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.
17 #include "qemu/queue.h"
18 #include "hw/core/cpu.h"
19 #include "exec/memattrs.h"
20 #include "qemu/accel.h"
21 #include "qom/object.h"
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_eventfds_allowed
;
40 extern bool kvm_irqfds_allowed
;
41 extern bool kvm_resamplefds_allowed
;
42 extern bool kvm_msi_via_irqfd_allowed
;
43 extern bool kvm_gsi_routing_allowed
;
44 extern bool kvm_gsi_direct_mapping
;
45 extern bool kvm_readonly_mem_allowed
;
46 extern bool kvm_direct_msi_allowed
;
47 extern bool kvm_ioeventfd_any_length_allowed
;
48 extern bool kvm_msi_use_devid
;
50 #define kvm_enabled() (kvm_allowed)
52 * kvm_irqchip_in_kernel:
54 * Returns: true if an in-kernel irqchip was created.
55 * What this actually means is architecture and machine model
56 * specific: on PC, for instance, it means that the LAPIC
57 * is in kernel. This function should never be used from generic
58 * target-independent code: use one of the following functions or
59 * some other specific check instead.
61 #define kvm_irqchip_in_kernel() (kvm_kernel_irqchip)
64 * kvm_irqchip_is_split:
66 * Returns: true if the irqchip implementation is split between
67 * user and kernel space. The details are architecture and
68 * machine specific. On PC, it means that the PIC, IOAPIC, and
69 * PIT are in user space while the LAPIC is in the kernel.
71 #define kvm_irqchip_is_split() (kvm_split_irqchip)
74 * kvm_async_interrupts_enabled:
76 * Returns: true if we can deliver interrupts to KVM
77 * asynchronously (ie by ioctl from any thread at any time)
78 * rather than having to do interrupt delivery synchronously
79 * (where the vcpu must be stopped at a suitable point first).
81 #define kvm_async_interrupts_enabled() (kvm_async_interrupts_allowed)
86 * Returns: true if halted cpus should still get a KVM_RUN ioctl to run
87 * inside of kernel space. This only works if MP state is implemented.
89 #define kvm_halt_in_kernel() (kvm_halt_in_kernel_allowed)
92 * kvm_eventfds_enabled:
94 * Returns: true if we can use eventfds to receive notifications
95 * from a KVM CPU (ie the kernel supports eventds and we are running
96 * with a configuration where it is meaningful to use them).
98 #define kvm_eventfds_enabled() (kvm_eventfds_allowed)
101 * kvm_irqfds_enabled:
103 * Returns: true if we can use irqfds to inject interrupts into
104 * a KVM CPU (ie the kernel supports irqfds and we are running
105 * with a configuration where it is meaningful to use them).
107 #define kvm_irqfds_enabled() (kvm_irqfds_allowed)
110 * kvm_resamplefds_enabled:
112 * Returns: true if we can use resamplefds to inject interrupts into
113 * a KVM CPU (ie the kernel supports resamplefds and we are running
114 * with a configuration where it is meaningful to use them).
116 #define kvm_resamplefds_enabled() (kvm_resamplefds_allowed)
119 * kvm_msi_via_irqfd_enabled:
121 * Returns: true if we can route a PCI MSI (Message Signaled Interrupt)
122 * to a KVM CPU via an irqfd. This requires that the kernel supports
123 * this and that we're running in a configuration that permits it.
125 #define kvm_msi_via_irqfd_enabled() (kvm_msi_via_irqfd_allowed)
128 * kvm_gsi_routing_enabled:
130 * Returns: true if GSI routing is enabled (ie the kernel supports
131 * it and we're running in a configuration that permits it).
133 #define kvm_gsi_routing_enabled() (kvm_gsi_routing_allowed)
136 * kvm_gsi_direct_mapping:
138 * Returns: true if GSI direct mapping is enabled.
140 #define kvm_gsi_direct_mapping() (kvm_gsi_direct_mapping)
143 * kvm_readonly_mem_enabled:
145 * Returns: true if KVM readonly memory is enabled (ie the kernel
146 * supports it and we're running in a configuration that permits it).
148 #define kvm_readonly_mem_enabled() (kvm_readonly_mem_allowed)
151 * kvm_direct_msi_enabled:
153 * Returns: true if KVM allows direct MSI injection.
155 #define kvm_direct_msi_enabled() (kvm_direct_msi_allowed)
158 * kvm_ioeventfd_any_length_enabled:
159 * Returns: true if KVM allows any length io eventfd.
161 #define kvm_ioeventfd_any_length_enabled() (kvm_ioeventfd_any_length_allowed)
164 * kvm_msi_devid_required:
165 * Returns: true if KVM requires a device id to be provided while
166 * defining an MSI routing entry.
168 #define kvm_msi_devid_required() (kvm_msi_use_devid)
172 #define kvm_enabled() (0)
173 #define kvm_irqchip_in_kernel() (false)
174 #define kvm_irqchip_is_split() (false)
175 #define kvm_async_interrupts_enabled() (false)
176 #define kvm_halt_in_kernel() (false)
177 #define kvm_eventfds_enabled() (false)
178 #define kvm_irqfds_enabled() (false)
179 #define kvm_resamplefds_enabled() (false)
180 #define kvm_msi_via_irqfd_enabled() (false)
181 #define kvm_gsi_routing_allowed() (false)
182 #define kvm_gsi_direct_mapping() (false)
183 #define kvm_readonly_mem_enabled() (false)
184 #define kvm_direct_msi_enabled() (false)
185 #define kvm_ioeventfd_any_length_enabled() (false)
186 #define kvm_msi_devid_required() (false)
188 #endif /* CONFIG_KVM_IS_POSSIBLE */
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 }
204 #define TYPE_KVM_ACCEL ACCEL_CLASS_NAME("kvm")
205 typedef struct KVMState KVMState
;
206 DECLARE_INSTANCE_CHECKER(KVMState
, KVM_STATE
,
209 extern KVMState
*kvm_state
;
210 typedef struct Notifier Notifier
;
212 typedef struct KVMRouteChange
{
219 bool kvm_has_free_slot(MachineState
*ms
);
220 bool kvm_has_sync_mmu(void);
221 int kvm_has_vcpu_events(void);
222 int kvm_has_robust_singlestep(void);
223 int kvm_has_debugregs(void);
224 int kvm_max_nested_state_length(void);
225 int kvm_has_pit_state2(void);
226 int kvm_has_many_ioeventfds(void);
227 int kvm_has_gsi_routing(void);
228 int kvm_has_intx_set_mask(void);
231 * kvm_arm_supports_user_irq
233 * Not all KVM implementations support notifications for kernel generated
234 * interrupt events to user space. This function indicates whether the current
235 * KVM implementation does support them.
237 * Returns: true if KVM supports using kernel generated IRQs from user space
239 bool kvm_arm_supports_user_irq(void);
242 int kvm_on_sigbus_vcpu(CPUState
*cpu
, int code
, void *addr
);
243 int kvm_on_sigbus(int code
, void *addr
);
248 void kvm_flush_coalesced_mmio_buffer(void);
251 * kvm_update_guest_debug(): ensure KVM debug structures updated
252 * @cs: the CPUState for this cpu
253 * @reinject_trap: KVM trap injection control
255 * There are usually per-arch specifics which will be handled by
256 * calling down to kvm_arch_update_guest_debug after the generic
257 * fields have been set.
259 #ifdef KVM_CAP_SET_GUEST_DEBUG
260 int kvm_update_guest_debug(CPUState
*cpu
, unsigned long reinject_trap
);
262 static inline int kvm_update_guest_debug(CPUState
*cpu
, unsigned long reinject_trap
)
270 int kvm_ioctl(KVMState
*s
, int type
, ...);
272 int kvm_vm_ioctl(KVMState
*s
, int type
, ...);
274 int kvm_vcpu_ioctl(CPUState
*cpu
, int type
, ...);
277 * kvm_device_ioctl - call an ioctl on a kvm device
278 * @fd: The KVM device file descriptor as returned from KVM_CREATE_DEVICE
279 * @type: The device-ctrl ioctl number
281 * Returns: -errno on error, nonnegative on success
283 int kvm_device_ioctl(int fd
, int type
, ...);
286 * kvm_vm_check_attr - check for existence of a specific vm attribute
287 * @s: The KVMState pointer
289 * @attr: the attribute of that group to query for
291 * Returns: 1 if the attribute exists
292 * 0 if the attribute either does not exist or if the vm device
293 * interface is unavailable
295 int kvm_vm_check_attr(KVMState
*s
, uint32_t group
, uint64_t attr
);
298 * kvm_device_check_attr - check for existence of a specific device attribute
299 * @fd: The device file descriptor
301 * @attr: the attribute of that group to query for
303 * Returns: 1 if the attribute exists
304 * 0 if the attribute either does not exist or if the vm device
305 * interface is unavailable
307 int kvm_device_check_attr(int fd
, uint32_t group
, uint64_t attr
);
310 * kvm_device_access - set or get value of a specific device attribute
311 * @fd: The device file descriptor
313 * @attr: the attribute of that group to set or get
314 * @val: pointer to a storage area for the value
315 * @write: true for set and false for get operation
316 * @errp: error object handle
318 * Returns: 0 on success
320 * Use kvm_device_check_attr() in order to check for the availability
321 * of optional attributes.
323 int kvm_device_access(int fd
, int group
, uint64_t attr
,
324 void *val
, bool write
, Error
**errp
);
327 * kvm_create_device - create a KVM device for the device control API
328 * @KVMState: The KVMState pointer
329 * @type: The KVM device type (see Documentation/virtual/kvm/devices in the
331 * @test: If true, only test if device can be created, but don't actually
334 * Returns: -errno on error, nonnegative on success: @test ? 0 : device fd;
336 int kvm_create_device(KVMState
*s
, uint64_t type
, bool test
);
339 * kvm_device_supported - probe whether KVM supports specific device
341 * @vmfd: The fd handler for VM
342 * @type: type of device
344 * @return: true if supported, otherwise false.
346 bool kvm_device_supported(int vmfd
, uint64_t type
);
348 /* Arch specific hooks */
350 extern const KVMCapabilityInfo kvm_arch_required_capabilities
[];
352 void kvm_arch_accel_class_init(ObjectClass
*oc
);
354 void kvm_arch_pre_run(CPUState
*cpu
, struct kvm_run
*run
);
355 MemTxAttrs
kvm_arch_post_run(CPUState
*cpu
, struct kvm_run
*run
);
357 int kvm_arch_handle_exit(CPUState
*cpu
, struct kvm_run
*run
);
359 int kvm_arch_process_async_events(CPUState
*cpu
);
361 int kvm_arch_get_registers(CPUState
*cpu
);
363 /* state subset only touched by the VCPU itself during runtime */
364 #define KVM_PUT_RUNTIME_STATE 1
365 /* state subset modified during VCPU reset */
366 #define KVM_PUT_RESET_STATE 2
367 /* full state set, modified during initialization or on vmload */
368 #define KVM_PUT_FULL_STATE 3
370 int kvm_arch_put_registers(CPUState
*cpu
, int level
);
372 int kvm_arch_init(MachineState
*ms
, KVMState
*s
);
374 int kvm_arch_init_vcpu(CPUState
*cpu
);
375 int kvm_arch_destroy_vcpu(CPUState
*cpu
);
377 bool kvm_vcpu_id_is_valid(int vcpu_id
);
379 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */
380 unsigned long kvm_arch_vcpu_id(CPUState
*cpu
);
382 #ifdef KVM_HAVE_MCE_INJECTION
383 void kvm_arch_on_sigbus_vcpu(CPUState
*cpu
, int code
, void *addr
);
386 void kvm_arch_init_irq_routing(KVMState
*s
);
388 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry
*route
,
389 uint64_t address
, uint32_t data
, PCIDevice
*dev
);
391 /* Notify arch about newly added MSI routes */
392 int kvm_arch_add_msi_route_post(struct kvm_irq_routing_entry
*route
,
393 int vector
, PCIDevice
*dev
);
394 /* Notify arch about released MSI routes */
395 int kvm_arch_release_virq_post(int virq
);
397 int kvm_arch_msi_data_to_gsi(uint32_t data
);
399 int kvm_set_irq(KVMState
*s
, int irq
, int level
);
400 int kvm_irqchip_send_msi(KVMState
*s
, MSIMessage msg
);
402 void kvm_irqchip_add_irq_route(KVMState
*s
, int gsi
, int irqchip
, int pin
);
404 void kvm_irqchip_add_change_notifier(Notifier
*n
);
405 void kvm_irqchip_remove_change_notifier(Notifier
*n
);
406 void kvm_irqchip_change_notify(void);
408 void kvm_get_apic_state(DeviceState
*d
, struct kvm_lapic_state
*kapic
);
410 struct kvm_guest_debug
;
411 struct kvm_debug_exit_arch
;
413 struct kvm_sw_breakpoint
{
415 target_ulong saved_insn
;
417 QTAILQ_ENTRY(kvm_sw_breakpoint
) entry
;
420 struct kvm_sw_breakpoint
*kvm_find_sw_breakpoint(CPUState
*cpu
,
423 int kvm_sw_breakpoints_active(CPUState
*cpu
);
425 int kvm_arch_insert_sw_breakpoint(CPUState
*cpu
,
426 struct kvm_sw_breakpoint
*bp
);
427 int kvm_arch_remove_sw_breakpoint(CPUState
*cpu
,
428 struct kvm_sw_breakpoint
*bp
);
429 int kvm_arch_insert_hw_breakpoint(target_ulong addr
,
430 target_ulong len
, int type
);
431 int kvm_arch_remove_hw_breakpoint(target_ulong addr
,
432 target_ulong len
, int type
);
433 void kvm_arch_remove_all_hw_breakpoints(void);
435 void kvm_arch_update_guest_debug(CPUState
*cpu
, struct kvm_guest_debug
*dbg
);
437 bool kvm_arch_stop_on_emulation_error(CPUState
*cpu
);
439 int kvm_check_extension(KVMState
*s
, unsigned int extension
);
441 int kvm_vm_check_extension(KVMState
*s
, unsigned int extension
);
443 #define kvm_vm_enable_cap(s, capability, cap_flags, ...) \
445 struct kvm_enable_cap cap = { \
447 .flags = cap_flags, \
449 uint64_t args_tmp[] = { __VA_ARGS__ }; \
450 size_t n = MIN(ARRAY_SIZE(args_tmp), ARRAY_SIZE(cap.args)); \
451 memcpy(cap.args, args_tmp, n * sizeof(cap.args[0])); \
452 kvm_vm_ioctl(s, KVM_ENABLE_CAP, &cap); \
455 #define kvm_vcpu_enable_cap(cpu, capability, cap_flags, ...) \
457 struct kvm_enable_cap cap = { \
459 .flags = cap_flags, \
461 uint64_t args_tmp[] = { __VA_ARGS__ }; \
462 size_t n = MIN(ARRAY_SIZE(args_tmp), ARRAY_SIZE(cap.args)); \
463 memcpy(cap.args, args_tmp, n * sizeof(cap.args[0])); \
464 kvm_vcpu_ioctl(cpu, KVM_ENABLE_CAP, &cap); \
467 uint32_t kvm_arch_get_supported_cpuid(KVMState
*env
, uint32_t function
,
468 uint32_t index
, int reg
);
469 uint64_t kvm_arch_get_supported_msr_feature(KVMState
*s
, uint32_t index
);
472 void kvm_set_sigmask_len(KVMState
*s
, unsigned int sigmask_len
);
474 int kvm_physical_memory_addr_from_host(KVMState
*s
, void *ram_addr
,
477 #endif /* NEED_CPU_H */
479 void kvm_cpu_synchronize_state(CPUState
*cpu
);
481 void kvm_init_cpu_signals(CPUState
*cpu
);
484 * kvm_irqchip_add_msi_route - Add MSI route for specific vector
485 * @c: KVMRouteChange instance.
486 * @vector: which vector to add. This can be either MSI/MSIX
487 * vector. The function will automatically detect whether
488 * MSI/MSIX is enabled, and fetch corresponding MSI
490 * @dev: Owner PCI device to add the route. If @dev is specified
491 * as @NULL, an empty MSI message will be inited.
492 * @return: virq (>=0) when success, errno (<0) when failed.
494 int kvm_irqchip_add_msi_route(KVMRouteChange
*c
, int vector
, PCIDevice
*dev
);
495 int kvm_irqchip_update_msi_route(KVMState
*s
, int virq
, MSIMessage msg
,
497 void kvm_irqchip_commit_routes(KVMState
*s
);
499 static inline KVMRouteChange
kvm_irqchip_begin_route_changes(KVMState
*s
)
501 return (KVMRouteChange
) { .s
= s
, .changes
= 0 };
504 static inline void kvm_irqchip_commit_route_changes(KVMRouteChange
*c
)
507 kvm_irqchip_commit_routes(c
->s
);
512 void kvm_irqchip_release_virq(KVMState
*s
, int virq
);
514 int kvm_irqchip_add_adapter_route(KVMState
*s
, AdapterInfo
*adapter
);
515 int kvm_irqchip_add_hv_sint_route(KVMState
*s
, uint32_t vcpu
, uint32_t sint
);
517 int kvm_irqchip_add_irqfd_notifier_gsi(KVMState
*s
, EventNotifier
*n
,
518 EventNotifier
*rn
, int virq
);
519 int kvm_irqchip_remove_irqfd_notifier_gsi(KVMState
*s
, EventNotifier
*n
,
521 int kvm_irqchip_add_irqfd_notifier(KVMState
*s
, EventNotifier
*n
,
522 EventNotifier
*rn
, qemu_irq irq
);
523 int kvm_irqchip_remove_irqfd_notifier(KVMState
*s
, EventNotifier
*n
,
525 void kvm_irqchip_set_qemuirq_gsi(KVMState
*s
, qemu_irq irq
, int gsi
);
526 void kvm_pc_setup_irq_routing(bool pci_enabled
);
527 void kvm_init_irq_routing(KVMState
*s
);
529 bool kvm_kernel_irqchip_allowed(void);
530 bool kvm_kernel_irqchip_required(void);
531 bool kvm_kernel_irqchip_split(void);
534 * kvm_arch_irqchip_create:
535 * @KVMState: The KVMState pointer
537 * Allow architectures to create an in-kernel irq chip themselves.
539 * Returns: < 0: error
540 * 0: irq chip was not created
541 * > 0: irq chip was created
543 int kvm_arch_irqchip_create(KVMState
*s
);
546 * kvm_set_one_reg - set a register value in KVM via KVM_SET_ONE_REG ioctl
547 * @id: The register ID
548 * @source: The pointer to the value to be set. It must point to a variable
549 * of the correct type/size for the register being accessed.
551 * Returns: 0 on success, or a negative errno on failure.
553 int kvm_set_one_reg(CPUState
*cs
, uint64_t id
, void *source
);
556 * kvm_get_one_reg - get a register value from KVM via KVM_GET_ONE_REG ioctl
557 * @id: The register ID
558 * @target: The pointer where the value is to be stored. It must point to a
559 * variable of the correct type/size for the register being accessed.
561 * Returns: 0 on success, or a negative errno on failure.
563 int kvm_get_one_reg(CPUState
*cs
, uint64_t id
, void *target
);
564 struct ppc_radix_page_info
*kvm_get_radix_page_info(void);
565 int kvm_get_max_memslots(void);
567 /* Notify resamplefd for EOI of specific interrupts. */
568 void kvm_resample_fd_notify(int gsi
);
571 * kvm_cpu_check_are_resettable - return whether CPUs can be reset
573 * Returns: true: CPUs are resettable
574 * false: CPUs are not resettable
576 bool kvm_cpu_check_are_resettable(void);
578 bool kvm_arch_cpu_check_are_resettable(void);
580 bool kvm_dirty_ring_enabled(void);
582 uint32_t kvm_dirty_ring_size(void);