KVM: x86 emulator: drop vcpu argument from pio callbacks
[deliverable/linux.git] / arch / x86 / kvm / emulate.c
index 0ad47b819a8b2eb1743810e9b9dd19c9d5b59a3d..8af08a16f4dd82342fd4732d9ffdf36a910991bf 100644 (file)
 #define Stack       (1<<13)     /* Stack instruction (push/pop) */
 #define Group       (1<<14)     /* Bits 3:5 of modrm byte extend opcode */
 #define GroupDual   (1<<15)     /* Alternate decoding of mod == 3 */
+#define Prefix      (1<<16)     /* Instruction varies with 66/f2/f3 prefix */
+#define Sse         (1<<17)     /* SSE Vector instruction */
+#define RMExt       (1<<18)     /* Opcode extension in ModRM r/m if mod == 3 */
 /* Misc flags */
+#define Prot        (1<<21) /* instruction generates #UD if not in prot-mode */
 #define VendorSpecific (1<<22) /* Vendor specific instruction */
 #define NoAccess    (1<<23) /* Don't access memory (lea/invlpg/verr etc) */
 #define Op3264      (1<<24) /* Operand is 64b in long mode, 32b otherwise */
 
 struct opcode {
        u32 flags;
+       u8 intercept;
        union {
                int (*execute)(struct x86_emulate_ctxt *ctxt);
                struct opcode *group;
                struct group_dual *gdual;
+               struct gprefix *gprefix;
        } u;
+       int (*check_perm)(struct x86_emulate_ctxt *ctxt);
 };
 
 struct group_dual {
@@ -114,6 +121,13 @@ struct group_dual {
        struct opcode mod3[8];
 };
 
+struct gprefix {
+       struct opcode pfx_no;
+       struct opcode pfx_66;
+       struct opcode pfx_f2;
+       struct opcode pfx_f3;
+};
+
 /* EFLAGS bit definitions. */
 #define EFLG_ID (1<<21)
 #define EFLG_VIP (1<<20)
@@ -248,42 +262,42 @@ struct group_dual {
                             "w", "r", _LO32, "r", "", "r")
 
 /* Instruction has three operands and one operand is stored in ECX register */
-#define __emulate_2op_cl(_op, _cl, _src, _dst, _eflags, _suffix, _type)        \
-       do {                                                                    \
-               unsigned long _tmp;                                             \
-               _type _clv  = (_cl).val;                                        \
-               _type _srcv = (_src).val;                                       \
-               _type _dstv = (_dst).val;                                       \
-                                                                               \
-               __asm__ __volatile__ (                                          \
-                       _PRE_EFLAGS("0", "5", "2")                              \
-                       _op _suffix " %4,%1 \n"                                 \
-                       _POST_EFLAGS("0", "5", "2")                             \
-                       : "=m" (_eflags), "+r" (_dstv), "=&r" (_tmp)            \
-                       : "c" (_clv) , "r" (_srcv), "i" (EFLAGS_MASK)           \
-                       );                                                      \
-                                                                               \
-               (_cl).val  = (unsigned long) _clv;                              \
-               (_src).val = (unsigned long) _srcv;                             \
-               (_dst).val = (unsigned long) _dstv;                             \
+#define __emulate_2op_cl(_op, _cl, _src, _dst, _eflags, _suffix, _type)        \
+       do {                                                            \
+               unsigned long _tmp;                                     \
+               _type _clv  = (_cl).val;                                \
+               _type _srcv = (_src).val;                               \
+               _type _dstv = (_dst).val;                               \
+                                                                       \
+               __asm__ __volatile__ (                                  \
+                       _PRE_EFLAGS("0", "5", "2")                      \
+                       _op _suffix " %4,%1 \n"                         \
+                       _POST_EFLAGS("0", "5", "2")                     \
+                       : "=m" (_eflags), "+r" (_dstv), "=&r" (_tmp)    \
+                       : "c" (_clv) , "r" (_srcv), "i" (EFLAGS_MASK)   \
+                       );                                              \
+                                                                       \
+               (_cl).val  = (unsigned long) _clv;                      \
+               (_src).val = (unsigned long) _srcv;                     \
+               (_dst).val = (unsigned long) _dstv;                     \
        } while (0)
 
-#define emulate_2op_cl(_op, _cl, _src, _dst, _eflags)                          \
-       do {                                                                    \
-               switch ((_dst).bytes) {                                         \
-               case 2:                                                         \
-                       __emulate_2op_cl(_op, _cl, _src, _dst, _eflags,         \
-                                               "w", unsigned short);           \
-                       break;                                                  \
-               case 4:                                                         \
-                       __emulate_2op_cl(_op, _cl, _src, _dst, _eflags,         \
-                                               "l", unsigned int);             \
-                       break;                                                  \
-               case 8:                                                         \
-                       ON64(__emulate_2op_cl(_op, _cl, _src, _dst, _eflags,    \
-                                               "q", unsigned long));           \
-                       break;                                                  \
-               }                                                               \
+#define emulate_2op_cl(_op, _cl, _src, _dst, _eflags)                  \
+       do {                                                            \
+               switch ((_dst).bytes) {                                 \
+               case 2:                                                 \
+                       __emulate_2op_cl(_op, _cl, _src, _dst, _eflags, \
+                                        "w", unsigned short);          \
+                       break;                                          \
+               case 4:                                                 \
+                       __emulate_2op_cl(_op, _cl, _src, _dst, _eflags, \
+                                        "l", unsigned int);            \
+                       break;                                          \
+               case 8:                                                 \
+                       ON64(__emulate_2op_cl(_op, _cl, _src, _dst, _eflags, \
+                                             "q", unsigned long));     \
+                       break;                                          \
+               }                                                       \
        } while (0)
 
 #define __emulate_1op(_op, _dst, _eflags, _suffix)                     \
@@ -346,13 +360,25 @@ struct group_dual {
        } while (0)
 
 /* instruction has only one source operand, destination is implicit (e.g. mul, div, imul, idiv) */
-#define emulate_1op_rax_rdx(_op, _src, _rax, _rdx, _eflags)                    \
-       do {                                                                    \
-               switch((_src).bytes) {                                          \
-               case 1: __emulate_1op_rax_rdx(_op, _src, _rax, _rdx, _eflags, "b"); break; \
-               case 2: __emulate_1op_rax_rdx(_op, _src, _rax, _rdx,  _eflags, "w"); break; \
-               case 4: __emulate_1op_rax_rdx(_op, _src, _rax, _rdx, _eflags, "l"); break; \
-               case 8: ON64(__emulate_1op_rax_rdx(_op, _src, _rax, _rdx, _eflags, "q")); break; \
+#define emulate_1op_rax_rdx(_op, _src, _rax, _rdx, _eflags)            \
+       do {                                                            \
+               switch((_src).bytes) {                                  \
+               case 1:                                                 \
+                       __emulate_1op_rax_rdx(_op, _src, _rax, _rdx,    \
+                                             _eflags, "b");            \
+                       break;                                          \
+               case 2:                                                 \
+                       __emulate_1op_rax_rdx(_op, _src, _rax, _rdx,    \
+                                             _eflags, "w");            \
+                       break;                                          \
+               case 4:                                                 \
+                       __emulate_1op_rax_rdx(_op, _src, _rax, _rdx,    \
+                                             _eflags, "l");            \
+                       break;                                          \
+               case 8:                                                 \
+                       ON64(__emulate_1op_rax_rdx(_op, _src, _rax, _rdx, \
+                                                  _eflags, "q"));      \
+                       break;                                          \
                }                                                       \
        } while (0)
 
@@ -388,13 +414,33 @@ struct group_dual {
        (_type)_x;                                                      \
 })
 
-#define insn_fetch_arr(_arr, _size, _eip)                                \
+#define insn_fetch_arr(_arr, _size, _eip)                              \
 ({     rc = do_insn_fetch(ctxt, ops, (_eip), _arr, (_size));           \
        if (rc != X86EMUL_CONTINUE)                                     \
                goto done;                                              \
        (_eip) += (_size);                                              \
 })
 
+static int emulator_check_intercept(struct x86_emulate_ctxt *ctxt,
+                                   enum x86_intercept intercept,
+                                   enum x86_intercept_stage stage)
+{
+       struct x86_instruction_info info = {
+               .intercept  = intercept,
+               .rep_prefix = ctxt->decode.rep_prefix,
+               .modrm_mod  = ctxt->decode.modrm_mod,
+               .modrm_reg  = ctxt->decode.modrm_reg,
+               .modrm_rm   = ctxt->decode.modrm_rm,
+               .src_val    = ctxt->decode.src.val64,
+               .src_bytes  = ctxt->decode.src.bytes,
+               .dst_bytes  = ctxt->decode.dst.bytes,
+               .ad_bytes   = ctxt->decode.ad_bytes,
+               .next_rip   = ctxt->eip,
+       };
+
+       return ctxt->ops->intercept(ctxt->vcpu, &info, stage);
+}
+
 static inline unsigned long ad_mask(struct decode_cache *c)
 {
        return (1UL << (c->ad_bytes << 3)) - 1;
@@ -430,6 +476,13 @@ static inline void jmp_rel(struct decode_cache *c, int rel)
        register_address_increment(c, &c->eip, rel);
 }
 
+static u32 desc_limit_scaled(struct desc_struct *desc)
+{
+       u32 limit = get_desc_limit(desc);
+
+       return desc->g ? (limit << 12) | 0xfff : limit;
+}
+
 static void set_seg_override(struct decode_cache *c, int seg)
 {
        c->has_seg_override = true;
@@ -455,18 +508,6 @@ static unsigned seg_override(struct x86_emulate_ctxt *ctxt,
        return c->seg_override;
 }
 
-static ulong linear(struct x86_emulate_ctxt *ctxt,
-                   struct segmented_address addr)
-{
-       struct decode_cache *c = &ctxt->decode;
-       ulong la;
-
-       la = seg_base(ctxt, ctxt->ops, addr.seg) + addr.ea;
-       if (c->ad_bytes != 8)
-               la &= (u32)-1;
-       return la;
-}
-
 static int emulate_exception(struct x86_emulate_ctxt *ctxt, int vec,
                             u32 error, bool valid)
 {
@@ -476,11 +517,21 @@ static int emulate_exception(struct x86_emulate_ctxt *ctxt, int vec,
        return X86EMUL_PROPAGATE_FAULT;
 }
 
+static int emulate_db(struct x86_emulate_ctxt *ctxt)
+{
+       return emulate_exception(ctxt, DB_VECTOR, 0, false);
+}
+
 static int emulate_gp(struct x86_emulate_ctxt *ctxt, int err)
 {
        return emulate_exception(ctxt, GP_VECTOR, err, true);
 }
 
+static int emulate_ss(struct x86_emulate_ctxt *ctxt, int err)
+{
+       return emulate_exception(ctxt, SS_VECTOR, err, true);
+}
+
 static int emulate_ud(struct x86_emulate_ctxt *ctxt)
 {
        return emulate_exception(ctxt, UD_VECTOR, 0, false);
@@ -496,6 +547,107 @@ static int emulate_de(struct x86_emulate_ctxt *ctxt)
        return emulate_exception(ctxt, DE_VECTOR, 0, false);
 }
 
+static int emulate_nm(struct x86_emulate_ctxt *ctxt)
+{
+       return emulate_exception(ctxt, NM_VECTOR, 0, false);
+}
+
+static int __linearize(struct x86_emulate_ctxt *ctxt,
+                    struct segmented_address addr,
+                    unsigned size, bool write, bool fetch,
+                    ulong *linear)
+{
+       struct decode_cache *c = &ctxt->decode;
+       struct desc_struct desc;
+       bool usable;
+       ulong la;
+       u32 lim;
+       unsigned cpl, rpl;
+
+       la = seg_base(ctxt, ctxt->ops, addr.seg) + addr.ea;
+       switch (ctxt->mode) {
+       case X86EMUL_MODE_REAL:
+               break;
+       case X86EMUL_MODE_PROT64:
+               if (((signed long)la << 16) >> 16 != la)
+                       return emulate_gp(ctxt, 0);
+               break;
+       default:
+               usable = ctxt->ops->get_cached_descriptor(&desc, NULL, addr.seg,
+                                                         ctxt->vcpu);
+               if (!usable)
+                       goto bad;
+               /* code segment or read-only data segment */
+               if (((desc.type & 8) || !(desc.type & 2)) && write)
+                       goto bad;
+               /* unreadable code segment */
+               if (!fetch && (desc.type & 8) && !(desc.type & 2))
+                       goto bad;
+               lim = desc_limit_scaled(&desc);
+               if ((desc.type & 8) || !(desc.type & 4)) {
+                       /* expand-up segment */
+                       if (addr.ea > lim || (u32)(addr.ea + size - 1) > lim)
+                               goto bad;
+               } else {
+                       /* exapand-down segment */
+                       if (addr.ea <= lim || (u32)(addr.ea + size - 1) <= lim)
+                               goto bad;
+                       lim = desc.d ? 0xffffffff : 0xffff;
+                       if (addr.ea > lim || (u32)(addr.ea + size - 1) > lim)
+                               goto bad;
+               }
+               cpl = ctxt->ops->cpl(ctxt->vcpu);
+               rpl = ctxt->ops->get_segment_selector(addr.seg, ctxt->vcpu) & 3;
+               cpl = max(cpl, rpl);
+               if (!(desc.type & 8)) {
+                       /* data segment */
+                       if (cpl > desc.dpl)
+                               goto bad;
+               } else if ((desc.type & 8) && !(desc.type & 4)) {
+                       /* nonconforming code segment */
+                       if (cpl != desc.dpl)
+                               goto bad;
+               } else if ((desc.type & 8) && (desc.type & 4)) {
+                       /* conforming code segment */
+                       if (cpl < desc.dpl)
+                               goto bad;
+               }
+               break;
+       }
+       if (fetch ? ctxt->mode != X86EMUL_MODE_PROT64 : c->ad_bytes != 8)
+               la &= (u32)-1;
+       *linear = la;
+       return X86EMUL_CONTINUE;
+bad:
+       if (addr.seg == VCPU_SREG_SS)
+               return emulate_ss(ctxt, addr.seg);
+       else
+               return emulate_gp(ctxt, addr.seg);
+}
+
+static int linearize(struct x86_emulate_ctxt *ctxt,
+                    struct segmented_address addr,
+                    unsigned size, bool write,
+                    ulong *linear)
+{
+       return __linearize(ctxt, addr, size, write, false, linear);
+}
+
+
+static int segmented_read_std(struct x86_emulate_ctxt *ctxt,
+                             struct segmented_address addr,
+                             void *data,
+                             unsigned size)
+{
+       int rc;
+       ulong linear;
+
+       rc = linearize(ctxt, addr, size, false, &linear);
+       if (rc != X86EMUL_CONTINUE)
+               return rc;
+       return ctxt->ops->read_std(ctxt, linear, data, size, &ctxt->exception);
+}
+
 static int do_fetch_insn_byte(struct x86_emulate_ctxt *ctxt,
                              struct x86_emulate_ops *ops,
                              unsigned long eip, u8 *dest)
@@ -505,10 +657,15 @@ static int do_fetch_insn_byte(struct x86_emulate_ctxt *ctxt,
        int size, cur_size;
 
        if (eip == fc->end) {
+               unsigned long linear;
+               struct segmented_address addr = { .seg=VCPU_SREG_CS, .ea=eip};
                cur_size = fc->end - fc->start;
                size = min(15UL - cur_size, PAGE_SIZE - offset_in_page(eip));
-               rc = ops->fetch(ctxt->cs_base + eip, fc->data + cur_size,
-                               size, ctxt->vcpu, &ctxt->exception);
+               rc = __linearize(ctxt, addr, size, false, true, &linear);
+               if (rc != X86EMUL_CONTINUE)
+                       return rc;
+               rc = ops->fetch(ctxt, linear, fc->data + cur_size,
+                               size, &ctxt->exception);
                if (rc != X86EMUL_CONTINUE)
                        return rc;
                fc->end += size;
@@ -560,13 +717,11 @@ static int read_descriptor(struct x86_emulate_ctxt *ctxt,
        if (op_bytes == 2)
                op_bytes = 3;
        *address = 0;
-       rc = ops->read_std(linear(ctxt, addr), (unsigned long *)size, 2,
-                          ctxt->vcpu, &ctxt->exception);
+       rc = segmented_read_std(ctxt, addr, size, 2);
        if (rc != X86EMUL_CONTINUE)
                return rc;
        addr.ea += 2;
-       rc = ops->read_std(linear(ctxt, addr), address, op_bytes,
-                          ctxt->vcpu, &ctxt->exception);
+       rc = segmented_read_std(ctxt, addr, address, op_bytes);
        return rc;
 }
 
@@ -623,7 +778,63 @@ static void fetch_register_operand(struct operand *op)
        }
 }
 
-static void decode_register_operand(struct operand *op,
+static void read_sse_reg(struct x86_emulate_ctxt *ctxt, sse128_t *data, int reg)
+{
+       ctxt->ops->get_fpu(ctxt);
+       switch (reg) {
+       case 0: asm("movdqu %%xmm0, %0" : "=m"(*data)); break;
+       case 1: asm("movdqu %%xmm1, %0" : "=m"(*data)); break;
+       case 2: asm("movdqu %%xmm2, %0" : "=m"(*data)); break;
+       case 3: asm("movdqu %%xmm3, %0" : "=m"(*data)); break;
+       case 4: asm("movdqu %%xmm4, %0" : "=m"(*data)); break;
+       case 5: asm("movdqu %%xmm5, %0" : "=m"(*data)); break;
+       case 6: asm("movdqu %%xmm6, %0" : "=m"(*data)); break;
+       case 7: asm("movdqu %%xmm7, %0" : "=m"(*data)); break;
+#ifdef CONFIG_X86_64
+       case 8: asm("movdqu %%xmm8, %0" : "=m"(*data)); break;
+       case 9: asm("movdqu %%xmm9, %0" : "=m"(*data)); break;
+       case 10: asm("movdqu %%xmm10, %0" : "=m"(*data)); break;
+       case 11: asm("movdqu %%xmm11, %0" : "=m"(*data)); break;
+       case 12: asm("movdqu %%xmm12, %0" : "=m"(*data)); break;
+       case 13: asm("movdqu %%xmm13, %0" : "=m"(*data)); break;
+       case 14: asm("movdqu %%xmm14, %0" : "=m"(*data)); break;
+       case 15: asm("movdqu %%xmm15, %0" : "=m"(*data)); break;
+#endif
+       default: BUG();
+       }
+       ctxt->ops->put_fpu(ctxt);
+}
+
+static void write_sse_reg(struct x86_emulate_ctxt *ctxt, sse128_t *data,
+                         int reg)
+{
+       ctxt->ops->get_fpu(ctxt);
+       switch (reg) {
+       case 0: asm("movdqu %0, %%xmm0" : : "m"(*data)); break;
+       case 1: asm("movdqu %0, %%xmm1" : : "m"(*data)); break;
+       case 2: asm("movdqu %0, %%xmm2" : : "m"(*data)); break;
+       case 3: asm("movdqu %0, %%xmm3" : : "m"(*data)); break;
+       case 4: asm("movdqu %0, %%xmm4" : : "m"(*data)); break;
+       case 5: asm("movdqu %0, %%xmm5" : : "m"(*data)); break;
+       case 6: asm("movdqu %0, %%xmm6" : : "m"(*data)); break;
+       case 7: asm("movdqu %0, %%xmm7" : : "m"(*data)); break;
+#ifdef CONFIG_X86_64
+       case 8: asm("movdqu %0, %%xmm8" : : "m"(*data)); break;
+       case 9: asm("movdqu %0, %%xmm9" : : "m"(*data)); break;
+       case 10: asm("movdqu %0, %%xmm10" : : "m"(*data)); break;
+       case 11: asm("movdqu %0, %%xmm11" : : "m"(*data)); break;
+       case 12: asm("movdqu %0, %%xmm12" : : "m"(*data)); break;
+       case 13: asm("movdqu %0, %%xmm13" : : "m"(*data)); break;
+       case 14: asm("movdqu %0, %%xmm14" : : "m"(*data)); break;
+       case 15: asm("movdqu %0, %%xmm15" : : "m"(*data)); break;
+#endif
+       default: BUG();
+       }
+       ctxt->ops->put_fpu(ctxt);
+}
+
+static void decode_register_operand(struct x86_emulate_ctxt *ctxt,
+                                   struct operand *op,
                                    struct decode_cache *c,
                                    int inhibit_bytereg)
 {
@@ -632,6 +843,15 @@ static void decode_register_operand(struct operand *op,
 
        if (!(c->d & ModRM))
                reg = (c->b & 7) | ((c->rex_prefix & 1) << 3);
+
+       if (c->d & Sse) {
+               op->type = OP_XMM;
+               op->bytes = 16;
+               op->addr.xmm = reg;
+               read_sse_reg(ctxt, &op->vec_val, reg);
+               return;
+       }
+
        op->type = OP_REG;
        if ((c->d & ByteOp) && !inhibit_bytereg) {
                op->addr.reg = decode_register(reg, c->regs, highbyte_regs);
@@ -671,6 +891,13 @@ static int decode_modrm(struct x86_emulate_ctxt *ctxt,
                op->bytes = (c->d & ByteOp) ? 1 : c->op_bytes;
                op->addr.reg = decode_register(c->modrm_rm,
                                               c->regs, c->d & ByteOp);
+               if (c->d & Sse) {
+                       op->type = OP_XMM;
+                       op->bytes = 16;
+                       op->addr.xmm = c->modrm_rm;
+                       read_sse_reg(ctxt, &op->vec_val, c->modrm_rm);
+                       return rc;
+               }
                fetch_register_operand(op);
                return rc;
        }
@@ -819,8 +1046,8 @@ static int read_emulated(struct x86_emulate_ctxt *ctxt,
                if (mc->pos < mc->end)
                        goto read_cached;
 
-               rc = ops->read_emulated(addr, mc->data + mc->end, n,
-                                       &ctxt->exception, ctxt->vcpu);
+               rc = ops->read_emulated(ctxt, addr, mc->data + mc->end, n,
+                                       &ctxt->exception);
                if (rc != X86EMUL_CONTINUE)
                        return rc;
                mc->end += n;
@@ -834,6 +1061,50 @@ static int read_emulated(struct x86_emulate_ctxt *ctxt,
        return X86EMUL_CONTINUE;
 }
 
+static int segmented_read(struct x86_emulate_ctxt *ctxt,
+                         struct segmented_address addr,
+                         void *data,
+                         unsigned size)
+{
+       int rc;
+       ulong linear;
+
+       rc = linearize(ctxt, addr, size, false, &linear);
+       if (rc != X86EMUL_CONTINUE)
+               return rc;
+       return read_emulated(ctxt, ctxt->ops, linear, data, size);
+}
+
+static int segmented_write(struct x86_emulate_ctxt *ctxt,
+                          struct segmented_address addr,
+                          const void *data,
+                          unsigned size)
+{
+       int rc;
+       ulong linear;
+
+       rc = linearize(ctxt, addr, size, true, &linear);
+       if (rc != X86EMUL_CONTINUE)
+               return rc;
+       return ctxt->ops->write_emulated(ctxt, linear, data, size,
+                                        &ctxt->exception);
+}
+
+static int segmented_cmpxchg(struct x86_emulate_ctxt *ctxt,
+                            struct segmented_address addr,
+                            const void *orig_data, const void *data,
+                            unsigned size)
+{
+       int rc;
+       ulong linear;
+
+       rc = linearize(ctxt, addr, size, true, &linear);
+       if (rc != X86EMUL_CONTINUE)
+               return rc;
+       return ctxt->ops->cmpxchg_emulated(ctxt, linear, orig_data, data,
+                                          size, &ctxt->exception);
+}
+
 static int pio_in_emulated(struct x86_emulate_ctxt *ctxt,
                           struct x86_emulate_ops *ops,
                           unsigned int size, unsigned short port,
@@ -854,7 +1125,7 @@ static int pio_in_emulated(struct x86_emulate_ctxt *ctxt,
                if (n == 0)
                        n = 1;
                rc->pos = rc->end = 0;
-               if (!ops->pio_in_emulated(size, port, rc->data, n, ctxt->vcpu))
+               if (!ops->pio_in_emulated(ctxt, size, port, rc->data, n))
                        return 0;
                rc->end = n * size;
        }
@@ -864,13 +1135,6 @@ static int pio_in_emulated(struct x86_emulate_ctxt *ctxt,
        return 1;
 }
 
-static u32 desc_limit_scaled(struct desc_struct *desc)
-{
-       u32 limit = get_desc_limit(desc);
-
-       return desc->g ? (limit << 12) | 0xfff : limit;
-}
-
 static void get_descriptor_table_ptr(struct x86_emulate_ctxt *ctxt,
                                     struct x86_emulate_ops *ops,
                                     u16 selector, struct desc_ptr *dt)
@@ -903,8 +1167,7 @@ static int read_segment_descriptor(struct x86_emulate_ctxt *ctxt,
        if (dt.size < index * 8 + 7)
                return emulate_gp(ctxt, selector & 0xfffc);
        addr = dt.address + index * 8;
-       ret = ops->read_std(addr, desc, sizeof *desc, ctxt->vcpu,
-                           &ctxt->exception);
+       ret = ops->read_std(ctxt, addr, desc, sizeof *desc, &ctxt->exception);
 
        return ret;
 }
@@ -925,8 +1188,7 @@ static int write_segment_descriptor(struct x86_emulate_ctxt *ctxt,
                return emulate_gp(ctxt, selector & 0xfffc);
 
        addr = dt.address + index * 8;
-       ret = ops->write_std(addr, desc, sizeof *desc, ctxt->vcpu,
-                            &ctxt->exception);
+       ret = ops->write_std(ctxt, addr, desc, sizeof *desc, &ctxt->exception);
 
        return ret;
 }
@@ -1081,23 +1343,22 @@ static inline int writeback(struct x86_emulate_ctxt *ctxt,
                break;
        case OP_MEM:
                if (c->lock_prefix)
-                       rc = ops->cmpxchg_emulated(
-                                       linear(ctxt, c->dst.addr.mem),
-                                       &c->dst.orig_val,
-                                       &c->dst.val,
-                                       c->dst.bytes,
-                                       &ctxt->exception,
-                                       ctxt->vcpu);
+                       rc = segmented_cmpxchg(ctxt,
+                                              c->dst.addr.mem,
+                                              &c->dst.orig_val,
+                                              &c->dst.val,
+                                              c->dst.bytes);
                else
-                       rc = ops->write_emulated(
-                                       linear(ctxt, c->dst.addr.mem),
-                                       &c->dst.val,
-                                       c->dst.bytes,
-                                       &ctxt->exception,
-                                       ctxt->vcpu);
+                       rc = segmented_write(ctxt,
+                                            c->dst.addr.mem,
+                                            &c->dst.val,
+                                            c->dst.bytes);
                if (rc != X86EMUL_CONTINUE)
                        return rc;
                break;
+       case OP_XMM:
+               write_sse_reg(ctxt, &c->dst.vec_val, c->dst.addr.xmm);
+               break;
        case OP_NONE:
                /* no writeback */
                break;
@@ -1107,17 +1368,18 @@ static inline int writeback(struct x86_emulate_ctxt *ctxt,
        return X86EMUL_CONTINUE;
 }
 
-static inline void emulate_push(struct x86_emulate_ctxt *ctxt,
-                               struct x86_emulate_ops *ops)
+static int em_push(struct x86_emulate_ctxt *ctxt)
 {
        struct decode_cache *c = &ctxt->decode;
+       struct segmented_address addr;
 
-       c->dst.type  = OP_MEM;
-       c->dst.bytes = c->op_bytes;
-       c->dst.val = c->src.val;
        register_address_increment(c, &c->regs[VCPU_REGS_RSP], -c->op_bytes);
-       c->dst.addr.mem.ea = register_address(c, c->regs[VCPU_REGS_RSP]);
-       c->dst.addr.mem.seg = VCPU_SREG_SS;
+       addr.ea = register_address(c, c->regs[VCPU_REGS_RSP]);
+       addr.seg = VCPU_SREG_SS;
+
+       /* Disable writeback. */
+       c->dst.type = OP_NONE;
+       return segmented_write(ctxt, addr, &c->src.val, c->op_bytes);
 }
 
 static int emulate_pop(struct x86_emulate_ctxt *ctxt,
@@ -1130,7 +1392,7 @@ static int emulate_pop(struct x86_emulate_ctxt *ctxt,
 
        addr.ea = register_address(c, c->regs[VCPU_REGS_RSP]);
        addr.seg = VCPU_SREG_SS;
-       rc = read_emulated(ctxt, ops, linear(ctxt, addr), dest, len);
+       rc = segmented_read(ctxt, addr, dest, len);
        if (rc != X86EMUL_CONTINUE)
                return rc;
 
@@ -1179,14 +1441,14 @@ static int emulate_popf(struct x86_emulate_ctxt *ctxt,
        return rc;
 }
 
-static void emulate_push_sreg(struct x86_emulate_ctxt *ctxt,
-                             struct x86_emulate_ops *ops, int seg)
+static int emulate_push_sreg(struct x86_emulate_ctxt *ctxt,
+                            struct x86_emulate_ops *ops, int seg)
 {
        struct decode_cache *c = &ctxt->decode;
 
        c->src.val = ops->get_segment_selector(seg, ctxt->vcpu);
 
-       emulate_push(ctxt, ops);
+       return em_push(ctxt);
 }
 
 static int emulate_pop_sreg(struct x86_emulate_ctxt *ctxt,
@@ -1204,8 +1466,7 @@ static int emulate_pop_sreg(struct x86_emulate_ctxt *ctxt,
        return rc;
 }
 
-static int emulate_pusha(struct x86_emulate_ctxt *ctxt,
-                         struct x86_emulate_ops *ops)
+static int emulate_pusha(struct x86_emulate_ctxt *ctxt)
 {
        struct decode_cache *c = &ctxt->decode;
        unsigned long old_esp = c->regs[VCPU_REGS_RSP];
@@ -1216,18 +1477,13 @@ static int emulate_pusha(struct x86_emulate_ctxt *ctxt,
                (reg == VCPU_REGS_RSP) ?
                (c->src.val = old_esp) : (c->src.val = c->regs[reg]);
 
-               emulate_push(ctxt, ops);
-
-               rc = writeback(ctxt, ops);
+               rc = em_push(ctxt);
                if (rc != X86EMUL_CONTINUE)
                        return rc;
 
                ++reg;
        }
 
-       /* Disable writeback. */
-       c->dst.type = OP_NONE;
-
        return rc;
 }
 
@@ -1265,37 +1521,32 @@ int emulate_int_real(struct x86_emulate_ctxt *ctxt,
 
        /* TODO: Add limit checks */
        c->src.val = ctxt->eflags;
-       emulate_push(ctxt, ops);
-       rc = writeback(ctxt, ops);
+       rc = em_push(ctxt);
        if (rc != X86EMUL_CONTINUE)
                return rc;
 
        ctxt->eflags &= ~(EFLG_IF | EFLG_TF | EFLG_AC);
 
        c->src.val = ops->get_segment_selector(VCPU_SREG_CS, ctxt->vcpu);
-       emulate_push(ctxt, ops);
-       rc = writeback(ctxt, ops);
+       rc = em_push(ctxt);
        if (rc != X86EMUL_CONTINUE)
                return rc;
 
        c->src.val = c->eip;
-       emulate_push(ctxt, ops);
-       rc = writeback(ctxt, ops);
+       rc = em_push(ctxt);
        if (rc != X86EMUL_CONTINUE)
                return rc;
 
-       c->dst.type = OP_NONE;
-
        ops->get_idt(&dt, ctxt->vcpu);
 
        eip_addr = dt.address + (irq << 2);
        cs_addr = dt.address + (irq << 2) + 2;
 
-       rc = ops->read_std(cs_addr, &cs, 2, ctxt->vcpu, &ctxt->exception);
+       rc = ops->read_std(ctxt, cs_addr, &cs, 2, &ctxt->exception);
        if (rc != X86EMUL_CONTINUE)
                return rc;
 
-       rc = ops->read_std(eip_addr, &eip, 2, ctxt->vcpu, &ctxt->exception);
+       rc = ops->read_std(ctxt, eip_addr, &eip, 2, &ctxt->exception);
        if (rc != X86EMUL_CONTINUE)
                return rc;
 
@@ -1471,10 +1722,10 @@ static inline int emulate_grp3(struct x86_emulate_ctxt *ctxt,
        return X86EMUL_CONTINUE;
 }
 
-static inline int emulate_grp45(struct x86_emulate_ctxt *ctxt,
-                              struct x86_emulate_ops *ops)
+static int emulate_grp45(struct x86_emulate_ctxt *ctxt)
 {
        struct decode_cache *c = &ctxt->decode;
+       int rc = X86EMUL_CONTINUE;
 
        switch (c->modrm_reg) {
        case 0: /* inc */
@@ -1488,17 +1739,17 @@ static inline int emulate_grp45(struct x86_emulate_ctxt *ctxt,
                old_eip = c->eip;
                c->eip = c->src.val;
                c->src.val = old_eip;
-               emulate_push(ctxt, ops);
+               rc = em_push(ctxt);
                break;
        }
        case 4: /* jmp abs */
                c->eip = c->src.val;
                break;
        case 6: /* push */
-               emulate_push(ctxt, ops);
+               rc = em_push(ctxt);
                break;
        }
-       return X86EMUL_CONTINUE;
+       return rc;
 }
 
 static inline int emulate_grp9(struct x86_emulate_ctxt *ctxt,
@@ -1782,13 +2033,12 @@ static bool emulator_io_port_access_allowed(struct x86_emulate_ctxt *ctxt,
 #ifdef CONFIG_X86_64
        base |= ((u64)base3) << 32;
 #endif
-       r = ops->read_std(base + 102, &io_bitmap_ptr, 2, ctxt->vcpu, NULL);
+       r = ops->read_std(ctxt, base + 102, &io_bitmap_ptr, 2, NULL);
        if (r != X86EMUL_CONTINUE)
                return false;
        if (io_bitmap_ptr + port/8 > desc_limit_scaled(&tr_seg))
                return false;
-       r = ops->read_std(base + io_bitmap_ptr + port/8, &perm, 2, ctxt->vcpu,
-                         NULL);
+       r = ops->read_std(ctxt, base + io_bitmap_ptr + port/8, &perm, 2, NULL);
        if (r != X86EMUL_CONTINUE)
                return false;
        if ((perm >> bit_idx) & mask)
@@ -1896,7 +2146,7 @@ static int task_switch_16(struct x86_emulate_ctxt *ctxt,
        int ret;
        u32 new_tss_base = get_desc_base(new_desc);
 
-       ret = ops->read_std(old_tss_base, &tss_seg, sizeof tss_seg, ctxt->vcpu,
+       ret = ops->read_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
                            &ctxt->exception);
        if (ret != X86EMUL_CONTINUE)
                /* FIXME: need to provide precise fault address */
@@ -1904,13 +2154,13 @@ static int task_switch_16(struct x86_emulate_ctxt *ctxt,
 
        save_state_to_tss16(ctxt, ops, &tss_seg);
 
-       ret = ops->write_std(old_tss_base, &tss_seg, sizeof tss_seg, ctxt->vcpu,
+       ret = ops->write_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
                             &ctxt->exception);
        if (ret != X86EMUL_CONTINUE)
                /* FIXME: need to provide precise fault address */
                return ret;
 
-       ret = ops->read_std(new_tss_base, &tss_seg, sizeof tss_seg, ctxt->vcpu,
+       ret = ops->read_std(ctxt, new_tss_base, &tss_seg, sizeof tss_seg,
                            &ctxt->exception);
        if (ret != X86EMUL_CONTINUE)
                /* FIXME: need to provide precise fault address */
@@ -1919,10 +2169,10 @@ static int task_switch_16(struct x86_emulate_ctxt *ctxt,
        if (old_tss_sel != 0xffff) {
                tss_seg.prev_task_link = old_tss_sel;
 
-               ret = ops->write_std(new_tss_base,
+               ret = ops->write_std(ctxt, new_tss_base,
                                     &tss_seg.prev_task_link,
                                     sizeof tss_seg.prev_task_link,
-                                    ctxt->vcpu, &ctxt->exception);
+                                    &ctxt->exception);
                if (ret != X86EMUL_CONTINUE)
                        /* FIXME: need to provide precise fault address */
                        return ret;
@@ -2028,7 +2278,7 @@ static int task_switch_32(struct x86_emulate_ctxt *ctxt,
        int ret;
        u32 new_tss_base = get_desc_base(new_desc);
 
-       ret = ops->read_std(old_tss_base, &tss_seg, sizeof tss_seg, ctxt->vcpu,
+       ret = ops->read_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
                            &ctxt->exception);
        if (ret != X86EMUL_CONTINUE)
                /* FIXME: need to provide precise fault address */
@@ -2036,13 +2286,13 @@ static int task_switch_32(struct x86_emulate_ctxt *ctxt,
 
        save_state_to_tss32(ctxt, ops, &tss_seg);
 
-       ret = ops->write_std(old_tss_base, &tss_seg, sizeof tss_seg, ctxt->vcpu,
+       ret = ops->write_std(ctxt, old_tss_base, &tss_seg, sizeof tss_seg,
                             &ctxt->exception);
        if (ret != X86EMUL_CONTINUE)
                /* FIXME: need to provide precise fault address */
                return ret;
 
-       ret = ops->read_std(new_tss_base, &tss_seg, sizeof tss_seg, ctxt->vcpu,
+       ret = ops->read_std(ctxt, new_tss_base, &tss_seg, sizeof tss_seg,
                            &ctxt->exception);
        if (ret != X86EMUL_CONTINUE)
                /* FIXME: need to provide precise fault address */
@@ -2051,10 +2301,10 @@ static int task_switch_32(struct x86_emulate_ctxt *ctxt,
        if (old_tss_sel != 0xffff) {
                tss_seg.prev_task_link = old_tss_sel;
 
-               ret = ops->write_std(new_tss_base,
+               ret = ops->write_std(ctxt, new_tss_base,
                                     &tss_seg.prev_task_link,
                                     sizeof tss_seg.prev_task_link,
-                                    ctxt->vcpu, &ctxt->exception);
+                                    &ctxt->exception);
                if (ret != X86EMUL_CONTINUE)
                        /* FIXME: need to provide precise fault address */
                        return ret;
@@ -2142,7 +2392,7 @@ static int emulator_do_task_switch(struct x86_emulate_ctxt *ctxt,
                c->op_bytes = c->ad_bytes = (next_tss_desc.type & 8) ? 4 : 2;
                c->lock_prefix = 0;
                c->src.val = (unsigned long) error_code;
-               emulate_push(ctxt, ops);
+               ret = em_push(ctxt);
        }
 
        return ret;
@@ -2162,13 +2412,10 @@ int emulator_task_switch(struct x86_emulate_ctxt *ctxt,
        rc = emulator_do_task_switch(ctxt, ops, tss_selector, reason,
                                     has_error_code, error_code);
 
-       if (rc == X86EMUL_CONTINUE) {
-               rc = writeback(ctxt, ops);
-               if (rc == X86EMUL_CONTINUE)
-                       ctxt->eip = c->eip;
-       }
+       if (rc == X86EMUL_CONTINUE)
+               ctxt->eip = c->eip;
 
-       return (rc == X86EMUL_UNHANDLEABLE) ? -1 : 0;
+       return (rc == X86EMUL_UNHANDLEABLE) ? EMULATION_FAILED : EMULATION_OK;
 }
 
 static void string_addr_inc(struct x86_emulate_ctxt *ctxt, unsigned seg,
@@ -2182,12 +2429,6 @@ static void string_addr_inc(struct x86_emulate_ctxt *ctxt, unsigned seg,
        op->addr.mem.seg = seg;
 }
 
-static int em_push(struct x86_emulate_ctxt *ctxt)
-{
-       emulate_push(ctxt, ctxt->ops);
-       return X86EMUL_CONTINUE;
-}
-
 static int em_das(struct x86_emulate_ctxt *ctxt)
 {
        struct decode_cache *c = &ctxt->decode;
@@ -2245,20 +2486,12 @@ static int em_call_far(struct x86_emulate_ctxt *ctxt)
        memcpy(&c->eip, c->src.valptr, c->op_bytes);
 
        c->src.val = old_cs;
-       emulate_push(ctxt, ctxt->ops);
-       rc = writeback(ctxt, ctxt->ops);
+       rc = em_push(ctxt);
        if (rc != X86EMUL_CONTINUE)
                return rc;
 
        c->src.val = old_eip;
-       emulate_push(ctxt, ctxt->ops);
-       rc = writeback(ctxt, ctxt->ops);
-       if (rc != X86EMUL_CONTINUE)
-               return rc;
-
-       c->dst.type = OP_NONE;
-
-       return X86EMUL_CONTINUE;
+       return em_push(ctxt);
 }
 
 static int em_ret_near_imm(struct x86_emulate_ctxt *ctxt)
@@ -2306,12 +2539,9 @@ static int em_cwd(struct x86_emulate_ctxt *ctxt)
 
 static int em_rdtsc(struct x86_emulate_ctxt *ctxt)
 {
-       unsigned cpl = ctxt->ops->cpl(ctxt->vcpu);
        struct decode_cache *c = &ctxt->decode;
        u64 tsc = 0;
 
-       if (cpl > 0 && (ctxt->ops->get_cr(4, ctxt->vcpu) & X86_CR4_TSD))
-               return emulate_gp(ctxt, 0);
        ctxt->ops->get_msr(ctxt->vcpu, MSR_IA32_TSC, &tsc);
        c->regs[VCPU_REGS_RAX] = (u32)tsc;
        c->regs[VCPU_REGS_RDX] = tsc >> 32;
@@ -2325,20 +2555,270 @@ static int em_mov(struct x86_emulate_ctxt *ctxt)
        return X86EMUL_CONTINUE;
 }
 
+static int em_movdqu(struct x86_emulate_ctxt *ctxt)
+{
+       struct decode_cache *c = &ctxt->decode;
+       memcpy(&c->dst.vec_val, &c->src.vec_val, c->op_bytes);
+       return X86EMUL_CONTINUE;
+}
+
+static int em_invlpg(struct x86_emulate_ctxt *ctxt)
+{
+       struct decode_cache *c = &ctxt->decode;
+       int rc;
+       ulong linear;
+
+       rc = linearize(ctxt, c->src.addr.mem, 1, false, &linear);
+       if (rc == X86EMUL_CONTINUE)
+               emulate_invlpg(ctxt->vcpu, linear);
+       /* Disable writeback. */
+       c->dst.type = OP_NONE;
+       return X86EMUL_CONTINUE;
+}
+
+static bool valid_cr(int nr)
+{
+       switch (nr) {
+       case 0:
+       case 2 ... 4:
+       case 8:
+               return true;
+       default:
+               return false;
+       }
+}
+
+static int check_cr_read(struct x86_emulate_ctxt *ctxt)
+{
+       struct decode_cache *c = &ctxt->decode;
+
+       if (!valid_cr(c->modrm_reg))
+               return emulate_ud(ctxt);
+
+       return X86EMUL_CONTINUE;
+}
+
+static int check_cr_write(struct x86_emulate_ctxt *ctxt)
+{
+       struct decode_cache *c = &ctxt->decode;
+       u64 new_val = c->src.val64;
+       int cr = c->modrm_reg;
+
+       static u64 cr_reserved_bits[] = {
+               0xffffffff00000000ULL,
+               0, 0, 0, /* CR3 checked later */
+               CR4_RESERVED_BITS,
+               0, 0, 0,
+               CR8_RESERVED_BITS,
+       };
+
+       if (!valid_cr(cr))
+               return emulate_ud(ctxt);
+
+       if (new_val & cr_reserved_bits[cr])
+               return emulate_gp(ctxt, 0);
+
+       switch (cr) {
+       case 0: {
+               u64 cr4, efer;
+               if (((new_val & X86_CR0_PG) && !(new_val & X86_CR0_PE)) ||
+                   ((new_val & X86_CR0_NW) && !(new_val & X86_CR0_CD)))
+                       return emulate_gp(ctxt, 0);
+
+               cr4 = ctxt->ops->get_cr(4, ctxt->vcpu);
+               ctxt->ops->get_msr(ctxt->vcpu, MSR_EFER, &efer);
+
+               if ((new_val & X86_CR0_PG) && (efer & EFER_LME) &&
+                   !(cr4 & X86_CR4_PAE))
+                       return emulate_gp(ctxt, 0);
+
+               break;
+               }
+       case 3: {
+               u64 rsvd = 0;
+
+               if (is_long_mode(ctxt->vcpu))
+                       rsvd = CR3_L_MODE_RESERVED_BITS;
+               else if (is_pae(ctxt->vcpu))
+                       rsvd = CR3_PAE_RESERVED_BITS;
+               else if (is_paging(ctxt->vcpu))
+                       rsvd = CR3_NONPAE_RESERVED_BITS;
+
+               if (new_val & rsvd)
+                       return emulate_gp(ctxt, 0);
+
+               break;
+               }
+       case 4: {
+               u64 cr4, efer;
+
+               cr4 = ctxt->ops->get_cr(4, ctxt->vcpu);
+               ctxt->ops->get_msr(ctxt->vcpu, MSR_EFER, &efer);
+
+               if ((efer & EFER_LMA) && !(new_val & X86_CR4_PAE))
+                       return emulate_gp(ctxt, 0);
+
+               break;
+               }
+       }
+
+       return X86EMUL_CONTINUE;
+}
+
+static int check_dr7_gd(struct x86_emulate_ctxt *ctxt)
+{
+       unsigned long dr7;
+
+       ctxt->ops->get_dr(7, &dr7, ctxt->vcpu);
+
+       /* Check if DR7.Global_Enable is set */
+       return dr7 & (1 << 13);
+}
+
+static int check_dr_read(struct x86_emulate_ctxt *ctxt)
+{
+       struct decode_cache *c = &ctxt->decode;
+       int dr = c->modrm_reg;
+       u64 cr4;
+
+       if (dr > 7)
+               return emulate_ud(ctxt);
+
+       cr4 = ctxt->ops->get_cr(4, ctxt->vcpu);
+       if ((cr4 & X86_CR4_DE) && (dr == 4 || dr == 5))
+               return emulate_ud(ctxt);
+
+       if (check_dr7_gd(ctxt))
+               return emulate_db(ctxt);
+
+       return X86EMUL_CONTINUE;
+}
+
+static int check_dr_write(struct x86_emulate_ctxt *ctxt)
+{
+       struct decode_cache *c = &ctxt->decode;
+       u64 new_val = c->src.val64;
+       int dr = c->modrm_reg;
+
+       if ((dr == 6 || dr == 7) && (new_val & 0xffffffff00000000ULL))
+               return emulate_gp(ctxt, 0);
+
+       return check_dr_read(ctxt);
+}
+
+static int check_svme(struct x86_emulate_ctxt *ctxt)
+{
+       u64 efer;
+
+       ctxt->ops->get_msr(ctxt->vcpu, MSR_EFER, &efer);
+
+       if (!(efer & EFER_SVME))
+               return emulate_ud(ctxt);
+
+       return X86EMUL_CONTINUE;
+}
+
+static int check_svme_pa(struct x86_emulate_ctxt *ctxt)
+{
+       u64 rax = kvm_register_read(ctxt->vcpu, VCPU_REGS_RAX);
+
+       /* Valid physical address? */
+       if (rax & 0xffff000000000000)
+               return emulate_gp(ctxt, 0);
+
+       return check_svme(ctxt);
+}
+
+static int check_rdtsc(struct x86_emulate_ctxt *ctxt)
+{
+       u64 cr4 = ctxt->ops->get_cr(4, ctxt->vcpu);
+
+       if (cr4 & X86_CR4_TSD && ctxt->ops->cpl(ctxt->vcpu))
+               return emulate_ud(ctxt);
+
+       return X86EMUL_CONTINUE;
+}
+
+static int check_rdpmc(struct x86_emulate_ctxt *ctxt)
+{
+       u64 cr4 = ctxt->ops->get_cr(4, ctxt->vcpu);
+       u64 rcx = kvm_register_read(ctxt->vcpu, VCPU_REGS_RCX);
+
+       if ((!(cr4 & X86_CR4_PCE) && ctxt->ops->cpl(ctxt->vcpu)) ||
+           (rcx > 3))
+               return emulate_gp(ctxt, 0);
+
+       return X86EMUL_CONTINUE;
+}
+
+static int check_perm_in(struct x86_emulate_ctxt *ctxt)
+{
+       struct decode_cache *c = &ctxt->decode;
+
+       c->dst.bytes = min(c->dst.bytes, 4u);
+       if (!emulator_io_permited(ctxt, ctxt->ops, c->src.val, c->dst.bytes))
+               return emulate_gp(ctxt, 0);
+
+       return X86EMUL_CONTINUE;
+}
+
+static int check_perm_out(struct x86_emulate_ctxt *ctxt)
+{
+       struct decode_cache *c = &ctxt->decode;
+
+       c->src.bytes = min(c->src.bytes, 4u);
+       if (!emulator_io_permited(ctxt, ctxt->ops, c->dst.val, c->src.bytes))
+               return emulate_gp(ctxt, 0);
+
+       return X86EMUL_CONTINUE;
+}
+
 #define D(_y) { .flags = (_y) }
+#define DI(_y, _i) { .flags = (_y), .intercept = x86_intercept_##_i }
+#define DIP(_y, _i, _p) { .flags = (_y), .intercept = x86_intercept_##_i, \
+                     .check_perm = (_p) }
 #define N    D(0)
+#define EXT(_f, _e) { .flags = ((_f) | RMExt), .u.group = (_e) }
 #define G(_f, _g) { .flags = ((_f) | Group), .u.group = (_g) }
 #define GD(_f, _g) { .flags = ((_f) | Group | GroupDual), .u.gdual = (_g) }
 #define I(_f, _e) { .flags = (_f), .u.execute = (_e) }
+#define II(_f, _e, _i) \
+       { .flags = (_f), .u.execute = (_e), .intercept = x86_intercept_##_i }
+#define IIP(_f, _e, _i, _p) \
+       { .flags = (_f), .u.execute = (_e), .intercept = x86_intercept_##_i, \
+         .check_perm = (_p) }
+#define GP(_f, _g) { .flags = ((_f) | Prefix), .u.gprefix = (_g) }
 
 #define D2bv(_f)      D((_f) | ByteOp), D(_f)
+#define D2bvIP(_f, _i, _p) DIP((_f) | ByteOp, _i, _p), DIP(_f, _i, _p)
 #define I2bv(_f, _e)  I((_f) | ByteOp, _e), I(_f, _e)
 
 #define D6ALU(_f) D2bv((_f) | DstMem | SrcReg | ModRM),                        \
                D2bv(((_f) | DstReg | SrcMem | ModRM) & ~Lock),         \
                D2bv(((_f) & ~Lock) | DstAcc | SrcImm)
 
+static struct opcode group7_rm1[] = {
+       DI(SrcNone | ModRM | Priv, monitor),
+       DI(SrcNone | ModRM | Priv, mwait),
+       N, N, N, N, N, N,
+};
+
+static struct opcode group7_rm3[] = {
+       DIP(SrcNone | ModRM | Prot | Priv, vmrun,   check_svme_pa),
+       DI(SrcNone | ModRM | Prot | VendorSpecific, vmmcall),
+       DIP(SrcNone | ModRM | Prot | Priv, vmload,  check_svme_pa),
+       DIP(SrcNone | ModRM | Prot | Priv, vmsave,  check_svme_pa),
+       DIP(SrcNone | ModRM | Prot | Priv, stgi,    check_svme),
+       DIP(SrcNone | ModRM | Prot | Priv, clgi,    check_svme),
+       DIP(SrcNone | ModRM | Prot | Priv, skinit,  check_svme),
+       DIP(SrcNone | ModRM | Prot | Priv, invlpga, check_svme),
+};
 
+static struct opcode group7_rm7[] = {
+       N,
+       DIP(SrcNone | ModRM, rdtscp, check_rdtsc),
+       N, N, N, N, N, N,
+};
 static struct opcode group1[] = {
        X7(D(Lock)), N
 };
@@ -2366,16 +2846,26 @@ static struct opcode group5[] = {
        D(SrcMem | ModRM | Stack), N,
 };
 
+static struct opcode group6[] = {
+       DI(ModRM | Prot,        sldt),
+       DI(ModRM | Prot,        str),
+       DI(ModRM | Prot | Priv, lldt),
+       DI(ModRM | Prot | Priv, ltr),
+       N, N, N, N,
+};
+
 static struct group_dual group7 = { {
-       N, N, D(ModRM | SrcMem | Priv), D(ModRM | SrcMem | Priv),
-       D(SrcNone | ModRM | DstMem | Mov), N,
-       D(SrcMem16 | ModRM | Mov | Priv),
-       D(SrcMem | ModRM | ByteOp | Priv | NoAccess),
+       DI(ModRM | Mov | DstMem | Priv, sgdt),
+       DI(ModRM | Mov | DstMem | Priv, sidt),
+       DI(ModRM | SrcMem | Priv, lgdt), DI(ModRM | SrcMem | Priv, lidt),
+       DI(SrcNone | ModRM | DstMem | Mov, smsw), N,
+       DI(SrcMem16 | ModRM | Mov | Priv, lmsw),
+       DI(SrcMem | ModRM | ByteOp | Priv | NoAccess, invlpg),
 }, {
-       D(SrcNone | ModRM | Priv | VendorSpecific), N,
-       N, D(SrcNone | ModRM | Priv | VendorSpecific),
-       D(SrcNone | ModRM | DstMem | Mov), N,
-       D(SrcMem16 | ModRM | Mov | Priv), N,
+       D(SrcNone | ModRM | Priv | VendorSpecific), EXT(0, group7_rm1),
+       N, EXT(0, group7_rm3),
+       DI(SrcNone | ModRM | DstMem | Mov, smsw), N,
+       DI(SrcMem16 | ModRM | Mov | Priv, lmsw), EXT(0, group7_rm7),
 } };
 
 static struct opcode group8[] = {
@@ -2394,6 +2884,10 @@ static struct opcode group11[] = {
        I(DstMem | SrcImm | ModRM | Mov, em_mov), X7(D(Undefined)),
 };
 
+static struct gprefix pfx_0f_6f_0f_7f = {
+       N, N, N, I(Sse, em_movdqu),
+};
+
 static struct opcode opcode_table[256] = {
        /* 0x00 - 0x07 */
        D6ALU(Lock),
@@ -2430,8 +2924,8 @@ static struct opcode opcode_table[256] = {
        I(DstReg | SrcMem | ModRM | Src2Imm, em_imul_3op),
        I(SrcImmByte | Mov | Stack, em_push),
        I(DstReg | SrcMem | ModRM | Src2ImmByte, em_imul_3op),
-       D2bv(DstDI | Mov | String), /* insb, insw/insd */
-       D2bv(SrcSI | ImplicitOps | String), /* outsb, outsw/outsd */
+       D2bvIP(DstDI | Mov | String, ins, check_perm_in), /* insb, insw/insd */
+       D2bvIP(SrcSI | ImplicitOps | String, outs, check_perm_out), /* outsb, outsw/outsd */
        /* 0x70 - 0x7F */
        X16(D(SrcImmByte)),
        /* 0x80 - 0x87 */
@@ -2446,11 +2940,11 @@ static struct opcode opcode_table[256] = {
        D(DstMem | SrcNone | ModRM | Mov), D(ModRM | SrcMem | NoAccess | DstReg),
        D(ImplicitOps | SrcMem16 | ModRM), G(0, group1A),
        /* 0x90 - 0x97 */
-       X8(D(SrcAcc | DstReg)),
+       DI(SrcAcc | DstReg, pause), X7(D(SrcAcc | DstReg)),
        /* 0x98 - 0x9F */
        D(DstAcc | SrcNone), I(ImplicitOps | SrcAcc, em_cwd),
        I(SrcImmFAddr | No64, em_call_far), N,
-       D(ImplicitOps | Stack), D(ImplicitOps | Stack), N, N,
+       DI(ImplicitOps | Stack, pushf), DI(ImplicitOps | Stack, popf), N, N,
        /* 0xA0 - 0xA7 */
        I2bv(DstAcc | SrcMem | Mov | MemAbs, em_mov),
        I2bv(DstMem | SrcAcc | Mov | MemAbs, em_mov),
@@ -2473,7 +2967,8 @@ static struct opcode opcode_table[256] = {
        G(ByteOp, group11), G(0, group11),
        /* 0xC8 - 0xCF */
        N, N, N, D(ImplicitOps | Stack),
-       D(ImplicitOps), D(SrcImmByte), D(ImplicitOps | No64), D(ImplicitOps),
+       D(ImplicitOps), DI(SrcImmByte, intn),
+       D(ImplicitOps | No64), DI(ImplicitOps, iret),
        /* 0xD0 - 0xD7 */
        D2bv(DstMem | SrcOne | ModRM), D2bv(DstMem | ModRM),
        N, N, N, N,
@@ -2481,14 +2976,17 @@ static struct opcode opcode_table[256] = {
        N, N, N, N, N, N, N, N,
        /* 0xE0 - 0xE7 */
        X4(D(SrcImmByte)),
-       D2bv(SrcImmUByte | DstAcc), D2bv(SrcAcc | DstImmUByte),
+       D2bvIP(SrcImmUByte | DstAcc, in,  check_perm_in),
+       D2bvIP(SrcAcc | DstImmUByte, out, check_perm_out),
        /* 0xE8 - 0xEF */
        D(SrcImm | Stack), D(SrcImm | ImplicitOps),
        D(SrcImmFAddr | No64), D(SrcImmByte | ImplicitOps),
-       D2bv(SrcNone | DstAcc), D2bv(SrcAcc | ImplicitOps),
+       D2bvIP(SrcNone | DstAcc,     in,  check_perm_in),
+       D2bvIP(SrcAcc | ImplicitOps, out, check_perm_out),
        /* 0xF0 - 0xF7 */
-       N, N, N, N,
-       D(ImplicitOps | Priv), D(ImplicitOps), G(ByteOp, group3), G(0, group3),
+       N, DI(ImplicitOps, icebp), N, N,
+       DI(ImplicitOps | Priv, hlt), D(ImplicitOps),
+       G(ByteOp, group3), G(0, group3),
        /* 0xF8 - 0xFF */
        D(ImplicitOps), D(ImplicitOps), D(ImplicitOps), D(ImplicitOps),
        D(ImplicitOps), D(ImplicitOps), G(0, group4), G(0, group5),
@@ -2496,20 +2994,24 @@ static struct opcode opcode_table[256] = {
 
 static struct opcode twobyte_table[256] = {
        /* 0x00 - 0x0F */
-       N, GD(0, &group7), N, N,
-       N, D(ImplicitOps | VendorSpecific), D(ImplicitOps | Priv), N,
-       D(ImplicitOps | Priv), D(ImplicitOps | Priv), N, N,
+       G(0, group6), GD(0, &group7), N, N,
+       N, D(ImplicitOps | VendorSpecific), DI(ImplicitOps | Priv, clts), N,
+       DI(ImplicitOps | Priv, invd), DI(ImplicitOps | Priv, wbinvd), N, N,
        N, D(ImplicitOps | ModRM), N, N,
        /* 0x10 - 0x1F */
        N, N, N, N, N, N, N, N, D(ImplicitOps | ModRM), N, N, N, N, N, N, N,
        /* 0x20 - 0x2F */
-       D(ModRM | DstMem | Priv | Op3264), D(ModRM | DstMem | Priv | Op3264),
-       D(ModRM | SrcMem | Priv | Op3264), D(ModRM | SrcMem | Priv | Op3264),
+       DIP(ModRM | DstMem | Priv | Op3264, cr_read, check_cr_read),
+       DIP(ModRM | DstMem | Priv | Op3264, dr_read, check_dr_read),
+       DIP(ModRM | SrcMem | Priv | Op3264, cr_write, check_cr_write),
+       DIP(ModRM | SrcMem | Priv | Op3264, dr_write, check_dr_write),
        N, N, N, N,
        N, N, N, N, N, N, N, N,
        /* 0x30 - 0x3F */
-       D(ImplicitOps | Priv), I(ImplicitOps, em_rdtsc),
-       D(ImplicitOps | Priv), N,
+       DI(ImplicitOps | Priv, wrmsr),
+       IIP(ImplicitOps, em_rdtsc, rdtsc, check_rdtsc),
+       DI(ImplicitOps | Priv, rdmsr),
+       DIP(ImplicitOps | Priv, rdpmc, check_rdpmc),
        D(ImplicitOps | VendorSpecific), D(ImplicitOps | Priv | VendorSpecific),
        N, N,
        N, N, N, N, N, N, N, N,
@@ -2518,21 +3020,27 @@ static struct opcode twobyte_table[256] = {
        /* 0x50 - 0x5F */
        N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
        /* 0x60 - 0x6F */
-       N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
+       N, N, N, N,
+       N, N, N, N,
+       N, N, N, N,
+       N, N, N, GP(SrcMem | DstReg | ModRM | Mov, &pfx_0f_6f_0f_7f),
        /* 0x70 - 0x7F */
-       N, N, N, N, N, N, N, N, N, N, N, N, N, N, N, N,
+       N, N, N, N,
+       N, N, N, N,
+       N, N, N, N,
+       N, N, N, GP(SrcReg | DstMem | ModRM | Mov, &pfx_0f_6f_0f_7f),
        /* 0x80 - 0x8F */
        X16(D(SrcImm)),
        /* 0x90 - 0x9F */
        X16(D(ByteOp | DstMem | SrcNone | ModRM| Mov)),
        /* 0xA0 - 0xA7 */
        D(ImplicitOps | Stack), D(ImplicitOps | Stack),
-       N, D(DstMem | SrcReg | ModRM | BitOp),
+       DI(ImplicitOps, cpuid), D(DstMem | SrcReg | ModRM | BitOp),
        D(DstMem | SrcReg | Src2ImmByte | ModRM),
        D(DstMem | SrcReg | Src2CL | ModRM), N, N,
        /* 0xA8 - 0xAF */
        D(ImplicitOps | Stack), D(ImplicitOps | Stack),
-       N, D(DstMem | SrcReg | ModRM | BitOp | Lock),
+       DI(ImplicitOps, rsm), D(DstMem | SrcReg | ModRM | BitOp | Lock),
        D(DstMem | SrcReg | Src2ImmByte | ModRM),
        D(DstMem | SrcReg | Src2CL | ModRM),
        D(ModRM), I(DstReg | SrcMem | ModRM, em_imul),
@@ -2564,8 +3072,11 @@ static struct opcode twobyte_table[256] = {
 #undef G
 #undef GD
 #undef I
+#undef GP
+#undef EXT
 
 #undef D2bv
+#undef D2bvIP
 #undef I2bv
 #undef D6ALU
 
@@ -2625,7 +3136,8 @@ x86_decode_insn(struct x86_emulate_ctxt *ctxt, void *insn, int insn_len)
        struct decode_cache *c = &ctxt->decode;
        int rc = X86EMUL_CONTINUE;
        int mode = ctxt->mode;
-       int def_op_bytes, def_ad_bytes, dual, goffset;
+       int def_op_bytes, def_ad_bytes, dual, goffset, simd_prefix;
+       bool op_prefix = false;
        struct opcode opcode, *g_mod012, *g_mod3;
        struct operand memop = { .type = OP_NONE };
 
@@ -2634,7 +3146,6 @@ x86_decode_insn(struct x86_emulate_ctxt *ctxt, void *insn, int insn_len)
        c->fetch.end = c->fetch.start + insn_len;
        if (insn_len > 0)
                memcpy(c->fetch.data, insn, insn_len);
-       ctxt->cs_base = seg_base(ctxt, ops, VCPU_SREG_CS);
 
        switch (mode) {
        case X86EMUL_MODE_REAL:
@@ -2662,6 +3173,7 @@ x86_decode_insn(struct x86_emulate_ctxt *ctxt, void *insn, int insn_len)
        for (;;) {
                switch (c->b = insn_fetch(u8, 1, c->eip)) {
                case 0x66:      /* operand-size override */
+                       op_prefix = true;
                        /* switch between 2/4 bytes */
                        c->op_bytes = def_op_bytes ^ 6;
                        break;
@@ -2692,10 +3204,8 @@ x86_decode_insn(struct x86_emulate_ctxt *ctxt, void *insn, int insn_len)
                        c->lock_prefix = 1;
                        break;
                case 0xf2:      /* REPNE/REPNZ */
-                       c->rep_prefix = REPNE_PREFIX;
-                       break;
                case 0xf3:      /* REP/REPE/REPZ */
-                       c->rep_prefix = REPE_PREFIX;
+                       c->rep_prefix = c->b;
                        break;
                default:
                        goto done_prefixes;
@@ -2741,10 +3251,31 @@ done_prefixes:
                        opcode = g_mod3[goffset];
                else
                        opcode = g_mod012[goffset];
+
+               if (opcode.flags & RMExt) {
+                       goffset = c->modrm & 7;
+                       opcode = opcode.u.group[goffset];
+               }
+
+               c->d |= opcode.flags;
+       }
+
+       if (c->d & Prefix) {
+               if (c->rep_prefix && op_prefix)
+                       return X86EMUL_UNHANDLEABLE;
+               simd_prefix = op_prefix ? 0x66 : c->rep_prefix;
+               switch (simd_prefix) {
+               case 0x00: opcode = opcode.u.gprefix->pfx_no; break;
+               case 0x66: opcode = opcode.u.gprefix->pfx_66; break;
+               case 0xf2: opcode = opcode.u.gprefix->pfx_f2; break;
+               case 0xf3: opcode = opcode.u.gprefix->pfx_f3; break;
+               }
                c->d |= opcode.flags;
        }
 
        c->execute = opcode.u.execute;
+       c->check_perm = opcode.check_perm;
+       c->intercept = opcode.intercept;
 
        /* Unrecognised? */
        if (c->d == 0 || (c->d & Undefined))
@@ -2763,6 +3294,9 @@ done_prefixes:
                        c->op_bytes = 4;
        }
 
+       if (c->d & Sse)
+               c->op_bytes = 16;
+
        /* ModRM and SIB bytes. */
        if (c->d & ModRM) {
                rc = decode_modrm(ctxt, ops, &memop);
@@ -2792,7 +3326,7 @@ done_prefixes:
        case SrcNone:
                break;
        case SrcReg:
-               decode_register_operand(&c->src, c, 0);
+               decode_register_operand(ctxt, &c->src, c, 0);
                break;
        case SrcMem16:
                memop.bytes = 2;
@@ -2883,7 +3417,7 @@ done_prefixes:
        /* Decode and fetch the destination operand: register or memory. */
        switch (c->d & DstMask) {
        case DstReg:
-               decode_register_operand(&c->dst, c,
+               decode_register_operand(ctxt, &c->dst, c,
                         c->twobyte && (c->b == 0xb6 || c->b == 0xb7));
                break;
        case DstImmUByte:
@@ -2926,7 +3460,7 @@ done_prefixes:
        }
 
 done:
-       return (rc == X86EMUL_UNHANDLEABLE) ? -1 : 0;
+       return (rc == X86EMUL_UNHANDLEABLE) ? EMULATION_FAILED : EMULATION_OK;
 }
 
 static bool string_insn_completed(struct x86_emulate_ctxt *ctxt)
@@ -2979,12 +3513,51 @@ x86_emulate_insn(struct x86_emulate_ctxt *ctxt)
                goto done;
        }
 
+       if ((c->d & Sse)
+           && ((ops->get_cr(0, ctxt->vcpu) & X86_CR0_EM)
+               || !(ops->get_cr(4, ctxt->vcpu) & X86_CR4_OSFXSR))) {
+               rc = emulate_ud(ctxt);
+               goto done;
+       }
+
+       if ((c->d & Sse) && (ops->get_cr(0, ctxt->vcpu) & X86_CR0_TS)) {
+               rc = emulate_nm(ctxt);
+               goto done;
+       }
+
+       if (unlikely(ctxt->guest_mode) && c->intercept) {
+               rc = emulator_check_intercept(ctxt, c->intercept,
+                                             X86_ICPT_PRE_EXCEPT);
+               if (rc != X86EMUL_CONTINUE)
+                       goto done;
+       }
+
        /* Privileged instruction can be executed only in CPL=0 */
        if ((c->d & Priv) && ops->cpl(ctxt->vcpu)) {
                rc = emulate_gp(ctxt, 0);
                goto done;
        }
 
+       /* Instruction can only be executed in protected mode */
+       if ((c->d & Prot) && !(ctxt->mode & X86EMUL_MODE_PROT)) {
+               rc = emulate_ud(ctxt);
+               goto done;
+       }
+
+       /* Do instruction specific permission checks */
+       if (c->check_perm) {
+               rc = c->check_perm(ctxt);
+               if (rc != X86EMUL_CONTINUE)
+                       goto done;
+       }
+
+       if (unlikely(ctxt->guest_mode) && c->intercept) {
+               rc = emulator_check_intercept(ctxt, c->intercept,
+                                             X86_ICPT_POST_EXCEPT);
+               if (rc != X86EMUL_CONTINUE)
+                       goto done;
+       }
+
        if (c->rep_prefix && (c->d & String)) {
                /* All REP prefixes have the same first termination condition */
                if (address_mask(c, c->regs[VCPU_REGS_RCX]) == 0) {
@@ -2994,16 +3567,16 @@ x86_emulate_insn(struct x86_emulate_ctxt *ctxt)
        }
 
        if ((c->src.type == OP_MEM) && !(c->d & NoAccess)) {
-               rc = read_emulated(ctxt, ops, linear(ctxt, c->src.addr.mem),
-                                       c->src.valptr, c->src.bytes);
+               rc = segmented_read(ctxt, c->src.addr.mem,
+                                   c->src.valptr, c->src.bytes);
                if (rc != X86EMUL_CONTINUE)
                        goto done;
                c->src.orig_val64 = c->src.val64;
        }
 
        if (c->src2.type == OP_MEM) {
-               rc = read_emulated(ctxt, ops, linear(ctxt, c->src2.addr.mem),
-                                       &c->src2.val, c->src2.bytes);
+               rc = segmented_read(ctxt, c->src2.addr.mem,
+                                   &c->src2.val, c->src2.bytes);
                if (rc != X86EMUL_CONTINUE)
                        goto done;
        }
@@ -3014,7 +3587,7 @@ x86_emulate_insn(struct x86_emulate_ctxt *ctxt)
 
        if ((c->dst.type == OP_MEM) && !(c->d & Mov)) {
                /* optimisation - avoid slow emulated read if Mov */
-               rc = read_emulated(ctxt, ops, linear(ctxt, c->dst.addr.mem),
+               rc = segmented_read(ctxt, c->dst.addr.mem,
                                   &c->dst.val, c->dst.bytes);
                if (rc != X86EMUL_CONTINUE)
                        goto done;
@@ -3023,6 +3596,13 @@ x86_emulate_insn(struct x86_emulate_ctxt *ctxt)
 
 special_insn:
 
+       if (unlikely(ctxt->guest_mode) && c->intercept) {
+               rc = emulator_check_intercept(ctxt, c->intercept,
+                                             X86_ICPT_POST_MEMACCESS);
+               if (rc != X86EMUL_CONTINUE)
+                       goto done;
+       }
+
        if (c->execute) {
                rc = c->execute(ctxt);
                if (rc != X86EMUL_CONTINUE)
@@ -3039,7 +3619,7 @@ special_insn:
                emulate_2op_SrcV("add", c->src, c->dst, ctxt->eflags);
                break;
        case 0x06:              /* push es */
-               emulate_push_sreg(ctxt, ops, VCPU_SREG_ES);
+               rc = emulate_push_sreg(ctxt, ops, VCPU_SREG_ES);
                break;
        case 0x07:              /* pop es */
                rc = emulate_pop_sreg(ctxt, ops, VCPU_SREG_ES);
@@ -3049,14 +3629,14 @@ special_insn:
                emulate_2op_SrcV("or", c->src, c->dst, ctxt->eflags);
                break;
        case 0x0e:              /* push cs */
-               emulate_push_sreg(ctxt, ops, VCPU_SREG_CS);
+               rc = emulate_push_sreg(ctxt, ops, VCPU_SREG_CS);
                break;
        case 0x10 ... 0x15:
              adc:              /* adc */
                emulate_2op_SrcV("adc", c->src, c->dst, ctxt->eflags);
                break;
        case 0x16:              /* push ss */
-               emulate_push_sreg(ctxt, ops, VCPU_SREG_SS);
+               rc = emulate_push_sreg(ctxt, ops, VCPU_SREG_SS);
                break;
        case 0x17:              /* pop ss */
                rc = emulate_pop_sreg(ctxt, ops, VCPU_SREG_SS);
@@ -3066,7 +3646,7 @@ special_insn:
                emulate_2op_SrcV("sbb", c->src, c->dst, ctxt->eflags);
                break;
        case 0x1e:              /* push ds */
-               emulate_push_sreg(ctxt, ops, VCPU_SREG_DS);
+               rc = emulate_push_sreg(ctxt, ops, VCPU_SREG_DS);
                break;
        case 0x1f:              /* pop ds */
                rc = emulate_pop_sreg(ctxt, ops, VCPU_SREG_DS);
@@ -3085,6 +3665,7 @@ special_insn:
                break;
        case 0x38 ... 0x3d:
              cmp:              /* cmp */
+               c->dst.type = OP_NONE; /* Disable writeback. */
                emulate_2op_SrcV("cmp", c->src, c->dst, ctxt->eflags);
                break;
        case 0x40 ... 0x47: /* inc r16/r32 */
@@ -3098,7 +3679,7 @@ special_insn:
                rc = emulate_pop(ctxt, ops, &c->dst.val, c->op_bytes);
                break;
        case 0x60:      /* pusha */
-               rc = emulate_pusha(ctxt, ops);
+               rc = emulate_pusha(ctxt);
                break;
        case 0x61:      /* popa */
                rc = emulate_popa(ctxt, ops);
@@ -3202,7 +3783,7 @@ special_insn:
                break;
        case 0x9c: /* pushf */
                c->src.val =  (unsigned long) ctxt->eflags;
-               emulate_push(ctxt, ops);
+               rc = em_push(ctxt);
                break;
        case 0x9d: /* popf */
                c->dst.type = OP_REG;
@@ -3211,7 +3792,6 @@ special_insn:
                rc = emulate_popf(ctxt, ops, &c->dst.val, c->op_bytes);
                break;
        case 0xa6 ... 0xa7:     /* cmps */
-               c->dst.type = OP_NONE; /* Disable writeback. */
                goto cmp;
        case 0xa8 ... 0xa9:     /* test ax, imm */
                goto test;
@@ -3278,7 +3858,7 @@ special_insn:
                long int rel = c->src.val;
                c->src.val = (unsigned long) c->eip;
                jmp_rel(c, rel);
-               emulate_push(ctxt, ops);
+               rc = em_push(ctxt);
                break;
        }
        case 0xe9: /* jmp rel */
@@ -3304,11 +3884,6 @@ special_insn:
        case 0xed: /* in (e/r)ax,dx */
                c->src.val = c->regs[VCPU_REGS_RDX];
        do_io_in:
-               c->dst.bytes = min(c->dst.bytes, 4u);
-               if (!emulator_io_permited(ctxt, ops, c->src.val, c->dst.bytes)) {
-                       rc = emulate_gp(ctxt, 0);
-                       goto done;
-               }
                if (!pio_in_emulated(ctxt, ops, c->dst.bytes, c->src.val,
                                     &c->dst.val))
                        goto done; /* IO is needed */
@@ -3317,14 +3892,8 @@ special_insn:
        case 0xef: /* out dx,(e/r)ax */
                c->dst.val = c->regs[VCPU_REGS_RDX];
        do_io_out:
-               c->src.bytes = min(c->src.bytes, 4u);
-               if (!emulator_io_permited(ctxt, ops, c->dst.val,
-                                         c->src.bytes)) {
-                       rc = emulate_gp(ctxt, 0);
-                       goto done;
-               }
-               ops->pio_out_emulated(c->src.bytes, c->dst.val,
-                                     &c->src.val, 1, ctxt->vcpu);
+               ops->pio_out_emulated(ctxt, c->src.bytes, c->dst.val,
+                                     &c->src.val, 1);
                c->dst.type = OP_NONE;  /* Disable writeback. */
                break;
        case 0xf4:              /* hlt */
@@ -3367,7 +3936,7 @@ special_insn:
                break;
        case 0xfe: /* Grp4 */
        grp45:
-               rc = emulate_grp45(ctxt, ops);
+               rc = emulate_grp45(ctxt);
                break;
        case 0xff: /* Grp5 */
                if (c->modrm_reg == 5)
@@ -3427,6 +3996,9 @@ writeback:
 done:
        if (rc == X86EMUL_PROPAGATE_FAULT)
                ctxt->have_exception = true;
+       if (rc == X86EMUL_INTERCEPTED)
+               return EMULATION_INTERCEPTED;
+
        return (rc == X86EMUL_UNHANDLEABLE) ? EMULATION_FAILED : EMULATION_OK;
 
 twobyte_insn:
@@ -3492,10 +4064,7 @@ twobyte_insn:
                        rc = X86EMUL_PROPAGATE_FAULT;
                        goto done;
                case 7: /* invlpg*/
-                       emulate_invlpg(ctxt->vcpu,
-                                      linear(ctxt, c->src.addr.mem));
-                       /* Disable writeback. */
-                       c->dst.type = OP_NONE;
+                       rc = em_invlpg(ctxt);
                        break;
                default:
                        goto cannot_emulate;
@@ -3515,23 +4084,9 @@ twobyte_insn:
        case 0x18:              /* Grp16 (prefetch/nop) */
                break;
        case 0x20: /* mov cr, reg */
-               switch (c->modrm_reg) {
-               case 1:
-               case 5 ... 7:
-               case 9 ... 15:
-                       emulate_ud(ctxt);
-                       rc = X86EMUL_PROPAGATE_FAULT;
-                       goto done;
-               }
                c->dst.val = ops->get_cr(c->modrm_reg, ctxt->vcpu);
                break;
        case 0x21: /* mov from dr to reg */
-               if ((ops->get_cr(4, ctxt->vcpu) & X86_CR4_DE) &&
-                   (c->modrm_reg == 4 || c->modrm_reg == 5)) {
-                       emulate_ud(ctxt);
-                       rc = X86EMUL_PROPAGATE_FAULT;
-                       goto done;
-               }
                ops->get_dr(c->modrm_reg, &c->dst.val, ctxt->vcpu);
                break;
        case 0x22: /* mov reg, cr */
@@ -3543,13 +4098,6 @@ twobyte_insn:
                c->dst.type = OP_NONE;
                break;
        case 0x23: /* mov from reg to dr */
-               if ((ops->get_cr(4, ctxt->vcpu) & X86_CR4_DE) &&
-                   (c->modrm_reg == 4 || c->modrm_reg == 5)) {
-                       emulate_ud(ctxt);
-                       rc = X86EMUL_PROPAGATE_FAULT;
-                       goto done;
-               }
-
                if (ops->set_dr(c->modrm_reg, c->src.val &
                                ((ctxt->mode == X86EMUL_MODE_PROT64) ?
                                 ~0ULL : ~0U), ctxt->vcpu) < 0) {
@@ -3603,7 +4151,7 @@ twobyte_insn:
                c->dst.val = test_cc(c->b, ctxt->eflags);
                break;
        case 0xa0:        /* push fs */
-               emulate_push_sreg(ctxt, ops, VCPU_SREG_FS);
+               rc = emulate_push_sreg(ctxt, ops, VCPU_SREG_FS);
                break;
        case 0xa1:       /* pop fs */
                rc = emulate_pop_sreg(ctxt, ops, VCPU_SREG_FS);
@@ -3620,7 +4168,7 @@ twobyte_insn:
                emulate_2op_cl("shld", c->src2, c->src, c->dst, ctxt->eflags);
                break;
        case 0xa8:      /* push gs */
-               emulate_push_sreg(ctxt, ops, VCPU_SREG_GS);
+               rc = emulate_push_sreg(ctxt, ops, VCPU_SREG_GS);
                break;
        case 0xa9:      /* pop gs */
                rc = emulate_pop_sreg(ctxt, ops, VCPU_SREG_GS);
@@ -3739,5 +4287,5 @@ twobyte_insn:
        goto writeback;
 
 cannot_emulate:
-       return -1;
+       return EMULATION_FAILED;
 }
This page took 0.047176 seconds and 5 git commands to generate.