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.


Previous software interrupt support was added to Xed, but I wasn't using the kernel
[palacios.git] / palacios / src / palacios / vmm_v3dec.c
index 47a2487..343be38 100644 (file)
 #include <palacios/vmm_decoder.h>
 #include <palacios/vmm_instr_decoder.h>
 
-
-/* Disgusting mask hack...
-   I can't think right now, so we'll do it this way...
-*/
-static const ullong_t mask_1 = 0x00000000000000ffLL;
-static const ullong_t mask_2 = 0x000000000000ffffLL;
-static const ullong_t mask_4 = 0x00000000ffffffffLL;
-static const ullong_t mask_8 = 0xffffffffffffffffLL;
-
-
-#define MASK(val, length) ({                   \
-           ullong_t mask = 0x0LL;              \
-           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;                         \
-       })
+#ifndef V3_CONFIG_DEBUG_DECODER
+#undef PrintDebug
+#define PrintDebug(fmt, args...)
+#endif
+
+
+#define MASK(val, length) ({                                           \
+            uint64_t mask = 0x0LL;                                     \
+            switch (length) {                                          \
+               case 1:                                                 \
+                   mask = 0x00000000000000ffLL;                        \
+                    break;                                             \
+                case 2:                                                        \
+                    mask = 0x000000000000ffffLL;                       \
+                    break;                                             \
+                case 4:                                                        \
+                    mask = 0x00000000ffffffffLL;                       \
+                    break;                                             \
+                case 8:                                                        \
+                    mask = 0xffffffffffffffffLL;                       \
+                    break;                                             \
+            }                                                          \
+            val & mask;                                                        \
+        })
 
 static v3_op_type_t op_form_to_type(op_form_t form);
 static int parse_operands(struct guest_info * core, uint8_t * instr_ptr, struct x86_instr * instr, op_form_t form);
 
 
 int v3_disasm(struct guest_info * info, void *instr_ptr, addr_t * rip, int mark) {
-    return 0;
+    return -1;
 }
 
 
@@ -69,7 +65,7 @@ int v3_deinit_decoder(struct guest_info * core) {
 
 
 int v3_encode(struct guest_info * info, struct x86_instr * instr, uint8_t * instr_buf) {
-    return 0;
+    return -1;
 }
 
 
@@ -78,17 +74,36 @@ int v3_decode(struct guest_info * core, addr_t instr_ptr, struct x86_instr * ins
     int ret = 0;
     int length = 0;
 
+
+    PrintDebug("Decoding Instruction at %p\n", (void *)instr_ptr);
+
     memset(instr, 0, sizeof(struct x86_instr));
 
     // scan for prefixes
     length = v3_get_prefixes((uint8_t *)instr_ptr, &(instr->prefixes));
 
 
-    // check for REX prefix
+    // REX prefix
+    if (v3_get_vm_cpu_mode(core) == LONG) {
+       uint8_t prefix = *(uint8_t *)(instr_ptr + length);
+
+       if ((prefix & 0xf0) == 0x40) {
+           instr->prefixes.rex = 1;
+
+           instr->prefixes.rex_rm = (prefix & 0x01);
+           instr->prefixes.rex_sib_idx = ((prefix & 0x02) >> 1);
+           instr->prefixes.rex_reg = ((prefix & 0x04) >> 2);
+           instr->prefixes.rex_op_size = ((prefix & 0x08) >> 3);
+
+           length += 1;
+       }
+    }
 
 
     form = op_code_to_form((uint8_t *)(instr_ptr + length), &length);
 
+    PrintDebug("\t decoded as (%s)\n", op_form_to_str(form));
+
     if (form == INVALID_INSTR) {
        PrintError("Could not find instruction form (%x)\n", *(uint32_t *)(instr_ptr + length));
        return -1;
@@ -103,10 +118,14 @@ int v3_decode(struct guest_info * core, addr_t instr_ptr, struct x86_instr * ins
        return -1;
     }
     length += ret;
-    
 
     instr->instr_length += length;
 
+#ifdef V3_CONFIG_DEBUG_DECODER
+    V3_Print("Decoding Instr at %p\n", (void *)core->rip);
+    v3_print_instr(instr);
+    V3_Print("CS DB FLag=%x\n", core->segments.cs.db);
+#endif
 
     return 0;
 }
@@ -116,11 +135,13 @@ static int parse_operands(struct guest_info * core, uint8_t * instr_ptr,
                          struct x86_instr * instr, op_form_t form) {
     // get operational mode of the guest for operand width
     uint8_t operand_width = get_operand_width(core, instr, form);
-    uint8_t addr_width = get_addr_width(core, instr, form);;
+    uint8_t addr_width = get_addr_width(core, instr);
     int ret = 0;
     uint8_t * instr_start = instr_ptr;
     
 
+    PrintDebug("\tOperand width=%d, Addr width=%d\n", operand_width, addr_width);
+
     switch (form) {
        case ADC_IMM2_8:
        case ADD_IMM2_8:
@@ -134,13 +155,11 @@ static int parse_operands(struct guest_info * core, uint8_t * instr_ptr,
        case AND_IMM2:  
        case OR_IMM2:
        case SUB_IMM2:
-       case XOR_IMM2: 
-       case MOV_IMM2:{
+       case XOR_IMM2:
+       case MOV_IMM2: {
            uint8_t reg_code = 0;
 
-           instr->dst_operand.size = operand_width;
-
-           ret = decode_rm_operand(core, instr_ptr, instr, &(instr->dst_operand), &reg_code);
+           ret = decode_rm_operand(core, instr_ptr, form, instr, &(instr->dst_operand), &reg_code);
 
            if (ret == -1) {
                PrintError("Error decoding operand\n");
@@ -152,17 +171,23 @@ static int parse_operands(struct guest_info * core, uint8_t * instr_ptr,
            instr->src_operand.type = IMM_OPERAND;
            instr->src_operand.size = operand_width;
 
+
            if (operand_width == 1) {
                instr->src_operand.operand = *(uint8_t *)instr_ptr;
            } else if (operand_width == 2) {
                instr->src_operand.operand = *(uint16_t *)instr_ptr;
            } else if (operand_width == 4) {
                instr->src_operand.operand = *(uint32_t *)instr_ptr;
+           } else if (operand_width == 8) {
+               instr->src_operand.operand = *(sint32_t *)instr_ptr; // This is a special case for sign extended 64bit ops
            } else {
                PrintError("Illegal operand width (%d)\n", operand_width);
                return -1;
            }
 
+           instr->src_operand.read = 1;
+           instr->dst_operand.write = 1;
+
            instr_ptr += operand_width;
 
            instr->num_operands = 2;
@@ -185,9 +210,7 @@ static int parse_operands(struct guest_info * core, uint8_t * instr_ptr,
        case MOV_2MEM: {
            uint8_t reg_code = 0;
 
-           instr->dst_operand.size = operand_width;
-
-           ret = decode_rm_operand(core, instr_ptr, instr, &(instr->dst_operand), &reg_code);
+           ret = decode_rm_operand(core, instr_ptr, form, instr, &(instr->dst_operand), &reg_code);
 
            if (ret == -1) {
                PrintError("Error decoding operand\n");
@@ -199,7 +222,11 @@ static int parse_operands(struct guest_info * core, uint8_t * instr_ptr,
            instr->src_operand.type = REG_OPERAND;
            instr->src_operand.size = operand_width;
 
-           decode_gpr(&(core->vm_regs), reg_code, &(instr->src_operand));
+           instr->src_operand.read = 1;
+           instr->dst_operand.write = 1;
+
+
+           decode_gpr(core, reg_code, &(instr->src_operand));
 
            instr->num_operands = 2;
            break;
@@ -219,9 +246,8 @@ static int parse_operands(struct guest_info * core, uint8_t * instr_ptr,
        case XOR_MEM2:
        case MOV_MEM2: {
            uint8_t reg_code = 0;
-           instr->src_operand.size = operand_width;
 
-           ret = decode_rm_operand(core, instr_ptr, instr, &(instr->src_operand), &reg_code);
+           ret = decode_rm_operand(core, instr_ptr, form, instr, &(instr->src_operand), &reg_code);
 
            if (ret == -1) {
                PrintError("Error decoding operand\n");
@@ -232,7 +258,60 @@ static int parse_operands(struct guest_info * core, uint8_t * instr_ptr,
 
            instr->dst_operand.size = operand_width;
            instr->dst_operand.type = REG_OPERAND;
-           decode_gpr(&(core->vm_regs), reg_code, &(instr->dst_operand));
+           decode_gpr(core, reg_code, &(instr->dst_operand));
+
+           instr->src_operand.read = 1;
+           instr->dst_operand.write = 1;
+
+           instr->num_operands = 2;
+
+           break;
+       }
+       case MOVSX_8:
+       case MOVZX_8: {
+           uint8_t reg_code = 0;
+
+           ret = decode_rm_operand(core, instr_ptr, form, instr, &(instr->src_operand), &reg_code);
+           instr->src_operand.size = 1;
+
+           if (ret == -1) {
+               PrintError("Error decoding operand\n");
+               return -1;
+           }
+
+           instr_ptr += ret;
+
+           instr->dst_operand.size = operand_width;
+           instr->dst_operand.type = REG_OPERAND;
+           decode_gpr(core, reg_code, &(instr->dst_operand));
+
+           instr->src_operand.read = 1;
+           instr->dst_operand.write = 1;
+
+           instr->num_operands = 2;
+
+           break;
+       }
+       case MOVSX:
+       case MOVZX: {
+           uint8_t reg_code = 0;
+
+           ret = decode_rm_operand(core, instr_ptr, form, instr, &(instr->src_operand), &reg_code);
+           instr->src_operand.size = 2;
+
+           if (ret == -1) {
+               PrintError("Error decoding operand\n");
+               return -1;
+           }
+
+           instr_ptr += ret;
+
+           instr->dst_operand.size = operand_width;
+           instr->dst_operand.type = REG_OPERAND;
+           decode_gpr(core, reg_code, &(instr->dst_operand));
+
+           instr->src_operand.read = 1;
+           instr->dst_operand.write = 1;
 
            instr->num_operands = 2;
 
@@ -245,9 +324,8 @@ static int parse_operands(struct guest_info * core, uint8_t * instr_ptr,
        case SUB_IMM2SX_8:
        case XOR_IMM2SX_8: {
            uint8_t reg_code = 0;
-           instr->src_operand.size = operand_width;
 
-           ret = decode_rm_operand(core, instr_ptr, instr, &(instr->src_operand), &reg_code);
+           ret = decode_rm_operand(core, instr_ptr, form, instr, &(instr->dst_operand), &reg_code);
 
            if (ret == -1) {
                PrintError("Error decoding operand\n");
@@ -258,7 +336,10 @@ static int parse_operands(struct guest_info * core, uint8_t * instr_ptr,
 
            instr->src_operand.type = IMM_OPERAND;
            instr->src_operand.size = operand_width;
-           instr->src_operand.operand = *(sint8_t *)instr_ptr;  // sign extend.
+           instr->src_operand.operand = (addr_t)MASK((sint64_t)*(sint8_t *)instr_ptr, operand_width);  // sign extend.
+
+           instr->src_operand.read = 1;
+           instr->dst_operand.write = 1;
 
            instr_ptr += 1;
 
@@ -271,28 +352,150 @@ static int parse_operands(struct guest_info * core, uint8_t * instr_ptr,
            instr->is_str_op = 1;
            
            if (instr->prefixes.rep == 1) {
-               instr->str_op_length = MASK(core->vm_regs.rcx, operand_width);
+               instr->str_op_length = MASK(core->vm_regs.rcx, addr_width);
            } else {
                instr->str_op_length = 1;
            }
 
            // Source: DS:(E)SI
-           // Source: ES:(E)DI
+           // Destination: ES:(E)DI
 
            instr->src_operand.type = MEM_OPERAND;
            instr->src_operand.size = operand_width;
-           instr->src_operand.operand = core->segments.ds.base + MASK(core->vm_regs.rsi, addr_width);
+           instr->src_operand.operand = get_addr_linear(core,  MASK(core->vm_regs.rsi, addr_width), &(core->segments.ds));
 
-           instr->src_operand.type = MEM_OPERAND;
+
+           instr->dst_operand.type = MEM_OPERAND;
+           instr->dst_operand.size = operand_width;
+           instr->dst_operand.operand = get_addr_linear(core, MASK(core->vm_regs.rdi, addr_width), &(core->segments.es));
+
+
+           instr->src_operand.read = 1;
+           instr->dst_operand.write = 1;
+
+           instr->num_operands = 2;
+
+           break;
+       }
+       case MOV_2CR: {
+           uint8_t reg_code = 0;
+           
+           ret = decode_rm_operand(core, instr_ptr, form, instr, &(instr->src_operand),
+                                   &reg_code);
+
+           if (ret == -1) {
+               PrintError("Error decoding operand for (%s)\n", op_form_to_str(form));
+               return -1;
+           }
+               
+           instr_ptr += ret;
+
+           instr->dst_operand.type = REG_OPERAND;
+           instr->dst_operand.size = operand_width;
+           decode_cr(core, reg_code, &(instr->dst_operand));
+           
+           instr->src_operand.read = 1;
+           instr->dst_operand.write = 1;
+
+           instr->num_operands = 2;
+           break;
+       }
+       case MOV_CR2: {
+           uint8_t reg_code = 0;
+           
+           ret = decode_rm_operand(core, instr_ptr, form, instr, &(instr->dst_operand),
+                                   &reg_code);
+           
+           if (ret == -1) {
+               PrintError("Error decoding operand for (%s)\n", op_form_to_str(form));
+               return -1;
+           }
+
+           instr_ptr += ret;
+           
+           instr->src_operand.type = REG_OPERAND;
            instr->src_operand.size = operand_width;
-           instr->src_operand.operand = core->segments.es.base + MASK(core->vm_regs.rdi, addr_width);
+           decode_cr(core, reg_code, &(instr->src_operand));
+
+           instr->src_operand.read = 1;
+           instr->dst_operand.write = 1;
 
            instr->num_operands = 2;
+           break;
+       }
+       case STOS:
+       case STOS_8: {
+           instr->is_str_op = 1;
+
+           if (instr->prefixes.rep == 1) {
+               instr->str_op_length = MASK(core->vm_regs.rcx, addr_width);
+           } else {
+               instr->str_op_length = 1;
+           }
+
+           instr->src_operand.size = operand_width;
+           instr->src_operand.type = REG_OPERAND;
+           instr->src_operand.operand = (addr_t)&(core->vm_regs.rax);
+
+           instr->dst_operand.type = MEM_OPERAND;
+           instr->dst_operand.size = operand_width;
+           instr->dst_operand.operand = get_addr_linear(core, MASK(core->vm_regs.rdi, addr_width), &(core->segments.es));
+
+           instr->src_operand.read = 1;
+           instr->dst_operand.write = 1;
+
+           instr->num_operands = 2;
+
+           break;
+       }
+       case INT: {
+           instr->dst_operand.type = IMM_OPERAND;
+           instr->dst_operand.size = operand_width;
+               instr->dst_operand.operand = *(uint8_t *)instr_ptr;
+           instr_ptr += operand_width;
+           instr->num_operands = 1;
 
            break;
        }
+       case INVLPG: {
+           uint8_t reg_code = 0;
+
+           ret = decode_rm_operand(core, instr_ptr, form, instr, &(instr->dst_operand), &reg_code);
+
+           if (ret == -1) {
+               PrintError("Error decoding operand for (%s)\n", op_form_to_str(form));
+               return -1;
+           }
+
+           instr_ptr += ret;
+
+           instr->num_operands = 1;
+           break;
+       }
+       case LMSW: 
+       case SMSW: {
+           uint8_t reg_code = 0;
+
+           ret = decode_rm_operand(core, instr_ptr, form, instr, &(instr->dst_operand), &reg_code);
+
+           if (ret == -1) {
+               PrintError("Error decoding operand for (%s)\n", op_form_to_str(form));
+               return -1;
+           }
+
+           instr_ptr += ret;
+
+           instr->dst_operand.read = 1;
+
+           instr->num_operands = 1;
+           break;
+       }
+       case CLTS: {
+           // no operands. 
+           break;
+       }
        default:
-           PrintError("Invalid Instruction form: %d\n", form);
+           PrintError("Invalid Instruction form: %s\n", op_form_to_str(form));
            return -1;
     }
 
@@ -311,12 +514,14 @@ static v3_op_type_t op_form_to_type(op_form_t form) {
        case INVLPG:
            return V3_OP_INVLPG;
 
+       case INT:
+           return V3_OP_INT;
+           
        case MOV_CR2:
            return V3_OP_MOVCR2;
        case MOV_2CR:
            return V3_OP_MOV2CR;
 
-
        case MOV_MEM2_8:
        case MOV_MEM2:
        case MOV_2MEM_8:
@@ -342,7 +547,6 @@ static v3_op_type_t op_form_to_type(op_form_t form) {
            return V3_OP_MOVZX;
 
 
-           
        case ADC_2MEM_8:
        case ADC_2MEM:
        case ADC_MEM2_8: