4 * Copyright IBM, Corp. 2008
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.
18 #include "config-host.h"
19 #include "qemu/queue.h"
21 #include "exec/memattrs.h"
25 #include <linux/kvm.h>
26 #include <linux/kvm_para.h>
28 /* These constants must never be used at runtime if kvm_enabled() is false.
29 * They exist so we don't need #ifdefs around KVM-specific code that already
30 * checks kvm_enabled() properly.
32 #define KVM_CPUID_SIGNATURE 0
33 #define KVM_CPUID_FEATURES 0
34 #define KVM_FEATURE_CLOCKSOURCE 0
35 #define KVM_FEATURE_NOP_IO_DELAY 0
36 #define KVM_FEATURE_MMU_OP 0
37 #define KVM_FEATURE_CLOCKSOURCE2 0
38 #define KVM_FEATURE_ASYNC_PF 0
39 #define KVM_FEATURE_STEAL_TIME 0
40 #define KVM_FEATURE_PV_EOI 0
41 #define KVM_FEATURE_CLOCKSOURCE_STABLE_BIT 0
44 extern bool kvm_allowed;
45 extern bool kvm_kernel_irqchip;
46 extern bool kvm_split_irqchip;
47 extern bool kvm_async_interrupts_allowed;
48 extern bool kvm_halt_in_kernel_allowed;
49 extern bool kvm_eventfds_allowed;
50 extern bool kvm_irqfds_allowed;
51 extern bool kvm_resamplefds_allowed;
52 extern bool kvm_msi_via_irqfd_allowed;
53 extern bool kvm_gsi_routing_allowed;
54 extern bool kvm_gsi_direct_mapping;
55 extern bool kvm_readonly_mem_allowed;
56 extern bool kvm_direct_msi_allowed;
57 extern bool kvm_ioeventfd_any_length_allowed;
59 #if defined CONFIG_KVM || !defined NEED_CPU_H
60 #define kvm_enabled() (kvm_allowed)
62 * kvm_irqchip_in_kernel:
64 * Returns: true if the user asked us to create an in-kernel
65 * irqchip via the "kernel_irqchip=on" machine option.
66 * What this actually means is architecture and machine model
67 * specific: on PC, for instance, it means that the LAPIC,
68 * IOAPIC and PIT are all in kernel. This function should never
69 * be used from generic target-independent code: use one of the
70 * following functions or some other specific check instead.
72 #define kvm_irqchip_in_kernel() (kvm_kernel_irqchip)
75 * kvm_irqchip_is_split:
77 * Returns: true if the user asked us to split the irqchip
78 * implementation between user and kernel space. The details are
79 * architecture and machine specific. On PC, it means that the PIC,
80 * IOAPIC, and PIT are in user space while the LAPIC is in the kernel.
82 #define kvm_irqchip_is_split() (kvm_split_irqchip)
85 * kvm_async_interrupts_enabled:
87 * Returns: true if we can deliver interrupts to KVM
88 * asynchronously (ie by ioctl from any thread at any time)
89 * rather than having to do interrupt delivery synchronously
90 * (where the vcpu must be stopped at a suitable point first).
92 #define kvm_async_interrupts_enabled() (kvm_async_interrupts_allowed)
97 * Returns: true if halted cpus should still get a KVM_RUN ioctl to run
98 * inside of kernel space. This only works if MP state is implemented.
100 #define kvm_halt_in_kernel() (kvm_halt_in_kernel_allowed)
103 * kvm_eventfds_enabled:
105 * Returns: true if we can use eventfds to receive notifications
106 * from a KVM CPU (ie the kernel supports eventds and we are running
107 * with a configuration where it is meaningful to use them).
109 #define kvm_eventfds_enabled() (kvm_eventfds_allowed)
112 * kvm_irqfds_enabled:
114 * Returns: true if we can use irqfds to inject interrupts into
115 * a KVM CPU (ie the kernel supports irqfds and we are running
116 * with a configuration where it is meaningful to use them).
118 #define kvm_irqfds_enabled() (kvm_irqfds_allowed)
121 * kvm_resamplefds_enabled:
123 * Returns: true if we can use resamplefds to inject interrupts into
124 * a KVM CPU (ie the kernel supports resamplefds and we are running
125 * with a configuration where it is meaningful to use them).
127 #define kvm_resamplefds_enabled() (kvm_resamplefds_allowed)
130 * kvm_msi_via_irqfd_enabled:
132 * Returns: true if we can route a PCI MSI (Message Signaled Interrupt)
133 * to a KVM CPU via an irqfd. This requires that the kernel supports
134 * this and that we're running in a configuration that permits it.
136 #define kvm_msi_via_irqfd_enabled() (kvm_msi_via_irqfd_allowed)
139 * kvm_gsi_routing_enabled:
141 * Returns: true if GSI routing is enabled (ie the kernel supports
142 * it and we're running in a configuration that permits it).
144 #define kvm_gsi_routing_enabled() (kvm_gsi_routing_allowed)
147 * kvm_gsi_direct_mapping:
149 * Returns: true if GSI direct mapping is enabled.
151 #define kvm_gsi_direct_mapping() (kvm_gsi_direct_mapping)
154 * kvm_readonly_mem_enabled:
156 * Returns: true if KVM readonly memory is enabled (ie the kernel
157 * supports it and we're running in a configuration that permits it).
159 #define kvm_readonly_mem_enabled() (kvm_readonly_mem_allowed)
162 * kvm_direct_msi_enabled:
164 * Returns: true if KVM allows direct MSI injection.
166 #define kvm_direct_msi_enabled() (kvm_direct_msi_allowed)
169 * kvm_ioeventfd_any_length_enabled:
170 * Returns: true if KVM allows any length io eventfd.
172 #define kvm_ioeventfd_any_length_enabled() (kvm_ioeventfd_any_length_allowed)
175 #define kvm_enabled() (0)
176 #define kvm_irqchip_in_kernel() (false)
177 #define kvm_async_interrupts_enabled() (false)
178 #define kvm_halt_in_kernel() (false)
179 #define kvm_eventfds_enabled() (false)
180 #define kvm_irqfds_enabled() (false)
181 #define kvm_resamplefds_enabled() (false)
182 #define kvm_msi_via_irqfd_enabled() (false)
183 #define kvm_gsi_routing_allowed() (false)
184 #define kvm_gsi_direct_mapping() (false)
185 #define kvm_readonly_mem_enabled() (false)
186 #define kvm_direct_msi_enabled() (false)
187 #define kvm_ioeventfd_any_length_enabled() (false)
191 struct kvm_lapic_state;
192 struct kvm_irq_routing_entry;
194 typedef struct KVMCapabilityInfo {
199 #define KVM_CAP_INFO(CAP) { "KVM_CAP_" stringify(CAP), KVM_CAP_##CAP }
200 #define KVM_CAP_LAST_INFO { NULL, 0 }
203 typedef struct KVMState KVMState;
204 extern KVMState *kvm_state;
208 bool kvm_has_free_slot(MachineState *ms);
209 int kvm_has_sync_mmu(void);
210 int kvm_has_vcpu_events(void);
211 int kvm_has_robust_singlestep(void);
212 int kvm_has_debugregs(void);
213 int kvm_has_pit_state2(void);
214 int kvm_has_many_ioeventfds(void);
215 int kvm_has_gsi_routing(void);
216 int kvm_has_intx_set_mask(void);
218 int kvm_init_vcpu(CPUState *cpu);
219 int kvm_cpu_exec(CPUState *cpu);
223 void kvm_setup_guest_memory(void *start, size_t size);
224 void kvm_flush_coalesced_mmio_buffer(void);
226 int kvm_insert_breakpoint(CPUState *cpu, target_ulong addr,
227 target_ulong len, int type);
228 int kvm_remove_breakpoint(CPUState *cpu, target_ulong addr,
229 target_ulong len, int type);
230 void kvm_remove_all_breakpoints(CPUState *cpu);
231 int kvm_update_guest_debug(CPUState *cpu, unsigned long reinject_trap);
233 int kvm_set_signal_mask(CPUState *cpu, const sigset_t *sigset);
236 int kvm_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
237 int kvm_on_sigbus(int code, void *addr);
239 /* interface with exec.c */
241 void phys_mem_set_alloc(void *(*alloc)(size_t, uint64_t *align));
245 int kvm_ioctl(KVMState *s, int type, ...);
247 int kvm_vm_ioctl(KVMState *s, int type, ...);
249 int kvm_vcpu_ioctl(CPUState *cpu, int type, ...);
252 * kvm_device_ioctl - call an ioctl on a kvm device
253 * @fd: The KVM device file descriptor as returned from KVM_CREATE_DEVICE
254 * @type: The device-ctrl ioctl number
256 * Returns: -errno on error, nonnegative on success
258 int kvm_device_ioctl(int fd, int type, ...);
261 * kvm_vm_check_attr - check for existence of a specific vm attribute
262 * @s: The KVMState pointer
264 * @attr: the attribute of that group to query for
266 * Returns: 1 if the attribute exists
267 * 0 if the attribute either does not exist or if the vm device
268 * interface is unavailable
270 int kvm_vm_check_attr(KVMState *s, uint32_t group, uint64_t attr);
273 * kvm_device_check_attr - check for existence of a specific device attribute
274 * @fd: The device file descriptor
276 * @attr: the attribute of that group to query for
278 * Returns: 1 if the attribute exists
279 * 0 if the attribute either does not exist or if the vm device
280 * interface is unavailable
282 int kvm_device_check_attr(int fd, uint32_t group, uint64_t attr);
285 * kvm_device_access - set or get value of a specific vm attribute
286 * @fd: The device file descriptor
288 * @attr: the attribute of that group to set or get
289 * @val: pointer to a storage area for the value
290 * @write: true for set and false for get operation
292 * This function is not allowed to fail. Use kvm_device_check_attr()
293 * in order to check for the availability of optional attributes.
295 void kvm_device_access(int fd, int group, uint64_t attr,
296 void *val, bool write);
299 * kvm_create_device - create a KVM device for the device control API
300 * @KVMState: The KVMState pointer
301 * @type: The KVM device type (see Documentation/virtual/kvm/devices in the
303 * @test: If true, only test if device can be created, but don't actually
306 * Returns: -errno on error, nonnegative on success: @test ? 0 : device fd;
308 int kvm_create_device(KVMState *s, uint64_t type, bool test);
311 /* Arch specific hooks */
313 extern const KVMCapabilityInfo kvm_arch_required_capabilities[];
315 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run);
316 MemTxAttrs kvm_arch_post_run(CPUState *cpu, struct kvm_run *run);
318 int kvm_arch_handle_exit(CPUState *cpu, struct kvm_run *run);
320 int kvm_arch_process_async_events(CPUState *cpu);
322 int kvm_arch_get_registers(CPUState *cpu);
324 /* state subset only touched by the VCPU itself during runtime */
325 #define KVM_PUT_RUNTIME_STATE 1
326 /* state subset modified during VCPU reset */
327 #define KVM_PUT_RESET_STATE 2
328 /* full state set, modified during initialization or on vmload */
329 #define KVM_PUT_FULL_STATE 3
331 int kvm_arch_put_registers(CPUState *cpu, int level);
333 int kvm_arch_init(MachineState *ms, KVMState *s);
335 int kvm_arch_init_vcpu(CPUState *cpu);
337 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */
338 unsigned long kvm_arch_vcpu_id(CPUState *cpu);
340 int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
341 int kvm_arch_on_sigbus(int code, void *addr);
343 void kvm_arch_init_irq_routing(KVMState *s);
345 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry *route,
346 uint64_t address, uint32_t data, PCIDevice *dev);
348 int kvm_arch_msi_data_to_gsi(uint32_t data);
350 int kvm_set_irq(KVMState *s, int irq, int level);
351 int kvm_irqchip_send_msi(KVMState *s, MSIMessage msg);
353 void kvm_irqchip_add_irq_route(KVMState *s, int gsi, int irqchip, int pin);
354 void kvm_irqchip_commit_routes(KVMState *s);
356 void kvm_put_apic_state(DeviceState *d, struct kvm_lapic_state *kapic);
357 void kvm_get_apic_state(DeviceState *d, struct kvm_lapic_state *kapic);
359 struct kvm_guest_debug;
360 struct kvm_debug_exit_arch;
362 struct kvm_sw_breakpoint {
364 target_ulong saved_insn;
366 QTAILQ_ENTRY(kvm_sw_breakpoint) entry;
369 QTAILQ_HEAD(kvm_sw_breakpoint_head, kvm_sw_breakpoint);
371 struct kvm_sw_breakpoint *kvm_find_sw_breakpoint(CPUState *cpu,
374 int kvm_sw_breakpoints_active(CPUState *cpu);
376 int kvm_arch_insert_sw_breakpoint(CPUState *cpu,
377 struct kvm_sw_breakpoint *bp);
378 int kvm_arch_remove_sw_breakpoint(CPUState *cpu,
379 struct kvm_sw_breakpoint *bp);
380 int kvm_arch_insert_hw_breakpoint(target_ulong addr,
381 target_ulong len, int type);
382 int kvm_arch_remove_hw_breakpoint(target_ulong addr,
383 target_ulong len, int type);
384 void kvm_arch_remove_all_hw_breakpoints(void);
386 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg);
388 bool kvm_arch_stop_on_emulation_error(CPUState *cpu);
390 int kvm_check_extension(KVMState *s, unsigned int extension);
392 int kvm_vm_check_extension(KVMState *s, unsigned int extension);
394 #define kvm_vm_enable_cap(s, capability, cap_flags, ...) \
396 struct kvm_enable_cap cap = { \
398 .flags = cap_flags, \
400 uint64_t args_tmp[] = { __VA_ARGS__ }; \
402 for (i = 0; i < (int)ARRAY_SIZE(args_tmp) && \
403 i < ARRAY_SIZE(cap.args); i++) { \
404 cap.args[i] = args_tmp[i]; \
406 kvm_vm_ioctl(s, KVM_ENABLE_CAP, &cap); \
409 #define kvm_vcpu_enable_cap(cpu, capability, cap_flags, ...) \
411 struct kvm_enable_cap cap = { \
413 .flags = cap_flags, \
415 uint64_t args_tmp[] = { __VA_ARGS__ }; \
417 for (i = 0; i < (int)ARRAY_SIZE(args_tmp) && \
418 i < ARRAY_SIZE(cap.args); i++) { \
419 cap.args[i] = args_tmp[i]; \
421 kvm_vcpu_ioctl(cpu, KVM_ENABLE_CAP, &cap); \
424 uint32_t kvm_arch_get_supported_cpuid(KVMState *env, uint32_t function,
425 uint32_t index, int reg);
427 void kvm_set_sigmask_len(KVMState *s, unsigned int sigmask_len);
429 #if !defined(CONFIG_USER_ONLY)
430 int kvm_physical_memory_addr_from_host(KVMState *s, void *ram_addr,
434 #endif /* NEED_CPU_H */
436 void kvm_cpu_synchronize_state(CPUState *cpu);
437 void kvm_cpu_synchronize_post_reset(CPUState *cpu);
438 void kvm_cpu_synchronize_post_init(CPUState *cpu);
440 /* generic hooks - to be moved/refactored once there are more users */
442 static inline void cpu_synchronize_state(CPUState *cpu)
445 kvm_cpu_synchronize_state(cpu);
449 static inline void cpu_synchronize_post_reset(CPUState *cpu)
452 kvm_cpu_synchronize_post_reset(cpu);
456 static inline void cpu_synchronize_post_init(CPUState *cpu)
459 kvm_cpu_synchronize_post_init(cpu);
463 int kvm_irqchip_add_msi_route(KVMState *s, MSIMessage msg, PCIDevice *dev);
464 int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg,
466 void kvm_irqchip_release_virq(KVMState *s, int virq);
468 int kvm_irqchip_add_adapter_route(KVMState *s, AdapterInfo *adapter);
469 int kvm_irqchip_add_hv_sint_route(KVMState *s, uint32_t vcpu, uint32_t sint);
471 int kvm_irqchip_add_irqfd_notifier_gsi(KVMState *s, EventNotifier *n,
472 EventNotifier *rn, int virq);
473 int kvm_irqchip_remove_irqfd_notifier_gsi(KVMState *s, EventNotifier *n,
475 int kvm_irqchip_add_irqfd_notifier(KVMState *s, EventNotifier *n,
476 EventNotifier *rn, qemu_irq irq);
477 int kvm_irqchip_remove_irqfd_notifier(KVMState *s, EventNotifier *n,
479 void kvm_irqchip_set_qemuirq_gsi(KVMState *s, qemu_irq irq, int gsi);
480 void kvm_pc_gsi_handler(void *opaque, int n, int level);
481 void kvm_pc_setup_irq_routing(bool pci_enabled);
482 void kvm_init_irq_routing(KVMState *s);
485 * kvm_arch_irqchip_create:
486 * @KVMState: The KVMState pointer
488 * Allow architectures to create an in-kernel irq chip themselves.
490 * Returns: < 0: error
491 * 0: irq chip was not created
492 * > 0: irq chip was created
494 int kvm_arch_irqchip_create(KVMState *s);
497 * kvm_set_one_reg - set a register value in KVM via KVM_SET_ONE_REG ioctl
498 * @id: The register ID
499 * @source: The pointer to the value to be set. It must point to a variable
500 * of the correct type/size for the register being accessed.
502 * Returns: 0 on success, or a negative errno on failure.
504 int kvm_set_one_reg(CPUState *cs, uint64_t id, void *source);
507 * kvm_get_one_reg - get a register value from KVM via KVM_GET_ONE_REG ioctl
508 * @id: The register ID
509 * @target: The pointer where the value is to be stored. It must point to a
510 * variable of the correct type/size for the register being accessed.
512 * Returns: 0 on success, or a negative errno on failure.
514 int kvm_get_one_reg(CPUState *cs, uint64_t id, void *target);