1 /* 2 * QEMU KVM support 3 * 4 * Copyright IBM, Corp. 2008 5 * 6 * Authors: 7 * Anthony Liguori <aliguori@us.ibm.com> 8 * 9 * This work is licensed under the terms of the GNU GPL, version 2 or later. 10 * See the COPYING file in the top-level directory. 11 * 12 */ 13 14 #ifndef QEMU_KVM_H 15 #define QEMU_KVM_H 16 17 #include <errno.h> 18 #include "config-host.h" 19 #include "qemu/queue.h" 20 #include "qom/cpu.h" 21 22 #ifdef CONFIG_KVM 23 #include <linux/kvm.h> 24 #include <linux/kvm_para.h> 25 #else 26 /* These constants must never be used at runtime if kvm_enabled() is false. 27 * They exist so we don't need #ifdefs around KVM-specific code that already 28 * checks kvm_enabled() properly. 29 */ 30 #define KVM_CPUID_SIGNATURE 0 31 #define KVM_CPUID_FEATURES 0 32 #define KVM_FEATURE_CLOCKSOURCE 0 33 #define KVM_FEATURE_NOP_IO_DELAY 0 34 #define KVM_FEATURE_MMU_OP 0 35 #define KVM_FEATURE_CLOCKSOURCE2 0 36 #define KVM_FEATURE_ASYNC_PF 0 37 #define KVM_FEATURE_STEAL_TIME 0 38 #define KVM_FEATURE_PV_EOI 0 39 #define KVM_FEATURE_CLOCKSOURCE_STABLE_BIT 0 40 #endif 41 42 extern bool kvm_allowed; 43 extern bool kvm_kernel_irqchip; 44 extern bool kvm_async_interrupts_allowed; 45 extern bool kvm_halt_in_kernel_allowed; 46 extern bool kvm_irqfds_allowed; 47 extern bool kvm_msi_via_irqfd_allowed; 48 extern bool kvm_gsi_routing_allowed; 49 extern bool kvm_gsi_direct_mapping; 50 extern bool kvm_readonly_mem_allowed; 51 52 #if defined CONFIG_KVM || !defined NEED_CPU_H 53 #define kvm_enabled() (kvm_allowed) 54 /** 55 * kvm_irqchip_in_kernel: 56 * 57 * Returns: true if the user asked us to create an in-kernel 58 * irqchip via the "kernel_irqchip=on" machine option. 59 * What this actually means is architecture and machine model 60 * specific: on PC, for instance, it means that the LAPIC, 61 * IOAPIC and PIT are all in kernel. This function should never 62 * be used from generic target-independent code: use one of the 63 * following functions or some other specific check instead. 64 */ 65 #define kvm_irqchip_in_kernel() (kvm_kernel_irqchip) 66 67 /** 68 * kvm_async_interrupts_enabled: 69 * 70 * Returns: true if we can deliver interrupts to KVM 71 * asynchronously (ie by ioctl from any thread at any time) 72 * rather than having to do interrupt delivery synchronously 73 * (where the vcpu must be stopped at a suitable point first). 74 */ 75 #define kvm_async_interrupts_enabled() (kvm_async_interrupts_allowed) 76 77 /** 78 * kvm_halt_in_kernel 79 * 80 * Returns: true if halted cpus should still get a KVM_RUN ioctl to run 81 * inside of kernel space. This only works if MP state is implemented. 82 */ 83 #define kvm_halt_in_kernel() (kvm_halt_in_kernel_allowed) 84 85 /** 86 * kvm_irqfds_enabled: 87 * 88 * Returns: true if we can use irqfds to inject interrupts into 89 * a KVM CPU (ie the kernel supports irqfds and we are running 90 * with a configuration where it is meaningful to use them). 91 */ 92 #define kvm_irqfds_enabled() (kvm_irqfds_allowed) 93 94 /** 95 * kvm_msi_via_irqfd_enabled: 96 * 97 * Returns: true if we can route a PCI MSI (Message Signaled Interrupt) 98 * to a KVM CPU via an irqfd. This requires that the kernel supports 99 * this and that we're running in a configuration that permits it. 100 */ 101 #define kvm_msi_via_irqfd_enabled() (kvm_msi_via_irqfd_allowed) 102 103 /** 104 * kvm_gsi_routing_enabled: 105 * 106 * Returns: true if GSI routing is enabled (ie the kernel supports 107 * it and we're running in a configuration that permits it). 108 */ 109 #define kvm_gsi_routing_enabled() (kvm_gsi_routing_allowed) 110 111 /** 112 * kvm_gsi_direct_mapping: 113 * 114 * Returns: true if GSI direct mapping is enabled. 115 */ 116 #define kvm_gsi_direct_mapping() (kvm_gsi_direct_mapping) 117 118 /** 119 * kvm_readonly_mem_enabled: 120 * 121 * Returns: true if KVM readonly memory is enabled (ie the kernel 122 * supports it and we're running in a configuration that permits it). 123 */ 124 #define kvm_readonly_mem_enabled() (kvm_readonly_mem_allowed) 125 126 #else 127 #define kvm_enabled() (0) 128 #define kvm_irqchip_in_kernel() (false) 129 #define kvm_async_interrupts_enabled() (false) 130 #define kvm_halt_in_kernel() (false) 131 #define kvm_irqfds_enabled() (false) 132 #define kvm_msi_via_irqfd_enabled() (false) 133 #define kvm_gsi_routing_allowed() (false) 134 #define kvm_gsi_direct_mapping() (false) 135 #define kvm_readonly_mem_enabled() (false) 136 #endif 137 138 struct kvm_run; 139 struct kvm_lapic_state; 140 141 typedef struct KVMCapabilityInfo { 142 const char *name; 143 int value; 144 } KVMCapabilityInfo; 145 146 #define KVM_CAP_INFO(CAP) { "KVM_CAP_" stringify(CAP), KVM_CAP_##CAP } 147 #define KVM_CAP_LAST_INFO { NULL, 0 } 148 149 struct KVMState; 150 typedef struct KVMState KVMState; 151 extern KVMState *kvm_state; 152 153 /* external API */ 154 155 int kvm_init(QEMUMachine *machine); 156 157 int kvm_has_sync_mmu(void); 158 int kvm_has_vcpu_events(void); 159 int kvm_has_robust_singlestep(void); 160 int kvm_has_debugregs(void); 161 int kvm_has_xsave(void); 162 int kvm_has_xcrs(void); 163 int kvm_has_pit_state2(void); 164 int kvm_has_many_ioeventfds(void); 165 int kvm_has_gsi_routing(void); 166 int kvm_has_intx_set_mask(void); 167 168 int kvm_init_vcpu(CPUState *cpu); 169 int kvm_cpu_exec(CPUState *cpu); 170 171 #ifdef NEED_CPU_H 172 173 void kvm_setup_guest_memory(void *start, size_t size); 174 void kvm_flush_coalesced_mmio_buffer(void); 175 176 int kvm_insert_breakpoint(CPUState *cpu, target_ulong addr, 177 target_ulong len, int type); 178 int kvm_remove_breakpoint(CPUState *cpu, target_ulong addr, 179 target_ulong len, int type); 180 void kvm_remove_all_breakpoints(CPUState *cpu); 181 int kvm_update_guest_debug(CPUState *cpu, unsigned long reinject_trap); 182 #ifndef _WIN32 183 int kvm_set_signal_mask(CPUState *cpu, const sigset_t *sigset); 184 #endif 185 186 int kvm_on_sigbus_vcpu(CPUState *cpu, int code, void *addr); 187 int kvm_on_sigbus(int code, void *addr); 188 189 /* internal API */ 190 191 int kvm_ioctl(KVMState *s, int type, ...); 192 193 int kvm_vm_ioctl(KVMState *s, int type, ...); 194 195 int kvm_vcpu_ioctl(CPUState *cpu, int type, ...); 196 197 /** 198 * kvm_device_ioctl - call an ioctl on a kvm device 199 * @fd: The KVM device file descriptor as returned from KVM_CREATE_DEVICE 200 * @type: The device-ctrl ioctl number 201 * 202 * Returns: -errno on error, nonnegative on success 203 */ 204 int kvm_device_ioctl(int fd, int type, ...); 205 206 /** 207 * kvm_create_device - create a KVM device for the device control API 208 * @KVMState: The KVMState pointer 209 * @type: The KVM device type (see Documentation/virtual/kvm/devices in the 210 * kernel source) 211 * @test: If true, only test if device can be created, but don't actually 212 * create the device. 213 * 214 * Returns: -errno on error, nonnegative on success: @test ? 0 : device fd; 215 */ 216 int kvm_create_device(KVMState *s, uint64_t type, bool test); 217 218 219 /* Arch specific hooks */ 220 221 extern const KVMCapabilityInfo kvm_arch_required_capabilities[]; 222 223 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run); 224 void kvm_arch_post_run(CPUState *cpu, struct kvm_run *run); 225 226 int kvm_arch_handle_exit(CPUState *cpu, struct kvm_run *run); 227 228 int kvm_arch_process_async_events(CPUState *cpu); 229 230 int kvm_arch_get_registers(CPUState *cpu); 231 232 /* state subset only touched by the VCPU itself during runtime */ 233 #define KVM_PUT_RUNTIME_STATE 1 234 /* state subset modified during VCPU reset */ 235 #define KVM_PUT_RESET_STATE 2 236 /* full state set, modified during initialization or on vmload */ 237 #define KVM_PUT_FULL_STATE 3 238 239 int kvm_arch_put_registers(CPUState *cpu, int level); 240 241 int kvm_arch_init(KVMState *s); 242 243 int kvm_arch_init_vcpu(CPUState *cpu); 244 245 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */ 246 unsigned long kvm_arch_vcpu_id(CPUState *cpu); 247 248 void kvm_arch_reset_vcpu(CPUState *cpu); 249 250 int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr); 251 int kvm_arch_on_sigbus(int code, void *addr); 252 253 void kvm_arch_init_irq_routing(KVMState *s); 254 255 int kvm_set_irq(KVMState *s, int irq, int level); 256 int kvm_irqchip_send_msi(KVMState *s, MSIMessage msg); 257 258 void kvm_irqchip_add_irq_route(KVMState *s, int gsi, int irqchip, int pin); 259 void kvm_irqchip_commit_routes(KVMState *s); 260 261 void kvm_put_apic_state(DeviceState *d, struct kvm_lapic_state *kapic); 262 void kvm_get_apic_state(DeviceState *d, struct kvm_lapic_state *kapic); 263 264 struct kvm_guest_debug; 265 struct kvm_debug_exit_arch; 266 267 struct kvm_sw_breakpoint { 268 target_ulong pc; 269 target_ulong saved_insn; 270 int use_count; 271 QTAILQ_ENTRY(kvm_sw_breakpoint) entry; 272 }; 273 274 QTAILQ_HEAD(kvm_sw_breakpoint_head, kvm_sw_breakpoint); 275 276 struct kvm_sw_breakpoint *kvm_find_sw_breakpoint(CPUState *cpu, 277 target_ulong pc); 278 279 int kvm_sw_breakpoints_active(CPUState *cpu); 280 281 int kvm_arch_insert_sw_breakpoint(CPUState *cpu, 282 struct kvm_sw_breakpoint *bp); 283 int kvm_arch_remove_sw_breakpoint(CPUState *cpu, 284 struct kvm_sw_breakpoint *bp); 285 int kvm_arch_insert_hw_breakpoint(target_ulong addr, 286 target_ulong len, int type); 287 int kvm_arch_remove_hw_breakpoint(target_ulong addr, 288 target_ulong len, int type); 289 void kvm_arch_remove_all_hw_breakpoints(void); 290 291 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg); 292 293 bool kvm_arch_stop_on_emulation_error(CPUState *cpu); 294 295 int kvm_check_extension(KVMState *s, unsigned int extension); 296 297 #define kvm_vm_enable_cap(s, capability, cap_flags, ...) \ 298 ({ \ 299 struct kvm_enable_cap cap = { \ 300 .cap = capability, \ 301 .flags = cap_flags, \ 302 }; \ 303 uint64_t args_tmp[] = { __VA_ARGS__ }; \ 304 int i; \ 305 for (i = 0; i < ARRAY_SIZE(args_tmp) && \ 306 i < ARRAY_SIZE(cap.args); i++) { \ 307 cap.args[i] = args_tmp[i]; \ 308 } \ 309 kvm_vm_ioctl(s, KVM_ENABLE_CAP, &cap); \ 310 }) 311 312 #define kvm_vcpu_enable_cap(cpu, capability, cap_flags, ...) \ 313 ({ \ 314 struct kvm_enable_cap cap = { \ 315 .cap = capability, \ 316 .flags = cap_flags, \ 317 }; \ 318 uint64_t args_tmp[] = { __VA_ARGS__ }; \ 319 int i; \ 320 for (i = 0; i < ARRAY_SIZE(args_tmp) && \ 321 i < ARRAY_SIZE(cap.args); i++) { \ 322 cap.args[i] = args_tmp[i]; \ 323 } \ 324 kvm_vcpu_ioctl(cpu, KVM_ENABLE_CAP, &cap); \ 325 }) 326 327 uint32_t kvm_arch_get_supported_cpuid(KVMState *env, uint32_t function, 328 uint32_t index, int reg); 329 330 #if !defined(CONFIG_USER_ONLY) 331 int kvm_physical_memory_addr_from_host(KVMState *s, void *ram_addr, 332 hwaddr *phys_addr); 333 #endif 334 335 #endif /* NEED_CPU_H */ 336 337 void kvm_cpu_synchronize_state(CPUState *cpu); 338 void kvm_cpu_synchronize_post_reset(CPUState *cpu); 339 void kvm_cpu_synchronize_post_init(CPUState *cpu); 340 341 /* generic hooks - to be moved/refactored once there are more users */ 342 343 static inline void cpu_synchronize_state(CPUState *cpu) 344 { 345 if (kvm_enabled()) { 346 kvm_cpu_synchronize_state(cpu); 347 } 348 } 349 350 static inline void cpu_synchronize_post_reset(CPUState *cpu) 351 { 352 if (kvm_enabled()) { 353 kvm_cpu_synchronize_post_reset(cpu); 354 } 355 } 356 357 static inline void cpu_synchronize_post_init(CPUState *cpu) 358 { 359 if (kvm_enabled()) { 360 kvm_cpu_synchronize_post_init(cpu); 361 } 362 } 363 364 int kvm_irqchip_add_msi_route(KVMState *s, MSIMessage msg); 365 int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg); 366 void kvm_irqchip_release_virq(KVMState *s, int virq); 367 368 int kvm_irqchip_add_irqfd_notifier(KVMState *s, EventNotifier *n, 369 EventNotifier *rn, int virq); 370 int kvm_irqchip_remove_irqfd_notifier(KVMState *s, EventNotifier *n, int virq); 371 void kvm_pc_gsi_handler(void *opaque, int n, int level); 372 void kvm_pc_setup_irq_routing(bool pci_enabled); 373 void kvm_init_irq_routing(KVMState *s); 374 375 /** 376 * kvm_arch_irqchip_create: 377 * @KVMState: The KVMState pointer 378 * 379 * Allow architectures to create an in-kernel irq chip themselves. 380 * 381 * Returns: < 0: error 382 * 0: irq chip was not created 383 * > 0: irq chip was created 384 */ 385 int kvm_arch_irqchip_create(KVMState *s); 386 #endif 387