/* * 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 * Copyright (c) 2008, The V3VEE Project * All rights reserved. * * Author: Jack Lange * * This is free software. You are permitted to use, * redistribute, and modify it as specified in the file "V3VEE_LICENSE". */ #include #include #include #include #include #include /* Segmentation is a problem here... * * When we get a memory operand, presumably we use the default segment (which is?) * unless an alternate segment was specfied in the prefix... */ #ifndef DEBUG_CTRL_REGS #undef PrintDebug #define PrintDebug(fmt, args...) #endif static int handle_lmsw(struct guest_info * info, struct x86_instr * dec_instr); static int handle_clts(struct guest_info * info, struct x86_instr * dec_instr); static int handle_mov_to_cr0(struct guest_info * info, struct x86_instr * dec_instr); static int handle_mov_to_cr0_32(struct guest_info * info, struct x86_instr * dec_instr); static int handle_mov_to_cr0_32pae(struct guest_info * info, struct x86_instr * dec_instr); static int handle_mov_to_cr0_64(struct guest_info * info, struct x86_instr * dec_instr); static int handle_mov_to_cr0_64compat(struct guest_info * info, struct x86_instr * dec_instr); static int handle_mov_to_cr3_32(struct guest_info * info, struct x86_instr * dec_instr); static int handle_mov_to_cr3_32pae(struct guest_info * info, struct x86_instr * dec_instr); static int handle_mov_to_cr3_64(struct guest_info * info, struct x86_instr * dec_instr); static int handle_mov_to_cr3_64compat(struct guest_info * info, struct x86_instr * dec_instr); // First Attempt = 494 lines // current = 106 lines int v3_handle_cr0_write(struct guest_info * info) { uchar_t instr[15]; int ret; struct x86_instr dec_instr; 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); } /* The IFetch will already have faulted in the necessary bytes for the full instruction if (ret != 15) { // I think we should inject a GPF into the guest PrintError("Could not read instruction (ret=%d)\n", ret); return -1; } */ if (v3_decode(info, (addr_t)instr, &dec_instr) == -1) { PrintError("Could not decode instruction\n"); return -1; } 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) { 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) { if (handle_clts(info, &dec_instr) == -1) { return -1; } } else { PrintError("Unhandled opcode in handle_cr0_write\n"); return -1; } info->rip += dec_instr.instr_length; return 0; } static int handle_mov_to_cr0(struct guest_info * info, struct x86_instr * dec_instr) { PrintDebug("MOV2CR0\n"); switch (info->cpu_mode) { case REAL: case PROTECTED: return handle_mov_to_cr0_32(info, dec_instr); case PROTECTED_PAE: return handle_mov_to_cr0_32pae(info, dec_instr); case LONG: return handle_mov_to_cr0_64(info, dec_instr); case LONG_32_COMPAT: return handle_mov_to_cr0_64compat(info, dec_instr); default: PrintError("Invalid CPU Operating Mode: %d\n", info->cpu_mode); return -1; } } static int handle_mov_to_cr0_32pae(struct guest_info * info, struct x86_instr * dec_instr) { PrintError("32 bit PAE mov to CR0 not implemented\n"); return -1; } static int handle_mov_to_cr0_64(struct guest_info * info, struct x86_instr * dec_instr) { PrintError("64 bit mov to CR0 not implemented\n"); return -1; } static int handle_mov_to_cr0_64compat(struct guest_info * info, struct x86_instr * dec_instr) { PrintError("64 bit compatibility mode move to CR0 not implemented\n"); return -1; } static int handle_mov_to_cr0_32(struct guest_info * info, struct x86_instr * dec_instr) { // 32 bit registers struct cr0_32 *real_cr0 = (struct cr0_32*)&(info->ctrl_regs.cr0); struct cr0_32 *new_cr0= (struct cr0_32 *)(dec_instr->src_operand.operand); PrintDebug("OperandVal = %x, length=%d\n", *(uint_t *)new_cr0, dec_instr->src_operand.size); PrintDebug("Old CR0=%x\n", *(uint_t *)real_cr0); *real_cr0 = *new_cr0; if (info->shdw_pg_mode == SHADOW_PAGING) { struct cr0_32 * shadow_cr0 = (struct cr0_32 *)&(info->shdw_pg_state.guest_cr0); PrintDebug("Old Shadow CR0=%x\n", *(uint_t *)shadow_cr0); real_cr0->et = 1; *shadow_cr0 = *new_cr0; shadow_cr0->et = 1; if (v3_get_mem_mode(info) == VIRTUAL_MEM) { struct cr3_32 * shadow_cr3 = (struct cr3_32 *)&(info->shdw_pg_state.shadow_cr3); PrintDebug("Setting up Shadow Page Table\n"); info->ctrl_regs.cr3 = *(addr_t*)shadow_cr3; } else { info->ctrl_regs.cr3 = *(addr_t*)&(info->direct_map_pt); real_cr0->pg = 1; } PrintDebug("New Shadow CR0=%x\n",*(uint_t *)shadow_cr0); } PrintDebug("New CR0=%x\n", *(uint_t *)real_cr0); return 0; } static int handle_clts(struct guest_info * info, struct x86_instr * dec_instr) { // CLTS struct cr0_32 *real_cr0 = (struct cr0_32*)&(info->ctrl_regs.cr0); real_cr0->ts = 0; if (info->shdw_pg_mode == SHADOW_PAGING) { struct cr0_32 * shadow_cr0 = (struct cr0_32 *)&(info->shdw_pg_state.guest_cr0); shadow_cr0->ts = 0; } return 0; } static int handle_lmsw(struct guest_info * info, struct x86_instr * dec_instr) { struct cr0_real *real_cr0 = (struct cr0_real*)&(info->ctrl_regs.cr0); struct cr0_real *new_cr0 = (struct cr0_real *)(dec_instr->src_operand.operand); uchar_t new_cr0_val; PrintDebug("LMSW\n"); new_cr0_val = (*(char*)(new_cr0)) & 0x0f; PrintDebug("OperandVal = %x\n", new_cr0_val); PrintDebug("Old CR0=%x\n", *(uint_t *)real_cr0); *(uchar_t*)real_cr0 &= 0xf0; *(uchar_t*)real_cr0 |= new_cr0_val; PrintDebug("New CR0=%x\n", *(uint_t *)real_cr0); if (info->shdw_pg_mode == SHADOW_PAGING) { struct cr0_real * shadow_cr0 = (struct cr0_real*)&(info->shdw_pg_state.guest_cr0); PrintDebug(" Old Shadow CR0=%x\n", *(uint_t *)shadow_cr0); *(uchar_t*)shadow_cr0 &= 0xf0; *(uchar_t*)shadow_cr0 |= new_cr0_val; PrintDebug("New Shadow CR0=%x\n", *(uint_t *)shadow_cr0); } return 0; } // First attempt = 253 lines // current = 51 lines int v3_handle_cr0_read(struct guest_info * info) { uchar_t instr[15]; int ret; struct x86_instr dec_instr; 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); } /* The IFetch will already have faulted in the necessary bytes for the full instruction if (ret != 15) { // I think we should inject a GPF into the guest PrintError("Could not read instruction (ret=%d)\n", ret); return -1; } */ if (v3_decode(info, (addr_t)instr, &dec_instr) == -1) { PrintError("Could not decode instruction\n"); return -1; } if (v3_opcode_cmp(V3_OPCODE_MOVCR2, (const uchar_t *)(dec_instr.opcode)) == 0) { struct cr0_32 * virt_cr0 = (struct cr0_32 *)(dec_instr.dst_operand.operand); struct cr0_32 * real_cr0 = (struct cr0_32 *)&(info->ctrl_regs.cr0); PrintDebug("MOVCR2\n"); PrintDebug("CR0 at 0x%p\n", (void *)real_cr0); if (info->shdw_pg_mode == SHADOW_PAGING) { *virt_cr0 = *(struct cr0_32 *)&(info->shdw_pg_state.guest_cr0); } else { *virt_cr0 = *real_cr0; } PrintDebug("real CR0: %x\n", *(uint_t*)real_cr0); PrintDebug("returned CR0: %x\n", *(uint_t*)virt_cr0); } else if (v3_opcode_cmp(V3_OPCODE_SMSW, (const uchar_t *)(dec_instr.opcode)) == 0) { struct cr0_real *real_cr0= (struct cr0_real*)&(info->ctrl_regs.cr0); struct cr0_real *virt_cr0 = (struct cr0_real *)(dec_instr.dst_operand.operand); char cr0_val = *(char*)real_cr0 & 0x0f; PrintDebug("SMSW\n"); PrintDebug("CR0 at 0x%p\n", real_cr0); *(char *)virt_cr0 &= 0xf0; *(char *)virt_cr0 |= cr0_val; } else { PrintError("Unhandled opcode in handle_cr0_read\n"); return -1; } info->rip += dec_instr.instr_length; return 0; } // First Attempt = 256 lines // current = 65 lines int v3_handle_cr3_write(struct guest_info * info) { int ret; uchar_t instr[15]; struct x86_instr dec_instr; 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 (v3_decode(info, (addr_t)instr, &dec_instr) == -1) { PrintError("Could not decode instruction\n"); return -1; } if (v3_opcode_cmp(V3_OPCODE_MOV2CR, (const uchar_t *)(dec_instr.opcode)) == 0) { PrintDebug("MOV2CR3\n"); if (info->mem_mode == PHYSICAL_MEM) { // All we do is update the guest CR3 if (info->cpu_mode == LONG) { struct cr3_64 * new_cr3 = (struct cr3_64 *)(dec_instr.src_operand.operand); struct cr3_64 * guest_cr3 = (struct cr3_64 *)&(info->shdw_pg_state.guest_cr3); *guest_cr3 = *new_cr3; } else { struct cr3_32 * new_cr3 = (struct cr3_32 *)(dec_instr.src_operand.operand); struct cr3_32 * guest_cr3 = (struct cr3_32 *)&(info->shdw_pg_state.guest_cr3); *guest_cr3 = *new_cr3; } } else { switch (info->cpu_mode) { case PROTECTED: if (handle_mov_to_cr3_32(info, &dec_instr) == -1) { return -1; } case PROTECTED_PAE: if (handle_mov_to_cr3_32pae(info, &dec_instr) == -1) { return -1; } case LONG: if (handle_mov_to_cr3_64(info, &dec_instr) == -1) { return -1; } case LONG_32_COMPAT: if (handle_mov_to_cr3_64compat(info, &dec_instr) == -1) { return -1; } default: PrintError("Unhandled CPU mode: %d\n", info->cpu_mode); return -1; } } } else { PrintError("Unhandled opcode in handle_cr3_write\n"); return -1; } info->rip += dec_instr.instr_length; return 0; } static int handle_mov_to_cr3_32(struct guest_info * info, struct x86_instr * dec_instr) { PrintDebug("CR3 at 0x%p\n", &(info->ctrl_regs.cr3)); if (info->shdw_pg_mode == SHADOW_PAGING) { struct cr3_32 * new_cr3 = (struct cr3_32 *)(dec_instr->src_operand.operand); struct cr3_32 * guest_cr3 = (struct cr3_32 *)&(info->shdw_pg_state.guest_cr3); struct cr3_32 * shadow_cr3 = (struct cr3_32 *)&(info->shdw_pg_state.shadow_cr3); int cached = 0; PrintDebug("Old Shadow CR3=%x; Old Guest CR3=%x\n", *(uint_t*)shadow_cr3, *(uint_t*)guest_cr3); cached = v3_cache_page_tables32(info, (addr_t)V3_PAddr((void *)(addr_t)CR3_TO_PDE32((void *)*(addr_t *)new_cr3))); if (cached == -1) { PrintError("CR3 Cache failed\n"); return -1; } else if (cached == 0) { addr_t shadow_pt; if(info->mem_mode == VIRTUAL_MEM) { PrintDebug("New CR3 is different - flushing shadow page table %p\n", shadow_cr3 ); delete_page_tables_32((pde32_t *)CR3_TO_PDE32(*(uint_t*)shadow_cr3)); } shadow_pt = v3_create_new_shadow_pt(); shadow_cr3->pdt_base_addr = (addr_t)V3_PAddr((void *)(addr_t)PD32_BASE_ADDR(shadow_pt)); PrintDebug( "Created new shadow page table %p\n", (void *)(addr_t)shadow_cr3->pdt_base_addr ); //PrintDebugPageTables( (pde32_t *)CR3_TO_PDE32(*(uint_t*)shadow_cr3) ); } else { PrintDebug("Reusing cached shadow Page table\n"); } shadow_cr3->pwt = new_cr3->pwt; shadow_cr3->pcd = new_cr3->pcd; // What the hell... *guest_cr3 = *new_cr3; PrintDebug("New Shadow CR3=%x; New Guest CR3=%x\n", *(uint_t*)shadow_cr3, *(uint_t*)guest_cr3); if (info->mem_mode == VIRTUAL_MEM) { // If we aren't in paged mode then we have to preserve the identity mapped CR3 info->ctrl_regs.cr3 = *(addr_t*)shadow_cr3; } } return 0; } static int handle_mov_to_cr3_32pae(struct guest_info * info, struct x86_instr * dec_instr) { PrintError("32 Bit PAE mode Mov to CR3 not implemented\n"); return -1; } static int handle_mov_to_cr3_64(struct guest_info * info, struct x86_instr * dec_instr) { PrintError("Long mode Mov to CR3 not implemented\n"); return -1; } static int handle_mov_to_cr3_64compat(struct guest_info * info, struct x86_instr * dec_instr) { PrintError("Long compatiblity mode move to CR3 not implemented\n"); return -1; } // first attempt = 156 lines // current = 36 lines int v3_handle_cr3_read(struct guest_info * info) { uchar_t instr[15]; int ret; struct x86_instr dec_instr; 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); } /* The IFetch will already have faulted in the necessary bytes for the full instruction if (ret != 15) { // I think we should inject a GPF into the guest PrintError("Could not read instruction (ret=%d)\n", ret); return -1; } */ if (v3_decode(info, (addr_t)instr, &dec_instr) == -1) { PrintError("Could not decode instruction\n"); return -1; } if (v3_opcode_cmp(V3_OPCODE_MOVCR2, (const uchar_t *)(dec_instr.opcode)) == 0) { PrintDebug("MOVCR32\n"); struct cr3_32 * virt_cr3 = (struct cr3_32 *)(dec_instr.dst_operand.operand); PrintDebug("CR3 at 0x%p\n", &(info->ctrl_regs.cr3)); if (info->shdw_pg_mode == SHADOW_PAGING) { *virt_cr3 = *(struct cr3_32 *)&(info->shdw_pg_state.guest_cr3); } else { *virt_cr3 = *(struct cr3_32 *)&(info->ctrl_regs.cr3); } } else { PrintError("Unhandled opcode in handle_cr3_read\n"); return -1; } info->rip += dec_instr.instr_length; return 0; } int v3_handle_cr4_read(struct guest_info * info) { PrintError("CR4 Read not handled\n"); return -1; } int v3_handle_cr4_write(struct guest_info * info) { uchar_t instr[15]; int ret; struct x86_instr dec_instr; 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 (v3_decode(info, (addr_t)instr, &dec_instr) == -1) { PrintError("Could not decode instruction\n"); return -1; } if (v3_opcode_cmp(V3_OPCODE_MOV2CR, (const uchar_t *)(dec_instr.opcode)) != 0) { PrintError("Invalid opcode in write to CR4\n"); return -1; } if ((info->cpu_mode == PROTECTED) || (info->cpu_mode == PROTECTED_PAE)) { struct cr4_32 * new_cr4 = (struct cr4_32 *)(dec_instr.src_operand.operand); struct cr4_32 * old_cr4 = (struct cr4_32 *)&(info->ctrl_regs.cr4); PrintDebug("OperandVal = %x, length = %d\n", *(uint_t *)new_cr4, dec_instr.src_operand.size); PrintDebug("Old CR4=%x\n", *(uint_t *)old_cr4); if ((info->shdw_pg_mode == SHADOW_PAGING) && (v3_get_mem_mode(info) == PHYSICAL_MEM)) { if ((old_cr4->pae == 0) && (new_cr4->pae == 1)) { PrintDebug("Creating PAE passthrough tables\n"); // Delete the old 32 bit direct map page tables delete_page_tables_32((pde32_t *)V3_VAddr((void *)(info->direct_map_pt))); // create 32 bit PAE direct map page table info->direct_map_pt = (addr_t)V3_PAddr(create_passthrough_pts_32PAE(info)); // reset cr3 to new page tables info->ctrl_regs.cr3 = *(addr_t*)&(info->direct_map_pt); } else if ((old_cr4->pae == 1) && (new_cr4->pae == 0)) { // Create passthrough standard 32bit pagetables return -1; } } *old_cr4 = *new_cr4; PrintDebug("New CR4=%x\n", *(uint_t *)old_cr4); } else { return -1; } info->rip += dec_instr.instr_length; return 0; } int v3_handle_efer_read(uint_t msr, struct v3_msr * dst, void * priv_data) { PrintError("EFER Read not handled\n"); return -1; } int v3_handle_efer_write(uint_t msr, struct v3_msr src, void * priv_data) { struct guest_info * info = (struct guest_info *)(priv_data); struct efer_64 * new_efer = (struct efer_64 *)&(src.value); struct efer_64 * old_efer = (struct efer_64 *)&(info->ctrl_regs.efer); PrintDebug("Old EFER=%p\n", (void *)*(addr_t*)(old_efer)); if ((info->shdw_pg_mode == SHADOW_PAGING) && (v3_get_mem_mode(info) == PHYSICAL_MEM)) { if ((old_efer->lme == 0) && (new_efer->lme == 1)) { PrintDebug("Transition to longmode\n"); PrintDebug("Creating Passthrough 64 bit page tables\n"); // Delete the old 32 bit direct map page tables /* * JRL BUG? * Will these page tables always be in PAE format?? */ PrintDebug("Deleting old PAE Page tables\n"); PrintError("JRL BUG?: Will the old page tables always be in PAE format??\n"); delete_page_tables_32PAE((pdpe32pae_t *)V3_VAddr((void *)(info->direct_map_pt))); // create 64 bit direct map page table info->direct_map_pt = (addr_t)V3_PAddr(create_passthrough_pts_64(info)); // reset cr3 to new page tables info->ctrl_regs.cr3 = *(addr_t*)&(info->direct_map_pt); // Does this mean we will have to fully virtualize a shadow EFER?? (yes it does) ((struct efer_64 *)&(info->guest_efer.value))->lme = 1; new_efer->lma = 1; } else if ((old_efer->lme == 1) && (new_efer->lme == 0)) { // transition out of long mode //((struct efer_64 *)&(info->guest_efer.value))->lme = 0; //((struct efer_64 *)&(info->guest_efer.value))->lma = 0; return -1; } *old_efer = *new_efer; PrintDebug("New EFER=%p\n", (void *)*(addr_t *)(old_efer)); } else { return -1; } info->rip += 2; // WRMSR/RDMSR are two byte operands return 0; }