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 2.1 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 #include "sysemu/hvf.h" 34 #include "sysemu/hvf_int.h" 35 36 #include "exec/address-spaces.h" 37 38 static inline uint64_t rreg(hv_vcpuid_t vcpu, hv_x86_reg_t reg) 39 { 40 uint64_t v; 41 42 if (hv_vcpu_read_register(vcpu, reg, &v)) { 43 abort(); 44 } 45 46 return v; 47 } 48 49 /* write GPR */ 50 static inline void wreg(hv_vcpuid_t vcpu, hv_x86_reg_t reg, uint64_t v) 51 { 52 if (hv_vcpu_write_register(vcpu, reg, v)) { 53 abort(); 54 } 55 } 56 57 /* read VMCS field */ 58 static inline uint64_t rvmcs(hv_vcpuid_t vcpu, uint32_t field) 59 { 60 uint64_t v; 61 62 hv_vmx_vcpu_read_vmcs(vcpu, field, &v); 63 64 return v; 65 } 66 67 /* write VMCS field */ 68 static inline void wvmcs(hv_vcpuid_t vcpu, uint32_t field, uint64_t v) 69 { 70 hv_vmx_vcpu_write_vmcs(vcpu, field, v); 71 } 72 73 /* desired control word constrained by hardware/hypervisor capabilities */ 74 static inline uint64_t cap2ctrl(uint64_t cap, uint64_t ctrl) 75 { 76 return (ctrl | (cap & 0xffffffff)) & (cap >> 32); 77 } 78 79 #define VM_ENTRY_GUEST_LMA (1LL << 9) 80 81 #define AR_TYPE_ACCESSES_MASK 1 82 #define AR_TYPE_READABLE_MASK (1 << 1) 83 #define AR_TYPE_WRITABLE_MASK (1 << 2) 84 #define AR_TYPE_CODE_MASK (1 << 3) 85 #define AR_TYPE_MASK 0x0f 86 #define AR_TYPE_BUSY_64_TSS 11 87 #define AR_TYPE_BUSY_32_TSS 11 88 #define AR_TYPE_BUSY_16_TSS 3 89 #define AR_TYPE_LDT 2 90 91 static void enter_long_mode(hv_vcpuid_t vcpu, uint64_t cr0, uint64_t efer) 92 { 93 uint64_t entry_ctls; 94 95 efer |= MSR_EFER_LMA; 96 wvmcs(vcpu, VMCS_GUEST_IA32_EFER, efer); 97 entry_ctls = rvmcs(vcpu, VMCS_ENTRY_CTLS); 98 wvmcs(vcpu, VMCS_ENTRY_CTLS, rvmcs(vcpu, VMCS_ENTRY_CTLS) | 99 VM_ENTRY_GUEST_LMA); 100 101 uint64_t guest_tr_ar = rvmcs(vcpu, VMCS_GUEST_TR_ACCESS_RIGHTS); 102 if ((efer & MSR_EFER_LME) && 103 (guest_tr_ar & AR_TYPE_MASK) != AR_TYPE_BUSY_64_TSS) { 104 wvmcs(vcpu, VMCS_GUEST_TR_ACCESS_RIGHTS, 105 (guest_tr_ar & ~AR_TYPE_MASK) | AR_TYPE_BUSY_64_TSS); 106 } 107 } 108 109 static void exit_long_mode(hv_vcpuid_t vcpu, uint64_t cr0, uint64_t efer) 110 { 111 uint64_t entry_ctls; 112 113 entry_ctls = rvmcs(vcpu, VMCS_ENTRY_CTLS); 114 wvmcs(vcpu, VMCS_ENTRY_CTLS, entry_ctls & ~VM_ENTRY_GUEST_LMA); 115 116 efer &= ~MSR_EFER_LMA; 117 wvmcs(vcpu, VMCS_GUEST_IA32_EFER, efer); 118 } 119 120 static inline void macvm_set_cr0(hv_vcpuid_t vcpu, uint64_t cr0) 121 { 122 int i; 123 uint64_t pdpte[4] = {0, 0, 0, 0}; 124 uint64_t efer = rvmcs(vcpu, VMCS_GUEST_IA32_EFER); 125 uint64_t old_cr0 = rvmcs(vcpu, VMCS_GUEST_CR0); 126 uint64_t changed_cr0 = old_cr0 ^ cr0; 127 uint64_t mask = CR0_PG_MASK | CR0_CD_MASK | CR0_NW_MASK | 128 CR0_NE_MASK | CR0_ET_MASK; 129 uint64_t entry_ctls; 130 131 if ((cr0 & CR0_PG_MASK) && (rvmcs(vcpu, VMCS_GUEST_CR4) & CR4_PAE_MASK) && 132 !(efer & MSR_EFER_LME)) { 133 address_space_read(&address_space_memory, 134 rvmcs(vcpu, VMCS_GUEST_CR3) & ~0x1f, 135 MEMTXATTRS_UNSPECIFIED, pdpte, 32); 136 /* Only set PDPTE when appropriate. */ 137 for (i = 0; i < 4; i++) { 138 wvmcs(vcpu, VMCS_GUEST_PDPTE0 + i * 2, pdpte[i]); 139 } 140 } 141 142 wvmcs(vcpu, VMCS_CR0_MASK, mask); 143 wvmcs(vcpu, VMCS_CR0_SHADOW, cr0); 144 145 if (efer & MSR_EFER_LME) { 146 if (changed_cr0 & CR0_PG_MASK) { 147 if (cr0 & CR0_PG_MASK) { 148 enter_long_mode(vcpu, cr0, efer); 149 } else { 150 exit_long_mode(vcpu, cr0, efer); 151 } 152 } 153 } else { 154 entry_ctls = rvmcs(vcpu, VMCS_ENTRY_CTLS); 155 wvmcs(vcpu, VMCS_ENTRY_CTLS, entry_ctls & ~VM_ENTRY_GUEST_LMA); 156 } 157 158 /* Filter new CR0 after we are finished examining it above. */ 159 cr0 = (cr0 & ~(mask & ~CR0_PG_MASK)); 160 wvmcs(vcpu, VMCS_GUEST_CR0, cr0 | CR0_NE_MASK | CR0_ET_MASK); 161 162 hv_vcpu_invalidate_tlb(vcpu); 163 } 164 165 static inline void macvm_set_cr4(hv_vcpuid_t vcpu, uint64_t cr4) 166 { 167 uint64_t guest_cr4 = cr4 | CR4_VMXE_MASK; 168 169 wvmcs(vcpu, VMCS_GUEST_CR4, guest_cr4); 170 wvmcs(vcpu, VMCS_CR4_SHADOW, cr4); 171 wvmcs(vcpu, VMCS_CR4_MASK, CR4_VMXE_MASK); 172 173 hv_vcpu_invalidate_tlb(vcpu); 174 } 175 176 static inline void macvm_set_rip(CPUState *cpu, uint64_t rip) 177 { 178 X86CPU *x86_cpu = X86_CPU(cpu); 179 CPUX86State *env = &x86_cpu->env; 180 uint64_t val; 181 182 /* BUG, should take considering overlap.. */ 183 wreg(cpu->hvf->fd, HV_X86_RIP, rip); 184 env->eip = rip; 185 186 /* after moving forward in rip, we need to clean INTERRUPTABILITY */ 187 val = rvmcs(cpu->hvf->fd, VMCS_GUEST_INTERRUPTIBILITY); 188 if (val & (VMCS_INTERRUPTIBILITY_STI_BLOCKING | 189 VMCS_INTERRUPTIBILITY_MOVSS_BLOCKING)) { 190 env->hflags &= ~HF_INHIBIT_IRQ_MASK; 191 wvmcs(cpu->hvf->fd, VMCS_GUEST_INTERRUPTIBILITY, 192 val & ~(VMCS_INTERRUPTIBILITY_STI_BLOCKING | 193 VMCS_INTERRUPTIBILITY_MOVSS_BLOCKING)); 194 } 195 } 196 197 static inline void vmx_clear_nmi_blocking(CPUState *cpu) 198 { 199 X86CPU *x86_cpu = X86_CPU(cpu); 200 CPUX86State *env = &x86_cpu->env; 201 202 env->hflags2 &= ~HF2_NMI_MASK; 203 uint32_t gi = (uint32_t) rvmcs(cpu->hvf->fd, VMCS_GUEST_INTERRUPTIBILITY); 204 gi &= ~VMCS_INTERRUPTIBILITY_NMI_BLOCKING; 205 wvmcs(cpu->hvf->fd, VMCS_GUEST_INTERRUPTIBILITY, gi); 206 } 207 208 static inline void vmx_set_nmi_blocking(CPUState *cpu) 209 { 210 X86CPU *x86_cpu = X86_CPU(cpu); 211 CPUX86State *env = &x86_cpu->env; 212 213 env->hflags2 |= HF2_NMI_MASK; 214 uint32_t gi = (uint32_t)rvmcs(cpu->hvf->fd, VMCS_GUEST_INTERRUPTIBILITY); 215 gi |= VMCS_INTERRUPTIBILITY_NMI_BLOCKING; 216 wvmcs(cpu->hvf->fd, VMCS_GUEST_INTERRUPTIBILITY, gi); 217 } 218 219 static inline void vmx_set_nmi_window_exiting(CPUState *cpu) 220 { 221 uint64_t val; 222 val = rvmcs(cpu->hvf->fd, VMCS_PRI_PROC_BASED_CTLS); 223 wvmcs(cpu->hvf->fd, VMCS_PRI_PROC_BASED_CTLS, val | 224 VMCS_PRI_PROC_BASED_CTLS_NMI_WINDOW_EXITING); 225 226 } 227 228 static inline void vmx_clear_nmi_window_exiting(CPUState *cpu) 229 { 230 231 uint64_t val; 232 val = rvmcs(cpu->hvf->fd, VMCS_PRI_PROC_BASED_CTLS); 233 wvmcs(cpu->hvf->fd, VMCS_PRI_PROC_BASED_CTLS, val & 234 ~VMCS_PRI_PROC_BASED_CTLS_NMI_WINDOW_EXITING); 235 } 236 237 #endif 238