N, N, N, N,
3515 	/* 0xD8 - 0xDF */
3516 	N, N, N, N, N, N, N, N,
3517 	/* 0xE0 - 0xE7 */
3518 	X3(I(SrcImmByte, em_loop)),
3519 	I(SrcImmByte, em_jcxz),
3520 	I2bvIP(SrcImmUByte | DstAcc, em_in,  in,  check_perm_in),
3521 	I2bvIP(SrcAcc | DstImmUByte, em_out, out, check_perm_out),
3522 	/* 0xE8 - 0xEF */
3523 	I(SrcImm | Stack, em_call), D(SrcImm | ImplicitOps),
3524 	I(SrcImmFAddr | No64, em_jmp_far), D(SrcImmByte | ImplicitOps),
3525 	I2bvIP(SrcDX | DstAcc, em_in,  in,  check_perm_in),
3526 	I2bvIP(SrcAcc | DstDX, em_out, out, check_perm_out),
3527 	/* 0xF0 - 0xF7 */
3528 	N, DI(ImplicitOps, icebp), N, N,
3529 	DI(ImplicitOps | Priv, hlt), D(ImplicitOps),
3530 	G(ByteOp, group3), G(0, group3),
3531 	/* 0xF8 - 0xFF */
3532 	D(ImplicitOps), D(ImplicitOps),
3533 	I(ImplicitOps, em_cli), I(ImplicitOps, em_sti),
3534 	D(ImplicitOps), D(ImplicitOps), G(0, group4), G(0, group5),
3535 };
3536 
3537 static struct opcode twobyte_table[256] = {
3538 	/* 0x00 - 0x0F */
3539 	G(0, group6), GD(0, &group7), N, N,
3540 	N, I(ImplicitOps | VendorSpecific, em_syscall),
3541 	II(ImplicitOps | Priv, em_clts, clts), N,
3542 	DI(ImplicitOps | Priv, invd), DI(ImplicitOps | Priv, wbinvd), N, N,
3543 	N, D(ImplicitOps | ModRM), N, N,
3544 	/* 0x10 - 0x1F */
3545 	N, N, N, N, N, N, N, N, D(ImplicitOps | ModRM), N, N, N, N, N, N, N,
3546 	/* 0x20 - 0x2F */
3547 	DIP(ModRM | DstMem | Priv | Op3264, cr_read, check_cr_read),
3548 	DIP(ModRM | DstMem | Priv | Op3264, dr_read, check_dr_read),
3549 	IIP(ModRM | SrcMem | Priv | Op3264, em_cr_write, cr_write, check_cr_write),
3550 	IIP(ModRM | SrcMem | Priv | Op3264, em_dr_write, dr_write, check_dr_write),
3551 	N, N, N, N,
3552 	N, N, N, N, N, N, N, N,
3553 	/* 0x30 - 0x3F */
3554 	II(ImplicitOps | Priv, em_wrmsr, wrmsr),
3555 	IIP(ImplicitOps, em_rdtsc, rdtsc, check_rdtsc),
3556 	II(ImplicitOps | Priv, em_rdmsr, rdmsr),
3557 	IIP(ImplicitOps, em_rdpmc, rdpmc, check_rdpmc),
3558 	I(ImplicitOps | VendorSpecific, em_sysenter),
3559 	I(ImplicitOps | Priv | VendorSpecific, em_sysexit),
3560 	N, N,
3561 	N, N, N, N, N, N, N, N,
3562 	/* 0x40 - 0x4F */
3563 	X16(D(DstReg | SrcMem | ModRM | Mov)),
3564 	/* 0x50 - 0x5F */
3565 	N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
3566 	/* 0x60 - 0x6F */
3567 	N, N, N, N,
3568 	N, N, N, N,
3569 	N, N, N, N,
3570 	N, N, N, GP(SrcMem | DstReg | ModRM | Mov, &pfx_0f_6f_0f_7f),
3571 	/* 0x70 - 0x7F */
3572 	N, N, N, N,
3573 	N, N, N, N,
3574 	N, N, N, N,
3575 	N, N, N, GP(SrcReg | DstMem | ModRM | Mov, &pfx_0f_6f_0f_7f),
3576 	/* 0x80 - 0x8F */
3577 	X16(D(SrcImm)),
3578 	/* 0x90 - 0x9F */
3579 	X16(D(ByteOp | DstMem | SrcNone | ModRM| Mov)),
3580 	/* 0xA0 - 0xA7 */
3581 	I(Stack | Src2FS, em_push_sreg), I(Stack | Src2FS, em_pop_sreg),
3582 	DI(ImplicitOps, cpuid), I(DstMem | SrcReg | ModRM | BitOp, em_bt),
3583 	D(DstMem | SrcReg | Src2ImmByte | ModRM),
3584 	D(DstMem | SrcReg | Src2CL | ModRM), N, N,
3585 	/* 0xA8 - 0xAF */
3586 	I(Stack | Src2GS, em_push_sreg), I(Stack | Src2GS, em_pop_sreg),
3587 	DI(ImplicitOps, rsm),
3588 	I(DstMem | SrcReg | ModRM | BitOp | Lock | PageTable, em_bts),
3589 	D(DstMem | SrcReg | Src2ImmByte | ModRM),
3590 	D(DstMem | SrcReg | Src2CL | ModRM),
3591 	D(ModRM), I(DstReg | SrcMem | ModRM, em_imul),
3592 	/* 0xB0 - 0xB7 */
3593 	I2bv(DstMem | SrcReg | ModRM | Lock | PageTable, em_cmpxchg),
3594 	I(DstReg | SrcMemFAddr | ModRM | Src2SS, em_lseg),
3595 	I(DstMem | SrcReg | ModRM | BitOp | Lock, em_btr),
3596 	I(DstReg | SrcMemFAddr | ModRM | Src2FS, em_lseg),
3597 	I(DstReg | SrcMemFAddr | ModRM | Src2GS, em_lseg),
3598 	D(DstReg | SrcMem8 | ModRM | Mov), D(DstReg | SrcMem16 | ModRM | Mov),
3599 	/* 0xB8 - 0xBF */
3600 	N, N,
3601 	G(BitOp, group8),
3602 	I(DstMem | SrcReg | ModRM | BitOp | Lock | PageTable, em_btc),
3603 	I(DstReg | SrcMem | ModRM, em_bsf), I(DstReg | SrcMem | ModRM, em_bsr),
3604 	D(DstReg | SrcMem8 | ModRM | Mov), D(DstReg | SrcMem16 | ModRM | Mov),
3605 	/* 0xC0 - 0xCF */
3606 	D2bv(DstMem | SrcReg | ModRM | Lock),
3607 	N, D(DstMem | SrcReg | ModRM | Mov),
3608 	N, N, N, GD(0, &group9),
3609 	N, N, N, N, N, N, N, N,
3610 	/* 0xD0 - 0xDF */
3611 	N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
3612 	/* 0xE0 - 0xEF */
3613 	N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
3614 	/* 0xF0 - 0xFF */
3615 	N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N
3616 };
3617 
3618 #undef D
3619 #undef N
3620 #undef G
3621 #undef GD
3622 #undef I
3623 #undef GP
3624 #undef EXT
3625 
3626 #undef D2bv
3627 #undef D2bvIP
3628 #undef I2bv
3629 #undef I2bvIP
3630 #undef I6ALU
3631 
3632 static unsigned imm_size(struct x86_emulate_ctxt *ctxt)
3633 {
3634 	unsigned size;
3635 
3636 	size = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
3637 	if (size == 8)
3638 		size = 4;
3639 	return size;
3640 }
3641 
3642 static int decode_imm(struct x86_emulate_ctxt *ctxt, struct operand *op,
3643 		      unsigned size, bool sign_extension)
3644 {
3645 	int rc = X86EMUL_CONTINUE;
3646 
3647 	op->type = OP_IMM;
3648 	op->bytes = size;
3649 	op->addr.mem.ea = ctxt->_eip;
3650 	/* NB. Immediates are sign-extended as necessary. */
3651 	switch (op->bytes) {
3652 	case 1:
3653 		op->val = insn_fetch(s8, ctxt);
3654 		break;
3655 	case 2:
3656 		op->val = insn_fetch(s16, ctxt);
3657 		break;
3658 	case 4:
3659 		op->val = insn_fetch(s32, ctxt);
3660 		break;
3661 	}
3662 	if (!sign_extension) {
3663 		switch (op->bytes) {
3664 		case 1:
3665 			op->val &= 0xff;
3666 			break;
3667 		case 2:
3668 			op->val &= 0xffff;
3669 			break;
3670 		case 4:
3671 			op->val &= 0xffffffff;
3672 			break;
3673 		}
3674 	}
3675 done:
3676 	return rc;
3677 }
3678 
3679 static int decode_operand(struct x86_emulate_ctxt *ctxt, struct operand *op,
3680 			  unsigned d)
3681 {
3682 	int rc = X86EMUL_CONTINUE;
3683 
3684 	switch (d) {
3685 	case OpReg:
3686 		decode_register_operand(ctxt, op);
3687 		break;
3688 	case OpImmUByte:
3689 		rc = decode_imm(ctxt, op, 1, false);
3690 		break;
3691 	case OpMem:
3692 		ctxt->memop.bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
3693 	mem_common:
3694 		*op = ctxt->memop;
3695 		ctxt->memopp = op;
3696 		if ((ctxt->d & BitOp) && op == &ctxt->dst)
3697 			fetch_bit_operand(ctxt);
3698 		op->orig_val = op->val;
3699 		break;
3700 	case OpMem64:
3701 		ctxt->memop.bytes = 8;
3702 		goto mem_common;
3703 	case OpAcc:
3704 		op->type = OP_REG;
3705 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
3706 		op->addr.reg = &ctxt->regs[VCPU_REGS_RAX];
3707 		fetch_register_operand(op);
3708 		op->orig_val = op->val;
3709 		break;
3710 	case OpDI:
3711 		op->type = OP_MEM;
3712 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
3713 		op->addr.mem.ea =
3714 			register_address(ctxt, ctxt->regs[VCPU_REGS_RDI]);
3715 		op->addr.mem.seg = VCPU_SREG_ES;
3716 		op->val = 0;
3717 		break;
3718 	case OpDX:
3719 		op->type = OP_REG;
3720 		op->bytes = 2;
3721 		op->addr.reg = &ctxt->regs[VCPU_REGS_RDX];
3722 		fetch_register_operand(op);
3723 		break;
3724 	case OpCL:
3725 		op->bytes = 1;
3726 		op->val = ctxt->regs[VCPU_REGS_RCX] & 0xff;
3727 		break;
3728 	case OpImmByte:
3729 		rc = decode_imm(ctxt, op, 1, true);
3730 		break;
3731 	case OpOne:
3732 		op->bytes = 1;
3733 		op->val = 1;
3734 		break;
3735 	case OpImm:
3736 		rc = decode_imm(ctxt, op, imm_size(ctxt), true);
3737 		break;
3738 	case OpMem8:
3739 		ctxt->memop.bytes = 1;
3740 		goto mem_common;
3741 	case OpMem16:
3742 		ctxt->memop.bytes = 2;
3743 		goto mem_common;
3744 	case OpMem32:
3745 		ctxt->memop.bytes = 4;
3746 		goto mem_common;
3747 	case OpImmU16:
3748 		rc = decode_imm(ctxt, op, 2, false);
3749 		break;
3750 	case OpImmU:
3751 		rc = decode_imm(ctxt, op, imm_size(ctxt), false);
3752 		break;
3753 	case OpSI:
3754 		op->type = OP_MEM;
3755 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
3756 		op->addr.mem.ea =
3757 			register_address(ctxt, ctxt->regs[VCPU_REGS_RSI]);
3758 		op->addr.mem.seg = seg_override(ctxt);
3759 		op->val = 0;
3760 		break;
3761 	case OpImmFAddr:
3762 		op->type = OP_IMM;
3763 		op->addr.mem.ea = ctxt->_eip;
3764 		op->bytes = ctxt->op_bytes + 2;
3765 		insn_fetch_arr(op->valptr, op->bytes, ctxt);
3766 		break;
3767 	case OpMemFAddr:
3768 		ctxt->memop.bytes = ctxt->op_bytes + 2;
3769 		goto mem_common;
3770 	case OpES:
3771 		op->val = VCPU_SREG_ES;
3772 		break;
3773 	case OpCS:
3774 		op->val = VCPU_SREG_CS;
3775 		break;
3776 	case OpSS:
3777 		op->val = VCPU_SREG_SS;
3778 		break;
3779 	case OpDS:
3780 		op->val = VCPU_SREG_DS;
3781 		break;
3782 	case OpFS:
3783 		op->val = VCPU_SREG_FS;
3784 		break;
3785 	case OpGS:
3786 		op->val = VCPU_SREG_GS;
3787 		break;
3788 	case OpImplicit:
3789 		/* Special instructions do their own operand decoding. */
3790 	default:
3791 		op->type = OP_NONE; /* Disable writeback. */
3792 		break;
3793 	}
3794 
3795 done:
3796 	return rc;
3797 }
3798 
3799 int x86_decode_insn(struct x86_emulate_ctxt *ctxt, void *insn, int insn_len)
3800 {
3801 	int rc = X86EMUL_CONTINUE;
3802 	int mode = ctxt->mode;
3803 	int def_op_bytes, def_ad_bytes, goffset, simd_prefix;
3804 	bool op_prefix = false;
3805 	struct opcode opcode;
3806 
3807 	ctxt->memop.type = OP_NONE;
3808 	ctxt->memopp = NULL;
3809 	ctxt->_eip = ctxt->eip;
3810 	ctxt->fetch.start = ctxt->_eip;
3811 	ctxt->fetch.end = ctxt->fetch.start + insn_len;
3812 	if (insn_len > 0)
3813 		memcpy(ctxt->fetch.data, insn, insn_len);
3814 
3815 	switch (mode) {
3816 	case X86EMUL_MODE_REAL:
3817 	case X86EMUL_MODE_VM86:
3818 	case X86EMUL_MODE_PROT16:
3819 		def_op_bytes = def_ad_bytes = 2;
3820 		break;
3821 	case X86EMUL_MODE_PROT32:
3822 		def_op_bytes = def_ad_bytes = 4;
3823 		break;
3824 #ifdef CONFIG_X86_64
3825 	case X86EMUL_MODE_PROT64:
3826 		def_op_bytes = 4;
3827 		def_ad_bytes = 8;
3828 		break;
3829 #endif
3830 	default:
3831 		return EMULATION_FAILED;
3832 	}
3833 
3834 	ctxt->op_bytes = def_op_bytes;
3835 	ctxt->ad_bytes = def_ad_bytes;
3836 
3837 	/* Legacy prefixes. */
3838 	for (;;) {
3839 		switch (ctxt->b = insn_fetch(u8, ctxt)) {
3840 		case 0x66:	/* operand-size override */
3841 			op_prefix = true;
3842 			/* switch between 2/4 bytes */
3843 			ctxt->op_bytes = def_op_bytes ^ 6;
3844 			break;
3845 		case 0x67:	/* address-size override */
3846 			if (mode == X86EMUL_MODE_PROT64)
3847 				/* switch between 4/8 bytes */
3848 				ctxt->ad_bytes = def_ad_bytes ^ 12;
3849 			else
3850 				/* switch between 2/4 bytes */
3851 				ctxt->ad_bytes = def_ad_bytes ^ 6;
3852 			break;
3853 		case 0x26:	/* ES override */
3854 		case 0x2e:	/* CS override */
3855 		case 0x36:	/* SS override */
3856 		case 0x3e:	/* DS override */
3857 			set_seg_override(ctxt, (ctxt->b >> 3) & 3);
3858 			break;
3859 		case 0x64:	/* FS override */
3860 		case 0x65:	/* GS override */
3861 			set_seg_override(ctxt, ctxt->b & 7);
3862 			break;
3863 		case 0x40 ... 0x4f: /* REX */
3864 			if (mode != X86EMUL_MODE_PROT64)
3865 				goto done_prefixes;
3866 			ctxt->rex_prefix = ctxt->b;
3867 			continue;
3868 		case 0xf0:	/* LOCK */
3869 			ctxt->lock_prefix = 1;
3870 			break;
3871 		case 0xf2:	/* REPNE/REPNZ */
3872 		case 0xf3:	/* REP/REPE/REPZ */
3873 			ctxt->rep_prefix = ctxt->b;
3874 			break;
3875 		default:
3876 			goto done_prefixes;
3877 		}
3878 
3879 		/* Any legacy prefix after a REX prefix nullifies its effect. */
3880 
3881 		ctxt->rex_prefix = 0;
3882 	}
3883 
3884 done_prefixes:
3885 
3886 	/* REX prefix. */
3887 	if (ctxt->rex_prefix & 8)
3888 		ctxt->op_bytes = 8;	/* REX.W */
3889 
3890 	/* Opcode byte(s). */
3891 	opcode = opcode_table[ctxt->b];
3892 	/* Two-byte opcode? */
3893 	if (ctxt->b == 0x0f) {
3894 		ctxt->twobyte = 1;
3895 		ctxt->b = insn_fetch(u8, ctxt);
3896 		opcode = twobyte_table[ctxt->b];
3897 	}
3898 	ctxt->d = opcode.flags;
3899 
3900 	while (ctxt->d & GroupMask) {
3901 		switch (ctxt->d & GroupMask) {
3902 		case Group:
3903 			ctxt->modrm = insn_fetch(u8, ctxt);
3904 			--ctxt->_eip;
3905 			goffset = (ctxt->modrm >> 3) & 7;
3906 			opcode = opcode.u.group[goffset];
3907 			break;
3908 		case GroupDual:
3909 			ctxt->modrm = insn_fetch(u8, ctxt);
3910 			--ctxt->_eip;
3911 			goffset = (ctxt->modrm >> 3) & 7;
3912 			if ((ctxt->modrm >> 6) == 3)
3913 				opcode = opcode.u.gdual->mod3[goffset];
3914 			else
3915 				opcode = opcode.u.gdual->mod012[goffset];
3916 			break;
3917 		case RMExt:
3918 			goffset = ctxt->modrm & 7;
3919 			opcode = opcode.u.group[goffset];
3920 			break;
3921 		case Prefix:
3922 			if (ctxt->rep_prefix && op_prefix)
3923 				return EMULATION_FAILED;
3924 			simd_prefix = op_prefix ? 0x66 : ctxt->rep_prefix;
3925 			switch (simd_prefix) {
3926 			case 0x00: opcode = opcode.u.gprefix->pfx_no; break;
3927 			case 0x66: opcode = opcode.u.gprefix->pfx_66; break;
3928 			case 0xf2: opcode = opcode.u.gprefix->pfx_f2; break;
3929 			case 0xf3: opcode = opcode.u.gprefix->pfx_f3; break;
3930 			}
3931 			break;
3932 		default:
3933 			return EMULATION_FAILED;
3934 		}
3935 
3936 		ctxt->d &= ~(u64)GroupMask;
3937 		ctxt->d |= opcode.flags;
3938 	}
3939 
3940 	ctxt->execute = opcode.u.execute;
3941 	ctxt->check_perm = opcode.check_perm;
3942 	ctxt->intercept = opcode.intercept;
3943 
3944 	/* Unrecognised? */
3945 	if (ctxt->d == 0 || (ctxt->d & Undefined))
3946 		return EMULATION_FAILED;
3947 
3948 	if (!(ctxt->d & VendorSpecific) && ctxt->only_vendor_specific_insn)
3949 		return EMULATION_FAILED;
3950 
3951 	if (mode == X86EMUL_MODE_PROT64 && (ctxt->d & Stack))
3952 		ctxt->op_bytes = 8;
3953 
3954 	if (ctxt->d & Op3264) {
3955 		if (mode == X86EMUL_MODE_PROT64)
3956 			ctxt->op_bytes = 8;
3957 		else
3958 			ctxt->op_bytes = 4;
3959 	}
3960 
3961 	if (ctxt->d & Sse)
3962 		ctxt->op_bytes = 16;
3963 
3964 	/* ModRM and SIB bytes. */
3965 	if (ctxt->d & ModRM) {
3966 		rc = decode_modrm(ctxt, &ctxt->memop);
3967 		if (!ctxt->has_seg_override)
3968 			set_seg_override(ctxt, ctxt->modrm_seg);
3969 	} else if (ctxt->d & MemAbs)
3970 		rc = decode_abs(ctxt, &ctxt->memop);
3971 	if (rc != X86EMUL_CONTINUE)
3972 		goto done;
3973 
3974 	if (!ctxt->has_seg_override)
3975 		set_seg_override(ctxt, VCPU_SREG_DS);
3976 
3977 	ctxt->memop.addr.mem.seg = seg_override(ctxt);
3978 
3979 	if (ctxt->memop.type == OP_MEM && ctxt->ad_bytes != 8)
3980 		ctxt->memop.addr.mem.ea = (u32)ctxt->memop.addr.mem.ea;
3981 
3982 	/*
3983 	 * Decode and fetch the source operand: register, memory
3984 	 * or immediate.
3985 	 */
3986 	rc = decode_operand(ctxt, &ctxt->src, (ctxt->d >> SrcShift) & OpMask);
3987 	if (rc != X86EMUL_CONTINUE)
3988 		goto done;
3989 
3990 	/*
3991 	 * Decode and fetch the second source operand: register, memory
3992 	 * or immediate.
3993 	 */
3994 	rc = decode_operand(ctxt, &ctxt->src2, (ctxt->d >> Src2Shift) & OpMask);
3995 	if (rc != X86EMUL_CONTINUE)
3996 		goto done;
3997 
3998 	/* Decode and fetch the destination operand: register or memory. */
3999 	rc = decode_operand(ctxt, &ctxt->dst, (ctxt->d >> DstShift) & OpMask);
4000 
4001 done:
4002 	if (ctxt->memopp && ctxt->memopp->type == OP_MEM && ctxt->rip_relative)
4003 		ctxt->memopp->addr.mem.ea += ctxt->_eip;
4004 
4005 	return (rc != X86EMUL_CONTINUE) ? EMULATION_FAILED : EMULATION_OK;
4006 }
4007 
4008 bool x86_page_table_writing_insn(struct x86_emulate_ctxt *ctxt)
4009 {
4010 	return ctxt->d & PageTable;
4011 }
4012 
4013 static bool string_insn_completed(struct x86_emulate_ctxt *ctxt)
4014 {
4015 	/* The second termination condition only applies for REPE
4016 	 * and REPNE. Test if the repeat string operation prefix is
4017 	 * REPE/REPZ or REPNE/REPNZ and if it's the case it tests the
4018 	 * corresponding termination condition according to:
4019 	 * 	- if REPE/REPZ and ZF = 0 then done
4020 	 * 	- if REPNE/REPNZ and ZF = 1 then done
4021 	 */
4022 	if (((ctxt->b == 0xa6) || (ctxt->b == 0xa7) ||
4023 	     (ctxt->b == 0xae) || (ctxt->b == 0xaf))
4024 	    && (((ctxt->rep_prefix == REPE_PREFIX) &&
4025 		 ((ctxt->eflags & EFLG_ZF) == 0))
4026 		|| ((ctxt->rep_prefix == REPNE_PREFIX) &&
4027 		    ((ctxt->eflags & EFLG_ZF) == EFLG_ZF))))
4028 		return true;
4029 
4030 	return false;
4031 }
4032 
4033 int x86_emulate_insn(struct x86_emulate_ctxt *ctxt)
4034 {
4035 	struct x86_emulate_ops *ops = ctxt->ops;
4036 	int rc = X86EMUL_CONTINUE;
4037 	int saved_dst_type = ctxt->dst.type;
4038 
4039 	ctxt->mem_read.pos = 0;
4040 
4041 	if (ctxt->mode == X86EMUL_MODE_PROT64 && (ctxt->d & No64)) {
4042 		rc = emulate_ud(ctxt);
4043 		goto done;
4044 	}
4045 
4046 	/* LOCK prefix is allowed only with some instructions */
4047 	if (ctxt->lock_prefix && (!(ctxt->d & Lock) || ctxt->dst.type != OP_MEM)) {
4048 		rc = emulate_ud(ctxt);
4049 		goto done;
4050 	}
4051 
4052 	if ((ctxt->d & SrcMask) == SrcMemFAddr && ctxt->src.type != OP_MEM) {
4053 		rc = emulate_ud(ctxt);
4054 		goto done;
4055 	}
4056 
4057 	if ((ctxt->d & Sse)
4058 	    && ((ops->get_cr(ctxt, 0) & X86_CR0_EM)
4059 		|| !(ops->get_cr(ctxt, 4) & X86_CR4_OSFXSR))) {
4060 		rc = emulate_ud(ctxt);
4061 		goto done;
4062 	}
4063 
4064 	if ((ctxt->d & Sse) && (ops->get_cr(ctxt, 0) & X86_CR0_TS)) {
4065 		rc = emulate_nm(ctxt);
4066 		goto done;
4067 	}
4068 
4069 	if (unlikely(ctxt->guest_mode) && ctxt->intercept) {
4070 		rc = emulator_check_intercept(ctxt, ctxt->intercept,
4071 					      X86_ICPT_PRE_EXCEPT);
4072 		if (rc != X86EMUL_CONTINUE)
4073 			goto done;
4074 	}
4075 
4076 	/* Privileged instruction can be executed only in CPL=0 */
4077 	if ((ctxt->d & Priv) && ops->cpl(ctxt)) {
4078 		rc = emulate_gp(ctxt, 0);
4079 		goto done;
4080 	}
4081 
4082 	/* Instruction can only be executed in protected mode */
4083 	if ((ctxt->d & Prot) && !(ctxt->mode & X86EMUL_MODE_PROT)) {
4084 		rc = emulate_ud(ctxt);
4085 		goto done;
4086 	}
4087 
4088 	/* Do instruction specific permission checks */
4089 	if (ctxt->check_perm) {
4090 		rc = ctxt->check_perm(ctxt);
4091 		if (rc != X86EMUL_CONTINUE)
4092 			goto done;
4093 	}
4094 
4095 	if (unlikely(ctxt->guest_mode) && ctxt->intercept) {
4096 		rc = emulator_check_intercept(ctxt, ctxt->intercept,
4097 					      X86_ICPT_POST_EXCEPT);
4098 		if (rc != X86EMUL_CONTINUE)
4099 			goto done;
4100 	}
4101 
4102 	if (ctxt->rep_prefix && (ctxt->d & String)) {
4103 		/* All REP prefixes have the same first termination condition */
4104 		if (address_mask(ctxt, ctxt->regs[VCPU_REGS_RCX]) == 0) {
4105 			ctxt->eip = ctxt->_eip;
4106 			goto done;
4107 		}
4108 	}
4109 
4110 	if ((ctxt->src.type == OP_MEM) && !(ctxt->d & NoAccess)) {
4111 		rc = segmented_read(ctxt, ctxt->src.addr.mem,
4112 				    ctxt->src.valptr, ctxt->src.bytes);
4113 		if (rc != X86EMUL_CONTINUE)
4114 			goto done;
4115 		ctxt->src.orig_val64 = ctxt->src.val64;
4116 	}
4117 
4118 	if (ctxt->src2.type == OP_MEM) {
4119 		rc = segmented_read(ctxt, ctxt->src2.addr.mem,
4120 				    &ctxt->src2.val, ctxt->src2.bytes);
4121 		if (rc != X86EMUL_CONTINUE)
4122 			goto done;
4123 	}
4124 
4125 	if ((ctxt->d & DstMask) == ImplicitOps)
4126 		goto special_insn;
4127 
4128 
4129 	if ((ctxt->dst.type == OP_MEM) && !(ctxt->d & Mov)) {
4130 		/* optimisation - avoid slow emulated read if Mov */
4131 		rc = segmented_read(ctxt, ctxt->dst.addr.mem,
4132 				   &ctxt->dst.val, ctxt->dst.bytes);
4133 		if (rc != X86EMUL_CONTINUE)
4134 			goto done;
4135 	}
4136 	ctxt->dst.orig_val = ctxt->dst.val;
4137 
4138 special_insn:
4139 
4140 	if (unlikely(ctxt->guest_mode) && ctxt->intercept) {
4141 		rc = emulator_check_intercept(ctxt, ctxt->intercept,
4142 					      X86_ICPT_POST_MEMACCESS);
4143 		if (rc != X86EMUL_CONTINUE)
4144 			goto done;
4145 	}
4146 
4147 	if (ctxt->execute) {
4148 		rc = ctxt->execute(ctxt);
4149 		if (rc != X86EMUL_CONTINUE)
4150 			goto done;
4151 		goto writeback;
4152 	}
4153 
4154 	if (ctxt->twobyte)
4155 		goto twobyte_insn;
4156 
4157 	switch (ctxt->b) {
4158 	case 0x40 ... 0x47: /* inc r16/r32 */
4159 		emulate_1op(ctxt, "inc");
4160 		break;
4161 	case 0x48 ... 0x4f: /* dec r16/r32 */
4162 		emulate_1op(ctxt, "dec");
4163 		break;
4164 	case 0x63:		/* movsxd */
4165 		if (ctxt->mode != X86EMUL_MODE_PROT64)
4166 			goto cannot_emulate;
4167 		ctxt->dst.val = (s32) ctxt->src.val;
4168 		break;
4169 	case 0x70 ... 0x7f: /* jcc (short) */
4170 		if (test_cc(ctxt->b, ctxt->eflags))
4171 			jmp_rel(ctxt, ctxt->src.val);
4172 		break;
4173 	case 0x8d: /* lea r16/r32, m */
4174 		ctxt->dst.val = ctxt->src.addr.mem.ea;
4175 		break;
4176 	case 0x90 ... 0x97: /* nop / xchg reg, rax */
4177 		if (ctxt->dst.addr.reg == &ctxt->regs[VCPU_REGS_RAX])
4178 			break;
4179 		rc = em_xchg(ctxt);
4180 		break;
4181 	case 0x98: /* cbw/cwde/cdqe */
4182 		switch (ctxt->op_bytes) {
4183 		case 2: ctxt->dst.val = (s8)ctxt->dst.val; break;
4184 		case 4: ctxt->dst.val = (s16)ctxt->dst.val; break;
4185 		case 8: ctxt->dst.val = (s32)ctxt->dst.val; break;
4186 		}
4187 		break;
4188 	case 0xc0 ... 0xc1:
4189 		rc = em_grp2(ctxt);
4190 		break;
4191 	case 0xcc:		/* int3 */
4192 		rc = emulate_int(ctxt, 3);
4193 		break;
4194 	case 0xcd:		/* int n */
4195 		rc = emulate_int(ctxt, ctxt->src.val);
4196 		break;
4197 	case 0xce:		/* into */
4198 		if (ctxt->eflags & EFLG_OF)
4199 			rc = emulate_int(ctxt, 4);
4200 		break;
4201 	case 0xd0 ... 0xd1:	/* Grp2 */
4202 		rc = em_grp2(ctxt);
4203 		break;
4204 	case 0xd2 ... 0xd3:	/* Grp2 */
4205 		ctxt->src.val = ctxt->regs[VCPU_REGS_RCX];
4206 		rc = em_grp2(ctxt);
4207 		break;
4208 	case 0xe9: /* jmp rel */
4209 	case 0xeb: /* jmp rel short */
4210 		jmp_rel(ctxt, ctxt->src.val);
4211 		ctxt->dst.type = OP_NONE; /* Disable writeback. */
4212 		break;
4213 	case 0xf4:              /* hlt */
4214 		ctxt->ops->halt(ctxt);
4215 		break;
4216 	case 0xf5:	/* cmc */
4217 		/* complement carry flag from eflags reg */
4218 		ctxt->eflags ^= EFLG_CF;
4219 		break;
4220 	case 0xf8: /* clc */
4221 		ctxt->eflags &= ~EFLG_CF;
4222 		break;
4223 	case 0xf9: /* stc */
4224 		ctxt->eflags |= EFLG_CF;
4225 		break;
4226 	case 0xfc: /* cld */
4227 		ctxt->eflags &= ~EFLG_DF;
4228 		break;
4229 	case 0xfd: /* std */
4230 		ctxt->eflags |= EFLG_DF;
4231 		break;
4232 	default:
4233 		goto cannot_emulate;
4234 	}
4235 
4236 	if (rc != X86EMUL_CONTINUE)
4237 		goto done;
4238 
4239 writeback:
4240 	rc = writeback(ctxt);
4241 	if (rc != X86EMUL_CONTINUE)
4242 		goto done;
4243 
4244 	/*
4245 	 * restore dst type in case the decoding will be reused
4246 	 * (happens for string instruction )
4247 	 */
4248 	ctxt->dst.type = saved_dst_type;
4249 
4250 	if ((ctxt->d & SrcMask) == SrcSI)
4251 		string_addr_inc(ctxt, seg_override(ctxt),
4252 				VCPU_REGS_RSI, &ctxt->src);
4253 
4254 	if ((ctxt->d & DstMask) == DstDI)
4255 		string_addr_inc(ctxt, VCPU_SREG_ES, VCPU_REGS_RDI,
4256 				&ctxt->dst);
4257 
4258 	if (ctxt->rep_prefix && (ctxt->d & String)) {
4259 		struct read_cache *r = &ctxt->io_read;
4260 		register_address_increment(ctxt, &ctxt->regs[VCPU_REGS_RCX], -1);
4261 
4262 		if (!string_insn_completed(ctxt)) {
4263 			/*
4264 			 * Re-enter guest when pio read ahead buffer is empty
4265 			 * or, if it is not used, after each 1024 iteration.
4266 			 */
4267 			if ((r->end != 0 || ctxt->regs[VCPU_REGS_RCX] & 0x3ff) &&
4268 			    (r->end == 0 || r->end != r->pos)) {
4269 				/*
4270 				 * Reset read cache. Usually happens before
4271 				 * decode, but since instruction is restarted
4272 				 * we have to do it here.
4273 				 */
4274 				ctxt->mem_read.end = 0;
4275 				return EMULATION_RESTART;
4276 			}
4277 			goto done; /* skip rip writeback */
4278 		}
4279 	}
4280 
4281 	ctxt->eip = ctxt->_eip;
4282 
4283 done:
4284 	if (rc == X86EMUL_PROPAGATE_FAULT)
4285 		ctxt->have_exception = true;
4286 	if (rc == X86EMUL_INTERCEPTED)
4287 		return EMULATION_INTERCEPTED;
4288 
4289 	return (rc == X86EMUL_UNHANDLEABLE) ? EMULATION_FAILED : EMULATION_OK;
4290 
4291 twobyte_insn:
4292 	switch (ctxt->b) {
4293 	case 0x09:		/* wbinvd */
4294 		(ctxt->ops->wbinvd)(ctxt);
4295 		break;
4296 	case 0x08:		/* invd */
4297 	case 0x0d:		/* GrpP (prefetch) */
4298 	case 0x18:		/* Grp16 (prefetch/nop) */
4299 		break;
4300 	case 0x20: /* mov cr, reg */
4301 		ctxt->dst.val = ops->get_cr(ctxt, ctxt->modrm_reg);
4302 		break;
4303 	case 0x21: /* mov from dr to reg */
4304 		ops->get_dr(ctxt, ctxt->modrm_reg, &ctxt->dst.val);
4305 		break;
4306 	case 0x40 ... 0x4f:	/* cmov */
4307 		ctxt->dst.val = ctxt->dst.orig_val = ctxt->src.val;
4308 		if (!test_cc(ctxt->b, ctxt->eflags))
4309 			ctxt->dst.type = OP_NONE; /* no writeback */
4310 		break;
4311 	case 0x80 ... 0x8f: /* jnz rel, etc*/
4312 		if (test_cc(ctxt->b, ctxt->eflags))
4313 			jmp_rel(ctxt, ctxt->src.val);
4314 		break;
4315 	case 0x90 ... 0x9f:     /* setcc r/m8 */
4316 		ctxt->dst.val = test_cc(ctxt->b, ctxt->eflags);
4317 		break;
4318 	case 0xa4: /* shld imm8, r, r/m */
4319 	case 0xa5: /* shld cl, r, r/m */
4320 		emulate_2op_cl(ctxt, "shld");
4321 		break;
4322 	case 0xac: /* shrd imm8, r, r/m */
4323 	case 0xad: /* shrd cl, r, r/m */
4324 		emulate_2op_cl(ctxt, "shrd");
4325 		break;
4326 	case 0xae:              /* clflush */
4327 		break;
4328 	case 0xb6 ... 0xb7:	/* movzx */
4329 		ctxt->dst.bytes = ctxt->op_bytes;
4330 		ctxt->dst.val = (ctxt->d & ByteOp) ? (u8) ctxt->src.val
4331 						       : (u16) ctxt->src.val;
4332 		break;
4333 	case 0xbe ... 0xbf:	/* movsx */
4334 		ctxt->dst.bytes = ctxt->op_bytes;
4335 		ctxt->dst.val = (ctxt->d & ByteOp) ? (s8) ctxt->src.val :
4336 							(s16) ctxt->src.val;
4337 		break;
4338 	case 0xc0 ... 0xc1:	/* xadd */
4339 		emulate_2op_SrcV(ctxt, "add");
4340 		/* Write back the register source. */
4341 		ctxt->src.val = ctxt->dst.orig_val;
4342 		write_register_operand(&ctxt->src);
4343 		break;
4344 	case 0xc3:		/* movnti */
4345 		ctxt->dst.bytes = ctxt->op_bytes;
4346 		ctxt->dst.val = (ctxt->op_bytes == 4) ? (u32) ctxt->src.val :
4347 							(u64) ctxt->src.val;
4348 		break;
4349 	default:
4350 		goto cannot_emulate;
4351 	}
4352 
4353 	if (rc != X86EMUL_CONTINUE)
4354 		goto done;
4355 
4356 	goto writeback;
4357 
4358 cannot_emulate:
4359 	return EMULATION_FAILED;
4360 }
4361