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.
14 /* header to be included in non-KVM-specific code */
19 #include "exec/memattrs.h"
20 #include "qemu/accel.h"
21 #include "qom/object.h"
23 #ifdef COMPILING_PER_TARGET
25 # include <linux/kvm.h>
26 # define CONFIG_KVM_IS_POSSIBLE
29 # define CONFIG_KVM_IS_POSSIBLE
32 #ifdef CONFIG_KVM_IS_POSSIBLE
34 extern bool kvm_allowed
;
35 extern bool kvm_kernel_irqchip
;
36 extern bool kvm_split_irqchip
;
37 extern bool kvm_async_interrupts_allowed
;
38 extern bool kvm_halt_in_kernel_allowed
;
39 extern bool kvm_resamplefds_allowed
;
40 extern bool kvm_msi_via_irqfd_allowed
;
41 extern bool kvm_gsi_routing_allowed
;
42 extern bool kvm_gsi_direct_mapping
;
43 extern bool kvm_readonly_mem_allowed
;
44 extern bool kvm_msi_use_devid
;
46 #define kvm_enabled() (kvm_allowed)
48 * kvm_irqchip_in_kernel:
50 * Returns: true if an in-kernel irqchip was created.
51 * What this actually means is architecture and machine model
52 * specific: on PC, for instance, it means that the LAPIC
53 * is in kernel. This function should never be used from generic
54 * target-independent code: use one of the following functions or
55 * some other specific check instead.
57 #define kvm_irqchip_in_kernel() (kvm_kernel_irqchip)
60 * kvm_irqchip_is_split:
62 * Returns: true if the irqchip implementation is split between
63 * user and kernel space. The details are architecture and
64 * machine specific. On PC, it means that the PIC, IOAPIC, and
65 * PIT are in user space while the LAPIC is in the kernel.
67 #define kvm_irqchip_is_split() (kvm_split_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)
90 * Returns: true if we can use irqfds to inject interrupts into
91 * a KVM CPU (ie the kernel supports irqfds and we are running
92 * with a configuration where it is meaningful to use them).
94 * Always available if running with in-kernel irqchip.
96 #define kvm_irqfds_enabled() kvm_irqchip_in_kernel()
99 * kvm_resamplefds_enabled:
101 * Returns: true if we can use resamplefds to inject interrupts into
102 * a KVM CPU (ie the kernel supports resamplefds and we are running
103 * with a configuration where it is meaningful to use them).
105 #define kvm_resamplefds_enabled() (kvm_resamplefds_allowed)
108 * kvm_msi_via_irqfd_enabled:
110 * Returns: true if we can route a PCI MSI (Message Signaled Interrupt)
111 * to a KVM CPU via an irqfd. This requires that the kernel supports
112 * this and that we're running in a configuration that permits it.
114 #define kvm_msi_via_irqfd_enabled() (kvm_msi_via_irqfd_allowed)
117 * kvm_gsi_routing_enabled:
119 * Returns: true if GSI routing is enabled (ie the kernel supports
120 * it and we're running in a configuration that permits it).
122 #define kvm_gsi_routing_enabled() (kvm_gsi_routing_allowed)
125 * kvm_gsi_direct_mapping:
127 * Returns: true if GSI direct mapping is enabled.
129 #define kvm_gsi_direct_mapping() (kvm_gsi_direct_mapping)
132 * kvm_readonly_mem_enabled:
134 * Returns: true if KVM readonly memory is enabled (ie the kernel
135 * supports it and we're running in a configuration that permits it).
137 #define kvm_readonly_mem_enabled() (kvm_readonly_mem_allowed)
140 * kvm_msi_devid_required:
141 * Returns: true if KVM requires a device id to be provided while
142 * defining an MSI routing entry.
144 #define kvm_msi_devid_required() (kvm_msi_use_devid)
148 #define kvm_enabled() (0)
149 #define kvm_irqchip_in_kernel() (false)
150 #define kvm_irqchip_is_split() (false)
151 #define kvm_async_interrupts_enabled() (false)
152 #define kvm_halt_in_kernel() (false)
153 #define kvm_irqfds_enabled() (false)
154 #define kvm_resamplefds_enabled() (false)
155 #define kvm_msi_via_irqfd_enabled() (false)
156 #define kvm_gsi_routing_allowed() (false)
157 #define kvm_gsi_direct_mapping() (false)
158 #define kvm_readonly_mem_enabled() (false)
159 #define kvm_msi_devid_required() (false)
161 #endif /* CONFIG_KVM_IS_POSSIBLE */
164 struct kvm_irq_routing_entry
;
166 typedef struct KVMCapabilityInfo
{
171 #define KVM_CAP_INFO(CAP) { "KVM_CAP_" stringify(CAP), KVM_CAP_##CAP }
172 #define KVM_CAP_LAST_INFO { NULL, 0 }
176 #define TYPE_KVM_ACCEL ACCEL_CLASS_NAME("kvm")
177 typedef struct KVMState KVMState
;
178 DECLARE_INSTANCE_CHECKER(KVMState
, KVM_STATE
,
181 extern KVMState
*kvm_state
;
182 typedef struct Notifier Notifier
;
184 typedef struct KVMRouteChange
{
191 unsigned int kvm_get_max_memslots(void);
192 unsigned int kvm_get_free_memslots(void);
193 bool kvm_has_sync_mmu(void);
194 int kvm_has_vcpu_events(void);
195 int kvm_max_nested_state_length(void);
196 int kvm_has_gsi_routing(void);
199 * kvm_arm_supports_user_irq
201 * Not all KVM implementations support notifications for kernel generated
202 * interrupt events to user space. This function indicates whether the current
203 * KVM implementation does support them.
205 * Returns: true if KVM supports using kernel generated IRQs from user space
207 bool kvm_arm_supports_user_irq(void);
210 int kvm_on_sigbus_vcpu(CPUState
*cpu
, int code
, void *addr
);
211 int kvm_on_sigbus(int code
, void *addr
);
213 #ifdef COMPILING_PER_TARGET
216 void kvm_flush_coalesced_mmio_buffer(void);
219 * kvm_update_guest_debug(): ensure KVM debug structures updated
220 * @cs: the CPUState for this cpu
221 * @reinject_trap: KVM trap injection control
223 * There are usually per-arch specifics which will be handled by
224 * calling down to kvm_arch_update_guest_debug after the generic
225 * fields have been set.
227 #ifdef TARGET_KVM_HAVE_GUEST_DEBUG
228 int kvm_update_guest_debug(CPUState
*cpu
, unsigned long reinject_trap
);
230 static inline int kvm_update_guest_debug(CPUState
*cpu
, unsigned long reinject_trap
)
238 int kvm_ioctl(KVMState
*s
, int type
, ...);
240 int kvm_vm_ioctl(KVMState
*s
, int type
, ...);
242 int kvm_vcpu_ioctl(CPUState
*cpu
, int type
, ...);
245 * kvm_device_ioctl - call an ioctl on a kvm device
246 * @fd: The KVM device file descriptor as returned from KVM_CREATE_DEVICE
247 * @type: The device-ctrl ioctl number
249 * Returns: -errno on error, nonnegative on success
251 int kvm_device_ioctl(int fd
, int type
, ...);
254 * kvm_vm_check_attr - check for existence of a specific vm attribute
255 * @s: The KVMState pointer
257 * @attr: the attribute of that group to query for
259 * Returns: 1 if the attribute exists
260 * 0 if the attribute either does not exist or if the vm device
261 * interface is unavailable
263 int kvm_vm_check_attr(KVMState
*s
, uint32_t group
, uint64_t attr
);
266 * kvm_device_check_attr - check for existence of a specific device attribute
267 * @fd: The device file descriptor
269 * @attr: the attribute of that group to query for
271 * Returns: 1 if the attribute exists
272 * 0 if the attribute either does not exist or if the vm device
273 * interface is unavailable
275 int kvm_device_check_attr(int fd
, uint32_t group
, uint64_t attr
);
278 * kvm_device_access - set or get value of a specific device attribute
279 * @fd: The device file descriptor
281 * @attr: the attribute of that group to set or get
282 * @val: pointer to a storage area for the value
283 * @write: true for set and false for get operation
284 * @errp: error object handle
286 * Returns: 0 on success
288 * Use kvm_device_check_attr() in order to check for the availability
289 * of optional attributes.
291 int kvm_device_access(int fd
, int group
, uint64_t attr
,
292 void *val
, bool write
, Error
**errp
);
295 * kvm_create_device - create a KVM device for the device control API
296 * @KVMState: The KVMState pointer
297 * @type: The KVM device type (see Documentation/virtual/kvm/devices in the
299 * @test: If true, only test if device can be created, but don't actually
302 * Returns: -errno on error, nonnegative on success: @test ? 0 : device fd;
304 int kvm_create_device(KVMState
*s
, uint64_t type
, bool test
);
307 * kvm_device_supported - probe whether KVM supports specific device
309 * @vmfd: The fd handler for VM
310 * @type: type of device
312 * @return: true if supported, otherwise false.
314 bool kvm_device_supported(int vmfd
, uint64_t type
);
317 * kvm_create_vcpu - Gets a parked KVM vCPU or creates a KVM vCPU
318 * @cpu: QOM CPUState object for which KVM vCPU has to be fetched/created.
320 * @returns: 0 when success, errno (<0) when failed.
322 int kvm_create_vcpu(CPUState
*cpu
);
325 * kvm_park_vcpu - Park QEMU KVM vCPU context
326 * @cpu: QOM CPUState object for which QEMU KVM vCPU context has to be parked.
330 void kvm_park_vcpu(CPUState
*cpu
);
333 * kvm_unpark_vcpu - unpark QEMU KVM vCPU context
335 * @vcpu_id: Architecture vCPU ID of the parked vCPU
339 int kvm_unpark_vcpu(KVMState
*s
, unsigned long vcpu_id
);
342 * kvm_create_and_park_vcpu - Create and park a KVM vCPU
343 * @cpu: QOM CPUState object for which KVM vCPU has to be created and parked.
345 * @returns: 0 when success, errno (<0) when failed.
347 int kvm_create_and_park_vcpu(CPUState
*cpu
);
349 /* Arch specific hooks */
351 extern const KVMCapabilityInfo kvm_arch_required_capabilities
[];
353 void kvm_arch_accel_class_init(ObjectClass
*oc
);
355 void kvm_arch_pre_run(CPUState
*cpu
, struct kvm_run
*run
);
356 MemTxAttrs
kvm_arch_post_run(CPUState
*cpu
, struct kvm_run
*run
);
358 int kvm_arch_handle_exit(CPUState
*cpu
, struct kvm_run
*run
);
360 int kvm_arch_process_async_events(CPUState
*cpu
);
362 int kvm_arch_get_registers(CPUState
*cpu
);
364 /* state subset only touched by the VCPU itself during runtime */
365 #define KVM_PUT_RUNTIME_STATE 1
366 /* state subset modified during VCPU reset */
367 #define KVM_PUT_RESET_STATE 2
368 /* full state set, modified during initialization or on vmload */
369 #define KVM_PUT_FULL_STATE 3
371 int kvm_arch_put_registers(CPUState
*cpu
, int level
);
373 int kvm_arch_get_default_type(MachineState
*ms
);
375 int kvm_arch_init(MachineState
*ms
, KVMState
*s
);
377 int kvm_arch_init_vcpu(CPUState
*cpu
);
378 int kvm_arch_destroy_vcpu(CPUState
*cpu
);
380 bool kvm_vcpu_id_is_valid(int vcpu_id
);
382 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */
383 unsigned long kvm_arch_vcpu_id(CPUState
*cpu
);
385 #ifdef KVM_HAVE_MCE_INJECTION
386 void kvm_arch_on_sigbus_vcpu(CPUState
*cpu
, int code
, void *addr
);
389 void kvm_arch_init_irq_routing(KVMState
*s
);
391 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry
*route
,
392 uint64_t address
, uint32_t data
, PCIDevice
*dev
);
394 /* Notify arch about newly added MSI routes */
395 int kvm_arch_add_msi_route_post(struct kvm_irq_routing_entry
*route
,
396 int vector
, PCIDevice
*dev
);
397 /* Notify arch about released MSI routes */
398 int kvm_arch_release_virq_post(int virq
);
400 int kvm_arch_msi_data_to_gsi(uint32_t data
);
402 int kvm_set_irq(KVMState
*s
, int irq
, int level
);
403 int kvm_irqchip_send_msi(KVMState
*s
, MSIMessage msg
);
405 void kvm_irqchip_add_irq_route(KVMState
*s
, int gsi
, int irqchip
, int pin
);
407 void kvm_irqchip_add_change_notifier(Notifier
*n
);
408 void kvm_irqchip_remove_change_notifier(Notifier
*n
);
409 void kvm_irqchip_change_notify(void);
411 struct kvm_guest_debug
;
412 struct kvm_debug_exit_arch
;
414 struct kvm_sw_breakpoint
{
418 QTAILQ_ENTRY(kvm_sw_breakpoint
) entry
;
421 struct kvm_sw_breakpoint
*kvm_find_sw_breakpoint(CPUState
*cpu
,
424 int kvm_sw_breakpoints_active(CPUState
*cpu
);
426 int kvm_arch_insert_sw_breakpoint(CPUState
*cpu
,
427 struct kvm_sw_breakpoint
*bp
);
428 int kvm_arch_remove_sw_breakpoint(CPUState
*cpu
,
429 struct kvm_sw_breakpoint
*bp
);
430 int kvm_arch_insert_hw_breakpoint(vaddr addr
, vaddr len
, int type
);
431 int kvm_arch_remove_hw_breakpoint(vaddr addr
, vaddr len
, int type
);
432 void kvm_arch_remove_all_hw_breakpoints(void);
434 void kvm_arch_update_guest_debug(CPUState
*cpu
, struct kvm_guest_debug
*dbg
);
436 bool kvm_arch_stop_on_emulation_error(CPUState
*cpu
);
438 int kvm_check_extension(KVMState
*s
, unsigned int extension
);
440 int kvm_vm_check_extension(KVMState
*s
, unsigned int extension
);
442 #define kvm_vm_enable_cap(s, capability, cap_flags, ...) \
444 struct kvm_enable_cap cap = { \
446 .flags = cap_flags, \
448 uint64_t args_tmp[] = { __VA_ARGS__ }; \
449 size_t n = MIN(ARRAY_SIZE(args_tmp), ARRAY_SIZE(cap.args)); \
450 memcpy(cap.args, args_tmp, n * sizeof(cap.args[0])); \
451 kvm_vm_ioctl(s, KVM_ENABLE_CAP, &cap); \
454 #define kvm_vcpu_enable_cap(cpu, capability, cap_flags, ...) \
456 struct kvm_enable_cap cap = { \
458 .flags = cap_flags, \
460 uint64_t args_tmp[] = { __VA_ARGS__ }; \
461 size_t n = MIN(ARRAY_SIZE(args_tmp), ARRAY_SIZE(cap.args)); \
462 memcpy(cap.args, args_tmp, n * sizeof(cap.args[0])); \
463 kvm_vcpu_ioctl(cpu, KVM_ENABLE_CAP, &cap); \
466 void kvm_set_sigmask_len(KVMState
*s
, unsigned int sigmask_len
);
468 int kvm_physical_memory_addr_from_host(KVMState
*s
, void *ram_addr
,
471 #endif /* COMPILING_PER_TARGET */
473 void kvm_cpu_synchronize_state(CPUState
*cpu
);
475 void kvm_init_cpu_signals(CPUState
*cpu
);
478 * kvm_irqchip_add_msi_route - Add MSI route for specific vector
479 * @c: KVMRouteChange instance.
480 * @vector: which vector to add. This can be either MSI/MSIX
481 * vector. The function will automatically detect whether
482 * MSI/MSIX is enabled, and fetch corresponding MSI
484 * @dev: Owner PCI device to add the route. If @dev is specified
485 * as @NULL, an empty MSI message will be inited.
486 * @return: virq (>=0) when success, errno (<0) when failed.
488 int kvm_irqchip_add_msi_route(KVMRouteChange
*c
, int vector
, PCIDevice
*dev
);
489 int kvm_irqchip_update_msi_route(KVMState
*s
, int virq
, MSIMessage msg
,
491 void kvm_irqchip_commit_routes(KVMState
*s
);
493 static inline KVMRouteChange
kvm_irqchip_begin_route_changes(KVMState
*s
)
495 return (KVMRouteChange
) { .s
= s
, .changes
= 0 };
498 static inline void kvm_irqchip_commit_route_changes(KVMRouteChange
*c
)
501 kvm_irqchip_commit_routes(c
->s
);
506 int kvm_irqchip_get_virq(KVMState
*s
);
507 void kvm_irqchip_release_virq(KVMState
*s
, int virq
);
509 void kvm_add_routing_entry(KVMState
*s
,
510 struct kvm_irq_routing_entry
*entry
);
512 int kvm_irqchip_add_irqfd_notifier_gsi(KVMState
*s
, EventNotifier
*n
,
513 EventNotifier
*rn
, int virq
);
514 int kvm_irqchip_remove_irqfd_notifier_gsi(KVMState
*s
, EventNotifier
*n
,
516 int kvm_irqchip_add_irqfd_notifier(KVMState
*s
, EventNotifier
*n
,
517 EventNotifier
*rn
, qemu_irq irq
);
518 int kvm_irqchip_remove_irqfd_notifier(KVMState
*s
, EventNotifier
*n
,
520 void kvm_irqchip_set_qemuirq_gsi(KVMState
*s
, qemu_irq irq
, int gsi
);
521 void kvm_init_irq_routing(KVMState
*s
);
523 bool kvm_kernel_irqchip_allowed(void);
524 bool kvm_kernel_irqchip_required(void);
525 bool kvm_kernel_irqchip_split(void);
528 * kvm_arch_irqchip_create:
529 * @KVMState: The KVMState pointer
531 * Allow architectures to create an in-kernel irq chip themselves.
533 * Returns: < 0: error
534 * 0: irq chip was not created
535 * > 0: irq chip was created
537 int kvm_arch_irqchip_create(KVMState
*s
);
540 * kvm_set_one_reg - set a register value in KVM via KVM_SET_ONE_REG ioctl
541 * @id: The register ID
542 * @source: The pointer to the value to be set. It must point to a variable
543 * of the correct type/size for the register being accessed.
545 * Returns: 0 on success, or a negative errno on failure.
547 int kvm_set_one_reg(CPUState
*cs
, uint64_t id
, void *source
);
550 * kvm_get_one_reg - get a register value from KVM via KVM_GET_ONE_REG ioctl
551 * @id: The register ID
552 * @target: The pointer where the value is to be stored. It must point to a
553 * variable of the correct type/size for the register being accessed.
555 * Returns: 0 on success, or a negative errno on failure.
557 int kvm_get_one_reg(CPUState
*cs
, uint64_t id
, void *target
);
559 /* Notify resamplefd for EOI of specific interrupts. */
560 void kvm_resample_fd_notify(int gsi
);
562 bool kvm_dirty_ring_enabled(void);
564 uint32_t kvm_dirty_ring_size(void);
566 void kvm_mark_guest_state_protected(void);
569 * kvm_hwpoisoned_mem - indicate if there is any hwpoisoned page
570 * reported for the VM.
572 bool kvm_hwpoisoned_mem(void);
574 int kvm_create_guest_memfd(uint64_t size
, uint64_t flags
, Error
**errp
);
576 int kvm_set_memory_attributes_private(hwaddr start
, uint64_t size
);
577 int kvm_set_memory_attributes_shared(hwaddr start
, uint64_t size
);
579 int kvm_convert_memory(hwaddr start
, hwaddr size
, bool to_private
);