Palacios Public Git Repository

To checkout Palacios execute

  git clone http://v3vee.org/palacios/palacios.web/palacios.git
This will give you the master branch. You probably want the devel branch or one of the release branches. To switch to the devel branch, simply execute
  cd palacios
  git checkout --track -b devel origin/devel
The other branches are similar.


Merge branch 'devel' of ssh://palacios@newskysaw.cs.northwestern.edu/home/palacios...
Lei Xia [Tue, 3 Feb 2009 22:09:59 +0000 (16:09 -0600)]
46 files changed:
build/Makefile
geekos/build/Makefile
geekos/src/geekos/vm.c
palacios/build/Makefile
palacios/include/palacios/svm_handler.h
palacios/include/palacios/vm_guest.h
palacios/include/palacios/vmcb.h
palacios/include/palacios/vmm.h
palacios/include/palacios/vmm_decoder.h
palacios/include/palacios/vmm_emulator.h
palacios/include/palacios/vmm_hashtable.h
palacios/include/palacios/vmm_instr_emulator.h [new file with mode: 0644]
palacios/include/palacios/vmm_io.h
palacios/include/palacios/vmm_list.h
palacios/include/palacios/vmm_mem.h
palacios/include/palacios/vmm_paging.h
palacios/include/palacios/vmm_profiler.h [new file with mode: 0644]
palacios/include/palacios/vmm_rbtree.h [new file with mode: 0644]
palacios/include/palacios/vmm_shadow_paging.h
palacios/include/palacios/vmm_xed.h
palacios/src/devices/8259a.c
palacios/src/devices/keyboard.c
palacios/src/devices/nvram.c
palacios/src/palacios/svm.c
palacios/src/palacios/svm_handler.c
palacios/src/palacios/svm_io.c
palacios/src/palacios/vm_guest_mem.c
palacios/src/palacios/vmm.c
palacios/src/palacios/vmm_config.c
palacios/src/palacios/vmm_ctrl_regs.c
palacios/src/palacios/vmm_decoder.c
palacios/src/palacios/vmm_emulator.c
palacios/src/palacios/vmm_hashtable.c
palacios/src/palacios/vmm_instr_emulator.c [deleted file]
palacios/src/palacios/vmm_io.c
palacios/src/palacios/vmm_mem.c
palacios/src/palacios/vmm_paging.c
palacios/src/palacios/vmm_paging_debug.h
palacios/src/palacios/vmm_profiler.c [new file with mode: 0644]
palacios/src/palacios/vmm_rbtree.c [new file with mode: 0644]
palacios/src/palacios/vmm_shadow_paging.c
palacios/src/palacios/vmm_shadow_paging_32.h [new file with mode: 0644]
palacios/src/palacios/vmm_shadow_paging_32pae.h [new file with mode: 0644]
palacios/src/palacios/vmm_shadow_paging_64.h [new file with mode: 0644]
palacios/src/palacios/vmm_xed.c
palacios/src/test/instr_emulator.c [new file with mode: 0644]

index 27d9909..a217d7d 100644 (file)
@@ -1,4 +1,4 @@
-# Makefile for GeekOS kernel, userspace, and tools
+# Top level Makefile for V3Vee
 #
 #  Northwestern University 
 # (c) 2008, Jack Lange <jarusl@cs.northwestern.edu>
@@ -6,38 +6,17 @@
 # (c) 2008, Lei Xia <xiaxlei@gmail.com>
 # (c) 2008, The V3VEE Project <http://www.v3vee.org> 
 #
-# Based on GeekOS Makefile:
-# Copyright (c) 2004,2005 David H. Hovemeyer <daveho@cs.umd.edu>
-# $Revision: 1.71 $
-
-
-# This is free software.  You are permitted to use,
-# redistribute, and modify it as specified in the file "COPYING".
-
-# Required software to build GeekOS:
+# Required software to build V3Vee:
 # - GNU Make (http://www.gnu.org/software/make)
-# - gcc 2.95.2 generating code for target (i386/ELF) and host platforms
 # - nasm (http://nasm.sourceforge.net)
 # - Perl5, AWK (any version), egrep
 #
-# Cygwin (http://cygwin.com) may be used to build GeekOS.
-# Make sure that gcc, binutils, nasm, and perl are installed.
 
-# NOTES:
-# - This makefile has been written carefully to work correctly
-#   with the -j (parallel make) option.  I regularly use "make -j 2"
-#   to speed the build process on 2 processor systems.
 
 
 
 
-# ----------------------------------------------------------------------
-# Configuration -
-#   Various options specifying how GeekOS should be built,
-#   what source files to build, which user programs to build,
-#   etc.  This is generally the only section of the makefile
-#   that will need to be modified.
-# ----------------------------------------------------------------------
+
 PROJECT_ROOT := ..
 PALACIOS_BUILD_DIR := $(PROJECT_ROOT)/palacios/build
 GEEKOS_BUILD_DIR := $(PROJECT_ROOT)/geekos/build
@@ -177,6 +156,11 @@ endif
 endif
 
 
+ifeq ($(PROFILE_VMM),1)
+  GEEKOS_FLAGS:= $(GEEKOS_FLAGS) PROFILE_VMM=1
+endif
+
+
 # ----------------------------------------------------------------------
 # Targets -
 #   Specifies files to be built
@@ -208,14 +192,14 @@ geekos: palacios32
        cp $(PALACIOS_BUILD_DIR)/libv3vee.a $(GEEKOS_BUILD_DIR)/palacios/
        cp $(PALACIOS_BUILD_DIR)/../lib/xed/libxed.a $(GEEKOS_BUILD_DIR)/palacios/
        cp $(PALACIOS_BUILD_DIR)/vm_kernel $(GEEKOS_BUILD_DIR)/palacios/
-       (cd $(GEEKOS_BUILD_DIR) && make)
+       (cd $(GEEKOS_BUILD_DIR) && make $(GEEKOS_FLAGS))
 
 
 geekos-full: palacios-full32
        cp $(PALACIOS_BUILD_DIR)/libv3vee.a $(GEEKOS_BUILD_DIR)/palacios/
        cp $(PALACIOS_BUILD_DIR)/../lib/xed/libxed.a $(GEEKOS_BUILD_DIR)/palacios/
        cp $(PALACIOS_BUILD_DIR)/vm_kernel $(GEEKOS_BUILD_DIR)/palacios/
-       (cd $(GEEKOS_BUILD_DIR) && make clean && make)
+       (cd $(GEEKOS_BUILD_DIR) && make clean && make $(GEEKOS_FLAGS))
 
 
 world: geekos-full palacios-full64
index c62d482..352199c 100644 (file)
@@ -51,7 +51,6 @@ VPATH := $(PROJECT_ROOT)/src
 
 
 
-
 #
 #uIP, ON -- used, OFF -- not used
 #
@@ -79,6 +78,9 @@ LWIP=OFF
 #endif
 
 
+ifeq ($(PROFILE_VMM),1)
+  EXTRA_C_OPTS:= -DPROFILE_VMM
+endif
 
 
 # ----------------------------------------------------------------------
index c391c63..da04444 100644 (file)
@@ -106,7 +106,13 @@ int RunVMM(struct Boot_Info * bootInfo) {
     vm_config.rombios_size = rombios->length;
     
     region_start += rombios->length;
-    
+
+#ifdef PROFILE_VMM
+    vm_config.enable_profiling = 1;
+#else
+    vm_config.enable_profiling = 0;
+#endif
+
     vm_config.vgabios = region_start;
     vm_config.vgabios_size = vgabios->length;
   }
index 604c266..1b2cf5d 100644 (file)
@@ -53,11 +53,11 @@ VPATH := $(PROJECT_ROOT)/src
 # 
 
 ifeq ($(LEAN_AND_MEAN),1) 
-DEBUG=0
-DEBUG_SECTIONS=
+DEBUG := 0
+DEBUG_SECTIONS :=
 else
-DEBUG=1
-DEBUG_SECTIONS= 
+DEBUG := 1
+DEBUG_SECTIONS := 
 endif
 
 
@@ -265,6 +265,9 @@ VMM_OBJS := \
        palacios/vmm_msr.o \
        palacios/svm_msr.o \
        palacios/vmm_socket.o \
+       palacios/vmm_xed.o \
+       palacios/vmm_rbtree.o \
+       palacios/vmm_profiler.o \
 
 #              vmx.c vmcs_gen.c vmcs.c
 
@@ -278,8 +281,8 @@ XED_OBJS := \
        xed/v3-xed-compat.o \
        xed/v3-udiv-compat.o \
 
-$(XED_OBJS) :: EXTRA_CFLAGS =
-
+$(XED_OBJS) :: EXTRA_CFLAGS = \
+       $(JRLDEBUG) \
 
 DEVICES_OBJS := \
        devices/generic.o \
@@ -294,19 +297,8 @@ DEVICES_OBJS := \
        devices/cdrom.o \
        devices/bochs_debug.o \
 
-$(DEVICES_OBJS) :: EXTRA_CFLAGS =
-
-#
-# DECODER is the decoder that will be used 
-# currently we only support xed
-#
-DECODER=XED
-
-ifeq ($(DECODER),XED)
-VMM_OBJS += palacios/vmm_xed.o
-else
-# This is an error
-endif
+$(DEVICES_OBJS) :: EXTRA_CFLAGS = \
+       $(JRLDEBUG) \
 
 
 
@@ -399,11 +391,12 @@ CC_GENERAL_OPTS = \
        $(EXTRA_C_OPTS) \
        $(VMM_FLAGS) \
        -I$(PROJECT_ROOT)/include \
-       -fPIC \
        -Werror \
+       -fPIC \
        -Wp,-MD,$(@D)/.$(@F).d \
        -Wp,-MT,$@ \
 
+#-fPIC \
 #-fvisibility=hidden
 
 
index e2425ac..8f9dc93 100644 (file)
 
 
 int v3_handle_svm_exit(struct guest_info * info);
+const char * vmexit_code_to_str(uint_t exit_code);
+
 
 #endif // ! __V3VEE__
 
index e1a9199..74e986e 100644 (file)
@@ -32,6 +32,7 @@
 #include <palacios/vmm_emulator.h>
 #include <palacios/vmm_host_events.h>
 #include <palacios/vmm_msr.h>
+#include <palacios/vmm_profiler.h>
 
 
 
@@ -80,7 +81,7 @@ struct v3_segment {
   uint_t long_mode      : 1;
   uint_t db             : 1;
   uint_t granularity    : 1;
-};
+} __attribute__((packed));
 
 
 struct v3_segments {
@@ -94,14 +95,12 @@ struct v3_segments {
   struct v3_segment gdtr;
   struct v3_segment idtr;
   struct v3_segment tr;
-};
+} ;
 
 struct shadow_page_state;
-struct shadow_map;
-struct vmm_io_map;
 struct emulation_state;
 struct v3_intr_state;
-
+struct v3_profiler;
 
 
 
@@ -112,7 +111,8 @@ struct guest_info {
   uint_t cpl;
 
   addr_t mem_size; // Probably in bytes for now....
-  struct shadow_map mem_map;
+  v3_shdw_map_t mem_map;
+
 
   struct vm_time time_state;
 
@@ -126,7 +126,7 @@ struct guest_info {
   // This structure is how we get interrupts for the guest
   struct v3_intr_state intr_state;
 
-  struct vmm_io_map io_map;
+  v3_io_map_t io_map;
 
   struct v3_msr_map msr_map;
   // device_map
@@ -144,12 +144,15 @@ struct guest_info {
   struct v3_dbg_regs dbg_regs;
   struct v3_segments segments;
 
-  struct emulation_state emulator;
-
   v3_vm_operating_mode_t run_state;
   void * vmm_data;
 
 
+  uint_t enable_profiler;
+  struct v3_profiler profiler;
+
+  void * decoder_state;
+
   struct v3_msr guest_efer;
 
   /* TEMP */
index ed85de5..968873a 100644 (file)
@@ -324,7 +324,7 @@ struct VMCB_State_Save_Area {
 
   //offset 0x0cb
   uchar_t cpl; // if the guest is real-mode then the CPL is forced to 0
-                                             // if the guest is virtual-mode then the CPL is forced to 3
+               // if the guest is virtual-mode then the CPL is forced to 3
 
   uint_t rsvd2;
 
index b164213..9fae52e 100644 (file)
@@ -35,7 +35,7 @@
 
 /* utility definitions */
 
-#ifdef VMM_DEBUG
+
 #define PrintDebug(fmt, args...)                       \
   do {                                                 \
     extern struct v3_os_hooks * os_hooks;              \
@@ -43,6 +43,8 @@
       (os_hooks)->print_debug((fmt), ##args);          \
     }                                                  \
   } while (0)                                          
+
+#if 1
 #else
 #define PrintDebug(fmt,args ...)
 #endif
@@ -256,6 +258,9 @@ struct v3_vm_config {
                           // so we can specify maximum physical address size
                           // (We're screwed if we want to do 32 bit host/64 bit guest)
 
+
+  int enable_profiling;
+
   int use_ramdisk;
   void * ramdisk;
   int ramdisk_size;
index cfe5ff7..6aa0833 100644 (file)
@@ -1,4 +1,3 @@
-
 /*
  * This file is part of the Palacios Virtual Machine Monitor developed
  * by the V3VEE Project with funding from the United States National 
 #include <palacios/vmm.h>
 
 
+typedef enum { V3_INVALID_OP,
+              V3_OP_MOVCR2, V3_OP_MOV2CR, V3_OP_SMSW, V3_OP_LMSW, V3_OP_CLTS,
+              V3_OP_INVLPG,
+              V3_OP_ADC, V3_OP_ADD, V3_OP_AND, V3_OP_OR, V3_OP_XOR, V3_OP_SUB,
+              V3_OP_INC, V3_OP_DEC, V3_OP_NEG, V3_OP_MOV, V3_OP_NOT, V3_OP_XCHG, 
+              V3_OP_SETB, V3_OP_SETBE, V3_OP_SETL, V3_OP_SETLE, V3_OP_SETNB, 
+              V3_OP_SETNBE, V3_OP_SETNL, V3_OP_SETNLE, V3_OP_SETNO, V3_OP_SETNP,
+              V3_OP_SETNS, V3_OP_SETNZ, V3_OP_SETO, V3_OP_SETP, V3_OP_SETS, 
+              V3_OP_SETZ, V3_OP_MOVS} v3_op_type_t;
+
+
 typedef enum {INVALID_OPERAND, REG_OPERAND, MEM_OPERAND, IMM_OPERAND} v3_operand_type_t;
 
 struct x86_operand {
@@ -49,7 +59,7 @@ struct x86_prefixes {
   uint_t fs_override : 1;  // 0x64
   uint_t gs_override : 1;  // 0x65
   uint_t br_not_taken : 1;  // 0x2E
-  uint_t br_takend   : 1;  // 0x3E
+  uint_t br_taken   : 1;  // 0x3E
   uint_t op_size     : 1;  // 0x66
   uint_t addr_size   : 1;  // 0x67
 };
@@ -58,11 +68,13 @@ struct x86_prefixes {
 struct x86_instr {
   struct x86_prefixes prefixes;
   uint_t instr_length;
-  addr_t opcode;    // a pointer to the V3_OPCODE_[*] arrays defined below
+  v3_op_type_t op_type;
   uint_t num_operands;
   struct x86_operand dst_operand;
   struct x86_operand src_operand;
   struct x86_operand third_operand;
+  addr_t str_op_length;
+  addr_t is_str_op;
   void * decoder_data;
 };
 
@@ -119,7 +131,7 @@ int v3_basic_mem_decode(struct guest_info * info, addr_t instr_ptr, struct basic
 
 /* Removes a rep prefix in place */
 void v3_strip_rep_prefix(uchar_t * instr, int length);
-
+void v3_get_prefixes(uchar_t * instr, struct x86_prefixes * prefixes);
 
 
 /* 
@@ -476,7 +488,7 @@ static inline v3_operand_type_t decode_operands32(struct v3_gprs * gprs, // inpu
       base_addr = gprs->rax;
       break;
     case 1:
-      base_addr = gprs->rcx;
+       base_addr = gprs->rcx;
       break;
     case 2:
       base_addr = gprs->rdx;
index 45cfa44..86c85c8 100644 (file)
 
 #ifdef __V3VEE__
 
-#include <palacios/vmm_list.h>
 #include <palacios/vmm_shadow_paging.h>
 #include <palacios/vmm_paging.h>
 
 
 
 
-
-struct emulated_page {
-  addr_t page_addr;
-  addr_t va;
-  pte32_t pte;
-  struct list_head page_list;
-};
-
-struct saved_page {
-  addr_t va;
-  pte32_t pte;
-  struct list_head page_list;
-};
-
-
-struct write_region {
-  void * write_data;
-  
-  uint_t length;
-  int (*write)(addr_t write_addr, void * src, uint_t length, void * priv_data);
-  addr_t write_addr;
-  void * private_data;
-  
-  struct list_head write_list;
-};
-
-
-struct emulation_state {
-  uint_t num_emulated_pages;
-  struct list_head emulated_pages;
-
-  uint_t num_saved_pages;
-  struct list_head saved_pages;
-
-  uint_t num_write_regions;
-  struct list_head write_regions;
-
-  uint_t running : 1;
-  uint_t instr_length;
-
-  uint_t tf_enabled : 1;
-};
-
-
-int v3_init_emulator(struct guest_info * info);
-
-
-int v3_emulation_exit_handler(struct guest_info * info);
-
-int v3_emulate_memory_write(struct guest_info * info, addr_t fault_gva,
-                           int (*write)(addr_t write_addr, void * src, uint_t length, void * priv_data), 
-                           addr_t write_addr, void * private_data);
-int v3_emulate_memory_read(struct guest_info * info, addr_t fault_gva, 
-                          int (*read)(addr_t read_addr, void * dst, uint_t length, void * priv_data), 
-                          addr_t read_addr, void * private_data);
+int v3_emulate_write_op(struct guest_info * info, addr_t fault_gva, addr_t write_gpa, addr_t * dst_addr);
 
 
 #endif // !__V3VEE__
index a51c53a..5acb89e 100644 (file)
@@ -104,18 +104,18 @@ ulong_t hash_buffer(uchar_t * msg, uint_t length);
 
 
 #define DEFINE_HASHTABLE_INSERT(fnname, keytype, valuetype)            \
-  int fnname (struct hashtable * htable, keytype key, valuetype value) { \
-    return hashtable_insert(htable, (addr_t)key, (addr_t)value);               \
+  static int fnname (struct hashtable * htable, keytype key, valuetype value) { \
+    return hashtable_insert(htable, (addr_t)key, (addr_t)value);       \
   }
 
 #define DEFINE_HASHTABLE_SEARCH(fnname, keytype, valuetype)            \
-  valuetype * fnname (struct hashtable * htable, keytype  key) {       \
-    return (valuetype *) (hashtable_search(htable, (addr_t)key));              \
+  static valuetype * fnname (struct hashtable * htable, keytype  key) {        \
+    return (valuetype *) (hashtable_search(htable, (addr_t)key));      \
   }
 
 #define DEFINE_HASHTABLE_REMOVE(fnname, keytype, valuetype, free_key)  \
-  valuetype * fnname (struct hashtable * htable, keytype key) {        \
-    return (valuetype *) (hashtable_remove(htable, (addr_t)key, free_key));    \
+  static valuetype * fnname (struct hashtable * htable, keytype key) { \
+    return (valuetype *) (hashtable_remove(htable, (addr_t)key, free_key)); \
   }
 
 
@@ -154,6 +154,11 @@ addr_t hashtable_remove(struct hashtable * htable, addr_t key, int free_key);
 
 uint_t hashtable_count(struct hashtable * htable);
 
+// Specialty functions for a counting hashtable 
+int hashtable_inc(struct hashtable * htable, addr_t key, addr_t value);
+int hashtable_dec(struct hashtable * htable, addr_t key, addr_t value);
+
+
   /* ************ */
  /* ITERATOR API */
 /* ************ */
diff --git a/palacios/include/palacios/vmm_instr_emulator.h b/palacios/include/palacios/vmm_instr_emulator.h
new file mode 100644 (file)
index 0000000..a971805
--- /dev/null
@@ -0,0 +1,442 @@
+/*
+ * This file is part of the Palacios Virtual Machine Monitor developed
+ * by the V3VEE Project with funding from the United States National 
+ * Science Foundation and the Department of Energy.  
+ *
+ * The V3VEE Project is a joint project between Northwestern University
+ * and the University of New Mexico.  You can find out more at 
+ * http://www.v3vee.org
+ *
+ * Copyright (c) 2008, Jack Lange <jarusl@cs.northwestern.edu> 
+ * Copyright (c) 2008, The V3VEE Project <http://www.v3vee.org> 
+ * All rights reserved.
+ *
+ * Author: Jack Lange <jarusl@cs.northwestern.edu>
+ *
+ * This is free software.  You are permitted to use,
+ * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
+ */
+
+#include <palacios/vmm_types.h>
+
+
+
+#define MAKE_1OP_8FLAGS_INST(iname) static inline void iname##8(addr_t * dst,  addr_t * flags) { \
+    uchar_t tmp_dst = *dst;                                            \
+                                                                       \
+    /* Some of the flags values are not copied out in a pushf, we save them here */ \
+    addr_t flags_rsvd = *flags & ~0xfffe7fff;                          \
+                                                                       \
+    asm volatile (                                                             \
+        "pushf; "                                                      \
+        "push %2; "                                                    \
+        "popf; "                                                       \
+        #iname"b %0; "                                                 \
+        "pushf; "                                                      \
+        "pop %1; "                                                     \
+        "popf; "                                                       \
+        : "=q"(tmp_dst),"=q"(*flags)                                   \
+        : "q"(*flags), "0"(tmp_dst)                                    \
+        );                                                             \
+    *dst = tmp_dst;                                                    \
+    *flags |= flags_rsvd;                                              \
+                                                                       \
+  }
+
+#define MAKE_1OP_16FLAGS_INST(iname) static inline void iname##16(addr_t * dst,  addr_t * flags) { \
+    ushort_t tmp_dst = *dst;                                           \
+                                                                       \
+    /* Some of the flags values are not copied out in a pushf, we save them here */ \
+    addr_t flags_rsvd = *flags & ~0xfffe7fff;                          \
+                                                                       \
+    asm volatile (                                                             \
+        "pushf; "                                                      \
+        "push %2; "                                                    \
+        "popf; "                                                       \
+        #iname"w %0; "                                                 \
+        "pushf; "                                                      \
+        "pop %1; "                                                     \
+        "popf; "                                                       \
+        : "=q"(tmp_dst),"=q"(*flags)                                   \
+        : "q"(*flags), "0"(tmp_dst)                                    \
+        );                                                             \
+    *dst = tmp_dst;                                                    \
+    *flags |= flags_rsvd;                                              \
+                                                                       \
+  }
+
+#define MAKE_1OP_32FLAGS_INST(iname) static inline void iname##32(addr_t * dst,  addr_t * flags) { \
+    uint_t tmp_dst = *dst;                                             \
+                                                                       \
+    /* Some of the flags values are not copied out in a pushf, we save them here */ \
+    addr_t flags_rsvd = *flags & ~0xfffe7fff;                          \
+                                                                       \
+    asm volatile (                                                             \
+        "pushf; "                                                      \
+        "push %2; "                                                    \
+        "popf; "                                                       \
+        #iname"l %0; "                                                 \
+        "pushf; "                                                      \
+        "pop %1; "                                                     \
+        "popf; "                                                       \
+        : "=q"(tmp_dst),"=q"(*flags)                                   \
+        : "q"(*flags), "0"(tmp_dst)                                    \
+        );                                                             \
+    *dst = tmp_dst;                                                    \
+    *flags |= flags_rsvd;                                              \
+                                                                       \
+  }
+
+#define MAKE_1OP_64FLAGS_INST(iname) static inline void iname##64(addr_t * dst,  addr_t * flags) { \
+    ullong_t tmp_dst = *dst;                                           \
+                                                                       \
+    /* Some of the flags values are not copied out in a pushf, we save them here */ \
+    addr_t flags_rsvd = *flags & ~0xfffe7fff;                          \
+                                                                       \
+    asm volatile (                                                             \
+        "pushfq; "                                                     \
+        "push %2; "                                                    \
+        "popfq; "                                                      \
+        #iname"q %0; "                                                 \
+        "pushfq; "                                                     \
+        "pop %1; "                                                     \
+        "popfq; "                                                      \
+        : "=q"(tmp_dst),"=q"(*flags)                                   \
+        : "q"(*flags), "0"(tmp_dst)                                    \
+        );                                                             \
+    *dst = tmp_dst;                                                    \
+    *flags |= flags_rsvd;                                              \
+                                                                       \
+  }
+
+
+
+#define MAKE_1OP_8_INST(iname) static inline void iname##8(addr_t * dst) { \
+    uchar_t tmp_dst = *dst;                                            \
+                                                                       \
+    asm volatile (                                                     \
+                 #iname"b %0; "                                        \
+                 : "=q"(tmp_dst)                                       \
+                 : "0"(tmp_dst)                                        \
+                 );                                                    \
+    *dst = tmp_dst;                                                    \
+  }
+
+#define MAKE_1OP_16_INST(iname) static inline void iname##16(addr_t * dst) { \
+    ushort_t tmp_dst = *dst;                                           \
+                                                                       \
+    asm volatile (                                                     \
+        #iname"w %0; "                                                 \
+        : "=q"(tmp_dst)                                                \
+        :  "0"(tmp_dst)                                                \
+        );                                                             \
+    *dst = tmp_dst;                                                    \
+  }
+
+#define MAKE_1OP_32_INST(iname) static inline void iname##32(addr_t * dst) { \
+    uint_t tmp_dst = *dst;                                             \
+                                                                       \
+    asm volatile (                                                     \
+        #iname"l %0; "                                                 \
+        : "=q"(tmp_dst)                                                \
+        : "0"(tmp_dst)                                                 \
+        );                                                             \
+    *dst = tmp_dst;                                                    \
+  }
+
+#define MAKE_1OP_64_INST(iname) static inline void iname##64(addr_t * dst) { \
+    ullong_t tmp_dst = *dst;                                           \
+                                                                       \
+    asm volatile (                                                     \
+                 #iname"q %0; "                                        \
+                 : "=q"(tmp_dst)                                       \
+                 : "0"(tmp_dst)                                        \
+                 );                                                    \
+    *dst = tmp_dst;                                                    \
+  }
+
+
+#define MAKE_2OP_64FLAGS_INST(iname) static inline void iname##64(addr_t * dst, addr_t * src, addr_t * flags) { \
+    uint64_t tmp_dst = *dst, tmp_src = *src;                                   \
+    addr_t tmp_flags = *flags;                                         \
+                                                                       \
+    /* Some of the flags values are not copied out in a pushf, we save them here */ \
+    addr_t flags_rsvd = *flags & ~0xfffe7fff;                          \
+                                                                       \
+    asm volatile (                                                             \
+        "pushfq\r\n"                                                   \
+        "push %3\r\n"                                                  \
+        "popfq\r\n"                                                    \
+        #iname"q %2, %0\r\n"                                           \
+        "pushfq\r\n"                                                   \
+        "pop %1\r\n"                                                   \
+        "popfq\r\n"                                                    \
+        : "=q"(tmp_dst),"=q"(tmp_flags)                                \
+        : "q"(tmp_src),"q"(tmp_flags), "0"(tmp_dst)                    \
+        );                                                             \
+                                                                       \
+    *dst = tmp_dst;                                                    \
+    *flags = tmp_flags;                                                        \
+    *flags |= flags_rsvd;                                              \
+                                                                       \
+  }
+
+
+
+
+#define MAKE_2OP_32FLAGS_INST(iname) static inline void iname##32(addr_t * dst, addr_t * src, addr_t * flags) { \
+    uint32_t tmp_dst = *dst, tmp_src = *src;                           \
+                                                                       \
+    /* Some of the flags values are not copied out in a pushf, we save them here */ \
+    addr_t flags_rsvd = *flags & ~0xfffe7fff;                          \
+                                                                       \
+    asm volatile (                                                     \
+        "pushf; "                                                      \
+        "push %3; "                                                    \
+        "popf; "                                                       \
+        #iname"l %2, %0; "                                             \
+        "pushf; "                                                      \
+        "pop %1; "                                                     \
+        "popf; "                                                       \
+        : "=q"(tmp_dst),"=q"(*flags)                                   \
+        : "q"(tmp_src),"q"(*flags), "0"(tmp_dst)                       \
+        );                                                             \
+    *dst = tmp_dst;                                                    \
+    *flags |= flags_rsvd;                                              \
+                                                                       \
+  }
+
+
+#define MAKE_2OP_16FLAGS_INST(iname) static inline void iname##16(addr_t * dst, addr_t * src, addr_t * flags) { \
+    ushort_t tmp_dst = *dst, tmp_src = *src;                           \
+                                                                       \
+    /* Some of the flags values are not copied out in a pushf, we save them here */ \
+    addr_t flags_rsvd = *flags & ~0xfffe7fff;                          \
+                                                                       \
+    asm volatile (                                                     \
+        "pushf; "                                                      \
+        "push %3; "                                                    \
+        "popf; "                                                       \
+        #iname"w %2, %0; "                                             \
+        "pushf; "                                                      \
+        "pop %1; "                                                     \
+        "popf; "                                                       \
+        : "=q"(tmp_dst),"=q"(*flags)                                   \
+        : "q"(tmp_src),"q"(*flags), "0"(tmp_dst)                       \
+        );                                                             \
+    *dst = tmp_dst;                                                    \
+    *flags |= flags_rsvd;                                              \
+                                                                       \
+  }
+
+#define MAKE_2OP_8FLAGS_INST(iname) static inline void iname##8(addr_t * dst, addr_t * src, addr_t * flags) { \
+    uchar_t tmp_dst = *dst, tmp_src = *src;                            \
+                                                                       \
+    /* Some of the flags values are not copied out in a pushf, we save them here */ \
+    addr_t flags_rsvd = *flags & ~0xfffe7fff;                          \
+                                                                       \
+    asm volatile (                                                     \
+        "pushf; "                                                      \
+        "push %3; "                                                    \
+        "popf; "                                                       \
+        #iname"b %2, %0; "                                             \
+        "pushf; "                                                      \
+        "pop %1; "                                                     \
+        "popf; "                                                       \
+        : "=q"(tmp_dst),"=q"(*flags)                                   \
+        : "q"(tmp_src),"q"(*flags), "0"(tmp_dst)                       \
+        );                                                             \
+    *dst = tmp_dst;                                                    \
+    *flags |= flags_rsvd;                                              \
+                                                                       \
+  }
+
+
+
+
+#define MAKE_2OP_32STR_INST(iname) static inline void iname##32(addr_t * dst, \
+                                                               addr_t * src, \
+                                                               addr_t * ecx, addr_t * flags) { \
+    /* Some of the flags values are not copied out in a pushf, we save them here */ \
+    addr_t flags_rsvd = *flags & ~0xfffe7fff;                          \
+                                                                       \
+    asm volatile (                                                     \
+        "pushf; "                                                      \
+        "push %4; "                                                    \
+        "popf; "                                                       \
+        "rep; "                                                        \
+        #iname"l; "                                                    \
+        "pushf; "                                                      \
+        "pop %0; "                                                     \
+        "popf; "                                                       \
+        : "=q"(*flags)                                                 \
+        : "D"(*dst),"S"(*src),"c"(*ecx),"q"(*flags)                    \
+        );                                                             \
+                                                                       \
+    /*  : "=D"(*dst),"=S"(*src),"=c"(*ecx),"=q"(*flags)*/              \
+    *flags |= flags_rsvd;                                              \
+  }
+
+#define MAKE_2OP_16STR_INST(iname) static inline void iname##16(addr_t * dst, \
+                                                               addr_t * src, \
+                                                               addr_t * ecx, addr_t * flags) { \
+     /* Some of the flags values are not copied out in a pushf, we save them here */ \
+    addr_t flags_rsvd = *flags & ~0xfffe7fff;                          \
+                                                                       \
+    asm volatile (                                                     \
+        "pushf; "                                                      \
+        "push %4; "                                                    \
+        "popf; "                                                       \
+        "rep; "                                                        \
+        #iname"w; "                                                    \
+        "pushf; "                                                      \
+        "pop %0; "                                                     \
+        "popf; "                                                       \
+        : "=q"(*flags)                                                 \
+        : "D"(*dst),"S"(*src),"c"(*ecx),"q"(*flags)                    \
+        );                                                             \
+    *flags |= flags_rsvd;                                              \
+  }
+
+
+
+#define MAKE_2OP_8STR_INST(iname) static inline void iname##8(addr_t * dst, \
+                                                             addr_t * src, \
+                                                             addr_t * ecx, addr_t * flags) { \
+    /* Some of the flags values are not copied out in a pushf, we save them here */ \
+    addr_t flags_rsvd = *flags & ~0xfffe7fff;                          \
+                                                                       \
+    asm volatile (                                                     \
+        "pushf; "                                                      \
+        "push %4; "                                                    \
+        "popf; "                                                       \
+        "rep; "                                                        \
+        #iname"b; "                                                    \
+        "pushf; "                                                      \
+        "pop %0; "                                                     \
+        "popf; "                                                       \
+        : "=q"(*flags)                                                 \
+        : "D"(*dst),"S"(*src),"c"(*ecx),"q"(*flags)                    \
+        );                                                             \
+    *flags |= flags_rsvd;                                              \
+  }
+
+
+
+
+#define MAKE_2OP_32_INST(iname) static inline void iname##32(addr_t * dst, addr_t * src) { \
+    uint32_t tmp_dst = *dst, tmp_src = *src;                           \
+                                                                       \
+    asm volatile (                                                     \
+        #iname"l %1, %0; "                                             \
+        : "=q"(tmp_dst)                                                \
+        : "q"(tmp_src), "0"(tmp_dst)                                   \
+        );                                                             \
+    *dst = tmp_dst;                                                    \
+  }
+
+#define MAKE_2OP_16_INST(iname) static inline void iname##16(addr_t * dst, addr_t * src) { \
+    ushort_t tmp_dst = *dst, tmp_src = *src;                           \
+                                                                       \
+    asm volatile (                                                     \
+        #iname"w %1, %0; "                                             \
+        : "=q"(tmp_dst)                                                \
+        : "q"(tmp_src), "0"(tmp_dst)                                   \
+        );                                                             \
+    *dst = tmp_dst;                                                    \
+  }
+
+#define MAKE_2OP_8_INST(iname) static inline void iname##8(addr_t * dst, addr_t * src) { \
+    uchar_t tmp_dst = *dst, tmp_src = *src;                            \
+                                                                       \
+    asm volatile (                                                     \
+        #iname"b %1, %0; "                                             \
+        : "=q"(tmp_dst)                                                \
+        : "q"(tmp_src), "0"(tmp_dst)                                   \
+        );                                                             \
+    *dst = tmp_dst;                                                    \
+  }
+
+
+
+
+
+
+
+MAKE_2OP_8FLAGS_INST(adc);
+MAKE_2OP_8FLAGS_INST(add);
+MAKE_2OP_8FLAGS_INST(and);
+MAKE_2OP_8FLAGS_INST(or);
+MAKE_2OP_8FLAGS_INST(xor);
+MAKE_2OP_8FLAGS_INST(sub);
+
+
+MAKE_1OP_8FLAGS_INST(inc);
+MAKE_1OP_8FLAGS_INST(dec);
+MAKE_1OP_8FLAGS_INST(neg);
+MAKE_1OP_8FLAGS_INST(setb);
+MAKE_1OP_8FLAGS_INST(setbe);
+MAKE_1OP_8FLAGS_INST(setl);
+MAKE_1OP_8FLAGS_INST(setle);
+MAKE_1OP_8FLAGS_INST(setnb);
+MAKE_1OP_8FLAGS_INST(setnbe);
+MAKE_1OP_8FLAGS_INST(setnl);
+MAKE_1OP_8FLAGS_INST(setnle);
+MAKE_1OP_8FLAGS_INST(setno);
+MAKE_1OP_8FLAGS_INST(setnp);
+MAKE_1OP_8FLAGS_INST(setns);
+MAKE_1OP_8FLAGS_INST(setnz);
+MAKE_1OP_8FLAGS_INST(seto);
+MAKE_1OP_8FLAGS_INST(setp);
+MAKE_1OP_8FLAGS_INST(sets);
+MAKE_1OP_8FLAGS_INST(setz);
+
+
+MAKE_1OP_8_INST(not);
+
+MAKE_2OP_8_INST(mov);
+MAKE_2OP_8_INST(xchg);
+
+
+
+MAKE_2OP_16FLAGS_INST(adc);
+MAKE_2OP_16FLAGS_INST(add);
+MAKE_2OP_16FLAGS_INST(and);
+MAKE_2OP_16FLAGS_INST(or);
+MAKE_2OP_16FLAGS_INST(xor);
+MAKE_2OP_16FLAGS_INST(sub);
+
+
+MAKE_1OP_16FLAGS_INST(inc);
+MAKE_1OP_16FLAGS_INST(dec);
+MAKE_1OP_16FLAGS_INST(neg);
+
+MAKE_1OP_16_INST(not);
+
+MAKE_2OP_16_INST(mov);
+MAKE_2OP_16_INST(xchg);
+
+
+
+
+
+MAKE_2OP_32FLAGS_INST(adc);
+MAKE_2OP_32FLAGS_INST(add);
+MAKE_2OP_32FLAGS_INST(and);
+MAKE_2OP_32FLAGS_INST(or);
+MAKE_2OP_32FLAGS_INST(xor);
+MAKE_2OP_32FLAGS_INST(sub);
+
+
+MAKE_1OP_32FLAGS_INST(inc);
+MAKE_1OP_32FLAGS_INST(dec);
+MAKE_1OP_32FLAGS_INST(neg);
+
+MAKE_1OP_32_INST(not);
+
+MAKE_2OP_32_INST(mov);
+MAKE_2OP_32_INST(xchg);
+
+MAKE_2OP_8STR_INST(movs);
+MAKE_2OP_16STR_INST(movs);
+MAKE_2OP_32STR_INST(movs);
index 9bfc197..8db4f71 100644 (file)
 
 #include <palacios/vmm_types.h>
 #include <palacios/vmm_util.h>
+#include <palacios/vmm_rbtree.h>
+
+typedef struct rb_root v3_io_map_t;
 
 
 struct guest_info;
 
-int v3_unhook_io_port(struct guest_info * info, uint_t port);
+void v3_init_io_map(struct guest_info * info);
+
+
 
 
 /* External API */
@@ -38,27 +43,12 @@ int v3_hook_io_port(struct guest_info * info, uint_t port,
                    int (*write)(ushort_t port, void * src, uint_t length, void * priv_data), 
                    void * priv_data);
 
+int v3_unhook_io_port(struct guest_info * info, uint_t port);
 
 
 
 
-
-struct vmm_io_hook;
-
-struct vmm_io_map {
-  uint_t num_ports;
-  struct vmm_io_hook * head;
-
-};
-
-
-void v3_init_vmm_io_map(struct guest_info * info);
-
-// FOREACH_IO_HOOK(vmm_io_map_t * io_map, vmm_io_hook_t * io_hook)
-#define FOREACH_IO_HOOK(io_map, io_hook) for (io_hook = (io_map).head; io_hook != NULL; io_hook = (io_hook)->next)
-
-
-struct vmm_io_hook {
+struct v3_io_hook {
   ushort_t port;
 
   // Reads data into the IO port (IN, INS)
@@ -68,17 +58,16 @@ struct vmm_io_hook {
   int (*write)(ushort_t port, void * src, uint_t length, void * priv_data);
 
   void * priv_data;
-
-  struct vmm_io_hook * next;
-  struct vmm_io_hook * prev;
+  
+  struct rb_node tree_node;
 
 };
 
 
-struct vmm_io_hook * v3_get_io_hook(struct vmm_io_map * io_map, uint_t port);
+struct v3_io_hook * v3_get_io_hook(struct guest_info * info, uint_t port);
 
 
-void v3_print_io_map(struct vmm_io_map * io_map);
+void v3_print_io_map(struct guest_info * info);
 
 
 
index 0d8747e..0d5b907 100644 (file)
@@ -233,6 +233,20 @@ static inline void list_splice_init(struct list_head *list,
        container_of(ptr, type, member)
 
 /**
+ * list_entry - get the struct for the tail entry
+ * @ptr:       the list_head head pointer.
+ * @type:      the type of the struct this is embedded in.
+ * @member:    the name of the list_struct within the struct.
+ */
+#define list_tail_entry(head, type, member) ({ \
+      type * tail = NULL;                      \
+      if ((head)->prev != (head)) {                    \
+       tail = list_entry((head)->prev, type, member);  \
+      }                                                        \
+      tail;                                            \
+})
+
+/**
  * list_for_each       -       iterate over a list
  * @pos:       the &struct list_head to use as a loop counter.
  * @head:      the head for your list.
index c21155b..ae4ca05 100644 (file)
 #include <palacios/vmm_types.h>
 
 #include <palacios/vmm_paging.h>
+#include <palacios/vmm_rbtree.h>
 
-struct guest_info;
-
-
-/*
 
-        Guest                  Shadow                 Host
-  Virtual   Physical    Virtual     Physical   Virtual     Physical
-               OK                      OK
-               OK                      NOK
-               NOK                     OK
-               NOK                     NOK
+struct guest_info;
 
-*/
 
-// These are the types of physical memory address regions
-// from the perspective of the guest
-typedef enum guest_region_type { 
-  GUEST_REGION_NOTHING, 
-  GUEST_REGION_PHYSICAL_MEMORY, 
-  GUEST_REGION_MEMORY_MAPPED_DEVICE} guest_region_type_t;
 
 // These are the types of physical memory address regions
 // from the perspective of the HOST
-typedef enum host_region_type { 
-  HOST_REGION_INVALID,                    // This region is INVALID (this is a return type, to denote errors)
-  HOST_REGION_HOOK,                       // This region is mapped as not present (always generate page faults)
-  HOST_REGION_PHYSICAL_MEMORY,            // Region is a section of host memory
-  HOST_REGION_MEMORY_MAPPED_DEVICE,       // Region is allocated for DMA
-  HOST_REGION_UNALLOCATED,                // Region is mapped on demand
-  HOST_REGION_REMOTE,                     // Region is located on a remote machine
-  HOST_REGION_SWAPPED,                    // Region is swapped
-} host_region_type_t;
+typedef enum shdw_region_type { 
+  SHDW_REGION_INVALID,                    // This region is INVALID (this is a return type to denote errors)
+  SHDW_REGION_WRITE_HOOK,                 // This region is mapped as read-only (page faults on write)
+  SHDW_REGION_FULL_HOOK,                  // This region is mapped as not present (always generate page faults)
+  SHDW_REGION_ALLOCATED,                  // Region is a section of host memory
+} v3_shdw_region_type_t;
 
 
+typedef struct rb_root v3_shdw_map_t;
 
-#define shadow_mem_type_t host_region_type_t
 
-struct shadow_region {
-  guest_region_type_t     guest_type;
+struct v3_shadow_region {
   addr_t                  guest_start; 
   addr_t                  guest_end; 
 
-  host_region_type_t      host_type;
-  addr_t                  host_addr; // This either points to a host address mapping, 
-                                     // or a structure holding the map info 
-
-  struct shadow_region *next, *prev;
-};
+  v3_shdw_region_type_t   host_type;
+  
+  addr_t                  host_addr; // This either points to a host address mapping
 
 
+  // Called when data is read from a memory page
+  int (*read_hook)(addr_t guest_addr, void * dst, uint_t length, void * priv_data);
+  // Called when data is written to a memory page
+  int (*write_hook)(addr_t guest_addr, void * src, uint_t length, void * priv_data);
 
-struct shadow_map {
-  uint_t num_regions;
+  void * priv_data;
 
-  struct shadow_region * head;
+  struct rb_node tree_node;
 };
 
 
-void init_shadow_region(struct shadow_region * entry,
-                          addr_t               guest_addr_start,
-                          addr_t               guest_addr_end,
-                          guest_region_type_t  guest_region_type,
-                          host_region_type_t   host_region_type);
 
-/*
-void init_shadow_region_physical(struct shadow_region * entry,
-                                   addr_t               guest_addr_start,
-                                   addr_t               guest_addr_end,
-                                   guest_region_type_t  guest_region_type,
-                                   addr_t               host_addr_start,
-                                   host_region_type_t   host_region_type);
-*/
+void v3_init_shadow_map(struct guest_info * info);
+void v3_delete_shadow_map(struct guest_info * info);
 
-int add_shadow_region_passthrough(struct guest_info * guest_info, 
-                                 addr_t guest_addr_start,
-                                 addr_t guest_addr_end,
-                                 addr_t host_addr);
 
-void init_shadow_map(struct guest_info * info);
-void free_shadow_map(struct shadow_map * map);
+int v3_add_shadow_mem(struct guest_info * guest_info, 
+                     addr_t guest_addr_start,
+                     addr_t guest_addr_end,
+                     addr_t host_addr);
 
-struct shadow_region * get_shadow_region_by_addr(struct shadow_map * map, addr_t guest_addr);
+int v3_hook_full_mem(struct guest_info * info, addr_t guest_addr_start, addr_t guest_addr_end,
+                    int (*read)(addr_t guest_addr, void * dst, uint_t length, void * priv_data),
+                    int (*write)(addr_t guest_addr, void * src, uint_t length, void * priv_data),
+                    void * priv_data);
 
-struct shadow_region * get_shadow_region_by_index(struct shadow_map * map, uint_t index);
+int v3_hook_write_mem(struct guest_info * info, addr_t guest_addr_start, addr_t guest_addr_end,
+                     addr_t host_addr,
+                     int (*write)(addr_t guest_addr, void * src, uint_t length, void * priv_data),
+                     void * priv_data);
 
-host_region_type_t lookup_shadow_map_addr(struct shadow_map * map, addr_t guest_addr, addr_t * host_addr);
 
-host_region_type_t get_shadow_addr_type(struct guest_info * info, addr_t guest_addr);
-addr_t get_shadow_addr(struct guest_info * info, addr_t guest_addr);
 
-// Semantics:
-// Adding a region that overlaps with an existing region results is undefined
-// and will probably fail
-int add_shadow_region(struct shadow_map * map, struct shadow_region * entry);
 
-// Semantics:
-// Deletions result in splitting
-int delete_shadow_region(struct shadow_map * map,
-                            addr_t guest_start, 
-                            addr_t guest_end);
 
 
-void print_shadow_map(struct shadow_map * map);
+void v3_delete_shadow_region(struct guest_info * info, struct v3_shadow_region * reg);
 
 
+struct v3_shadow_region * v3_get_shadow_region(struct guest_info * info, addr_t guest_addr);
+addr_t v3_get_shadow_addr(struct v3_shadow_region * reg, addr_t guest_addr);
 
-struct vmm_mem_hook {
-  // Called when data is read from a memory page
-  int (*read)(addr_t guest_addr, void * dst, uint_t length, void * priv_data);
-  
-  // Called when data is written to a memory page
-  int (*write)(addr_t guest_addr, void * src, uint_t length, void * priv_data);
 
-  void * priv_data;
-  struct shadow_region * region;
-};
 
 
 
-struct vmm_mem_hook * get_mem_hook(struct guest_info * info, addr_t guest_addr);
+void print_shadow_map(struct guest_info * info);
+
+
+
 
-int hook_guest_mem(struct guest_info * info, addr_t guest_addr_start, addr_t guest_addr_end,
-                  int (*read)(addr_t guest_addr, void * dst, uint_t length, void * priv_data),
-                  int (*write)(addr_t guest_addr, void * src, uint_t length, void * priv_data),
-                  void * priv_data);
-int unhook_guest_mem(struct guest_info * info, addr_t guest_addr);
 
+const uchar_t * v3_shdw_region_type_to_str(v3_shdw_region_type_t type);
 
 
 
 int handle_special_page_fault(struct guest_info * info, addr_t fault_addr, addr_t gp_addr, pf_error_t access_info);
 
+int v3_handle_mem_wr_hook(struct guest_info * info, addr_t guest_va, addr_t guest_pa, 
+                         struct v3_shadow_region * reg, pf_error_t access_info);
+int v3_handle_mem_full_hook(struct guest_info * info, addr_t guest_va, addr_t guest_pa, 
+                           struct v3_shadow_region * reg, pf_error_t access_info);
 
 #endif // ! __V3VEE__
 
index e7f68f1..7535a3e 100644 (file)
@@ -129,23 +129,7 @@ typedef enum {PAGE_4KB, PAGE_2MB, PAGE_4MB, PAGE_1GB,
 
 
 /* Gets the base address needed for a Page Table entry */
-/* Deprecate these :*/
-/*
-  #define PD32_BASE_ADDR(x) (((uint_t)x) >> 12)
-  #define PT32_BASE_ADDR(x) (((uint_t)x) >> 12)
-  #define PD32_4MB_BASE_ADDR(x) (((uint_t)x) >> 22)
-  
-  #define PML4E64_BASE_ADDR(x) (((ullong_t)x) >> 12)
-  #define PDPE64_BASE_ADDR(x) (((ullong_t)x) >> 12)
-  #define PDE64_BASE_ADDR(x) (((ullong_t)x) >> 12)
-  #define PTE64_BASE_ADDR(x) (((ullong_t)x) >> 12)
-  
-  // Accessor functions for the page table structures 
-  #define PDE32_T_ADDR(x) (((x).pt_base_addr) << 12)
-  #define PTE32_T_ADDR(x) (((x).page_base_addr) << 12)
-  #define PDE32_4MB_T_ADDR(x) (((x).page_base_addr) << 22)
-*/
-/* Replace The above with these... */
+
 #define PAGE_BASE_ADDR(x) ((x) >> 12)
 #define PAGE_BASE_ADDR_4KB(x) ((x) >> 12)
 #define PAGE_BASE_ADDR_2MB(x) ((x) >> 21)
@@ -159,21 +143,7 @@ typedef enum {PAGE_4KB, PAGE_2MB, PAGE_4MB, PAGE_1GB,
 #define BASE_TO_PAGE_ADDR_1GB(x) (((addr_t)x) << 30)
 /* *** */
 
-/* Deprecated */
-/*
-  #define PT32_PAGE_OFFSET(x) (((uint_t)x) & 0xfff)
-  #define PD32_4MB_PAGE_OFFSET(x) (((uint_t)x) & 0x003fffff)
-  
-  #define PT32_PAGE_ADDR(x)   (((uint_t)x) & 0xfffff000)
-  #define PD32_4MB_PAGE_ADDR(x) (((uint_t)x) & 0xffc00000)
-  
-  #define PT32_PAGE_POWER 12
-  #define PAGE_ALIGNED_ADDR(x)   (((uint_t) (x)) >> 12)
-  //#define PAGE_ADDR(x)   (PAGE_ALIGNED_ADDR(x) << 12)
-  #define PAGE_POWER 12
-  #define PAGE_SIZE 4096
-*/
-/* use these instead */
+
 #define PAGE_OFFSET(x) ((x) & 0xfff)
 #define PAGE_OFFSET_4KB(x) ((x) & 0xfff)
 #define PAGE_OFFSET_2MB(x) ((x) & 0x1fffff)
diff --git a/palacios/include/palacios/vmm_profiler.h b/palacios/include/palacios/vmm_profiler.h
new file mode 100644 (file)
index 0000000..7861a7b
--- /dev/null
@@ -0,0 +1,51 @@
+/*
+ * This file is part of the Palacios Virtual Machine Monitor developed
+ * by the V3VEE Project with funding from the United States National 
+ * Science Foundation and the Department of Energy.  
+ *
+ * The V3VEE Project is a joint project between Northwestern University
+ * and the University of New Mexico.  You can find out more at 
+ * http://www.v3vee.org
+ *
+ * Copyright (c) 2008, Jack Lange <jarusl@cs.northwestern.edu> 
+ * Copyright (c) 2008, The V3VEE Project <http://www.v3vee.org> 
+ * All rights reserved.
+ *
+ * Author: Jack Lange <jarusl@cs.northwestern.edu>
+ *
+ * This is free software.  You are permitted to use,
+ * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
+ */
+
+#ifndef __VMM_PROFILER_H__
+#define __VMM_PROFILER_H__
+
+#ifdef __V3VEE__
+
+#include <palacios/vmm_rbtree.h>
+
+struct guest_info;
+
+
+struct v3_profiler {
+  uint_t total_exits;
+
+  ullong_t start_time;
+  ullong_t end_time;
+
+  uint_t guest_pf_cnt;
+
+  struct rb_root root;
+};
+
+
+void v3_init_profiler(struct guest_info * info);
+
+void v3_profile_exit(struct guest_info * info, uint_t exit_code);
+
+void v3_print_profile(struct guest_info * info);
+
+
+#endif
+
+#endif
diff --git a/palacios/include/palacios/vmm_rbtree.h b/palacios/include/palacios/vmm_rbtree.h
new file mode 100644 (file)
index 0000000..6b72d6d
--- /dev/null
@@ -0,0 +1,175 @@
+/*
+  Red Black Trees
+  (C) 1999  Andrea Arcangeli <andrea@suse.de>
+  
+  This program is free software; you can redistribute it and/or modify
+  it under the terms of the GNU General Public License as published by
+  the Free Software Foundation; either version 2 of the License, or
+  (at your option) any later version.
+
+  This program is distributed in the hope that it will be useful,
+  but WITHOUT ANY WARRANTY; without even the implied warranty of
+  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+  GNU General Public License for more details.
+
+  You should have received a copy of the GNU General Public License
+  along with this program; if not, write to the Free Software
+  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+
+  linux/include/linux/rbtree.h
+
+  To use rbtrees you'll have to implement your own insert and search cores.
+  This will avoid us to use callbacks and to drop drammatically performances.
+  I know it's not the cleaner way,  but in C (not in C++) to get
+  performances and genericity...
+
+  Some example of insert and search follows here. The search is a plain
+  normal search over an ordered tree. The insert instead must be implemented
+  int two steps: as first thing the code must insert the element in
+  order as a red leaf in the tree, then the support library function
+  rb_insert_color() must be called. Such function will do the
+  not trivial work to rebalance the rbtree if necessary.
+
+-----------------------------------------------------------------------
+static inline struct page * rb_search_page_cache(struct inode * inode,
+                                                unsigned long offset)
+{
+       struct rb_node * n = inode->i_rb_page_cache.rb_node;
+       struct page * page;
+
+       while (n)
+       {
+               page = rb_entry(n, struct page, rb_page_cache);
+
+               if (offset < page->offset)
+                       n = n->rb_left;
+               else if (offset > page->offset)
+                       n = n->rb_right;
+               else
+                       return page;
+       }
+       return NULL;
+}
+
+static inline struct page * __rb_insert_page_cache(struct inode * inode,
+                                                  unsigned long offset,
+                                                  struct rb_node * node)
+{
+       struct rb_node ** p = &inode->i_rb_page_cache.rb_node;
+       struct rb_node * parent = NULL;
+       struct page * page;
+
+       while (*p)
+       {
+               parent = *p;
+               page = rb_entry(parent, struct page, rb_page_cache);
+
+               if (offset < page->offset)
+                       p = &(*p)->rb_left;
+               else if (offset > page->offset)
+                       p = &(*p)->rb_right;
+               else
+                       return page;
+       }
+
+       rb_link_node(node, parent, p);
+
+       return NULL;
+}
+
+static inline struct page * rb_insert_page_cache(struct inode * inode,
+                                                unsigned long offset,
+                                                struct rb_node * node)
+{
+       struct page * ret;
+       if ((ret = __rb_insert_page_cache(inode, offset, node)))
+               goto out;
+       rb_insert_color(node, &inode->i_rb_page_cache);
+ out:
+       return ret;
+}
+-----------------------------------------------------------------------
+*/
+
+#ifndef        _VMM_RBTREE_H
+#define        _VMM_RBTREE_H
+
+#ifdef __V3VEE__
+
+#include <palacios/vmm_types.h>
+
+
+#undef offsetof
+#define offsetof(TYPE, MEMBER) ((size_t) &((TYPE *)0)->MEMBER)
+
+#define container_of(ptr, type, member) ({                      \
+        const typeof( ((type *)0)->member ) *__mptr = (ptr);    \
+        (type *)( (char *)__mptr - offsetof(type,member) );})
+
+
+
+struct rb_node
+{
+       unsigned long  rb_parent_color;
+#define        RB_RED          0
+#define        RB_BLACK        1
+       struct rb_node *rb_right;
+       struct rb_node *rb_left;
+} __attribute__((aligned(sizeof(long))));
+    /* The alignment might seem pointless, but allegedly CRIS needs it */
+
+struct rb_root
+{
+       struct rb_node *rb_node;
+};
+
+
+#define rb_parent(r)   ((struct rb_node *)((r)->rb_parent_color & ~3))
+#define rb_color(r)   ((r)->rb_parent_color & 1)
+#define rb_is_red(r)   (!rb_color(r))
+#define rb_is_black(r) rb_color(r)
+#define rb_set_red(r)  do { (r)->rb_parent_color &= ~1; } while (0)
+#define rb_set_black(r)  do { (r)->rb_parent_color |= 1; } while (0)
+
+static inline void rb_set_parent(struct rb_node *rb, struct rb_node *p)
+{
+       rb->rb_parent_color = (rb->rb_parent_color & 3) | (unsigned long)p;
+}
+static inline void rb_set_color(struct rb_node *rb, int color)
+{
+       rb->rb_parent_color = (rb->rb_parent_color & ~1) | color;
+}
+
+#define RB_ROOT        (struct rb_root) { NULL, }
+#define        rb_entry(ptr, type, member) container_of(ptr, type, member)
+
+#define RB_EMPTY_ROOT(root)    ((root)->rb_node == NULL)
+#define RB_EMPTY_NODE(node)    (rb_parent(node) != node)
+#define RB_CLEAR_NODE(node)    (rb_set_parent(node, node))
+
+extern void v3_rb_insert_color(struct rb_node *, struct rb_root *);
+extern void v3_rb_erase(struct rb_node *, struct rb_root *);
+
+/* Find logical next and previous nodes in a tree */
+extern struct rb_node *v3_rb_next(struct rb_node *);
+extern struct rb_node *v3_rb_prev(struct rb_node *);
+extern struct rb_node *v3_rb_first(struct rb_root *);
+extern struct rb_node *v3_rb_last(struct rb_root *);
+
+/* Fast replacement of a single node without remove/rebalance/add/rebalance */
+extern void v3_rb_replace_node(struct rb_node *victim, struct rb_node *new, 
+                           struct rb_root *root);
+
+static inline void rb_link_node(struct rb_node * node, struct rb_node * parent,
+                               struct rb_node ** rb_link)
+{
+       node->rb_parent_color = (unsigned long )parent;
+       node->rb_left = node->rb_right = NULL;
+
+       *rb_link = node;
+}
+
+
+#endif
+
+#endif /* _LINUX_RBTREE_H */
index bea54aa..da9c76d 100644 (file)
@@ -35,14 +35,6 @@ struct shadow_page_state {
   v3_reg_t guest_cr0;
 
 
-
-
-  // Page table struct lookup table
-  struct hashtable * pt_cache;
-
-  // Guest CR3 to Shadow CR3 Lookup Table
-  struct hashtable *  cr3_cache;
-
   /* SOON TO BE DEPRECATED */
   // Hash table that contains a mapping of guest pte addresses to host pte addresses
   struct hashtable *  cached_ptes;
@@ -69,15 +61,6 @@ int v3_handle_shadow_invlpg(struct guest_info * info);
 int v3_activate_shadow_pt(struct guest_info * info);
 int v3_activate_passthrough_pt(struct guest_info * info);
 
-/* TODO: Change to static functions
- * External visibility not needed
- */
-addr_t v3_create_new_shadow_pt();
-int v3_replace_shdw_page32(struct guest_info * info, addr_t location, pte32_t * new_page, pte32_t * old_page); 
-/* *** */
-
-
-int v3_replace_shdw_page(struct guest_info * info, addr_t location, void * new_page, void * old_page);
 
 #endif // ! __V3VEE__
 
index 4099d22..95e68ec 100644 (file)
@@ -23,7 +23,6 @@
 #ifdef __V3VEE__
 
 
-
 #endif // ! __V3VEE__
 
 #endif
index f13dc1d..4b9649b 100644 (file)
@@ -242,8 +242,8 @@ static int pic_get_intr_number(void * private_data) {
   int i = 0;
   int irq = -1;
 
-  PrintDebug("8259 PIC: getnum: master_irr: 0x%x master_imr: 0x%x\n", i, state->master_irr, state->master_imr);
-  PrintDebug("8259 PIC: getnum: slave_irr: 0x%x slave_imr: 0x%x\n", i, state->slave_irr, state->slave_imr);
+  PrintDebug("8259 PIC: getnum: master_irr: 0x%x master_imr: 0x%x\n", state->master_irr, state->master_imr);
+  PrintDebug("8259 PIC: getnum: slave_irr: 0x%x slave_imr: 0x%x\n", state->slave_irr, state->slave_imr);
 
   for (i = 0; i < 16; i++) {
     if (i <= 7) {
index 8f25c0c..afcfc25 100644 (file)
@@ -1066,7 +1066,7 @@ static int keyboard_write_output(ushort_t port,
       case 0xff: // reset
        PushToOutputQueue(dev, 0xfa, OVERWRITE, COMMAND, KEYBOARD); // ack
        state->state = RESET;
-       PrintDebug("keyboard: reset complete and acked\n", data);
+       PrintDebug("keyboard: reset complete and acked\n");
        break;
       case 0xf5: // disable scanning
       case 0xf4: // enable scanning
@@ -1074,7 +1074,7 @@ static int keyboard_write_output(ushort_t port,
        PushToOutputQueue(dev, 0xfa, OVERWRITE, COMMAND, KEYBOARD);
        // should do something here... PAD
        state->state = NORMAL;
-       PrintDebug("keyboard: %s scanning done and acked\n",data==0xf5 ? "disable" : "enable", data);
+       PrintDebug("keyboard: %s scanning done and acked\n",data==0xf5 ? "disable" : "enable");
        break;
       case 0xfe: // resend
       case 0xfd: // set key type make
index 59e2003..5d433bb 100644 (file)
@@ -652,7 +652,7 @@ struct vm_device * v3_create_nvram() {
 
   nvram_state = (struct nvram_internal *)V3_Malloc(sizeof(struct nvram_internal) + 1000);
 
-  PrintDebug("nvram: internal at %x\n", nvram_state);
+  PrintDebug("nvram: internal at %p\n", (void *)nvram_state);
 
   struct vm_device * device = v3_create_device("NVRAM", &dev_ops, nvram_state);
 
index 96802d4..653643f 100644 (file)
@@ -37,6 +37,9 @@
 #include <palacios/vmm_lowlevel.h>
 #include <palacios/svm_msr.h>
 
+#include <palacios/vmm_rbtree.h>
+
+#include <palacios/vmm_profiler.h>
 
 
 extern void v3_stgi();
@@ -191,10 +194,12 @@ static void Init_VMCB_BIOS(vmcb_t * vmcb, struct guest_info *vm_info) {
 
 
 
-  if (vm_info->io_map.num_ports > 0) {
-    struct vmm_io_hook * iter;
+  if ( !RB_EMPTY_ROOT(&(vm_info->io_map)) ) {
+    struct v3_io_hook * iter;
+    struct rb_node * io_node = v3_rb_first(&(vm_info->io_map));
     addr_t io_port_bitmap;
-    
+    int i = 0;
+
     io_port_bitmap = (addr_t)V3_VAddr(V3_AllocPages(3));
     memset((uchar_t*)io_port_bitmap, 0, PAGE_SIZE * 3);
     
@@ -202,14 +207,19 @@ static void Init_VMCB_BIOS(vmcb_t * vmcb, struct guest_info *vm_info) {
 
     //PrintDebug("Setting up IO Map at 0x%x\n", io_port_bitmap);
 
-    FOREACH_IO_HOOK(vm_info->io_map, iter) {
+    do {
+      iter = rb_entry(io_node, struct v3_io_hook, tree_node);
+
       ushort_t port = iter->port;
       uchar_t * bitmap = (uchar_t *)io_port_bitmap;
+      PrintDebug("%d: Hooking Port %d\n", i, port);
 
       bitmap += (port / 8);
       //      PrintDebug("Setting Bit for port 0x%x\n", port);
       *bitmap |= 1 << (port % 8);
-    }
+
+      i++;
+    } while ((io_node = v3_rb_next(io_node)));
 
 
     //PrintDebugMemDump((uchar_t*)io_port_bitmap, PAGE_SIZE *2);
@@ -344,12 +354,15 @@ static int start_svm_guest(struct guest_info *info) {
     ullong_t tmp_tsc;
     uint_t vm_cr_low = 0, vm_cr_high = 0;
 
-
     v3_enable_ints();
     v3_clgi();
 
 
-    //PrintDebug("SVM Entry to rip=%p...\n", (void *)info->rip);
+    /*
+    PrintDebug("SVM Entry to CS=%p  rip=%p...\n", 
+              (void *)(addr_t)info->segments.cs.base, 
+              (void *)(addr_t)info->rip);
+    */
 
     v3_get_msr(0xc0000101, &vm_cr_high, &vm_cr_low);
 
@@ -372,10 +385,13 @@ static int start_svm_guest(struct guest_info *info) {
     v3_stgi();
 
 
-    if (num_exits % 25 == 0) {
+    if ((num_exits % 5000) == 0) {
       PrintDebug("SVM Exit number %d\n", num_exits);
+      if (info->enable_profiler) 
+       v3_print_profile(info);
     }
 
+
      
     if (v3_handle_svm_exit(info) != 0) {
       vmcb_ctrl_t * guest_ctrl = GET_VMCB_CTRL_AREA((vmcb_t*)(info->vmm_data));
@@ -404,10 +420,6 @@ static int start_svm_guest(struct guest_info *info) {
       }
       v3_print_GPRs(info);
 
-
-      
-
-
       PrintDebug("SVM Exit Code: %p\n", (void *)(addr_t)guest_ctrl->exit_code); 
       
       PrintDebug("exit_info1 low = 0x%.8x\n", *(uint_t*)&(guest_ctrl->exit_info1));
@@ -430,6 +442,7 @@ static int start_svm_guest(struct guest_info *info) {
 
       break;
     }
+
   }
   return 0;
 }
index 8495a17..893f49c 100644 (file)
 #include <palacios/vmm_intr.h>
 #include <palacios/vmm_emulator.h>
 #include <palacios/svm_msr.h>
+#include <palacios/vmm_profiler.h>
 
 
 
-static const uchar_t * vmexit_code_to_str(uint_t exit_code);
-
 
 int v3_handle_svm_exit(struct guest_info * info) {
   vmcb_ctrl_t * guest_ctrl = 0;
@@ -69,7 +68,10 @@ int v3_handle_svm_exit(struct guest_info * info) {
 
 
   exit_code = guest_ctrl->exit_code;
+
+
+
+  
 
   // Disable printing io exits due to bochs debug messages
   //if (!((exit_code == VMEXIT_IOIO) && ((ushort_t)(guest_ctrl->exit_info1 >> 16) == 0x402))) {
@@ -106,12 +108,10 @@ int v3_handle_svm_exit(struct guest_info * info) {
   }
 
 
-    //  }
-  // PrintDebugVMCB((vmcb_t*)(info->vmm_data));
-
 
-  // PrintDebug("SVM Returned:(VMCB=%x)\n", info->vmm_data); 
-  //PrintDebug("RIP: %x\n", guest_state->rip);
+  if (info->enable_profiler) {
+    rdtscll(info->profiler.start_time);
+  }
 
   
   //PrintDebug("SVM Returned: Exit Code: %x\n",exit_code); 
@@ -293,6 +293,9 @@ int v3_handle_svm_exit(struct guest_info * info) {
   } 
     break;
 
+
+#if 0
+    // Emulation handlers currently not used
   case VMEXIT_EXCP1: 
     {
 #ifdef DEBUG_EMULATOR
@@ -308,7 +311,7 @@ int v3_handle_svm_exit(struct guest_info * info) {
       }
       break;
     } 
-
+    
 
   case VMEXIT_VMMCALL: 
     {
@@ -340,7 +343,7 @@ int v3_handle_svm_exit(struct guest_info * info) {
       }
       break;
     } 
-    
+#endif
 
 
   case VMEXIT_WBINVD: 
@@ -426,6 +429,13 @@ int v3_handle_svm_exit(struct guest_info * info) {
   // END OF SWITCH (EXIT_CODE)
 
 
+  if (info->enable_profiler) {
+    rdtscll(info->profiler.end_time);
+    v3_profile_exit(info, exit_code);
+  }
+      
+
+
   // Update the low level state
 
   if (v3_intr_pending(info)) {
@@ -530,153 +540,153 @@ int v3_handle_svm_exit(struct guest_info * info) {
 }
 
 
-static const uchar_t VMEXIT_CR0_READ_STR[] = "VMEXIT_CR0_READ";
-static const uchar_t VMEXIT_CR1_READ_STR[] = "VMEXIT_CR1_READ";
-static const uchar_t VMEXIT_CR2_READ_STR[] = "VMEXIT_CR2_READ";
-static const uchar_t VMEXIT_CR3_READ_STR[] = "VMEXIT_CR3_READ";
-static const uchar_t VMEXIT_CR4_READ_STR[] = "VMEXIT_CR4_READ";
-static const uchar_t VMEXIT_CR5_READ_STR[] = "VMEXIT_CR5_READ";
-static const uchar_t VMEXIT_CR6_READ_STR[] = "VMEXIT_CR6_READ";
-static const uchar_t VMEXIT_CR7_READ_STR[] = "VMEXIT_CR7_READ";
-static const uchar_t VMEXIT_CR8_READ_STR[] = "VMEXIT_CR8_READ";
-static const uchar_t VMEXIT_CR9_READ_STR[] = "VMEXIT_CR9_READ";
-static const uchar_t VMEXIT_CR10_READ_STR[] = "VMEXIT_CR10_READ";
-static const uchar_t VMEXIT_CR11_READ_STR[] = "VMEXIT_CR11_READ";
-static const uchar_t VMEXIT_CR12_READ_STR[] = "VMEXIT_CR12_READ";
-static const uchar_t VMEXIT_CR13_READ_STR[] = "VMEXIT_CR13_READ";
-static const uchar_t VMEXIT_CR14_READ_STR[] = "VMEXIT_CR14_READ";
-static const uchar_t VMEXIT_CR15_READ_STR[] = "VMEXIT_CR15_READ";
-static const uchar_t VMEXIT_CR0_WRITE_STR[] = "VMEXIT_CR0_WRITE";
-static const uchar_t VMEXIT_CR1_WRITE_STR[] = "VMEXIT_CR1_WRITE";
-static const uchar_t VMEXIT_CR2_WRITE_STR[] = "VMEXIT_CR2_WRITE";
-static const uchar_t VMEXIT_CR3_WRITE_STR[] = "VMEXIT_CR3_WRITE";
-static const uchar_t VMEXIT_CR4_WRITE_STR[] = "VMEXIT_CR4_WRITE";
-static const uchar_t VMEXIT_CR5_WRITE_STR[] = "VMEXIT_CR5_WRITE";
-static const uchar_t VMEXIT_CR6_WRITE_STR[] = "VMEXIT_CR6_WRITE";
-static const uchar_t VMEXIT_CR7_WRITE_STR[] = "VMEXIT_CR7_WRITE";
-static const uchar_t VMEXIT_CR8_WRITE_STR[] = "VMEXIT_CR8_WRITE";
-static const uchar_t VMEXIT_CR9_WRITE_STR[] = "VMEXIT_CR9_WRITE";
-static const uchar_t VMEXIT_CR10_WRITE_STR[] = "VMEXIT_CR10_WRITE";
-static const uchar_t VMEXIT_CR11_WRITE_STR[] = "VMEXIT_CR11_WRITE";
-static const uchar_t VMEXIT_CR12_WRITE_STR[] = "VMEXIT_CR12_WRITE";
-static const uchar_t VMEXIT_CR13_WRITE_STR[] = "VMEXIT_CR13_WRITE";
-static const uchar_t VMEXIT_CR14_WRITE_STR[] = "VMEXIT_CR14_WRITE";
-static const uchar_t VMEXIT_CR15_WRITE_STR[] = "VMEXIT_CR15_WRITE";
-static const uchar_t VMEXIT_DR0_READ_STR[] = "VMEXIT_DR0_READ";
-static const uchar_t VMEXIT_DR1_READ_STR[] = "VMEXIT_DR1_READ";
-static const uchar_t VMEXIT_DR2_READ_STR[] = "VMEXIT_DR2_READ";
-static const uchar_t VMEXIT_DR3_READ_STR[] = "VMEXIT_DR3_READ";
-static const uchar_t VMEXIT_DR4_READ_STR[] = "VMEXIT_DR4_READ";
-static const uchar_t VMEXIT_DR5_READ_STR[] = "VMEXIT_DR5_READ";
-static const uchar_t VMEXIT_DR6_READ_STR[] = "VMEXIT_DR6_READ";
-static const uchar_t VMEXIT_DR7_READ_STR[] = "VMEXIT_DR7_READ";
-static const uchar_t VMEXIT_DR8_READ_STR[] = "VMEXIT_DR8_READ";
-static const uchar_t VMEXIT_DR9_READ_STR[] = "VMEXIT_DR9_READ";
-static const uchar_t VMEXIT_DR10_READ_STR[] = "VMEXIT_DR10_READ";
-static const uchar_t VMEXIT_DR11_READ_STR[] = "VMEXIT_DR11_READ";
-static const uchar_t VMEXIT_DR12_READ_STR[] = "VMEXIT_DR12_READ";
-static const uchar_t VMEXIT_DR13_READ_STR[] = "VMEXIT_DR13_READ";
-static const uchar_t VMEXIT_DR14_READ_STR[] = "VMEXIT_DR14_READ";
-static const uchar_t VMEXIT_DR15_READ_STR[] = "VMEXIT_DR15_READ";
-static const uchar_t VMEXIT_DR0_WRITE_STR[] = "VMEXIT_DR0_WRITE";
-static const uchar_t VMEXIT_DR1_WRITE_STR[] = "VMEXIT_DR1_WRITE";
-static const uchar_t VMEXIT_DR2_WRITE_STR[] = "VMEXIT_DR2_WRITE";
-static const uchar_t VMEXIT_DR3_WRITE_STR[] = "VMEXIT_DR3_WRITE";
-static const uchar_t VMEXIT_DR4_WRITE_STR[] = "VMEXIT_DR4_WRITE";
-static const uchar_t VMEXIT_DR5_WRITE_STR[] = "VMEXIT_DR5_WRITE";
-static const uchar_t VMEXIT_DR6_WRITE_STR[] = "VMEXIT_DR6_WRITE";
-static const uchar_t VMEXIT_DR7_WRITE_STR[] = "VMEXIT_DR7_WRITE";
-static const uchar_t VMEXIT_DR8_WRITE_STR[] = "VMEXIT_DR8_WRITE";
-static const uchar_t VMEXIT_DR9_WRITE_STR[] = "VMEXIT_DR9_WRITE";
-static const uchar_t VMEXIT_DR10_WRITE_STR[] = "VMEXIT_DR10_WRITE";
-static const uchar_t VMEXIT_DR11_WRITE_STR[] = "VMEXIT_DR11_WRITE";
-static const uchar_t VMEXIT_DR12_WRITE_STR[] = "VMEXIT_DR12_WRITE";
-static const uchar_t VMEXIT_DR13_WRITE_STR[] = "VMEXIT_DR13_WRITE";
-static const uchar_t VMEXIT_DR14_WRITE_STR[] = "VMEXIT_DR14_WRITE";
-static const uchar_t VMEXIT_DR15_WRITE_STR[] = "VMEXIT_DR15_WRITE";
-static const uchar_t VMEXIT_EXCP0_STR[] = "VMEXIT_EXCP0";
-static const uchar_t VMEXIT_EXCP1_STR[] = "VMEXIT_EXCP1";
-static const uchar_t VMEXIT_EXCP2_STR[] = "VMEXIT_EXCP2";
-static const uchar_t VMEXIT_EXCP3_STR[] = "VMEXIT_EXCP3";
-static const uchar_t VMEXIT_EXCP4_STR[] = "VMEXIT_EXCP4";
-static const uchar_t VMEXIT_EXCP5_STR[] = "VMEXIT_EXCP5";
-static const uchar_t VMEXIT_EXCP6_STR[] = "VMEXIT_EXCP6";
-static const uchar_t VMEXIT_EXCP7_STR[] = "VMEXIT_EXCP7";
-static const uchar_t VMEXIT_EXCP8_STR[] = "VMEXIT_EXCP8";
-static const uchar_t VMEXIT_EXCP9_STR[] = "VMEXIT_EXCP9";
-static const uchar_t VMEXIT_EXCP10_STR[] = "VMEXIT_EXCP10";
-static const uchar_t VMEXIT_EXCP11_STR[] = "VMEXIT_EXCP11";
-static const uchar_t VMEXIT_EXCP12_STR[] = "VMEXIT_EXCP12";
-static const uchar_t VMEXIT_EXCP13_STR[] = "VMEXIT_EXCP13";
-static const uchar_t VMEXIT_EXCP14_STR[] = "VMEXIT_EXCP14";
-static const uchar_t VMEXIT_EXCP15_STR[] = "VMEXIT_EXCP15";
-static const uchar_t VMEXIT_EXCP16_STR[] = "VMEXIT_EXCP16";
-static const uchar_t VMEXIT_EXCP17_STR[] = "VMEXIT_EXCP17";
-static const uchar_t VMEXIT_EXCP18_STR[] = "VMEXIT_EXCP18";
-static const uchar_t VMEXIT_EXCP19_STR[] = "VMEXIT_EXCP19";
-static const uchar_t VMEXIT_EXCP20_STR[] = "VMEXIT_EXCP20";
-static const uchar_t VMEXIT_EXCP21_STR[] = "VMEXIT_EXCP21";
-static const uchar_t VMEXIT_EXCP22_STR[] = "VMEXIT_EXCP22";
-static const uchar_t VMEXIT_EXCP23_STR[] = "VMEXIT_EXCP23";
-static const uchar_t VMEXIT_EXCP24_STR[] = "VMEXIT_EXCP24";
-static const uchar_t VMEXIT_EXCP25_STR[] = "VMEXIT_EXCP25";
-static const uchar_t VMEXIT_EXCP26_STR[] = "VMEXIT_EXCP26";
-static const uchar_t VMEXIT_EXCP27_STR[] = "VMEXIT_EXCP27";
-static const uchar_t VMEXIT_EXCP28_STR[] = "VMEXIT_EXCP28";
-static const uchar_t VMEXIT_EXCP29_STR[] = "VMEXIT_EXCP29";
-static const uchar_t VMEXIT_EXCP30_STR[] = "VMEXIT_EXCP30";
-static const uchar_t VMEXIT_EXCP31_STR[] = "VMEXIT_EXCP31";
-static const uchar_t VMEXIT_INTR_STR[] = "VMEXIT_INTR";
-static const uchar_t VMEXIT_NMI_STR[] = "VMEXIT_NMI";
-static const uchar_t VMEXIT_SMI_STR[] = "VMEXIT_SMI";
-static const uchar_t VMEXIT_INIT_STR[] = "VMEXIT_INIT";
-static const uchar_t VMEXIT_VINITR_STR[] = "VMEXIT_VINITR";
-static const uchar_t VMEXIT_CR0_SEL_WRITE_STR[] = "VMEXIT_CR0_SEL_WRITE";
-static const uchar_t VMEXIT_IDTR_READ_STR[] = "VMEXIT_IDTR_READ";
-static const uchar_t VMEXIT_GDTR_READ_STR[] = "VMEXIT_GDTR_READ";
-static const uchar_t VMEXIT_LDTR_READ_STR[] = "VMEXIT_LDTR_READ";
-static const uchar_t VMEXIT_TR_READ_STR[] = "VMEXIT_TR_READ";
-static const uchar_t VMEXIT_IDTR_WRITE_STR[] = "VMEXIT_IDTR_WRITE";
-static const uchar_t VMEXIT_GDTR_WRITE_STR[] = "VMEXIT_GDTR_WRITE";
-static const uchar_t VMEXIT_LDTR_WRITE_STR[] = "VMEXIT_LDTR_WRITE";
-static const uchar_t VMEXIT_TR_WRITE_STR[] = "VMEXIT_TR_WRITE";
-static const uchar_t VMEXIT_RDTSC_STR[] = "VMEXIT_RDTSC";
-static const uchar_t VMEXIT_RDPMC_STR[] = "VMEXIT_RDPMC";
-static const uchar_t VMEXIT_PUSHF_STR[] = "VMEXIT_PUSHF";
-static const uchar_t VMEXIT_POPF_STR[] = "VMEXIT_POPF";
-static const uchar_t VMEXIT_CPUID_STR[] = "VMEXIT_CPUID";
-static const uchar_t VMEXIT_RSM_STR[] = "VMEXIT_RSM";
-static const uchar_t VMEXIT_IRET_STR[] = "VMEXIT_IRET";
-static const uchar_t VMEXIT_SWINT_STR[] = "VMEXIT_SWINT";
-static const uchar_t VMEXIT_INVD_STR[] = "VMEXIT_INVD";
-static const uchar_t VMEXIT_PAUSE_STR[] = "VMEXIT_PAUSE";
-static const uchar_t VMEXIT_HLT_STR[] = "VMEXIT_HLT";
-static const uchar_t VMEXIT_INVLPG_STR[] = "VMEXIT_INVLPG";
-static const uchar_t VMEXIT_INVLPGA_STR[] = "VMEXIT_INVLPGA";
-static const uchar_t VMEXIT_IOIO_STR[] = "VMEXIT_IOIO";
-static const uchar_t VMEXIT_MSR_STR[] = "VMEXIT_MSR";
-static const uchar_t VMEXIT_TASK_SWITCH_STR[] = "VMEXIT_TASK_SWITCH";
-static const uchar_t VMEXIT_FERR_FREEZE_STR[] = "VMEXIT_FERR_FREEZE";
-static const uchar_t VMEXIT_SHUTDOWN_STR[] = "VMEXIT_SHUTDOWN";
-static const uchar_t VMEXIT_VMRUN_STR[] = "VMEXIT_VMRUN";
-static const uchar_t VMEXIT_VMMCALL_STR[] = "VMEXIT_VMMCALL";
-static const uchar_t VMEXIT_VMLOAD_STR[] = "VMEXIT_VMLOAD";
-static const uchar_t VMEXIT_VMSAVE_STR[] = "VMEXIT_VMSAVE";
-static const uchar_t VMEXIT_STGI_STR[] = "VMEXIT_STGI";
-static const uchar_t VMEXIT_CLGI_STR[] = "VMEXIT_CLGI";
-static const uchar_t VMEXIT_SKINIT_STR[] = "VMEXIT_SKINIT";
-static const uchar_t VMEXIT_RDTSCP_STR[] = "VMEXIT_RDTSCP";
-static const uchar_t VMEXIT_ICEBP_STR[] = "VMEXIT_ICEBP";
-static const uchar_t VMEXIT_WBINVD_STR[] = "VMEXIT_WBINVD";
-static const uchar_t VMEXIT_MONITOR_STR[] = "VMEXIT_MONITOR";
-static const uchar_t VMEXIT_MWAIT_STR[] = "VMEXIT_MWAIT";
-static const uchar_t VMEXIT_MWAIT_CONDITIONAL_STR[] = "VMEXIT_MWAIT_CONDITIONAL";
-static const uchar_t VMEXIT_NPF_STR[] = "VMEXIT_NPF";
-static const uchar_t VMEXIT_INVALID_VMCB_STR[] = "VMEXIT_INVALID_VMCB";
-
-
-
-const uchar_t * vmexit_code_to_str(uint_t exit_code) {
+static const char VMEXIT_CR0_READ_STR[] = "VMEXIT_CR0_READ";
+static const char VMEXIT_CR1_READ_STR[] = "VMEXIT_CR1_READ";
+static const char VMEXIT_CR2_READ_STR[] = "VMEXIT_CR2_READ";
+static const char VMEXIT_CR3_READ_STR[] = "VMEXIT_CR3_READ";
+static const char VMEXIT_CR4_READ_STR[] = "VMEXIT_CR4_READ";
+static const char VMEXIT_CR5_READ_STR[] = "VMEXIT_CR5_READ";
+static const char VMEXIT_CR6_READ_STR[] = "VMEXIT_CR6_READ";
+static const char VMEXIT_CR7_READ_STR[] = "VMEXIT_CR7_READ";
+static const char VMEXIT_CR8_READ_STR[] = "VMEXIT_CR8_READ";
+static const char VMEXIT_CR9_READ_STR[] = "VMEXIT_CR9_READ";
+static const char VMEXIT_CR10_READ_STR[] = "VMEXIT_CR10_READ";
+static const char VMEXIT_CR11_READ_STR[] = "VMEXIT_CR11_READ";
+static const char VMEXIT_CR12_READ_STR[] = "VMEXIT_CR12_READ";
+static const char VMEXIT_CR13_READ_STR[] = "VMEXIT_CR13_READ";
+static const char VMEXIT_CR14_READ_STR[] = "VMEXIT_CR14_READ";
+static const char VMEXIT_CR15_READ_STR[] = "VMEXIT_CR15_READ";
+static const char VMEXIT_CR0_WRITE_STR[] = "VMEXIT_CR0_WRITE";
+static const char VMEXIT_CR1_WRITE_STR[] = "VMEXIT_CR1_WRITE";
+static const char VMEXIT_CR2_WRITE_STR[] = "VMEXIT_CR2_WRITE";
+static const char VMEXIT_CR3_WRITE_STR[] = "VMEXIT_CR3_WRITE";
+static const char VMEXIT_CR4_WRITE_STR[] = "VMEXIT_CR4_WRITE";
+static const char VMEXIT_CR5_WRITE_STR[] = "VMEXIT_CR5_WRITE";
+static const char VMEXIT_CR6_WRITE_STR[] = "VMEXIT_CR6_WRITE";
+static const char VMEXIT_CR7_WRITE_STR[] = "VMEXIT_CR7_WRITE";
+static const char VMEXIT_CR8_WRITE_STR[] = "VMEXIT_CR8_WRITE";
+static const char VMEXIT_CR9_WRITE_STR[] = "VMEXIT_CR9_WRITE";
+static const char VMEXIT_CR10_WRITE_STR[] = "VMEXIT_CR10_WRITE";
+static const char VMEXIT_CR11_WRITE_STR[] = "VMEXIT_CR11_WRITE";
+static const char VMEXIT_CR12_WRITE_STR[] = "VMEXIT_CR12_WRITE";
+static const char VMEXIT_CR13_WRITE_STR[] = "VMEXIT_CR13_WRITE";
+static const char VMEXIT_CR14_WRITE_STR[] = "VMEXIT_CR14_WRITE";
+static const char VMEXIT_CR15_WRITE_STR[] = "VMEXIT_CR15_WRITE";
+static const char VMEXIT_DR0_READ_STR[] = "VMEXIT_DR0_READ";
+static const char VMEXIT_DR1_READ_STR[] = "VMEXIT_DR1_READ";
+static const char VMEXIT_DR2_READ_STR[] = "VMEXIT_DR2_READ";
+static const char VMEXIT_DR3_READ_STR[] = "VMEXIT_DR3_READ";
+static const char VMEXIT_DR4_READ_STR[] = "VMEXIT_DR4_READ";
+static const char VMEXIT_DR5_READ_STR[] = "VMEXIT_DR5_READ";
+static const char VMEXIT_DR6_READ_STR[] = "VMEXIT_DR6_READ";
+static const char VMEXIT_DR7_READ_STR[] = "VMEXIT_DR7_READ";
+static const char VMEXIT_DR8_READ_STR[] = "VMEXIT_DR8_READ";
+static const char VMEXIT_DR9_READ_STR[] = "VMEXIT_DR9_READ";
+static const char VMEXIT_DR10_READ_STR[] = "VMEXIT_DR10_READ";
+static const char VMEXIT_DR11_READ_STR[] = "VMEXIT_DR11_READ";
+static const char VMEXIT_DR12_READ_STR[] = "VMEXIT_DR12_READ";
+static const char VMEXIT_DR13_READ_STR[] = "VMEXIT_DR13_READ";
+static const char VMEXIT_DR14_READ_STR[] = "VMEXIT_DR14_READ";
+static const char VMEXIT_DR15_READ_STR[] = "VMEXIT_DR15_READ";
+static const char VMEXIT_DR0_WRITE_STR[] = "VMEXIT_DR0_WRITE";
+static const char VMEXIT_DR1_WRITE_STR[] = "VMEXIT_DR1_WRITE";
+static const char VMEXIT_DR2_WRITE_STR[] = "VMEXIT_DR2_WRITE";
+static const char VMEXIT_DR3_WRITE_STR[] = "VMEXIT_DR3_WRITE";
+static const char VMEXIT_DR4_WRITE_STR[] = "VMEXIT_DR4_WRITE";
+static const char VMEXIT_DR5_WRITE_STR[] = "VMEXIT_DR5_WRITE";
+static const char VMEXIT_DR6_WRITE_STR[] = "VMEXIT_DR6_WRITE";
+static const char VMEXIT_DR7_WRITE_STR[] = "VMEXIT_DR7_WRITE";
+static const char VMEXIT_DR8_WRITE_STR[] = "VMEXIT_DR8_WRITE";
+static const char VMEXIT_DR9_WRITE_STR[] = "VMEXIT_DR9_WRITE";
+static const char VMEXIT_DR10_WRITE_STR[] = "VMEXIT_DR10_WRITE";
+static const char VMEXIT_DR11_WRITE_STR[] = "VMEXIT_DR11_WRITE";
+static const char VMEXIT_DR12_WRITE_STR[] = "VMEXIT_DR12_WRITE";
+static const char VMEXIT_DR13_WRITE_STR[] = "VMEXIT_DR13_WRITE";
+static const char VMEXIT_DR14_WRITE_STR[] = "VMEXIT_DR14_WRITE";
+static const char VMEXIT_DR15_WRITE_STR[] = "VMEXIT_DR15_WRITE";
+static const char VMEXIT_EXCP0_STR[] = "VMEXIT_EXCP0";
+static const char VMEXIT_EXCP1_STR[] = "VMEXIT_EXCP1";
+static const char VMEXIT_EXCP2_STR[] = "VMEXIT_EXCP2";
+static const char VMEXIT_EXCP3_STR[] = "VMEXIT_EXCP3";
+static const char VMEXIT_EXCP4_STR[] = "VMEXIT_EXCP4";
+static const char VMEXIT_EXCP5_STR[] = "VMEXIT_EXCP5";
+static const char VMEXIT_EXCP6_STR[] = "VMEXIT_EXCP6";
+static const char VMEXIT_EXCP7_STR[] = "VMEXIT_EXCP7";
+static const char VMEXIT_EXCP8_STR[] = "VMEXIT_EXCP8";
+static const char VMEXIT_EXCP9_STR[] = "VMEXIT_EXCP9";
+static const char VMEXIT_EXCP10_STR[] = "VMEXIT_EXCP10";
+static const char VMEXIT_EXCP11_STR[] = "VMEXIT_EXCP11";
+static const char VMEXIT_EXCP12_STR[] = "VMEXIT_EXCP12";
+static const char VMEXIT_EXCP13_STR[] = "VMEXIT_EXCP13";
+static const char VMEXIT_EXCP14_STR[] = "VMEXIT_EXCP14";
+static const char VMEXIT_EXCP15_STR[] = "VMEXIT_EXCP15";
+static const char VMEXIT_EXCP16_STR[] = "VMEXIT_EXCP16";
+static const char VMEXIT_EXCP17_STR[] = "VMEXIT_EXCP17";
+static const char VMEXIT_EXCP18_STR[] = "VMEXIT_EXCP18";
+static const char VMEXIT_EXCP19_STR[] = "VMEXIT_EXCP19";
+static const char VMEXIT_EXCP20_STR[] = "VMEXIT_EXCP20";
+static const char VMEXIT_EXCP21_STR[] = "VMEXIT_EXCP21";
+static const char VMEXIT_EXCP22_STR[] = "VMEXIT_EXCP22";
+static const char VMEXIT_EXCP23_STR[] = "VMEXIT_EXCP23";
+static const char VMEXIT_EXCP24_STR[] = "VMEXIT_EXCP24";
+static const char VMEXIT_EXCP25_STR[] = "VMEXIT_EXCP25";
+static const char VMEXIT_EXCP26_STR[] = "VMEXIT_EXCP26";
+static const char VMEXIT_EXCP27_STR[] = "VMEXIT_EXCP27";
+static const char VMEXIT_EXCP28_STR[] = "VMEXIT_EXCP28";
+static const char VMEXIT_EXCP29_STR[] = "VMEXIT_EXCP29";
+static const char VMEXIT_EXCP30_STR[] = "VMEXIT_EXCP30";
+static const char VMEXIT_EXCP31_STR[] = "VMEXIT_EXCP31";
+static const char VMEXIT_INTR_STR[] = "VMEXIT_INTR";
+static const char VMEXIT_NMI_STR[] = "VMEXIT_NMI";
+static const char VMEXIT_SMI_STR[] = "VMEXIT_SMI";
+static const char VMEXIT_INIT_STR[] = "VMEXIT_INIT";
+static const char VMEXIT_VINITR_STR[] = "VMEXIT_VINITR";
+static const char VMEXIT_CR0_SEL_WRITE_STR[] = "VMEXIT_CR0_SEL_WRITE";
+static const char VMEXIT_IDTR_READ_STR[] = "VMEXIT_IDTR_READ";
+static const char VMEXIT_GDTR_READ_STR[] = "VMEXIT_GDTR_READ";
+static const char VMEXIT_LDTR_READ_STR[] = "VMEXIT_LDTR_READ";
+static const char VMEXIT_TR_READ_STR[] = "VMEXIT_TR_READ";
+static const char VMEXIT_IDTR_WRITE_STR[] = "VMEXIT_IDTR_WRITE";
+static const char VMEXIT_GDTR_WRITE_STR[] = "VMEXIT_GDTR_WRITE";
+static const char VMEXIT_LDTR_WRITE_STR[] = "VMEXIT_LDTR_WRITE";
+static const char VMEXIT_TR_WRITE_STR[] = "VMEXIT_TR_WRITE";
+static const char VMEXIT_RDTSC_STR[] = "VMEXIT_RDTSC";
+static const char VMEXIT_RDPMC_STR[] = "VMEXIT_RDPMC";
+static const char VMEXIT_PUSHF_STR[] = "VMEXIT_PUSHF";
+static const char VMEXIT_POPF_STR[] = "VMEXIT_POPF";
+static const char VMEXIT_CPUID_STR[] = "VMEXIT_CPUID";
+static const char VMEXIT_RSM_STR[] = "VMEXIT_RSM";
+static const char VMEXIT_IRET_STR[] = "VMEXIT_IRET";
+static const char VMEXIT_SWINT_STR[] = "VMEXIT_SWINT";
+static const char VMEXIT_INVD_STR[] = "VMEXIT_INVD";
+static const char VMEXIT_PAUSE_STR[] = "VMEXIT_PAUSE";
+static const char VMEXIT_HLT_STR[] = "VMEXIT_HLT";
+static const char VMEXIT_INVLPG_STR[] = "VMEXIT_INVLPG";
+static const char VMEXIT_INVLPGA_STR[] = "VMEXIT_INVLPGA";
+static const char VMEXIT_IOIO_STR[] = "VMEXIT_IOIO";
+static const char VMEXIT_MSR_STR[] = "VMEXIT_MSR";
+static const char VMEXIT_TASK_SWITCH_STR[] = "VMEXIT_TASK_SWITCH";
+static const char VMEXIT_FERR_FREEZE_STR[] = "VMEXIT_FERR_FREEZE";
+static const char VMEXIT_SHUTDOWN_STR[] = "VMEXIT_SHUTDOWN";
+static const char VMEXIT_VMRUN_STR[] = "VMEXIT_VMRUN";
+static const char VMEXIT_VMMCALL_STR[] = "VMEXIT_VMMCALL";
+static const char VMEXIT_VMLOAD_STR[] = "VMEXIT_VMLOAD";
+static const char VMEXIT_VMSAVE_STR[] = "VMEXIT_VMSAVE";
+static const char VMEXIT_STGI_STR[] = "VMEXIT_STGI";
+static const char VMEXIT_CLGI_STR[] = "VMEXIT_CLGI";
+static const char VMEXIT_SKINIT_STR[] = "VMEXIT_SKINIT";
+static const char VMEXIT_RDTSCP_STR[] = "VMEXIT_RDTSCP";
+static const char VMEXIT_ICEBP_STR[] = "VMEXIT_ICEBP";
+static const char VMEXIT_WBINVD_STR[] = "VMEXIT_WBINVD";
+static const char VMEXIT_MONITOR_STR[] = "VMEXIT_MONITOR";
+static const char VMEXIT_MWAIT_STR[] = "VMEXIT_MWAIT";
+static const char VMEXIT_MWAIT_CONDITIONAL_STR[] = "VMEXIT_MWAIT_CONDITIONAL";
+static const char VMEXIT_NPF_STR[] = "VMEXIT_NPF";
+static const char VMEXIT_INVALID_VMCB_STR[] = "VMEXIT_INVALID_VMCB";
+
+
+
+const char * vmexit_code_to_str(uint_t exit_code) {
   switch(exit_code) {
   case VMEXIT_CR0_READ:
     return VMEXIT_CR0_READ_STR;
index 04b488b..e43deb0 100644 (file)
@@ -17,9 +17,6 @@
  * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
  */
 
-
-
-
 #include <palacios/svm_io.h>
 #include <palacios/vmm_io.h>
 #include <palacios/vmm_ctrl_regs.h>
@@ -41,7 +38,7 @@ int v3_handle_svm_io_in(struct guest_info * info) {
   //  vmcb_saved_state_t * guest_state = GET_VMCB_SAVE_STATE_AREA((vmcb_t*)(info->vmm_data));
   struct svm_io_info * io_info = (struct svm_io_info *)&(ctrl_area->exit_info1);
 
-  struct vmm_io_hook * hook = v3_get_io_hook(&(info->io_map), io_info->port);
+  struct v3_io_hook * hook = v3_get_io_hook(info, io_info->port);
   int read_size = 0;
 
   if (hook == NULL) {
@@ -85,7 +82,7 @@ int v3_handle_svm_io_ins(struct guest_info * info) {
   
   struct svm_io_info * io_info = (struct svm_io_info *)&(ctrl_area->exit_info1);
   
-  struct vmm_io_hook * hook = v3_get_io_hook(&(info->io_map), io_info->port);
+  struct v3_io_hook * hook = v3_get_io_hook(info, io_info->port);
   int read_size = 0;
 
   addr_t dst_addr = 0;
@@ -221,7 +218,7 @@ int v3_handle_svm_io_out(struct guest_info * info) {
   //  vmcb_saved_state_t * guest_state = GET_VMCB_SAVE_STATE_AREA((vmcb_t*)(info->vmm_data));
   struct svm_io_info * io_info = (struct svm_io_info *)&(ctrl_area->exit_info1);
 
-  struct vmm_io_hook * hook = v3_get_io_hook(&(info->io_map), io_info->port);
+  struct v3_io_hook * hook = v3_get_io_hook(info, io_info->port);
   int write_size = 0;
 
   if (hook == NULL) {
@@ -264,7 +261,7 @@ int v3_handle_svm_io_outs(struct guest_info * info) {
   
   struct svm_io_info * io_info = (struct svm_io_info *)&(ctrl_area->exit_info1);
   
-  struct vmm_io_hook * hook = v3_get_io_hook(&(info->io_map), io_info->port);
+  struct v3_io_hook * hook = v3_get_io_hook(info, io_info->port);
   int write_size = 0;
 
   addr_t dst_addr = 0;
index 876f0b3..b8db954 100644 (file)
@@ -66,12 +66,15 @@ int host_pa_to_host_va(addr_t host_pa, addr_t * host_va) {
 
 
 int guest_pa_to_host_pa(struct guest_info * guest_info, addr_t guest_pa, addr_t * host_pa) {
-  // we use the shadow map here...
-  host_region_type_t reg_type = lookup_shadow_map_addr(&(guest_info->mem_map), guest_pa, host_pa);
+  struct v3_shadow_region * shdw_reg = v3_get_shadow_region(guest_info, guest_pa);
 
-  if (reg_type != HOST_REGION_PHYSICAL_MEMORY) {
-    PrintError("In GPA->HPA: Could not find address in shadow map (addr=%p) (reg_type=%d)\n", 
-               (void *)guest_pa, reg_type);
+  *host_pa = v3_get_shadow_addr(shdw_reg, guest_pa);
+
+  if ((shdw_reg == NULL) ||
+      (shdw_reg->host_type == SHDW_REGION_INVALID) ||
+      (shdw_reg->host_type == SHDW_REGION_FULL_HOOK)){
+    PrintError("In GPA->HPA: Could not find address in shadow map (addr=%p) (reg_type=%s)\n", 
+              (void *)guest_pa, v3_shdw_region_type_to_str(shdw_reg->host_type));
     return -1;
   }
 
index 4502a80..6ef7fa1 100644 (file)
@@ -23,7 +23,7 @@
 #include <palacios/vmm_intr.h>
 #include <palacios/vmm_config.h>
 #include <palacios/vm_guest.h>
-#include <palacios/vmm_decoder.h>
+
 
 v3_cpu_arch_t v3_cpu_type;
 struct v3_os_hooks * os_hooks = NULL;
@@ -43,7 +43,7 @@ void Init_V3(struct v3_os_hooks * hooks, struct v3_ctrl_ops * vmm_ops) {
 
   v3_cpu_type = V3_INVALID_CPU;
 
-  v3_init_decoder();
+
 
   if (v3_is_svm_capable()) {
 
index fd472dd..2a93ce1 100644 (file)
@@ -21,7 +21,9 @@
 #include <palacios/vmm.h>
 #include <palacios/vmm_debug.h>
 #include <palacios/vmm_msr.h>
-
+#include <palacios/vmm_decoder.h>
+#include <palacios/vmm_profiler.h>
+#include <palacios/vmm_mem.h>
 
 #include <devices/serial.h>
 #include <devices/keyboard.h>
@@ -34,6 +36,7 @@
 #include <devices/bochs_debug.h>
 
 
+
 #include <palacios/vmm_host_events.h>
 
 #define USE_GENERIC 1
@@ -50,23 +53,14 @@ static struct vm_device *  configure_generic(struct guest_info * info, struct v3
 
 
 
-static int mem_test_read(addr_t guest_addr, void * dst, uint_t length, void * priv_data) {
-  int foo = 20;
 
+static int passthrough_mem_write(addr_t guest_addr, void * src, uint_t length, void * priv_data) {
 
-  memcpy(dst, &foo, length);
-
-  PrintDebug("Passthrough mem read returning: %p (length=%d)\n", (void *)(foo + (guest_addr & 0xfff)), length);
   return length;
-}
-
-static int passthrough_mem_read(addr_t guest_addr, void * dst, uint_t length, void * priv_data) {
-    memcpy(dst, (void*)guest_addr, length);
-    return length;
-}
+  //  memcpy((void*)guest_addr, src, length);
+  PrintDebug("Write of %d bytes to %p\n", length, (void *)guest_addr);
+  PrintDebug("Write Value = %p\n", (void *)*(addr_t *)src);
 
-static int passthrough_mem_write(addr_t guest_addr, void * src, uint_t length, void * priv_data) {
-  memcpy((void*)guest_addr, src, length);
   return length;
 }
 
@@ -77,14 +71,15 @@ int v3_config_guest(struct guest_info * info, struct v3_vm_config * config_ptr)
 
   // Initialize the subsystem data strutures
   v3_init_time(info);
-  v3_init_vmm_io_map(info);
+  v3_init_io_map(info);
   v3_init_msr_map(info);
   v3_init_interrupt_state(info);
   v3_init_dev_mgr(info);
-  v3_init_emulator(info);
   v3_init_host_events(info);
 
-  init_shadow_map(info);
+  v3_init_decoder(info);
+
+  v3_init_shadow_map(info);
   
   if (v3_cpu_type == V3_SVM_REV3_CPU) {
     info->shdw_pg_mode = NESTED_PAGING;
@@ -107,6 +102,14 @@ int v3_config_guest(struct guest_info * info, struct v3_vm_config * config_ptr)
   setup_devices(info, config_ptr);
  
 
+
+  if (config_ptr->enable_profiling) {
+    info->enable_profiler = 1;
+    v3_init_profiler(info);
+  } else {
+    info->enable_profiler = 0;
+  }
+
   //v3_hook_io_port(info, 1234, &IO_Read, NULL, info);
 
   // Setup initial cpu register state
@@ -132,7 +135,7 @@ static int setup_memory_map(struct guest_info * info, struct v3_vm_config * conf
     PrintDebug("Layout Region %d bytes\n", config_ptr->rombios_size);
     memcpy(V3_VAddr(guest_mem), config_ptr->rombios, config_ptr->rombios_size);
 
-    add_shadow_region_passthrough(info, ROMBIOS_START, ROMBIOS_START + (num_pages * PAGE_SIZE), (addr_t)guest_mem);
+    v3_add_shadow_mem(info, ROMBIOS_START, ROMBIOS_START + (num_pages * PAGE_SIZE) - 1, (addr_t)guest_mem);
     
     PrintDebug("Adding Shadow Region (0x%p-0x%p) -> 0x%p\n", 
               (void *)ROMBIOS_START, 
@@ -149,7 +152,7 @@ static int setup_memory_map(struct guest_info * info, struct v3_vm_config * conf
     PrintDebug("Layout Region %d bytes\n", config_ptr->vgabios_size);
     memcpy(V3_VAddr(guest_mem), config_ptr->vgabios, config_ptr->vgabios_size);
 
-    add_shadow_region_passthrough(info, VGABIOS_START, VGABIOS_START + (num_pages * PAGE_SIZE), (addr_t)guest_mem);
+    v3_add_shadow_mem(info, VGABIOS_START, VGABIOS_START + (num_pages * PAGE_SIZE) - 1, (addr_t)guest_mem);
     
     PrintDebug("Adding Shadow Region (0x%p-0x%p) -> 0x%p\n", 
               (void *)VGABIOS_START, 
@@ -158,44 +161,44 @@ static int setup_memory_map(struct guest_info * info, struct v3_vm_config * conf
   }
 
       //     
-  add_shadow_region_passthrough(info, 0x0, 0xa0000, (addr_t)V3_AllocPages(160));
+  v3_add_shadow_mem(info, 0x0, 0x9ffff, (addr_t)V3_AllocPages(160));
   
   if (1) {
-    add_shadow_region_passthrough(info, 0xa0000, 0xc0000, 0xa0000); 
+    v3_add_shadow_mem(info, 0xa0000, 0xbffff, 0xa0000); 
   } else {
-    hook_guest_mem(info, 0xa0000, 0xc0000, passthrough_mem_read, passthrough_mem_write, NULL);
+    v3_hook_write_mem(info, 0xa0000, 0xbffff, 0xa0000,  passthrough_mem_write, NULL);
   }  
   
   // TEMP
   //add_shadow_region_passthrough(info, 0xc0000, 0xc8000, 0xc0000);
   
   if (1) {
-    add_shadow_region_passthrough(info, 0xc7000, 0xc8000, (addr_t)V3_AllocPages(1));
-    if (add_shadow_region_passthrough(info, 0xc8000, 0xf0000, (addr_t)V3_AllocPages(40)) == -1) {
+    v3_add_shadow_mem(info, 0xc7000, 0xc8000, (addr_t)V3_AllocPages(1));
+    if (v3_add_shadow_mem(info, 0xc8000, 0xf0000, (addr_t)V3_AllocPages(40)) == -1) {
       PrintDebug("Error adding shadow region\n");
     }
   } else {
-    add_shadow_region_passthrough(info, 0xc0000, 0xc8000, 0xc0000);
-    add_shadow_region_passthrough(info, 0xc8000, 0xf0000, 0xc8000);
+    v3_add_shadow_mem(info, 0xc0000, 0xc8000, 0xc0000);
+    v3_add_shadow_mem(info, 0xc8000, 0xf0000, 0xc8000);
   }
   
   
   if (1) {
-  add_shadow_region_passthrough(info, 0x100000, 0x1000000, (addr_t)V3_AllocPages(4096));
+    v3_add_shadow_mem(info, 0x100000, 0x1000000, (addr_t)V3_AllocPages(4096));
   } else {
     /* MEMORY HOOK TEST */
-    add_shadow_region_passthrough(info, 0x100000, 0xa00000, (addr_t)V3_AllocPages(2304));
-    hook_guest_mem(info, 0xa00000, 0xa01000, mem_test_read, passthrough_mem_write, NULL); 
-    add_shadow_region_passthrough(info, 0xa01000, 0x1000000, (addr_t)V3_AllocPages(1791));
+    v3_add_shadow_mem(info, 0x100000, 0xa00000, (addr_t)V3_AllocPages(2304));
+    v3_hook_write_mem(info, 0xa00000, 0xa01000, (addr_t)V3_AllocPages(1), passthrough_mem_write, NULL); 
+    v3_add_shadow_mem(info, 0xa01000, 0x1000000, (addr_t)V3_AllocPages(1791));
   }
 
-    add_shadow_region_passthrough(info, 0x1000000, 0x8000000, (addr_t)V3_AllocPages(32768));
+  v3_add_shadow_mem(info, 0x1000000, 0x8000000, (addr_t)V3_AllocPages(32768));
  
   // test - give linux accesss to PCI space - PAD
-  add_shadow_region_passthrough(info, 0xc0000000,0xffffffff,0xc0000000);
+  v3_add_shadow_mem(info, 0xc0000000,0xffffffff,0xc0000000);
   
   
-  print_shadow_map(&(info->mem_map));
+  print_shadow_map(info);
 
   return 0;
 }
index cc11c04..7b637ed 100644 (file)
@@ -62,19 +62,21 @@ int v3_handle_cr0_write(struct guest_info * info) {
     return -1;
   }
 
-  if (v3_opcode_cmp(V3_OPCODE_LMSW, (const uchar_t *)(dec_instr.opcode)) == 0) {
+  if (dec_instr.op_type == V3_OP_LMSW) {
+  //  if (v3_opcode_cmp(V3_OPCODE_LMSW, (const uchar_t *)(dec_instr.opcode)) == 0) {
 
     if (handle_lmsw(info, &dec_instr) == -1) {
       return -1;
     }
 
-  } else if (v3_opcode_cmp(V3_OPCODE_MOV2CR, (const uchar_t *)(dec_instr.opcode)) == 0) {
-
+    //  } else if (v3_opcode_cmp(V3_OPCODE_MOV2CR, (const uchar_t *)(dec_instr.opcode)) == 0) {
+  } else if (dec_instr.op_type == V3_OP_MOV2CR) {
     if (handle_mov_to_cr0(info, &dec_instr) == -1) {
       return -1;
     }
 
-  } else if (v3_opcode_cmp(V3_OPCODE_CLTS, (const uchar_t *)(dec_instr.opcode)) == 0) {
+    //  } else if (v3_opcode_cmp(V3_OPCODE_CLTS, (const uchar_t *)(dec_instr.opcode)) == 0) {
+  } else if (dec_instr.op_type == V3_OP_CLTS) {
 
     if (handle_clts(info, &dec_instr) == -1) {
       return -1;
@@ -225,7 +227,8 @@ int v3_handle_cr0_read(struct guest_info * info) {
     return -1;
   }
   
-  if (v3_opcode_cmp(V3_OPCODE_MOVCR2, (const uchar_t *)(dec_instr.opcode)) == 0) {
+  //  if (v3_opcode_cmp(V3_OPCODE_MOVCR2, (const uchar_t *)(dec_instr.opcode)) == 0) {
+  if (dec_instr.op_type == V3_OP_MOVCR2) {
     struct cr0_32 * dst_reg = (struct cr0_32 *)(dec_instr.dst_operand.operand);
     struct cr0_32 * shadow_cr0 = (struct cr0_32 *)&(info->ctrl_regs.cr0);
 
@@ -240,7 +243,8 @@ int v3_handle_cr0_read(struct guest_info * info) {
 
     PrintDebug("Shadow CR0: %x\n", *(uint_t*)shadow_cr0);    
     PrintDebug("returned CR0: %x\n", *(uint_t*)dst_reg);
-  } else if (v3_opcode_cmp(V3_OPCODE_SMSW, (const uchar_t *)(dec_instr.opcode)) == 0) {
+    //  } else if (v3_opcode_cmp(V3_OPCODE_SMSW, (const uchar_t *)(dec_instr.opcode)) == 0) {
+  } else if (dec_instr.op_type == V3_OP_SMSW) {
     struct cr0_real * shadow_cr0 = (struct cr0_real *)&(info->ctrl_regs.cr0);
     struct cr0_real * dst_reg = (struct cr0_real *)(dec_instr.dst_operand.operand);
     char cr0_val = *(char*)shadow_cr0 & 0x0f;
@@ -283,7 +287,8 @@ int v3_handle_cr3_write(struct guest_info * info) {
     return -1;
   }
 
-  if (v3_opcode_cmp(V3_OPCODE_MOV2CR, (const uchar_t *)(dec_instr.opcode)) == 0) {
+  //  if (v3_opcode_cmp(V3_OPCODE_MOV2CR, (const uchar_t *)(dec_instr.opcode)) == 0) {
+  if (dec_instr.op_type == V3_OP_MOV2CR) {
     PrintDebug("MOV2CR3 (cpu_mode=%s)\n", v3_cpu_mode_to_str(info->cpu_mode));
 
     if (info->shdw_pg_mode == SHADOW_PAGING) {
@@ -359,7 +364,8 @@ int v3_handle_cr3_read(struct guest_info * info) {
     return -1;
   }
 
-  if (v3_opcode_cmp(V3_OPCODE_MOVCR2, (const uchar_t *)(dec_instr.opcode)) == 0) {
+  //  if (v3_opcode_cmp(V3_OPCODE_MOVCR2, (const uchar_t *)(dec_instr.opcode)) == 0) {
+  if (dec_instr.op_type == V3_OP_MOVCR2) {
     PrintDebug("MOVCR32 (mode=%s)\n", v3_cpu_mode_to_str(info->cpu_mode));
 
     if (info->shdw_pg_mode == SHADOW_PAGING) {
@@ -422,7 +428,8 @@ int v3_handle_cr4_write(struct guest_info * info) {
     return -1;
   }
 
-  if (v3_opcode_cmp(V3_OPCODE_MOV2CR, (const uchar_t *)(dec_instr.opcode)) != 0) {
+  //  if (v3_opcode_cmp(V3_OPCODE_MOV2CR, (const uchar_t *)(dec_instr.opcode)) != 0) {
+  if (dec_instr.op_type != V3_OP_MOV2CR) {
     PrintError("Invalid opcode in write to CR4\n");
     return -1;
   }
index 593b038..54c77c4 100644 (file)
@@ -30,6 +30,67 @@ int v3_opcode_cmp(const uchar_t * op1, const uchar_t * op2) {
 }
 
 
+void v3_get_prefixes(uchar_t * instr, struct x86_prefixes * prefixes) {
+  while (1) {
+    switch (*instr) {
+    case 0xF0:      // lock
+      prefixes->lock = 1;
+      break;
+
+    case 0xF2:      // REPNE/REPNZ
+      prefixes->repnz = 1;
+      prefixes->repne = 1;
+      break;
+
+    case 0xF3:      // REP or REPE/REPZ
+      prefixes->rep = 1;
+      prefixes->repe = 1;
+      prefixes->repz = 1; 
+      break;
+
+    case 0x2E:      // CS override or Branch hint not taken (with Jcc instrs)
+      prefixes->cs_override = 1;
+      prefixes->br_not_taken = 1;
+      break;
+
+    case 0x36:      // SS override
+      prefixes->ss_override = 1;
+      break;
+
+    case 0x3E:      // DS override or Branch hint taken (with Jcc instrs)
+      prefixes->ds_override = 1;
+      prefixes->br_taken = 1;
+      break;
+
+    case 0x26:      // ES override
+      prefixes->es_override = 1;
+      break;
+
+    case 0x64:      // FS override
+      prefixes->fs_override = 1;
+      break;
+      
+    case 0x65:      // GS override
+      prefixes->gs_override = 1;
+      break;
+
+    case 0x66:      // operand size override
+      prefixes->op_size = 1;
+      break;
+
+    case 0x67:    // address size override
+      prefixes->addr_size = 1;
+      break;
+
+    default:
+      return;
+    }
+
+    instr++;
+  }
+
+}
+
 void v3_strip_rep_prefix(uchar_t * instr, int length) {
   int read_ctr = 0;
   int write_ctr = 0;
index 97549eb..bb67a5f 100644 (file)
 #include <palacios/vmm_emulator.h>
 #include <palacios/vm_guest_mem.h>
 #include <palacios/vmm_decoder.h>
-#include <palacios/vmm_debug.h>
-#include <palacios/vmcb.h>
-#include <palacios/vmm_ctrl_regs.h>
-
-static const char VMMCALL[3] = {0x0f, 0x01, 0xd9};
+#include <palacios/vmm_paging.h>
+#include <palacios/vmm_instr_emulator.h>
 
 #ifndef DEBUG_EMULATOR
 #undef PrintDebug
@@ -33,120 +30,121 @@ static const char VMMCALL[3] = {0x0f, 0x01, 0xd9};
 #endif
 
 
-int v3_init_emulator(struct guest_info * info) {
-  struct emulation_state * emulator = &(info->emulator);
-
-  emulator->num_emulated_pages = 0;
-  INIT_LIST_HEAD(&(emulator->emulated_pages));
-
 
-  emulator->num_saved_pages = 0;
-  INIT_LIST_HEAD(&(emulator->saved_pages));
-  
-  emulator->num_write_regions = 0;
-  INIT_LIST_HEAD(&(emulator->write_regions));
 
-  emulator->running = 0;
-  emulator->instr_length = 0;
 
-  emulator->tf_enabled = 0;
 
-  return 0;
-}
+// We emulate up to the next 4KB page boundry
+static int emulate_string_write_op(struct guest_info * info, struct x86_instr * dec_instr, 
+                                  addr_t write_gva, addr_t write_gpa, addr_t * dst_addr) {
+  uint_t emulation_length = 0;
+  addr_t tmp_rcx = 0;
+  addr_t src_addr = 0;
 
-static addr_t get_new_page() {
-  void * page = V3_VAddr(V3_AllocPages(1));
-  memset(page, 0, PAGE_SIZE);
+  if (dec_instr->op_type == V3_OP_MOVS) {
+    PrintDebug("MOVS emulation\n");
 
-  return (addr_t)page;
-}
+    if (dec_instr->dst_operand.operand != write_gva) {
+      PrintError("Inconsistency between Pagefault and Instruction Decode XED_ADDR=%p, PF_ADDR=%p\n",
+                (void *)dec_instr->dst_operand.operand, (void *)write_gva);
+      return -1;
+    }
 
-/*
-static int setup_code_page(struct guest_info * info, char * instr, struct basic_instr_info * instr_info ) {
-  addr_t code_page_offset = PT32_PAGE_OFFSET(info->rip);
-  addr_t code_page = get_new_page();
-  struct emulated_page * new_code_page = V3_Malloc(sizeof(struct emulated_page));
-  struct saved_page * saved_code_page = V3_Malloc(sizeof(struct saved_page));
+    emulation_length = ( (dec_instr->str_op_length  < (0x1000 - PAGE_OFFSET_4KB(write_gva))) ? 
+                        dec_instr->str_op_length :
+                        (0x1000 - PAGE_OFFSET_4KB(write_gva)));
+    /* ** Fix emulation length so that it doesn't overrun over the src page either ** */
 
 
-  saved_code_page->va = PT32_PAGE_ADDR(info->rip);
-
-  new_code_page->page_addr = code_page; 
-  new_code_page->va = PT32_PAGE_ADDR(info->rip);
-
-  new_code_page->pte.present = 1;
-  new_code_page->pte.writable = 0;
-  new_code_page->pte.user_page = 1;
-  new_code_page->pte.page_base_addr = PT32_BASE_ADDR(code_page);
-
-  memcpy((void *)(code_page + code_page_offset), instr, instr_info->instr_length);
-  memcpy((void *)(code_page + code_page_offset + instr_info->instr_length), VMMCALL, 3);
-
-#ifdef DEBUG_EMULATOR
-  PrintDebug("New Instr Stream:\n");
-  PrintTraceMemDump((void *)(code_page + code_page_offset), 32);
-  PrintDebug("rip =%x\n", info->rip);
-#endif
+    PrintDebug("STR_OP_LEN: %d, Page Len: %d\n", 
+              (uint_t)dec_instr->str_op_length, 
+              (uint_t)(0x1000 - PAGE_OFFSET_4KB(write_gva)));
+    PrintDebug("Emulation length: %d\n", emulation_length);
+    tmp_rcx = emulation_length;
+    if (guest_pa_to_host_va(info, write_gpa, dst_addr) == -1) {
+      PrintError("Could not translate write destination to host VA\n");
+      return -1;
+    }
 
+    // figure out addresses here....
+    if (info->mem_mode == PHYSICAL_MEM) {
+      if (guest_pa_to_host_va(info, dec_instr->src_operand.operand, &src_addr) == -1) {
+       PrintError("Could not translate write Source (Physical) to host VA\n");
+       return -1;
+      }
+    } else {
+      if (guest_va_to_host_va(info, dec_instr->src_operand.operand, &src_addr) == -1) {
+       PrintError("Could not translate write Source (Virtual) to host VA\n");
+       return -1;
+      }
+    }
 
 
-  v3_replace_shdw_page32(info, new_code_page->va, &(new_code_page->pte), &(saved_code_page->pte));
+    PrintDebug("Dst Operand: %p (size=%d), Src Operand: %p\n", 
+              (void *)dec_instr->dst_operand.operand, 
+              dec_instr->dst_operand.size,
+              (void *)dec_instr->src_operand.operand);
+    PrintDebug("Dst Addr: %p, Src Addr: %p\n", (void *)(addr_t *)*dst_addr, (void *)src_addr);
 
+    //return -1;
 
-  list_add(&(new_code_page->page_list), &(info->emulator.emulated_pages));
-  info->emulator.num_emulated_pages++;
 
-  list_add(&(saved_code_page->page_list), &(info->emulator.saved_pages));
-  info->emulator.num_saved_pages++;
 
-  return 0;
-}
-*/
 
+    if (dec_instr->dst_operand.size == 1) {
+      movs8(dst_addr, &src_addr, &tmp_rcx, (addr_t *)&(info->ctrl_regs.rflags));
+    } else if (dec_instr->dst_operand.size == 2) {
+      movs16(dst_addr, &src_addr, &tmp_rcx, (addr_t *)&(info->ctrl_regs.rflags));
+    } else if (dec_instr->dst_operand.size == 4) {
+      movs32(dst_addr, &src_addr, &tmp_rcx, (addr_t *)&(info->ctrl_regs.rflags));
+    } else {
+      PrintError("Invalid operand length\n");
+      return -1;
+    }
 
-static int set_stepping(struct guest_info * info) {
-  vmcb_ctrl_t * ctrl_area = GET_VMCB_CTRL_AREA((vmcb_t*)(info->vmm_data));
-  ctrl_area->exceptions.db = 1;
 
-  info->emulator.tf_enabled = ((struct rflags *)&(info->ctrl_regs.rflags))->tf;
 
-  ((struct rflags *)&(info->ctrl_regs.rflags))->tf = 1;
 
-  return 0;
-}
 
+    PrintDebug("RDI=%p, RSI=%p, RCX=%p\n", 
+              (void *)*(addr_t *)&(info->vm_regs.rdi), 
+              (void *)*(addr_t *)&(info->vm_regs.rsi), 
+              (void *)*(addr_t *)&(info->vm_regs.rcx)); 
 
-static int unset_stepping(struct guest_info * info) {
-  vmcb_ctrl_t * ctrl_area = GET_VMCB_CTRL_AREA((vmcb_t*)(info->vmm_data));
-  ctrl_area->exceptions.db = 0;
+    info->vm_regs.rdi += emulation_length;
+    info->vm_regs.rsi += emulation_length;
+    info->vm_regs.rcx -= emulation_length;
+    
+    PrintDebug("RDI=%p, RSI=%p, RCX=%p\n", 
+              (void *)*(addr_t *)&(info->vm_regs.rdi), 
+              (void *)*(addr_t *)&(info->vm_regs.rsi), 
+              (void *)*(addr_t *)&(info->vm_regs.rcx)); 
 
-  ((struct rflags *)&(info->ctrl_regs.rflags))->tf = info->emulator.tf_enabled;
+    if (emulation_length == dec_instr->str_op_length) {
+      info->rip += dec_instr->instr_length;
+    }
 
-  if (info->emulator.tf_enabled) {
-    // Inject breakpoint exception into guest
+    return emulation_length;
   }
 
-  return 0;
+  
+
 
+  return -1;
 }
 
 
-// get the current instr
-// check if rep + remove
-// put into new page, vmexit after
-// replace new page with current eip page
-// 
-int v3_emulate_memory_read(struct guest_info * info, addr_t read_gva, 
-                          int (*read)(addr_t read_addr, void * dst, uint_t length, void * priv_data), 
-                          addr_t read_gpa, void * private_data) {
-  struct basic_instr_info instr_info;
+int v3_emulate_write_op(struct guest_info * info, addr_t write_gva, addr_t write_gpa, addr_t * dst_addr) {
+  struct x86_instr dec_instr;
   uchar_t instr[15];
-  int ret;
-  struct emulated_page * data_page = V3_Malloc(sizeof(struct emulated_page));
-  addr_t data_addr_offset = PAGE_OFFSET(read_gva);
-  pte32_t saved_pte;
+  int ret = 0;
+  addr_t src_addr = 0;
+
 
-  PrintDebug("Emulating Read\n");
+  PrintDebug("Emulating Write for instruction at %p\n", (void *)(addr_t)(info->rip));
+  PrintDebug("GPA=%p, GVA=%p\n", (void *)write_gpa, (void *)write_gva);
 
   if (info->mem_mode == PHYSICAL_MEM) { 
     ret = read_guest_pa_memory(info, get_addr_linear(info, info->rip, &(info->segments.cs)), 15, instr);
@@ -155,233 +153,258 @@ int v3_emulate_memory_read(struct guest_info * info, addr_t read_gva,
   }
 
   if (ret == -1) {
-    PrintError("Could not read guest memory\n");
     return -1;
   }
 
-#ifdef DEBUG_EMULATOR
-  PrintDebug("Instr (15 bytes) at %p:\n", (void *)(addr_t)instr);
-  PrintTraceMemDump(instr, 15);
-#endif  
 
-
-  if (v3_basic_mem_decode(info, (addr_t)instr, &instr_info) == -1) {
-    PrintError("Could not do a basic memory instruction decode\n");
-    V3_Free(data_page);
+  if (guest_pa_to_host_va(info, write_gpa, dst_addr) == -1) {
+    PrintError("Could not translate write destination to host VA\n");
     return -1;
   }
 
-  /*
-  if (instr_info.has_rep == 1) {
-    PrintError("We currently don't handle rep* instructions\n");
-    V3_Free(data_page);
+  if (v3_decode(info, (addr_t)instr, &dec_instr) == -1) {
+    PrintError("Decoding Error\n");
+    // Kick off single step emulator
     return -1;
   }
-  */
-
-  data_page->page_addr = get_new_page();
-  data_page->va = PAGE_ADDR(read_gva);
-  data_page->pte.present = 1;
-  data_page->pte.writable = 0;
-  data_page->pte.user_page = 1;
-  data_page->pte.page_base_addr = PAGE_BASE_ADDR((addr_t)V3_PAddr((void *)(addr_t)(data_page->page_addr)));
-
-
-  // Read the data directly onto the emulated page
-  ret = read(read_gpa, (void *)(data_page->page_addr + data_addr_offset), instr_info.op_size, private_data);
-  if ((ret == -1) || ((uint_t)ret != instr_info.op_size)) {
-    PrintError("Read error in emulator\n");
-    V3_FreePage((void *)V3_PAddr((void *)(data_page->page_addr)));
-    V3_Free(data_page);
-    return -1;
+  
+  if (dec_instr.is_str_op) {
+    return emulate_string_write_op(info, &dec_instr, write_gva, write_gpa, dst_addr);
   }
 
-  v3_replace_shdw_page32(info, data_page->va, &(data_page->pte), &saved_pte);
-
-
-  list_add(&(data_page->page_list), &(info->emulator.emulated_pages));
-  info->emulator.num_emulated_pages++;
 
-  if (saved_pte.present == 1) {
-    struct saved_page * saved_data_page = V3_Malloc(sizeof(struct saved_page));
-    saved_data_page->pte = saved_pte;
-    saved_data_page->va = PAGE_ADDR(read_gva);
-
-    list_add(&(saved_data_page->page_list), &(info->emulator.saved_pages));
-    info->emulator.num_saved_pages++;
+  if ((dec_instr.dst_operand.type != MEM_OPERAND) ||
+      (dec_instr.dst_operand.operand != write_gva)) {
+    PrintError("Inconsistency between Pagefault and Instruction Decode XED_ADDR=%p, PF_ADDR=%p\n",
+              (void *)dec_instr.dst_operand.operand, (void *)write_gva);
+    return -1;
   }
 
 
-  // setup_code_page(info, instr, &instr_info);
-  set_stepping(info);
-
-  info->emulator.running = 1;
-  info->run_state = VM_EMULATING;
-  info->emulator.instr_length = instr_info.instr_length;
-
-  return 0;
-}
-
-
-
-int v3_emulate_memory_write(struct guest_info * info, addr_t write_gva,
-                           int (*write)(addr_t write_addr, void * src, uint_t length, void * priv_data), 
-                           addr_t write_gpa, void * private_data) {
-
-  struct basic_instr_info instr_info;
-  uchar_t instr[15];
-  int ret;
-  struct write_region * write_op = V3_Malloc(sizeof(struct write_region ));
-  struct emulated_page * data_page = V3_Malloc(sizeof(struct emulated_page));
-  addr_t data_addr_offset = PAGE_OFFSET(write_gva);
-  pte32_t saved_pte;
-  int i;
-
-  PrintDebug("Emulating Write for instruction at 0x%p\n", (void *)(addr_t)(info->rip));
-
-  if (info->mem_mode == PHYSICAL_MEM) { 
-    ret = read_guest_pa_memory(info, get_addr_linear(info, info->rip, &(info->segments.cs)), 15, instr);
-  } else { 
-    ret = read_guest_va_memory(info, get_addr_linear(info, info->rip, &(info->segments.cs)), 15, instr);
+  if (dec_instr.src_operand.type == MEM_OPERAND) {
+    if (info->mem_mode == PHYSICAL_MEM) {
+      if (guest_pa_to_host_va(info, dec_instr.src_operand.operand, &src_addr) == -1) {
+       PrintError("Could not translate write Source (Physical) to host VA\n");
+       return -1;
+      }
+    } else {
+      if (guest_va_to_host_va(info, dec_instr.src_operand.operand, &src_addr) == -1) {
+       PrintError("Could not translate write Source (Virtual) to host VA\n");
+       return -1;
+      }
+    }
+  } else if (dec_instr.src_operand.type == REG_OPERAND) {
+    src_addr = dec_instr.src_operand.operand;
+  } else {
+    src_addr = (addr_t)&(dec_instr.src_operand.operand);
   }
 
 
-  PrintDebug("Instruction is");
-  for (i=0;i<15;i++) { PrintDebug(" 0x%x",instr[i]); } 
-  PrintDebug("\n");
-  
-  if (v3_basic_mem_decode(info, (addr_t)instr, &instr_info) == -1) {
-    PrintError("Could not do a basic memory instruction decode\n");
-    V3_Free(write_op);
-    V3_Free(data_page);
+  PrintDebug("Dst_Addr Ptr = %p (val=%p), SRC operand = %p\n", 
+            (void *)dst_addr, (void *)*dst_addr, (void *)src_addr);
+
+
+  if (dec_instr.dst_operand.size == 1) {
+
+    switch (dec_instr.op_type) {
+    case V3_OP_ADC:
+      adc8((addr_t *)*dst_addr, (addr_t *)src_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_ADD:
+      add8((addr_t *)*dst_addr, (addr_t *)src_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_AND:
+      and8((addr_t *)*dst_addr, (addr_t *)src_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_OR:
+      or8((addr_t *)*dst_addr, (addr_t *)src_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_XOR:
+      xor8((addr_t *)*dst_addr, (addr_t *)src_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SUB:
+      sub8((addr_t *)*dst_addr, (addr_t *)src_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+
+    case V3_OP_MOV:
+      mov8((addr_t *)*dst_addr, (addr_t *)src_addr);
+      break;
+    case V3_OP_NOT:
+      not8((addr_t *)*dst_addr);
+      break;
+    case V3_OP_XCHG:
+      xchg8((addr_t *)*dst_addr, (addr_t *)src_addr);
+      break;
+      
+
+    case V3_OP_INC:
+      inc8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_DEC:
+      dec8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_NEG:
+      neg8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SETB:
+      setb8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SETBE:
+      setbe8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SETL:
+      setl8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SETLE:
+      setle8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SETNB:
+      setnb8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SETNBE:
+      setnbe8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SETNL:
+      setnl8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SETNLE:
+      setnle8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SETNO:
+      setno8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SETNP:
+      setnp8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SETNS:
+      setns8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SETNZ:
+      setnz8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SETO:
+      seto8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SETP:
+      setp8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SETS:
+      sets8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SETZ:
+      setz8((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+
+    default:
+      PrintError("Unknown 8 bit instruction\n");
+      return -1;
+    }
+
+  } else if (dec_instr.dst_operand.size == 2) {
+
+    switch (dec_instr.op_type) {
+    case V3_OP_ADC:
+      adc16((addr_t *)*dst_addr, (addr_t *)src_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_ADD:
+      add16((addr_t *)*dst_addr, (addr_t *)src_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_AND:
+      and16((addr_t *)*dst_addr, (addr_t *)src_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_OR:
+      or16((addr_t *)*dst_addr, (addr_t *)src_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_XOR:
+      xor16((addr_t *)*dst_addr, (addr_t *)src_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SUB:
+      sub16((addr_t *)*dst_addr, (addr_t *)src_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+
+
+    case V3_OP_INC:
+      inc16((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_DEC:
+      dec16((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_NEG:
+      neg16((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+
+    case V3_OP_MOV:
+      mov16((addr_t *)*dst_addr, (addr_t *)src_addr);
+      break;
+    case V3_OP_NOT:
+      not16((addr_t *)*dst_addr);
+      break;
+    case V3_OP_XCHG:
+      xchg16((addr_t *)*dst_addr, (addr_t *)src_addr);
+      break;
+      
+    default:
+      PrintError("Unknown 16 bit instruction\n");
+      return -1;
+    }
+
+  } else if (dec_instr.dst_operand.size == 4) {
+
+    switch (dec_instr.op_type) {
+    case V3_OP_ADC:
+      adc32((addr_t *)*dst_addr, (addr_t *)src_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_ADD:
+      add32((addr_t *)*dst_addr, (addr_t *)src_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_AND:
+      and32((addr_t *)*dst_addr, (addr_t *)src_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_OR:
+      or32((addr_t *)*dst_addr, (addr_t *)src_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_XOR:
+      xor32((addr_t *)*dst_addr, (addr_t *)src_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_SUB:
+      sub32((addr_t *)*dst_addr, (addr_t *)src_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+
+    case V3_OP_INC:
+      inc32((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_DEC:
+      dec32((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+    case V3_OP_NEG:
+      neg32((addr_t *)*dst_addr, (addr_t *)&(info->ctrl_regs.rflags));
+      break;
+
+    case V3_OP_MOV:
+      mov32((addr_t *)*dst_addr, (addr_t *)src_addr);
+      break;
+    case V3_OP_NOT:
+      not32((addr_t *)*dst_addr);
+      break;
+    case V3_OP_XCHG:
+      xchg32((addr_t *)*dst_addr, (addr_t *)src_addr);
+      break;
+      
+    default:
+      PrintError("Unknown 32 bit instruction\n");
+      return -1;
+    }
+
+  } else if (dec_instr.dst_operand.size == 8) {
+    PrintError("64 bit instructions not handled\n");
     return -1;
-  }
-
-  if (instr_info.has_rep==1) { 
-    PrintDebug("Emulated instruction has rep\n");
-  }
-
-  /*
-  if (instr_info.has_rep == 1) {
-    PrintError("We currently don't handle rep* instructions\n");
-    V3_Free(write_op);
-    V3_Free(data_page);
+  } else {
+    PrintError("Invalid Operation Size\n");
     return -1;
   }
-  */
 
-  data_page->page_addr = get_new_page();
-  data_page->va = PAGE_ADDR(write_gva);
-  data_page->pte.present = 1;
-  data_page->pte.writable = 1;
-  data_page->pte.user_page = 1;
-  data_page->pte.page_base_addr = PAGE_BASE_ADDR((addr_t)V3_PAddr((void *)(addr_t)(data_page->page_addr)));
+  info->rip += dec_instr.instr_length;
 
-
-
-  write_op->write = write;
-  write_op->write_addr = write_gpa;
-  write_op->length = instr_info.op_size;
-  write_op->private_data = private_data;
-
-  write_op->write_data = (void *)(data_page->page_addr + data_addr_offset);
-
-  list_add(&(write_op->write_list), &(info->emulator.write_regions));
-  info->emulator.num_write_regions--;
-
-  v3_replace_shdw_page32(info, data_page->va, &(data_page->pte), &saved_pte);
-
-
-  list_add(&(data_page->page_list), &(info->emulator.emulated_pages));
-  info->emulator.num_emulated_pages++;
-
-  if (saved_pte.present == 1) {
-    struct saved_page * saved_data_page = V3_Malloc(sizeof(struct saved_page));
-    saved_data_page->pte = saved_pte;
-    saved_data_page->va = PAGE_ADDR(write_gva);
-
-    list_add(&(saved_data_page->page_list), &(info->emulator.saved_pages));
-    info->emulator.num_saved_pages++;
-  }
-
-
-  if (info->emulator.running == 0) {
-    //    setup_code_page(info, instr, &instr_info);
-    set_stepping(info);
-    info->emulator.running = 1;
-    info->run_state = VM_EMULATING;
-    info->emulator.instr_length = instr_info.instr_length;
-  }
-
-  return 0;
+  return dec_instr.dst_operand.size;
 }
 
-
-// end emulation
-int v3_emulation_exit_handler(struct guest_info * info) {
-  struct saved_page * svpg, * p_svpg;
-  struct emulated_page * empg, * p_empg;
-  struct write_region * wr_reg, * p_wr_reg;
-  pte32_t dummy_pte;
-
-  // Complete the writes
-  // delete writes
-  // swap out emulated pages with blank dummies
-  // swap in saved pages
-  // increment rip
-  
-  PrintDebug("V3 Emulation Exit Handler\n");
-
-  list_for_each_entry_safe(wr_reg, p_wr_reg, &(info->emulator.write_regions), write_list) {
-    wr_reg->write(wr_reg->write_addr, wr_reg->write_data, wr_reg->length, wr_reg->private_data);
-    PrintDebug("Writing \n");
-    
-    list_del(&(wr_reg->write_list));
-    V3_Free(wr_reg);
-
-  }
-  info->emulator.num_write_regions = 0;
-
-
-  *(uint_t *)&dummy_pte = 0;
-  
-  list_for_each_entry_safe(empg, p_empg, &(info->emulator.emulated_pages), page_list) {
-    pte32_t empte32_t;
-
-    PrintDebug("wiping page %p\n", (void *)(addr_t)(empg->va)); 
-
-    v3_replace_shdw_page32(info, empg->va, &dummy_pte, &empte32_t);
-    V3_FreePage((void *)(V3_PAddr((void *)(empg->page_addr))));
-
-    list_del(&(empg->page_list));
-    V3_Free(empg);
-  }
-  info->emulator.num_emulated_pages = 0;
-
-  list_for_each_entry_safe(svpg, p_svpg, &(info->emulator.saved_pages), page_list) {
-
-    PrintDebug("Setting Saved page %p back\n", (void *)(addr_t)(svpg->va)); 
-    v3_replace_shdw_page32(info, empg->va, &(svpg->pte), &dummy_pte);
-    
-    list_del(&(svpg->page_list));
-    V3_Free(svpg);
-  }
-  info->emulator.num_saved_pages = 0;
-
-  info->run_state = VM_RUNNING;
-  info->emulator.running = 0;
-  //info->rip += info->emulator.instr_length;
-
-
-  PrintDebug("Returning to rip: 0x%p\n", (void *)(addr_t)(info->rip));
-
-  info->emulator.instr_length = 0;
-  
-  
-  unset_stepping(info);
-
-
-  PrintDebug("returning from emulation\n");
-
-  return 0;
-}
index 612ae78..fa9aba3 100644 (file)
@@ -394,6 +394,55 @@ int hashtable_change(struct hashtable * htable, addr_t key, addr_t value, int fr
 
 
 
+int hashtable_inc(struct hashtable * htable, addr_t key, addr_t value) {
+    struct hash_entry * tmp_entry;
+    uint_t hash_value;
+    uint_t index;
+
+    hash_value = do_hash(htable, key);
+
+    index = indexFor(htable->table_length, hash_value);
+
+    tmp_entry = htable->table[index];
+
+    while (tmp_entry != NULL) {
+        /* Check hash value to short circuit heavier comparison */
+        if ((hash_value == tmp_entry->hash) && (htable->eq_fn(key, tmp_entry->key))) {
+
+         tmp_entry->value += value;
+         return -1;
+        }
+        tmp_entry = tmp_entry->next;
+    }
+    return 0;
+}
+
+
+int hashtable_dec(struct hashtable * htable, addr_t key, addr_t value) {
+    struct hash_entry * tmp_entry;
+    uint_t hash_value;
+    uint_t index;
+
+    hash_value = do_hash(htable, key);
+
+    index = indexFor(htable->table_length, hash_value);
+
+    tmp_entry = htable->table[index];
+
+    while (tmp_entry != NULL) {
+        /* Check hash value to short circuit heavier comparison */
+        if ((hash_value == tmp_entry->hash) && (htable->eq_fn(key, tmp_entry->key))) {
+
+         tmp_entry->value -= value;
+         return -1;
+        }
+        tmp_entry = tmp_entry->next;
+    }
+    return 0;
+}
+
+
+
 
 /*****************************************************************************/
 /* returns value associated with key */
diff --git a/palacios/src/palacios/vmm_instr_emulator.c b/palacios/src/palacios/vmm_instr_emulator.c
deleted file mode 100644 (file)
index f89d4e6..0000000
+++ /dev/null
@@ -1,101 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-
-#define addr_t unsigned long
-
-
-
-
-void test() {
-  int a = 10;
-  int b;
-
-  asm ("movl %1, %%eax\n\t"
-       "movl %%eax, %0\n\t"
-       :"=r"(b)        /* output */
-       :"r"(a)         /* input */
-       :"%eax"         /* clobbered register */
-       );       
-
-}
-
-void get_flags(addr_t * flags) {
-  addr_t tmp;
-  asm  ("pushfq\n\t"
-       "pop %0\n\t"
-       :"=r"(tmp)
-       : 
-       );
-
-  *flags = tmp;
-}
-
-void adc64(int * dst, int * src, addr_t * flags) {
-  int tmp_dst = *dst, tmp_src = *src;
-  addr_t tmp_flags = *flags;
-
-  char * inst = "adcl";
-
-  // Some of the flags values are not copied out in a pushf, we save them here
-  addr_t flags_rsvd = *flags & ~0xfffe7fff;
-
-  asm volatile (
-       "pushfq\r\n"
-       "push %3\r\n"
-       "popfq\r\n"
-       "adcl %2, %0\r\n"
-       "pushfq\r\n"
-       "pop %1\r\n"
-       "popfq\r\n"
-       : "=a"(tmp_dst),"=c"(tmp_flags)
-       : "b"(tmp_src),"c"(tmp_flags), "0"(tmp_dst)
-    );
-
-  *dst = tmp_dst;
-  *flags = tmp_flags;
-  *flags |= flags_rsvd;
-
-}
-
-
-void adc32(int * dst, int * src, addr_t * flags) {
-  int tmp_dst = *dst, tmp_src = *src;
-  addr_t tmp_flags = *flags;
-
-  
-  asm volatile (
-       "pushfd\r\n"
-       "push %3\r\n"
-       "popfd\r\n"
-       "adcl %2, %0\r\n"
-       "pushfd\r\n"
-       "pop %1\r\n"
-       "popfd\r\n"
-       : "=a"(tmp_dst),"=c"(tmp_flags)
-       : "b"(tmp_src),"c"(tmp_flags), "0"(tmp_dst)
-    );
-
-  *dst = tmp_dst;
-  *flags = tmp_flags;
-
-}
-
-
-int main(int argc, char ** argv) {
-  addr_t flags;
-  int dest = 4;
-  int src = 5;
-  
-  printf("sizeof ulong: %d\n", sizeof(unsigned long));
-
-  printf("Getting flags\n");
-  get_flags(&flags);
-  flags = flags | 0x1;
-
-  printf("Flags=0x%x\n", flags);
-  test();
-  printf("Adding\n");
-  adc64(&dest, &src, &flags);
-  printf("Result=%d\n", dest);
-
-}
index 514d8e1..f7bea5a 100644 (file)
@@ -34,89 +34,79 @@ static int default_write(ushort_t port, void *src, uint_t length, void * priv_da
 static int default_read(ushort_t port, void * dst, uint_t length, void * priv_data);
 
 
-void v3_init_vmm_io_map(struct guest_info * info) {
-  struct vmm_io_map * io_map = &(info->io_map);
-  io_map->num_ports = 0;
-  io_map->head = NULL;
+void v3_init_io_map(struct guest_info * info) {
+  info->io_map.rb_node = NULL;
 }
 
 
 
 
 
-static int add_io_hook(struct vmm_io_map * io_map, struct vmm_io_hook * io_hook) {
+static inline struct v3_io_hook * __insert_io_hook(struct guest_info * info, struct v3_io_hook * hook) {
+  struct rb_node ** p = &(info->io_map.rb_node);
+  struct rb_node * parent = NULL;
+  struct v3_io_hook * tmp_hook = NULL;
 
-  if (!(io_map->head)) {
-    io_map->head = io_hook;
-    io_map->num_ports = 1;
-    return 0;
-  } else if (io_map->head->port > io_hook->port) {
-    io_hook->next = io_map->head;
+  while (*p) {
+    parent = *p;
+    tmp_hook = rb_entry(parent, struct v3_io_hook, tree_node);
 
-    io_map->head->prev = io_hook;
-    io_map->head = io_hook;
-    io_map->num_ports++;
-
-    return 0;
-  } else {
-    struct vmm_io_hook * tmp_hook = io_map->head;
-    
-    while ((tmp_hook->next)  && 
-          (tmp_hook->next->port <= io_hook->port)) {
-       tmp_hook = tmp_hook->next;
-    }
-    
-    if (tmp_hook->port == io_hook->port) {
-      //tmp_hook->read = io_hook->read;
-      //tmp_hook->write = io_hook->write;
-      //V3_Free(io_hook);
-      return -1;
+    if (hook->port < tmp_hook->port) {
+      p = &(*p)->rb_left;
+    } else if (hook->port > tmp_hook->port) {
+      p = &(*p)->rb_right;
     } else {
-      io_hook->prev = tmp_hook;
-      io_hook->next = tmp_hook->next;
+      return tmp_hook;
+    }
+  }
+  rb_link_node(&(hook->tree_node), parent, p);
 
-      if (tmp_hook->next) {
-       tmp_hook->next->prev = io_hook;
-      }
+  return NULL;
+}
 
-      tmp_hook->next = io_hook;
 
-      io_map->num_ports++;
-      return 0;
-    }
+static inline struct v3_io_hook * insert_io_hook(struct guest_info * info, struct v3_io_hook * hook) {
+  struct v3_io_hook * ret;
+
+  if ((ret = __insert_io_hook(info, hook))) {
+    return ret;
   }
-  return -1;
+
+  v3_rb_insert_color(&(hook->tree_node), &(info->io_map));
+
+  return NULL;
 }
 
-static int remove_io_hook(struct vmm_io_map * io_map, struct vmm_io_hook * io_hook) {
-  if (io_map->head == io_hook) {
-    io_map->head = io_hook->next;
-  } else if (io_hook->prev) {
-    io_hook->prev->next = io_hook->next;
-  } else {
-    return -1;
-    // data corruption failure
-  }
-  
-  if (io_hook->next) {
-    io_hook->next->prev = io_hook->prev;
-  }
 
-  io_map->num_ports--;
+struct v3_io_hook * v3_get_io_hook(struct guest_info * info, uint_t port) {
+  struct rb_node * n = info->io_map.rb_node;
+  struct v3_io_hook * hook = NULL;
 
-  return 0;
+  while (n) {
+    hook = rb_entry(n, struct v3_io_hook, tree_node);
+    
+    if (port < hook->port) {
+      n = n->rb_left;
+    } else if (port > hook->port) {
+      n = n->rb_right;
+    } else {
+      return hook;
+    }
+  }
+
+  return NULL;
 }
 
 
 
 
 
+
 int v3_hook_io_port(struct guest_info * info, uint_t port, 
                    int (*read)(ushort_t port, void * dst, uint_t length, void * priv_data),
                    int (*write)(ushort_t port, void * src, uint_t length, void * priv_data), 
                    void * priv_data) {
-  struct vmm_io_map * io_map = &(info->io_map);
-  struct vmm_io_hook * io_hook = (struct vmm_io_hook *)V3_Malloc(sizeof(struct vmm_io_hook));
+  struct v3_io_hook * io_hook = (struct v3_io_hook *)V3_Malloc(sizeof(struct v3_io_hook));
 
   io_hook->port = port;
 
@@ -132,12 +122,10 @@ int v3_hook_io_port(struct guest_info * info, uint_t port,
     io_hook->write = write;
   }
 
-  io_hook->next = NULL;
-  io_hook->prev = NULL;
 
   io_hook->priv_data = priv_data;
 
-  if (add_io_hook(io_map, io_hook) != 0) {
+  if (insert_io_hook(info, io_hook)) {
     V3_Free(io_hook);
     return -1;
   }
@@ -146,40 +134,37 @@ int v3_hook_io_port(struct guest_info * info, uint_t port,
 }
 
 int v3_unhook_io_port(struct guest_info * info, uint_t port) {
-  struct vmm_io_map * io_map = &(info->io_map);
-  struct vmm_io_hook * hook = v3_get_io_hook(io_map, port);
+  struct v3_io_hook * hook = v3_get_io_hook(info, port);
 
   if (hook == NULL) {
     return -1;
   }
 
-  remove_io_hook(io_map, hook);
+  v3_rb_erase(&(hook->tree_node), &(info->io_map));
+
+  V3_Free(hook);
+
   return 0;
 }
 
 
-struct vmm_io_hook * v3_get_io_hook(struct vmm_io_map * io_map, uint_t port) {
-  struct vmm_io_hook * tmp_hook;
-  FOREACH_IO_HOOK(*io_map, tmp_hook) {
-    if (tmp_hook->port == port) {
-      return tmp_hook;
-    }
-  }
-  return NULL;
-}
 
 
 
-void v3_print_io_map(struct vmm_io_map * io_map) {
-  struct vmm_io_hook * iter = io_map->head;
+
+void v3_print_io_map(struct guest_info * info) {
+  struct v3_io_hook * tmp_hook = NULL;
+  struct rb_node * node = v3_rb_first(&(info->io_map));
 
   PrintDebug("VMM IO Map (Entries=%d)\n", io_map->num_ports);
 
-  while (iter) {
+  do {
+    tmp_hook = rb_entry(node, struct v3_io_hook, tree_node);
+
     PrintDebug("IO Port: %hu (Read=%p) (Write=%p)\n", 
-              iter->port, 
-              (void *)(iter->read), (void *)(iter->write));
-  }
+              tmp_hook->port, 
+              (void *)(tmp_hook->read), (void *)(tmp_hook->write));
+  } while ((node = v3_rb_next(node)));
 }
 
 
index 1b9f16d..62bca66 100644 (file)
 
 
 
-void init_shadow_region(struct shadow_region * entry,
-                       addr_t               guest_addr_start,
-                       addr_t               guest_addr_end,
-                       guest_region_type_t  guest_region_type,
-                       host_region_type_t   host_region_type)
-{
-  entry->guest_type = guest_region_type;
-  entry->guest_start = guest_addr_start;
-  entry->guest_end = guest_addr_end;
-  entry->host_type = host_region_type;
-  entry->host_addr = 0;
-  entry->next=entry->prev = NULL;
-}
-
-int add_shadow_region_passthrough( struct guest_info *  guest_info,
-                                  addr_t               guest_addr_start,
-                                  addr_t               guest_addr_end,
-                                  addr_t               host_addr)
-{
-  struct shadow_region * entry = (struct shadow_region *)V3_Malloc(sizeof(struct shadow_region));
-
-  init_shadow_region(entry, guest_addr_start, guest_addr_end, 
-                    GUEST_REGION_PHYSICAL_MEMORY, HOST_REGION_PHYSICAL_MEMORY);
-  entry->host_addr = host_addr;
-
-  return add_shadow_region(&(guest_info->mem_map), entry);
-}
 
-int hook_guest_mem(struct guest_info * info, addr_t guest_addr_start, addr_t guest_addr_end,
-                  int (*read)(addr_t guest_addr, void * dst, uint_t length, void * priv_data),
-                  int (*write)(addr_t guest_addr, void * src, uint_t length, void * priv_data),
-                  void * priv_data) {
-  
-  struct shadow_region * entry = (struct shadow_region *)V3_Malloc(sizeof(struct shadow_region));
-  struct vmm_mem_hook * hook = (struct vmm_mem_hook *)V3_Malloc(sizeof(struct vmm_mem_hook));
 
-  memset(hook, 0, sizeof(struct vmm_mem_hook));
 
-  hook->read = read;
-  hook->write = write;
-  hook->region = entry;
-  hook->priv_data = priv_data;
 
+static inline
+struct v3_shadow_region * insert_shadow_region(struct guest_info * info, 
+                                              struct v3_shadow_region * region);
 
-  init_shadow_region(entry, guest_addr_start, guest_addr_end, 
-                    GUEST_REGION_PHYSICAL_MEMORY, HOST_REGION_HOOK);
 
-  entry->host_addr = (addr_t)hook;
 
-  return add_shadow_region(&(info->mem_map), entry);
+void v3_init_shadow_map(struct guest_info * info) {
+  info->mem_map.rb_node = NULL;
 }
 
+void v3_delete_shadow_map(struct guest_info * info) {
+  struct rb_node * node = v3_rb_first(&(info->mem_map));
+  struct v3_shadow_region * reg;
+  struct rb_node * tmp_node = NULL;
+  
+   while (node) {
+    reg = rb_entry(node, struct v3_shadow_region, tree_node);
+    tmp_node = node;
+    node = v3_rb_next(node);
 
-struct vmm_mem_hook * get_mem_hook(struct guest_info * info, addr_t guest_addr) {
-  struct shadow_region * region = get_shadow_region_by_addr(&(info->mem_map), guest_addr);
-
-  if (region == NULL) {
-    PrintDebug("Could not find shadow region for addr: %p\n", (void *)guest_addr);
-    return NULL;
+    v3_delete_shadow_region(info, reg);
   }
-
-  return (struct vmm_mem_hook *)(region->host_addr);
 }
 
 
-/* mem_addr is the guest physical memory address */
-static int mem_hook_dispatch(struct guest_info * info, 
-                            addr_t fault_gva, addr_t fault_gpa,  
-                            pf_error_t access_info, struct vmm_mem_hook * hook) 
-{
-
-  // emulate and then dispatch 
-  // or dispatch and emulate
-
-
-  if (access_info.write == 1) {
-    if (v3_emulate_memory_write(info, fault_gva, hook->write, fault_gpa, hook->priv_data) == -1) {
-      PrintError("Memory write emulation failed\n");
-      return -1;
-    }
-    
-  } else {
-    if (v3_emulate_memory_read(info, fault_gva, hook->read, fault_gpa, hook->priv_data) == -1) {
-      PrintError("Memory read emulation failed\n");
-      return -1;
-    }
-  }    
-
-  return 0;
-}
 
 
-int handle_special_page_fault(struct guest_info * info, 
-                             addr_t fault_gva, addr_t fault_gpa, 
-                             pf_error_t access_info) 
+int v3_add_shadow_mem( struct guest_info *  info,
+                      addr_t               guest_addr_start,
+                      addr_t               guest_addr_end,
+                      addr_t               host_addr)
 {
-  struct shadow_region * reg = get_shadow_region_by_addr(&(info->mem_map), fault_gpa);
+  struct v3_shadow_region * entry = (struct v3_shadow_region *)V3_Malloc(sizeof(struct v3_shadow_region));
 
-  PrintDebug("Handling Special Page Fault\n");
+  entry->guest_start = guest_addr_start;
+  entry->guest_end = guest_addr_end;
+  entry->host_type = SHDW_REGION_ALLOCATED;
+  entry->host_addr = host_addr;
+  entry->write_hook = NULL;
+  entry->read_hook = NULL;
+  entry->priv_data = NULL;
 
-  switch (reg->host_type) {
-  case HOST_REGION_HOOK:
-    return mem_hook_dispatch(info, fault_gva, fault_gpa, access_info, (struct vmm_mem_hook *)(reg->host_addr));
-  default:
+  if (insert_shadow_region(info, entry)) {
+    V3_Free(entry);
     return -1;
   }
 
   return 0;
-
 }
 
 
 
-void init_shadow_map(struct guest_info * info) {
-  struct shadow_map * map = &(info->mem_map);
+int v3_hook_write_mem(struct guest_info * info, addr_t guest_addr_start, addr_t guest_addr_end, 
+                     addr_t host_addr,
+                     int (*write)(addr_t guest_addr, void * src, uint_t length, void * priv_data),
+                     void * priv_data) {
 
-  map->num_regions = 0;
-
-  map->head = NULL;
-}
+  struct v3_shadow_region * entry = (struct v3_shadow_region *)V3_Malloc(sizeof(struct v3_shadow_region));
 
 
-void free_shadow_map(struct shadow_map * map) {
-  struct shadow_region * cursor = map->head;
-  struct shadow_region * tmp = NULL;
+  entry->guest_start = guest_addr_start;
+  entry->guest_end = guest_addr_end;
+  entry->host_type = SHDW_REGION_WRITE_HOOK;
+  entry->host_addr = host_addr;
+  entry->write_hook = write;
+  entry->read_hook = NULL;
+  entry->priv_data = priv_data;
 
-  while(cursor) {
-    tmp = cursor;
-    cursor = cursor->next;
-    V3_Free(tmp);
+  if (insert_shadow_region(info, entry)) {
+    V3_Free(entry);
+    return -1;
   }
 
-  V3_Free(map);
+  return 0;  
 }
 
-
-
-
-int add_shadow_region(struct shadow_map * map,
-                     struct shadow_region * region) 
-{
-  struct shadow_region * cursor = map->head;
-
-  PrintDebug("Adding Shadow Region: (0x%p-0x%p)\n", 
-            (void *)region->guest_start, (void *)region->guest_end);
-
-  if ((!cursor) || (cursor->guest_start >= region->guest_end)) {
-    region->prev = NULL;
-    region->next = cursor;
-    map->num_regions++;
-    map->head = region;
-    return 0;
-  }
-
-  while (cursor) {
-    // Check if it overlaps with the current cursor
-    if ((cursor->guest_end > region->guest_start) && (cursor->guest_start < region->guest_start)) {
-      // overlaps not allowed
-      return -1;
-    }
-    
-    if (!(cursor->next)) {
-      // add to the end of the list
-      cursor->next = region;
-      region->prev = cursor;
-      region->next = NULL;
-      map->num_regions++;
-      return 0;
-    } else if (cursor->next->guest_start >= region->guest_end) {
-      // add here
-      region->next = cursor->next;
-      region->prev = cursor;
-      
-      cursor->next->prev = region;
-      cursor->next = region;
-
-      map->num_regions++;
-      
-      return 0;
-    } else if (cursor->next->guest_end <= region->guest_start) {
-      cursor = cursor->next;
-    } else {
-      // This cannot happen!
-      // we should panic here
-      return -1;
-    }
-  }
+int v3_hook_full_mem(struct guest_info * info, addr_t guest_addr_start, addr_t guest_addr_end,
+                    int (*read)(addr_t guest_addr, void * dst, uint_t length, void * priv_data),
+                    int (*write)(addr_t guest_addr, void * src, uint_t length, void * priv_data),
+                    void * priv_data) {
   
-  // This cannot happen
-  // We should panic here
-  return -1;
-}
+  struct v3_shadow_region * entry = (struct v3_shadow_region *)V3_Malloc(sizeof(struct v3_shadow_region));
 
+  entry->guest_start = guest_addr_start;
+  entry->guest_end = guest_addr_end;
+  entry->host_type = SHDW_REGION_FULL_HOOK;
+  entry->host_addr = (addr_t)NULL;
+  entry->write_hook = write;
+  entry->read_hook = read;
+  entry->priv_data = priv_data;
+  
+  if (insert_shadow_region(info, entry)) {
+    V3_Free(entry);
+    return -1;
+  }
 
-int delete_shadow_region(struct shadow_map * map,
-                        addr_t guest_start,
-                        addr_t guest_end) {
-  return -1;
+  return 0;
 }
 
 
 
-struct shadow_region *get_shadow_region_by_index(struct shadow_map *  map,
-                                                uint_t index) {
-  struct shadow_region * reg = map->head;
-  uint_t i = 0;
-
-  while (reg) { 
-    if (i == index) { 
-      return reg;
-    }
-    reg = reg->next;
-    i++;
-  }
-  return NULL;
-}
 
+static inline 
+struct v3_shadow_region * __insert_shadow_region(struct guest_info * info, 
+                                                struct v3_shadow_region * region) {
+  struct rb_node ** p = &(info->mem_map.rb_node);
+  struct rb_node * parent = NULL;
+  struct v3_shadow_region * tmp_region;
 
-struct shadow_region * get_shadow_region_by_addr(struct shadow_map * map,
-                                                addr_t addr) {
-  struct shadow_region * reg = map->head;
+  while (*p) {
+    parent = *p;
+    tmp_region = rb_entry(parent, struct v3_shadow_region, tree_node);
 
-  while (reg) {
-    if ((reg->guest_start <= addr) && (reg->guest_end > addr)) {
-      return reg;
-    } else if (reg->guest_start > addr) {
-      return NULL;
+    if (region->guest_end <= tmp_region->guest_start) {
+      p = &(*p)->rb_left;
+    } else if (region->guest_start >= tmp_region->guest_end) {
+      p = &(*p)->rb_right;
     } else {
-      reg = reg->next;
+      return tmp_region;
     }
   }
-  return NULL;
-}
-
-
-host_region_type_t get_shadow_addr_type(struct guest_info * info, addr_t guest_addr) {
-  struct shadow_region * reg = get_shadow_region_by_addr(&(info->mem_map), guest_addr);
-
-  if (!reg) {
-    return HOST_REGION_INVALID;
-  } else {
-    return reg->host_type;
-  }
-}
 
-addr_t get_shadow_addr(struct guest_info * info, addr_t guest_addr) {
-  struct shadow_region * reg = get_shadow_region_by_addr(&(info->mem_map), guest_addr);
-
-  if (!reg) {
-    return 0;
-  } else {
-    return (guest_addr - reg->guest_start) + reg->host_addr;
-  }
+  rb_link_node(&(region->tree_node), parent, p);
+  
+  return NULL;
 }
 
 
-host_region_type_t lookup_shadow_map_addr(struct shadow_map * map, addr_t guest_addr, addr_t * host_addr) {
-  struct shadow_region * reg = get_shadow_region_by_addr(map, guest_addr);
+static inline
+struct v3_shadow_region * insert_shadow_region(struct guest_info * info, 
+                                              struct v3_shadow_region * region) {
+  struct v3_shadow_region * ret;
 
-  if (!reg) {
-    // No mapping exists
-    return HOST_REGION_INVALID;
-  } else {
-    switch (reg->host_type) {
-    case HOST_REGION_PHYSICAL_MEMORY:
-     *host_addr = (guest_addr - reg->guest_start) + reg->host_addr;
-     return reg->host_type;
-    case HOST_REGION_MEMORY_MAPPED_DEVICE:
-    case HOST_REGION_UNALLOCATED:
-      // ... 
-    default:
-      *host_addr = 0;
-      return reg->host_type;
-    }
+  if ((ret = __insert_shadow_region(info, region))) {
+    return ret;
   }
-}
-
-
-void print_shadow_map(struct shadow_map * map) {
-  struct shadow_region * cur = map->head;
-  int i = 0;
+  
+  v3_rb_insert_color(&(region->tree_node), &(info->mem_map));
 
-  PrintDebug("Memory Layout (regions: %d) \n", map->num_regions);
-
-  while (cur) {
-    PrintDebug("%d:  0x%p - 0x%p (%s) -> ", i, 
-              (void *)cur->guest_start, (void *)(cur->guest_end - 1),
-              cur->guest_type == GUEST_REGION_PHYSICAL_MEMORY ? "GUEST_REGION_PHYSICAL_MEMORY" :
-              cur->guest_type == GUEST_REGION_NOTHING ? "GUEST_REGION_NOTHING" :
-              cur->guest_type == GUEST_REGION_MEMORY_MAPPED_DEVICE ? "GUEST_REGION_MEMORY_MAPPED_DEVICE" :
-              "UNKNOWN");
-    if (cur->host_type == HOST_REGION_PHYSICAL_MEMORY || 
-       cur->host_type == HOST_REGION_UNALLOCATED ||
-       cur->host_type == HOST_REGION_MEMORY_MAPPED_DEVICE) { 
-      PrintDebug("0x%p", (void *)(cur->host_addr));
-    }
-    PrintDebug("(%s)\n",
-              cur->host_type == HOST_REGION_PHYSICAL_MEMORY ? "HOST_REGION_PHYSICAL_MEMORY" :
-              cur->host_type == HOST_REGION_UNALLOCATED ? "HOST_REGION_UNALLOACTED" :
-              cur->host_type == HOST_REGION_HOOK ? "HOST_REGION_HOOK" :
-              cur->host_type == HOST_REGION_MEMORY_MAPPED_DEVICE ? "HOST_REGION_MEMORY_MAPPED_DEVICE" :
-              cur->host_type == HOST_REGION_REMOTE ? "HOST_REGION_REMOTE" : 
-              cur->host_type == HOST_REGION_SWAPPED ? "HOST_REGION_SWAPPED" :
-              "UNKNOWN");
-    cur = cur->next;
-    i++;
-  }
+  return NULL;
 }
+                                                
 
 
 
@@ -339,159 +177,142 @@ void print_shadow_map(struct shadow_map * map) {
 
 
 
+int handle_special_page_fault(struct guest_info * info, 
+                             addr_t fault_gva, addr_t fault_gpa, 
+                             pf_error_t access_info) 
+{
+ struct v3_shadow_region * reg = v3_get_shadow_region(info, fault_gpa);
 
+  PrintDebug("Handling Special Page Fault\n");
 
+  switch (reg->host_type) {
+  case SHDW_REGION_WRITE_HOOK:
+    return v3_handle_mem_wr_hook(info, fault_gva, fault_gpa, reg, access_info);
+  case SHDW_REGION_FULL_HOOK:
+    return v3_handle_mem_full_hook(info, fault_gva, fault_gpa, reg, access_info);
+  default:
+    return -1;
+  }
 
-#ifdef VMM_MEM_TEST
-
-
-#include <stdlib.h>
-#include <stdio.h>
-#include <stdarg.h>
-
-
-
-
-
-struct vmm_os_hooks * os_hooks;
-
-void * TestMalloc(uint_t size) {
-  return malloc(size);
-}
-
-void * TestAllocatePages(int size) {
-  return malloc(4096 * size);
-}
-
-
-void TestPrint(const char * fmt, ...) {
-  va_list args;
+  return 0;
 
-  va_start(args, fmt);
-  vprintf(fmt, args);
-  va_end(args);
 }
 
-int mem_list_add_test_1(  vmm_mem_list_t * list) {
-
-  uint_t offset = 0;
-
-  PrintDebug("\n\nTesting Memory List\n");
-
-  init_mem_list(list);
-
-  offset = PAGE_SIZE * 6;
-  PrintDebug("Adding 0x%x - 0x%x\n", offset, offset + (PAGE_SIZE * 10));
-  add_mem_list_pages(list, offset, 10);
-  print_mem_list(list);
+int v3_handle_mem_wr_hook(struct guest_info * info, addr_t guest_va, addr_t guest_pa, 
+                         struct v3_shadow_region * reg, pf_error_t access_info) {
 
+  addr_t write_src_addr = 0;
 
-  offset = 0;
-  PrintDebug("Adding 0x%x - 0x%x\n", offset, offset + PAGE_SIZE * 4);
-  add_mem_list_pages(list, offset, 4);
-  print_mem_list(list);
-
-  offset = PAGE_SIZE * 20;
-  PrintDebug("Adding 0x%x - 0x%x\n", offset, offset + (PAGE_SIZE * 1));
-  add_mem_list_pages(list, offset, 1);
-  print_mem_list(list);
-
-  offset = PAGE_SIZE * 21;
-  PrintDebug("Adding 0x%x - 0x%x\n", offset, offset + (PAGE_SIZE * 3));
-  add_mem_list_pages(list, offset, 3);
-  print_mem_list(list);
-
-
-  offset = PAGE_SIZE * 10;
-  PrintDebug("Adding 0x%x - 0x%x\n", offset, offset + (PAGE_SIZE * 30));
-  add_mem_list_pages(list, offset, 30);
-  print_mem_list(list);
+  int write_len = v3_emulate_write_op(info, guest_va, guest_pa, &write_src_addr);
 
+  if (write_len == -1) {
+    PrintError("Emulation failure in write hook\n");
+    return -1;
+  }
 
-  offset = PAGE_SIZE * 5;
-  PrintDebug("Adding 0x%x - 0x%x\n", offset, offset + (PAGE_SIZE * 1));
-  add_mem_list_pages(list, offset, 1);
-  print_mem_list(list);
 
+  if (reg->write_hook(guest_pa, (void *)write_src_addr, write_len, reg->priv_data) != write_len) {
+    PrintError("Memory write hook did not return correct value\n");
+    return -1;
+  }
 
   return 0;
 }
 
+int v3_handle_mem_full_hook(struct guest_info * info, addr_t guest_va, addr_t guest_pa, 
+                           struct v3_shadow_region * reg, pf_error_t access_info) {
+  return -1;
+}
 
-int mem_layout_add_test_1(vmm_mem_layout_t * layout) {
-
-  
-  uint_t start = 0;
-  uint_t end = 0;
-
-  PrintDebug("\n\nTesting Memory Layout\n");
 
-  init_mem_layout(layout);
 
-  start = 0x6000;
-  end = 0x10000;;
-  PrintDebug("Adding 0x%x - 0x%x\n", start, end);
-  add_guest_mem_range(layout, start, end);
-  print_mem_layout(layout);
+struct v3_shadow_region * v3_get_shadow_region(struct guest_info * info, addr_t guest_addr) {
+  struct rb_node * n = info->mem_map.rb_node;
+  struct v3_shadow_region * reg = NULL;
 
+  while (n) {
+    reg = rb_entry(n, struct v3_shadow_region, tree_node);
 
-  start = 0x1000;
-  end = 0x3000;
-  PrintDebug("Adding 0x%x - 0x%x\n", start, end);
-  add_guest_mem_range(layout, start, end);
-  print_mem_layout(layout);
+    if (guest_addr < reg->guest_start) {
+      n = n->rb_left;
+    } else if (guest_addr >= reg->guest_end) {
+      n = n->rb_right;
+    } else {
+      return reg;
+    }
+  }
 
-  start = 0x2000;
-  end = 0x6000;
-  PrintDebug("Adding 0x%x - 0x%x\n", start, end);
-  add_guest_mem_range(layout, start, end);
-  print_mem_layout(layout);
+  return NULL;
+}
 
-  start = 0x4000;
-  end = 0x5000;
-  PrintDebug("Adding 0x%x - 0x%x\n", start, end);
-  add_guest_mem_range(layout, start, end);
-  print_mem_layout(layout);
 
 
-  start = 0x5000;
-  end = 0x7000;
-  PrintDebug("Adding 0x%x - 0x%x\n", start, end);
-  add_guest_mem_range(layout, start, end);
-  print_mem_layout(layout);
+addr_t v3_get_shadow_addr(struct v3_shadow_region * reg, addr_t guest_addr) {
+  if ((reg) && 
+      (reg->host_type != SHDW_REGION_FULL_HOOK) &&
+      (reg->host_type != SHDW_REGION_INVALID)) {
+    return (guest_addr - reg->guest_start) + reg->host_addr;
+  } else {
+    PrintError("MEM Region Invalid\n");
+    return 0;
+  }
+}
 
 
 
+void v3_delete_shadow_region(struct guest_info * info, struct v3_shadow_region * reg) {
+  if (reg != NULL) {
+    v3_rb_erase(&(reg->tree_node), &(info->mem_map));
 
-  return 0;
+    V3_Free(reg);
+  }
 }
 
 
 
-int main(int argc, char ** argv) {
-  struct vmm_os_hooks dummy_hooks;
-  os_hooks = &dummy_hooks;
-
-  vmm_mem_layout_t layout;
-  vmm_mem_list_t list;
 
-  os_hooks->malloc = &TestMalloc;
-  os_hooks->free = &free;
-  os_hooks->print_debug = &TestPrint;
-  os_hooks->allocate_pages = &TestAllocatePages;
+void print_shadow_map(struct guest_info * info) {
+  struct rb_node * node = v3_rb_first(&(info->mem_map));
+  struct v3_shadow_region * reg;
+  int i = 0;
 
+  PrintDebug("Memory Layout:\n");
 
+  do {
+    reg = rb_entry(node, struct v3_shadow_region, tree_node);
 
-  printf("mem_list_add_test_1: %d\n", mem_list_add_test_1(&list));
-  printf("layout_add_test_1: %d\n", mem_layout_add_test_1(&layout));
+    PrintDebug("%d:  0x%p - 0x%p -> 0x%p\n", i, 
+              (void *)(reg->guest_start), 
+              (void *)(reg->guest_end - 1), 
+              (void *)(reg->host_addr));
 
-  return 0;
+    PrintDebug("\t(%s) (WriteHook = 0x%p) (ReadHook = 0x%p)\n", 
+              v3_shdw_region_type_to_str(reg->host_type),
+              (void *)(reg->write_hook), 
+              (void *)(reg->read_hook));
+    
+    i++;
+  } while ((node = v3_rb_next(node)));
 }
-#endif
 
 
+static const uchar_t  SHDW_REGION_INVALID_STR[] = "SHDW_REGION_INVALID";
+static const uchar_t  SHDW_REGION_WRITE_HOOK_STR[] = "SHDW_REGION_WRITE_HOOK";
+static const uchar_t  SHDW_REGION_FULL_HOOK_STR[] = "SHDW_REGION_FULL_HOOK";
+static const uchar_t  SHDW_REGION_ALLOCATED_STR[] = "SHDW_REGION_ALLOCATED";
 
 
 
+const uchar_t * v3_shdw_region_type_to_str(v3_shdw_region_type_t type) {
+  switch (type) {
+  case SHDW_REGION_WRITE_HOOK:
+    return SHDW_REGION_WRITE_HOOK_STR;
+  case SHDW_REGION_FULL_HOOK:
+    return SHDW_REGION_FULL_HOOK_STR;
+  case SHDW_REGION_ALLOCATED:
+    return SHDW_REGION_ALLOCATED_STR;
+  default:
+    return SHDW_REGION_INVALID_STR;
+  }
+}
 
index a5ef2df..1a7dc7b 100644 (file)
@@ -46,6 +46,11 @@ static pt_entry_type_t pte64_lookup(pte64_t * pt, addr_t addr, addr_t * entry);
 
 
 
+#ifndef DEBUG_SHADOW_PAGING
+#undef PrintDebug
+#define PrintDebug(fmt, args...)
+#endif
+
 
 
 void delete_page_tables_32(pde32_t * pde) {
@@ -664,7 +669,6 @@ pt_access_status_t inline v3_can_access_pte64(pte64_t * pte, addr_t addr, pf_err
 pde32_t * create_passthrough_pts_32(struct guest_info * guest_info) {
   addr_t current_page_addr = 0;
   int i, j;
-  struct shadow_map * map = &(guest_info->mem_map);
 
   pde32_t * pde = V3_VAddr(V3_AllocPages(1));
 
@@ -674,14 +678,10 @@ pde32_t * create_passthrough_pts_32(struct guest_info * guest_info) {
     
 
     for (j = 0; j < MAX_PTE32_ENTRIES; j++) {
-      struct shadow_region * region = get_shadow_region_by_addr(map, current_page_addr);
+      struct v3_shadow_region * region =  v3_get_shadow_region(guest_info, current_page_addr);
 
       if (!region || 
-         (region->host_type == HOST_REGION_HOOK) || 
-         (region->host_type == HOST_REGION_UNALLOCATED) || 
-         (region->host_type == HOST_REGION_MEMORY_MAPPED_DEVICE) || 
-         (region->host_type == HOST_REGION_REMOTE) ||
-         (region->host_type == HOST_REGION_SWAPPED)) {
+         (region->host_type == SHDW_REGION_FULL_HOOK)) {
        pte[j].present = 0;
        pte[j].writable = 0;
        pte[j].user_page = 0;
@@ -696,7 +696,14 @@ pde32_t * create_passthrough_pts_32(struct guest_info * guest_info) {
       } else {
        addr_t host_addr;
        pte[j].present = 1;
-       pte[j].writable = 1;
+
+       if (region->host_type == SHDW_REGION_WRITE_HOOK) {
+         pte[j].writable = 0;
+         PrintDebug("Marking Write hook host_addr %p as RO\n", (void *)current_page_addr);
+       } else {
+         pte[j].writable = 1;
+       }
+         
        pte[j].user_page = 1;
        pte[j].write_through = 0;
        pte[j].cache_disable = 0;
@@ -761,7 +768,6 @@ pde32_t * create_passthrough_pts_32(struct guest_info * guest_info) {
 pdpe32pae_t * create_passthrough_pts_32PAE(struct guest_info * guest_info) {
   addr_t current_page_addr = 0;
   int i, j, k;
-  struct shadow_map * map = &(guest_info->mem_map);
 
   pdpe32pae_t * pdpe = V3_VAddr(V3_AllocPages(1));
   memset(pdpe, 0, PAGE_SIZE);
@@ -778,14 +784,10 @@ pdpe32pae_t * create_passthrough_pts_32PAE(struct guest_info * guest_info) {
       
       
       for (k = 0; k < MAX_PTE32PAE_ENTRIES; k++) {
-       struct shadow_region * region = get_shadow_region_by_addr(map, current_page_addr);
+       struct v3_shadow_region * region = v3_get_shadow_region(guest_info, current_page_addr);
        
        if (!region || 
-           (region->host_type == HOST_REGION_HOOK) || 
-           (region->host_type == HOST_REGION_UNALLOCATED) || 
-           (region->host_type == HOST_REGION_MEMORY_MAPPED_DEVICE) || 
-           (region->host_type == HOST_REGION_REMOTE) ||
-           (region->host_type == HOST_REGION_SWAPPED)) {
+           (region->host_type == SHDW_REGION_FULL_HOOK)) {
          pte[k].present = 0;
          pte[k].writable = 0;
          pte[k].user_page = 0;
@@ -801,7 +803,13 @@ pdpe32pae_t * create_passthrough_pts_32PAE(struct guest_info * guest_info) {
        } else {
          addr_t host_addr;
          pte[k].present = 1;
-         pte[k].writable = 1;
+       
+         if (region->host_type == SHDW_REGION_WRITE_HOOK) {
+           pte[k].writable = 0;
+         } else {
+           pte[k].writable = 1;
+         }
+
          pte[k].user_page = 1;
          pte[k].write_through = 0;
          pte[k].cache_disable = 0;
@@ -900,7 +908,6 @@ pdpe32pae_t * create_passthrough_pts_32PAE(struct guest_info * guest_info) {
 pml4e64_t * create_passthrough_pts_64(struct guest_info * info) {
   addr_t current_page_addr = 0;
   int i, j, k, m;
-  struct shadow_map * map = &(info->mem_map);
   
   pml4e64_t * pml = V3_VAddr(V3_AllocPages(1));
 
@@ -918,16 +925,12 @@ pml4e64_t * create_passthrough_pts_64(struct guest_info * info) {
 
 
        for (m = 0; m < MAX_PTE64_ENTRIES; m++) {
-         struct shadow_region * region = get_shadow_region_by_addr(map, current_page_addr);
+         struct v3_shadow_region * region = v3_get_shadow_region(info, current_page_addr);
          
 
          
          if (!region || 
-             (region->host_type == HOST_REGION_HOOK) || 
-             (region->host_type == HOST_REGION_UNALLOCATED) || 
-             (region->host_type == HOST_REGION_MEMORY_MAPPED_DEVICE) || 
-             (region->host_type == HOST_REGION_REMOTE) ||
-             (region->host_type == HOST_REGION_SWAPPED)) {
+             (region->host_type == SHDW_REGION_FULL_HOOK)) {
            pte[m].present = 0;
            pte[m].writable = 0;
            pte[m].user_page = 0;
@@ -942,7 +945,13 @@ pml4e64_t * create_passthrough_pts_64(struct guest_info * info) {
          } else {
            addr_t host_addr;
            pte[m].present = 1;
-           pte[m].writable = 1;
+
+           if (region->host_type == SHDW_REGION_WRITE_HOOK) {
+             pte[m].writable = 0;
+           } else {
+             pte[m].writable = 1;
+           }
+       
            pte[m].user_page = 1;
            pte[m].write_through = 0;
            pte[m].cache_disable = 0;
index cba22a8..21817a3 100644 (file)
@@ -1,3 +1,21 @@
+/*
+ * This file is part of the Palacios Virtual Machine Monitor developed
+ * by the V3VEE Project with funding from the United States National 
+ * Science Foundation and the Department of Energy.  
+ *
+ * The V3VEE Project is a joint project between Northwestern University
+ * and the University of New Mexico.  You can find out more at 
+ * http://www.v3vee.org
+ *
+ * Copyright (c) 2008, Jack Lange <jarusl@cs.northwestern.edu> 
+ * Copyright (c) 2008, The V3VEE Project <http://www.v3vee.org> 
+ * All rights reserved.
+ *
+ * Author: Jack Lange <jarusl@cs.northwestern.edu>
+ *
+ * This is free software.  You are permitted to use,
+ * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
+ */
 
 #ifdef USE_VMM_PAGING_DEBUG
 
diff --git a/palacios/src/palacios/vmm_profiler.c b/palacios/src/palacios/vmm_profiler.c
new file mode 100644 (file)
index 0000000..46a9e5e
--- /dev/null
@@ -0,0 +1,152 @@
+/*
+ * This file is part of the Palacios Virtual Machine Monitor developed
+ * by the V3VEE Project with funding from the United States National 
+ * Science Foundation and the Department of Energy.  
+ *
+ * The V3VEE Project is a joint project between Northwestern University
+ * and the University of New Mexico.  You can find out more at 
+ * http://www.v3vee.org
+ *
+ * Copyright (c) 2008, Jack Lange <jarusl@cs.northwestern.edu> 
+ * Copyright (c) 2008, The V3VEE Project <http://www.v3vee.org> 
+ * All rights reserved.
+ *
+ * Author: Jack Lange <jarusl@cs.northwestern.edu>
+ *
+ * This is free software.  You are permitted to use,
+ * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
+ */
+
+#include <palacios/vmm_types.h>
+#include <palacios/vmm_profiler.h>
+#include <palacios/svm_handler.h>
+#include <palacios/vmm_rbtree.h>
+
+
+struct exit_event {
+  uint_t exit_code;
+  uint_t exit_count;
+  uint_t handler_time;
+
+  struct rb_node tree_node;
+};
+
+
+void v3_init_profiler(struct guest_info * info) {
+  info->profiler.total_exits = 0;
+
+  info->profiler.start_time = 0;
+  info->profiler.end_time = 0;  
+  info->profiler.guest_pf_cnt = 0;
+
+  info->profiler.root.rb_node = NULL;
+}
+
+
+
+static inline struct exit_event * __insert_event(struct guest_info * info, 
+                                                struct exit_event * evt) {
+  struct rb_node ** p = &(info->profiler.root.rb_node);
+  struct rb_node * parent = NULL;
+  struct exit_event * tmp_evt = NULL;
+
+  while (*p) {
+    parent = *p;
+    tmp_evt = rb_entry(parent, struct exit_event, tree_node);
+
+    if (evt->exit_code < tmp_evt->exit_code) {
+      p = &(*p)->rb_left;
+    } else if (evt->exit_code > tmp_evt->exit_code) {
+      p = &(*p)->rb_right;
+    } else {
+      return tmp_evt;
+    }
+  }
+  rb_link_node(&(evt->tree_node), parent, p);
+
+  return NULL;
+}
+
+static inline struct exit_event * insert_event(struct guest_info * info, 
+                                              struct exit_event * evt) {
+  struct exit_event * ret;
+
+  if ((ret = __insert_event(info, evt))) {
+    return ret;
+  }
+
+  v3_rb_insert_color(&(evt->tree_node), &(info->profiler.root));
+
+  return NULL;
+}
+
+
+static struct exit_event * get_exit(struct guest_info * info, uint_t exit_code) {
+  struct rb_node * n = info->profiler.root.rb_node;
+  struct exit_event * evt = NULL;
+
+  while (n) {
+    evt = rb_entry(n, struct exit_event, tree_node);
+    
+    if (exit_code < evt->exit_code) {
+      n = n->rb_left;
+    } else if (exit_code > evt->exit_code) {
+      n = n->rb_right;
+    } else {
+      return evt;
+    }
+  }
+
+  return NULL;
+}
+
+
+static inline struct exit_event * create_exit(uint_t exit_code) {
+  struct exit_event * evt = V3_Malloc(sizeof(struct exit_event));
+
+  evt->exit_code = exit_code;
+  evt->exit_count = 0;
+  evt->handler_time = 0;
+
+  return evt;
+}
+
+void v3_profile_exit(struct guest_info * info, uint_t exit_code) {
+  uint_t time = (info->profiler.end_time - info->profiler.start_time);
+  struct exit_event * evt = get_exit(info, exit_code);
+
+  if (evt == NULL) {
+    evt = create_exit(exit_code);
+    insert_event(info, evt);
+  }
+
+  
+
+  evt->handler_time = (evt->handler_time * .99) + (time * .01);
+
+
+  evt->exit_count++;
+  
+  info->profiler.total_exits++;
+}
+
+
+void v3_print_profile(struct guest_info * info) {
+  struct exit_event * evt = NULL;
+  struct rb_node * node = v3_rb_first(&(info->profiler.root));
+  
+  PrintDebug("GUEST_PF: %u\n", info->profiler.guest_pf_cnt);
+
+  do {
+    evt = rb_entry(node, struct exit_event, tree_node);
+    const char * code_str = vmexit_code_to_str(evt->exit_code);
+
+    PrintDebug("%s:%sCnt=%u,%sTime=%u\n", 
+              code_str,
+              (strlen(code_str) > 14) ? "\t" : "\t\t",
+              evt->exit_count,
+              (evt->exit_count >= 100) ? "\t" : "\t\t",
+              evt->handler_time);
+              
+  } while ((node = v3_rb_next(node)));
+}
diff --git a/palacios/src/palacios/vmm_rbtree.c b/palacios/src/palacios/vmm_rbtree.c
new file mode 100644 (file)
index 0000000..520e227
--- /dev/null
@@ -0,0 +1,391 @@
+/*
+  Red Black Trees
+  (C) 1999  Andrea Arcangeli <andrea@suse.de>
+  (C) 2002  David Woodhouse <dwmw2@infradead.org>
+  
+  This program is free software; you can redistribute it and/or modify
+  it under the terms of the GNU General Public License as published by
+  the Free Software Foundation; either version 2 of the License, or
+  (at your option) any later version.
+
+  This program is distributed in the hope that it will be useful,
+  but WITHOUT ANY WARRANTY; without even the implied warranty of
+  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+  GNU General Public License for more details.
+
+  You should have received a copy of the GNU General Public License
+  along with this program; if not, write to the Free Software
+  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+
+  linux/lib/rbtree.c
+*/
+
+#include <palacios/vmm_rbtree.h>
+
+
+static void __rb_rotate_left(struct rb_node *node, struct rb_root *root)
+{
+       struct rb_node *right = node->rb_right;
+       struct rb_node *parent = rb_parent(node);
+
+       if ((node->rb_right = right->rb_left))
+               rb_set_parent(right->rb_left, node);
+       right->rb_left = node;
+
+       rb_set_parent(right, parent);
+
+       if (parent)
+       {
+               if (node == parent->rb_left)
+                       parent->rb_left = right;
+               else
+                       parent->rb_right = right;
+       }
+       else
+               root->rb_node = right;
+       rb_set_parent(node, right);
+}
+
+static void __rb_rotate_right(struct rb_node *node, struct rb_root *root)
+{
+       struct rb_node *left = node->rb_left;
+       struct rb_node *parent = rb_parent(node);
+
+       if ((node->rb_left = left->rb_right))
+               rb_set_parent(left->rb_right, node);
+       left->rb_right = node;
+
+       rb_set_parent(left, parent);
+
+       if (parent)
+       {
+               if (node == parent->rb_right)
+                       parent->rb_right = left;
+               else
+                       parent->rb_left = left;
+       }
+       else
+               root->rb_node = left;
+       rb_set_parent(node, left);
+}
+
+void v3_rb_insert_color(struct rb_node *node, struct rb_root *root)
+{
+       struct rb_node *parent, *gparent;
+
+       while ((parent = rb_parent(node)) && rb_is_red(parent))
+       {
+               gparent = rb_parent(parent);
+
+               if (parent == gparent->rb_left)
+               {
+                       {
+                               register struct rb_node *uncle = gparent->rb_right;
+                               if (uncle && rb_is_red(uncle))
+                               {
+                                       rb_set_black(uncle);
+                                       rb_set_black(parent);
+                                       rb_set_red(gparent);
+                                       node = gparent;
+                                       continue;
+                               }
+                       }
+
+                       if (parent->rb_right == node)
+                       {
+                               register struct rb_node *tmp;
+                               __rb_rotate_left(parent, root);
+                               tmp = parent;
+                               parent = node;
+                               node = tmp;
+                       }
+
+                       rb_set_black(parent);
+                       rb_set_red(gparent);
+                       __rb_rotate_right(gparent, root);
+               } else {
+                       {
+                               register struct rb_node *uncle = gparent->rb_left;
+                               if (uncle && rb_is_red(uncle))
+                               {
+                                       rb_set_black(uncle);
+                                       rb_set_black(parent);
+                                       rb_set_red(gparent);
+                                       node = gparent;
+                                       continue;
+                               }
+                       }
+
+                       if (parent->rb_left == node)
+                       {
+                               register struct rb_node *tmp;
+                               __rb_rotate_right(parent, root);
+                               tmp = parent;
+                               parent = node;
+                               node = tmp;
+                       }
+
+                       rb_set_black(parent);
+                       rb_set_red(gparent);
+                       __rb_rotate_left(gparent, root);
+               }
+       }
+
+       rb_set_black(root->rb_node);
+}
+
+
+static void __rb_erase_color(struct rb_node *node, struct rb_node *parent,
+                            struct rb_root *root)
+{
+       struct rb_node *other;
+
+       while ((!node || rb_is_black(node)) && node != root->rb_node)
+       {
+               if (parent->rb_left == node)
+               {
+                       other = parent->rb_right;
+                       if (rb_is_red(other))
+                       {
+                               rb_set_black(other);
+                               rb_set_red(parent);
+                               __rb_rotate_left(parent, root);
+                               other = parent->rb_right;
+                       }
+                       if ((!other->rb_left || rb_is_black(other->rb_left)) &&
+                           (!other->rb_right || rb_is_black(other->rb_right)))
+                       {
+                               rb_set_red(other);
+                               node = parent;
+                               parent = rb_parent(node);
+                       }
+                       else
+                       {
+                               if (!other->rb_right || rb_is_black(other->rb_right))
+                               {
+                                       struct rb_node *o_left;
+                                       if ((o_left = other->rb_left))
+                                               rb_set_black(o_left);
+                                       rb_set_red(other);
+                                       __rb_rotate_right(other, root);
+                                       other = parent->rb_right;
+                               }
+                               rb_set_color(other, rb_color(parent));
+                               rb_set_black(parent);
+                               if (other->rb_right)
+                                       rb_set_black(other->rb_right);
+                               __rb_rotate_left(parent, root);
+                               node = root->rb_node;
+                               break;
+                       }
+               }
+               else
+               {
+                       other = parent->rb_left;
+                       if (rb_is_red(other))
+                       {
+                               rb_set_black(other);
+                               rb_set_red(parent);
+                               __rb_rotate_right(parent, root);
+                               other = parent->rb_left;
+                       }
+                       if ((!other->rb_left || rb_is_black(other->rb_left)) &&
+                           (!other->rb_right || rb_is_black(other->rb_right)))
+                       {
+                               rb_set_red(other);
+                               node = parent;
+                               parent = rb_parent(node);
+                       }
+                       else
+                       {
+                               if (!other->rb_left || rb_is_black(other->rb_left))
+                               {
+                                       register struct rb_node *o_right;
+                                       if ((o_right = other->rb_right))
+                                               rb_set_black(o_right);
+                                       rb_set_red(other);
+                                       __rb_rotate_left(other, root);
+                                       other = parent->rb_left;
+                               }
+                               rb_set_color(other, rb_color(parent));
+                               rb_set_black(parent);
+                               if (other->rb_left)
+                                       rb_set_black(other->rb_left);
+                               __rb_rotate_right(parent, root);
+                               node = root->rb_node;
+                               break;
+                       }
+               }
+       }
+       if (node)
+               rb_set_black(node);
+}
+
+void v3_rb_erase(struct rb_node *node, struct rb_root *root)
+{
+       struct rb_node *child, *parent;
+       int color;
+
+       if (!node->rb_left)
+               child = node->rb_right;
+       else if (!node->rb_right)
+               child = node->rb_left;
+       else
+       {
+               struct rb_node *old = node, *left;
+
+               node = node->rb_right;
+               while ((left = node->rb_left) != NULL)
+                       node = left;
+               child = node->rb_right;
+               parent = rb_parent(node);
+               color = rb_color(node);
+
+               if (child)
+                       rb_set_parent(child, parent);
+               if (parent == old) {
+                       parent->rb_right = child;
+                       parent = node;
+               } else
+                       parent->rb_left = child;
+
+               node->rb_parent_color = old->rb_parent_color;
+               node->rb_right = old->rb_right;
+               node->rb_left = old->rb_left;
+
+               if (rb_parent(old))
+               {
+                       if (rb_parent(old)->rb_left == old)
+                               rb_parent(old)->rb_left = node;
+                       else
+                               rb_parent(old)->rb_right = node;
+               } else
+                       root->rb_node = node;
+
+               rb_set_parent(old->rb_left, node);
+               if (old->rb_right)
+                       rb_set_parent(old->rb_right, node);
+               goto color;
+       }
+
+       parent = rb_parent(node);
+       color = rb_color(node);
+
+       if (child)
+               rb_set_parent(child, parent);
+       if (parent)
+       {
+               if (parent->rb_left == node)
+                       parent->rb_left = child;
+               else
+                       parent->rb_right = child;
+       }
+       else
+               root->rb_node = child;
+
+ color:
+       if (color == RB_BLACK)
+               __rb_erase_color(child, parent, root);
+}
+
+
+/*
+ * This function returns the first node (in sort order) of the tree.
+ */
+struct rb_node *v3_rb_first(struct rb_root *root)
+{
+       struct rb_node  *n;
+
+       n = root->rb_node;
+       if (!n)
+               return NULL;
+       while (n->rb_left)
+               n = n->rb_left;
+       return n;
+}
+
+
+struct rb_node *v3_rb_last(struct rb_root *root)
+{
+       struct rb_node  *n;
+
+       n = root->rb_node;
+       if (!n)
+               return NULL;
+       while (n->rb_right)
+               n = n->rb_right;
+       return n;
+}
+
+
+struct rb_node *v3_rb_next(struct rb_node *node)
+{
+       struct rb_node *parent;
+
+       /* If we have a right-hand child, go down and then left as far
+          as we can. */
+       if (node->rb_right) {
+               node = node->rb_right; 
+               while (node->rb_left)
+                       node=node->rb_left;
+               return node;
+       }
+
+       /* No right-hand children.  Everything down and left is
+          smaller than us, so any 'next' node must be in the general
+          direction of our parent. Go up the tree; any time the
+          ancestor is a right-hand child of its parent, keep going
+          up. First time it's a left-hand child of its parent, said
+          parent is our 'next' node. */
+       while ((parent = rb_parent(node)) && node == parent->rb_right)
+               node = parent;
+
+       return parent;
+}
+
+
+struct rb_node *v3_rb_prev(struct rb_node *node)
+{
+       struct rb_node *parent;
+
+       /* If we have a left-hand child, go down and then right as far
+          as we can. */
+       if (node->rb_left) {
+               node = node->rb_left; 
+               while (node->rb_right)
+                       node=node->rb_right;
+               return node;
+       }
+
+       /* No left-hand children. Go up till we find an ancestor which
+          is a right-hand child of its parent */
+       while ((parent = rb_parent(node)) && node == parent->rb_left)
+               node = parent;
+
+       return parent;
+}
+
+
+void v3_rb_replace_node(struct rb_node *victim, struct rb_node *new,
+                    struct rb_root *root)
+{
+       struct rb_node *parent = rb_parent(victim);
+
+       /* Set the surrounding nodes to point to the replacement */
+       if (parent) {
+               if (victim == parent->rb_left)
+                       parent->rb_left = new;
+               else
+                       parent->rb_right = new;
+       } else {
+               root->rb_node = new;
+       }
+       if (victim->rb_left)
+               rb_set_parent(victim->rb_left, new);
+       if (victim->rb_right)
+               rb_set_parent(victim->rb_right, new);
+
+       /* Copy the pointers/colour from the victim to the replacement */
+       *new = *victim;
+}
+
index a427b13..611e0ea 100644 (file)
@@ -62,14 +62,14 @@ struct shadow_page_data {
 
 
 
-DEFINE_HASHTABLE_INSERT(add_cr3_to_cache, addr_t, struct hashtable *);
-DEFINE_HASHTABLE_SEARCH(find_cr3_in_cache, addr_t, struct hashtable *);
-DEFINE_HASHTABLE_REMOVE(del_cr3_from_cache, addr_t, struct hashtable *, 0);
+//DEFINE_HASHTABLE_INSERT(add_cr3_to_cache, addr_t, struct hashtable *);
+//DEFINE_HASHTABLE_SEARCH(find_cr3_in_cache, addr_t, struct hashtable *);
+//DEFINE_HASHTABLE_REMOVE(del_cr3_from_cache, addr_t, struct hashtable *, 0);
 
 
 DEFINE_HASHTABLE_INSERT(add_pte_map, addr_t, addr_t);
 DEFINE_HASHTABLE_SEARCH(find_pte_map, addr_t, addr_t);
-DEFINE_HASHTABLE_REMOVE(del_pte_map, addr_t, addr_t, 0);
+//DEFINE_HASHTABLE_REMOVE(del_pte_map, addr_t, addr_t, 0);
 
 
 
@@ -81,219 +81,29 @@ static int pte_equals(addr_t key1, addr_t key2) {
   return (key1 == key2);
 }
 
-static uint_t cr3_hash_fn(addr_t key) {
-  return hash_long(key, 32);
-}
-
-static int cr3_equals(addr_t key1, addr_t key2) {
-  return (key1 == key2);
-}
+static addr_t create_new_shadow_pt();
+static void inject_guest_pf(struct guest_info * info, addr_t fault_addr, pf_error_t error_code);
+static int is_guest_pf(pt_access_status_t guest_access, pt_access_status_t shadow_access);
 
 
+#include "vmm_shadow_paging_32.h"
+#include "vmm_shadow_paging_32pae.h"
+#include "vmm_shadow_paging_64.h"
 
-static int activate_shadow_pt_32(struct guest_info * info);
-static int activate_shadow_pt_32pae(struct guest_info * info);
-static int activate_shadow_pt_64(struct guest_info * info);
 
 
-static int handle_shadow_pagefault_32(struct guest_info * info, addr_t fault_addr, pf_error_t error_code);
-static int handle_shadow_pagefault_32pae(struct guest_info * info, addr_t fault_addr, pf_error_t error_code);
-static int handle_shadow_pagefault_64(struct guest_info * info, addr_t fault_addr, pf_error_t error_code);
-
-
-static int cache_page_tables_32(struct guest_info * info, addr_t pde);
-static int cache_page_tables_64(struct guest_info * info, addr_t pde);
-
 int v3_init_shadow_page_state(struct guest_info * info) {
   struct shadow_page_state * state = &(info->shdw_pg_state);
   
   state->guest_cr3 = 0;
   state->guest_cr0 = 0;
 
-  state->cr3_cache = create_hashtable(0, &cr3_hash_fn, &cr3_equals);
-
-  state->cached_cr3 = 0;
   state->cached_ptes = NULL;
-
-  return 0;
-}
-
-
-
-
-
-int v3_cache_page_tables(struct guest_info * info, addr_t cr3) {
-  switch(v3_get_cpu_mode(info)) {
-  case PROTECTED:
-    return cache_page_tables_32(info, CR3_TO_PDE32_PA(cr3));
-  default:
-    return -1;
-  }
-}
-
-static int cache_page_tables_32(struct guest_info * info, addr_t pde) {
-  struct shadow_page_state * state = &(info->shdw_pg_state);
-  addr_t pde_host_addr;
-  pde32_t * tmp_pde;
-  struct hashtable * pte_cache = NULL;
-  int i = 0;
-
-  if (pde == state->cached_cr3) {
-    return 1;
-  }
-
-  if (state->cached_ptes != NULL) {
-    hashtable_destroy(state->cached_ptes, 0, 0);
-    state->cached_ptes = NULL;
-  }
-
-  state->cached_cr3 = pde;
-
-  pte_cache = create_hashtable(0, &pte_hash_fn, &pte_equals);
-  state->cached_ptes = pte_cache;
-
-  if (guest_pa_to_host_va(info, pde, &pde_host_addr) == -1) {
-    PrintError("Could not lookup host address of guest PDE\n");
-    return -1;
-  }
-
-  tmp_pde = (pde32_t *)pde_host_addr;
-
-  add_pte_map(pte_cache, pde, pde_host_addr);
-
-
-  for (i = 0; i < MAX_PDE32_ENTRIES; i++) {
-    if ((tmp_pde[i].present) && (tmp_pde[i].large_page == 0)) {
-      addr_t pte_host_addr;
-
-      if (guest_pa_to_host_va(info, (addr_t)(BASE_TO_PAGE_ADDR(tmp_pde[i].pt_base_addr)), &pte_host_addr) == -1) {
-       PrintError("Could not lookup host address of guest PDE\n");
-       return -1;
-      }
-
-      add_pte_map(pte_cache, (addr_t)(BASE_TO_PAGE_ADDR(tmp_pde[i].pt_base_addr)), pte_host_addr); 
-    }
-  }
-
-  return 0;
-
-}
-
-
-static int cache_page_tables_64(struct guest_info * info, addr_t pde) {
-  return -1;
-}
-
-
-int v3_replace_shdw_page32(struct guest_info * info, addr_t location, pte32_t * new_page, pte32_t * old_page) {
-  pde32_t * shadow_pd = (pde32_t *)CR3_TO_PDE32_VA(info->ctrl_regs.cr3);
-  pde32_t * shadow_pde =  (pde32_t *)&(shadow_pd[PDE32_INDEX(location)]);
-
-  if (shadow_pde->large_page == 0) {
-    pte32_t * shadow_pt = (pte32_t *)(addr_t)BASE_TO_PAGE_ADDR(shadow_pde->pt_base_addr);
-    pte32_t * shadow_pte = (pte32_t *)&(shadow_pt[PTE32_INDEX(location)]);
-
-    //if (shadow_pte->present == 1) {
-    *(uint_t *)old_page = *(uint_t *)shadow_pte;
-    //}
-
-    *(uint_t *)shadow_pte = *(uint_t *)new_page;
-
-  } else {
-    // currently unhandled
-    PrintError("Replacing large shadow pages not implemented\n");
-    return -1;
-  }
-  
-  return 0;
-}
-
-
-
-
-
-// We assume that shdw_pg_state.guest_cr3 is pointing to the page tables we want to activate
-// We also assume that the CPU mode has not changed during this page table transition
-static int activate_shadow_pt_32(struct guest_info * info) {
-  struct cr3_32 * shadow_cr3 = (struct cr3_32 *)&(info->ctrl_regs.cr3);
-  struct cr3_32 * guest_cr3 = (struct cr3_32 *)&(info->shdw_pg_state.guest_cr3);
-  int cached = 0;
-  
-  // Check if shadow page tables are in the cache
-  cached = cache_page_tables_32(info, CR3_TO_PDE32_PA(*(addr_t *)guest_cr3));
-  
-  if (cached == -1) {
-    PrintError("CR3 Cache failed\n");
-    return -1;
-  } else if (cached == 0) {
-    addr_t shadow_pt;
-    
-    PrintDebug("New CR3 is different - flushing shadow page table %p\n", shadow_cr3 );
-    delete_page_tables_32(CR3_TO_PDE32_VA(*(uint_t*)shadow_cr3));
-    
-    shadow_pt = v3_create_new_shadow_pt();
-    
-    shadow_cr3->pdt_base_addr = (addr_t)V3_PAddr((void *)(addr_t)PAGE_BASE_ADDR(shadow_pt));
-    PrintDebug( "Created new shadow page table %p\n", (void *)(addr_t)shadow_cr3->pdt_base_addr );
-  } else {
-    PrintDebug("Reusing cached shadow Page table\n");
-  }
-  
-  shadow_cr3->pwt = guest_cr3->pwt;
-  shadow_cr3->pcd = guest_cr3->pcd;
+  state->cached_cr3 = 0;
   
   return 0;
 }
 
-static int activate_shadow_pt_32pae(struct guest_info * info) {
-  PrintError("Activating 32 bit PAE page tables not implemented\n");
-  return -1;
-}
-
-static int activate_shadow_pt_64_cb(page_type_t type, addr_t vaddr, addr_t page_ptr, addr_t page_pa, void * private_data) {
-  PrintDebug("CB: Page: %p->%p (host_ptr=%p), Type: %s\n", (void *)vaddr, (void *)page_pa, (void *)page_ptr, v3_page_type_to_str(type));
-  return 0;
-}
-
-
-static int activate_shadow_pt_64(struct guest_info * info) {
-  //  struct cr3_64 * shadow_cr3 = (struct cr3_64 *)&(info->ctrl_regs.cr3);
-  struct cr3_64 * guest_cr3 = (struct cr3_64 *)&(info->shdw_pg_state.guest_cr3);
-  int cached = 0;
-  
-  v3_walk_guest_pt_64(info, info->shdw_pg_state.guest_cr3, activate_shadow_pt_64_cb, NULL);
-
-  
-
-  return -1;
-
-
-  // Check if shadow page tables are in the cache
-  cached = cache_page_tables_64(info, CR3_TO_PDE32_PA(*(addr_t *)guest_cr3));
-             /*  
-  if (cached == -1) {
-    PrintError("CR3 Cache failed\n");
-    return -1;
-  } else if (cached == 0) {
-    addr_t shadow_pt;
-    
-    PrintDebug("New CR3 is different - flushing shadow page table %p\n", shadow_cr3 );
-    delete_page_tables_32(CR3_TO_PDE32_VA(*(uint_t*)shadow_cr3));
-    
-    shadow_pt = v3_create_new_shadow_pt();
-    
-    shadow_cr3->pml4t_base_addr = (addr_t)V3_PAddr((void *)(addr_t)PAGE_BASE_ADDR(shadow_pt));
-    PrintDebug( "Created new shadow page table %p\n", (void *)(addr_t)shadow_cr3->pml4t_base_addr );
-  } else {
-    PrintDebug("Reusing cached shadow Page table\n");
-  }
-  
-  shadow_cr3->pwt = guest_cr3->pwt;
-  shadow_cr3->pcd = guest_cr3->pcd;
-  
-  return 0;
-                     */
-}
 
 
 // Reads the guest CR3 register
@@ -335,12 +145,6 @@ int v3_handle_shadow_pagefault(struct guest_info * info, addr_t fault_addr, pf_e
   
   if (info->mem_mode == PHYSICAL_MEM) {
     // If paging is not turned on we need to handle the special cases
-
-#ifdef DEBUG_SHADOW_PAGING
-    PrintHostPageTree(info->cpu_mode, fault_addr, info->ctrl_regs.cr3);
-    PrintGuestPageTree(info, fault_addr, info->shdw_pg_state.guest_cr3);
-#endif
-
     return handle_special_page_fault(info, fault_addr, fault_addr, error_code);
   } else if (info->mem_mode == VIRTUAL_MEM) {
 
@@ -363,7 +167,66 @@ int v3_handle_shadow_pagefault(struct guest_info * info, addr_t fault_addr, pf_e
   }
 }
 
-addr_t v3_create_new_shadow_pt() {
+
+int v3_handle_shadow_invlpg(struct guest_info * info) {
+  uchar_t instr[15];
+  struct x86_instr dec_instr;
+  int ret = 0;
+  addr_t vaddr = 0;
+
+  if (info->mem_mode != VIRTUAL_MEM) {
+    // Paging must be turned on...
+    // should handle with some sort of fault I think
+    PrintError("ERROR: INVLPG called in non paged mode\n");
+    return -1;
+  }
+
+  if (info->mem_mode == PHYSICAL_MEM) { 
+    ret = read_guest_pa_memory(info, get_addr_linear(info, info->rip, &(info->segments.cs)), 15, instr);
+  } else { 
+    ret = read_guest_va_memory(info, get_addr_linear(info, info->rip, &(info->segments.cs)), 15, instr);
+  }
+
+  if (ret == -1) {
+    PrintError("Could not read instruction into buffer\n");
+    return -1;
+  }
+
+  if (v3_decode(info, (addr_t)instr, &dec_instr) == -1) {
+    PrintError("Decoding Error\n");
+    return -1;
+  }
+  
+  if ((dec_instr.op_type != V3_OP_INVLPG) || 
+      (dec_instr.num_operands != 1) ||
+      (dec_instr.dst_operand.type != MEM_OPERAND)) {
+    PrintError("Decoder Error: Not a valid INVLPG instruction...\n");
+    return -1;
+  }
+
+  vaddr = dec_instr.dst_operand.operand;
+
+  info->rip += dec_instr.instr_length;
+
+  switch (info->cpu_mode) {
+  case PROTECTED:
+    return handle_shadow_invlpg_32(info, vaddr);
+  case PROTECTED_PAE:
+    return handle_shadow_invlpg_32pae(info, vaddr);
+  case LONG:
+  case LONG_32_COMPAT:
+  case LONG_16_COMPAT:
+    return handle_shadow_invlpg_64(info, vaddr);
+  default:
+    PrintError("Invalid CPU mode: %d\n", info->cpu_mode);
+    return -1;
+  }
+}
+
+
+
+
+static addr_t create_new_shadow_pt() {
   void * host_pde = 0;
 
   host_pde = V3_VAddr(V3_AllocPages(1));
@@ -374,6 +237,10 @@ addr_t v3_create_new_shadow_pt() {
 
 
 static void inject_guest_pf(struct guest_info * info, addr_t fault_addr, pf_error_t error_code) {
+  if (info->enable_profiler) {
+    info->profiler.guest_pf_cnt++;
+  }
+
   info->ctrl_regs.cr2 = fault_addr;
   v3_raise_exception_with_error(info, PF_EXCEPTION, *(uint_t *)&error_code);
 }
@@ -411,533 +278,3 @@ static int is_guest_pf(pt_access_status_t guest_access, pt_access_status_t shado
 }
 
 
-
-
-/* 
- * *
- * * 
- * * 64 bit Page table fault handlers
- * *
- * *
- */
-
-static int handle_shadow_pagefault_64(struct guest_info * info, addr_t fault_addr, pf_error_t error_code) {
-  pt_access_status_t guest_access;
-  pt_access_status_t shadow_access;
-  int ret; 
-  PrintDebug("64 bit shadow page fault\n");
-
-  ret = v3_check_guest_pt_32(info, info->shdw_pg_state.guest_cr3, fault_addr, error_code, &guest_access);
-
-  PrintDebug("Guest Access Check: %d (access=%d)\n", ret, guest_access);
-
-  ret = v3_check_host_pt_32(info->ctrl_regs.cr3, fault_addr, error_code, &shadow_access);
-
-  PrintDebug("Shadow Access Check: %d (access=%d)\n", ret, shadow_access);
-  
-
-  PrintError("64 bit shadow paging not implemented\n");
-  return -1;
-}
-
-
-/* 
- * *
- * * 
- * * 32 bit PAE  Page table fault handlers
- * *
- * *
- */
-
-static int handle_shadow_pagefault_32pae(struct guest_info * info, addr_t fault_addr, pf_error_t error_code) {
-  PrintError("32 bit PAE shadow paging not implemented\n");
-  return -1;
-}
-
-
-
-
-
-
-
-/* 
- * *
- * * 
- * * 32 bit Page table fault handlers
- * *
- * *
- */
-static int handle_large_pagefault_32(struct guest_info * info, 
-                                   addr_t fault_addr, pf_error_t error_code, 
-                                    pte32_t * shadow_pt, pde32_4MB_t * large_guest_pde);
-
-static int handle_shadow_pte32_fault(struct guest_info * info, 
-                                    addr_t fault_addr, 
-                                    pf_error_t error_code,
-                                    pte32_t * shadow_pt, 
-                                    pte32_t * guest_pt);
-
-
-static int handle_shadow_pagefault_32(struct guest_info * info, addr_t fault_addr, pf_error_t error_code) {
-  pde32_t * guest_pd = NULL;
-  pde32_t * shadow_pd = CR3_TO_PDE32_VA(info->ctrl_regs.cr3);
-  addr_t guest_cr3 = CR3_TO_PDE32_PA(info->shdw_pg_state.guest_cr3);
-  pt_access_status_t guest_pde_access;
-  pt_access_status_t shadow_pde_access;
-  pde32_t * guest_pde = NULL;
-  pde32_t * shadow_pde = (pde32_t *)&(shadow_pd[PDE32_INDEX(fault_addr)]);
-
-  PrintDebug("Shadow page fault handler: %p\n", (void*) fault_addr );
-
-  if (guest_pa_to_host_va(info, guest_cr3, (addr_t*)&guest_pd) == -1) {
-    PrintError("Invalid Guest PDE Address: 0x%p\n",  (void *)guest_cr3);
-    return -1;
-  } 
-
-  guest_pde = (pde32_t *)&(guest_pd[PDE32_INDEX(fault_addr)]);
-
-
-  // Check the guest page permissions
-  guest_pde_access = v3_can_access_pde32(guest_pd, fault_addr, error_code);
-
-  // Check the shadow page permissions
-  shadow_pde_access = v3_can_access_pde32(shadow_pd, fault_addr, error_code);
-  
-  /* Was the page fault caused by the Guest's page tables? */
-  if (is_guest_pf(guest_pde_access, shadow_pde_access) == 1) {
-    PrintDebug("Injecting PDE pf to guest: (guest access error=%d) (pf error code=%d)\n", 
-              *(uint_t *)&guest_pde_access, *(uint_t *)&error_code);
-    inject_guest_pf(info, fault_addr, error_code);
-    return 0;
-  }
-
-  
-  if (shadow_pde_access == PT_ACCESS_NOT_PRESENT) 
-    {
-      pte32_t * shadow_pt =  (pte32_t *)v3_create_new_shadow_pt();
-
-      shadow_pde->present = 1;
-      shadow_pde->user_page = guest_pde->user_page;
-      //    shadow_pde->large_page = guest_pde->large_page;
-      shadow_pde->large_page = 0;
-      
-
-      // VMM Specific options
-      shadow_pde->write_through = 0;
-      shadow_pde->cache_disable = 0;
-      shadow_pde->global_page = 0;
-      //
-      
-      guest_pde->accessed = 1;
-      
-      shadow_pde->pt_base_addr = PAGE_BASE_ADDR((addr_t)V3_PAddr(shadow_pt));
-      
-      if (guest_pde->large_page == 0) {
-       shadow_pde->writable = guest_pde->writable;
-      } else {
-       // ??  What if guest pde is dirty a this point?
-       ((pde32_4MB_t *)guest_pde)->dirty = 0;
-       shadow_pde->writable = 0;
-      }
-    }
-  else if (shadow_pde_access == PT_ACCESS_OK) 
-    {
-      //
-      // PTE fault
-      //
-      pte32_t * shadow_pt = (pte32_t *)V3_VAddr( (void*)(addr_t) BASE_TO_PAGE_ADDR(shadow_pde->pt_base_addr) );
-
-      if (guest_pde->large_page == 0) {
-       pte32_t * guest_pt = NULL;
-       if (guest_pa_to_host_va(info, BASE_TO_PAGE_ADDR(guest_pde->pt_base_addr), (addr_t*)&guest_pt) == -1) {
-         // Machine check the guest
-         PrintDebug("Invalid Guest PTE Address: 0x%p\n", (void *)BASE_TO_PAGE_ADDR(guest_pde->pt_base_addr));
-         v3_raise_exception(info, MC_EXCEPTION);
-         return 0;
-       }
-       
-       if (handle_shadow_pte32_fault(info, fault_addr, error_code, shadow_pt, guest_pt)  == -1) {
-         PrintError("Error handling Page fault caused by PTE\n");
-         return -1;
-       }
-      } else if (guest_pde->large_page == 1) {
-       if (handle_large_pagefault_32(info, fault_addr, error_code, shadow_pt, (pde32_4MB_t *)guest_pde) == -1) {
-         PrintError("Error handling large pagefault\n");
-         return -1;
-       }
-      }
-    }
-  else if ((shadow_pde_access == PT_ACCESS_WRITE_ERROR) && 
-          (guest_pde->large_page == 1) && 
-          (((pde32_4MB_t *)guest_pde)->dirty == 0)) 
-    {
-      //
-      // Page Directory Entry marked read-only
-      // Its a large page and we need to update the dirty bit in the guest
-      //
-
-      PrintDebug("Large page write error... Setting dirty bit and returning\n");
-      ((pde32_4MB_t *)guest_pde)->dirty = 1;
-      shadow_pde->writable = guest_pde->writable;
-      return 0;
-      
-    } 
-  else if (shadow_pde_access == PT_ACCESS_USER_ERROR) 
-    {
-      //
-      // Page Directory Entry marked non-user
-      //      
-      PrintDebug("Shadow Paging User access error (shadow_pde_access=0x%x, guest_pde_access=0x%x)\n", 
-                shadow_pde_access, guest_pde_access);
-      inject_guest_pf(info, fault_addr, error_code);
-      return 0;
-    }
-  else 
-    {
-      // inject page fault in guest
-      inject_guest_pf(info, fault_addr, error_code);
-      PrintDebug("Unknown Error occurred (shadow_pde_access=%d)\n", shadow_pde_access);
-      PrintDebug("Manual Says to inject page fault into guest\n");
-#ifdef DEBUG_SHADOW_PAGING
-      PrintDebug("Guest PDE: (access=%d)\n\t", guest_pde_access);
-      PrintPTEntry(PAGE_PD32, fault_addr, guest_pde);
-      PrintDebug("Shadow PDE: (access=%d)\n\t", shadow_pde_access);
-      PrintPTEntry(PAGE_PD32, fault_addr, shadow_pde);
-#endif
-
-      return 0; 
-    }
-
-  PrintDebug("Returning end of PDE function (rip=%p)\n", (void *)(addr_t)(info->rip));
-  return 0;
-}
-
-
-
-/* The guest status checks have already been done,
- * only special case shadow checks remain
- */
-static int handle_large_pagefault_32(struct guest_info * info, 
-                                   addr_t fault_addr, pf_error_t error_code, 
-                                   pte32_t * shadow_pt, pde32_4MB_t * large_guest_pde) 
-{
-  pt_access_status_t shadow_pte_access = v3_can_access_pte32(shadow_pt, fault_addr, error_code);
-  pte32_t * shadow_pte = (pte32_t *)&(shadow_pt[PTE32_INDEX(fault_addr)]);
-  
-  if (shadow_pte_access == PT_ACCESS_OK) {
-    // Inconsistent state...
-    // Guest Re-Entry will flush tables and everything should now workd
-    PrintDebug("Inconsistent state... Guest re-entry should flush tlb\n");
-    return 0;
-  }
-
-  
-  if (shadow_pte_access == PT_ACCESS_NOT_PRESENT) {
-    // Get the guest physical address of the fault
-    addr_t guest_fault_pa = BASE_TO_PAGE_ADDR_4MB(large_guest_pde->page_base_addr) + PAGE_OFFSET_4MB(fault_addr);
-    host_region_type_t host_page_type = get_shadow_addr_type(info, guest_fault_pa);
-
-    if (host_page_type == HOST_REGION_INVALID) {
-      // Inject a machine check in the guest
-      PrintDebug("Invalid Guest Address in page table (0x%p)\n", (void *)guest_fault_pa);
-      v3_raise_exception(info, MC_EXCEPTION);
-      return 0;
-    }
-
-    if (host_page_type == HOST_REGION_PHYSICAL_MEMORY) {
-      struct shadow_page_state * state = &(info->shdw_pg_state);
-      addr_t shadow_pa = get_shadow_addr(info, guest_fault_pa);
-
-      shadow_pte->page_base_addr = PAGE_BASE_ADDR(shadow_pa);
-
-      shadow_pte->present = 1;
-
-      /* We are assuming that the PDE entry has precedence
-       * so the Shadow PDE will mirror the guest PDE settings, 
-       * and we don't have to worry about them here
-       * Allow everything
-       */
-      shadow_pte->user_page = 1;
-
-      if (find_pte_map(state->cached_ptes, PAGE_ADDR(guest_fault_pa)) != NULL) {
-       // Check if the entry is a page table...
-       PrintDebug("Marking page as Guest Page Table (large page)\n");
-       shadow_pte->vmm_info = PT32_GUEST_PT;
-       shadow_pte->writable = 0;
-      } else {
-       shadow_pte->writable = 1;
-      }
-
-
-      //set according to VMM policy
-      shadow_pte->write_through = 0;
-      shadow_pte->cache_disable = 0;
-      shadow_pte->global_page = 0;
-      //
-      
-    } else {
-      // Handle hooked pages as well as other special pages
-      if (handle_special_page_fault(info, fault_addr, guest_fault_pa, error_code) == -1) {
-       PrintError("Special Page Fault handler returned error for address: %p\n", (void *)fault_addr);
-       return -1;
-      }
-    }
-  } else if ((shadow_pte_access == PT_ACCESS_WRITE_ERROR) && 
-            (shadow_pte->vmm_info == PT32_GUEST_PT)) {
-
-    struct shadow_page_state * state = &(info->shdw_pg_state);
-    PrintDebug("Write operation on Guest PAge Table Page (large page)\n");
-    state->cached_cr3 = 0;
-    shadow_pte->writable = 1;
-
-  } else {
-    PrintError("Error in large page fault handler...\n");
-    PrintError("This case should have been handled at the top level handler\n");
-    return -1;
-  }
-
-  PrintDebug("Returning from large page fault handler\n");
-  return 0;
-}
-
-
-
-
-/* 
- * We assume the the guest pte pointer has already been translated to a host virtual address
- */
-static int handle_shadow_pte32_fault(struct guest_info * info, 
-                                    addr_t fault_addr, 
-                                    pf_error_t error_code,
-                                    pte32_t * shadow_pt, 
-                                    pte32_t * guest_pt) {
-
-  pt_access_status_t guest_pte_access;
-  pt_access_status_t shadow_pte_access;
-  pte32_t * guest_pte = (pte32_t *)&(guest_pt[PTE32_INDEX(fault_addr)]);;
-  pte32_t * shadow_pte = (pte32_t *)&(shadow_pt[PTE32_INDEX(fault_addr)]);
-
-
-  // Check the guest page permissions
-  guest_pte_access = v3_can_access_pte32(guest_pt, fault_addr, error_code);
-
-  // Check the shadow page permissions
-  shadow_pte_access = v3_can_access_pte32(shadow_pt, fault_addr, error_code);
-  
-#ifdef DEBUG_SHADOW_PAGING
-  PrintDebug("Guest PTE: (access=%d)\n\t", guest_pte_access);
-  PrintPTEntry(PAGE_PT32, fault_addr, guest_pte);
-  PrintDebug("Shadow PTE: (access=%d)\n\t", shadow_pte_access);
-  PrintPTEntry(PAGE_PT32, fault_addr, shadow_pte);
-#endif
-  
-  /* Was the page fault caused by the Guest's page tables? */
-  if (is_guest_pf(guest_pte_access, shadow_pte_access) == 1) {
-    PrintDebug("Access error injecting pf to guest (guest access error=%d) (pf error code=%d)\n", 
-              guest_pte_access, *(uint_t*)&error_code);    
-    inject_guest_pf(info, fault_addr, error_code);
-    return 0; 
-  }
-  
-  
-  if (shadow_pte_access == PT_ACCESS_OK) {
-    // Inconsistent state...
-    // Guest Re-Entry will flush page tables and everything should now work
-    PrintDebug("Inconsistent state... Guest re-entry should flush tlb\n");
-    return 0;
-  }
-
-
-  if (shadow_pte_access == PT_ACCESS_NOT_PRESENT) {
-
-    addr_t guest_pa = BASE_TO_PAGE_ADDR((addr_t)(guest_pte->page_base_addr)) +  PAGE_OFFSET(fault_addr);
-
-    // Page Table Entry Not Present
-    PrintDebug("guest_pa =%p\n", (void *)guest_pa);
-
-    host_region_type_t host_page_type = get_shadow_addr_type(info, guest_pa);
-
-    if (host_page_type == HOST_REGION_INVALID) {
-      // Inject a machine check in the guest
-      PrintDebug("Invalid Guest Address in page table (0x%p)\n", (void *)guest_pa);
-      v3_raise_exception(info, MC_EXCEPTION);
-      return 0;
-    }
-
-    // else...
-
-    if (host_page_type == HOST_REGION_PHYSICAL_MEMORY) {
-      struct shadow_page_state * state = &(info->shdw_pg_state);
-      addr_t shadow_pa = get_shadow_addr(info, guest_pa);
-      
-      shadow_pte->page_base_addr = PAGE_BASE_ADDR(shadow_pa);
-      
-      shadow_pte->present = guest_pte->present;
-      shadow_pte->user_page = guest_pte->user_page;
-      
-      //set according to VMM policy
-      shadow_pte->write_through = 0;
-      shadow_pte->cache_disable = 0;
-      shadow_pte->global_page = 0;
-      //
-      
-      guest_pte->accessed = 1;
-      
-      if (find_pte_map(state->cached_ptes, PAGE_ADDR(guest_pa)) != NULL) {
-       // Check if the entry is a page table...
-       PrintDebug("Marking page as Guest Page Table %d\n", shadow_pte->writable);
-       shadow_pte->vmm_info = PT32_GUEST_PT;
-      }
-
-      if (guest_pte->dirty == 1) {
-       shadow_pte->writable = guest_pte->writable;
-      } else if ((guest_pte->dirty == 0) && (error_code.write == 1)) {
-       shadow_pte->writable = guest_pte->writable;
-       guest_pte->dirty = 1;
-       
-       if (shadow_pte->vmm_info == PT32_GUEST_PT) {
-         // Well that was quick...
-         struct shadow_page_state * state = &(info->shdw_pg_state);
-         PrintDebug("Immediate Write operation on Guest PAge Table Page\n");
-         state->cached_cr3 = 0;
-       }
-
-      } else if ((guest_pte->dirty == 0) && (error_code.write == 0)) {  // was =
-       shadow_pte->writable = 0;
-      }
-
-
-
-    } else {
-      // Page fault handled by hook functions
-      if (handle_special_page_fault(info, fault_addr, guest_pa, error_code) == -1) {
-       PrintError("Special Page fault handler returned error for address: %p\n",  (void *)fault_addr);
-       return -1;
-      }
-    }
-
-  } else if ((shadow_pte_access == PT_ACCESS_WRITE_ERROR) &&
-            (guest_pte->dirty == 0)) {
-
-    PrintDebug("Shadow PTE Write Error\n");
-    guest_pte->dirty = 1;
-    shadow_pte->writable = guest_pte->writable;
-
-    if (shadow_pte->vmm_info == PT32_GUEST_PT) {
-      struct shadow_page_state * state = &(info->shdw_pg_state);
-      PrintDebug("Write operation on Guest PAge Table Page\n");
-      state->cached_cr3 = 0;
-    }
-    
-    return 0;
-
-  } else {
-    // Inject page fault into the guest        
-    inject_guest_pf(info, fault_addr, error_code);
-    PrintError("PTE Page fault fell through... Not sure if this should ever happen\n");
-    PrintError("Manual Says to inject page fault into guest\n");
-    return -1;
-  }
-
-  PrintDebug("Returning end of function\n");
-  return 0;
-}
-
-
-
-
-
-
-/* Currently Does not work with Segmentation!!! */
-int v3_handle_shadow_invlpg(struct guest_info * info)
-{
-  if (info->mem_mode != VIRTUAL_MEM) {
-    // Paging must be turned on...
-    // should handle with some sort of fault I think
-    PrintError("ERROR: INVLPG called in non paged mode\n");
-    return -1;
-  }
-  
-  
-  if (info->cpu_mode != PROTECTED) {
-    PrintError("Unsupported CPU mode (mode=%s)\n", v3_cpu_mode_to_str(info->cpu_mode));
-    return -1;
-  }
-  
-  uchar_t instr[15];
-  int index = 0;
-  
-  int ret = read_guest_va_memory(info, get_addr_linear(info, info->rip, &(info->segments.cs)), 15, instr);
-  if (ret != 15) {
-    PrintError("Could not read instruction 0x%p (ret=%d)\n",  (void *)(addr_t)(info->rip), ret);
-    return -1;
-  }
-  
-  
-  /* Can INVLPG work with Segments?? */
-  while (is_prefix_byte(instr[index])) {
-    index++;
-  }
-    
-    
-  if( (instr[index + 0] != (uchar_t) 0x0f) ||  
-      (instr[index + 1] != (uchar_t) 0x01) ) {
-    PrintError("invalid Instruction Opcode\n");
-    PrintTraceMemDump(instr, 15);
-    return -1;
-  }
-  
-  addr_t first_operand;
-  addr_t second_operand;
-  addr_t guest_cr3 =  CR3_TO_PDE32_PA(info->shdw_pg_state.guest_cr3);
-  
-  pde32_t * guest_pd = NULL;
-  
-  if (guest_pa_to_host_va(info, guest_cr3, (addr_t*)&guest_pd) == -1) {
-    PrintError("Invalid Guest PDE Address: 0x%p\n",  (void *)guest_cr3);
-    return -1;
-  }
-  
-  index += 2;
-
-  v3_operand_type_t addr_type = decode_operands32(&(info->vm_regs), instr + index, &index, &first_operand, &second_operand, REG32);
-  
-  if (addr_type != MEM_OPERAND) {
-    PrintError("Invalid Operand type\n");
-    return -1;
-  }
-  
-  pde32_t * shadow_pd = (pde32_t *)CR3_TO_PDE32_VA(info->ctrl_regs.cr3);
-  pde32_t * shadow_pde = (pde32_t *)&shadow_pd[PDE32_INDEX(first_operand)];
-  pde32_t * guest_pde;
-  
-  //PrintDebug("PDE Index=%d\n", PDE32_INDEX(first_operand));
-  //PrintDebug("FirstOperand = %x\n", first_operand);
-  
-  PrintDebug("Invalidating page for %p\n", (void *)first_operand);
-  
-  guest_pde = (pde32_t *)&(guest_pd[PDE32_INDEX(first_operand)]);
-  
-  if (guest_pde->large_page == 1) {
-    shadow_pde->present = 0;
-    PrintDebug("Invalidating Large Page\n");
-  } else
-    if (shadow_pde->present == 1) {
-      pte32_t * shadow_pt = (pte32_t *)(addr_t)BASE_TO_PAGE_ADDR(shadow_pde->pt_base_addr);
-      pte32_t * shadow_pte = (pte32_t *) V3_VAddr( (void*) &shadow_pt[PTE32_INDEX(first_operand)] );
-      
-#ifdef DEBUG_SHADOW_PAGING
-      PrintDebug("Setting not present\n");
-      PrintPTEntry(PAGE_PT32, first_operand, shadow_pte);
-#endif
-      
-      shadow_pte->present = 0;
-    }
-  
-  info->rip += index;
-  
-  return 0;
-}
-
-
diff --git a/palacios/src/palacios/vmm_shadow_paging_32.h b/palacios/src/palacios/vmm_shadow_paging_32.h
new file mode 100644 (file)
index 0000000..749d371
--- /dev/null
@@ -0,0 +1,529 @@
+
+static int cache_page_tables_32(struct guest_info * info, addr_t pde) {
+  struct shadow_page_state * state = &(info->shdw_pg_state);
+  addr_t pde_host_addr;
+  pde32_t * tmp_pde;
+  struct hashtable * pte_cache = NULL;
+  int i = 0;
+
+  if (pde == state->cached_cr3) {
+    return 1;
+  }
+
+  if (state->cached_ptes != NULL) {
+    hashtable_destroy(state->cached_ptes, 0, 0);
+    state->cached_ptes = NULL;
+  }
+
+  state->cached_cr3 = pde;
+
+  pte_cache = create_hashtable(0, &pte_hash_fn, &pte_equals);
+  state->cached_ptes = pte_cache;
+
+  if (guest_pa_to_host_va(info, pde, &pde_host_addr) == -1) {
+    PrintError("Could not lookup host address of guest PDE\n");
+    return -1;
+  }
+
+  tmp_pde = (pde32_t *)pde_host_addr;
+
+  add_pte_map(pte_cache, pde, pde_host_addr);
+
+
+  for (i = 0; i < MAX_PDE32_ENTRIES; i++) {
+    if ((tmp_pde[i].present) && (tmp_pde[i].large_page == 0)) {
+      addr_t pte_host_addr;
+
+      if (guest_pa_to_host_va(info, (addr_t)(BASE_TO_PAGE_ADDR(tmp_pde[i].pt_base_addr)), &pte_host_addr) == -1) {
+       PrintError("Could not lookup host address of guest PDE\n");
+       return -1;
+      }
+
+      add_pte_map(pte_cache, (addr_t)(BASE_TO_PAGE_ADDR(tmp_pde[i].pt_base_addr)), pte_host_addr); 
+    }
+  }
+
+  return 0;
+
+}
+
+
+
+// We assume that shdw_pg_state.guest_cr3 is pointing to the page tables we want to activate
+// We also assume that the CPU mode has not changed during this page table transition
+static inline int activate_shadow_pt_32(struct guest_info * info) {
+  struct cr3_32 * shadow_cr3 = (struct cr3_32 *)&(info->ctrl_regs.cr3);
+  struct cr3_32 * guest_cr3 = (struct cr3_32 *)&(info->shdw_pg_state.guest_cr3);
+  int cached = 0;
+  
+  // Check if shadow page tables are in the cache
+  cached = cache_page_tables_32(info, CR3_TO_PDE32_PA(*(addr_t *)guest_cr3));
+  
+  if (cached == -1) {
+    PrintError("CR3 Cache failed\n");
+    return -1;
+  } else if (cached == 0) {
+    addr_t shadow_pt;
+    
+    PrintDebug("New CR3 is different - flushing shadow page table %p\n", shadow_cr3 );
+    delete_page_tables_32(CR3_TO_PDE32_VA(*(uint_t*)shadow_cr3));
+    
+    shadow_pt = create_new_shadow_pt();
+    
+    shadow_cr3->pdt_base_addr = (addr_t)V3_PAddr((void *)(addr_t)PAGE_BASE_ADDR(shadow_pt));
+    PrintDebug( "Created new shadow page table %p\n", (void *)(addr_t)shadow_cr3->pdt_base_addr );
+  } else {
+    PrintDebug("Reusing cached shadow Page table\n");
+  }
+  
+  shadow_cr3->pwt = guest_cr3->pwt;
+  shadow_cr3->pcd = guest_cr3->pcd;
+  
+  return 0;
+}
+
+/* 
+ * *
+ * * 
+ * * 32 bit Page table fault handlers
+ * *
+ * *
+ */
+static int handle_large_pagefault_32(struct guest_info * info, 
+                                   addr_t fault_addr, pf_error_t error_code, 
+                                    pte32_t * shadow_pt, pde32_4MB_t * large_guest_pde);
+
+static int handle_shadow_pte32_fault(struct guest_info * info, 
+                                    addr_t fault_addr, 
+                                    pf_error_t error_code,
+                                    pte32_t * shadow_pt, 
+                                    pte32_t * guest_pt);
+
+
+static inline int handle_shadow_pagefault_32(struct guest_info * info, addr_t fault_addr, pf_error_t error_code) {
+  pde32_t * guest_pd = NULL;
+  pde32_t * shadow_pd = CR3_TO_PDE32_VA(info->ctrl_regs.cr3);
+  addr_t guest_cr3 = CR3_TO_PDE32_PA(info->shdw_pg_state.guest_cr3);
+  pt_access_status_t guest_pde_access;
+  pt_access_status_t shadow_pde_access;
+  pde32_t * guest_pde = NULL;
+  pde32_t * shadow_pde = (pde32_t *)&(shadow_pd[PDE32_INDEX(fault_addr)]);
+
+  PrintDebug("Shadow page fault handler: %p\n", (void*) fault_addr );
+
+  if (guest_pa_to_host_va(info, guest_cr3, (addr_t*)&guest_pd) == -1) {
+    PrintError("Invalid Guest PDE Address: 0x%p\n",  (void *)guest_cr3);
+    return -1;
+  } 
+
+  guest_pde = (pde32_t *)&(guest_pd[PDE32_INDEX(fault_addr)]);
+
+
+  // Check the guest page permissions
+  guest_pde_access = v3_can_access_pde32(guest_pd, fault_addr, error_code);
+
+  // Check the shadow page permissions
+  shadow_pde_access = v3_can_access_pde32(shadow_pd, fault_addr, error_code);
+  
+  /* Was the page fault caused by the Guest's page tables? */
+  if (is_guest_pf(guest_pde_access, shadow_pde_access) == 1) {
+    PrintDebug("Injecting PDE pf to guest: (guest access error=%d) (pf error code=%d)\n", 
+              *(uint_t *)&guest_pde_access, *(uint_t *)&error_code);
+    inject_guest_pf(info, fault_addr, error_code);
+    return 0;
+  }
+
+  
+  if (shadow_pde_access == PT_ACCESS_NOT_PRESENT) 
+    {
+      pte32_t * shadow_pt =  (pte32_t *)create_new_shadow_pt();
+
+      shadow_pde->present = 1;
+      shadow_pde->user_page = guest_pde->user_page;
+      //    shadow_pde->large_page = guest_pde->large_page;
+      shadow_pde->large_page = 0;
+      
+
+      // VMM Specific options
+      shadow_pde->write_through = 0;
+      shadow_pde->cache_disable = 0;
+      shadow_pde->global_page = 0;
+      //
+      
+      guest_pde->accessed = 1;
+      
+      shadow_pde->pt_base_addr = PAGE_BASE_ADDR((addr_t)V3_PAddr(shadow_pt));
+      
+      if (guest_pde->large_page == 0) {
+       pte32_t * guest_pt = NULL;
+       shadow_pde->writable = guest_pde->writable;
+
+       if (guest_pa_to_host_va(info, BASE_TO_PAGE_ADDR(guest_pde->pt_base_addr), (addr_t*)&guest_pt) == -1) {
+         // Machine check the guest
+         PrintDebug("Invalid Guest PTE Address: 0x%p\n", (void *)BASE_TO_PAGE_ADDR(guest_pde->pt_base_addr));
+         v3_raise_exception(info, MC_EXCEPTION);
+         return 0;
+       }
+
+       if (handle_shadow_pte32_fault(info, fault_addr, error_code, shadow_pt, guest_pt)  == -1) {
+         PrintError("Error handling Page fault caused by PTE\n");
+         return -1;
+       }
+      } else {
+       // ??  What if guest pde is dirty a this point?
+       ((pde32_4MB_t *)guest_pde)->dirty = 0;
+       shadow_pde->writable = 0;
+      }
+    }
+  else if (shadow_pde_access == PT_ACCESS_OK) 
+    {
+      //
+      // PTE fault
+      //
+      pte32_t * shadow_pt = (pte32_t *)V3_VAddr( (void*)(addr_t) BASE_TO_PAGE_ADDR(shadow_pde->pt_base_addr) );
+
+      if (guest_pde->large_page == 0) {
+       pte32_t * guest_pt = NULL;
+
+       if (guest_pa_to_host_va(info, BASE_TO_PAGE_ADDR(guest_pde->pt_base_addr), (addr_t*)&guest_pt) == -1) {
+         // Machine check the guest
+         PrintDebug("Invalid Guest PTE Address: 0x%p\n", (void *)BASE_TO_PAGE_ADDR(guest_pde->pt_base_addr));
+         v3_raise_exception(info, MC_EXCEPTION);
+         return 0;
+       }
+       
+       if (handle_shadow_pte32_fault(info, fault_addr, error_code, shadow_pt, guest_pt)  == -1) {
+         PrintError("Error handling Page fault caused by PTE\n");
+         return -1;
+       }
+      } else if (guest_pde->large_page == 1) {
+       if (handle_large_pagefault_32(info, fault_addr, error_code, shadow_pt, (pde32_4MB_t *)guest_pde) == -1) {
+         PrintError("Error handling large pagefault\n");
+         return -1;
+       }
+      }
+    }
+  else if ((shadow_pde_access == PT_ACCESS_WRITE_ERROR) && 
+          (guest_pde->large_page == 1) && 
+          (((pde32_4MB_t *)guest_pde)->dirty == 0)) 
+    {
+      //
+      // Page Directory Entry marked read-only
+      // Its a large page and we need to update the dirty bit in the guest
+      //
+
+      PrintDebug("Large page write error... Setting dirty bit and returning\n");
+      ((pde32_4MB_t *)guest_pde)->dirty = 1;
+      shadow_pde->writable = guest_pde->writable;
+      return 0;
+      
+    } 
+  else if (shadow_pde_access == PT_ACCESS_USER_ERROR) 
+    {
+      //
+      // Page Directory Entry marked non-user
+      //      
+      PrintDebug("Shadow Paging User access error (shadow_pde_access=0x%x, guest_pde_access=0x%x)\n", 
+                shadow_pde_access, guest_pde_access);
+      inject_guest_pf(info, fault_addr, error_code);
+      return 0;
+    }
+  else 
+    {
+      // inject page fault in guest
+      inject_guest_pf(info, fault_addr, error_code);
+      PrintDebug("Unknown Error occurred (shadow_pde_access=%d)\n", shadow_pde_access);
+      PrintDebug("Manual Says to inject page fault into guest\n");
+#ifdef DEBUG_SHADOW_PAGING
+      PrintDebug("Guest PDE: (access=%d)\n\t", guest_pde_access);
+      PrintPTEntry(PAGE_PD32, fault_addr, guest_pde);
+      PrintDebug("Shadow PDE: (access=%d)\n\t", shadow_pde_access);
+      PrintPTEntry(PAGE_PD32, fault_addr, shadow_pde);
+#endif
+
+      return 0; 
+    }
+
+  PrintDebug("Returning end of PDE function (rip=%p)\n", (void *)(addr_t)(info->rip));
+  return 0;
+}
+
+
+
+/* The guest status checks have already been done,
+ * only special case shadow checks remain
+ */
+static int handle_large_pagefault_32(struct guest_info * info, 
+                                   addr_t fault_addr, pf_error_t error_code, 
+                                   pte32_t * shadow_pt, pde32_4MB_t * large_guest_pde) 
+{
+  pt_access_status_t shadow_pte_access = v3_can_access_pte32(shadow_pt, fault_addr, error_code);
+  pte32_t * shadow_pte = (pte32_t *)&(shadow_pt[PTE32_INDEX(fault_addr)]);
+  addr_t guest_fault_pa = BASE_TO_PAGE_ADDR_4MB(large_guest_pde->page_base_addr) + PAGE_OFFSET_4MB(fault_addr);  
+
+  struct v3_shadow_region * shdw_reg = v3_get_shadow_region(info, guest_fault_pa);
+
+  if ((shdw_reg == NULL) || 
+      (shdw_reg->host_type == SHDW_REGION_INVALID)) {
+    // Inject a machine check in the guest
+    PrintDebug("Invalid Guest Address in page table (0x%p)\n", (void *)guest_fault_pa);
+    v3_raise_exception(info, MC_EXCEPTION);
+    return -1;
+  }
+
+  if (shadow_pte_access == PT_ACCESS_OK) {
+    // Inconsistent state...
+    // Guest Re-Entry will flush tables and everything should now workd
+    PrintDebug("Inconsistent state... Guest re-entry should flush tlb\n");
+    return 0;
+  }
+
+  
+  if (shadow_pte_access == PT_ACCESS_NOT_PRESENT) {
+    // Get the guest physical address of the fault
+
+    if ((shdw_reg->host_type == SHDW_REGION_ALLOCATED) || 
+       (shdw_reg->host_type == SHDW_REGION_WRITE_HOOK)) {
+      struct shadow_page_state * state = &(info->shdw_pg_state);
+      addr_t shadow_pa = v3_get_shadow_addr(shdw_reg, guest_fault_pa);
+
+      shadow_pte->page_base_addr = PAGE_BASE_ADDR(shadow_pa);
+
+      shadow_pte->present = 1;
+
+      /* We are assuming that the PDE entry has precedence
+       * so the Shadow PDE will mirror the guest PDE settings, 
+       * and we don't have to worry about them here
+       * Allow everything
+       */
+      shadow_pte->user_page = 1;
+
+      if (find_pte_map(state->cached_ptes, PAGE_ADDR(guest_fault_pa)) != NULL) {
+       // Check if the entry is a page table...
+       PrintDebug("Marking page as Guest Page Table (large page)\n");
+       shadow_pte->vmm_info = PT32_GUEST_PT;
+       shadow_pte->writable = 0;
+      } else if (shdw_reg->host_type == SHDW_REGION_WRITE_HOOK) {
+       shadow_pte->writable = 0;
+      } else {
+       shadow_pte->writable = 1;
+      }
+
+      //set according to VMM policy
+      shadow_pte->write_through = 0;
+      shadow_pte->cache_disable = 0;
+      shadow_pte->global_page = 0;
+      //
+      
+    } else {
+      // Handle hooked pages as well as other special pages
+      //      if (handle_special_page_fault(info, fault_addr, guest_fault_pa, error_code) == -1) {
+
+      if (v3_handle_mem_full_hook(info, fault_addr, guest_fault_pa, shdw_reg, error_code) == -1) {
+       PrintError("Special Page Fault handler returned error for address: %p\n", (void *)fault_addr);
+       return -1;
+      }
+    }
+  } else if (shadow_pte_access == PT_ACCESS_WRITE_ERROR) {
+
+    if (shdw_reg->host_type == SHDW_REGION_WRITE_HOOK) {
+
+      if (v3_handle_mem_wr_hook(info, fault_addr, guest_fault_pa, shdw_reg, error_code) == -1) {
+       PrintError("Special Page Fault handler returned error for address: %p\n", (void *)fault_addr);
+       return -1;
+      }
+    } else if (shadow_pte->vmm_info == PT32_GUEST_PT) {
+      struct shadow_page_state * state = &(info->shdw_pg_state);
+      PrintDebug("Write operation on Guest PAge Table Page (large page)\n");
+      state->cached_cr3 = 0;
+      shadow_pte->writable = 1;
+    }
+
+  } else {
+    PrintError("Error in large page fault handler...\n");
+    PrintError("This case should have been handled at the top level handler\n");
+    return -1;
+  }
+
+  PrintDebug("Returning from large page fault handler\n");
+  return 0;
+}
+
+
+
+
+/* 
+ * We assume the the guest pte pointer has already been translated to a host virtual address
+ */
+static int handle_shadow_pte32_fault(struct guest_info * info, 
+                                    addr_t fault_addr, 
+                                    pf_error_t error_code,
+                                    pte32_t * shadow_pt, 
+                                    pte32_t * guest_pt) {
+
+  pt_access_status_t guest_pte_access;
+  pt_access_status_t shadow_pte_access;
+  pte32_t * guest_pte = (pte32_t *)&(guest_pt[PTE32_INDEX(fault_addr)]);;
+  pte32_t * shadow_pte = (pte32_t *)&(shadow_pt[PTE32_INDEX(fault_addr)]);
+  addr_t guest_pa = BASE_TO_PAGE_ADDR((addr_t)(guest_pte->page_base_addr)) +  PAGE_OFFSET(fault_addr);
+
+  struct v3_shadow_region * shdw_reg =  v3_get_shadow_region(info, guest_pa);
+
+  if ((shdw_reg == NULL) || 
+      (shdw_reg->host_type == SHDW_REGION_INVALID)) {
+    // Inject a machine check in the guest
+    PrintDebug("Invalid Guest Address in page table (0x%p)\n", (void *)guest_pa);
+    v3_raise_exception(info, MC_EXCEPTION);
+    return 0;
+  }
+
+  // Check the guest page permissions
+  guest_pte_access = v3_can_access_pte32(guest_pt, fault_addr, error_code);
+
+  // Check the shadow page permissions
+  shadow_pte_access = v3_can_access_pte32(shadow_pt, fault_addr, error_code);
+  
+#ifdef DEBUG_SHADOW_PAGING
+  PrintDebug("Guest PTE: (access=%d)\n\t", guest_pte_access);
+  PrintPTEntry(PAGE_PT32, fault_addr, guest_pte);
+  PrintDebug("Shadow PTE: (access=%d)\n\t", shadow_pte_access);
+  PrintPTEntry(PAGE_PT32, fault_addr, shadow_pte);
+#endif
+  
+  /* Was the page fault caused by the Guest's page tables? */
+  if (is_guest_pf(guest_pte_access, shadow_pte_access) == 1) {
+    PrintDebug("Access error injecting pf to guest (guest access error=%d) (pf error code=%d)\n", 
+              guest_pte_access, *(uint_t*)&error_code);    
+    inject_guest_pf(info, fault_addr, error_code);
+    return 0; 
+  }
+  
+  
+  if (shadow_pte_access == PT_ACCESS_OK) {
+    // Inconsistent state...
+    // Guest Re-Entry will flush page tables and everything should now work
+    PrintDebug("Inconsistent state... Guest re-entry should flush tlb\n");
+    return 0;
+  }
+
+
+  if (shadow_pte_access == PT_ACCESS_NOT_PRESENT) {
+    // Page Table Entry Not Present
+    PrintDebug("guest_pa =%p\n", (void *)guest_pa);
+
+    if ((shdw_reg->host_type == SHDW_REGION_ALLOCATED) ||
+       (shdw_reg->host_type == SHDW_REGION_WRITE_HOOK)) {
+      struct shadow_page_state * state = &(info->shdw_pg_state);
+      addr_t shadow_pa = v3_get_shadow_addr(shdw_reg, guest_pa);
+      
+      shadow_pte->page_base_addr = PAGE_BASE_ADDR(shadow_pa);
+      
+      shadow_pte->present = guest_pte->present;
+      shadow_pte->user_page = guest_pte->user_page;
+      
+      //set according to VMM policy
+      shadow_pte->write_through = 0;
+      shadow_pte->cache_disable = 0;
+      shadow_pte->global_page = 0;
+      //
+      
+      guest_pte->accessed = 1;
+      
+      if (find_pte_map(state->cached_ptes, PAGE_ADDR(guest_pa)) != NULL) {
+       // Check if the entry is a page table...
+       PrintDebug("Marking page as Guest Page Table %d\n", shadow_pte->writable);
+       shadow_pte->vmm_info = PT32_GUEST_PT;
+      }
+
+      if (shdw_reg->host_type == SHDW_REGION_WRITE_HOOK) {
+       shadow_pte->writable = 0;
+      } else if (guest_pte->dirty == 1) {
+       shadow_pte->writable = guest_pte->writable;
+      } else if ((guest_pte->dirty == 0) && (error_code.write == 1)) {
+       shadow_pte->writable = guest_pte->writable;
+       guest_pte->dirty = 1;
+       
+       if (shadow_pte->vmm_info == PT32_GUEST_PT) {
+         // Well that was quick...
+         struct shadow_page_state * state = &(info->shdw_pg_state);
+         PrintDebug("Immediate Write operation on Guest PAge Table Page\n");
+         state->cached_cr3 = 0;
+       }
+
+      } else if ((guest_pte->dirty == 0) && (error_code.write == 0)) {  // was =
+       shadow_pte->writable = 0;
+      }
+
+    } else {
+      // Page fault handled by hook functions
+
+      if (v3_handle_mem_full_hook(info, fault_addr, guest_pa, shdw_reg, error_code) == -1) {
+       PrintError("Special Page fault handler returned error for address: %p\n",  (void *)fault_addr);
+       return -1;
+      }
+    }
+  } else if (shadow_pte_access == PT_ACCESS_WRITE_ERROR) {
+    guest_pte->dirty = 1;
+
+    if (shdw_reg->host_type == SHDW_REGION_WRITE_HOOK) {
+      if (v3_handle_mem_wr_hook(info, fault_addr, guest_pa, shdw_reg, error_code) == -1) {
+       PrintError("Special Page fault handler returned error for address: %p\n",  (void *)fault_addr);
+       return -1;
+      }
+    } else {
+      PrintDebug("Shadow PTE Write Error\n");
+      shadow_pte->writable = guest_pte->writable;
+    }
+
+    if (shadow_pte->vmm_info == PT32_GUEST_PT) {
+      struct shadow_page_state * state = &(info->shdw_pg_state);
+      PrintDebug("Write operation on Guest PAge Table Page\n");
+      state->cached_cr3 = 0;
+    }
+
+    return 0;
+
+  } else {
+    // Inject page fault into the guest        
+    inject_guest_pf(info, fault_addr, error_code);
+    PrintError("PTE Page fault fell through... Not sure if this should ever happen\n");
+    PrintError("Manual Says to inject page fault into guest\n");
+    return -1;
+  }
+
+  PrintDebug("Returning end of function\n");
+  return 0;
+}
+
+
+
+/* If we start to optimize we should look up the guest pages in the cache... */
+static inline int handle_shadow_invlpg_32(struct guest_info * info, addr_t vaddr) {
+  pde32_t * shadow_pd = (pde32_t *)CR3_TO_PDE32_VA(info->ctrl_regs.cr3);
+  pde32_t * shadow_pde = (pde32_t *)&shadow_pd[PDE32_INDEX(vaddr)];
+
+  addr_t guest_cr3 =  CR3_TO_PDE32_PA(info->shdw_pg_state.guest_cr3);
+  pde32_t * guest_pd = NULL;
+  pde32_t * guest_pde;
+
+  if (guest_pa_to_host_va(info, guest_cr3, (addr_t*)&guest_pd) == -1) {
+    PrintError("Invalid Guest PDE Address: 0x%p\n",  (void *)guest_cr3);
+    return -1;
+  }
+  
+  guest_pde = (pde32_t *)&(guest_pd[PDE32_INDEX(vaddr)]);
+  
+  if (guest_pde->large_page == 1) {
+    shadow_pde->present = 0;
+    PrintDebug("Invalidating Large Page\n");
+  } else if (shadow_pde->present == 1) {
+    pte32_t * shadow_pt = (pte32_t *)(addr_t)BASE_TO_PAGE_ADDR_4KB(shadow_pde->pt_base_addr);
+    pte32_t * shadow_pte = (pte32_t *) V3_VAddr( (void*) &shadow_pt[PTE32_INDEX(vaddr)] );
+    
+    PrintDebug("Setting not present\n");
+    
+    shadow_pte->present = 0;
+  }
+  return 0;
+}
diff --git a/palacios/src/palacios/vmm_shadow_paging_32pae.h b/palacios/src/palacios/vmm_shadow_paging_32pae.h
new file mode 100644 (file)
index 0000000..625552e
--- /dev/null
@@ -0,0 +1,35 @@
+
+
+
+static inline int activate_shadow_pt_32pae(struct guest_info * info) {
+  PrintError("Activating 32 bit PAE page tables not implemented\n");
+  return -1;
+}
+
+
+
+
+
+
+/* 
+ * *
+ * * 
+ * * 32 bit PAE  Page table fault handlers
+ * *
+ * *
+ */
+
+static inline int handle_shadow_pagefault_32pae(struct guest_info * info, addr_t fault_addr, pf_error_t error_code) {
+  PrintError("32 bit PAE shadow paging not implemented\n");
+  return -1;
+}
+
+
+static inline int handle_shadow_invlpg_32pae(struct guest_info * info, addr_t vaddr) {
+  PrintError("32 bit PAE shadow paging not implemented\n");
+  return -1;
+}
+
+
+
+
diff --git a/palacios/src/palacios/vmm_shadow_paging_64.h b/palacios/src/palacios/vmm_shadow_paging_64.h
new file mode 100644 (file)
index 0000000..fbd3f3f
--- /dev/null
@@ -0,0 +1,44 @@
+
+static inline int activate_shadow_pt_64(struct guest_info * info) {
+  //  struct cr3_64 * shadow_cr3 = (struct cr3_64 *)&(info->ctrl_regs.cr3);
+  return -1;
+}
+
+
+
+
+
+
+/* 
+ * *
+ * * 
+ * * 64 bit Page table fault handlers
+ * *
+ * *
+ */
+
+static inline int handle_shadow_pagefault_64(struct guest_info * info, addr_t fault_addr, pf_error_t error_code) {
+  pt_access_status_t guest_access;
+  pt_access_status_t shadow_access;
+  int ret; 
+  PrintDebug("64 bit shadow page fault\n");
+
+  ret = v3_check_guest_pt_32(info, info->shdw_pg_state.guest_cr3, fault_addr, error_code, &guest_access);
+
+  PrintDebug("Guest Access Check: %d (access=%d)\n", ret, guest_access);
+
+  ret = v3_check_host_pt_32(info->ctrl_regs.cr3, fault_addr, error_code, &shadow_access);
+
+  PrintDebug("Shadow Access Check: %d (access=%d)\n", ret, shadow_access);
+  
+
+  PrintError("64 bit shadow paging not implemented\n");
+  return -1;
+}
+
+
+static inline int handle_shadow_invlpg_64(struct guest_info * info, addr_t vaddr) {
+  PrintError("64 bit shadow paging not implemented\n");
+  return -1;
+}
index 9b16078..73dd4ce 100644 (file)
 #include <xed/xed-interface.h>
 #include "vm_guest.h"
 #include "test.h"
+
 #else
+
 #include <palacios/vmm_decoder.h>
 #include <palacios/vmm_xed.h>
 #include <xed/xed-interface.h>
 #include <palacios/vm_guest.h>
 #include <palacios/vmm.h>
-
-
 #endif
 
 
@@ -42,7 +42,9 @@
 
 
 
-static xed_state_t decoder_state;
+
+static uint_t tables_inited = 0;
+
 
 #define GPR_REGISTER     0
 #define SEGMENT_REGISTER 1
@@ -65,12 +67,16 @@ static const ullong_t mask_8 = 0xffffffffffffffffLL;
       switch (length) {                                \
       case 1:                                  \
        mask = mask_1;                          \
+       break;                                  \
       case 2:                                  \
        mask = mask_2;                          \
+       break;                                  \
       case 4:                                  \
        mask = mask_4;                          \
+       break;                                  \
       case 8:                                  \
        mask = mask_8;                          \
+       break;                                  \
       }                                                \
       val & mask;})                            \
 
@@ -88,8 +94,8 @@ struct memory_operand {
 
 
 
-// This returns a pointer to a V3_OPCODE_[*] array defined in vmm_decoder.h
-static int get_opcode(xed_iform_enum_t iform, addr_t * opcode);
+
+static v3_op_type_t get_opcode(xed_iform_enum_t iform);
 
 static int xed_reg_to_v3_reg(struct guest_info * info, xed_reg_enum_t xed_reg, addr_t * v3_reg, uint_t * reg_len);
 static int get_memory_operand(struct guest_info * info,  xed_decoded_inst_t * xed_instr, uint_t index, struct x86_operand * operand);
@@ -139,9 +145,18 @@ static int is_flags_reg(xed_reg_enum_t xed_reg) {
 
 
 
-int v3_init_decoder() {
-  xed_tables_init();
-  xed_state_zero(&decoder_state);
+int v3_init_decoder(struct guest_info * info) {
+  // Global library initialization, only do it once
+  if (tables_inited == 0) {
+    xed_tables_init();
+    tables_inited = 1;
+  }
+
+  xed_state_t * decoder_state = (xed_state_t *)V3_Malloc(sizeof(xed_state_t));
+  xed_state_zero(decoder_state);
+
+  info->decoder_state = decoder_state;
+
   return 0;
 }
 
@@ -152,13 +167,13 @@ int v3_basic_mem_decode(struct guest_info * info, addr_t instr_ptr, struct basic
   xed_error_enum_t xed_error;
   
 
-  if (set_decoder_mode(info, &decoder_state) == -1) {
+  if (set_decoder_mode(info, info->decoder_state) == -1) {
     PrintError("Could not set decoder mode\n");
     return -1;
   }
 
 
-  xed_decoded_inst_zero_set_mode(&xed_instr, &decoder_state);
+  xed_decoded_inst_zero_set_mode(&xed_instr, info->decoder_state);
 
   xed_error = xed_decode(&xed_instr, 
                         REINTERPRET_CAST(const xed_uint8_t *, instr_ptr), 
@@ -198,21 +213,55 @@ int v3_basic_mem_decode(struct guest_info * info, addr_t instr_ptr, struct basic
 }
 
 
+static int decode_string_op(struct guest_info * info, 
+                           xed_decoded_inst_t * xed_instr,  const xed_inst_t * xi,
+                           struct x86_instr * instr) {
+
+  PrintDebug("String operation\n");
+
+  if (instr->op_type == V3_OP_MOVS) {
+    instr->num_operands = 2;
+
+    if (get_memory_operand(info, xed_instr, 0, &(instr->dst_operand)) == -1) {
+      PrintError("Could not get Destination memory operand\n");
+      return -1;
+    }
+
+    if (get_memory_operand(info, xed_instr, 1, &(instr->src_operand)) == -1) {
+      PrintError("Could not get Source memory operand\n");
+      return -1;
+    }
+
+    if (instr->prefixes.rep == 1) {
+      addr_t reg_addr = 0;
+      uint_t reg_length = 0;
+
+      xed_reg_to_v3_reg(info, xed_decoded_inst_get_reg(xed_instr, XED_OPERAND_REG0), &reg_addr, &reg_length);
+      instr->str_op_length = MASK(*(addr_t *)reg_addr, reg_length);
+    } else {
+      instr->str_op_length = 1;
+    }
+
+  }
+
+  return 0;
+}
+
+
 
 int v3_decode(struct guest_info * info, addr_t instr_ptr, struct x86_instr * instr) {
   xed_decoded_inst_t xed_instr;
   xed_error_enum_t xed_error;
 
 
+  v3_get_prefixes((uchar_t *)instr_ptr, &(instr->prefixes));
 
-  if (set_decoder_mode(info, &decoder_state) == -1) {
+  if (set_decoder_mode(info, info->decoder_state) == -1) {
     PrintError("Could not set decoder mode\n");
     return -1;
   }
 
-
-
-  xed_decoded_inst_zero_set_mode(&xed_instr, &decoder_state);
+  xed_decoded_inst_zero_set_mode(&xed_instr, info->decoder_state);
 
   xed_error = xed_decode(&xed_instr, 
                         REINTERPRET_CAST(const xed_uint8_t *, instr_ptr), 
@@ -227,13 +276,34 @@ int v3_decode(struct guest_info * info, addr_t instr_ptr, struct x86_instr * ins
   const xed_inst_t * xi = xed_decoded_inst_inst(&xed_instr);
   
   instr->instr_length = xed_decoded_inst_get_length(&xed_instr);
-  instr->num_operands = xed_decoded_inst_noperands(&xed_instr);
+
 
   xed_iform_enum_t iform = xed_decoded_inst_get_iform_enum(&xed_instr);
 
+#ifdef DEBUG_XED
+  xed_iclass_enum_t iclass = xed_decoded_inst_get_iclass(&xed_instr);
+
+  PrintDebug("iform=%s, iclass=%s\n", xed_iform_enum_t2str(iform), xed_iclass_enum_t2str(iclass));
+#endif
 
-  PrintDebug("iform=%s\n", xed_iform_enum_t2str(iform));
 
+  if ((instr->op_type = get_opcode(iform)) == V3_INVALID_OP) {
+    PrintError("Could not get opcode. (iform=%s)\n", xed_iform_enum_t2str(iform));
+    return -1;
+  }
+
+
+  // We special case the string operations...
+  if (xed_decoded_inst_get_category(&xed_instr) == XED_CATEGORY_STRINGOP) {
+    instr->is_str_op = 1;
+    return decode_string_op(info, &xed_instr, xi, instr); 
+  } else {
+    instr->is_str_op = 0;
+    instr->str_op_length = 0;
+  }
+
+
+  instr->num_operands = xed_decoded_inst_noperands(&xed_instr);
 
   if (instr->num_operands > 3) {
     PrintDebug("Special Case Not Handled\n");
@@ -253,16 +323,6 @@ int v3_decode(struct guest_info * info, addr_t instr_ptr, struct x86_instr * ins
 
 
 
-
-
-  if (get_opcode(iform, &(instr->opcode)) == -1) {
-    PrintError("Could not get opcode. (iform=%s)\n", xed_iform_enum_t2str(iform));
-    return -1;
-  }
-
-
-
-
   //PrintDebug("Number of operands: %d\n", instr->num_operands);
   //PrintDebug("INSTR length: %d\n", instr->instr_length);
 
@@ -286,7 +346,7 @@ int v3_decode(struct guest_info * info, addr_t instr_ptr, struct x86_instr * ins
                                          xed_reg, 
                                          &(v3_op->operand), 
                                          &(v3_op->size));
-                                         
+
       if (v3_reg_type == -1) {
        PrintError("First operand is an Unhandled Operand: %s\n", xed_reg_enum_t2str(xed_reg));
        v3_op->type = INVALID_OPERAND;
@@ -303,16 +363,6 @@ int v3_decode(struct guest_info * info, addr_t instr_ptr, struct x86_instr * ins
 
       case XED_OPERAND_MEM0:
        {
-         /*
-           struct x86_operand * operand = &(instr->dst_operand);
-           
-           if (xed_decoded_inst_mem_read(&xed_instr, 0)) {
-           operand = &(instr->src_operand);
-           } else if (xed_decoded_inst_mem_written(&xed_instr, 0)) {
-           operand = &(instr->dst_operand);
-           }
-         */
-         
          if (get_memory_operand(info, &xed_instr, 0, v3_op) == -1) {
            PrintError("Could not get first memory operand\n");
            return -1;
@@ -379,15 +429,6 @@ int v3_decode(struct guest_info * info, addr_t instr_ptr, struct x86_instr * ins
 
       case XED_OPERAND_MEM0:
        {
-
-         /*
-         if (xed_decoded_inst_mem_read(&xed_instr, 0)) {
-           v3_op = &(instr->src_operand);
-         } else if (xed_decoded_inst_mem_written(&xed_instr, 0)) {
-           v3_op = &(instr->dst_operand);
-         }
-         */
-
          if (get_memory_operand(info, &xed_instr, 0, v3_op) == -1) {
            PrintError("Could not get first memory operand\n");
            return -1;
@@ -535,9 +576,8 @@ static int get_memory_operand(struct guest_info * info,  xed_decoded_inst_t * xe
   if (disp_bits) {
     xed_int64_t xed_disp = xed_decoded_inst_get_memory_displacement(xed_instr, op_index);
 
-    mem_op.displacement_size = disp_bits / 8;
+    mem_op.displacement_size = disp_bits;
     mem_op.displacement = xed_disp;
-    
   }
 
   operand->type = MEM_OPERAND;
@@ -545,17 +585,21 @@ static int get_memory_operand(struct guest_info * info,  xed_decoded_inst_t * xe
   
   
 
-  PrintDebug("Struct: Seg=%x, base=%x, index=%x, scale=%x, displacement=%x\n", 
-            mem_op.segment, mem_op.base, mem_op.index, mem_op.scale, mem_op.displacement);
+  PrintDebug("Struct: Seg=%p, base=%p, index=%p, scale=%p, displacement=%p (size=%d)\n", 
+            (void *)mem_op.segment, (void*)mem_op.base, (void *)mem_op.index, 
+            (void *)mem_op.scale, (void *)(addr_t)mem_op.displacement, mem_op.displacement_size);
+
 
+  PrintDebug("operand size: %d\n", operand->size);
 
-  seg = mem_op.segment;
+  seg = MASK(mem_op.segment, mem_op.segment_size);
   base = MASK(mem_op.base, mem_op.base_size);
   index = MASK(mem_op.index, mem_op.index_size);
   scale = mem_op.scale;
   displacement = MASK(mem_op.displacement, mem_op.displacement_size);
 
-  PrintDebug("Seg=%x, base=%x, index=%x, scale=%x, displacement=%x\n", seg, base, index, scale, displacement);
+  PrintDebug("Seg=%p, base=%p, index=%p, scale=%p, displacement=%p\n", 
+            (void *)seg, (void *)base, (void *)index, (void *)scale, (void *)(addr_t)displacement);
   
   operand->operand = seg + base + (scale * index) + displacement;
   return 0;
@@ -773,7 +817,7 @@ static int xed_reg_to_v3_reg(struct guest_info * info, xed_reg_enum_t xed_reg, a
     return CTRL_REGISTER;
   case XED_REG_CR4:
     *v3_reg = (addr_t)&(info->ctrl_regs.cr4);
-    *reg_len = 4;
+   *reg_len = 4;
     return CTRL_REGISTER;
   case XED_REG_CR8:
     *v3_reg = (addr_t)&(info->ctrl_regs.cr8);
@@ -974,37 +1018,154 @@ static int xed_reg_to_v3_reg(struct guest_info * info, xed_reg_enum_t xed_reg, a
 
 
 
-static int get_opcode(xed_iform_enum_t iform, addr_t * opcode) {
+static v3_op_type_t get_opcode(xed_iform_enum_t iform) {
 
   switch (iform) {
+
+    /* Control Instructions */
+
   case XED_IFORM_MOV_CR_GPR64_CR:
   case XED_IFORM_MOV_CR_GPR32_CR:
-    *opcode = (addr_t)&V3_OPCODE_MOVCR2;
-    break;
+    return V3_OP_MOVCR2;
 
   case XED_IFORM_MOV_CR_CR_GPR64:
   case XED_IFORM_MOV_CR_CR_GPR32:
-    *opcode = (addr_t)&V3_OPCODE_MOV2CR;
-    break;
+    return V3_OP_MOV2CR;
 
   case XED_IFORM_SMSW_GPRv:
-    *opcode = (addr_t)&V3_OPCODE_SMSW;
-    break;
+    return V3_OP_SMSW;
 
   case XED_IFORM_LMSW_GPR16:
-    *opcode = (addr_t)&V3_OPCODE_LMSW;
-    break;
+    return V3_OP_LMSW;
 
   case XED_IFORM_CLTS:
-    *opcode = (addr_t)&V3_OPCODE_CLTS;
-    break;
+    return V3_OP_CLTS;
+
+  case XED_IFORM_INVLPG_MEMb:
+    return V3_OP_INVLPG;
+
+
+    /* Data Instructions */
+
+  case XED_IFORM_ADC_MEMv_GPRv:
+  case XED_IFORM_ADC_MEMv_IMM:
+  case XED_IFORM_ADC_MEMb_GPR8:
+  case XED_IFORM_ADC_MEMb_IMM:
+    return V3_OP_ADC;
+
+  case XED_IFORM_ADD_MEMv_GPRv:
+  case XED_IFORM_ADD_MEMb_IMM:
+  case XED_IFORM_ADD_MEMb_GPR8:
+  case XED_IFORM_ADD_MEMv_IMM:
+    return V3_OP_ADD;
+
+  case XED_IFORM_AND_MEMv_IMM:
+  case XED_IFORM_AND_MEMb_GPR8:
+  case XED_IFORM_AND_MEMv_GPRv:
+  case XED_IFORM_AND_MEMb_IMM:
+    return V3_OP_AND;
+
+  case XED_IFORM_SUB_MEMv_IMM:
+  case XED_IFORM_SUB_MEMb_GPR8:
+  case XED_IFORM_SUB_MEMb_IMM:
+  case XED_IFORM_SUB_MEMv_GPRv:
+    return V3_OP_SUB;
+
+  case XED_IFORM_MOV_MEMv_GPRv:
+  case XED_IFORM_MOV_MEMb_GPR8:
+  case XED_IFORM_MOV_MEMb_AL:
+  case XED_IFORM_MOV_MEMv_IMM:
+  case XED_IFORM_MOV_MEMb_IMM:
+    return V3_OP_MOV;
+
+  case XED_IFORM_DEC_MEMv:
+  case XED_IFORM_DEC_MEMb:
+    return V3_OP_DEC;
 
+  case XED_IFORM_INC_MEMb:
+  case XED_IFORM_INC_MEMv:
+    return V3_OP_INC;
 
+  case XED_IFORM_OR_MEMv_IMM:
+  case XED_IFORM_OR_MEMb_IMM:
+  case XED_IFORM_OR_MEMv_GPRv:
+  case XED_IFORM_OR_MEMb_GPR8:
+    return V3_OP_OR;
+
+  case XED_IFORM_XOR_MEMv_GPRv:
+  case XED_IFORM_XOR_MEMb_IMM:
+  case XED_IFORM_XOR_MEMb_GPR8:
+  case XED_IFORM_XOR_MEMv_IMM:
+    return V3_OP_XOR;
+
+  case XED_IFORM_NEG_MEMb:
+  case XED_IFORM_NEG_MEMv:
+    return V3_OP_NEG;
+
+  case XED_IFORM_NOT_MEMv:
+  case XED_IFORM_NOT_MEMb:
+    return V3_OP_NOT;
+
+  case XED_IFORM_XCHG_MEMv_GPRv:
+  case XED_IFORM_XCHG_MEMb_GPR8:
+    return V3_OP_XCHG;
+
+  case XED_IFORM_SETB_MEMb:
+    return V3_OP_SETB;
+
+  case XED_IFORM_SETBE_MEMb:
+    return V3_OP_SETBE;
+
+  case XED_IFORM_SETL_MEMb:
+    return V3_OP_SETL;
+
+  case XED_IFORM_SETLE_MEMb:
+    return V3_OP_SETLE;
+
+  case XED_IFORM_SETNB_MEMb:
+    return V3_OP_SETNB;
+
+  case XED_IFORM_SETNBE_MEMb:
+    return V3_OP_SETNBE;
+
+  case XED_IFORM_SETNL_MEMb:
+    return V3_OP_SETNL;
+
+  case XED_IFORM_SETNLE_MEMb:
+    return V3_OP_SETNLE;
+
+  case XED_IFORM_SETNO_MEMb:
+    return V3_OP_SETNO;
+    
+  case XED_IFORM_SETNP_MEMb:
+    return V3_OP_SETNP;
+
+  case XED_IFORM_SETNS_MEMb:
+    return V3_OP_SETNS;
+
+  case XED_IFORM_SETNZ_MEMb:
+    return V3_OP_SETNZ;
+
+  case XED_IFORM_SETO_MEMb:
+    return V3_OP_SETO;
+    
+  case XED_IFORM_SETP_MEMb:
+    return V3_OP_SETP;
+
+  case XED_IFORM_SETS_MEMb:
+    return V3_OP_SETS;
+
+  case XED_IFORM_SETZ_MEMb:
+    return V3_OP_SETZ;
+
+
+  case XED_IFORM_MOVSB:
+  case XED_IFORM_MOVSW:
+  case XED_IFORM_MOVSD:
+  case XED_IFORM_MOVSQ:
+    return V3_OP_MOVS;
 
   default:
-    *opcode = 0;
-    return -1;
+    return V3_INVALID_OP;
   }
-
-  return 0;
 }
diff --git a/palacios/src/test/instr_emulator.c b/palacios/src/test/instr_emulator.c
new file mode 100644 (file)
index 0000000..e69de29