4 * Copyright IBM, Corp. 2008
7 * Anthony Liguori <aliguori@us.ibm.com>
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"
23 #include <linux/kvm.h>
24 #include <linux/kvm_para.h>
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.
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
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_eventfds_allowed
;
47 extern bool kvm_irqfds_allowed
;
48 extern bool kvm_resamplefds_allowed
;
49 extern bool kvm_msi_via_irqfd_allowed
;
50 extern bool kvm_gsi_routing_allowed
;
51 extern bool kvm_gsi_direct_mapping
;
52 extern bool kvm_readonly_mem_allowed
;
54 #if defined CONFIG_KVM || !defined NEED_CPU_H
55 #define kvm_enabled() (kvm_allowed)
57 * kvm_irqchip_in_kernel:
59 * Returns: true if the user asked us to create an in-kernel
60 * irqchip via the "kernel_irqchip=on" machine option.
61 * What this actually means is architecture and machine model
62 * specific: on PC, for instance, it means that the LAPIC,
63 * IOAPIC and PIT are all in kernel. This function should never
64 * be used from generic target-independent code: use one of the
65 * following functions or some other specific check instead.
67 #define kvm_irqchip_in_kernel() (kvm_kernel_irqchip)
70 * kvm_async_interrupts_enabled:
72 * Returns: true if we can deliver interrupts to KVM
73 * asynchronously (ie by ioctl from any thread at any time)
74 * rather than having to do interrupt delivery synchronously
75 * (where the vcpu must be stopped at a suitable point first).
77 #define kvm_async_interrupts_enabled() (kvm_async_interrupts_allowed)
82 * Returns: true if halted cpus should still get a KVM_RUN ioctl to run
83 * inside of kernel space. This only works if MP state is implemented.
85 #define kvm_halt_in_kernel() (kvm_halt_in_kernel_allowed)
88 * kvm_eventfds_enabled:
90 * Returns: true if we can use eventfds to receive notifications
91 * from a KVM CPU (ie the kernel supports eventds and we are running
92 * with a configuration where it is meaningful to use them).
94 #define kvm_eventfds_enabled() (kvm_eventfds_allowed)
99 * Returns: true if we can use irqfds to inject interrupts into
100 * a KVM CPU (ie the kernel supports irqfds and we are running
101 * with a configuration where it is meaningful to use them).
103 #define kvm_irqfds_enabled() (kvm_irqfds_allowed)
106 * kvm_resamplefds_enabled:
108 * Returns: true if we can use resamplefds to inject interrupts into
109 * a KVM CPU (ie the kernel supports resamplefds and we are running
110 * with a configuration where it is meaningful to use them).
112 #define kvm_resamplefds_enabled() (kvm_resamplefds_allowed)
115 * kvm_msi_via_irqfd_enabled:
117 * Returns: true if we can route a PCI MSI (Message Signaled Interrupt)
118 * to a KVM CPU via an irqfd. This requires that the kernel supports
119 * this and that we're running in a configuration that permits it.
121 #define kvm_msi_via_irqfd_enabled() (kvm_msi_via_irqfd_allowed)
124 * kvm_gsi_routing_enabled:
126 * Returns: true if GSI routing is enabled (ie the kernel supports
127 * it and we're running in a configuration that permits it).
129 #define kvm_gsi_routing_enabled() (kvm_gsi_routing_allowed)
132 * kvm_gsi_direct_mapping:
134 * Returns: true if GSI direct mapping is enabled.
136 #define kvm_gsi_direct_mapping() (kvm_gsi_direct_mapping)
139 * kvm_readonly_mem_enabled:
141 * Returns: true if KVM readonly memory is enabled (ie the kernel
142 * supports it and we're running in a configuration that permits it).
144 #define kvm_readonly_mem_enabled() (kvm_readonly_mem_allowed)
147 #define kvm_enabled() (0)
148 #define kvm_irqchip_in_kernel() (false)
149 #define kvm_async_interrupts_enabled() (false)
150 #define kvm_halt_in_kernel() (false)
151 #define kvm_eventfds_enabled() (false)
152 #define kvm_irqfds_enabled() (false)
153 #define kvm_msi_via_irqfd_enabled() (false)
154 #define kvm_gsi_routing_allowed() (false)
155 #define kvm_gsi_direct_mapping() (false)
156 #define kvm_readonly_mem_enabled() (false)
160 struct kvm_lapic_state
;
161 struct kvm_irq_routing_entry
;
163 typedef struct KVMCapabilityInfo
{
168 #define KVM_CAP_INFO(CAP) { "KVM_CAP_" stringify(CAP), KVM_CAP_##CAP }
169 #define KVM_CAP_LAST_INFO { NULL, 0 }
172 typedef struct KVMState KVMState
;
173 extern KVMState
*kvm_state
;
177 bool kvm_has_free_slot(MachineState
*ms
);
178 int kvm_has_sync_mmu(void);
179 int kvm_has_vcpu_events(void);
180 int kvm_has_robust_singlestep(void);
181 int kvm_has_debugregs(void);
182 int kvm_has_xsave(void);
183 int kvm_has_xcrs(void);
184 int kvm_has_pit_state2(void);
185 int kvm_has_many_ioeventfds(void);
186 int kvm_has_gsi_routing(void);
187 int kvm_has_intx_set_mask(void);
189 int kvm_init_vcpu(CPUState
*cpu
);
190 int kvm_cpu_exec(CPUState
*cpu
);
194 void kvm_setup_guest_memory(void *start
, size_t size
);
195 void kvm_flush_coalesced_mmio_buffer(void);
197 int kvm_insert_breakpoint(CPUState
*cpu
, target_ulong addr
,
198 target_ulong len
, int type
);
199 int kvm_remove_breakpoint(CPUState
*cpu
, target_ulong addr
,
200 target_ulong len
, int type
);
201 void kvm_remove_all_breakpoints(CPUState
*cpu
);
202 int kvm_update_guest_debug(CPUState
*cpu
, unsigned long reinject_trap
);
204 int kvm_set_signal_mask(CPUState
*cpu
, const sigset_t
*sigset
);
207 int kvm_on_sigbus_vcpu(CPUState
*cpu
, int code
, void *addr
);
208 int kvm_on_sigbus(int code
, void *addr
);
212 int kvm_ioctl(KVMState
*s
, int type
, ...);
214 int kvm_vm_ioctl(KVMState
*s
, int type
, ...);
216 int kvm_vcpu_ioctl(CPUState
*cpu
, int type
, ...);
219 * kvm_device_ioctl - call an ioctl on a kvm device
220 * @fd: The KVM device file descriptor as returned from KVM_CREATE_DEVICE
221 * @type: The device-ctrl ioctl number
223 * Returns: -errno on error, nonnegative on success
225 int kvm_device_ioctl(int fd
, int type
, ...);
228 * kvm_vm_check_attr - check for existence of a specific vm attribute
229 * @s: The KVMState pointer
231 * @attr: the attribute of that group to query for
233 * Returns: 1 if the attribute exists
234 * 0 if the attribute either does not exist or if the vm device
235 * interface is unavailable
237 int kvm_vm_check_attr(KVMState
*s
, uint32_t group
, uint64_t attr
);
240 * kvm_create_device - create a KVM device for the device control API
241 * @KVMState: The KVMState pointer
242 * @type: The KVM device type (see Documentation/virtual/kvm/devices in the
244 * @test: If true, only test if device can be created, but don't actually
247 * Returns: -errno on error, nonnegative on success: @test ? 0 : device fd;
249 int kvm_create_device(KVMState
*s
, uint64_t type
, bool test
);
252 /* Arch specific hooks */
254 extern const KVMCapabilityInfo kvm_arch_required_capabilities
[];
256 void kvm_arch_pre_run(CPUState
*cpu
, struct kvm_run
*run
);
257 void kvm_arch_post_run(CPUState
*cpu
, struct kvm_run
*run
);
259 int kvm_arch_handle_exit(CPUState
*cpu
, struct kvm_run
*run
);
261 int kvm_arch_process_async_events(CPUState
*cpu
);
263 int kvm_arch_get_registers(CPUState
*cpu
);
265 /* state subset only touched by the VCPU itself during runtime */
266 #define KVM_PUT_RUNTIME_STATE 1
267 /* state subset modified during VCPU reset */
268 #define KVM_PUT_RESET_STATE 2
269 /* full state set, modified during initialization or on vmload */
270 #define KVM_PUT_FULL_STATE 3
272 int kvm_arch_put_registers(CPUState
*cpu
, int level
);
274 int kvm_arch_init(MachineState
*ms
, KVMState
*s
);
276 int kvm_arch_init_vcpu(CPUState
*cpu
);
278 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */
279 unsigned long kvm_arch_vcpu_id(CPUState
*cpu
);
281 int kvm_arch_on_sigbus_vcpu(CPUState
*cpu
, int code
, void *addr
);
282 int kvm_arch_on_sigbus(int code
, void *addr
);
284 void kvm_arch_init_irq_routing(KVMState
*s
);
286 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry
*route
,
287 uint64_t address
, uint32_t data
);
289 int kvm_set_irq(KVMState
*s
, int irq
, int level
);
290 int kvm_irqchip_send_msi(KVMState
*s
, MSIMessage msg
);
292 void kvm_irqchip_add_irq_route(KVMState
*s
, int gsi
, int irqchip
, int pin
);
293 void kvm_irqchip_commit_routes(KVMState
*s
);
295 void kvm_put_apic_state(DeviceState
*d
, struct kvm_lapic_state
*kapic
);
296 void kvm_get_apic_state(DeviceState
*d
, struct kvm_lapic_state
*kapic
);
298 struct kvm_guest_debug
;
299 struct kvm_debug_exit_arch
;
301 struct kvm_sw_breakpoint
{
303 target_ulong saved_insn
;
305 QTAILQ_ENTRY(kvm_sw_breakpoint
) entry
;
308 QTAILQ_HEAD(kvm_sw_breakpoint_head
, kvm_sw_breakpoint
);
310 struct kvm_sw_breakpoint
*kvm_find_sw_breakpoint(CPUState
*cpu
,
313 int kvm_sw_breakpoints_active(CPUState
*cpu
);
315 int kvm_arch_insert_sw_breakpoint(CPUState
*cpu
,
316 struct kvm_sw_breakpoint
*bp
);
317 int kvm_arch_remove_sw_breakpoint(CPUState
*cpu
,
318 struct kvm_sw_breakpoint
*bp
);
319 int kvm_arch_insert_hw_breakpoint(target_ulong addr
,
320 target_ulong len
, int type
);
321 int kvm_arch_remove_hw_breakpoint(target_ulong addr
,
322 target_ulong len
, int type
);
323 void kvm_arch_remove_all_hw_breakpoints(void);
325 void kvm_arch_update_guest_debug(CPUState
*cpu
, struct kvm_guest_debug
*dbg
);
327 bool kvm_arch_stop_on_emulation_error(CPUState
*cpu
);
329 int kvm_check_extension(KVMState
*s
, unsigned int extension
);
331 int kvm_vm_check_extension(KVMState
*s
, unsigned int extension
);
333 #define kvm_vm_enable_cap(s, capability, cap_flags, ...) \
335 struct kvm_enable_cap cap = { \
337 .flags = cap_flags, \
339 uint64_t args_tmp[] = { __VA_ARGS__ }; \
341 for (i = 0; i < (int)ARRAY_SIZE(args_tmp) && \
342 i < ARRAY_SIZE(cap.args); i++) { \
343 cap.args[i] = args_tmp[i]; \
345 kvm_vm_ioctl(s, KVM_ENABLE_CAP, &cap); \
348 #define kvm_vcpu_enable_cap(cpu, capability, cap_flags, ...) \
350 struct kvm_enable_cap cap = { \
352 .flags = cap_flags, \
354 uint64_t args_tmp[] = { __VA_ARGS__ }; \
356 for (i = 0; i < (int)ARRAY_SIZE(args_tmp) && \
357 i < ARRAY_SIZE(cap.args); i++) { \
358 cap.args[i] = args_tmp[i]; \
360 kvm_vcpu_ioctl(cpu, KVM_ENABLE_CAP, &cap); \
363 uint32_t kvm_arch_get_supported_cpuid(KVMState
*env
, uint32_t function
,
364 uint32_t index
, int reg
);
366 void kvm_set_sigmask_len(KVMState
*s
, unsigned int sigmask_len
);
368 #if !defined(CONFIG_USER_ONLY)
369 int kvm_physical_memory_addr_from_host(KVMState
*s
, void *ram_addr
,
373 #endif /* NEED_CPU_H */
375 void kvm_cpu_synchronize_state(CPUState
*cpu
);
376 void kvm_cpu_synchronize_post_reset(CPUState
*cpu
);
377 void kvm_cpu_synchronize_post_init(CPUState
*cpu
);
378 void kvm_cpu_clean_state(CPUState
*cpu
);
380 /* generic hooks - to be moved/refactored once there are more users */
382 static inline void cpu_synchronize_state(CPUState
*cpu
)
385 kvm_cpu_synchronize_state(cpu
);
389 static inline void cpu_synchronize_post_reset(CPUState
*cpu
)
392 kvm_cpu_synchronize_post_reset(cpu
);
396 static inline void cpu_synchronize_post_init(CPUState
*cpu
)
399 kvm_cpu_synchronize_post_init(cpu
);
403 static inline void cpu_clean_state(CPUState
*cpu
)
406 kvm_cpu_clean_state(cpu
);
410 int kvm_irqchip_add_msi_route(KVMState
*s
, MSIMessage msg
);
411 int kvm_irqchip_update_msi_route(KVMState
*s
, int virq
, MSIMessage msg
);
412 void kvm_irqchip_release_virq(KVMState
*s
, int virq
);
414 int kvm_irqchip_add_adapter_route(KVMState
*s
, AdapterInfo
*adapter
);
416 int kvm_irqchip_add_irqfd_notifier(KVMState
*s
, EventNotifier
*n
,
417 EventNotifier
*rn
, int virq
);
418 int kvm_irqchip_remove_irqfd_notifier(KVMState
*s
, EventNotifier
*n
, int virq
);
419 void kvm_pc_gsi_handler(void *opaque
, int n
, int level
);
420 void kvm_pc_setup_irq_routing(bool pci_enabled
);
421 void kvm_init_irq_routing(KVMState
*s
);
424 * kvm_arch_irqchip_create:
425 * @KVMState: The KVMState pointer
427 * Allow architectures to create an in-kernel irq chip themselves.
429 * Returns: < 0: error
430 * 0: irq chip was not created
431 * > 0: irq chip was created
433 int kvm_arch_irqchip_create(KVMState
*s
);
436 * kvm_set_one_reg - set a register value in KVM via KVM_SET_ONE_REG ioctl
437 * @id: The register ID
438 * @source: The pointer to the value to be set. It must point to a variable
439 * of the correct type/size for the register being accessed.
441 * Returns: 0 on success, or a negative errno on failure.
443 int kvm_set_one_reg(CPUState
*cs
, uint64_t id
, void *source
);
446 * kvm_get_one_reg - get a register value from KVM via KVM_GET_ONE_REG ioctl
447 * @id: The register ID
448 * @target: The pointer where the value is to be stored. It must point to a
449 * variable of the correct type/size for the register being accessed.
451 * Returns: 0 on success, or a negative errno on failure.
453 int kvm_get_one_reg(CPUState
*cs
, uint64_t id
, void *target
);