xref: /linux/arch/x86/kvm/emulate.c (revision 0d456bad36d42d16022be045c8a53ddbb59ee478)
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 #define Mmx         ((u64)1 << 40)  /* MMX Vector instruction */
146 #define Aligned     ((u64)1 << 41)  /* Explicitly aligned (e.g. MOVDQA) */
147 #define Unaligned   ((u64)1 << 42)  /* Explicitly unaligned (e.g. MOVDQU) */
148 #define Avx         ((u64)1 << 43)  /* Advanced Vector Extensions */
149 
150 #define X2(x...) x, x
151 #define X3(x...) X2(x), x
152 #define X4(x...) X2(x), X2(x)
153 #define X5(x...) X4(x), x
154 #define X6(x...) X4(x), X2(x)
155 #define X7(x...) X4(x), X3(x)
156 #define X8(x...) X4(x), X4(x)
157 #define X16(x...) X8(x), X8(x)
158 
159 struct opcode {
160 	u64 flags : 56;
161 	u64 intercept : 8;
162 	union {
163 		int (*execute)(struct x86_emulate_ctxt *ctxt);
164 		const struct opcode *group;
165 		const struct group_dual *gdual;
166 		const struct gprefix *gprefix;
167 	} u;
168 	int (*check_perm)(struct x86_emulate_ctxt *ctxt);
169 };
170 
171 struct group_dual {
172 	struct opcode mod012[8];
173 	struct opcode mod3[8];
174 };
175 
176 struct gprefix {
177 	struct opcode pfx_no;
178 	struct opcode pfx_66;
179 	struct opcode pfx_f2;
180 	struct opcode pfx_f3;
181 };
182 
183 /* EFLAGS bit definitions. */
184 #define EFLG_ID (1<<21)
185 #define EFLG_VIP (1<<20)
186 #define EFLG_VIF (1<<19)
187 #define EFLG_AC (1<<18)
188 #define EFLG_VM (1<<17)
189 #define EFLG_RF (1<<16)
190 #define EFLG_IOPL (3<<12)
191 #define EFLG_NT (1<<14)
192 #define EFLG_OF (1<<11)
193 #define EFLG_DF (1<<10)
194 #define EFLG_IF (1<<9)
195 #define EFLG_TF (1<<8)
196 #define EFLG_SF (1<<7)
197 #define EFLG_ZF (1<<6)
198 #define EFLG_AF (1<<4)
199 #define EFLG_PF (1<<2)
200 #define EFLG_CF (1<<0)
201 
202 #define EFLG_RESERVED_ZEROS_MASK 0xffc0802a
203 #define EFLG_RESERVED_ONE_MASK 2
204 
205 static ulong reg_read(struct x86_emulate_ctxt *ctxt, unsigned nr)
206 {
207 	if (!(ctxt->regs_valid & (1 << nr))) {
208 		ctxt->regs_valid |= 1 << nr;
209 		ctxt->_regs[nr] = ctxt->ops->read_gpr(ctxt, nr);
210 	}
211 	return ctxt->_regs[nr];
212 }
213 
214 static ulong *reg_write(struct x86_emulate_ctxt *ctxt, unsigned nr)
215 {
216 	ctxt->regs_valid |= 1 << nr;
217 	ctxt->regs_dirty |= 1 << nr;
218 	return &ctxt->_regs[nr];
219 }
220 
221 static ulong *reg_rmw(struct x86_emulate_ctxt *ctxt, unsigned nr)
222 {
223 	reg_read(ctxt, nr);
224 	return reg_write(ctxt, nr);
225 }
226 
227 static void writeback_registers(struct x86_emulate_ctxt *ctxt)
228 {
229 	unsigned reg;
230 
231 	for_each_set_bit(reg, (ulong *)&ctxt->regs_dirty, 16)
232 		ctxt->ops->write_gpr(ctxt, reg, ctxt->_regs[reg]);
233 }
234 
235 static void invalidate_registers(struct x86_emulate_ctxt *ctxt)
236 {
237 	ctxt->regs_dirty = 0;
238 	ctxt->regs_valid = 0;
239 }
240 
241 /*
242  * Instruction emulation:
243  * Most instructions are emulated directly via a fragment of inline assembly
244  * code. This allows us to save/restore EFLAGS and thus very easily pick up
245  * any modified flags.
246  */
247 
248 #if defined(CONFIG_X86_64)
249 #define _LO32 "k"		/* force 32-bit operand */
250 #define _STK  "%%rsp"		/* stack pointer */
251 #elif defined(__i386__)
252 #define _LO32 ""		/* force 32-bit operand */
253 #define _STK  "%%esp"		/* stack pointer */
254 #endif
255 
256 /*
257  * These EFLAGS bits are restored from saved value during emulation, and
258  * any changes are written back to the saved value after emulation.
259  */
260 #define EFLAGS_MASK (EFLG_OF|EFLG_SF|EFLG_ZF|EFLG_AF|EFLG_PF|EFLG_CF)
261 
262 /* Before executing instruction: restore necessary bits in EFLAGS. */
263 #define _PRE_EFLAGS(_sav, _msk, _tmp)					\
264 	/* EFLAGS = (_sav & _msk) | (EFLAGS & ~_msk); _sav &= ~_msk; */ \
265 	"movl %"_sav",%"_LO32 _tmp"; "                                  \
266 	"push %"_tmp"; "                                                \
267 	"push %"_tmp"; "                                                \
268 	"movl %"_msk",%"_LO32 _tmp"; "                                  \
269 	"andl %"_LO32 _tmp",("_STK"); "                                 \
270 	"pushf; "                                                       \
271 	"notl %"_LO32 _tmp"; "                                          \
272 	"andl %"_LO32 _tmp",("_STK"); "                                 \
273 	"andl %"_LO32 _tmp","__stringify(BITS_PER_LONG/4)"("_STK"); "	\
274 	"pop  %"_tmp"; "                                                \
275 	"orl  %"_LO32 _tmp",("_STK"); "                                 \
276 	"popf; "                                                        \
277 	"pop  %"_sav"; "
278 
279 /* After executing instruction: write-back necessary bits in EFLAGS. */
280 #define _POST_EFLAGS(_sav, _msk, _tmp) \
281 	/* _sav |= EFLAGS & _msk; */		\
282 	"pushf; "				\
283 	"pop  %"_tmp"; "			\
284 	"andl %"_msk",%"_LO32 _tmp"; "		\
285 	"orl  %"_LO32 _tmp",%"_sav"; "
286 
287 #ifdef CONFIG_X86_64
288 #define ON64(x) x
289 #else
290 #define ON64(x)
291 #endif
292 
293 #define ____emulate_2op(ctxt, _op, _x, _y, _suffix, _dsttype)	\
294 	do {								\
295 		__asm__ __volatile__ (					\
296 			_PRE_EFLAGS("0", "4", "2")			\
297 			_op _suffix " %"_x"3,%1; "			\
298 			_POST_EFLAGS("0", "4", "2")			\
299 			: "=m" ((ctxt)->eflags),			\
300 			  "+q" (*(_dsttype*)&(ctxt)->dst.val),		\
301 			  "=&r" (_tmp)					\
302 			: _y ((ctxt)->src.val), "i" (EFLAGS_MASK));	\
303 	} while (0)
304 
305 
306 /* Raw emulation: instruction has two explicit operands. */
307 #define __emulate_2op_nobyte(ctxt,_op,_wx,_wy,_lx,_ly,_qx,_qy)		\
308 	do {								\
309 		unsigned long _tmp;					\
310 									\
311 		switch ((ctxt)->dst.bytes) {				\
312 		case 2:							\
313 			____emulate_2op(ctxt,_op,_wx,_wy,"w",u16);	\
314 			break;						\
315 		case 4:							\
316 			____emulate_2op(ctxt,_op,_lx,_ly,"l",u32);	\
317 			break;						\
318 		case 8:							\
319 			ON64(____emulate_2op(ctxt,_op,_qx,_qy,"q",u64)); \
320 			break;						\
321 		}							\
322 	} while (0)
323 
324 #define __emulate_2op(ctxt,_op,_bx,_by,_wx,_wy,_lx,_ly,_qx,_qy)		     \
325 	do {								     \
326 		unsigned long _tmp;					     \
327 		switch ((ctxt)->dst.bytes) {				     \
328 		case 1:							     \
329 			____emulate_2op(ctxt,_op,_bx,_by,"b",u8);	     \
330 			break;						     \
331 		default:						     \
332 			__emulate_2op_nobyte(ctxt, _op,			     \
333 					     _wx, _wy, _lx, _ly, _qx, _qy);  \
334 			break;						     \
335 		}							     \
336 	} while (0)
337 
338 /* Source operand is byte-sized and may be restricted to just %cl. */
339 #define emulate_2op_SrcB(ctxt, _op)					\
340 	__emulate_2op(ctxt, _op, "b", "c", "b", "c", "b", "c", "b", "c")
341 
342 /* Source operand is byte, word, long or quad sized. */
343 #define emulate_2op_SrcV(ctxt, _op)					\
344 	__emulate_2op(ctxt, _op, "b", "q", "w", "r", _LO32, "r", "", "r")
345 
346 /* Source operand is word, long or quad sized. */
347 #define emulate_2op_SrcV_nobyte(ctxt, _op)				\
348 	__emulate_2op_nobyte(ctxt, _op, "w", "r", _LO32, "r", "", "r")
349 
350 /* Instruction has three operands and one operand is stored in ECX register */
351 #define __emulate_2op_cl(ctxt, _op, _suffix, _type)		\
352 	do {								\
353 		unsigned long _tmp;					\
354 		_type _clv  = (ctxt)->src2.val;				\
355 		_type _srcv = (ctxt)->src.val;				\
356 		_type _dstv = (ctxt)->dst.val;				\
357 									\
358 		__asm__ __volatile__ (					\
359 			_PRE_EFLAGS("0", "5", "2")			\
360 			_op _suffix " %4,%1 \n"				\
361 			_POST_EFLAGS("0", "5", "2")			\
362 			: "=m" ((ctxt)->eflags), "+r" (_dstv), "=&r" (_tmp) \
363 			: "c" (_clv) , "r" (_srcv), "i" (EFLAGS_MASK)	\
364 			);						\
365 									\
366 		(ctxt)->src2.val  = (unsigned long) _clv;		\
367 		(ctxt)->src2.val = (unsigned long) _srcv;		\
368 		(ctxt)->dst.val = (unsigned long) _dstv;		\
369 	} while (0)
370 
371 #define emulate_2op_cl(ctxt, _op)					\
372 	do {								\
373 		switch ((ctxt)->dst.bytes) {				\
374 		case 2:							\
375 			__emulate_2op_cl(ctxt, _op, "w", u16);		\
376 			break;						\
377 		case 4:							\
378 			__emulate_2op_cl(ctxt, _op, "l", u32);		\
379 			break;						\
380 		case 8:							\
381 			ON64(__emulate_2op_cl(ctxt, _op, "q", ulong));	\
382 			break;						\
383 		}							\
384 	} while (0)
385 
386 #define __emulate_1op(ctxt, _op, _suffix)				\
387 	do {								\
388 		unsigned long _tmp;					\
389 									\
390 		__asm__ __volatile__ (					\
391 			_PRE_EFLAGS("0", "3", "2")			\
392 			_op _suffix " %1; "				\
393 			_POST_EFLAGS("0", "3", "2")			\
394 			: "=m" ((ctxt)->eflags), "+m" ((ctxt)->dst.val), \
395 			  "=&r" (_tmp)					\
396 			: "i" (EFLAGS_MASK));				\
397 	} while (0)
398 
399 /* Instruction has only one explicit operand (no source operand). */
400 #define emulate_1op(ctxt, _op)						\
401 	do {								\
402 		switch ((ctxt)->dst.bytes) {				\
403 		case 1:	__emulate_1op(ctxt, _op, "b"); break;		\
404 		case 2:	__emulate_1op(ctxt, _op, "w"); break;		\
405 		case 4:	__emulate_1op(ctxt, _op, "l"); break;		\
406 		case 8:	ON64(__emulate_1op(ctxt, _op, "q")); break;	\
407 		}							\
408 	} while (0)
409 
410 #define __emulate_1op_rax_rdx(ctxt, _op, _suffix, _ex)			\
411 	do {								\
412 		unsigned long _tmp;					\
413 		ulong *rax = reg_rmw((ctxt), VCPU_REGS_RAX);		\
414 		ulong *rdx = reg_rmw((ctxt), VCPU_REGS_RDX);		\
415 									\
416 		__asm__ __volatile__ (					\
417 			_PRE_EFLAGS("0", "5", "1")			\
418 			"1: \n\t"					\
419 			_op _suffix " %6; "				\
420 			"2: \n\t"					\
421 			_POST_EFLAGS("0", "5", "1")			\
422 			".pushsection .fixup,\"ax\" \n\t"		\
423 			"3: movb $1, %4 \n\t"				\
424 			"jmp 2b \n\t"					\
425 			".popsection \n\t"				\
426 			_ASM_EXTABLE(1b, 3b)				\
427 			: "=m" ((ctxt)->eflags), "=&r" (_tmp),		\
428 			  "+a" (*rax), "+d" (*rdx), "+qm"(_ex)		\
429 			: "i" (EFLAGS_MASK), "m" ((ctxt)->src.val));	\
430 	} while (0)
431 
432 /* instruction has only one source operand, destination is implicit (e.g. mul, div, imul, idiv) */
433 #define emulate_1op_rax_rdx(ctxt, _op, _ex)	\
434 	do {								\
435 		switch((ctxt)->src.bytes) {				\
436 		case 1:							\
437 			__emulate_1op_rax_rdx(ctxt, _op, "b", _ex);	\
438 			break;						\
439 		case 2:							\
440 			__emulate_1op_rax_rdx(ctxt, _op, "w", _ex);	\
441 			break;						\
442 		case 4:							\
443 			__emulate_1op_rax_rdx(ctxt, _op, "l", _ex);	\
444 			break;						\
445 		case 8: ON64(						\
446 			__emulate_1op_rax_rdx(ctxt, _op, "q", _ex));	\
447 			break;						\
448 		}							\
449 	} while (0)
450 
451 static int emulator_check_intercept(struct x86_emulate_ctxt *ctxt,
452 				    enum x86_intercept intercept,
453 				    enum x86_intercept_stage stage)
454 {
455 	struct x86_instruction_info info = {
456 		.intercept  = intercept,
457 		.rep_prefix = ctxt->rep_prefix,
458 		.modrm_mod  = ctxt->modrm_mod,
459 		.modrm_reg  = ctxt->modrm_reg,
460 		.modrm_rm   = ctxt->modrm_rm,
461 		.src_val    = ctxt->src.val64,
462 		.src_bytes  = ctxt->src.bytes,
463 		.dst_bytes  = ctxt->dst.bytes,
464 		.ad_bytes   = ctxt->ad_bytes,
465 		.next_rip   = ctxt->eip,
466 	};
467 
468 	return ctxt->ops->intercept(ctxt, &info, stage);
469 }
470 
471 static void assign_masked(ulong *dest, ulong src, ulong mask)
472 {
473 	*dest = (*dest & ~mask) | (src & mask);
474 }
475 
476 static inline unsigned long ad_mask(struct x86_emulate_ctxt *ctxt)
477 {
478 	return (1UL << (ctxt->ad_bytes << 3)) - 1;
479 }
480 
481 static ulong stack_mask(struct x86_emulate_ctxt *ctxt)
482 {
483 	u16 sel;
484 	struct desc_struct ss;
485 
486 	if (ctxt->mode == X86EMUL_MODE_PROT64)
487 		return ~0UL;
488 	ctxt->ops->get_segment(ctxt, &sel, &ss, NULL, VCPU_SREG_SS);
489 	return ~0U >> ((ss.d ^ 1) * 16);  /* d=0: 0xffff; d=1: 0xffffffff */
490 }
491 
492 static int stack_size(struct x86_emulate_ctxt *ctxt)
493 {
494 	return (__fls(stack_mask(ctxt)) + 1) >> 3;
495 }
496 
497 /* Access/update address held in a register, based on addressing mode. */
498 static inline unsigned long
499 address_mask(struct x86_emulate_ctxt *ctxt, unsigned long reg)
500 {
501 	if (ctxt->ad_bytes == sizeof(unsigned long))
502 		return reg;
503 	else
504 		return reg & ad_mask(ctxt);
505 }
506 
507 static inline unsigned long
508 register_address(struct x86_emulate_ctxt *ctxt, unsigned long reg)
509 {
510 	return address_mask(ctxt, reg);
511 }
512 
513 static void masked_increment(ulong *reg, ulong mask, int inc)
514 {
515 	assign_masked(reg, *reg + inc, mask);
516 }
517 
518 static inline void
519 register_address_increment(struct x86_emulate_ctxt *ctxt, unsigned long *reg, int inc)
520 {
521 	ulong mask;
522 
523 	if (ctxt->ad_bytes == sizeof(unsigned long))
524 		mask = ~0UL;
525 	else
526 		mask = ad_mask(ctxt);
527 	masked_increment(reg, mask, inc);
528 }
529 
530 static void rsp_increment(struct x86_emulate_ctxt *ctxt, int inc)
531 {
532 	masked_increment(reg_rmw(ctxt, VCPU_REGS_RSP), stack_mask(ctxt), inc);
533 }
534 
535 static inline void jmp_rel(struct x86_emulate_ctxt *ctxt, int rel)
536 {
537 	register_address_increment(ctxt, &ctxt->_eip, rel);
538 }
539 
540 static u32 desc_limit_scaled(struct desc_struct *desc)
541 {
542 	u32 limit = get_desc_limit(desc);
543 
544 	return desc->g ? (limit << 12) | 0xfff : limit;
545 }
546 
547 static void set_seg_override(struct x86_emulate_ctxt *ctxt, int seg)
548 {
549 	ctxt->has_seg_override = true;
550 	ctxt->seg_override = seg;
551 }
552 
553 static unsigned long seg_base(struct x86_emulate_ctxt *ctxt, int seg)
554 {
555 	if (ctxt->mode == X86EMUL_MODE_PROT64 && seg < VCPU_SREG_FS)
556 		return 0;
557 
558 	return ctxt->ops->get_cached_segment_base(ctxt, seg);
559 }
560 
561 static unsigned seg_override(struct x86_emulate_ctxt *ctxt)
562 {
563 	if (!ctxt->has_seg_override)
564 		return 0;
565 
566 	return ctxt->seg_override;
567 }
568 
569 static int emulate_exception(struct x86_emulate_ctxt *ctxt, int vec,
570 			     u32 error, bool valid)
571 {
572 	ctxt->exception.vector = vec;
573 	ctxt->exception.error_code = error;
574 	ctxt->exception.error_code_valid = valid;
575 	return X86EMUL_PROPAGATE_FAULT;
576 }
577 
578 static int emulate_db(struct x86_emulate_ctxt *ctxt)
579 {
580 	return emulate_exception(ctxt, DB_VECTOR, 0, false);
581 }
582 
583 static int emulate_gp(struct x86_emulate_ctxt *ctxt, int err)
584 {
585 	return emulate_exception(ctxt, GP_VECTOR, err, true);
586 }
587 
588 static int emulate_ss(struct x86_emulate_ctxt *ctxt, int err)
589 {
590 	return emulate_exception(ctxt, SS_VECTOR, err, true);
591 }
592 
593 static int emulate_ud(struct x86_emulate_ctxt *ctxt)
594 {
595 	return emulate_exception(ctxt, UD_VECTOR, 0, false);
596 }
597 
598 static int emulate_ts(struct x86_emulate_ctxt *ctxt, int err)
599 {
600 	return emulate_exception(ctxt, TS_VECTOR, err, true);
601 }
602 
603 static int emulate_de(struct x86_emulate_ctxt *ctxt)
604 {
605 	return emulate_exception(ctxt, DE_VECTOR, 0, false);
606 }
607 
608 static int emulate_nm(struct x86_emulate_ctxt *ctxt)
609 {
610 	return emulate_exception(ctxt, NM_VECTOR, 0, false);
611 }
612 
613 static u16 get_segment_selector(struct x86_emulate_ctxt *ctxt, unsigned seg)
614 {
615 	u16 selector;
616 	struct desc_struct desc;
617 
618 	ctxt->ops->get_segment(ctxt, &selector, &desc, NULL, seg);
619 	return selector;
620 }
621 
622 static void set_segment_selector(struct x86_emulate_ctxt *ctxt, u16 selector,
623 				 unsigned seg)
624 {
625 	u16 dummy;
626 	u32 base3;
627 	struct desc_struct desc;
628 
629 	ctxt->ops->get_segment(ctxt, &dummy, &desc, &base3, seg);
630 	ctxt->ops->set_segment(ctxt, selector, &desc, base3, seg);
631 }
632 
633 /*
634  * x86 defines three classes of vector instructions: explicitly
635  * aligned, explicitly unaligned, and the rest, which change behaviour
636  * depending on whether they're AVX encoded or not.
637  *
638  * Also included is CMPXCHG16B which is not a vector instruction, yet it is
639  * subject to the same check.
640  */
641 static bool insn_aligned(struct x86_emulate_ctxt *ctxt, unsigned size)
642 {
643 	if (likely(size < 16))
644 		return false;
645 
646 	if (ctxt->d & Aligned)
647 		return true;
648 	else if (ctxt->d & Unaligned)
649 		return false;
650 	else if (ctxt->d & Avx)
651 		return false;
652 	else
653 		return true;
654 }
655 
656 static int __linearize(struct x86_emulate_ctxt *ctxt,
657 		     struct segmented_address addr,
658 		     unsigned size, bool write, bool fetch,
659 		     ulong *linear)
660 {
661 	struct desc_struct desc;
662 	bool usable;
663 	ulong la;
664 	u32 lim;
665 	u16 sel;
666 	unsigned cpl, rpl;
667 
668 	la = seg_base(ctxt, addr.seg) + addr.ea;
669 	switch (ctxt->mode) {
670 	case X86EMUL_MODE_PROT64:
671 		if (((signed long)la << 16) >> 16 != la)
672 			return emulate_gp(ctxt, 0);
673 		break;
674 	default:
675 		usable = ctxt->ops->get_segment(ctxt, &sel, &desc, NULL,
676 						addr.seg);
677 		if (!usable)
678 			goto bad;
679 		/* code segment in protected mode or read-only data segment */
680 		if ((((ctxt->mode != X86EMUL_MODE_REAL) && (desc.type & 8))
681 					|| !(desc.type & 2)) && write)
682 			goto bad;
683 		/* unreadable code segment */
684 		if (!fetch && (desc.type & 8) && !(desc.type & 2))
685 			goto bad;
686 		lim = desc_limit_scaled(&desc);
687 		if ((desc.type & 8) || !(desc.type & 4)) {
688 			/* expand-up segment */
689 			if (addr.ea > lim || (u32)(addr.ea + size - 1) > lim)
690 				goto bad;
691 		} else {
692 			/* expand-down segment */
693 			if (addr.ea <= lim || (u32)(addr.ea + size - 1) <= lim)
694 				goto bad;
695 			lim = desc.d ? 0xffffffff : 0xffff;
696 			if (addr.ea > lim || (u32)(addr.ea + size - 1) > lim)
697 				goto bad;
698 		}
699 		cpl = ctxt->ops->cpl(ctxt);
700 		if (ctxt->mode == X86EMUL_MODE_REAL)
701 			rpl = 0;
702 		else
703 			rpl = sel & 3;
704 		cpl = max(cpl, rpl);
705 		if (!(desc.type & 8)) {
706 			/* data segment */
707 			if (cpl > desc.dpl)
708 				goto bad;
709 		} else if ((desc.type & 8) && !(desc.type & 4)) {
710 			/* nonconforming code segment */
711 			if (cpl != desc.dpl)
712 				goto bad;
713 		} else if ((desc.type & 8) && (desc.type & 4)) {
714 			/* conforming code segment */
715 			if (cpl < desc.dpl)
716 				goto bad;
717 		}
718 		break;
719 	}
720 	if (fetch ? ctxt->mode != X86EMUL_MODE_PROT64 : ctxt->ad_bytes != 8)
721 		la &= (u32)-1;
722 	if (insn_aligned(ctxt, size) && ((la & (size - 1)) != 0))
723 		return emulate_gp(ctxt, 0);
724 	*linear = la;
725 	return X86EMUL_CONTINUE;
726 bad:
727 	if (addr.seg == VCPU_SREG_SS)
728 		return emulate_ss(ctxt, sel);
729 	else
730 		return emulate_gp(ctxt, sel);
731 }
732 
733 static int linearize(struct x86_emulate_ctxt *ctxt,
734 		     struct segmented_address addr,
735 		     unsigned size, bool write,
736 		     ulong *linear)
737 {
738 	return __linearize(ctxt, addr, size, write, false, linear);
739 }
740 
741 
742 static int segmented_read_std(struct x86_emulate_ctxt *ctxt,
743 			      struct segmented_address addr,
744 			      void *data,
745 			      unsigned size)
746 {
747 	int rc;
748 	ulong linear;
749 
750 	rc = linearize(ctxt, addr, size, false, &linear);
751 	if (rc != X86EMUL_CONTINUE)
752 		return rc;
753 	return ctxt->ops->read_std(ctxt, linear, data, size, &ctxt->exception);
754 }
755 
756 /*
757  * Fetch the next byte of the instruction being emulated which is pointed to
758  * by ctxt->_eip, then increment ctxt->_eip.
759  *
760  * Also prefetch the remaining bytes of the instruction without crossing page
761  * boundary if they are not in fetch_cache yet.
762  */
763 static int do_insn_fetch_byte(struct x86_emulate_ctxt *ctxt, u8 *dest)
764 {
765 	struct fetch_cache *fc = &ctxt->fetch;
766 	int rc;
767 	int size, cur_size;
768 
769 	if (ctxt->_eip == fc->end) {
770 		unsigned long linear;
771 		struct segmented_address addr = { .seg = VCPU_SREG_CS,
772 						  .ea  = ctxt->_eip };
773 		cur_size = fc->end - fc->start;
774 		size = min(15UL - cur_size,
775 			   PAGE_SIZE - offset_in_page(ctxt->_eip));
776 		rc = __linearize(ctxt, addr, size, false, true, &linear);
777 		if (unlikely(rc != X86EMUL_CONTINUE))
778 			return rc;
779 		rc = ctxt->ops->fetch(ctxt, linear, fc->data + cur_size,
780 				      size, &ctxt->exception);
781 		if (unlikely(rc != X86EMUL_CONTINUE))
782 			return rc;
783 		fc->end += size;
784 	}
785 	*dest = fc->data[ctxt->_eip - fc->start];
786 	ctxt->_eip++;
787 	return X86EMUL_CONTINUE;
788 }
789 
790 static int do_insn_fetch(struct x86_emulate_ctxt *ctxt,
791 			 void *dest, unsigned size)
792 {
793 	int rc;
794 
795 	/* x86 instructions are limited to 15 bytes. */
796 	if (unlikely(ctxt->_eip + size - ctxt->eip > 15))
797 		return X86EMUL_UNHANDLEABLE;
798 	while (size--) {
799 		rc = do_insn_fetch_byte(ctxt, dest++);
800 		if (rc != X86EMUL_CONTINUE)
801 			return rc;
802 	}
803 	return X86EMUL_CONTINUE;
804 }
805 
806 /* Fetch next part of the instruction being emulated. */
807 #define insn_fetch(_type, _ctxt)					\
808 ({	unsigned long _x;						\
809 	rc = do_insn_fetch(_ctxt, &_x, sizeof(_type));			\
810 	if (rc != X86EMUL_CONTINUE)					\
811 		goto done;						\
812 	(_type)_x;							\
813 })
814 
815 #define insn_fetch_arr(_arr, _size, _ctxt)				\
816 ({	rc = do_insn_fetch(_ctxt, _arr, (_size));			\
817 	if (rc != X86EMUL_CONTINUE)					\
818 		goto done;						\
819 })
820 
821 /*
822  * Given the 'reg' portion of a ModRM byte, and a register block, return a
823  * pointer into the block that addresses the relevant register.
824  * @highbyte_regs specifies whether to decode AH,CH,DH,BH.
825  */
826 static void *decode_register(struct x86_emulate_ctxt *ctxt, u8 modrm_reg,
827 			     int highbyte_regs)
828 {
829 	void *p;
830 
831 	if (highbyte_regs && modrm_reg >= 4 && modrm_reg < 8)
832 		p = (unsigned char *)reg_rmw(ctxt, modrm_reg & 3) + 1;
833 	else
834 		p = reg_rmw(ctxt, modrm_reg);
835 	return p;
836 }
837 
838 static int read_descriptor(struct x86_emulate_ctxt *ctxt,
839 			   struct segmented_address addr,
840 			   u16 *size, unsigned long *address, int op_bytes)
841 {
842 	int rc;
843 
844 	if (op_bytes == 2)
845 		op_bytes = 3;
846 	*address = 0;
847 	rc = segmented_read_std(ctxt, addr, size, 2);
848 	if (rc != X86EMUL_CONTINUE)
849 		return rc;
850 	addr.ea += 2;
851 	rc = segmented_read_std(ctxt, addr, address, op_bytes);
852 	return rc;
853 }
854 
855 static int test_cc(unsigned int condition, unsigned int flags)
856 {
857 	int rc = 0;
858 
859 	switch ((condition & 15) >> 1) {
860 	case 0: /* o */
861 		rc |= (flags & EFLG_OF);
862 		break;
863 	case 1: /* b/c/nae */
864 		rc |= (flags & EFLG_CF);
865 		break;
866 	case 2: /* z/e */
867 		rc |= (flags & EFLG_ZF);
868 		break;
869 	case 3: /* be/na */
870 		rc |= (flags & (EFLG_CF|EFLG_ZF));
871 		break;
872 	case 4: /* s */
873 		rc |= (flags & EFLG_SF);
874 		break;
875 	case 5: /* p/pe */
876 		rc |= (flags & EFLG_PF);
877 		break;
878 	case 7: /* le/ng */
879 		rc |= (flags & EFLG_ZF);
880 		/* fall through */
881 	case 6: /* l/nge */
882 		rc |= (!(flags & EFLG_SF) != !(flags & EFLG_OF));
883 		break;
884 	}
885 
886 	/* Odd condition identifiers (lsb == 1) have inverted sense. */
887 	return (!!rc ^ (condition & 1));
888 }
889 
890 static void fetch_register_operand(struct operand *op)
891 {
892 	switch (op->bytes) {
893 	case 1:
894 		op->val = *(u8 *)op->addr.reg;
895 		break;
896 	case 2:
897 		op->val = *(u16 *)op->addr.reg;
898 		break;
899 	case 4:
900 		op->val = *(u32 *)op->addr.reg;
901 		break;
902 	case 8:
903 		op->val = *(u64 *)op->addr.reg;
904 		break;
905 	}
906 }
907 
908 static void read_sse_reg(struct x86_emulate_ctxt *ctxt, sse128_t *data, int reg)
909 {
910 	ctxt->ops->get_fpu(ctxt);
911 	switch (reg) {
912 	case 0: asm("movdqa %%xmm0, %0" : "=m"(*data)); break;
913 	case 1: asm("movdqa %%xmm1, %0" : "=m"(*data)); break;
914 	case 2: asm("movdqa %%xmm2, %0" : "=m"(*data)); break;
915 	case 3: asm("movdqa %%xmm3, %0" : "=m"(*data)); break;
916 	case 4: asm("movdqa %%xmm4, %0" : "=m"(*data)); break;
917 	case 5: asm("movdqa %%xmm5, %0" : "=m"(*data)); break;
918 	case 6: asm("movdqa %%xmm6, %0" : "=m"(*data)); break;
919 	case 7: asm("movdqa %%xmm7, %0" : "=m"(*data)); break;
920 #ifdef CONFIG_X86_64
921 	case 8: asm("movdqa %%xmm8, %0" : "=m"(*data)); break;
922 	case 9: asm("movdqa %%xmm9, %0" : "=m"(*data)); break;
923 	case 10: asm("movdqa %%xmm10, %0" : "=m"(*data)); break;
924 	case 11: asm("movdqa %%xmm11, %0" : "=m"(*data)); break;
925 	case 12: asm("movdqa %%xmm12, %0" : "=m"(*data)); break;
926 	case 13: asm("movdqa %%xmm13, %0" : "=m"(*data)); break;
927 	case 14: asm("movdqa %%xmm14, %0" : "=m"(*data)); break;
928 	case 15: asm("movdqa %%xmm15, %0" : "=m"(*data)); break;
929 #endif
930 	default: BUG();
931 	}
932 	ctxt->ops->put_fpu(ctxt);
933 }
934 
935 static void write_sse_reg(struct x86_emulate_ctxt *ctxt, sse128_t *data,
936 			  int reg)
937 {
938 	ctxt->ops->get_fpu(ctxt);
939 	switch (reg) {
940 	case 0: asm("movdqa %0, %%xmm0" : : "m"(*data)); break;
941 	case 1: asm("movdqa %0, %%xmm1" : : "m"(*data)); break;
942 	case 2: asm("movdqa %0, %%xmm2" : : "m"(*data)); break;
943 	case 3: asm("movdqa %0, %%xmm3" : : "m"(*data)); break;
944 	case 4: asm("movdqa %0, %%xmm4" : : "m"(*data)); break;
945 	case 5: asm("movdqa %0, %%xmm5" : : "m"(*data)); break;
946 	case 6: asm("movdqa %0, %%xmm6" : : "m"(*data)); break;
947 	case 7: asm("movdqa %0, %%xmm7" : : "m"(*data)); break;
948 #ifdef CONFIG_X86_64
949 	case 8: asm("movdqa %0, %%xmm8" : : "m"(*data)); break;
950 	case 9: asm("movdqa %0, %%xmm9" : : "m"(*data)); break;
951 	case 10: asm("movdqa %0, %%xmm10" : : "m"(*data)); break;
952 	case 11: asm("movdqa %0, %%xmm11" : : "m"(*data)); break;
953 	case 12: asm("movdqa %0, %%xmm12" : : "m"(*data)); break;
954 	case 13: asm("movdqa %0, %%xmm13" : : "m"(*data)); break;
955 	case 14: asm("movdqa %0, %%xmm14" : : "m"(*data)); break;
956 	case 15: asm("movdqa %0, %%xmm15" : : "m"(*data)); break;
957 #endif
958 	default: BUG();
959 	}
960 	ctxt->ops->put_fpu(ctxt);
961 }
962 
963 static void read_mmx_reg(struct x86_emulate_ctxt *ctxt, u64 *data, int reg)
964 {
965 	ctxt->ops->get_fpu(ctxt);
966 	switch (reg) {
967 	case 0: asm("movq %%mm0, %0" : "=m"(*data)); break;
968 	case 1: asm("movq %%mm1, %0" : "=m"(*data)); break;
969 	case 2: asm("movq %%mm2, %0" : "=m"(*data)); break;
970 	case 3: asm("movq %%mm3, %0" : "=m"(*data)); break;
971 	case 4: asm("movq %%mm4, %0" : "=m"(*data)); break;
972 	case 5: asm("movq %%mm5, %0" : "=m"(*data)); break;
973 	case 6: asm("movq %%mm6, %0" : "=m"(*data)); break;
974 	case 7: asm("movq %%mm7, %0" : "=m"(*data)); break;
975 	default: BUG();
976 	}
977 	ctxt->ops->put_fpu(ctxt);
978 }
979 
980 static void write_mmx_reg(struct x86_emulate_ctxt *ctxt, u64 *data, int reg)
981 {
982 	ctxt->ops->get_fpu(ctxt);
983 	switch (reg) {
984 	case 0: asm("movq %0, %%mm0" : : "m"(*data)); break;
985 	case 1: asm("movq %0, %%mm1" : : "m"(*data)); break;
986 	case 2: asm("movq %0, %%mm2" : : "m"(*data)); break;
987 	case 3: asm("movq %0, %%mm3" : : "m"(*data)); break;
988 	case 4: asm("movq %0, %%mm4" : : "m"(*data)); break;
989 	case 5: asm("movq %0, %%mm5" : : "m"(*data)); break;
990 	case 6: asm("movq %0, %%mm6" : : "m"(*data)); break;
991 	case 7: asm("movq %0, %%mm7" : : "m"(*data)); break;
992 	default: BUG();
993 	}
994 	ctxt->ops->put_fpu(ctxt);
995 }
996 
997 static void decode_register_operand(struct x86_emulate_ctxt *ctxt,
998 				    struct operand *op)
999 {
1000 	unsigned reg = ctxt->modrm_reg;
1001 	int highbyte_regs = ctxt->rex_prefix == 0;
1002 
1003 	if (!(ctxt->d & ModRM))
1004 		reg = (ctxt->b & 7) | ((ctxt->rex_prefix & 1) << 3);
1005 
1006 	if (ctxt->d & Sse) {
1007 		op->type = OP_XMM;
1008 		op->bytes = 16;
1009 		op->addr.xmm = reg;
1010 		read_sse_reg(ctxt, &op->vec_val, reg);
1011 		return;
1012 	}
1013 	if (ctxt->d & Mmx) {
1014 		reg &= 7;
1015 		op->type = OP_MM;
1016 		op->bytes = 8;
1017 		op->addr.mm = reg;
1018 		return;
1019 	}
1020 
1021 	op->type = OP_REG;
1022 	if (ctxt->d & ByteOp) {
1023 		op->addr.reg = decode_register(ctxt, reg, highbyte_regs);
1024 		op->bytes = 1;
1025 	} else {
1026 		op->addr.reg = decode_register(ctxt, reg, 0);
1027 		op->bytes = ctxt->op_bytes;
1028 	}
1029 	fetch_register_operand(op);
1030 	op->orig_val = op->val;
1031 }
1032 
1033 static void adjust_modrm_seg(struct x86_emulate_ctxt *ctxt, int base_reg)
1034 {
1035 	if (base_reg == VCPU_REGS_RSP || base_reg == VCPU_REGS_RBP)
1036 		ctxt->modrm_seg = VCPU_SREG_SS;
1037 }
1038 
1039 static int decode_modrm(struct x86_emulate_ctxt *ctxt,
1040 			struct operand *op)
1041 {
1042 	u8 sib;
1043 	int index_reg = 0, base_reg = 0, scale;
1044 	int rc = X86EMUL_CONTINUE;
1045 	ulong modrm_ea = 0;
1046 
1047 	if (ctxt->rex_prefix) {
1048 		ctxt->modrm_reg = (ctxt->rex_prefix & 4) << 1;	/* REX.R */
1049 		index_reg = (ctxt->rex_prefix & 2) << 2; /* REX.X */
1050 		ctxt->modrm_rm = base_reg = (ctxt->rex_prefix & 1) << 3; /* REG.B */
1051 	}
1052 
1053 	ctxt->modrm_mod |= (ctxt->modrm & 0xc0) >> 6;
1054 	ctxt->modrm_reg |= (ctxt->modrm & 0x38) >> 3;
1055 	ctxt->modrm_rm |= (ctxt->modrm & 0x07);
1056 	ctxt->modrm_seg = VCPU_SREG_DS;
1057 
1058 	if (ctxt->modrm_mod == 3) {
1059 		op->type = OP_REG;
1060 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
1061 		op->addr.reg = decode_register(ctxt, ctxt->modrm_rm, ctxt->d & ByteOp);
1062 		if (ctxt->d & Sse) {
1063 			op->type = OP_XMM;
1064 			op->bytes = 16;
1065 			op->addr.xmm = ctxt->modrm_rm;
1066 			read_sse_reg(ctxt, &op->vec_val, ctxt->modrm_rm);
1067 			return rc;
1068 		}
1069 		if (ctxt->d & Mmx) {
1070 			op->type = OP_MM;
1071 			op->bytes = 8;
1072 			op->addr.xmm = ctxt->modrm_rm & 7;
1073 			return rc;
1074 		}
1075 		fetch_register_operand(op);
1076 		return rc;
1077 	}
1078 
1079 	op->type = OP_MEM;
1080 
1081 	if (ctxt->ad_bytes == 2) {
1082 		unsigned bx = reg_read(ctxt, VCPU_REGS_RBX);
1083 		unsigned bp = reg_read(ctxt, VCPU_REGS_RBP);
1084 		unsigned si = reg_read(ctxt, VCPU_REGS_RSI);
1085 		unsigned di = reg_read(ctxt, VCPU_REGS_RDI);
1086 
1087 		/* 16-bit ModR/M decode. */
1088 		switch (ctxt->modrm_mod) {
1089 		case 0:
1090 			if (ctxt->modrm_rm == 6)
1091 				modrm_ea += insn_fetch(u16, ctxt);
1092 			break;
1093 		case 1:
1094 			modrm_ea += insn_fetch(s8, ctxt);
1095 			break;
1096 		case 2:
1097 			modrm_ea += insn_fetch(u16, ctxt);
1098 			break;
1099 		}
1100 		switch (ctxt->modrm_rm) {
1101 		case 0:
1102 			modrm_ea += bx + si;
1103 			break;
1104 		case 1:
1105 			modrm_ea += bx + di;
1106 			break;
1107 		case 2:
1108 			modrm_ea += bp + si;
1109 			break;
1110 		case 3:
1111 			modrm_ea += bp + di;
1112 			break;
1113 		case 4:
1114 			modrm_ea += si;
1115 			break;
1116 		case 5:
1117 			modrm_ea += di;
1118 			break;
1119 		case 6:
1120 			if (ctxt->modrm_mod != 0)
1121 				modrm_ea += bp;
1122 			break;
1123 		case 7:
1124 			modrm_ea += bx;
1125 			break;
1126 		}
1127 		if (ctxt->modrm_rm == 2 || ctxt->modrm_rm == 3 ||
1128 		    (ctxt->modrm_rm == 6 && ctxt->modrm_mod != 0))
1129 			ctxt->modrm_seg = VCPU_SREG_SS;
1130 		modrm_ea = (u16)modrm_ea;
1131 	} else {
1132 		/* 32/64-bit ModR/M decode. */
1133 		if ((ctxt->modrm_rm & 7) == 4) {
1134 			sib = insn_fetch(u8, ctxt);
1135 			index_reg |= (sib >> 3) & 7;
1136 			base_reg |= sib & 7;
1137 			scale = sib >> 6;
1138 
1139 			if ((base_reg & 7) == 5 && ctxt->modrm_mod == 0)
1140 				modrm_ea += insn_fetch(s32, ctxt);
1141 			else {
1142 				modrm_ea += reg_read(ctxt, base_reg);
1143 				adjust_modrm_seg(ctxt, base_reg);
1144 			}
1145 			if (index_reg != 4)
1146 				modrm_ea += reg_read(ctxt, index_reg) << scale;
1147 		} else if ((ctxt->modrm_rm & 7) == 5 && ctxt->modrm_mod == 0) {
1148 			if (ctxt->mode == X86EMUL_MODE_PROT64)
1149 				ctxt->rip_relative = 1;
1150 		} else {
1151 			base_reg = ctxt->modrm_rm;
1152 			modrm_ea += reg_read(ctxt, base_reg);
1153 			adjust_modrm_seg(ctxt, base_reg);
1154 		}
1155 		switch (ctxt->modrm_mod) {
1156 		case 0:
1157 			if (ctxt->modrm_rm == 5)
1158 				modrm_ea += insn_fetch(s32, ctxt);
1159 			break;
1160 		case 1:
1161 			modrm_ea += insn_fetch(s8, ctxt);
1162 			break;
1163 		case 2:
1164 			modrm_ea += insn_fetch(s32, ctxt);
1165 			break;
1166 		}
1167 	}
1168 	op->addr.mem.ea = modrm_ea;
1169 done:
1170 	return rc;
1171 }
1172 
1173 static int decode_abs(struct x86_emulate_ctxt *ctxt,
1174 		      struct operand *op)
1175 {
1176 	int rc = X86EMUL_CONTINUE;
1177 
1178 	op->type = OP_MEM;
1179 	switch (ctxt->ad_bytes) {
1180 	case 2:
1181 		op->addr.mem.ea = insn_fetch(u16, ctxt);
1182 		break;
1183 	case 4:
1184 		op->addr.mem.ea = insn_fetch(u32, ctxt);
1185 		break;
1186 	case 8:
1187 		op->addr.mem.ea = insn_fetch(u64, ctxt);
1188 		break;
1189 	}
1190 done:
1191 	return rc;
1192 }
1193 
1194 static void fetch_bit_operand(struct x86_emulate_ctxt *ctxt)
1195 {
1196 	long sv = 0, mask;
1197 
1198 	if (ctxt->dst.type == OP_MEM && ctxt->src.type == OP_REG) {
1199 		mask = ~(ctxt->dst.bytes * 8 - 1);
1200 
1201 		if (ctxt->src.bytes == 2)
1202 			sv = (s16)ctxt->src.val & (s16)mask;
1203 		else if (ctxt->src.bytes == 4)
1204 			sv = (s32)ctxt->src.val & (s32)mask;
1205 
1206 		ctxt->dst.addr.mem.ea += (sv >> 3);
1207 	}
1208 
1209 	/* only subword offset */
1210 	ctxt->src.val &= (ctxt->dst.bytes << 3) - 1;
1211 }
1212 
1213 static int read_emulated(struct x86_emulate_ctxt *ctxt,
1214 			 unsigned long addr, void *dest, unsigned size)
1215 {
1216 	int rc;
1217 	struct read_cache *mc = &ctxt->mem_read;
1218 
1219 	if (mc->pos < mc->end)
1220 		goto read_cached;
1221 
1222 	WARN_ON((mc->end + size) >= sizeof(mc->data));
1223 
1224 	rc = ctxt->ops->read_emulated(ctxt, addr, mc->data + mc->end, size,
1225 				      &ctxt->exception);
1226 	if (rc != X86EMUL_CONTINUE)
1227 		return rc;
1228 
1229 	mc->end += size;
1230 
1231 read_cached:
1232 	memcpy(dest, mc->data + mc->pos, size);
1233 	mc->pos += size;
1234 	return X86EMUL_CONTINUE;
1235 }
1236 
1237 static int segmented_read(struct x86_emulate_ctxt *ctxt,
1238 			  struct segmented_address addr,
1239 			  void *data,
1240 			  unsigned size)
1241 {
1242 	int rc;
1243 	ulong linear;
1244 
1245 	rc = linearize(ctxt, addr, size, false, &linear);
1246 	if (rc != X86EMUL_CONTINUE)
1247 		return rc;
1248 	return read_emulated(ctxt, linear, data, size);
1249 }
1250 
1251 static int segmented_write(struct x86_emulate_ctxt *ctxt,
1252 			   struct segmented_address addr,
1253 			   const void *data,
1254 			   unsigned size)
1255 {
1256 	int rc;
1257 	ulong linear;
1258 
1259 	rc = linearize(ctxt, addr, size, true, &linear);
1260 	if (rc != X86EMUL_CONTINUE)
1261 		return rc;
1262 	return ctxt->ops->write_emulated(ctxt, linear, data, size,
1263 					 &ctxt->exception);
1264 }
1265 
1266 static int segmented_cmpxchg(struct x86_emulate_ctxt *ctxt,
1267 			     struct segmented_address addr,
1268 			     const void *orig_data, const void *data,
1269 			     unsigned size)
1270 {
1271 	int rc;
1272 	ulong linear;
1273 
1274 	rc = linearize(ctxt, addr, size, true, &linear);
1275 	if (rc != X86EMUL_CONTINUE)
1276 		return rc;
1277 	return ctxt->ops->cmpxchg_emulated(ctxt, linear, orig_data, data,
1278 					   size, &ctxt->exception);
1279 }
1280 
1281 static int pio_in_emulated(struct x86_emulate_ctxt *ctxt,
1282 			   unsigned int size, unsigned short port,
1283 			   void *dest)
1284 {
1285 	struct read_cache *rc = &ctxt->io_read;
1286 
1287 	if (rc->pos == rc->end) { /* refill pio read ahead */
1288 		unsigned int in_page, n;
1289 		unsigned int count = ctxt->rep_prefix ?
1290 			address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) : 1;
1291 		in_page = (ctxt->eflags & EFLG_DF) ?
1292 			offset_in_page(reg_read(ctxt, VCPU_REGS_RDI)) :
1293 			PAGE_SIZE - offset_in_page(reg_read(ctxt, VCPU_REGS_RDI));
1294 		n = min(min(in_page, (unsigned int)sizeof(rc->data)) / size,
1295 			count);
1296 		if (n == 0)
1297 			n = 1;
1298 		rc->pos = rc->end = 0;
1299 		if (!ctxt->ops->pio_in_emulated(ctxt, size, port, rc->data, n))
1300 			return 0;
1301 		rc->end = n * size;
1302 	}
1303 
1304 	if (ctxt->rep_prefix && !(ctxt->eflags & EFLG_DF)) {
1305 		ctxt->dst.data = rc->data + rc->pos;
1306 		ctxt->dst.type = OP_MEM_STR;
1307 		ctxt->dst.count = (rc->end - rc->pos) / size;
1308 		rc->pos = rc->end;
1309 	} else {
1310 		memcpy(dest, rc->data + rc->pos, size);
1311 		rc->pos += size;
1312 	}
1313 	return 1;
1314 }
1315 
1316 static int read_interrupt_descriptor(struct x86_emulate_ctxt *ctxt,
1317 				     u16 index, struct desc_struct *desc)
1318 {
1319 	struct desc_ptr dt;
1320 	ulong addr;
1321 
1322 	ctxt->ops->get_idt(ctxt, &dt);
1323 
1324 	if (dt.size < index * 8 + 7)
1325 		return emulate_gp(ctxt, index << 3 | 0x2);
1326 
1327 	addr = dt.address + index * 8;
1328 	return ctxt->ops->read_std(ctxt, addr, desc, sizeof *desc,
1329 				   &ctxt->exception);
1330 }
1331 
1332 static void get_descriptor_table_ptr(struct x86_emulate_ctxt *ctxt,
1333 				     u16 selector, struct desc_ptr *dt)
1334 {
1335 	const struct x86_emulate_ops *ops = ctxt->ops;
1336 
1337 	if (selector & 1 << 2) {
1338 		struct desc_struct desc;
1339 		u16 sel;
1340 
1341 		memset (dt, 0, sizeof *dt);
1342 		if (!ops->get_segment(ctxt, &sel, &desc, NULL, VCPU_SREG_LDTR))
1343 			return;
1344 
1345 		dt->size = desc_limit_scaled(&desc); /* what if limit > 65535? */
1346 		dt->address = get_desc_base(&desc);
1347 	} else
1348 		ops->get_gdt(ctxt, dt);
1349 }
1350 
1351 /* allowed just for 8 bytes segments */
1352 static int read_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1353 				   u16 selector, struct desc_struct *desc,
1354 				   ulong *desc_addr_p)
1355 {
1356 	struct desc_ptr dt;
1357 	u16 index = selector >> 3;
1358 	ulong addr;
1359 
1360 	get_descriptor_table_ptr(ctxt, selector, &dt);
1361 
1362 	if (dt.size < index * 8 + 7)
1363 		return emulate_gp(ctxt, selector & 0xfffc);
1364 
1365 	*desc_addr_p = addr = dt.address + index * 8;
1366 	return ctxt->ops->read_std(ctxt, addr, desc, sizeof *desc,
1367 				   &ctxt->exception);
1368 }
1369 
1370 /* allowed just for 8 bytes segments */
1371 static int write_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1372 				    u16 selector, struct desc_struct *desc)
1373 {
1374 	struct desc_ptr dt;
1375 	u16 index = selector >> 3;
1376 	ulong addr;
1377 
1378 	get_descriptor_table_ptr(ctxt, selector, &dt);
1379 
1380 	if (dt.size < index * 8 + 7)
1381 		return emulate_gp(ctxt, selector & 0xfffc);
1382 
1383 	addr = dt.address + index * 8;
1384 	return ctxt->ops->write_std(ctxt, addr, desc, sizeof *desc,
1385 				    &ctxt->exception);
1386 }
1387 
1388 /* Does not support long mode */
1389 static int load_segment_descriptor(struct x86_emulate_ctxt *ctxt,
1390 				   u16 selector, int seg)
1391 {
1392 	struct desc_struct seg_desc, old_desc;
1393 	u8 dpl, rpl, cpl;
1394 	unsigned err_vec = GP_VECTOR;
1395 	u32 err_code = 0;
1396 	bool null_selector = !(selector & ~0x3); /* 0000-0003 are null */
1397 	ulong desc_addr;
1398 	int ret;
1399 	u16 dummy;
1400 
1401 	memset(&seg_desc, 0, sizeof seg_desc);
1402 
1403 	if ((seg <= VCPU_SREG_GS && ctxt->mode == X86EMUL_MODE_VM86)
1404 	    || ctxt->mode == X86EMUL_MODE_REAL) {
1405 		/* set real mode segment descriptor */
1406 		ctxt->ops->get_segment(ctxt, &dummy, &seg_desc, NULL, seg);
1407 		set_desc_base(&seg_desc, selector << 4);
1408 		goto load;
1409 	}
1410 
1411 	rpl = selector & 3;
1412 	cpl = ctxt->ops->cpl(ctxt);
1413 
1414 	/* NULL selector is not valid for TR, CS and SS (except for long mode) */
1415 	if ((seg == VCPU_SREG_CS
1416 	     || (seg == VCPU_SREG_SS
1417 		 && (ctxt->mode != X86EMUL_MODE_PROT64 || rpl != cpl))
1418 	     || seg == VCPU_SREG_TR)
1419 	    && null_selector)
1420 		goto exception;
1421 
1422 	/* TR should be in GDT only */
1423 	if (seg == VCPU_SREG_TR && (selector & (1 << 2)))
1424 		goto exception;
1425 
1426 	if (null_selector) /* for NULL selector skip all following checks */
1427 		goto load;
1428 
1429 	ret = read_segment_descriptor(ctxt, selector, &seg_desc, &desc_addr);
1430 	if (ret != X86EMUL_CONTINUE)
1431 		return ret;
1432 
1433 	err_code = selector & 0xfffc;
1434 	err_vec = GP_VECTOR;
1435 
1436 	/* can't load system descriptor into segment selector */
1437 	if (seg <= VCPU_SREG_GS && !seg_desc.s)
1438 		goto exception;
1439 
1440 	if (!seg_desc.p) {
1441 		err_vec = (seg == VCPU_SREG_SS) ? SS_VECTOR : NP_VECTOR;
1442 		goto exception;
1443 	}
1444 
1445 	dpl = seg_desc.dpl;
1446 
1447 	switch (seg) {
1448 	case VCPU_SREG_SS:
1449 		/*
1450 		 * segment is not a writable data segment or segment
1451 		 * selector's RPL != CPL or segment selector's RPL != CPL
1452 		 */
1453 		if (rpl != cpl || (seg_desc.type & 0xa) != 0x2 || dpl != cpl)
1454 			goto exception;
1455 		break;
1456 	case VCPU_SREG_CS:
1457 		if (!(seg_desc.type & 8))
1458 			goto exception;
1459 
1460 		if (seg_desc.type & 4) {
1461 			/* conforming */
1462 			if (dpl > cpl)
1463 				goto exception;
1464 		} else {
1465 			/* nonconforming */
1466 			if (rpl > cpl || dpl != cpl)
1467 				goto exception;
1468 		}
1469 		/* CS(RPL) <- CPL */
1470 		selector = (selector & 0xfffc) | cpl;
1471 		break;
1472 	case VCPU_SREG_TR:
1473 		if (seg_desc.s || (seg_desc.type != 1 && seg_desc.type != 9))
1474 			goto exception;
1475 		old_desc = seg_desc;
1476 		seg_desc.type |= 2; /* busy */
1477 		ret = ctxt->ops->cmpxchg_emulated(ctxt, desc_addr, &old_desc, &seg_desc,
1478 						  sizeof(seg_desc), &ctxt->exception);
1479 		if (ret != X86EMUL_CONTINUE)
1480 			return ret;
1481 		break;
1482 	case VCPU_SREG_LDTR:
1483 		if (seg_desc.s || seg_desc.type != 2)
1484 			goto exception;
1485 		break;
1486 	default: /*  DS, ES, FS, or GS */
1487 		/*
1488 		 * segment is not a data or readable code segment or
1489 		 * ((segment is a data or nonconforming code segment)
1490 		 * and (both RPL and CPL > DPL))
1491 		 */
1492 		if ((seg_desc.type & 0xa) == 0x8 ||
1493 		    (((seg_desc.type & 0xc) != 0xc) &&
1494 		     (rpl > dpl && cpl > dpl)))
1495 			goto exception;
1496 		break;
1497 	}
1498 
1499 	if (seg_desc.s) {
1500 		/* mark segment as accessed */
1501 		seg_desc.type |= 1;
1502 		ret = write_segment_descriptor(ctxt, selector, &seg_desc);
1503 		if (ret != X86EMUL_CONTINUE)
1504 			return ret;
1505 	}
1506 load:
1507 	ctxt->ops->set_segment(ctxt, selector, &seg_desc, 0, seg);
1508 	return X86EMUL_CONTINUE;
1509 exception:
1510 	emulate_exception(ctxt, err_vec, err_code, true);
1511 	return X86EMUL_PROPAGATE_FAULT;
1512 }
1513 
1514 static void write_register_operand(struct operand *op)
1515 {
1516 	/* The 4-byte case *is* correct: in 64-bit mode we zero-extend. */
1517 	switch (op->bytes) {
1518 	case 1:
1519 		*(u8 *)op->addr.reg = (u8)op->val;
1520 		break;
1521 	case 2:
1522 		*(u16 *)op->addr.reg = (u16)op->val;
1523 		break;
1524 	case 4:
1525 		*op->addr.reg = (u32)op->val;
1526 		break;	/* 64b: zero-extend */
1527 	case 8:
1528 		*op->addr.reg = op->val;
1529 		break;
1530 	}
1531 }
1532 
1533 static int writeback(struct x86_emulate_ctxt *ctxt)
1534 {
1535 	int rc;
1536 
1537 	switch (ctxt->dst.type) {
1538 	case OP_REG:
1539 		write_register_operand(&ctxt->dst);
1540 		break;
1541 	case OP_MEM:
1542 		if (ctxt->lock_prefix)
1543 			rc = segmented_cmpxchg(ctxt,
1544 					       ctxt->dst.addr.mem,
1545 					       &ctxt->dst.orig_val,
1546 					       &ctxt->dst.val,
1547 					       ctxt->dst.bytes);
1548 		else
1549 			rc = segmented_write(ctxt,
1550 					     ctxt->dst.addr.mem,
1551 					     &ctxt->dst.val,
1552 					     ctxt->dst.bytes);
1553 		if (rc != X86EMUL_CONTINUE)
1554 			return rc;
1555 		break;
1556 	case OP_MEM_STR:
1557 		rc = segmented_write(ctxt,
1558 				ctxt->dst.addr.mem,
1559 				ctxt->dst.data,
1560 				ctxt->dst.bytes * ctxt->dst.count);
1561 		if (rc != X86EMUL_CONTINUE)
1562 			return rc;
1563 		break;
1564 	case OP_XMM:
1565 		write_sse_reg(ctxt, &ctxt->dst.vec_val, ctxt->dst.addr.xmm);
1566 		break;
1567 	case OP_MM:
1568 		write_mmx_reg(ctxt, &ctxt->dst.mm_val, ctxt->dst.addr.mm);
1569 		break;
1570 	case OP_NONE:
1571 		/* no writeback */
1572 		break;
1573 	default:
1574 		break;
1575 	}
1576 	return X86EMUL_CONTINUE;
1577 }
1578 
1579 static int push(struct x86_emulate_ctxt *ctxt, void *data, int bytes)
1580 {
1581 	struct segmented_address addr;
1582 
1583 	rsp_increment(ctxt, -bytes);
1584 	addr.ea = reg_read(ctxt, VCPU_REGS_RSP) & stack_mask(ctxt);
1585 	addr.seg = VCPU_SREG_SS;
1586 
1587 	return segmented_write(ctxt, addr, data, bytes);
1588 }
1589 
1590 static int em_push(struct x86_emulate_ctxt *ctxt)
1591 {
1592 	/* Disable writeback. */
1593 	ctxt->dst.type = OP_NONE;
1594 	return push(ctxt, &ctxt->src.val, ctxt->op_bytes);
1595 }
1596 
1597 static int emulate_pop(struct x86_emulate_ctxt *ctxt,
1598 		       void *dest, int len)
1599 {
1600 	int rc;
1601 	struct segmented_address addr;
1602 
1603 	addr.ea = reg_read(ctxt, VCPU_REGS_RSP) & stack_mask(ctxt);
1604 	addr.seg = VCPU_SREG_SS;
1605 	rc = segmented_read(ctxt, addr, dest, len);
1606 	if (rc != X86EMUL_CONTINUE)
1607 		return rc;
1608 
1609 	rsp_increment(ctxt, len);
1610 	return rc;
1611 }
1612 
1613 static int em_pop(struct x86_emulate_ctxt *ctxt)
1614 {
1615 	return emulate_pop(ctxt, &ctxt->dst.val, ctxt->op_bytes);
1616 }
1617 
1618 static int emulate_popf(struct x86_emulate_ctxt *ctxt,
1619 			void *dest, int len)
1620 {
1621 	int rc;
1622 	unsigned long val, change_mask;
1623 	int iopl = (ctxt->eflags & X86_EFLAGS_IOPL) >> IOPL_SHIFT;
1624 	int cpl = ctxt->ops->cpl(ctxt);
1625 
1626 	rc = emulate_pop(ctxt, &val, len);
1627 	if (rc != X86EMUL_CONTINUE)
1628 		return rc;
1629 
1630 	change_mask = EFLG_CF | EFLG_PF | EFLG_AF | EFLG_ZF | EFLG_SF | EFLG_OF
1631 		| EFLG_TF | EFLG_DF | EFLG_NT | EFLG_RF | EFLG_AC | EFLG_ID;
1632 
1633 	switch(ctxt->mode) {
1634 	case X86EMUL_MODE_PROT64:
1635 	case X86EMUL_MODE_PROT32:
1636 	case X86EMUL_MODE_PROT16:
1637 		if (cpl == 0)
1638 			change_mask |= EFLG_IOPL;
1639 		if (cpl <= iopl)
1640 			change_mask |= EFLG_IF;
1641 		break;
1642 	case X86EMUL_MODE_VM86:
1643 		if (iopl < 3)
1644 			return emulate_gp(ctxt, 0);
1645 		change_mask |= EFLG_IF;
1646 		break;
1647 	default: /* real mode */
1648 		change_mask |= (EFLG_IOPL | EFLG_IF);
1649 		break;
1650 	}
1651 
1652 	*(unsigned long *)dest =
1653 		(ctxt->eflags & ~change_mask) | (val & change_mask);
1654 
1655 	return rc;
1656 }
1657 
1658 static int em_popf(struct x86_emulate_ctxt *ctxt)
1659 {
1660 	ctxt->dst.type = OP_REG;
1661 	ctxt->dst.addr.reg = &ctxt->eflags;
1662 	ctxt->dst.bytes = ctxt->op_bytes;
1663 	return emulate_popf(ctxt, &ctxt->dst.val, ctxt->op_bytes);
1664 }
1665 
1666 static int em_enter(struct x86_emulate_ctxt *ctxt)
1667 {
1668 	int rc;
1669 	unsigned frame_size = ctxt->src.val;
1670 	unsigned nesting_level = ctxt->src2.val & 31;
1671 	ulong rbp;
1672 
1673 	if (nesting_level)
1674 		return X86EMUL_UNHANDLEABLE;
1675 
1676 	rbp = reg_read(ctxt, VCPU_REGS_RBP);
1677 	rc = push(ctxt, &rbp, stack_size(ctxt));
1678 	if (rc != X86EMUL_CONTINUE)
1679 		return rc;
1680 	assign_masked(reg_rmw(ctxt, VCPU_REGS_RBP), reg_read(ctxt, VCPU_REGS_RSP),
1681 		      stack_mask(ctxt));
1682 	assign_masked(reg_rmw(ctxt, VCPU_REGS_RSP),
1683 		      reg_read(ctxt, VCPU_REGS_RSP) - frame_size,
1684 		      stack_mask(ctxt));
1685 	return X86EMUL_CONTINUE;
1686 }
1687 
1688 static int em_leave(struct x86_emulate_ctxt *ctxt)
1689 {
1690 	assign_masked(reg_rmw(ctxt, VCPU_REGS_RSP), reg_read(ctxt, VCPU_REGS_RBP),
1691 		      stack_mask(ctxt));
1692 	return emulate_pop(ctxt, reg_rmw(ctxt, VCPU_REGS_RBP), ctxt->op_bytes);
1693 }
1694 
1695 static int em_push_sreg(struct x86_emulate_ctxt *ctxt)
1696 {
1697 	int seg = ctxt->src2.val;
1698 
1699 	ctxt->src.val = get_segment_selector(ctxt, seg);
1700 
1701 	return em_push(ctxt);
1702 }
1703 
1704 static int em_pop_sreg(struct x86_emulate_ctxt *ctxt)
1705 {
1706 	int seg = ctxt->src2.val;
1707 	unsigned long selector;
1708 	int rc;
1709 
1710 	rc = emulate_pop(ctxt, &selector, ctxt->op_bytes);
1711 	if (rc != X86EMUL_CONTINUE)
1712 		return rc;
1713 
1714 	rc = load_segment_descriptor(ctxt, (u16)selector, seg);
1715 	return rc;
1716 }
1717 
1718 static int em_pusha(struct x86_emulate_ctxt *ctxt)
1719 {
1720 	unsigned long old_esp = reg_read(ctxt, VCPU_REGS_RSP);
1721 	int rc = X86EMUL_CONTINUE;
1722 	int reg = VCPU_REGS_RAX;
1723 
1724 	while (reg <= VCPU_REGS_RDI) {
1725 		(reg == VCPU_REGS_RSP) ?
1726 		(ctxt->src.val = old_esp) : (ctxt->src.val = reg_read(ctxt, reg));
1727 
1728 		rc = em_push(ctxt);
1729 		if (rc != X86EMUL_CONTINUE)
1730 			return rc;
1731 
1732 		++reg;
1733 	}
1734 
1735 	return rc;
1736 }
1737 
1738 static int em_pushf(struct x86_emulate_ctxt *ctxt)
1739 {
1740 	ctxt->src.val =  (unsigned long)ctxt->eflags;
1741 	return em_push(ctxt);
1742 }
1743 
1744 static int em_popa(struct x86_emulate_ctxt *ctxt)
1745 {
1746 	int rc = X86EMUL_CONTINUE;
1747 	int reg = VCPU_REGS_RDI;
1748 
1749 	while (reg >= VCPU_REGS_RAX) {
1750 		if (reg == VCPU_REGS_RSP) {
1751 			rsp_increment(ctxt, ctxt->op_bytes);
1752 			--reg;
1753 		}
1754 
1755 		rc = emulate_pop(ctxt, reg_rmw(ctxt, reg), ctxt->op_bytes);
1756 		if (rc != X86EMUL_CONTINUE)
1757 			break;
1758 		--reg;
1759 	}
1760 	return rc;
1761 }
1762 
1763 static int __emulate_int_real(struct x86_emulate_ctxt *ctxt, int irq)
1764 {
1765 	const struct x86_emulate_ops *ops = ctxt->ops;
1766 	int rc;
1767 	struct desc_ptr dt;
1768 	gva_t cs_addr;
1769 	gva_t eip_addr;
1770 	u16 cs, eip;
1771 
1772 	/* TODO: Add limit checks */
1773 	ctxt->src.val = ctxt->eflags;
1774 	rc = em_push(ctxt);
1775 	if (rc != X86EMUL_CONTINUE)
1776 		return rc;
1777 
1778 	ctxt->eflags &= ~(EFLG_IF | EFLG_TF | EFLG_AC);
1779 
1780 	ctxt->src.val = get_segment_selector(ctxt, VCPU_SREG_CS);
1781 	rc = em_push(ctxt);
1782 	if (rc != X86EMUL_CONTINUE)
1783 		return rc;
1784 
1785 	ctxt->src.val = ctxt->_eip;
1786 	rc = em_push(ctxt);
1787 	if (rc != X86EMUL_CONTINUE)
1788 		return rc;
1789 
1790 	ops->get_idt(ctxt, &dt);
1791 
1792 	eip_addr = dt.address + (irq << 2);
1793 	cs_addr = dt.address + (irq << 2) + 2;
1794 
1795 	rc = ops->read_std(ctxt, cs_addr, &cs, 2, &ctxt->exception);
1796 	if (rc != X86EMUL_CONTINUE)
1797 		return rc;
1798 
1799 	rc = ops->read_std(ctxt, eip_addr, &eip, 2, &ctxt->exception);
1800 	if (rc != X86EMUL_CONTINUE)
1801 		return rc;
1802 
1803 	rc = load_segment_descriptor(ctxt, cs, VCPU_SREG_CS);
1804 	if (rc != X86EMUL_CONTINUE)
1805 		return rc;
1806 
1807 	ctxt->_eip = eip;
1808 
1809 	return rc;
1810 }
1811 
1812 int emulate_int_real(struct x86_emulate_ctxt *ctxt, int irq)
1813 {
1814 	int rc;
1815 
1816 	invalidate_registers(ctxt);
1817 	rc = __emulate_int_real(ctxt, irq);
1818 	if (rc == X86EMUL_CONTINUE)
1819 		writeback_registers(ctxt);
1820 	return rc;
1821 }
1822 
1823 static int emulate_int(struct x86_emulate_ctxt *ctxt, int irq)
1824 {
1825 	switch(ctxt->mode) {
1826 	case X86EMUL_MODE_REAL:
1827 		return __emulate_int_real(ctxt, irq);
1828 	case X86EMUL_MODE_VM86:
1829 	case X86EMUL_MODE_PROT16:
1830 	case X86EMUL_MODE_PROT32:
1831 	case X86EMUL_MODE_PROT64:
1832 	default:
1833 		/* Protected mode interrupts unimplemented yet */
1834 		return X86EMUL_UNHANDLEABLE;
1835 	}
1836 }
1837 
1838 static int emulate_iret_real(struct x86_emulate_ctxt *ctxt)
1839 {
1840 	int rc = X86EMUL_CONTINUE;
1841 	unsigned long temp_eip = 0;
1842 	unsigned long temp_eflags = 0;
1843 	unsigned long cs = 0;
1844 	unsigned long mask = EFLG_CF | EFLG_PF | EFLG_AF | EFLG_ZF | EFLG_SF | EFLG_TF |
1845 			     EFLG_IF | EFLG_DF | EFLG_OF | EFLG_IOPL | EFLG_NT | EFLG_RF |
1846 			     EFLG_AC | EFLG_ID | (1 << 1); /* Last one is the reserved bit */
1847 	unsigned long vm86_mask = EFLG_VM | EFLG_VIF | EFLG_VIP;
1848 
1849 	/* TODO: Add stack limit check */
1850 
1851 	rc = emulate_pop(ctxt, &temp_eip, ctxt->op_bytes);
1852 
1853 	if (rc != X86EMUL_CONTINUE)
1854 		return rc;
1855 
1856 	if (temp_eip & ~0xffff)
1857 		return emulate_gp(ctxt, 0);
1858 
1859 	rc = emulate_pop(ctxt, &cs, ctxt->op_bytes);
1860 
1861 	if (rc != X86EMUL_CONTINUE)
1862 		return rc;
1863 
1864 	rc = emulate_pop(ctxt, &temp_eflags, ctxt->op_bytes);
1865 
1866 	if (rc != X86EMUL_CONTINUE)
1867 		return rc;
1868 
1869 	rc = load_segment_descriptor(ctxt, (u16)cs, VCPU_SREG_CS);
1870 
1871 	if (rc != X86EMUL_CONTINUE)
1872 		return rc;
1873 
1874 	ctxt->_eip = temp_eip;
1875 
1876 
1877 	if (ctxt->op_bytes == 4)
1878 		ctxt->eflags = ((temp_eflags & mask) | (ctxt->eflags & vm86_mask));
1879 	else if (ctxt->op_bytes == 2) {
1880 		ctxt->eflags &= ~0xffff;
1881 		ctxt->eflags |= temp_eflags;
1882 	}
1883 
1884 	ctxt->eflags &= ~EFLG_RESERVED_ZEROS_MASK; /* Clear reserved zeros */
1885 	ctxt->eflags |= EFLG_RESERVED_ONE_MASK;
1886 
1887 	return rc;
1888 }
1889 
1890 static int em_iret(struct x86_emulate_ctxt *ctxt)
1891 {
1892 	switch(ctxt->mode) {
1893 	case X86EMUL_MODE_REAL:
1894 		return emulate_iret_real(ctxt);
1895 	case X86EMUL_MODE_VM86:
1896 	case X86EMUL_MODE_PROT16:
1897 	case X86EMUL_MODE_PROT32:
1898 	case X86EMUL_MODE_PROT64:
1899 	default:
1900 		/* iret from protected mode unimplemented yet */
1901 		return X86EMUL_UNHANDLEABLE;
1902 	}
1903 }
1904 
1905 static int em_jmp_far(struct x86_emulate_ctxt *ctxt)
1906 {
1907 	int rc;
1908 	unsigned short sel;
1909 
1910 	memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
1911 
1912 	rc = load_segment_descriptor(ctxt, sel, VCPU_SREG_CS);
1913 	if (rc != X86EMUL_CONTINUE)
1914 		return rc;
1915 
1916 	ctxt->_eip = 0;
1917 	memcpy(&ctxt->_eip, ctxt->src.valptr, ctxt->op_bytes);
1918 	return X86EMUL_CONTINUE;
1919 }
1920 
1921 static int em_grp2(struct x86_emulate_ctxt *ctxt)
1922 {
1923 	switch (ctxt->modrm_reg) {
1924 	case 0:	/* rol */
1925 		emulate_2op_SrcB(ctxt, "rol");
1926 		break;
1927 	case 1:	/* ror */
1928 		emulate_2op_SrcB(ctxt, "ror");
1929 		break;
1930 	case 2:	/* rcl */
1931 		emulate_2op_SrcB(ctxt, "rcl");
1932 		break;
1933 	case 3:	/* rcr */
1934 		emulate_2op_SrcB(ctxt, "rcr");
1935 		break;
1936 	case 4:	/* sal/shl */
1937 	case 6:	/* sal/shl */
1938 		emulate_2op_SrcB(ctxt, "sal");
1939 		break;
1940 	case 5:	/* shr */
1941 		emulate_2op_SrcB(ctxt, "shr");
1942 		break;
1943 	case 7:	/* sar */
1944 		emulate_2op_SrcB(ctxt, "sar");
1945 		break;
1946 	}
1947 	return X86EMUL_CONTINUE;
1948 }
1949 
1950 static int em_not(struct x86_emulate_ctxt *ctxt)
1951 {
1952 	ctxt->dst.val = ~ctxt->dst.val;
1953 	return X86EMUL_CONTINUE;
1954 }
1955 
1956 static int em_neg(struct x86_emulate_ctxt *ctxt)
1957 {
1958 	emulate_1op(ctxt, "neg");
1959 	return X86EMUL_CONTINUE;
1960 }
1961 
1962 static int em_mul_ex(struct x86_emulate_ctxt *ctxt)
1963 {
1964 	u8 ex = 0;
1965 
1966 	emulate_1op_rax_rdx(ctxt, "mul", ex);
1967 	return X86EMUL_CONTINUE;
1968 }
1969 
1970 static int em_imul_ex(struct x86_emulate_ctxt *ctxt)
1971 {
1972 	u8 ex = 0;
1973 
1974 	emulate_1op_rax_rdx(ctxt, "imul", ex);
1975 	return X86EMUL_CONTINUE;
1976 }
1977 
1978 static int em_div_ex(struct x86_emulate_ctxt *ctxt)
1979 {
1980 	u8 de = 0;
1981 
1982 	emulate_1op_rax_rdx(ctxt, "div", de);
1983 	if (de)
1984 		return emulate_de(ctxt);
1985 	return X86EMUL_CONTINUE;
1986 }
1987 
1988 static int em_idiv_ex(struct x86_emulate_ctxt *ctxt)
1989 {
1990 	u8 de = 0;
1991 
1992 	emulate_1op_rax_rdx(ctxt, "idiv", de);
1993 	if (de)
1994 		return emulate_de(ctxt);
1995 	return X86EMUL_CONTINUE;
1996 }
1997 
1998 static int em_grp45(struct x86_emulate_ctxt *ctxt)
1999 {
2000 	int rc = X86EMUL_CONTINUE;
2001 
2002 	switch (ctxt->modrm_reg) {
2003 	case 0:	/* inc */
2004 		emulate_1op(ctxt, "inc");
2005 		break;
2006 	case 1:	/* dec */
2007 		emulate_1op(ctxt, "dec");
2008 		break;
2009 	case 2: /* call near abs */ {
2010 		long int old_eip;
2011 		old_eip = ctxt->_eip;
2012 		ctxt->_eip = ctxt->src.val;
2013 		ctxt->src.val = old_eip;
2014 		rc = em_push(ctxt);
2015 		break;
2016 	}
2017 	case 4: /* jmp abs */
2018 		ctxt->_eip = ctxt->src.val;
2019 		break;
2020 	case 5: /* jmp far */
2021 		rc = em_jmp_far(ctxt);
2022 		break;
2023 	case 6:	/* push */
2024 		rc = em_push(ctxt);
2025 		break;
2026 	}
2027 	return rc;
2028 }
2029 
2030 static int em_cmpxchg8b(struct x86_emulate_ctxt *ctxt)
2031 {
2032 	u64 old = ctxt->dst.orig_val64;
2033 
2034 	if (((u32) (old >> 0) != (u32) reg_read(ctxt, VCPU_REGS_RAX)) ||
2035 	    ((u32) (old >> 32) != (u32) reg_read(ctxt, VCPU_REGS_RDX))) {
2036 		*reg_write(ctxt, VCPU_REGS_RAX) = (u32) (old >> 0);
2037 		*reg_write(ctxt, VCPU_REGS_RDX) = (u32) (old >> 32);
2038 		ctxt->eflags &= ~EFLG_ZF;
2039 	} else {
2040 		ctxt->dst.val64 = ((u64)reg_read(ctxt, VCPU_REGS_RCX) << 32) |
2041 			(u32) reg_read(ctxt, VCPU_REGS_RBX);
2042 
2043 		ctxt->eflags |= EFLG_ZF;
2044 	}
2045 	return X86EMUL_CONTINUE;
2046 }
2047 
2048 static int em_ret(struct x86_emulate_ctxt *ctxt)
2049 {
2050 	ctxt->dst.type = OP_REG;
2051 	ctxt->dst.addr.reg = &ctxt->_eip;
2052 	ctxt->dst.bytes = ctxt->op_bytes;
2053 	return em_pop(ctxt);
2054 }
2055 
2056 static int em_ret_far(struct x86_emulate_ctxt *ctxt)
2057 {
2058 	int rc;
2059 	unsigned long cs;
2060 
2061 	rc = emulate_pop(ctxt, &ctxt->_eip, ctxt->op_bytes);
2062 	if (rc != X86EMUL_CONTINUE)
2063 		return rc;
2064 	if (ctxt->op_bytes == 4)
2065 		ctxt->_eip = (u32)ctxt->_eip;
2066 	rc = emulate_pop(ctxt, &cs, ctxt->op_bytes);
2067 	if (rc != X86EMUL_CONTINUE)
2068 		return rc;
2069 	rc = load_segment_descriptor(ctxt, (u16)cs, VCPU_SREG_CS);
2070 	return rc;
2071 }
2072 
2073 static int em_cmpxchg(struct x86_emulate_ctxt *ctxt)
2074 {
2075 	/* Save real source value, then compare EAX against destination. */
2076 	ctxt->src.orig_val = ctxt->src.val;
2077 	ctxt->src.val = reg_read(ctxt, VCPU_REGS_RAX);
2078 	emulate_2op_SrcV(ctxt, "cmp");
2079 
2080 	if (ctxt->eflags & EFLG_ZF) {
2081 		/* Success: write back to memory. */
2082 		ctxt->dst.val = ctxt->src.orig_val;
2083 	} else {
2084 		/* Failure: write the value we saw to EAX. */
2085 		ctxt->dst.type = OP_REG;
2086 		ctxt->dst.addr.reg = reg_rmw(ctxt, VCPU_REGS_RAX);
2087 	}
2088 	return X86EMUL_CONTINUE;
2089 }
2090 
2091 static int em_lseg(struct x86_emulate_ctxt *ctxt)
2092 {
2093 	int seg = ctxt->src2.val;
2094 	unsigned short sel;
2095 	int rc;
2096 
2097 	memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
2098 
2099 	rc = load_segment_descriptor(ctxt, sel, seg);
2100 	if (rc != X86EMUL_CONTINUE)
2101 		return rc;
2102 
2103 	ctxt->dst.val = ctxt->src.val;
2104 	return rc;
2105 }
2106 
2107 static void
2108 setup_syscalls_segments(struct x86_emulate_ctxt *ctxt,
2109 			struct desc_struct *cs, struct desc_struct *ss)
2110 {
2111 	cs->l = 0;		/* will be adjusted later */
2112 	set_desc_base(cs, 0);	/* flat segment */
2113 	cs->g = 1;		/* 4kb granularity */
2114 	set_desc_limit(cs, 0xfffff);	/* 4GB limit */
2115 	cs->type = 0x0b;	/* Read, Execute, Accessed */
2116 	cs->s = 1;
2117 	cs->dpl = 0;		/* will be adjusted later */
2118 	cs->p = 1;
2119 	cs->d = 1;
2120 	cs->avl = 0;
2121 
2122 	set_desc_base(ss, 0);	/* flat segment */
2123 	set_desc_limit(ss, 0xfffff);	/* 4GB limit */
2124 	ss->g = 1;		/* 4kb granularity */
2125 	ss->s = 1;
2126 	ss->type = 0x03;	/* Read/Write, Accessed */
2127 	ss->d = 1;		/* 32bit stack segment */
2128 	ss->dpl = 0;
2129 	ss->p = 1;
2130 	ss->l = 0;
2131 	ss->avl = 0;
2132 }
2133 
2134 static bool vendor_intel(struct x86_emulate_ctxt *ctxt)
2135 {
2136 	u32 eax, ebx, ecx, edx;
2137 
2138 	eax = ecx = 0;
2139 	ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
2140 	return ebx == X86EMUL_CPUID_VENDOR_GenuineIntel_ebx
2141 		&& ecx == X86EMUL_CPUID_VENDOR_GenuineIntel_ecx
2142 		&& edx == X86EMUL_CPUID_VENDOR_GenuineIntel_edx;
2143 }
2144 
2145 static bool em_syscall_is_enabled(struct x86_emulate_ctxt *ctxt)
2146 {
2147 	const struct x86_emulate_ops *ops = ctxt->ops;
2148 	u32 eax, ebx, ecx, edx;
2149 
2150 	/*
2151 	 * syscall should always be enabled in longmode - so only become
2152 	 * vendor specific (cpuid) if other modes are active...
2153 	 */
2154 	if (ctxt->mode == X86EMUL_MODE_PROT64)
2155 		return true;
2156 
2157 	eax = 0x00000000;
2158 	ecx = 0x00000000;
2159 	ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
2160 	/*
2161 	 * Intel ("GenuineIntel")
2162 	 * remark: Intel CPUs only support "syscall" in 64bit
2163 	 * longmode. Also an 64bit guest with a
2164 	 * 32bit compat-app running will #UD !! While this
2165 	 * behaviour can be fixed (by emulating) into AMD
2166 	 * response - CPUs of AMD can't behave like Intel.
2167 	 */
2168 	if (ebx == X86EMUL_CPUID_VENDOR_GenuineIntel_ebx &&
2169 	    ecx == X86EMUL_CPUID_VENDOR_GenuineIntel_ecx &&
2170 	    edx == X86EMUL_CPUID_VENDOR_GenuineIntel_edx)
2171 		return false;
2172 
2173 	/* AMD ("AuthenticAMD") */
2174 	if (ebx == X86EMUL_CPUID_VENDOR_AuthenticAMD_ebx &&
2175 	    ecx == X86EMUL_CPUID_VENDOR_AuthenticAMD_ecx &&
2176 	    edx == X86EMUL_CPUID_VENDOR_AuthenticAMD_edx)
2177 		return true;
2178 
2179 	/* AMD ("AMDisbetter!") */
2180 	if (ebx == X86EMUL_CPUID_VENDOR_AMDisbetterI_ebx &&
2181 	    ecx == X86EMUL_CPUID_VENDOR_AMDisbetterI_ecx &&
2182 	    edx == X86EMUL_CPUID_VENDOR_AMDisbetterI_edx)
2183 		return true;
2184 
2185 	/* default: (not Intel, not AMD), apply Intel's stricter rules... */
2186 	return false;
2187 }
2188 
2189 static int em_syscall(struct x86_emulate_ctxt *ctxt)
2190 {
2191 	const struct x86_emulate_ops *ops = ctxt->ops;
2192 	struct desc_struct cs, ss;
2193 	u64 msr_data;
2194 	u16 cs_sel, ss_sel;
2195 	u64 efer = 0;
2196 
2197 	/* syscall is not available in real mode */
2198 	if (ctxt->mode == X86EMUL_MODE_REAL ||
2199 	    ctxt->mode == X86EMUL_MODE_VM86)
2200 		return emulate_ud(ctxt);
2201 
2202 	if (!(em_syscall_is_enabled(ctxt)))
2203 		return emulate_ud(ctxt);
2204 
2205 	ops->get_msr(ctxt, MSR_EFER, &efer);
2206 	setup_syscalls_segments(ctxt, &cs, &ss);
2207 
2208 	if (!(efer & EFER_SCE))
2209 		return emulate_ud(ctxt);
2210 
2211 	ops->get_msr(ctxt, MSR_STAR, &msr_data);
2212 	msr_data >>= 32;
2213 	cs_sel = (u16)(msr_data & 0xfffc);
2214 	ss_sel = (u16)(msr_data + 8);
2215 
2216 	if (efer & EFER_LMA) {
2217 		cs.d = 0;
2218 		cs.l = 1;
2219 	}
2220 	ops->set_segment(ctxt, cs_sel, &cs, 0, VCPU_SREG_CS);
2221 	ops->set_segment(ctxt, ss_sel, &ss, 0, VCPU_SREG_SS);
2222 
2223 	*reg_write(ctxt, VCPU_REGS_RCX) = ctxt->_eip;
2224 	if (efer & EFER_LMA) {
2225 #ifdef CONFIG_X86_64
2226 		*reg_write(ctxt, VCPU_REGS_R11) = ctxt->eflags & ~EFLG_RF;
2227 
2228 		ops->get_msr(ctxt,
2229 			     ctxt->mode == X86EMUL_MODE_PROT64 ?
2230 			     MSR_LSTAR : MSR_CSTAR, &msr_data);
2231 		ctxt->_eip = msr_data;
2232 
2233 		ops->get_msr(ctxt, MSR_SYSCALL_MASK, &msr_data);
2234 		ctxt->eflags &= ~(msr_data | EFLG_RF);
2235 #endif
2236 	} else {
2237 		/* legacy mode */
2238 		ops->get_msr(ctxt, MSR_STAR, &msr_data);
2239 		ctxt->_eip = (u32)msr_data;
2240 
2241 		ctxt->eflags &= ~(EFLG_VM | EFLG_IF | EFLG_RF);
2242 	}
2243 
2244 	return X86EMUL_CONTINUE;
2245 }
2246 
2247 static int em_sysenter(struct x86_emulate_ctxt *ctxt)
2248 {
2249 	const struct x86_emulate_ops *ops = ctxt->ops;
2250 	struct desc_struct cs, ss;
2251 	u64 msr_data;
2252 	u16 cs_sel, ss_sel;
2253 	u64 efer = 0;
2254 
2255 	ops->get_msr(ctxt, MSR_EFER, &efer);
2256 	/* inject #GP if in real mode */
2257 	if (ctxt->mode == X86EMUL_MODE_REAL)
2258 		return emulate_gp(ctxt, 0);
2259 
2260 	/*
2261 	 * Not recognized on AMD in compat mode (but is recognized in legacy
2262 	 * mode).
2263 	 */
2264 	if ((ctxt->mode == X86EMUL_MODE_PROT32) && (efer & EFER_LMA)
2265 	    && !vendor_intel(ctxt))
2266 		return emulate_ud(ctxt);
2267 
2268 	/* XXX sysenter/sysexit have not been tested in 64bit mode.
2269 	* Therefore, we inject an #UD.
2270 	*/
2271 	if (ctxt->mode == X86EMUL_MODE_PROT64)
2272 		return emulate_ud(ctxt);
2273 
2274 	setup_syscalls_segments(ctxt, &cs, &ss);
2275 
2276 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_CS, &msr_data);
2277 	switch (ctxt->mode) {
2278 	case X86EMUL_MODE_PROT32:
2279 		if ((msr_data & 0xfffc) == 0x0)
2280 			return emulate_gp(ctxt, 0);
2281 		break;
2282 	case X86EMUL_MODE_PROT64:
2283 		if (msr_data == 0x0)
2284 			return emulate_gp(ctxt, 0);
2285 		break;
2286 	default:
2287 		break;
2288 	}
2289 
2290 	ctxt->eflags &= ~(EFLG_VM | EFLG_IF | EFLG_RF);
2291 	cs_sel = (u16)msr_data;
2292 	cs_sel &= ~SELECTOR_RPL_MASK;
2293 	ss_sel = cs_sel + 8;
2294 	ss_sel &= ~SELECTOR_RPL_MASK;
2295 	if (ctxt->mode == X86EMUL_MODE_PROT64 || (efer & EFER_LMA)) {
2296 		cs.d = 0;
2297 		cs.l = 1;
2298 	}
2299 
2300 	ops->set_segment(ctxt, cs_sel, &cs, 0, VCPU_SREG_CS);
2301 	ops->set_segment(ctxt, ss_sel, &ss, 0, VCPU_SREG_SS);
2302 
2303 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_EIP, &msr_data);
2304 	ctxt->_eip = msr_data;
2305 
2306 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_ESP, &msr_data);
2307 	*reg_write(ctxt, VCPU_REGS_RSP) = msr_data;
2308 
2309 	return X86EMUL_CONTINUE;
2310 }
2311 
2312 static int em_sysexit(struct x86_emulate_ctxt *ctxt)
2313 {
2314 	const struct x86_emulate_ops *ops = ctxt->ops;
2315 	struct desc_struct cs, ss;
2316 	u64 msr_data;
2317 	int usermode;
2318 	u16 cs_sel = 0, ss_sel = 0;
2319 
2320 	/* inject #GP if in real mode or Virtual 8086 mode */
2321 	if (ctxt->mode == X86EMUL_MODE_REAL ||
2322 	    ctxt->mode == X86EMUL_MODE_VM86)
2323 		return emulate_gp(ctxt, 0);
2324 
2325 	setup_syscalls_segments(ctxt, &cs, &ss);
2326 
2327 	if ((ctxt->rex_prefix & 0x8) != 0x0)
2328 		usermode = X86EMUL_MODE_PROT64;
2329 	else
2330 		usermode = X86EMUL_MODE_PROT32;
2331 
2332 	cs.dpl = 3;
2333 	ss.dpl = 3;
2334 	ops->get_msr(ctxt, MSR_IA32_SYSENTER_CS, &msr_data);
2335 	switch (usermode) {
2336 	case X86EMUL_MODE_PROT32:
2337 		cs_sel = (u16)(msr_data + 16);
2338 		if ((msr_data & 0xfffc) == 0x0)
2339 			return emulate_gp(ctxt, 0);
2340 		ss_sel = (u16)(msr_data + 24);
2341 		break;
2342 	case X86EMUL_MODE_PROT64:
2343 		cs_sel = (u16)(msr_data + 32);
2344 		if (msr_data == 0x0)
2345 			return emulate_gp(ctxt, 0);
2346 		ss_sel = cs_sel + 8;
2347 		cs.d = 0;
2348 		cs.l = 1;
2349 		break;
2350 	}
2351 	cs_sel |= SELECTOR_RPL_MASK;
2352 	ss_sel |= SELECTOR_RPL_MASK;
2353 
2354 	ops->set_segment(ctxt, cs_sel, &cs, 0, VCPU_SREG_CS);
2355 	ops->set_segment(ctxt, ss_sel, &ss, 0, VCPU_SREG_SS);
2356 
2357 	ctxt->_eip = reg_read(ctxt, VCPU_REGS_RDX);
2358 	*reg_write(ctxt, VCPU_REGS_RSP) = reg_read(ctxt, VCPU_REGS_RCX);
2359 
2360 	return X86EMUL_CONTINUE;
2361 }
2362 
2363 static bool emulator_bad_iopl(struct x86_emulate_ctxt *ctxt)
2364 {
2365 	int iopl;
2366 	if (ctxt->mode == X86EMUL_MODE_REAL)
2367 		return false;
2368 	if (ctxt->mode == X86EMUL_MODE_VM86)
2369 		return true;
2370 	iopl = (ctxt->eflags & X86_EFLAGS_IOPL) >> IOPL_SHIFT;
2371 	return ctxt->ops->cpl(ctxt) > iopl;
2372 }
2373 
2374 static bool emulator_io_port_access_allowed(struct x86_emulate_ctxt *ctxt,
2375 					    u16 port, u16 len)
2376 {
2377 	const struct x86_emulate_ops *ops = ctxt->ops;
2378 	struct desc_struct tr_seg;
2379 	u32 base3;
2380 	int r;
2381 	u16 tr, io_bitmap_ptr, perm, bit_idx = port & 0x7;
2382 	unsigned mask = (1 << len) - 1;
2383 	unsigned long base;
2384 
2385 	ops->get_segment(ctxt, &tr, &tr_seg, &base3, VCPU_SREG_TR);
2386 	if (!tr_seg.p)
2387 		return false;
2388 	if (desc_limit_scaled(&tr_seg) < 103)
2389 		return false;
2390 	base = get_desc_base(&tr_seg);
2391 #ifdef CONFIG_X86_64
2392 	base |= ((u64)base3) << 32;
2393 #endif
2394 	r = ops->read_std(ctxt, base + 102, &io_bitmap_ptr, 2, NULL);
2395 	if (r != X86EMUL_CONTINUE)
2396 		return false;
2397 	if (io_bitmap_ptr + port/8 > desc_limit_scaled(&tr_seg))
2398 		return false;
2399 	r = ops->read_std(ctxt, base + io_bitmap_ptr + port/8, &perm, 2, NULL);
2400 	if (r != X86EMUL_CONTINUE)
2401 		return false;
2402 	if ((perm >> bit_idx) & mask)
2403 		return false;
2404 	return true;
2405 }
2406 
2407 static bool emulator_io_permited(struct x86_emulate_ctxt *ctxt,
2408 				 u16 port, u16 len)
2409 {
2410 	if (ctxt->perm_ok)
2411 		return true;
2412 
2413 	if (emulator_bad_iopl(ctxt))
2414 		if (!emulator_io_port_access_allowed(ctxt, port, len))
2415 			return false;
2416 
2417 	ctxt->perm_ok = true;
2418 
2419 	return true;
2420 }
2421 
2422 static void save_state_to_tss16(struct x86_emulate_ctxt *ctxt,
2423 				struct tss_segment_16 *tss)
2424 {
2425 	tss->ip = ctxt->_eip;
2426 	tss->flag = ctxt->eflags;
2427 	tss->ax = reg_read(ctxt, VCPU_REGS_RAX);
2428 	tss->cx = reg_read(ctxt, VCPU_REGS_RCX);
2429 	tss->dx = reg_read(ctxt, VCPU_REGS_RDX);
2430 	tss->bx = reg_read(ctxt, VCPU_REGS_RBX);
2431 	tss->sp = reg_read(ctxt, VCPU_REGS_RSP);
2432 	tss->bp = reg_read(ctxt, VCPU_REGS_RBP);
2433 	tss->si = reg_read(ctxt, VCPU_REGS_RSI);
2434 	tss->di = reg_read(ctxt, VCPU_REGS_RDI);
2435 
2436 	tss->es = get_segment_selector(ctxt, VCPU_SREG_ES);
2437 	tss->cs = get_segment_selector(ctxt, VCPU_SREG_CS);
2438 	tss->ss = get_segment_selector(ctxt, VCPU_SREG_SS);
2439 	tss->ds = get_segment_selector(ctxt, VCPU_SREG_DS);
2440 	tss->ldt = get_segment_selector(ctxt, VCPU_SREG_LDTR);
2441 }
2442 
2443 static int load_state_from_tss16(struct x86_emulate_ctxt *ctxt,
2444 				 struct tss_segment_16 *tss)
2445 {
2446 	int ret;
2447 
2448 	ctxt->_eip = tss->ip;
2449 	ctxt->eflags = tss->flag | 2;
2450 	*reg_write(ctxt, VCPU_REGS_RAX) = tss->ax;
2451 	*reg_write(ctxt, VCPU_REGS_RCX) = tss->cx;
2452 	*reg_write(ctxt, VCPU_REGS_RDX) = tss->dx;
2453 	*reg_write(ctxt, VCPU_REGS_RBX) = tss->bx;
2454 	*reg_write(ctxt, VCPU_REGS_RSP) = tss->sp;
2455 	*reg_write(ctxt, VCPU_REGS_RBP) = tss->bp;
2456 	*reg_write(ctxt, VCPU_REGS_RSI) = tss->si;
2457 	*reg_write(ctxt, VCPU_REGS_RDI) = tss->di;
2458 
2459 	/*
2460 	 * SDM says that segment selectors are loaded before segment
2461 	 * descriptors
2462 	 */
2463 	set_segment_selector(ctxt, tss->ldt, VCPU_SREG_LDTR);
2464 	set_segment_selector(ctxt, tss->es, VCPU_SREG_ES);
2465 	set_segment_selector(ctxt, tss->cs, VCPU_SREG_CS);
2466 	set_segment_selector(ctxt, tss->ss, VCPU_SREG_SS);
2467 	set_segment_selector(ctxt, tss->ds, VCPU_SREG_DS);
2468 
2469 	/*
2470 	 * Now load segment descriptors. If fault happens at this stage
2471 	 * it is handled in a context of new task
2472 	 */
2473 	ret = load_segment_descriptor(ctxt, tss->ldt, VCPU_SREG_LDTR);
2474 	if (ret != X86EMUL_CONTINUE)
2475 		return ret;
2476 	ret = load_segment_descriptor(ctxt, tss->es, VCPU_SREG_ES);
2477 	if (ret != X86EMUL_CONTINUE)
2478 		return ret;
2479 	ret = load_segment_descriptor(ctxt, tss->cs, VCPU_SREG_CS);
2480 	if (ret != X86EMUL_CONTINUE)
2481 		return ret;
2482 	ret = load_segment_descriptor(ctxt, tss->ss, VCPU_SREG_SS);
2483 	if (ret != X86EMUL_CONTINUE)
2484 		return ret;
2485 	ret = load_segment_descriptor(ctxt, tss->ds, VCPU_SREG_DS);
2486 	if (ret != X86EMUL_CONTINUE)
2487 		return ret;
2488 
2489 	return X86EMUL_CONTINUE;
2490 }
2491 
2492 static int task_switch_16(struct x86_emulate_ctxt *ctxt,
2493 			  u16 tss_selector, u16 old_tss_sel,
2494 			  ulong old_tss_base, struct desc_struct *new_desc)
2495 {
2496 	const struct x86_emulate_ops *ops = ctxt->ops;
2497 	struct tss_segment_16 tss_seg;
2498 	int ret;
2499 	u32 new_tss_base = get_desc_base(new_desc);
2500 
2501 	ret = ops->read_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2502 			    &ctxt->exception);
2503 	if (ret != X86EMUL_CONTINUE)
2504 		/* FIXME: need to provide precise fault address */
2505 		return ret;
2506 
2507 	save_state_to_tss16(ctxt, &tss_seg);
2508 
2509 	ret = ops->write_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2510 			     &ctxt->exception);
2511 	if (ret != X86EMUL_CONTINUE)
2512 		/* FIXME: need to provide precise fault address */
2513 		return ret;
2514 
2515 	ret = ops->read_std(ctxt, new_tss_base, &tss_seg, sizeof tss_seg,
2516 			    &ctxt->exception);
2517 	if (ret != X86EMUL_CONTINUE)
2518 		/* FIXME: need to provide precise fault address */
2519 		return ret;
2520 
2521 	if (old_tss_sel != 0xffff) {
2522 		tss_seg.prev_task_link = old_tss_sel;
2523 
2524 		ret = ops->write_std(ctxt, new_tss_base,
2525 				     &tss_seg.prev_task_link,
2526 				     sizeof tss_seg.prev_task_link,
2527 				     &ctxt->exception);
2528 		if (ret != X86EMUL_CONTINUE)
2529 			/* FIXME: need to provide precise fault address */
2530 			return ret;
2531 	}
2532 
2533 	return load_state_from_tss16(ctxt, &tss_seg);
2534 }
2535 
2536 static void save_state_to_tss32(struct x86_emulate_ctxt *ctxt,
2537 				struct tss_segment_32 *tss)
2538 {
2539 	tss->cr3 = ctxt->ops->get_cr(ctxt, 3);
2540 	tss->eip = ctxt->_eip;
2541 	tss->eflags = ctxt->eflags;
2542 	tss->eax = reg_read(ctxt, VCPU_REGS_RAX);
2543 	tss->ecx = reg_read(ctxt, VCPU_REGS_RCX);
2544 	tss->edx = reg_read(ctxt, VCPU_REGS_RDX);
2545 	tss->ebx = reg_read(ctxt, VCPU_REGS_RBX);
2546 	tss->esp = reg_read(ctxt, VCPU_REGS_RSP);
2547 	tss->ebp = reg_read(ctxt, VCPU_REGS_RBP);
2548 	tss->esi = reg_read(ctxt, VCPU_REGS_RSI);
2549 	tss->edi = reg_read(ctxt, VCPU_REGS_RDI);
2550 
2551 	tss->es = get_segment_selector(ctxt, VCPU_SREG_ES);
2552 	tss->cs = get_segment_selector(ctxt, VCPU_SREG_CS);
2553 	tss->ss = get_segment_selector(ctxt, VCPU_SREG_SS);
2554 	tss->ds = get_segment_selector(ctxt, VCPU_SREG_DS);
2555 	tss->fs = get_segment_selector(ctxt, VCPU_SREG_FS);
2556 	tss->gs = get_segment_selector(ctxt, VCPU_SREG_GS);
2557 	tss->ldt_selector = get_segment_selector(ctxt, VCPU_SREG_LDTR);
2558 }
2559 
2560 static int load_state_from_tss32(struct x86_emulate_ctxt *ctxt,
2561 				 struct tss_segment_32 *tss)
2562 {
2563 	int ret;
2564 
2565 	if (ctxt->ops->set_cr(ctxt, 3, tss->cr3))
2566 		return emulate_gp(ctxt, 0);
2567 	ctxt->_eip = tss->eip;
2568 	ctxt->eflags = tss->eflags | 2;
2569 
2570 	/* General purpose registers */
2571 	*reg_write(ctxt, VCPU_REGS_RAX) = tss->eax;
2572 	*reg_write(ctxt, VCPU_REGS_RCX) = tss->ecx;
2573 	*reg_write(ctxt, VCPU_REGS_RDX) = tss->edx;
2574 	*reg_write(ctxt, VCPU_REGS_RBX) = tss->ebx;
2575 	*reg_write(ctxt, VCPU_REGS_RSP) = tss->esp;
2576 	*reg_write(ctxt, VCPU_REGS_RBP) = tss->ebp;
2577 	*reg_write(ctxt, VCPU_REGS_RSI) = tss->esi;
2578 	*reg_write(ctxt, VCPU_REGS_RDI) = tss->edi;
2579 
2580 	/*
2581 	 * SDM says that segment selectors are loaded before segment
2582 	 * descriptors
2583 	 */
2584 	set_segment_selector(ctxt, tss->ldt_selector, VCPU_SREG_LDTR);
2585 	set_segment_selector(ctxt, tss->es, VCPU_SREG_ES);
2586 	set_segment_selector(ctxt, tss->cs, VCPU_SREG_CS);
2587 	set_segment_selector(ctxt, tss->ss, VCPU_SREG_SS);
2588 	set_segment_selector(ctxt, tss->ds, VCPU_SREG_DS);
2589 	set_segment_selector(ctxt, tss->fs, VCPU_SREG_FS);
2590 	set_segment_selector(ctxt, tss->gs, VCPU_SREG_GS);
2591 
2592 	/*
2593 	 * If we're switching between Protected Mode and VM86, we need to make
2594 	 * sure to update the mode before loading the segment descriptors so
2595 	 * that the selectors are interpreted correctly.
2596 	 *
2597 	 * Need to get rflags to the vcpu struct immediately because it
2598 	 * influences the CPL which is checked at least when loading the segment
2599 	 * descriptors and when pushing an error code to the new kernel stack.
2600 	 *
2601 	 * TODO Introduce a separate ctxt->ops->set_cpl callback
2602 	 */
2603 	if (ctxt->eflags & X86_EFLAGS_VM)
2604 		ctxt->mode = X86EMUL_MODE_VM86;
2605 	else
2606 		ctxt->mode = X86EMUL_MODE_PROT32;
2607 
2608 	ctxt->ops->set_rflags(ctxt, ctxt->eflags);
2609 
2610 	/*
2611 	 * Now load segment descriptors. If fault happenes at this stage
2612 	 * it is handled in a context of new task
2613 	 */
2614 	ret = load_segment_descriptor(ctxt, tss->ldt_selector, VCPU_SREG_LDTR);
2615 	if (ret != X86EMUL_CONTINUE)
2616 		return ret;
2617 	ret = load_segment_descriptor(ctxt, tss->es, VCPU_SREG_ES);
2618 	if (ret != X86EMUL_CONTINUE)
2619 		return ret;
2620 	ret = load_segment_descriptor(ctxt, tss->cs, VCPU_SREG_CS);
2621 	if (ret != X86EMUL_CONTINUE)
2622 		return ret;
2623 	ret = load_segment_descriptor(ctxt, tss->ss, VCPU_SREG_SS);
2624 	if (ret != X86EMUL_CONTINUE)
2625 		return ret;
2626 	ret = load_segment_descriptor(ctxt, tss->ds, VCPU_SREG_DS);
2627 	if (ret != X86EMUL_CONTINUE)
2628 		return ret;
2629 	ret = load_segment_descriptor(ctxt, tss->fs, VCPU_SREG_FS);
2630 	if (ret != X86EMUL_CONTINUE)
2631 		return ret;
2632 	ret = load_segment_descriptor(ctxt, tss->gs, VCPU_SREG_GS);
2633 	if (ret != X86EMUL_CONTINUE)
2634 		return ret;
2635 
2636 	return X86EMUL_CONTINUE;
2637 }
2638 
2639 static int task_switch_32(struct x86_emulate_ctxt *ctxt,
2640 			  u16 tss_selector, u16 old_tss_sel,
2641 			  ulong old_tss_base, struct desc_struct *new_desc)
2642 {
2643 	const struct x86_emulate_ops *ops = ctxt->ops;
2644 	struct tss_segment_32 tss_seg;
2645 	int ret;
2646 	u32 new_tss_base = get_desc_base(new_desc);
2647 
2648 	ret = ops->read_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2649 			    &ctxt->exception);
2650 	if (ret != X86EMUL_CONTINUE)
2651 		/* FIXME: need to provide precise fault address */
2652 		return ret;
2653 
2654 	save_state_to_tss32(ctxt, &tss_seg);
2655 
2656 	ret = ops->write_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
2657 			     &ctxt->exception);
2658 	if (ret != X86EMUL_CONTINUE)
2659 		/* FIXME: need to provide precise fault address */
2660 		return ret;
2661 
2662 	ret = ops->read_std(ctxt, new_tss_base, &tss_seg, sizeof tss_seg,
2663 			    &ctxt->exception);
2664 	if (ret != X86EMUL_CONTINUE)
2665 		/* FIXME: need to provide precise fault address */
2666 		return ret;
2667 
2668 	if (old_tss_sel != 0xffff) {
2669 		tss_seg.prev_task_link = old_tss_sel;
2670 
2671 		ret = ops->write_std(ctxt, new_tss_base,
2672 				     &tss_seg.prev_task_link,
2673 				     sizeof tss_seg.prev_task_link,
2674 				     &ctxt->exception);
2675 		if (ret != X86EMUL_CONTINUE)
2676 			/* FIXME: need to provide precise fault address */
2677 			return ret;
2678 	}
2679 
2680 	return load_state_from_tss32(ctxt, &tss_seg);
2681 }
2682 
2683 static int emulator_do_task_switch(struct x86_emulate_ctxt *ctxt,
2684 				   u16 tss_selector, int idt_index, int reason,
2685 				   bool has_error_code, u32 error_code)
2686 {
2687 	const struct x86_emulate_ops *ops = ctxt->ops;
2688 	struct desc_struct curr_tss_desc, next_tss_desc;
2689 	int ret;
2690 	u16 old_tss_sel = get_segment_selector(ctxt, VCPU_SREG_TR);
2691 	ulong old_tss_base =
2692 		ops->get_cached_segment_base(ctxt, VCPU_SREG_TR);
2693 	u32 desc_limit;
2694 	ulong desc_addr;
2695 
2696 	/* FIXME: old_tss_base == ~0 ? */
2697 
2698 	ret = read_segment_descriptor(ctxt, tss_selector, &next_tss_desc, &desc_addr);
2699 	if (ret != X86EMUL_CONTINUE)
2700 		return ret;
2701 	ret = read_segment_descriptor(ctxt, old_tss_sel, &curr_tss_desc, &desc_addr);
2702 	if (ret != X86EMUL_CONTINUE)
2703 		return ret;
2704 
2705 	/* FIXME: check that next_tss_desc is tss */
2706 
2707 	/*
2708 	 * Check privileges. The three cases are task switch caused by...
2709 	 *
2710 	 * 1. jmp/call/int to task gate: Check against DPL of the task gate
2711 	 * 2. Exception/IRQ/iret: No check is performed
2712 	 * 3. jmp/call to TSS: Check against DPL of the TSS
2713 	 */
2714 	if (reason == TASK_SWITCH_GATE) {
2715 		if (idt_index != -1) {
2716 			/* Software interrupts */
2717 			struct desc_struct task_gate_desc;
2718 			int dpl;
2719 
2720 			ret = read_interrupt_descriptor(ctxt, idt_index,
2721 							&task_gate_desc);
2722 			if (ret != X86EMUL_CONTINUE)
2723 				return ret;
2724 
2725 			dpl = task_gate_desc.dpl;
2726 			if ((tss_selector & 3) > dpl || ops->cpl(ctxt) > dpl)
2727 				return emulate_gp(ctxt, (idt_index << 3) | 0x2);
2728 		}
2729 	} else if (reason != TASK_SWITCH_IRET) {
2730 		int dpl = next_tss_desc.dpl;
2731 		if ((tss_selector & 3) > dpl || ops->cpl(ctxt) > dpl)
2732 			return emulate_gp(ctxt, tss_selector);
2733 	}
2734 
2735 
2736 	desc_limit = desc_limit_scaled(&next_tss_desc);
2737 	if (!next_tss_desc.p ||
2738 	    ((desc_limit < 0x67 && (next_tss_desc.type & 8)) ||
2739 	     desc_limit < 0x2b)) {
2740 		emulate_ts(ctxt, tss_selector & 0xfffc);
2741 		return X86EMUL_PROPAGATE_FAULT;
2742 	}
2743 
2744 	if (reason == TASK_SWITCH_IRET || reason == TASK_SWITCH_JMP) {
2745 		curr_tss_desc.type &= ~(1 << 1); /* clear busy flag */
2746 		write_segment_descriptor(ctxt, old_tss_sel, &curr_tss_desc);
2747 	}
2748 
2749 	if (reason == TASK_SWITCH_IRET)
2750 		ctxt->eflags = ctxt->eflags & ~X86_EFLAGS_NT;
2751 
2752 	/* set back link to prev task only if NT bit is set in eflags
2753 	   note that old_tss_sel is not used after this point */
2754 	if (reason != TASK_SWITCH_CALL && reason != TASK_SWITCH_GATE)
2755 		old_tss_sel = 0xffff;
2756 
2757 	if (next_tss_desc.type & 8)
2758 		ret = task_switch_32(ctxt, tss_selector, old_tss_sel,
2759 				     old_tss_base, &next_tss_desc);
2760 	else
2761 		ret = task_switch_16(ctxt, tss_selector, old_tss_sel,
2762 				     old_tss_base, &next_tss_desc);
2763 	if (ret != X86EMUL_CONTINUE)
2764 		return ret;
2765 
2766 	if (reason == TASK_SWITCH_CALL || reason == TASK_SWITCH_GATE)
2767 		ctxt->eflags = ctxt->eflags | X86_EFLAGS_NT;
2768 
2769 	if (reason != TASK_SWITCH_IRET) {
2770 		next_tss_desc.type |= (1 << 1); /* set busy flag */
2771 		write_segment_descriptor(ctxt, tss_selector, &next_tss_desc);
2772 	}
2773 
2774 	ops->set_cr(ctxt, 0,  ops->get_cr(ctxt, 0) | X86_CR0_TS);
2775 	ops->set_segment(ctxt, tss_selector, &next_tss_desc, 0, VCPU_SREG_TR);
2776 
2777 	if (has_error_code) {
2778 		ctxt->op_bytes = ctxt->ad_bytes = (next_tss_desc.type & 8) ? 4 : 2;
2779 		ctxt->lock_prefix = 0;
2780 		ctxt->src.val = (unsigned long) error_code;
2781 		ret = em_push(ctxt);
2782 	}
2783 
2784 	return ret;
2785 }
2786 
2787 int emulator_task_switch(struct x86_emulate_ctxt *ctxt,
2788 			 u16 tss_selector, int idt_index, int reason,
2789 			 bool has_error_code, u32 error_code)
2790 {
2791 	int rc;
2792 
2793 	invalidate_registers(ctxt);
2794 	ctxt->_eip = ctxt->eip;
2795 	ctxt->dst.type = OP_NONE;
2796 
2797 	rc = emulator_do_task_switch(ctxt, tss_selector, idt_index, reason,
2798 				     has_error_code, error_code);
2799 
2800 	if (rc == X86EMUL_CONTINUE) {
2801 		ctxt->eip = ctxt->_eip;
2802 		writeback_registers(ctxt);
2803 	}
2804 
2805 	return (rc == X86EMUL_UNHANDLEABLE) ? EMULATION_FAILED : EMULATION_OK;
2806 }
2807 
2808 static void string_addr_inc(struct x86_emulate_ctxt *ctxt, int reg,
2809 		struct operand *op)
2810 {
2811 	int df = (ctxt->eflags & EFLG_DF) ? -op->count : op->count;
2812 
2813 	register_address_increment(ctxt, reg_rmw(ctxt, reg), df * op->bytes);
2814 	op->addr.mem.ea = register_address(ctxt, reg_read(ctxt, reg));
2815 }
2816 
2817 static int em_das(struct x86_emulate_ctxt *ctxt)
2818 {
2819 	u8 al, old_al;
2820 	bool af, cf, old_cf;
2821 
2822 	cf = ctxt->eflags & X86_EFLAGS_CF;
2823 	al = ctxt->dst.val;
2824 
2825 	old_al = al;
2826 	old_cf = cf;
2827 	cf = false;
2828 	af = ctxt->eflags & X86_EFLAGS_AF;
2829 	if ((al & 0x0f) > 9 || af) {
2830 		al -= 6;
2831 		cf = old_cf | (al >= 250);
2832 		af = true;
2833 	} else {
2834 		af = false;
2835 	}
2836 	if (old_al > 0x99 || old_cf) {
2837 		al -= 0x60;
2838 		cf = true;
2839 	}
2840 
2841 	ctxt->dst.val = al;
2842 	/* Set PF, ZF, SF */
2843 	ctxt->src.type = OP_IMM;
2844 	ctxt->src.val = 0;
2845 	ctxt->src.bytes = 1;
2846 	emulate_2op_SrcV(ctxt, "or");
2847 	ctxt->eflags &= ~(X86_EFLAGS_AF | X86_EFLAGS_CF);
2848 	if (cf)
2849 		ctxt->eflags |= X86_EFLAGS_CF;
2850 	if (af)
2851 		ctxt->eflags |= X86_EFLAGS_AF;
2852 	return X86EMUL_CONTINUE;
2853 }
2854 
2855 static int em_call(struct x86_emulate_ctxt *ctxt)
2856 {
2857 	long rel = ctxt->src.val;
2858 
2859 	ctxt->src.val = (unsigned long)ctxt->_eip;
2860 	jmp_rel(ctxt, rel);
2861 	return em_push(ctxt);
2862 }
2863 
2864 static int em_call_far(struct x86_emulate_ctxt *ctxt)
2865 {
2866 	u16 sel, old_cs;
2867 	ulong old_eip;
2868 	int rc;
2869 
2870 	old_cs = get_segment_selector(ctxt, VCPU_SREG_CS);
2871 	old_eip = ctxt->_eip;
2872 
2873 	memcpy(&sel, ctxt->src.valptr + ctxt->op_bytes, 2);
2874 	if (load_segment_descriptor(ctxt, sel, VCPU_SREG_CS))
2875 		return X86EMUL_CONTINUE;
2876 
2877 	ctxt->_eip = 0;
2878 	memcpy(&ctxt->_eip, ctxt->src.valptr, ctxt->op_bytes);
2879 
2880 	ctxt->src.val = old_cs;
2881 	rc = em_push(ctxt);
2882 	if (rc != X86EMUL_CONTINUE)
2883 		return rc;
2884 
2885 	ctxt->src.val = old_eip;
2886 	return em_push(ctxt);
2887 }
2888 
2889 static int em_ret_near_imm(struct x86_emulate_ctxt *ctxt)
2890 {
2891 	int rc;
2892 
2893 	ctxt->dst.type = OP_REG;
2894 	ctxt->dst.addr.reg = &ctxt->_eip;
2895 	ctxt->dst.bytes = ctxt->op_bytes;
2896 	rc = emulate_pop(ctxt, &ctxt->dst.val, ctxt->op_bytes);
2897 	if (rc != X86EMUL_CONTINUE)
2898 		return rc;
2899 	rsp_increment(ctxt, ctxt->src.val);
2900 	return X86EMUL_CONTINUE;
2901 }
2902 
2903 static int em_add(struct x86_emulate_ctxt *ctxt)
2904 {
2905 	emulate_2op_SrcV(ctxt, "add");
2906 	return X86EMUL_CONTINUE;
2907 }
2908 
2909 static int em_or(struct x86_emulate_ctxt *ctxt)
2910 {
2911 	emulate_2op_SrcV(ctxt, "or");
2912 	return X86EMUL_CONTINUE;
2913 }
2914 
2915 static int em_adc(struct x86_emulate_ctxt *ctxt)
2916 {
2917 	emulate_2op_SrcV(ctxt, "adc");
2918 	return X86EMUL_CONTINUE;
2919 }
2920 
2921 static int em_sbb(struct x86_emulate_ctxt *ctxt)
2922 {
2923 	emulate_2op_SrcV(ctxt, "sbb");
2924 	return X86EMUL_CONTINUE;
2925 }
2926 
2927 static int em_and(struct x86_emulate_ctxt *ctxt)
2928 {
2929 	emulate_2op_SrcV(ctxt, "and");
2930 	return X86EMUL_CONTINUE;
2931 }
2932 
2933 static int em_sub(struct x86_emulate_ctxt *ctxt)
2934 {
2935 	emulate_2op_SrcV(ctxt, "sub");
2936 	return X86EMUL_CONTINUE;
2937 }
2938 
2939 static int em_xor(struct x86_emulate_ctxt *ctxt)
2940 {
2941 	emulate_2op_SrcV(ctxt, "xor");
2942 	return X86EMUL_CONTINUE;
2943 }
2944 
2945 static int em_cmp(struct x86_emulate_ctxt *ctxt)
2946 {
2947 	emulate_2op_SrcV(ctxt, "cmp");
2948 	/* Disable writeback. */
2949 	ctxt->dst.type = OP_NONE;
2950 	return X86EMUL_CONTINUE;
2951 }
2952 
2953 static int em_test(struct x86_emulate_ctxt *ctxt)
2954 {
2955 	emulate_2op_SrcV(ctxt, "test");
2956 	/* Disable writeback. */
2957 	ctxt->dst.type = OP_NONE;
2958 	return X86EMUL_CONTINUE;
2959 }
2960 
2961 static int em_xchg(struct x86_emulate_ctxt *ctxt)
2962 {
2963 	/* Write back the register source. */
2964 	ctxt->src.val = ctxt->dst.val;
2965 	write_register_operand(&ctxt->src);
2966 
2967 	/* Write back the memory destination with implicit LOCK prefix. */
2968 	ctxt->dst.val = ctxt->src.orig_val;
2969 	ctxt->lock_prefix = 1;
2970 	return X86EMUL_CONTINUE;
2971 }
2972 
2973 static int em_imul(struct x86_emulate_ctxt *ctxt)
2974 {
2975 	emulate_2op_SrcV_nobyte(ctxt, "imul");
2976 	return X86EMUL_CONTINUE;
2977 }
2978 
2979 static int em_imul_3op(struct x86_emulate_ctxt *ctxt)
2980 {
2981 	ctxt->dst.val = ctxt->src2.val;
2982 	return em_imul(ctxt);
2983 }
2984 
2985 static int em_cwd(struct x86_emulate_ctxt *ctxt)
2986 {
2987 	ctxt->dst.type = OP_REG;
2988 	ctxt->dst.bytes = ctxt->src.bytes;
2989 	ctxt->dst.addr.reg = reg_rmw(ctxt, VCPU_REGS_RDX);
2990 	ctxt->dst.val = ~((ctxt->src.val >> (ctxt->src.bytes * 8 - 1)) - 1);
2991 
2992 	return X86EMUL_CONTINUE;
2993 }
2994 
2995 static int em_rdtsc(struct x86_emulate_ctxt *ctxt)
2996 {
2997 	u64 tsc = 0;
2998 
2999 	ctxt->ops->get_msr(ctxt, MSR_IA32_TSC, &tsc);
3000 	*reg_write(ctxt, VCPU_REGS_RAX) = (u32)tsc;
3001 	*reg_write(ctxt, VCPU_REGS_RDX) = tsc >> 32;
3002 	return X86EMUL_CONTINUE;
3003 }
3004 
3005 static int em_rdpmc(struct x86_emulate_ctxt *ctxt)
3006 {
3007 	u64 pmc;
3008 
3009 	if (ctxt->ops->read_pmc(ctxt, reg_read(ctxt, VCPU_REGS_RCX), &pmc))
3010 		return emulate_gp(ctxt, 0);
3011 	*reg_write(ctxt, VCPU_REGS_RAX) = (u32)pmc;
3012 	*reg_write(ctxt, VCPU_REGS_RDX) = pmc >> 32;
3013 	return X86EMUL_CONTINUE;
3014 }
3015 
3016 static int em_mov(struct x86_emulate_ctxt *ctxt)
3017 {
3018 	memcpy(ctxt->dst.valptr, ctxt->src.valptr, ctxt->op_bytes);
3019 	return X86EMUL_CONTINUE;
3020 }
3021 
3022 static int em_cr_write(struct x86_emulate_ctxt *ctxt)
3023 {
3024 	if (ctxt->ops->set_cr(ctxt, ctxt->modrm_reg, ctxt->src.val))
3025 		return emulate_gp(ctxt, 0);
3026 
3027 	/* Disable writeback. */
3028 	ctxt->dst.type = OP_NONE;
3029 	return X86EMUL_CONTINUE;
3030 }
3031 
3032 static int em_dr_write(struct x86_emulate_ctxt *ctxt)
3033 {
3034 	unsigned long val;
3035 
3036 	if (ctxt->mode == X86EMUL_MODE_PROT64)
3037 		val = ctxt->src.val & ~0ULL;
3038 	else
3039 		val = ctxt->src.val & ~0U;
3040 
3041 	/* #UD condition is already handled. */
3042 	if (ctxt->ops->set_dr(ctxt, ctxt->modrm_reg, val) < 0)
3043 		return emulate_gp(ctxt, 0);
3044 
3045 	/* Disable writeback. */
3046 	ctxt->dst.type = OP_NONE;
3047 	return X86EMUL_CONTINUE;
3048 }
3049 
3050 static int em_wrmsr(struct x86_emulate_ctxt *ctxt)
3051 {
3052 	u64 msr_data;
3053 
3054 	msr_data = (u32)reg_read(ctxt, VCPU_REGS_RAX)
3055 		| ((u64)reg_read(ctxt, VCPU_REGS_RDX) << 32);
3056 	if (ctxt->ops->set_msr(ctxt, reg_read(ctxt, VCPU_REGS_RCX), msr_data))
3057 		return emulate_gp(ctxt, 0);
3058 
3059 	return X86EMUL_CONTINUE;
3060 }
3061 
3062 static int em_rdmsr(struct x86_emulate_ctxt *ctxt)
3063 {
3064 	u64 msr_data;
3065 
3066 	if (ctxt->ops->get_msr(ctxt, reg_read(ctxt, VCPU_REGS_RCX), &msr_data))
3067 		return emulate_gp(ctxt, 0);
3068 
3069 	*reg_write(ctxt, VCPU_REGS_RAX) = (u32)msr_data;
3070 	*reg_write(ctxt, VCPU_REGS_RDX) = msr_data >> 32;
3071 	return X86EMUL_CONTINUE;
3072 }
3073 
3074 static int em_mov_rm_sreg(struct x86_emulate_ctxt *ctxt)
3075 {
3076 	if (ctxt->modrm_reg > VCPU_SREG_GS)
3077 		return emulate_ud(ctxt);
3078 
3079 	ctxt->dst.val = get_segment_selector(ctxt, ctxt->modrm_reg);
3080 	return X86EMUL_CONTINUE;
3081 }
3082 
3083 static int em_mov_sreg_rm(struct x86_emulate_ctxt *ctxt)
3084 {
3085 	u16 sel = ctxt->src.val;
3086 
3087 	if (ctxt->modrm_reg == VCPU_SREG_CS || ctxt->modrm_reg > VCPU_SREG_GS)
3088 		return emulate_ud(ctxt);
3089 
3090 	if (ctxt->modrm_reg == VCPU_SREG_SS)
3091 		ctxt->interruptibility = KVM_X86_SHADOW_INT_MOV_SS;
3092 
3093 	/* Disable writeback. */
3094 	ctxt->dst.type = OP_NONE;
3095 	return load_segment_descriptor(ctxt, sel, ctxt->modrm_reg);
3096 }
3097 
3098 static int em_lldt(struct x86_emulate_ctxt *ctxt)
3099 {
3100 	u16 sel = ctxt->src.val;
3101 
3102 	/* Disable writeback. */
3103 	ctxt->dst.type = OP_NONE;
3104 	return load_segment_descriptor(ctxt, sel, VCPU_SREG_LDTR);
3105 }
3106 
3107 static int em_ltr(struct x86_emulate_ctxt *ctxt)
3108 {
3109 	u16 sel = ctxt->src.val;
3110 
3111 	/* Disable writeback. */
3112 	ctxt->dst.type = OP_NONE;
3113 	return load_segment_descriptor(ctxt, sel, VCPU_SREG_TR);
3114 }
3115 
3116 static int em_invlpg(struct x86_emulate_ctxt *ctxt)
3117 {
3118 	int rc;
3119 	ulong linear;
3120 
3121 	rc = linearize(ctxt, ctxt->src.addr.mem, 1, false, &linear);
3122 	if (rc == X86EMUL_CONTINUE)
3123 		ctxt->ops->invlpg(ctxt, linear);
3124 	/* Disable writeback. */
3125 	ctxt->dst.type = OP_NONE;
3126 	return X86EMUL_CONTINUE;
3127 }
3128 
3129 static int em_clts(struct x86_emulate_ctxt *ctxt)
3130 {
3131 	ulong cr0;
3132 
3133 	cr0 = ctxt->ops->get_cr(ctxt, 0);
3134 	cr0 &= ~X86_CR0_TS;
3135 	ctxt->ops->set_cr(ctxt, 0, cr0);
3136 	return X86EMUL_CONTINUE;
3137 }
3138 
3139 static int em_vmcall(struct x86_emulate_ctxt *ctxt)
3140 {
3141 	int rc;
3142 
3143 	if (ctxt->modrm_mod != 3 || ctxt->modrm_rm != 1)
3144 		return X86EMUL_UNHANDLEABLE;
3145 
3146 	rc = ctxt->ops->fix_hypercall(ctxt);
3147 	if (rc != X86EMUL_CONTINUE)
3148 		return rc;
3149 
3150 	/* Let the processor re-execute the fixed hypercall */
3151 	ctxt->_eip = ctxt->eip;
3152 	/* Disable writeback. */
3153 	ctxt->dst.type = OP_NONE;
3154 	return X86EMUL_CONTINUE;
3155 }
3156 
3157 static int emulate_store_desc_ptr(struct x86_emulate_ctxt *ctxt,
3158 				  void (*get)(struct x86_emulate_ctxt *ctxt,
3159 					      struct desc_ptr *ptr))
3160 {
3161 	struct desc_ptr desc_ptr;
3162 
3163 	if (ctxt->mode == X86EMUL_MODE_PROT64)
3164 		ctxt->op_bytes = 8;
3165 	get(ctxt, &desc_ptr);
3166 	if (ctxt->op_bytes == 2) {
3167 		ctxt->op_bytes = 4;
3168 		desc_ptr.address &= 0x00ffffff;
3169 	}
3170 	/* Disable writeback. */
3171 	ctxt->dst.type = OP_NONE;
3172 	return segmented_write(ctxt, ctxt->dst.addr.mem,
3173 			       &desc_ptr, 2 + ctxt->op_bytes);
3174 }
3175 
3176 static int em_sgdt(struct x86_emulate_ctxt *ctxt)
3177 {
3178 	return emulate_store_desc_ptr(ctxt, ctxt->ops->get_gdt);
3179 }
3180 
3181 static int em_sidt(struct x86_emulate_ctxt *ctxt)
3182 {
3183 	return emulate_store_desc_ptr(ctxt, ctxt->ops->get_idt);
3184 }
3185 
3186 static int em_lgdt(struct x86_emulate_ctxt *ctxt)
3187 {
3188 	struct desc_ptr desc_ptr;
3189 	int rc;
3190 
3191 	if (ctxt->mode == X86EMUL_MODE_PROT64)
3192 		ctxt->op_bytes = 8;
3193 	rc = read_descriptor(ctxt, ctxt->src.addr.mem,
3194 			     &desc_ptr.size, &desc_ptr.address,
3195 			     ctxt->op_bytes);
3196 	if (rc != X86EMUL_CONTINUE)
3197 		return rc;
3198 	ctxt->ops->set_gdt(ctxt, &desc_ptr);
3199 	/* Disable writeback. */
3200 	ctxt->dst.type = OP_NONE;
3201 	return X86EMUL_CONTINUE;
3202 }
3203 
3204 static int em_vmmcall(struct x86_emulate_ctxt *ctxt)
3205 {
3206 	int rc;
3207 
3208 	rc = ctxt->ops->fix_hypercall(ctxt);
3209 
3210 	/* Disable writeback. */
3211 	ctxt->dst.type = OP_NONE;
3212 	return rc;
3213 }
3214 
3215 static int em_lidt(struct x86_emulate_ctxt *ctxt)
3216 {
3217 	struct desc_ptr desc_ptr;
3218 	int rc;
3219 
3220 	if (ctxt->mode == X86EMUL_MODE_PROT64)
3221 		ctxt->op_bytes = 8;
3222 	rc = read_descriptor(ctxt, ctxt->src.addr.mem,
3223 			     &desc_ptr.size, &desc_ptr.address,
3224 			     ctxt->op_bytes);
3225 	if (rc != X86EMUL_CONTINUE)
3226 		return rc;
3227 	ctxt->ops->set_idt(ctxt, &desc_ptr);
3228 	/* Disable writeback. */
3229 	ctxt->dst.type = OP_NONE;
3230 	return X86EMUL_CONTINUE;
3231 }
3232 
3233 static int em_smsw(struct x86_emulate_ctxt *ctxt)
3234 {
3235 	ctxt->dst.bytes = 2;
3236 	ctxt->dst.val = ctxt->ops->get_cr(ctxt, 0);
3237 	return X86EMUL_CONTINUE;
3238 }
3239 
3240 static int em_lmsw(struct x86_emulate_ctxt *ctxt)
3241 {
3242 	ctxt->ops->set_cr(ctxt, 0, (ctxt->ops->get_cr(ctxt, 0) & ~0x0eul)
3243 			  | (ctxt->src.val & 0x0f));
3244 	ctxt->dst.type = OP_NONE;
3245 	return X86EMUL_CONTINUE;
3246 }
3247 
3248 static int em_loop(struct x86_emulate_ctxt *ctxt)
3249 {
3250 	register_address_increment(ctxt, reg_rmw(ctxt, VCPU_REGS_RCX), -1);
3251 	if ((address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) != 0) &&
3252 	    (ctxt->b == 0xe2 || test_cc(ctxt->b ^ 0x5, ctxt->eflags)))
3253 		jmp_rel(ctxt, ctxt->src.val);
3254 
3255 	return X86EMUL_CONTINUE;
3256 }
3257 
3258 static int em_jcxz(struct x86_emulate_ctxt *ctxt)
3259 {
3260 	if (address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) == 0)
3261 		jmp_rel(ctxt, ctxt->src.val);
3262 
3263 	return X86EMUL_CONTINUE;
3264 }
3265 
3266 static int em_in(struct x86_emulate_ctxt *ctxt)
3267 {
3268 	if (!pio_in_emulated(ctxt, ctxt->dst.bytes, ctxt->src.val,
3269 			     &ctxt->dst.val))
3270 		return X86EMUL_IO_NEEDED;
3271 
3272 	return X86EMUL_CONTINUE;
3273 }
3274 
3275 static int em_out(struct x86_emulate_ctxt *ctxt)
3276 {
3277 	ctxt->ops->pio_out_emulated(ctxt, ctxt->src.bytes, ctxt->dst.val,
3278 				    &ctxt->src.val, 1);
3279 	/* Disable writeback. */
3280 	ctxt->dst.type = OP_NONE;
3281 	return X86EMUL_CONTINUE;
3282 }
3283 
3284 static int em_cli(struct x86_emulate_ctxt *ctxt)
3285 {
3286 	if (emulator_bad_iopl(ctxt))
3287 		return emulate_gp(ctxt, 0);
3288 
3289 	ctxt->eflags &= ~X86_EFLAGS_IF;
3290 	return X86EMUL_CONTINUE;
3291 }
3292 
3293 static int em_sti(struct x86_emulate_ctxt *ctxt)
3294 {
3295 	if (emulator_bad_iopl(ctxt))
3296 		return emulate_gp(ctxt, 0);
3297 
3298 	ctxt->interruptibility = KVM_X86_SHADOW_INT_STI;
3299 	ctxt->eflags |= X86_EFLAGS_IF;
3300 	return X86EMUL_CONTINUE;
3301 }
3302 
3303 static int em_bt(struct x86_emulate_ctxt *ctxt)
3304 {
3305 	/* Disable writeback. */
3306 	ctxt->dst.type = OP_NONE;
3307 	/* only subword offset */
3308 	ctxt->src.val &= (ctxt->dst.bytes << 3) - 1;
3309 
3310 	emulate_2op_SrcV_nobyte(ctxt, "bt");
3311 	return X86EMUL_CONTINUE;
3312 }
3313 
3314 static int em_bts(struct x86_emulate_ctxt *ctxt)
3315 {
3316 	emulate_2op_SrcV_nobyte(ctxt, "bts");
3317 	return X86EMUL_CONTINUE;
3318 }
3319 
3320 static int em_btr(struct x86_emulate_ctxt *ctxt)
3321 {
3322 	emulate_2op_SrcV_nobyte(ctxt, "btr");
3323 	return X86EMUL_CONTINUE;
3324 }
3325 
3326 static int em_btc(struct x86_emulate_ctxt *ctxt)
3327 {
3328 	emulate_2op_SrcV_nobyte(ctxt, "btc");
3329 	return X86EMUL_CONTINUE;
3330 }
3331 
3332 static int em_bsf(struct x86_emulate_ctxt *ctxt)
3333 {
3334 	emulate_2op_SrcV_nobyte(ctxt, "bsf");
3335 	return X86EMUL_CONTINUE;
3336 }
3337 
3338 static int em_bsr(struct x86_emulate_ctxt *ctxt)
3339 {
3340 	emulate_2op_SrcV_nobyte(ctxt, "bsr");
3341 	return X86EMUL_CONTINUE;
3342 }
3343 
3344 static int em_cpuid(struct x86_emulate_ctxt *ctxt)
3345 {
3346 	u32 eax, ebx, ecx, edx;
3347 
3348 	eax = reg_read(ctxt, VCPU_REGS_RAX);
3349 	ecx = reg_read(ctxt, VCPU_REGS_RCX);
3350 	ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
3351 	*reg_write(ctxt, VCPU_REGS_RAX) = eax;
3352 	*reg_write(ctxt, VCPU_REGS_RBX) = ebx;
3353 	*reg_write(ctxt, VCPU_REGS_RCX) = ecx;
3354 	*reg_write(ctxt, VCPU_REGS_RDX) = edx;
3355 	return X86EMUL_CONTINUE;
3356 }
3357 
3358 static int em_lahf(struct x86_emulate_ctxt *ctxt)
3359 {
3360 	*reg_rmw(ctxt, VCPU_REGS_RAX) &= ~0xff00UL;
3361 	*reg_rmw(ctxt, VCPU_REGS_RAX) |= (ctxt->eflags & 0xff) << 8;
3362 	return X86EMUL_CONTINUE;
3363 }
3364 
3365 static int em_bswap(struct x86_emulate_ctxt *ctxt)
3366 {
3367 	switch (ctxt->op_bytes) {
3368 #ifdef CONFIG_X86_64
3369 	case 8:
3370 		asm("bswap %0" : "+r"(ctxt->dst.val));
3371 		break;
3372 #endif
3373 	default:
3374 		asm("bswap %0" : "+r"(*(u32 *)&ctxt->dst.val));
3375 		break;
3376 	}
3377 	return X86EMUL_CONTINUE;
3378 }
3379 
3380 static bool valid_cr(int nr)
3381 {
3382 	switch (nr) {
3383 	case 0:
3384 	case 2 ... 4:
3385 	case 8:
3386 		return true;
3387 	default:
3388 		return false;
3389 	}
3390 }
3391 
3392 static int check_cr_read(struct x86_emulate_ctxt *ctxt)
3393 {
3394 	if (!valid_cr(ctxt->modrm_reg))
3395 		return emulate_ud(ctxt);
3396 
3397 	return X86EMUL_CONTINUE;
3398 }
3399 
3400 static int check_cr_write(struct x86_emulate_ctxt *ctxt)
3401 {
3402 	u64 new_val = ctxt->src.val64;
3403 	int cr = ctxt->modrm_reg;
3404 	u64 efer = 0;
3405 
3406 	static u64 cr_reserved_bits[] = {
3407 		0xffffffff00000000ULL,
3408 		0, 0, 0, /* CR3 checked later */
3409 		CR4_RESERVED_BITS,
3410 		0, 0, 0,
3411 		CR8_RESERVED_BITS,
3412 	};
3413 
3414 	if (!valid_cr(cr))
3415 		return emulate_ud(ctxt);
3416 
3417 	if (new_val & cr_reserved_bits[cr])
3418 		return emulate_gp(ctxt, 0);
3419 
3420 	switch (cr) {
3421 	case 0: {
3422 		u64 cr4;
3423 		if (((new_val & X86_CR0_PG) && !(new_val & X86_CR0_PE)) ||
3424 		    ((new_val & X86_CR0_NW) && !(new_val & X86_CR0_CD)))
3425 			return emulate_gp(ctxt, 0);
3426 
3427 		cr4 = ctxt->ops->get_cr(ctxt, 4);
3428 		ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3429 
3430 		if ((new_val & X86_CR0_PG) && (efer & EFER_LME) &&
3431 		    !(cr4 & X86_CR4_PAE))
3432 			return emulate_gp(ctxt, 0);
3433 
3434 		break;
3435 		}
3436 	case 3: {
3437 		u64 rsvd = 0;
3438 
3439 		ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3440 		if (efer & EFER_LMA)
3441 			rsvd = CR3_L_MODE_RESERVED_BITS;
3442 		else if (ctxt->ops->get_cr(ctxt, 4) & X86_CR4_PAE)
3443 			rsvd = CR3_PAE_RESERVED_BITS;
3444 		else if (ctxt->ops->get_cr(ctxt, 0) & X86_CR0_PG)
3445 			rsvd = CR3_NONPAE_RESERVED_BITS;
3446 
3447 		if (new_val & rsvd)
3448 			return emulate_gp(ctxt, 0);
3449 
3450 		break;
3451 		}
3452 	case 4: {
3453 		ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3454 
3455 		if ((efer & EFER_LMA) && !(new_val & X86_CR4_PAE))
3456 			return emulate_gp(ctxt, 0);
3457 
3458 		break;
3459 		}
3460 	}
3461 
3462 	return X86EMUL_CONTINUE;
3463 }
3464 
3465 static int check_dr7_gd(struct x86_emulate_ctxt *ctxt)
3466 {
3467 	unsigned long dr7;
3468 
3469 	ctxt->ops->get_dr(ctxt, 7, &dr7);
3470 
3471 	/* Check if DR7.Global_Enable is set */
3472 	return dr7 & (1 << 13);
3473 }
3474 
3475 static int check_dr_read(struct x86_emulate_ctxt *ctxt)
3476 {
3477 	int dr = ctxt->modrm_reg;
3478 	u64 cr4;
3479 
3480 	if (dr > 7)
3481 		return emulate_ud(ctxt);
3482 
3483 	cr4 = ctxt->ops->get_cr(ctxt, 4);
3484 	if ((cr4 & X86_CR4_DE) && (dr == 4 || dr == 5))
3485 		return emulate_ud(ctxt);
3486 
3487 	if (check_dr7_gd(ctxt))
3488 		return emulate_db(ctxt);
3489 
3490 	return X86EMUL_CONTINUE;
3491 }
3492 
3493 static int check_dr_write(struct x86_emulate_ctxt *ctxt)
3494 {
3495 	u64 new_val = ctxt->src.val64;
3496 	int dr = ctxt->modrm_reg;
3497 
3498 	if ((dr == 6 || dr == 7) && (new_val & 0xffffffff00000000ULL))
3499 		return emulate_gp(ctxt, 0);
3500 
3501 	return check_dr_read(ctxt);
3502 }
3503 
3504 static int check_svme(struct x86_emulate_ctxt *ctxt)
3505 {
3506 	u64 efer;
3507 
3508 	ctxt->ops->get_msr(ctxt, MSR_EFER, &efer);
3509 
3510 	if (!(efer & EFER_SVME))
3511 		return emulate_ud(ctxt);
3512 
3513 	return X86EMUL_CONTINUE;
3514 }
3515 
3516 static int check_svme_pa(struct x86_emulate_ctxt *ctxt)
3517 {
3518 	u64 rax = reg_read(ctxt, VCPU_REGS_RAX);
3519 
3520 	/* Valid physical address? */
3521 	if (rax & 0xffff000000000000ULL)
3522 		return emulate_gp(ctxt, 0);
3523 
3524 	return check_svme(ctxt);
3525 }
3526 
3527 static int check_rdtsc(struct x86_emulate_ctxt *ctxt)
3528 {
3529 	u64 cr4 = ctxt->ops->get_cr(ctxt, 4);
3530 
3531 	if (cr4 & X86_CR4_TSD && ctxt->ops->cpl(ctxt))
3532 		return emulate_ud(ctxt);
3533 
3534 	return X86EMUL_CONTINUE;
3535 }
3536 
3537 static int check_rdpmc(struct x86_emulate_ctxt *ctxt)
3538 {
3539 	u64 cr4 = ctxt->ops->get_cr(ctxt, 4);
3540 	u64 rcx = reg_read(ctxt, VCPU_REGS_RCX);
3541 
3542 	if ((!(cr4 & X86_CR4_PCE) && ctxt->ops->cpl(ctxt)) ||
3543 	    (rcx > 3))
3544 		return emulate_gp(ctxt, 0);
3545 
3546 	return X86EMUL_CONTINUE;
3547 }
3548 
3549 static int check_perm_in(struct x86_emulate_ctxt *ctxt)
3550 {
3551 	ctxt->dst.bytes = min(ctxt->dst.bytes, 4u);
3552 	if (!emulator_io_permited(ctxt, ctxt->src.val, ctxt->dst.bytes))
3553 		return emulate_gp(ctxt, 0);
3554 
3555 	return X86EMUL_CONTINUE;
3556 }
3557 
3558 static int check_perm_out(struct x86_emulate_ctxt *ctxt)
3559 {
3560 	ctxt->src.bytes = min(ctxt->src.bytes, 4u);
3561 	if (!emulator_io_permited(ctxt, ctxt->dst.val, ctxt->src.bytes))
3562 		return emulate_gp(ctxt, 0);
3563 
3564 	return X86EMUL_CONTINUE;
3565 }
3566 
3567 #define D(_y) { .flags = (_y) }
3568 #define DI(_y, _i) { .flags = (_y), .intercept = x86_intercept_##_i }
3569 #define DIP(_y, _i, _p) { .flags = (_y), .intercept = x86_intercept_##_i, \
3570 		      .check_perm = (_p) }
3571 #define N    D(0)
3572 #define EXT(_f, _e) { .flags = ((_f) | RMExt), .u.group = (_e) }
3573 #define G(_f, _g) { .flags = ((_f) | Group | ModRM), .u.group = (_g) }
3574 #define GD(_f, _g) { .flags = ((_f) | GroupDual | ModRM), .u.gdual = (_g) }
3575 #define I(_f, _e) { .flags = (_f), .u.execute = (_e) }
3576 #define II(_f, _e, _i) \
3577 	{ .flags = (_f), .u.execute = (_e), .intercept = x86_intercept_##_i }
3578 #define IIP(_f, _e, _i, _p) \
3579 	{ .flags = (_f), .u.execute = (_e), .intercept = x86_intercept_##_i, \
3580 	  .check_perm = (_p) }
3581 #define GP(_f, _g) { .flags = ((_f) | Prefix), .u.gprefix = (_g) }
3582 
3583 #define D2bv(_f)      D((_f) | ByteOp), D(_f)
3584 #define D2bvIP(_f, _i, _p) DIP((_f) | ByteOp, _i, _p), DIP(_f, _i, _p)
3585 #define I2bv(_f, _e)  I((_f) | ByteOp, _e), I(_f, _e)
3586 #define I2bvIP(_f, _e, _i, _p) \
3587 	IIP((_f) | ByteOp, _e, _i, _p), IIP(_f, _e, _i, _p)
3588 
3589 #define I6ALU(_f, _e) I2bv((_f) | DstMem | SrcReg | ModRM, _e),		\
3590 		I2bv(((_f) | DstReg | SrcMem | ModRM) & ~Lock, _e),	\
3591 		I2bv(((_f) & ~Lock) | DstAcc | SrcImm, _e)
3592 
3593 static const struct opcode group7_rm1[] = {
3594 	DI(SrcNone | Priv, monitor),
3595 	DI(SrcNone | Priv, mwait),
3596 	N, N, N, N, N, N,
3597 };
3598 
3599 static const struct opcode group7_rm3[] = {
3600 	DIP(SrcNone | Prot | Priv,		vmrun,		check_svme_pa),
3601 	II(SrcNone  | Prot | VendorSpecific,	em_vmmcall,	vmmcall),
3602 	DIP(SrcNone | Prot | Priv,		vmload,		check_svme_pa),
3603 	DIP(SrcNone | Prot | Priv,		vmsave,		check_svme_pa),
3604 	DIP(SrcNone | Prot | Priv,		stgi,		check_svme),
3605 	DIP(SrcNone | Prot | Priv,		clgi,		check_svme),
3606 	DIP(SrcNone | Prot | Priv,		skinit,		check_svme),
3607 	DIP(SrcNone | Prot | Priv,		invlpga,	check_svme),
3608 };
3609 
3610 static const struct opcode group7_rm7[] = {
3611 	N,
3612 	DIP(SrcNone, rdtscp, check_rdtsc),
3613 	N, N, N, N, N, N,
3614 };
3615 
3616 static const struct opcode group1[] = {
3617 	I(Lock, em_add),
3618 	I(Lock | PageTable, em_or),
3619 	I(Lock, em_adc),
3620 	I(Lock, em_sbb),
3621 	I(Lock | PageTable, em_and),
3622 	I(Lock, em_sub),
3623 	I(Lock, em_xor),
3624 	I(0, em_cmp),
3625 };
3626 
3627 static const struct opcode group1A[] = {
3628 	I(DstMem | SrcNone | Mov | Stack, em_pop), N, N, N, N, N, N, N,
3629 };
3630 
3631 static const struct opcode group3[] = {
3632 	I(DstMem | SrcImm, em_test),
3633 	I(DstMem | SrcImm, em_test),
3634 	I(DstMem | SrcNone | Lock, em_not),
3635 	I(DstMem | SrcNone | Lock, em_neg),
3636 	I(SrcMem, em_mul_ex),
3637 	I(SrcMem, em_imul_ex),
3638 	I(SrcMem, em_div_ex),
3639 	I(SrcMem, em_idiv_ex),
3640 };
3641 
3642 static const struct opcode group4[] = {
3643 	I(ByteOp | DstMem | SrcNone | Lock, em_grp45),
3644 	I(ByteOp | DstMem | SrcNone | Lock, em_grp45),
3645 	N, N, N, N, N, N,
3646 };
3647 
3648 static const struct opcode group5[] = {
3649 	I(DstMem | SrcNone | Lock,		em_grp45),
3650 	I(DstMem | SrcNone | Lock,		em_grp45),
3651 	I(SrcMem | Stack,			em_grp45),
3652 	I(SrcMemFAddr | ImplicitOps | Stack,	em_call_far),
3653 	I(SrcMem | Stack,			em_grp45),
3654 	I(SrcMemFAddr | ImplicitOps,		em_grp45),
3655 	I(SrcMem | Stack,			em_grp45), N,
3656 };
3657 
3658 static const struct opcode group6[] = {
3659 	DI(Prot,	sldt),
3660 	DI(Prot,	str),
3661 	II(Prot | Priv | SrcMem16, em_lldt, lldt),
3662 	II(Prot | Priv | SrcMem16, em_ltr, ltr),
3663 	N, N, N, N,
3664 };
3665 
3666 static const struct group_dual group7 = { {
3667 	II(Mov | DstMem | Priv,			em_sgdt, sgdt),
3668 	II(Mov | DstMem | Priv,			em_sidt, sidt),
3669 	II(SrcMem | Priv,			em_lgdt, lgdt),
3670 	II(SrcMem | Priv,			em_lidt, lidt),
3671 	II(SrcNone | DstMem | Mov,		em_smsw, smsw), N,
3672 	II(SrcMem16 | Mov | Priv,		em_lmsw, lmsw),
3673 	II(SrcMem | ByteOp | Priv | NoAccess,	em_invlpg, invlpg),
3674 }, {
3675 	I(SrcNone | Priv | VendorSpecific,	em_vmcall),
3676 	EXT(0, group7_rm1),
3677 	N, EXT(0, group7_rm3),
3678 	II(SrcNone | DstMem | Mov,		em_smsw, smsw), N,
3679 	II(SrcMem16 | Mov | Priv,		em_lmsw, lmsw),
3680 	EXT(0, group7_rm7),
3681 } };
3682 
3683 static const struct opcode group8[] = {
3684 	N, N, N, N,
3685 	I(DstMem | SrcImmByte,				em_bt),
3686 	I(DstMem | SrcImmByte | Lock | PageTable,	em_bts),
3687 	I(DstMem | SrcImmByte | Lock,			em_btr),
3688 	I(DstMem | SrcImmByte | Lock | PageTable,	em_btc),
3689 };
3690 
3691 static const struct group_dual group9 = { {
3692 	N, I(DstMem64 | Lock | PageTable, em_cmpxchg8b), N, N, N, N, N, N,
3693 }, {
3694 	N, N, N, N, N, N, N, N,
3695 } };
3696 
3697 static const struct opcode group11[] = {
3698 	I(DstMem | SrcImm | Mov | PageTable, em_mov),
3699 	X7(D(Undefined)),
3700 };
3701 
3702 static const struct gprefix pfx_0f_6f_0f_7f = {
3703 	I(Mmx, em_mov), I(Sse | Aligned, em_mov), N, I(Sse | Unaligned, em_mov),
3704 };
3705 
3706 static const struct gprefix pfx_vmovntpx = {
3707 	I(0, em_mov), N, N, N,
3708 };
3709 
3710 static const struct opcode opcode_table[256] = {
3711 	/* 0x00 - 0x07 */
3712 	I6ALU(Lock, em_add),
3713 	I(ImplicitOps | Stack | No64 | Src2ES, em_push_sreg),
3714 	I(ImplicitOps | Stack | No64 | Src2ES, em_pop_sreg),
3715 	/* 0x08 - 0x0F */
3716 	I6ALU(Lock | PageTable, em_or),
3717 	I(ImplicitOps | Stack | No64 | Src2CS, em_push_sreg),
3718 	N,
3719 	/* 0x10 - 0x17 */
3720 	I6ALU(Lock, em_adc),
3721 	I(ImplicitOps | Stack | No64 | Src2SS, em_push_sreg),
3722 	I(ImplicitOps | Stack | No64 | Src2SS, em_pop_sreg),
3723 	/* 0x18 - 0x1F */
3724 	I6ALU(Lock, em_sbb),
3725 	I(ImplicitOps | Stack | No64 | Src2DS, em_push_sreg),
3726 	I(ImplicitOps | Stack | No64 | Src2DS, em_pop_sreg),
3727 	/* 0x20 - 0x27 */
3728 	I6ALU(Lock | PageTable, em_and), N, N,
3729 	/* 0x28 - 0x2F */
3730 	I6ALU(Lock, em_sub), N, I(ByteOp | DstAcc | No64, em_das),
3731 	/* 0x30 - 0x37 */
3732 	I6ALU(Lock, em_xor), N, N,
3733 	/* 0x38 - 0x3F */
3734 	I6ALU(0, em_cmp), N, N,
3735 	/* 0x40 - 0x4F */
3736 	X16(D(DstReg)),
3737 	/* 0x50 - 0x57 */
3738 	X8(I(SrcReg | Stack, em_push)),
3739 	/* 0x58 - 0x5F */
3740 	X8(I(DstReg | Stack, em_pop)),
3741 	/* 0x60 - 0x67 */
3742 	I(ImplicitOps | Stack | No64, em_pusha),
3743 	I(ImplicitOps | Stack | No64, em_popa),
3744 	N, D(DstReg | SrcMem32 | ModRM | Mov) /* movsxd (x86/64) */ ,
3745 	N, N, N, N,
3746 	/* 0x68 - 0x6F */
3747 	I(SrcImm | Mov | Stack, em_push),
3748 	I(DstReg | SrcMem | ModRM | Src2Imm, em_imul_3op),
3749 	I(SrcImmByte | Mov | Stack, em_push),
3750 	I(DstReg | SrcMem | ModRM | Src2ImmByte, em_imul_3op),
3751 	I2bvIP(DstDI | SrcDX | Mov | String | Unaligned, em_in, ins, check_perm_in), /* insb, insw/insd */
3752 	I2bvIP(SrcSI | DstDX | String, em_out, outs, check_perm_out), /* outsb, outsw/outsd */
3753 	/* 0x70 - 0x7F */
3754 	X16(D(SrcImmByte)),
3755 	/* 0x80 - 0x87 */
3756 	G(ByteOp | DstMem | SrcImm, group1),
3757 	G(DstMem | SrcImm, group1),
3758 	G(ByteOp | DstMem | SrcImm | No64, group1),
3759 	G(DstMem | SrcImmByte, group1),
3760 	I2bv(DstMem | SrcReg | ModRM, em_test),
3761 	I2bv(DstMem | SrcReg | ModRM | Lock | PageTable, em_xchg),
3762 	/* 0x88 - 0x8F */
3763 	I2bv(DstMem | SrcReg | ModRM | Mov | PageTable, em_mov),
3764 	I2bv(DstReg | SrcMem | ModRM | Mov, em_mov),
3765 	I(DstMem | SrcNone | ModRM | Mov | PageTable, em_mov_rm_sreg),
3766 	D(ModRM | SrcMem | NoAccess | DstReg),
3767 	I(ImplicitOps | SrcMem16 | ModRM, em_mov_sreg_rm),
3768 	G(0, group1A),
3769 	/* 0x90 - 0x97 */
3770 	DI(SrcAcc | DstReg, pause), X7(D(SrcAcc | DstReg)),
3771 	/* 0x98 - 0x9F */
3772 	D(DstAcc | SrcNone), I(ImplicitOps | SrcAcc, em_cwd),
3773 	I(SrcImmFAddr | No64, em_call_far), N,
3774 	II(ImplicitOps | Stack, em_pushf, pushf),
3775 	II(ImplicitOps | Stack, em_popf, popf), N, I(ImplicitOps, em_lahf),
3776 	/* 0xA0 - 0xA7 */
3777 	I2bv(DstAcc | SrcMem | Mov | MemAbs, em_mov),
3778 	I2bv(DstMem | SrcAcc | Mov | MemAbs | PageTable, em_mov),
3779 	I2bv(SrcSI | DstDI | Mov | String, em_mov),
3780 	I2bv(SrcSI | DstDI | String, em_cmp),
3781 	/* 0xA8 - 0xAF */
3782 	I2bv(DstAcc | SrcImm, em_test),
3783 	I2bv(SrcAcc | DstDI | Mov | String, em_mov),
3784 	I2bv(SrcSI | DstAcc | Mov | String, em_mov),
3785 	I2bv(SrcAcc | DstDI | String, em_cmp),
3786 	/* 0xB0 - 0xB7 */
3787 	X8(I(ByteOp | DstReg | SrcImm | Mov, em_mov)),
3788 	/* 0xB8 - 0xBF */
3789 	X8(I(DstReg | SrcImm | Mov, em_mov)),
3790 	/* 0xC0 - 0xC7 */
3791 	D2bv(DstMem | SrcImmByte | ModRM),
3792 	I(ImplicitOps | Stack | SrcImmU16, em_ret_near_imm),
3793 	I(ImplicitOps | Stack, em_ret),
3794 	I(DstReg | SrcMemFAddr | ModRM | No64 | Src2ES, em_lseg),
3795 	I(DstReg | SrcMemFAddr | ModRM | No64 | Src2DS, em_lseg),
3796 	G(ByteOp, group11), G(0, group11),
3797 	/* 0xC8 - 0xCF */
3798 	I(Stack | SrcImmU16 | Src2ImmByte, em_enter), I(Stack, em_leave),
3799 	N, I(ImplicitOps | Stack, em_ret_far),
3800 	D(ImplicitOps), DI(SrcImmByte, intn),
3801 	D(ImplicitOps | No64), II(ImplicitOps, em_iret, iret),
3802 	/* 0xD0 - 0xD7 */
3803 	D2bv(DstMem | SrcOne | ModRM), D2bv(DstMem | ModRM),
3804 	N, N, N, N,
3805 	/* 0xD8 - 0xDF */
3806 	N, N, N, N, N, N, N, N,
3807 	/* 0xE0 - 0xE7 */
3808 	X3(I(SrcImmByte, em_loop)),
3809 	I(SrcImmByte, em_jcxz),
3810 	I2bvIP(SrcImmUByte | DstAcc, em_in,  in,  check_perm_in),
3811 	I2bvIP(SrcAcc | DstImmUByte, em_out, out, check_perm_out),
3812 	/* 0xE8 - 0xEF */
3813 	I(SrcImm | Stack, em_call), D(SrcImm | ImplicitOps),
3814 	I(SrcImmFAddr | No64, em_jmp_far), D(SrcImmByte | ImplicitOps),
3815 	I2bvIP(SrcDX | DstAcc, em_in,  in,  check_perm_in),
3816 	I2bvIP(SrcAcc | DstDX, em_out, out, check_perm_out),
3817 	/* 0xF0 - 0xF7 */
3818 	N, DI(ImplicitOps, icebp), N, N,
3819 	DI(ImplicitOps | Priv, hlt), D(ImplicitOps),
3820 	G(ByteOp, group3), G(0, group3),
3821 	/* 0xF8 - 0xFF */
3822 	D(ImplicitOps), D(ImplicitOps),
3823 	I(ImplicitOps, em_cli), I(ImplicitOps, em_sti),
3824 	D(ImplicitOps), D(ImplicitOps), G(0, group4), G(0, group5),
3825 };
3826 
3827 static const struct opcode twobyte_table[256] = {
3828 	/* 0x00 - 0x0F */
3829 	G(0, group6), GD(0, &group7), N, N,
3830 	N, I(ImplicitOps | VendorSpecific, em_syscall),
3831 	II(ImplicitOps | Priv, em_clts, clts), N,
3832 	DI(ImplicitOps | Priv, invd), DI(ImplicitOps | Priv, wbinvd), N, N,
3833 	N, D(ImplicitOps | ModRM), N, N,
3834 	/* 0x10 - 0x1F */
3835 	N, N, N, N, N, N, N, N, D(ImplicitOps | ModRM), N, N, N, N, N, N, N,
3836 	/* 0x20 - 0x2F */
3837 	DIP(ModRM | DstMem | Priv | Op3264, cr_read, check_cr_read),
3838 	DIP(ModRM | DstMem | Priv | Op3264, dr_read, check_dr_read),
3839 	IIP(ModRM | SrcMem | Priv | Op3264, em_cr_write, cr_write, check_cr_write),
3840 	IIP(ModRM | SrcMem | Priv | Op3264, em_dr_write, dr_write, check_dr_write),
3841 	N, N, N, N,
3842 	N, N, N, GP(ModRM | DstMem | SrcReg | Sse | Mov | Aligned, &pfx_vmovntpx),
3843 	N, N, N, N,
3844 	/* 0x30 - 0x3F */
3845 	II(ImplicitOps | Priv, em_wrmsr, wrmsr),
3846 	IIP(ImplicitOps, em_rdtsc, rdtsc, check_rdtsc),
3847 	II(ImplicitOps | Priv, em_rdmsr, rdmsr),
3848 	IIP(ImplicitOps, em_rdpmc, rdpmc, check_rdpmc),
3849 	I(ImplicitOps | VendorSpecific, em_sysenter),
3850 	I(ImplicitOps | Priv | VendorSpecific, em_sysexit),
3851 	N, N,
3852 	N, N, N, N, N, N, N, N,
3853 	/* 0x40 - 0x4F */
3854 	X16(D(DstReg | SrcMem | ModRM | Mov)),
3855 	/* 0x50 - 0x5F */
3856 	N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
3857 	/* 0x60 - 0x6F */
3858 	N, N, N, N,
3859 	N, N, N, N,
3860 	N, N, N, N,
3861 	N, N, N, GP(SrcMem | DstReg | ModRM | Mov, &pfx_0f_6f_0f_7f),
3862 	/* 0x70 - 0x7F */
3863 	N, N, N, N,
3864 	N, N, N, N,
3865 	N, N, N, N,
3866 	N, N, N, GP(SrcReg | DstMem | ModRM | Mov, &pfx_0f_6f_0f_7f),
3867 	/* 0x80 - 0x8F */
3868 	X16(D(SrcImm)),
3869 	/* 0x90 - 0x9F */
3870 	X16(D(ByteOp | DstMem | SrcNone | ModRM| Mov)),
3871 	/* 0xA0 - 0xA7 */
3872 	I(Stack | Src2FS, em_push_sreg), I(Stack | Src2FS, em_pop_sreg),
3873 	II(ImplicitOps, em_cpuid, cpuid), I(DstMem | SrcReg | ModRM | BitOp, em_bt),
3874 	D(DstMem | SrcReg | Src2ImmByte | ModRM),
3875 	D(DstMem | SrcReg | Src2CL | ModRM), N, N,
3876 	/* 0xA8 - 0xAF */
3877 	I(Stack | Src2GS, em_push_sreg), I(Stack | Src2GS, em_pop_sreg),
3878 	DI(ImplicitOps, rsm),
3879 	I(DstMem | SrcReg | ModRM | BitOp | Lock | PageTable, em_bts),
3880 	D(DstMem | SrcReg | Src2ImmByte | ModRM),
3881 	D(DstMem | SrcReg | Src2CL | ModRM),
3882 	D(ModRM), I(DstReg | SrcMem | ModRM, em_imul),
3883 	/* 0xB0 - 0xB7 */
3884 	I2bv(DstMem | SrcReg | ModRM | Lock | PageTable, em_cmpxchg),
3885 	I(DstReg | SrcMemFAddr | ModRM | Src2SS, em_lseg),
3886 	I(DstMem | SrcReg | ModRM | BitOp | Lock, em_btr),
3887 	I(DstReg | SrcMemFAddr | ModRM | Src2FS, em_lseg),
3888 	I(DstReg | SrcMemFAddr | ModRM | Src2GS, em_lseg),
3889 	D(DstReg | SrcMem8 | ModRM | Mov), D(DstReg | SrcMem16 | ModRM | Mov),
3890 	/* 0xB8 - 0xBF */
3891 	N, N,
3892 	G(BitOp, group8),
3893 	I(DstMem | SrcReg | ModRM | BitOp | Lock | PageTable, em_btc),
3894 	I(DstReg | SrcMem | ModRM, em_bsf), I(DstReg | SrcMem | ModRM, em_bsr),
3895 	D(DstReg | SrcMem8 | ModRM | Mov), D(DstReg | SrcMem16 | ModRM | Mov),
3896 	/* 0xC0 - 0xC7 */
3897 	D2bv(DstMem | SrcReg | ModRM | Lock),
3898 	N, D(DstMem | SrcReg | ModRM | Mov),
3899 	N, N, N, GD(0, &group9),
3900 	/* 0xC8 - 0xCF */
3901 	X8(I(DstReg, em_bswap)),
3902 	/* 0xD0 - 0xDF */
3903 	N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
3904 	/* 0xE0 - 0xEF */
3905 	N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
3906 	/* 0xF0 - 0xFF */
3907 	N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N
3908 };
3909 
3910 #undef D
3911 #undef N
3912 #undef G
3913 #undef GD
3914 #undef I
3915 #undef GP
3916 #undef EXT
3917 
3918 #undef D2bv
3919 #undef D2bvIP
3920 #undef I2bv
3921 #undef I2bvIP
3922 #undef I6ALU
3923 
3924 static unsigned imm_size(struct x86_emulate_ctxt *ctxt)
3925 {
3926 	unsigned size;
3927 
3928 	size = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
3929 	if (size == 8)
3930 		size = 4;
3931 	return size;
3932 }
3933 
3934 static int decode_imm(struct x86_emulate_ctxt *ctxt, struct operand *op,
3935 		      unsigned size, bool sign_extension)
3936 {
3937 	int rc = X86EMUL_CONTINUE;
3938 
3939 	op->type = OP_IMM;
3940 	op->bytes = size;
3941 	op->addr.mem.ea = ctxt->_eip;
3942 	/* NB. Immediates are sign-extended as necessary. */
3943 	switch (op->bytes) {
3944 	case 1:
3945 		op->val = insn_fetch(s8, ctxt);
3946 		break;
3947 	case 2:
3948 		op->val = insn_fetch(s16, ctxt);
3949 		break;
3950 	case 4:
3951 		op->val = insn_fetch(s32, ctxt);
3952 		break;
3953 	}
3954 	if (!sign_extension) {
3955 		switch (op->bytes) {
3956 		case 1:
3957 			op->val &= 0xff;
3958 			break;
3959 		case 2:
3960 			op->val &= 0xffff;
3961 			break;
3962 		case 4:
3963 			op->val &= 0xffffffff;
3964 			break;
3965 		}
3966 	}
3967 done:
3968 	return rc;
3969 }
3970 
3971 static int decode_operand(struct x86_emulate_ctxt *ctxt, struct operand *op,
3972 			  unsigned d)
3973 {
3974 	int rc = X86EMUL_CONTINUE;
3975 
3976 	switch (d) {
3977 	case OpReg:
3978 		decode_register_operand(ctxt, op);
3979 		break;
3980 	case OpImmUByte:
3981 		rc = decode_imm(ctxt, op, 1, false);
3982 		break;
3983 	case OpMem:
3984 		ctxt->memop.bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
3985 	mem_common:
3986 		*op = ctxt->memop;
3987 		ctxt->memopp = op;
3988 		if ((ctxt->d & BitOp) && op == &ctxt->dst)
3989 			fetch_bit_operand(ctxt);
3990 		op->orig_val = op->val;
3991 		break;
3992 	case OpMem64:
3993 		ctxt->memop.bytes = 8;
3994 		goto mem_common;
3995 	case OpAcc:
3996 		op->type = OP_REG;
3997 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
3998 		op->addr.reg = reg_rmw(ctxt, VCPU_REGS_RAX);
3999 		fetch_register_operand(op);
4000 		op->orig_val = op->val;
4001 		break;
4002 	case OpDI:
4003 		op->type = OP_MEM;
4004 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4005 		op->addr.mem.ea =
4006 			register_address(ctxt, reg_read(ctxt, VCPU_REGS_RDI));
4007 		op->addr.mem.seg = VCPU_SREG_ES;
4008 		op->val = 0;
4009 		op->count = 1;
4010 		break;
4011 	case OpDX:
4012 		op->type = OP_REG;
4013 		op->bytes = 2;
4014 		op->addr.reg = reg_rmw(ctxt, VCPU_REGS_RDX);
4015 		fetch_register_operand(op);
4016 		break;
4017 	case OpCL:
4018 		op->bytes = 1;
4019 		op->val = reg_read(ctxt, VCPU_REGS_RCX) & 0xff;
4020 		break;
4021 	case OpImmByte:
4022 		rc = decode_imm(ctxt, op, 1, true);
4023 		break;
4024 	case OpOne:
4025 		op->bytes = 1;
4026 		op->val = 1;
4027 		break;
4028 	case OpImm:
4029 		rc = decode_imm(ctxt, op, imm_size(ctxt), true);
4030 		break;
4031 	case OpMem8:
4032 		ctxt->memop.bytes = 1;
4033 		goto mem_common;
4034 	case OpMem16:
4035 		ctxt->memop.bytes = 2;
4036 		goto mem_common;
4037 	case OpMem32:
4038 		ctxt->memop.bytes = 4;
4039 		goto mem_common;
4040 	case OpImmU16:
4041 		rc = decode_imm(ctxt, op, 2, false);
4042 		break;
4043 	case OpImmU:
4044 		rc = decode_imm(ctxt, op, imm_size(ctxt), false);
4045 		break;
4046 	case OpSI:
4047 		op->type = OP_MEM;
4048 		op->bytes = (ctxt->d & ByteOp) ? 1 : ctxt->op_bytes;
4049 		op->addr.mem.ea =
4050 			register_address(ctxt, reg_read(ctxt, VCPU_REGS_RSI));
4051 		op->addr.mem.seg = seg_override(ctxt);
4052 		op->val = 0;
4053 		op->count = 1;
4054 		break;
4055 	case OpImmFAddr:
4056 		op->type = OP_IMM;
4057 		op->addr.mem.ea = ctxt->_eip;
4058 		op->bytes = ctxt->op_bytes + 2;
4059 		insn_fetch_arr(op->valptr, op->bytes, ctxt);
4060 		break;
4061 	case OpMemFAddr:
4062 		ctxt->memop.bytes = ctxt->op_bytes + 2;
4063 		goto mem_common;
4064 	case OpES:
4065 		op->val = VCPU_SREG_ES;
4066 		break;
4067 	case OpCS:
4068 		op->val = VCPU_SREG_CS;
4069 		break;
4070 	case OpSS:
4071 		op->val = VCPU_SREG_SS;
4072 		break;
4073 	case OpDS:
4074 		op->val = VCPU_SREG_DS;
4075 		break;
4076 	case OpFS:
4077 		op->val = VCPU_SREG_FS;
4078 		break;
4079 	case OpGS:
4080 		op->val = VCPU_SREG_GS;
4081 		break;
4082 	case OpImplicit:
4083 		/* Special instructions do their own operand decoding. */
4084 	default:
4085 		op->type = OP_NONE; /* Disable writeback. */
4086 		break;
4087 	}
4088 
4089 done:
4090 	return rc;
4091 }
4092 
4093 int x86_decode_insn(struct x86_emulate_ctxt *ctxt, void *insn, int insn_len)
4094 {
4095 	int rc = X86EMUL_CONTINUE;
4096 	int mode = ctxt->mode;
4097 	int def_op_bytes, def_ad_bytes, goffset, simd_prefix;
4098 	bool op_prefix = false;
4099 	struct opcode opcode;
4100 
4101 	ctxt->memop.type = OP_NONE;
4102 	ctxt->memopp = NULL;
4103 	ctxt->_eip = ctxt->eip;
4104 	ctxt->fetch.start = ctxt->_eip;
4105 	ctxt->fetch.end = ctxt->fetch.start + insn_len;
4106 	if (insn_len > 0)
4107 		memcpy(ctxt->fetch.data, insn, insn_len);
4108 
4109 	switch (mode) {
4110 	case X86EMUL_MODE_REAL:
4111 	case X86EMUL_MODE_VM86:
4112 	case X86EMUL_MODE_PROT16:
4113 		def_op_bytes = def_ad_bytes = 2;
4114 		break;
4115 	case X86EMUL_MODE_PROT32:
4116 		def_op_bytes = def_ad_bytes = 4;
4117 		break;
4118 #ifdef CONFIG_X86_64
4119 	case X86EMUL_MODE_PROT64:
4120 		def_op_bytes = 4;
4121 		def_ad_bytes = 8;
4122 		break;
4123 #endif
4124 	default:
4125 		return EMULATION_FAILED;
4126 	}
4127 
4128 	ctxt->op_bytes = def_op_bytes;
4129 	ctxt->ad_bytes = def_ad_bytes;
4130 
4131 	/* Legacy prefixes. */
4132 	for (;;) {
4133 		switch (ctxt->b = insn_fetch(u8, ctxt)) {
4134 		case 0x66:	/* operand-size override */
4135 			op_prefix = true;
4136 			/* switch between 2/4 bytes */
4137 			ctxt->op_bytes = def_op_bytes ^ 6;
4138 			break;
4139 		case 0x67:	/* address-size override */
4140 			if (mode == X86EMUL_MODE_PROT64)
4141 				/* switch between 4/8 bytes */
4142 				ctxt->ad_bytes = def_ad_bytes ^ 12;
4143 			else
4144 				/* switch between 2/4 bytes */
4145 				ctxt->ad_bytes = def_ad_bytes ^ 6;
4146 			break;
4147 		case 0x26:	/* ES override */
4148 		case 0x2e:	/* CS override */
4149 		case 0x36:	/* SS override */
4150 		case 0x3e:	/* DS override */
4151 			set_seg_override(ctxt, (ctxt->b >> 3) & 3);
4152 			break;
4153 		case 0x64:	/* FS override */
4154 		case 0x65:	/* GS override */
4155 			set_seg_override(ctxt, ctxt->b & 7);
4156 			break;
4157 		case 0x40 ... 0x4f: /* REX */
4158 			if (mode != X86EMUL_MODE_PROT64)
4159 				goto done_prefixes;
4160 			ctxt->rex_prefix = ctxt->b;
4161 			continue;
4162 		case 0xf0:	/* LOCK */
4163 			ctxt->lock_prefix = 1;
4164 			break;
4165 		case 0xf2:	/* REPNE/REPNZ */
4166 		case 0xf3:	/* REP/REPE/REPZ */
4167 			ctxt->rep_prefix = ctxt->b;
4168 			break;
4169 		default:
4170 			goto done_prefixes;
4171 		}
4172 
4173 		/* Any legacy prefix after a REX prefix nullifies its effect. */
4174 
4175 		ctxt->rex_prefix = 0;
4176 	}
4177 
4178 done_prefixes:
4179 
4180 	/* REX prefix. */
4181 	if (ctxt->rex_prefix & 8)
4182 		ctxt->op_bytes = 8;	/* REX.W */
4183 
4184 	/* Opcode byte(s). */
4185 	opcode = opcode_table[ctxt->b];
4186 	/* Two-byte opcode? */
4187 	if (ctxt->b == 0x0f) {
4188 		ctxt->twobyte = 1;
4189 		ctxt->b = insn_fetch(u8, ctxt);
4190 		opcode = twobyte_table[ctxt->b];
4191 	}
4192 	ctxt->d = opcode.flags;
4193 
4194 	if (ctxt->d & ModRM)
4195 		ctxt->modrm = insn_fetch(u8, ctxt);
4196 
4197 	while (ctxt->d & GroupMask) {
4198 		switch (ctxt->d & GroupMask) {
4199 		case Group:
4200 			goffset = (ctxt->modrm >> 3) & 7;
4201 			opcode = opcode.u.group[goffset];
4202 			break;
4203 		case GroupDual:
4204 			goffset = (ctxt->modrm >> 3) & 7;
4205 			if ((ctxt->modrm >> 6) == 3)
4206 				opcode = opcode.u.gdual->mod3[goffset];
4207 			else
4208 				opcode = opcode.u.gdual->mod012[goffset];
4209 			break;
4210 		case RMExt:
4211 			goffset = ctxt->modrm & 7;
4212 			opcode = opcode.u.group[goffset];
4213 			break;
4214 		case Prefix:
4215 			if (ctxt->rep_prefix && op_prefix)
4216 				return EMULATION_FAILED;
4217 			simd_prefix = op_prefix ? 0x66 : ctxt->rep_prefix;
4218 			switch (simd_prefix) {
4219 			case 0x00: opcode = opcode.u.gprefix->pfx_no; break;
4220 			case 0x66: opcode = opcode.u.gprefix->pfx_66; break;
4221 			case 0xf2: opcode = opcode.u.gprefix->pfx_f2; break;
4222 			case 0xf3: opcode = opcode.u.gprefix->pfx_f3; break;
4223 			}
4224 			break;
4225 		default:
4226 			return EMULATION_FAILED;
4227 		}
4228 
4229 		ctxt->d &= ~(u64)GroupMask;
4230 		ctxt->d |= opcode.flags;
4231 	}
4232 
4233 	ctxt->execute = opcode.u.execute;
4234 	ctxt->check_perm = opcode.check_perm;
4235 	ctxt->intercept = opcode.intercept;
4236 
4237 	/* Unrecognised? */
4238 	if (ctxt->d == 0 || (ctxt->d & Undefined))
4239 		return EMULATION_FAILED;
4240 
4241 	if (!(ctxt->d & VendorSpecific) && ctxt->only_vendor_specific_insn)
4242 		return EMULATION_FAILED;
4243 
4244 	if (mode == X86EMUL_MODE_PROT64 && (ctxt->d & Stack))
4245 		ctxt->op_bytes = 8;
4246 
4247 	if (ctxt->d & Op3264) {
4248 		if (mode == X86EMUL_MODE_PROT64)
4249 			ctxt->op_bytes = 8;
4250 		else
4251 			ctxt->op_bytes = 4;
4252 	}
4253 
4254 	if (ctxt->d & Sse)
4255 		ctxt->op_bytes = 16;
4256 	else if (ctxt->d & Mmx)
4257 		ctxt->op_bytes = 8;
4258 
4259 	/* ModRM and SIB bytes. */
4260 	if (ctxt->d & ModRM) {
4261 		rc = decode_modrm(ctxt, &ctxt->memop);
4262 		if (!ctxt->has_seg_override)
4263 			set_seg_override(ctxt, ctxt->modrm_seg);
4264 	} else if (ctxt->d & MemAbs)
4265 		rc = decode_abs(ctxt, &ctxt->memop);
4266 	if (rc != X86EMUL_CONTINUE)
4267 		goto done;
4268 
4269 	if (!ctxt->has_seg_override)
4270 		set_seg_override(ctxt, VCPU_SREG_DS);
4271 
4272 	ctxt->memop.addr.mem.seg = seg_override(ctxt);
4273 
4274 	if (ctxt->memop.type == OP_MEM && ctxt->ad_bytes != 8)
4275 		ctxt->memop.addr.mem.ea = (u32)ctxt->memop.addr.mem.ea;
4276 
4277 	/*
4278 	 * Decode and fetch the source operand: register, memory
4279 	 * or immediate.
4280 	 */
4281 	rc = decode_operand(ctxt, &ctxt->src, (ctxt->d >> SrcShift) & OpMask);
4282 	if (rc != X86EMUL_CONTINUE)
4283 		goto done;
4284 
4285 	/*
4286 	 * Decode and fetch the second source operand: register, memory
4287 	 * or immediate.
4288 	 */
4289 	rc = decode_operand(ctxt, &ctxt->src2, (ctxt->d >> Src2Shift) & OpMask);
4290 	if (rc != X86EMUL_CONTINUE)
4291 		goto done;
4292 
4293 	/* Decode and fetch the destination operand: register or memory. */
4294 	rc = decode_operand(ctxt, &ctxt->dst, (ctxt->d >> DstShift) & OpMask);
4295 
4296 done:
4297 	if (ctxt->memopp && ctxt->memopp->type == OP_MEM && ctxt->rip_relative)
4298 		ctxt->memopp->addr.mem.ea += ctxt->_eip;
4299 
4300 	return (rc != X86EMUL_CONTINUE) ? EMULATION_FAILED : EMULATION_OK;
4301 }
4302 
4303 bool x86_page_table_writing_insn(struct x86_emulate_ctxt *ctxt)
4304 {
4305 	return ctxt->d & PageTable;
4306 }
4307 
4308 static bool string_insn_completed(struct x86_emulate_ctxt *ctxt)
4309 {
4310 	/* The second termination condition only applies for REPE
4311 	 * and REPNE. Test if the repeat string operation prefix is
4312 	 * REPE/REPZ or REPNE/REPNZ and if it's the case it tests the
4313 	 * corresponding termination condition according to:
4314 	 * 	- if REPE/REPZ and ZF = 0 then done
4315 	 * 	- if REPNE/REPNZ and ZF = 1 then done
4316 	 */
4317 	if (((ctxt->b == 0xa6) || (ctxt->b == 0xa7) ||
4318 	     (ctxt->b == 0xae) || (ctxt->b == 0xaf))
4319 	    && (((ctxt->rep_prefix == REPE_PREFIX) &&
4320 		 ((ctxt->eflags & EFLG_ZF) == 0))
4321 		|| ((ctxt->rep_prefix == REPNE_PREFIX) &&
4322 		    ((ctxt->eflags & EFLG_ZF) == EFLG_ZF))))
4323 		return true;
4324 
4325 	return false;
4326 }
4327 
4328 static int flush_pending_x87_faults(struct x86_emulate_ctxt *ctxt)
4329 {
4330 	bool fault = false;
4331 
4332 	ctxt->ops->get_fpu(ctxt);
4333 	asm volatile("1: fwait \n\t"
4334 		     "2: \n\t"
4335 		     ".pushsection .fixup,\"ax\" \n\t"
4336 		     "3: \n\t"
4337 		     "movb $1, %[fault] \n\t"
4338 		     "jmp 2b \n\t"
4339 		     ".popsection \n\t"
4340 		     _ASM_EXTABLE(1b, 3b)
4341 		     : [fault]"+qm"(fault));
4342 	ctxt->ops->put_fpu(ctxt);
4343 
4344 	if (unlikely(fault))
4345 		return emulate_exception(ctxt, MF_VECTOR, 0, false);
4346 
4347 	return X86EMUL_CONTINUE;
4348 }
4349 
4350 static void fetch_possible_mmx_operand(struct x86_emulate_ctxt *ctxt,
4351 				       struct operand *op)
4352 {
4353 	if (op->type == OP_MM)
4354 		read_mmx_reg(ctxt, &op->mm_val, op->addr.mm);
4355 }
4356 
4357 
4358 int x86_emulate_insn(struct x86_emulate_ctxt *ctxt)
4359 {
4360 	const struct x86_emulate_ops *ops = ctxt->ops;
4361 	int rc = X86EMUL_CONTINUE;
4362 	int saved_dst_type = ctxt->dst.type;
4363 
4364 	ctxt->mem_read.pos = 0;
4365 
4366 	if (ctxt->mode == X86EMUL_MODE_PROT64 && (ctxt->d & No64)) {
4367 		rc = emulate_ud(ctxt);
4368 		goto done;
4369 	}
4370 
4371 	/* LOCK prefix is allowed only with some instructions */
4372 	if (ctxt->lock_prefix && (!(ctxt->d & Lock) || ctxt->dst.type != OP_MEM)) {
4373 		rc = emulate_ud(ctxt);
4374 		goto done;
4375 	}
4376 
4377 	if ((ctxt->d & SrcMask) == SrcMemFAddr && ctxt->src.type != OP_MEM) {
4378 		rc = emulate_ud(ctxt);
4379 		goto done;
4380 	}
4381 
4382 	if (((ctxt->d & (Sse|Mmx)) && ((ops->get_cr(ctxt, 0) & X86_CR0_EM)))
4383 	    || ((ctxt->d & Sse) && !(ops->get_cr(ctxt, 4) & X86_CR4_OSFXSR))) {
4384 		rc = emulate_ud(ctxt);
4385 		goto done;
4386 	}
4387 
4388 	if ((ctxt->d & (Sse|Mmx)) && (ops->get_cr(ctxt, 0) & X86_CR0_TS)) {
4389 		rc = emulate_nm(ctxt);
4390 		goto done;
4391 	}
4392 
4393 	if (ctxt->d & Mmx) {
4394 		rc = flush_pending_x87_faults(ctxt);
4395 		if (rc != X86EMUL_CONTINUE)
4396 			goto done;
4397 		/*
4398 		 * Now that we know the fpu is exception safe, we can fetch
4399 		 * operands from it.
4400 		 */
4401 		fetch_possible_mmx_operand(ctxt, &ctxt->src);
4402 		fetch_possible_mmx_operand(ctxt, &ctxt->src2);
4403 		if (!(ctxt->d & Mov))
4404 			fetch_possible_mmx_operand(ctxt, &ctxt->dst);
4405 	}
4406 
4407 	if (unlikely(ctxt->guest_mode) && ctxt->intercept) {
4408 		rc = emulator_check_intercept(ctxt, ctxt->intercept,
4409 					      X86_ICPT_PRE_EXCEPT);
4410 		if (rc != X86EMUL_CONTINUE)
4411 			goto done;
4412 	}
4413 
4414 	/* Privileged instruction can be executed only in CPL=0 */
4415 	if ((ctxt->d & Priv) && ops->cpl(ctxt)) {
4416 		rc = emulate_gp(ctxt, 0);
4417 		goto done;
4418 	}
4419 
4420 	/* Instruction can only be executed in protected mode */
4421 	if ((ctxt->d & Prot) && ctxt->mode < X86EMUL_MODE_PROT16) {
4422 		rc = emulate_ud(ctxt);
4423 		goto done;
4424 	}
4425 
4426 	/* Do instruction specific permission checks */
4427 	if (ctxt->check_perm) {
4428 		rc = ctxt->check_perm(ctxt);
4429 		if (rc != X86EMUL_CONTINUE)
4430 			goto done;
4431 	}
4432 
4433 	if (unlikely(ctxt->guest_mode) && ctxt->intercept) {
4434 		rc = emulator_check_intercept(ctxt, ctxt->intercept,
4435 					      X86_ICPT_POST_EXCEPT);
4436 		if (rc != X86EMUL_CONTINUE)
4437 			goto done;
4438 	}
4439 
4440 	if (ctxt->rep_prefix && (ctxt->d & String)) {
4441 		/* All REP prefixes have the same first termination condition */
4442 		if (address_mask(ctxt, reg_read(ctxt, VCPU_REGS_RCX)) == 0) {
4443 			ctxt->eip = ctxt->_eip;
4444 			goto done;
4445 		}
4446 	}
4447 
4448 	if ((ctxt->src.type == OP_MEM) && !(ctxt->d & NoAccess)) {
4449 		rc = segmented_read(ctxt, ctxt->src.addr.mem,
4450 				    ctxt->src.valptr, ctxt->src.bytes);
4451 		if (rc != X86EMUL_CONTINUE)
4452 			goto done;
4453 		ctxt->src.orig_val64 = ctxt->src.val64;
4454 	}
4455 
4456 	if (ctxt->src2.type == OP_MEM) {
4457 		rc = segmented_read(ctxt, ctxt->src2.addr.mem,
4458 				    &ctxt->src2.val, ctxt->src2.bytes);
4459 		if (rc != X86EMUL_CONTINUE)
4460 			goto done;
4461 	}
4462 
4463 	if ((ctxt->d & DstMask) == ImplicitOps)
4464 		goto special_insn;
4465 
4466 
4467 	if ((ctxt->dst.type == OP_MEM) && !(ctxt->d & Mov)) {
4468 		/* optimisation - avoid slow emulated read if Mov */
4469 		rc = segmented_read(ctxt, ctxt->dst.addr.mem,
4470 				   &ctxt->dst.val, ctxt->dst.bytes);
4471 		if (rc != X86EMUL_CONTINUE)
4472 			goto done;
4473 	}
4474 	ctxt->dst.orig_val = ctxt->dst.val;
4475 
4476 special_insn:
4477 
4478 	if (unlikely(ctxt->guest_mode) && ctxt->intercept) {
4479 		rc = emulator_check_intercept(ctxt, ctxt->intercept,
4480 					      X86_ICPT_POST_MEMACCESS);
4481 		if (rc != X86EMUL_CONTINUE)
4482 			goto done;
4483 	}
4484 
4485 	if (ctxt->execute) {
4486 		rc = ctxt->execute(ctxt);
4487 		if (rc != X86EMUL_CONTINUE)
4488 			goto done;
4489 		goto writeback;
4490 	}
4491 
4492 	if (ctxt->twobyte)
4493 		goto twobyte_insn;
4494 
4495 	switch (ctxt->b) {
4496 	case 0x40 ... 0x47: /* inc r16/r32 */
4497 		emulate_1op(ctxt, "inc");
4498 		break;
4499 	case 0x48 ... 0x4f: /* dec r16/r32 */
4500 		emulate_1op(ctxt, "dec");
4501 		break;
4502 	case 0x63:		/* movsxd */
4503 		if (ctxt->mode != X86EMUL_MODE_PROT64)
4504 			goto cannot_emulate;
4505 		ctxt->dst.val = (s32) ctxt->src.val;
4506 		break;
4507 	case 0x70 ... 0x7f: /* jcc (short) */
4508 		if (test_cc(ctxt->b, ctxt->eflags))
4509 			jmp_rel(ctxt, ctxt->src.val);
4510 		break;
4511 	case 0x8d: /* lea r16/r32, m */
4512 		ctxt->dst.val = ctxt->src.addr.mem.ea;
4513 		break;
4514 	case 0x90 ... 0x97: /* nop / xchg reg, rax */
4515 		if (ctxt->dst.addr.reg == reg_rmw(ctxt, VCPU_REGS_RAX))
4516 			break;
4517 		rc = em_xchg(ctxt);
4518 		break;
4519 	case 0x98: /* cbw/cwde/cdqe */
4520 		switch (ctxt->op_bytes) {
4521 		case 2: ctxt->dst.val = (s8)ctxt->dst.val; break;
4522 		case 4: ctxt->dst.val = (s16)ctxt->dst.val; break;
4523 		case 8: ctxt->dst.val = (s32)ctxt->dst.val; break;
4524 		}
4525 		break;
4526 	case 0xc0 ... 0xc1:
4527 		rc = em_grp2(ctxt);
4528 		break;
4529 	case 0xcc:		/* int3 */
4530 		rc = emulate_int(ctxt, 3);
4531 		break;
4532 	case 0xcd:		/* int n */
4533 		rc = emulate_int(ctxt, ctxt->src.val);
4534 		break;
4535 	case 0xce:		/* into */
4536 		if (ctxt->eflags & EFLG_OF)
4537 			rc = emulate_int(ctxt, 4);
4538 		break;
4539 	case 0xd0 ... 0xd1:	/* Grp2 */
4540 		rc = em_grp2(ctxt);
4541 		break;
4542 	case 0xd2 ... 0xd3:	/* Grp2 */
4543 		ctxt->src.val = reg_read(ctxt, VCPU_REGS_RCX);
4544 		rc = em_grp2(ctxt);
4545 		break;
4546 	case 0xe9: /* jmp rel */
4547 	case 0xeb: /* jmp rel short */
4548 		jmp_rel(ctxt, ctxt->src.val);
4549 		ctxt->dst.type = OP_NONE; /* Disable writeback. */
4550 		break;
4551 	case 0xf4:              /* hlt */
4552 		ctxt->ops->halt(ctxt);
4553 		break;
4554 	case 0xf5:	/* cmc */
4555 		/* complement carry flag from eflags reg */
4556 		ctxt->eflags ^= EFLG_CF;
4557 		break;
4558 	case 0xf8: /* clc */
4559 		ctxt->eflags &= ~EFLG_CF;
4560 		break;
4561 	case 0xf9: /* stc */
4562 		ctxt->eflags |= EFLG_CF;
4563 		break;
4564 	case 0xfc: /* cld */
4565 		ctxt->eflags &= ~EFLG_DF;
4566 		break;
4567 	case 0xfd: /* std */
4568 		ctxt->eflags |= EFLG_DF;
4569 		break;
4570 	default:
4571 		goto cannot_emulate;
4572 	}
4573 
4574 	if (rc != X86EMUL_CONTINUE)
4575 		goto done;
4576 
4577 writeback:
4578 	rc = writeback(ctxt);
4579 	if (rc != X86EMUL_CONTINUE)
4580 		goto done;
4581 
4582 	/*
4583 	 * restore dst type in case the decoding will be reused
4584 	 * (happens for string instruction )
4585 	 */
4586 	ctxt->dst.type = saved_dst_type;
4587 
4588 	if ((ctxt->d & SrcMask) == SrcSI)
4589 		string_addr_inc(ctxt, VCPU_REGS_RSI, &ctxt->src);
4590 
4591 	if ((ctxt->d & DstMask) == DstDI)
4592 		string_addr_inc(ctxt, VCPU_REGS_RDI, &ctxt->dst);
4593 
4594 	if (ctxt->rep_prefix && (ctxt->d & String)) {
4595 		unsigned int count;
4596 		struct read_cache *r = &ctxt->io_read;
4597 		if ((ctxt->d & SrcMask) == SrcSI)
4598 			count = ctxt->src.count;
4599 		else
4600 			count = ctxt->dst.count;
4601 		register_address_increment(ctxt, reg_rmw(ctxt, VCPU_REGS_RCX),
4602 				-count);
4603 
4604 		if (!string_insn_completed(ctxt)) {
4605 			/*
4606 			 * Re-enter guest when pio read ahead buffer is empty
4607 			 * or, if it is not used, after each 1024 iteration.
4608 			 */
4609 			if ((r->end != 0 || reg_read(ctxt, VCPU_REGS_RCX) & 0x3ff) &&
4610 			    (r->end == 0 || r->end != r->pos)) {
4611 				/*
4612 				 * Reset read cache. Usually happens before
4613 				 * decode, but since instruction is restarted
4614 				 * we have to do it here.
4615 				 */
4616 				ctxt->mem_read.end = 0;
4617 				writeback_registers(ctxt);
4618 				return EMULATION_RESTART;
4619 			}
4620 			goto done; /* skip rip writeback */
4621 		}
4622 	}
4623 
4624 	ctxt->eip = ctxt->_eip;
4625 
4626 done:
4627 	if (rc == X86EMUL_PROPAGATE_FAULT)
4628 		ctxt->have_exception = true;
4629 	if (rc == X86EMUL_INTERCEPTED)
4630 		return EMULATION_INTERCEPTED;
4631 
4632 	if (rc == X86EMUL_CONTINUE)
4633 		writeback_registers(ctxt);
4634 
4635 	return (rc == X86EMUL_UNHANDLEABLE) ? EMULATION_FAILED : EMULATION_OK;
4636 
4637 twobyte_insn:
4638 	switch (ctxt->b) {
4639 	case 0x09:		/* wbinvd */
4640 		(ctxt->ops->wbinvd)(ctxt);
4641 		break;
4642 	case 0x08:		/* invd */
4643 	case 0x0d:		/* GrpP (prefetch) */
4644 	case 0x18:		/* Grp16 (prefetch/nop) */
4645 		break;
4646 	case 0x20: /* mov cr, reg */
4647 		ctxt->dst.val = ops->get_cr(ctxt, ctxt->modrm_reg);
4648 		break;
4649 	case 0x21: /* mov from dr to reg */
4650 		ops->get_dr(ctxt, ctxt->modrm_reg, &ctxt->dst.val);
4651 		break;
4652 	case 0x40 ... 0x4f:	/* cmov */
4653 		ctxt->dst.val = ctxt->dst.orig_val = ctxt->src.val;
4654 		if (!test_cc(ctxt->b, ctxt->eflags))
4655 			ctxt->dst.type = OP_NONE; /* no writeback */
4656 		break;
4657 	case 0x80 ... 0x8f: /* jnz rel, etc*/
4658 		if (test_cc(ctxt->b, ctxt->eflags))
4659 			jmp_rel(ctxt, ctxt->src.val);
4660 		break;
4661 	case 0x90 ... 0x9f:     /* setcc r/m8 */
4662 		ctxt->dst.val = test_cc(ctxt->b, ctxt->eflags);
4663 		break;
4664 	case 0xa4: /* shld imm8, r, r/m */
4665 	case 0xa5: /* shld cl, r, r/m */
4666 		emulate_2op_cl(ctxt, "shld");
4667 		break;
4668 	case 0xac: /* shrd imm8, r, r/m */
4669 	case 0xad: /* shrd cl, r, r/m */
4670 		emulate_2op_cl(ctxt, "shrd");
4671 		break;
4672 	case 0xae:              /* clflush */
4673 		break;
4674 	case 0xb6 ... 0xb7:	/* movzx */
4675 		ctxt->dst.bytes = ctxt->op_bytes;
4676 		ctxt->dst.val = (ctxt->src.bytes == 1) ? (u8) ctxt->src.val
4677 						       : (u16) ctxt->src.val;
4678 		break;
4679 	case 0xbe ... 0xbf:	/* movsx */
4680 		ctxt->dst.bytes = ctxt->op_bytes;
4681 		ctxt->dst.val = (ctxt->src.bytes == 1) ? (s8) ctxt->src.val :
4682 							(s16) ctxt->src.val;
4683 		break;
4684 	case 0xc0 ... 0xc1:	/* xadd */
4685 		emulate_2op_SrcV(ctxt, "add");
4686 		/* Write back the register source. */
4687 		ctxt->src.val = ctxt->dst.orig_val;
4688 		write_register_operand(&ctxt->src);
4689 		break;
4690 	case 0xc3:		/* movnti */
4691 		ctxt->dst.bytes = ctxt->op_bytes;
4692 		ctxt->dst.val = (ctxt->op_bytes == 4) ? (u32) ctxt->src.val :
4693 							(u64) ctxt->src.val;
4694 		break;
4695 	default:
4696 		goto cannot_emulate;
4697 	}
4698 
4699 	if (rc != X86EMUL_CONTINUE)
4700 		goto done;
4701 
4702 	goto writeback;
4703 
4704 cannot_emulate:
4705 	return EMULATION_FAILED;
4706 }
4707 
4708 void emulator_invalidate_register_cache(struct x86_emulate_ctxt *ctxt)
4709 {
4710 	invalidate_registers(ctxt);
4711 }
4712 
4713 void emulator_writeback_register_cache(struct x86_emulate_ctxt *ctxt)
4714 {
4715 	writeback_registers(ctxt);
4716 }
4717