1 /* 2 * Copyright (C) 2016 Veertu Inc, 3 * Copyright (C) 2017 Google Inc, 4 * Based on Veertu vddh/vmm/vmx.h 5 * 6 * Interfaces to Hypervisor.framework to read/write X86 registers and VMCS. 7 * 8 * This program is free software; you can redistribute it and/or 9 * modify it under the terms of the GNU Lesser General Public 10 * License as published by the Free Software Foundation; either 11 * version 2 of the License, or (at your option) any later version. 12 * 13 * This program is distributed in the hope that it will be useful, 14 * but WITHOUT ANY WARRANTY; without even the implied warranty of 15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 16 * Lesser General Public License for more details. 17 * 18 * You should have received a copy of the GNU Lesser General Public 19 * License along with this program; if not, see <http://www.gnu.org/licenses/>. 20 * 21 * This file contain code under public domain from the hvdos project: 22 * https://github.com/mist64/hvdos 23 */ 24 25 #ifndef VMX_H 26 #define VMX_H 27 28 #include <Hypervisor/hv.h> 29 #include <Hypervisor/hv_vmx.h> 30 #include "vmcs.h" 31 #include "cpu.h" 32 #include "x86.h" 33 34 #include "exec/address-spaces.h" 35 36 static inline uint64_t rreg(hv_vcpuid_t vcpu, hv_x86_reg_t reg) 37 { 38 uint64_t v; 39 40 if (hv_vcpu_read_register(vcpu, reg, &v)) { 41 abort(); 42 } 43 44 return v; 45 } 46 47 /* write GPR */ 48 static inline void wreg(hv_vcpuid_t vcpu, hv_x86_reg_t reg, uint64_t v) 49 { 50 if (hv_vcpu_write_register(vcpu, reg, v)) { 51 abort(); 52 } 53 } 54 55 /* read VMCS field */ 56 static inline uint64_t rvmcs(hv_vcpuid_t vcpu, uint32_t field) 57 { 58 uint64_t v; 59 60 hv_vmx_vcpu_read_vmcs(vcpu, field, &v); 61 62 return v; 63 } 64 65 /* write VMCS field */ 66 static inline void wvmcs(hv_vcpuid_t vcpu, uint32_t field, uint64_t v) 67 { 68 hv_vmx_vcpu_write_vmcs(vcpu, field, v); 69 } 70 71 /* desired control word constrained by hardware/hypervisor capabilities */ 72 static inline uint64_t cap2ctrl(uint64_t cap, uint64_t ctrl) 73 { 74 return (ctrl | (cap & 0xffffffff)) & (cap >> 32); 75 } 76 77 #define VM_ENTRY_GUEST_LMA (1LL << 9) 78 79 #define AR_TYPE_ACCESSES_MASK 1 80 #define AR_TYPE_READABLE_MASK (1 << 1) 81 #define AR_TYPE_WRITEABLE_MASK (1 << 2) 82 #define AR_TYPE_CODE_MASK (1 << 3) 83 #define AR_TYPE_MASK 0x0f 84 #define AR_TYPE_BUSY_64_TSS 11 85 #define AR_TYPE_BUSY_32_TSS 11 86 #define AR_TYPE_BUSY_16_TSS 3 87 #define AR_TYPE_LDT 2 88 89 static void enter_long_mode(hv_vcpuid_t vcpu, uint64_t cr0, uint64_t efer) 90 { 91 uint64_t entry_ctls; 92 93 efer |= MSR_EFER_LMA; 94 wvmcs(vcpu, VMCS_GUEST_IA32_EFER, efer); 95 entry_ctls = rvmcs(vcpu, VMCS_ENTRY_CTLS); 96 wvmcs(vcpu, VMCS_ENTRY_CTLS, rvmcs(vcpu, VMCS_ENTRY_CTLS) | 97 VM_ENTRY_GUEST_LMA); 98 99 uint64_t guest_tr_ar = rvmcs(vcpu, VMCS_GUEST_TR_ACCESS_RIGHTS); 100 if ((efer & MSR_EFER_LME) && 101 (guest_tr_ar & AR_TYPE_MASK) != AR_TYPE_BUSY_64_TSS) { 102 wvmcs(vcpu, VMCS_GUEST_TR_ACCESS_RIGHTS, 103 (guest_tr_ar & ~AR_TYPE_MASK) | AR_TYPE_BUSY_64_TSS); 104 } 105 } 106 107 static void exit_long_mode(hv_vcpuid_t vcpu, uint64_t cr0, uint64_t efer) 108 { 109 uint64_t entry_ctls; 110 111 entry_ctls = rvmcs(vcpu, VMCS_ENTRY_CTLS); 112 wvmcs(vcpu, VMCS_ENTRY_CTLS, entry_ctls & ~VM_ENTRY_GUEST_LMA); 113 114 efer &= ~MSR_EFER_LMA; 115 wvmcs(vcpu, VMCS_GUEST_IA32_EFER, efer); 116 } 117 118 static inline void macvm_set_cr0(hv_vcpuid_t vcpu, uint64_t cr0) 119 { 120 int i; 121 uint64_t pdpte[4] = {0, 0, 0, 0}; 122 uint64_t efer = rvmcs(vcpu, VMCS_GUEST_IA32_EFER); 123 uint64_t old_cr0 = rvmcs(vcpu, VMCS_GUEST_CR0); 124 uint64_t changed_cr0 = old_cr0 ^ cr0; 125 uint64_t mask = CR0_PG | CR0_CD | CR0_NW | CR0_NE | CR0_ET; 126 uint64_t entry_ctls; 127 128 if ((cr0 & CR0_PG) && (rvmcs(vcpu, VMCS_GUEST_CR4) & CR4_PAE) && 129 !(efer & MSR_EFER_LME)) { 130 address_space_read(&address_space_memory, 131 rvmcs(vcpu, VMCS_GUEST_CR3) & ~0x1f, 132 MEMTXATTRS_UNSPECIFIED, pdpte, 32); 133 /* Only set PDPTE when appropriate. */ 134 for (i = 0; i < 4; i++) { 135 wvmcs(vcpu, VMCS_GUEST_PDPTE0 + i * 2, pdpte[i]); 136 } 137 } 138 139 wvmcs(vcpu, VMCS_CR0_MASK, mask); 140 wvmcs(vcpu, VMCS_CR0_SHADOW, cr0); 141 142 if (efer & MSR_EFER_LME) { 143 if (changed_cr0 & CR0_PG) { 144 if (cr0 & CR0_PG) { 145 enter_long_mode(vcpu, cr0, efer); 146 } else { 147 exit_long_mode(vcpu, cr0, efer); 148 } 149 } 150 } else { 151 entry_ctls = rvmcs(vcpu, VMCS_ENTRY_CTLS); 152 wvmcs(vcpu, VMCS_ENTRY_CTLS, entry_ctls & ~VM_ENTRY_GUEST_LMA); 153 } 154 155 /* Filter new CR0 after we are finished examining it above. */ 156 cr0 = (cr0 & ~(mask & ~CR0_PG)); 157 wvmcs(vcpu, VMCS_GUEST_CR0, cr0 | CR0_NE | CR0_ET); 158 159 hv_vcpu_invalidate_tlb(vcpu); 160 hv_vcpu_flush(vcpu); 161 } 162 163 static inline void macvm_set_cr4(hv_vcpuid_t vcpu, uint64_t cr4) 164 { 165 uint64_t guest_cr4 = cr4 | CR4_VMXE; 166 167 wvmcs(vcpu, VMCS_GUEST_CR4, guest_cr4); 168 wvmcs(vcpu, VMCS_CR4_SHADOW, cr4); 169 170 hv_vcpu_invalidate_tlb(vcpu); 171 hv_vcpu_flush(vcpu); 172 } 173 174 static inline void macvm_set_rip(CPUState *cpu, uint64_t rip) 175 { 176 X86CPU *x86_cpu = X86_CPU(cpu); 177 CPUX86State *env = &x86_cpu->env; 178 uint64_t val; 179 180 /* BUG, should take considering overlap.. */ 181 wreg(cpu->hvf_fd, HV_X86_RIP, rip); 182 env->eip = rip; 183 184 /* after moving forward in rip, we need to clean INTERRUPTABILITY */ 185 val = rvmcs(cpu->hvf_fd, VMCS_GUEST_INTERRUPTIBILITY); 186 if (val & (VMCS_INTERRUPTIBILITY_STI_BLOCKING | 187 VMCS_INTERRUPTIBILITY_MOVSS_BLOCKING)) { 188 env->hflags &= ~HF_INHIBIT_IRQ_MASK; 189 wvmcs(cpu->hvf_fd, VMCS_GUEST_INTERRUPTIBILITY, 190 val & ~(VMCS_INTERRUPTIBILITY_STI_BLOCKING | 191 VMCS_INTERRUPTIBILITY_MOVSS_BLOCKING)); 192 } 193 } 194 195 static inline void vmx_clear_nmi_blocking(CPUState *cpu) 196 { 197 X86CPU *x86_cpu = X86_CPU(cpu); 198 CPUX86State *env = &x86_cpu->env; 199 200 env->hflags2 &= ~HF2_NMI_MASK; 201 uint32_t gi = (uint32_t) rvmcs(cpu->hvf_fd, VMCS_GUEST_INTERRUPTIBILITY); 202 gi &= ~VMCS_INTERRUPTIBILITY_NMI_BLOCKING; 203 wvmcs(cpu->hvf_fd, VMCS_GUEST_INTERRUPTIBILITY, gi); 204 } 205 206 static inline void vmx_set_nmi_blocking(CPUState *cpu) 207 { 208 X86CPU *x86_cpu = X86_CPU(cpu); 209 CPUX86State *env = &x86_cpu->env; 210 211 env->hflags2 |= HF2_NMI_MASK; 212 uint32_t gi = (uint32_t)rvmcs(cpu->hvf_fd, VMCS_GUEST_INTERRUPTIBILITY); 213 gi |= VMCS_INTERRUPTIBILITY_NMI_BLOCKING; 214 wvmcs(cpu->hvf_fd, VMCS_GUEST_INTERRUPTIBILITY, gi); 215 } 216 217 static inline void vmx_set_nmi_window_exiting(CPUState *cpu) 218 { 219 uint64_t val; 220 val = rvmcs(cpu->hvf_fd, VMCS_PRI_PROC_BASED_CTLS); 221 wvmcs(cpu->hvf_fd, VMCS_PRI_PROC_BASED_CTLS, val | 222 VMCS_PRI_PROC_BASED_CTLS_NMI_WINDOW_EXITING); 223 224 } 225 226 static inline void vmx_clear_nmi_window_exiting(CPUState *cpu) 227 { 228 229 uint64_t val; 230 val = rvmcs(cpu->hvf_fd, VMCS_PRI_PROC_BASED_CTLS); 231 wvmcs(cpu->hvf_fd, VMCS_PRI_PROC_BASED_CTLS, val & 232 ~VMCS_PRI_PROC_BASED_CTLS_NMI_WINDOW_EXITING); 233 } 234 235 #endif 236