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
;
49 extern bool kvm_has_guest_debug
;
50 extern int kvm_sstep_flags
;
52 #define kvm_enabled() (kvm_allowed)
54 * kvm_irqchip_in_kernel:
56 * Returns: true if an in-kernel irqchip was created.
57 * What this actually means is architecture and machine model
58 * specific: on PC, for instance, it means that the LAPIC
59 * is in kernel. This function should never be used from generic
60 * target-independent code: use one of the following functions or
61 * some other specific check instead.
63 #define kvm_irqchip_in_kernel() (kvm_kernel_irqchip)
66 * kvm_irqchip_is_split:
68 * Returns: true if the irqchip implementation is split between
69 * user and kernel space. The details are architecture and
70 * machine specific. On PC, it means that the PIC, IOAPIC, and
71 * PIT are in user space while the LAPIC is in the kernel.
73 #define kvm_irqchip_is_split() (kvm_split_irqchip)
76 * kvm_async_interrupts_enabled:
78 * Returns: true if we can deliver interrupts to KVM
79 * asynchronously (ie by ioctl from any thread at any time)
80 * rather than having to do interrupt delivery synchronously
81 * (where the vcpu must be stopped at a suitable point first).
83 #define kvm_async_interrupts_enabled() (kvm_async_interrupts_allowed)
88 * Returns: true if halted cpus should still get a KVM_RUN ioctl to run
89 * inside of kernel space. This only works if MP state is implemented.
91 #define kvm_halt_in_kernel() (kvm_halt_in_kernel_allowed)
94 * kvm_eventfds_enabled:
96 * Returns: true if we can use eventfds to receive notifications
97 * from a KVM CPU (ie the kernel supports eventds and we are running
98 * with a configuration where it is meaningful to use them).
100 #define kvm_eventfds_enabled() (kvm_eventfds_allowed)
103 * kvm_irqfds_enabled:
105 * Returns: true if we can use irqfds to inject interrupts into
106 * a KVM CPU (ie the kernel supports irqfds and we are running
107 * with a configuration where it is meaningful to use them).
109 #define kvm_irqfds_enabled() (kvm_irqfds_allowed)
112 * kvm_resamplefds_enabled:
114 * Returns: true if we can use resamplefds to inject interrupts into
115 * a KVM CPU (ie the kernel supports resamplefds and we are running
116 * with a configuration where it is meaningful to use them).
118 #define kvm_resamplefds_enabled() (kvm_resamplefds_allowed)
121 * kvm_msi_via_irqfd_enabled:
123 * Returns: true if we can route a PCI MSI (Message Signaled Interrupt)
124 * to a KVM CPU via an irqfd. This requires that the kernel supports
125 * this and that we're running in a configuration that permits it.
127 #define kvm_msi_via_irqfd_enabled() (kvm_msi_via_irqfd_allowed)
130 * kvm_gsi_routing_enabled:
132 * Returns: true if GSI routing is enabled (ie the kernel supports
133 * it and we're running in a configuration that permits it).
135 #define kvm_gsi_routing_enabled() (kvm_gsi_routing_allowed)
138 * kvm_gsi_direct_mapping:
140 * Returns: true if GSI direct mapping is enabled.
142 #define kvm_gsi_direct_mapping() (kvm_gsi_direct_mapping)
145 * kvm_readonly_mem_enabled:
147 * Returns: true if KVM readonly memory is enabled (ie the kernel
148 * supports it and we're running in a configuration that permits it).
150 #define kvm_readonly_mem_enabled() (kvm_readonly_mem_allowed)
153 * kvm_direct_msi_enabled:
155 * Returns: true if KVM allows direct MSI injection.
157 #define kvm_direct_msi_enabled() (kvm_direct_msi_allowed)
160 * kvm_ioeventfd_any_length_enabled:
161 * Returns: true if KVM allows any length io eventfd.
163 #define kvm_ioeventfd_any_length_enabled() (kvm_ioeventfd_any_length_allowed)
166 * kvm_msi_devid_required:
167 * Returns: true if KVM requires a device id to be provided while
168 * defining an MSI routing entry.
170 #define kvm_msi_devid_required() (kvm_msi_use_devid)
173 * Does KVM support guest debugging
175 #define kvm_supports_guest_debug() (kvm_has_guest_debug)
178 * kvm_supported_sstep_flags
179 * Returns: SSTEP_* flags that KVM supports for guest debug
181 #define kvm_get_supported_sstep_flags() (kvm_sstep_flags)
185 #define kvm_enabled() (0)
186 #define kvm_irqchip_in_kernel() (false)
187 #define kvm_irqchip_is_split() (false)
188 #define kvm_async_interrupts_enabled() (false)
189 #define kvm_halt_in_kernel() (false)
190 #define kvm_eventfds_enabled() (false)
191 #define kvm_irqfds_enabled() (false)
192 #define kvm_resamplefds_enabled() (false)
193 #define kvm_msi_via_irqfd_enabled() (false)
194 #define kvm_gsi_routing_allowed() (false)
195 #define kvm_gsi_direct_mapping() (false)
196 #define kvm_readonly_mem_enabled() (false)
197 #define kvm_direct_msi_enabled() (false)
198 #define kvm_ioeventfd_any_length_enabled() (false)
199 #define kvm_msi_devid_required() (false)
200 #define kvm_supports_guest_debug() (false)
201 #define kvm_get_supported_sstep_flags() (0)
203 #endif /* CONFIG_KVM_IS_POSSIBLE */
206 struct kvm_lapic_state
;
207 struct kvm_irq_routing_entry
;
209 typedef struct KVMCapabilityInfo
{
214 #define KVM_CAP_INFO(CAP) { "KVM_CAP_" stringify(CAP), KVM_CAP_##CAP }
215 #define KVM_CAP_LAST_INFO { NULL, 0 }
219 #define TYPE_KVM_ACCEL ACCEL_CLASS_NAME("kvm")
220 typedef struct KVMState KVMState
;
221 DECLARE_INSTANCE_CHECKER(KVMState
, KVM_STATE
,
224 extern KVMState
*kvm_state
;
225 typedef struct Notifier Notifier
;
229 bool kvm_has_free_slot(MachineState
*ms
);
230 bool kvm_has_sync_mmu(void);
231 int kvm_has_vcpu_events(void);
232 int kvm_has_robust_singlestep(void);
233 int kvm_has_debugregs(void);
234 int kvm_max_nested_state_length(void);
235 int kvm_has_pit_state2(void);
236 int kvm_has_many_ioeventfds(void);
237 int kvm_has_gsi_routing(void);
238 int kvm_has_intx_set_mask(void);
241 * kvm_arm_supports_user_irq
243 * Not all KVM implementations support notifications for kernel generated
244 * interrupt events to user space. This function indicates whether the current
245 * KVM implementation does support them.
247 * Returns: true if KVM supports using kernel generated IRQs from user space
249 bool kvm_arm_supports_user_irq(void);
252 int kvm_on_sigbus_vcpu(CPUState
*cpu
, int code
, void *addr
);
253 int kvm_on_sigbus(int code
, void *addr
);
258 void kvm_flush_coalesced_mmio_buffer(void);
260 int kvm_insert_breakpoint(CPUState
*cpu
, target_ulong addr
,
261 target_ulong len
, int type
);
262 int kvm_remove_breakpoint(CPUState
*cpu
, target_ulong addr
,
263 target_ulong len
, int type
);
264 void kvm_remove_all_breakpoints(CPUState
*cpu
);
265 int kvm_update_guest_debug(CPUState
*cpu
, unsigned long reinject_trap
);
269 int kvm_ioctl(KVMState
*s
, int type
, ...);
271 int kvm_vm_ioctl(KVMState
*s
, int type
, ...);
273 int kvm_vcpu_ioctl(CPUState
*cpu
, int type
, ...);
276 * kvm_device_ioctl - call an ioctl on a kvm device
277 * @fd: The KVM device file descriptor as returned from KVM_CREATE_DEVICE
278 * @type: The device-ctrl ioctl number
280 * Returns: -errno on error, nonnegative on success
282 int kvm_device_ioctl(int fd
, int type
, ...);
285 * kvm_vm_check_attr - check for existence of a specific vm attribute
286 * @s: The KVMState pointer
288 * @attr: the attribute of that group to query for
290 * Returns: 1 if the attribute exists
291 * 0 if the attribute either does not exist or if the vm device
292 * interface is unavailable
294 int kvm_vm_check_attr(KVMState
*s
, uint32_t group
, uint64_t attr
);
297 * kvm_device_check_attr - check for existence of a specific device attribute
298 * @fd: The device file descriptor
300 * @attr: the attribute of that group to query for
302 * Returns: 1 if the attribute exists
303 * 0 if the attribute either does not exist or if the vm device
304 * interface is unavailable
306 int kvm_device_check_attr(int fd
, uint32_t group
, uint64_t attr
);
309 * kvm_device_access - set or get value of a specific device attribute
310 * @fd: The device file descriptor
312 * @attr: the attribute of that group to set or get
313 * @val: pointer to a storage area for the value
314 * @write: true for set and false for get operation
315 * @errp: error object handle
317 * Returns: 0 on success
319 * Use kvm_device_check_attr() in order to check for the availability
320 * of optional attributes.
322 int kvm_device_access(int fd
, int group
, uint64_t attr
,
323 void *val
, bool write
, Error
**errp
);
326 * kvm_create_device - create a KVM device for the device control API
327 * @KVMState: The KVMState pointer
328 * @type: The KVM device type (see Documentation/virtual/kvm/devices in the
330 * @test: If true, only test if device can be created, but don't actually
333 * Returns: -errno on error, nonnegative on success: @test ? 0 : device fd;
335 int kvm_create_device(KVMState
*s
, uint64_t type
, bool test
);
338 * kvm_device_supported - probe whether KVM supports specific device
340 * @vmfd: The fd handler for VM
341 * @type: type of device
343 * @return: true if supported, otherwise false.
345 bool kvm_device_supported(int vmfd
, uint64_t type
);
347 /* Arch specific hooks */
349 extern const KVMCapabilityInfo kvm_arch_required_capabilities
[];
351 void kvm_arch_pre_run(CPUState
*cpu
, struct kvm_run
*run
);
352 MemTxAttrs
kvm_arch_post_run(CPUState
*cpu
, struct kvm_run
*run
);
354 int kvm_arch_handle_exit(CPUState
*cpu
, struct kvm_run
*run
);
356 int kvm_arch_process_async_events(CPUState
*cpu
);
358 int kvm_arch_get_registers(CPUState
*cpu
);
360 /* state subset only touched by the VCPU itself during runtime */
361 #define KVM_PUT_RUNTIME_STATE 1
362 /* state subset modified during VCPU reset */
363 #define KVM_PUT_RESET_STATE 2
364 /* full state set, modified during initialization or on vmload */
365 #define KVM_PUT_FULL_STATE 3
367 int kvm_arch_put_registers(CPUState
*cpu
, int level
);
369 int kvm_arch_init(MachineState
*ms
, KVMState
*s
);
371 int kvm_arch_init_vcpu(CPUState
*cpu
);
372 int kvm_arch_destroy_vcpu(CPUState
*cpu
);
374 bool kvm_vcpu_id_is_valid(int vcpu_id
);
376 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */
377 unsigned long kvm_arch_vcpu_id(CPUState
*cpu
);
379 #ifdef KVM_HAVE_MCE_INJECTION
380 void kvm_arch_on_sigbus_vcpu(CPUState
*cpu
, int code
, void *addr
);
383 void kvm_arch_init_irq_routing(KVMState
*s
);
385 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry
*route
,
386 uint64_t address
, uint32_t data
, PCIDevice
*dev
);
388 /* Notify arch about newly added MSI routes */
389 int kvm_arch_add_msi_route_post(struct kvm_irq_routing_entry
*route
,
390 int vector
, PCIDevice
*dev
);
391 /* Notify arch about released MSI routes */
392 int kvm_arch_release_virq_post(int virq
);
394 int kvm_arch_msi_data_to_gsi(uint32_t data
);
396 int kvm_set_irq(KVMState
*s
, int irq
, int level
);
397 int kvm_irqchip_send_msi(KVMState
*s
, MSIMessage msg
);
399 void kvm_irqchip_add_irq_route(KVMState
*s
, int gsi
, int irqchip
, int pin
);
401 void kvm_irqchip_add_change_notifier(Notifier
*n
);
402 void kvm_irqchip_remove_change_notifier(Notifier
*n
);
403 void kvm_irqchip_change_notify(void);
405 void kvm_get_apic_state(DeviceState
*d
, struct kvm_lapic_state
*kapic
);
407 struct kvm_guest_debug
;
408 struct kvm_debug_exit_arch
;
410 struct kvm_sw_breakpoint
{
412 target_ulong saved_insn
;
414 QTAILQ_ENTRY(kvm_sw_breakpoint
) entry
;
417 struct kvm_sw_breakpoint
*kvm_find_sw_breakpoint(CPUState
*cpu
,
420 int kvm_sw_breakpoints_active(CPUState
*cpu
);
422 int kvm_arch_insert_sw_breakpoint(CPUState
*cpu
,
423 struct kvm_sw_breakpoint
*bp
);
424 int kvm_arch_remove_sw_breakpoint(CPUState
*cpu
,
425 struct kvm_sw_breakpoint
*bp
);
426 int kvm_arch_insert_hw_breakpoint(target_ulong addr
,
427 target_ulong len
, int type
);
428 int kvm_arch_remove_hw_breakpoint(target_ulong addr
,
429 target_ulong len
, int type
);
430 void kvm_arch_remove_all_hw_breakpoints(void);
432 void kvm_arch_update_guest_debug(CPUState
*cpu
, struct kvm_guest_debug
*dbg
);
434 bool kvm_arch_stop_on_emulation_error(CPUState
*cpu
);
436 int kvm_check_extension(KVMState
*s
, unsigned int extension
);
438 int kvm_vm_check_extension(KVMState
*s
, unsigned int extension
);
440 #define kvm_vm_enable_cap(s, capability, cap_flags, ...) \
442 struct kvm_enable_cap cap = { \
444 .flags = cap_flags, \
446 uint64_t args_tmp[] = { __VA_ARGS__ }; \
447 size_t n = MIN(ARRAY_SIZE(args_tmp), ARRAY_SIZE(cap.args)); \
448 memcpy(cap.args, args_tmp, n * sizeof(cap.args[0])); \
449 kvm_vm_ioctl(s, KVM_ENABLE_CAP, &cap); \
452 #define kvm_vcpu_enable_cap(cpu, capability, cap_flags, ...) \
454 struct kvm_enable_cap cap = { \
456 .flags = cap_flags, \
458 uint64_t args_tmp[] = { __VA_ARGS__ }; \
459 size_t n = MIN(ARRAY_SIZE(args_tmp), ARRAY_SIZE(cap.args)); \
460 memcpy(cap.args, args_tmp, n * sizeof(cap.args[0])); \
461 kvm_vcpu_ioctl(cpu, KVM_ENABLE_CAP, &cap); \
464 uint32_t kvm_arch_get_supported_cpuid(KVMState
*env
, uint32_t function
,
465 uint32_t index
, int reg
);
466 uint64_t kvm_arch_get_supported_msr_feature(KVMState
*s
, uint32_t index
);
469 void kvm_set_sigmask_len(KVMState
*s
, unsigned int sigmask_len
);
471 #if !defined(CONFIG_USER_ONLY)
472 int kvm_physical_memory_addr_from_host(KVMState
*s
, void *ram_addr
,
476 #endif /* NEED_CPU_H */
478 void kvm_cpu_synchronize_state(CPUState
*cpu
);
480 void kvm_init_cpu_signals(CPUState
*cpu
);
483 * kvm_irqchip_add_msi_route - Add MSI route for specific vector
485 * @vector: which vector to add. This can be either MSI/MSIX
486 * vector. The function will automatically detect whether
487 * MSI/MSIX is enabled, and fetch corresponding MSI
489 * @dev: Owner PCI device to add the route. If @dev is specified
490 * as @NULL, an empty MSI message will be inited.
491 * @return: virq (>=0) when success, errno (<0) when failed.
493 int kvm_irqchip_add_msi_route(KVMState
*s
, int vector
, PCIDevice
*dev
);
494 int kvm_irqchip_update_msi_route(KVMState
*s
, int virq
, MSIMessage msg
,
496 void kvm_irqchip_commit_routes(KVMState
*s
);
497 void kvm_irqchip_release_virq(KVMState
*s
, int virq
);
499 int kvm_irqchip_add_adapter_route(KVMState
*s
, AdapterInfo
*adapter
);
500 int kvm_irqchip_add_hv_sint_route(KVMState
*s
, uint32_t vcpu
, uint32_t sint
);
502 int kvm_irqchip_add_irqfd_notifier_gsi(KVMState
*s
, EventNotifier
*n
,
503 EventNotifier
*rn
, int virq
);
504 int kvm_irqchip_remove_irqfd_notifier_gsi(KVMState
*s
, EventNotifier
*n
,
506 int kvm_irqchip_add_irqfd_notifier(KVMState
*s
, EventNotifier
*n
,
507 EventNotifier
*rn
, qemu_irq irq
);
508 int kvm_irqchip_remove_irqfd_notifier(KVMState
*s
, EventNotifier
*n
,
510 void kvm_irqchip_set_qemuirq_gsi(KVMState
*s
, qemu_irq irq
, int gsi
);
511 void kvm_pc_setup_irq_routing(bool pci_enabled
);
512 void kvm_init_irq_routing(KVMState
*s
);
514 bool kvm_kernel_irqchip_allowed(void);
515 bool kvm_kernel_irqchip_required(void);
516 bool kvm_kernel_irqchip_split(void);
519 * kvm_arch_irqchip_create:
520 * @KVMState: The KVMState pointer
522 * Allow architectures to create an in-kernel irq chip themselves.
524 * Returns: < 0: error
525 * 0: irq chip was not created
526 * > 0: irq chip was created
528 int kvm_arch_irqchip_create(KVMState
*s
);
531 * kvm_set_one_reg - set a register value in KVM via KVM_SET_ONE_REG ioctl
532 * @id: The register ID
533 * @source: The pointer to the value to be set. It must point to a variable
534 * of the correct type/size for the register being accessed.
536 * Returns: 0 on success, or a negative errno on failure.
538 int kvm_set_one_reg(CPUState
*cs
, uint64_t id
, void *source
);
541 * kvm_get_one_reg - get a register value from KVM via KVM_GET_ONE_REG ioctl
542 * @id: The register ID
543 * @target: The pointer where the value is to be stored. It must point to a
544 * variable of the correct type/size for the register being accessed.
546 * Returns: 0 on success, or a negative errno on failure.
548 int kvm_get_one_reg(CPUState
*cs
, uint64_t id
, void *target
);
549 struct ppc_radix_page_info
*kvm_get_radix_page_info(void);
550 int kvm_get_max_memslots(void);
552 /* Notify resamplefd for EOI of specific interrupts. */
553 void kvm_resample_fd_notify(int gsi
);
556 * kvm_cpu_check_are_resettable - return whether CPUs can be reset
558 * Returns: true: CPUs are resettable
559 * false: CPUs are not resettable
561 bool kvm_cpu_check_are_resettable(void);
563 bool kvm_arch_cpu_check_are_resettable(void);
565 bool kvm_dirty_ring_enabled(void);