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"
19 #include "exec/memattrs.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_split_irqchip
;
45 extern bool kvm_async_interrupts_allowed
;
46 extern bool kvm_halt_in_kernel_allowed
;
47 extern bool kvm_eventfds_allowed
;
48 extern bool kvm_irqfds_allowed
;
49 extern bool kvm_resamplefds_allowed
;
50 extern bool kvm_msi_via_irqfd_allowed
;
51 extern bool kvm_gsi_routing_allowed
;
52 extern bool kvm_gsi_direct_mapping
;
53 extern bool kvm_readonly_mem_allowed
;
54 extern bool kvm_direct_msi_allowed
;
55 extern bool kvm_ioeventfd_any_length_allowed
;
56 extern bool kvm_msi_use_devid
;
58 #if defined CONFIG_KVM || !defined NEED_CPU_H
59 #define kvm_enabled() (kvm_allowed)
61 * kvm_irqchip_in_kernel:
63 * Returns: true if the user asked us to create an in-kernel
64 * irqchip via the "kernel_irqchip=on" machine option.
65 * What this actually means is architecture and machine model
66 * specific: on PC, for instance, it means that the LAPIC,
67 * IOAPIC and PIT are all in kernel. This function should never
68 * be used from generic target-independent code: use one of the
69 * following functions or some other specific check instead.
71 #define kvm_irqchip_in_kernel() (kvm_kernel_irqchip)
74 * kvm_irqchip_is_split:
76 * Returns: true if the user asked us to split the irqchip
77 * implementation between user and kernel space. The details are
78 * architecture and machine specific. On PC, it means that the PIC,
79 * IOAPIC, and PIT are in user space while the LAPIC is in the kernel.
81 #define kvm_irqchip_is_split() (kvm_split_irqchip)
84 * kvm_async_interrupts_enabled:
86 * Returns: true if we can deliver interrupts to KVM
87 * asynchronously (ie by ioctl from any thread at any time)
88 * rather than having to do interrupt delivery synchronously
89 * (where the vcpu must be stopped at a suitable point first).
91 #define kvm_async_interrupts_enabled() (kvm_async_interrupts_allowed)
96 * Returns: true if halted cpus should still get a KVM_RUN ioctl to run
97 * inside of kernel space. This only works if MP state is implemented.
99 #define kvm_halt_in_kernel() (kvm_halt_in_kernel_allowed)
102 * kvm_eventfds_enabled:
104 * Returns: true if we can use eventfds to receive notifications
105 * from a KVM CPU (ie the kernel supports eventds and we are running
106 * with a configuration where it is meaningful to use them).
108 #define kvm_eventfds_enabled() (kvm_eventfds_allowed)
111 * kvm_irqfds_enabled:
113 * Returns: true if we can use irqfds to inject interrupts into
114 * a KVM CPU (ie the kernel supports irqfds and we are running
115 * with a configuration where it is meaningful to use them).
117 #define kvm_irqfds_enabled() (kvm_irqfds_allowed)
120 * kvm_resamplefds_enabled:
122 * Returns: true if we can use resamplefds to inject interrupts into
123 * a KVM CPU (ie the kernel supports resamplefds and we are running
124 * with a configuration where it is meaningful to use them).
126 #define kvm_resamplefds_enabled() (kvm_resamplefds_allowed)
129 * kvm_msi_via_irqfd_enabled:
131 * Returns: true if we can route a PCI MSI (Message Signaled Interrupt)
132 * to a KVM CPU via an irqfd. This requires that the kernel supports
133 * this and that we're running in a configuration that permits it.
135 #define kvm_msi_via_irqfd_enabled() (kvm_msi_via_irqfd_allowed)
138 * kvm_gsi_routing_enabled:
140 * Returns: true if GSI routing is enabled (ie the kernel supports
141 * it and we're running in a configuration that permits it).
143 #define kvm_gsi_routing_enabled() (kvm_gsi_routing_allowed)
146 * kvm_gsi_direct_mapping:
148 * Returns: true if GSI direct mapping is enabled.
150 #define kvm_gsi_direct_mapping() (kvm_gsi_direct_mapping)
153 * kvm_readonly_mem_enabled:
155 * Returns: true if KVM readonly memory is enabled (ie the kernel
156 * supports it and we're running in a configuration that permits it).
158 #define kvm_readonly_mem_enabled() (kvm_readonly_mem_allowed)
161 * kvm_direct_msi_enabled:
163 * Returns: true if KVM allows direct MSI injection.
165 #define kvm_direct_msi_enabled() (kvm_direct_msi_allowed)
168 * kvm_ioeventfd_any_length_enabled:
169 * Returns: true if KVM allows any length io eventfd.
171 #define kvm_ioeventfd_any_length_enabled() (kvm_ioeventfd_any_length_allowed)
174 * kvm_msi_devid_required:
175 * Returns: true if KVM requires a device id to be provided while
176 * defining an MSI routing entry.
178 #define kvm_msi_devid_required() (kvm_msi_use_devid)
181 #define kvm_enabled() (0)
182 #define kvm_irqchip_in_kernel() (false)
183 #define kvm_irqchip_is_split() (false)
184 #define kvm_async_interrupts_enabled() (false)
185 #define kvm_halt_in_kernel() (false)
186 #define kvm_eventfds_enabled() (false)
187 #define kvm_irqfds_enabled() (false)
188 #define kvm_resamplefds_enabled() (false)
189 #define kvm_msi_via_irqfd_enabled() (false)
190 #define kvm_gsi_routing_allowed() (false)
191 #define kvm_gsi_direct_mapping() (false)
192 #define kvm_readonly_mem_enabled() (false)
193 #define kvm_direct_msi_enabled() (false)
194 #define kvm_ioeventfd_any_length_enabled() (false)
195 #define kvm_msi_devid_required() (false)
199 struct kvm_lapic_state
;
200 struct kvm_irq_routing_entry
;
202 typedef struct KVMCapabilityInfo
{
207 #define KVM_CAP_INFO(CAP) { "KVM_CAP_" stringify(CAP), KVM_CAP_##CAP }
208 #define KVM_CAP_LAST_INFO { NULL, 0 }
211 typedef struct KVMState KVMState
;
212 extern KVMState
*kvm_state
;
216 bool kvm_has_free_slot(MachineState
*ms
);
217 int kvm_has_sync_mmu(void);
218 int kvm_has_vcpu_events(void);
219 int kvm_has_robust_singlestep(void);
220 int kvm_has_debugregs(void);
221 int kvm_has_pit_state2(void);
222 int kvm_has_many_ioeventfds(void);
223 int kvm_has_gsi_routing(void);
224 int kvm_has_intx_set_mask(void);
226 int kvm_init_vcpu(CPUState
*cpu
);
227 int kvm_cpu_exec(CPUState
*cpu
);
228 int kvm_destroy_vcpu(CPUState
*cpu
);
233 void kvm_flush_coalesced_mmio_buffer(void);
235 int kvm_insert_breakpoint(CPUState
*cpu
, target_ulong addr
,
236 target_ulong len
, int type
);
237 int kvm_remove_breakpoint(CPUState
*cpu
, target_ulong addr
,
238 target_ulong len
, int type
);
239 void kvm_remove_all_breakpoints(CPUState
*cpu
);
240 int kvm_update_guest_debug(CPUState
*cpu
, unsigned long reinject_trap
);
242 int kvm_on_sigbus_vcpu(CPUState
*cpu
, int code
, void *addr
);
243 int kvm_on_sigbus(int code
, void *addr
);
245 /* interface with exec.c */
247 void phys_mem_set_alloc(void *(*alloc
)(size_t, uint64_t *align
));
251 int kvm_ioctl(KVMState
*s
, int type
, ...);
253 int kvm_vm_ioctl(KVMState
*s
, int type
, ...);
255 int kvm_vcpu_ioctl(CPUState
*cpu
, int type
, ...);
258 * kvm_device_ioctl - call an ioctl on a kvm device
259 * @fd: The KVM device file descriptor as returned from KVM_CREATE_DEVICE
260 * @type: The device-ctrl ioctl number
262 * Returns: -errno on error, nonnegative on success
264 int kvm_device_ioctl(int fd
, int type
, ...);
267 * kvm_vm_check_attr - check for existence of a specific vm attribute
268 * @s: The KVMState pointer
270 * @attr: the attribute of that group to query for
272 * Returns: 1 if the attribute exists
273 * 0 if the attribute either does not exist or if the vm device
274 * interface is unavailable
276 int kvm_vm_check_attr(KVMState
*s
, uint32_t group
, uint64_t attr
);
279 * kvm_device_check_attr - check for existence of a specific device attribute
280 * @fd: The device file descriptor
282 * @attr: the attribute of that group to query for
284 * Returns: 1 if the attribute exists
285 * 0 if the attribute either does not exist or if the vm device
286 * interface is unavailable
288 int kvm_device_check_attr(int fd
, uint32_t group
, uint64_t attr
);
291 * kvm_device_access - set or get value of a specific vm attribute
292 * @fd: The device file descriptor
294 * @attr: the attribute of that group to set or get
295 * @val: pointer to a storage area for the value
296 * @write: true for set and false for get operation
298 * This function is not allowed to fail. Use kvm_device_check_attr()
299 * in order to check for the availability of optional attributes.
301 void kvm_device_access(int fd
, int group
, uint64_t attr
,
302 void *val
, bool write
);
305 * kvm_create_device - create a KVM device for the device control API
306 * @KVMState: The KVMState pointer
307 * @type: The KVM device type (see Documentation/virtual/kvm/devices in the
309 * @test: If true, only test if device can be created, but don't actually
312 * Returns: -errno on error, nonnegative on success: @test ? 0 : device fd;
314 int kvm_create_device(KVMState
*s
, uint64_t type
, bool test
);
317 * kvm_device_supported - probe whether KVM supports specific device
319 * @vmfd: The fd handler for VM
320 * @type: type of device
322 * @return: true if supported, otherwise false.
324 bool kvm_device_supported(int vmfd
, uint64_t type
);
326 /* Arch specific hooks */
328 extern const KVMCapabilityInfo kvm_arch_required_capabilities
[];
330 void kvm_arch_pre_run(CPUState
*cpu
, struct kvm_run
*run
);
331 MemTxAttrs
kvm_arch_post_run(CPUState
*cpu
, struct kvm_run
*run
);
333 int kvm_arch_handle_exit(CPUState
*cpu
, struct kvm_run
*run
);
335 int kvm_arch_process_async_events(CPUState
*cpu
);
337 int kvm_arch_get_registers(CPUState
*cpu
);
339 /* state subset only touched by the VCPU itself during runtime */
340 #define KVM_PUT_RUNTIME_STATE 1
341 /* state subset modified during VCPU reset */
342 #define KVM_PUT_RESET_STATE 2
343 /* full state set, modified during initialization or on vmload */
344 #define KVM_PUT_FULL_STATE 3
346 int kvm_arch_put_registers(CPUState
*cpu
, int level
);
348 int kvm_arch_init(MachineState
*ms
, KVMState
*s
);
350 int kvm_arch_init_vcpu(CPUState
*cpu
);
352 bool kvm_vcpu_id_is_valid(int vcpu_id
);
354 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */
355 unsigned long kvm_arch_vcpu_id(CPUState
*cpu
);
358 #define KVM_HAVE_MCE_INJECTION 1
359 void kvm_arch_on_sigbus_vcpu(CPUState
*cpu
, int code
, void *addr
);
362 void kvm_arch_init_irq_routing(KVMState
*s
);
364 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry
*route
,
365 uint64_t address
, uint32_t data
, PCIDevice
*dev
);
367 /* Notify arch about newly added MSI routes */
368 int kvm_arch_add_msi_route_post(struct kvm_irq_routing_entry
*route
,
369 int vector
, PCIDevice
*dev
);
370 /* Notify arch about released MSI routes */
371 int kvm_arch_release_virq_post(int virq
);
373 int kvm_arch_msi_data_to_gsi(uint32_t data
);
375 int kvm_set_irq(KVMState
*s
, int irq
, int level
);
376 int kvm_irqchip_send_msi(KVMState
*s
, MSIMessage msg
);
378 void kvm_irqchip_add_irq_route(KVMState
*s
, int gsi
, int irqchip
, int pin
);
380 void kvm_get_apic_state(DeviceState
*d
, struct kvm_lapic_state
*kapic
);
382 struct kvm_guest_debug
;
383 struct kvm_debug_exit_arch
;
385 struct kvm_sw_breakpoint
{
387 target_ulong saved_insn
;
389 QTAILQ_ENTRY(kvm_sw_breakpoint
) entry
;
392 QTAILQ_HEAD(kvm_sw_breakpoint_head
, kvm_sw_breakpoint
);
394 struct kvm_sw_breakpoint
*kvm_find_sw_breakpoint(CPUState
*cpu
,
397 int kvm_sw_breakpoints_active(CPUState
*cpu
);
399 int kvm_arch_insert_sw_breakpoint(CPUState
*cpu
,
400 struct kvm_sw_breakpoint
*bp
);
401 int kvm_arch_remove_sw_breakpoint(CPUState
*cpu
,
402 struct kvm_sw_breakpoint
*bp
);
403 int kvm_arch_insert_hw_breakpoint(target_ulong addr
,
404 target_ulong len
, int type
);
405 int kvm_arch_remove_hw_breakpoint(target_ulong addr
,
406 target_ulong len
, int type
);
407 void kvm_arch_remove_all_hw_breakpoints(void);
409 void kvm_arch_update_guest_debug(CPUState
*cpu
, struct kvm_guest_debug
*dbg
);
411 bool kvm_arch_stop_on_emulation_error(CPUState
*cpu
);
413 int kvm_check_extension(KVMState
*s
, unsigned int extension
);
415 int kvm_vm_check_extension(KVMState
*s
, unsigned int extension
);
417 #define kvm_vm_enable_cap(s, capability, cap_flags, ...) \
419 struct kvm_enable_cap cap = { \
421 .flags = cap_flags, \
423 uint64_t args_tmp[] = { __VA_ARGS__ }; \
425 for (i = 0; i < (int)ARRAY_SIZE(args_tmp) && \
426 i < ARRAY_SIZE(cap.args); i++) { \
427 cap.args[i] = args_tmp[i]; \
429 kvm_vm_ioctl(s, KVM_ENABLE_CAP, &cap); \
432 #define kvm_vcpu_enable_cap(cpu, capability, cap_flags, ...) \
434 struct kvm_enable_cap cap = { \
436 .flags = cap_flags, \
438 uint64_t args_tmp[] = { __VA_ARGS__ }; \
440 for (i = 0; i < (int)ARRAY_SIZE(args_tmp) && \
441 i < ARRAY_SIZE(cap.args); i++) { \
442 cap.args[i] = args_tmp[i]; \
444 kvm_vcpu_ioctl(cpu, KVM_ENABLE_CAP, &cap); \
447 uint32_t kvm_arch_get_supported_cpuid(KVMState
*env
, uint32_t function
,
448 uint32_t index
, int reg
);
450 void kvm_set_sigmask_len(KVMState
*s
, unsigned int sigmask_len
);
452 #if !defined(CONFIG_USER_ONLY)
453 int kvm_physical_memory_addr_from_host(KVMState
*s
, void *ram_addr
,
457 #endif /* NEED_CPU_H */
459 void kvm_cpu_synchronize_state(CPUState
*cpu
);
460 void kvm_cpu_synchronize_post_reset(CPUState
*cpu
);
461 void kvm_cpu_synchronize_post_init(CPUState
*cpu
);
462 void kvm_cpu_synchronize_pre_loadvm(CPUState
*cpu
);
464 void kvm_init_cpu_signals(CPUState
*cpu
);
467 * kvm_irqchip_add_msi_route - Add MSI route for specific vector
469 * @vector: which vector to add. This can be either MSI/MSIX
470 * vector. The function will automatically detect whether
471 * MSI/MSIX is enabled, and fetch corresponding MSI
473 * @dev: Owner PCI device to add the route. If @dev is specified
474 * as @NULL, an empty MSI message will be inited.
475 * @return: virq (>=0) when success, errno (<0) when failed.
477 int kvm_irqchip_add_msi_route(KVMState
*s
, int vector
, PCIDevice
*dev
);
478 int kvm_irqchip_update_msi_route(KVMState
*s
, int virq
, MSIMessage msg
,
480 void kvm_irqchip_commit_routes(KVMState
*s
);
481 void kvm_irqchip_release_virq(KVMState
*s
, int virq
);
483 int kvm_irqchip_add_adapter_route(KVMState
*s
, AdapterInfo
*adapter
);
484 int kvm_irqchip_add_hv_sint_route(KVMState
*s
, uint32_t vcpu
, uint32_t sint
);
486 int kvm_irqchip_add_irqfd_notifier_gsi(KVMState
*s
, EventNotifier
*n
,
487 EventNotifier
*rn
, int virq
);
488 int kvm_irqchip_remove_irqfd_notifier_gsi(KVMState
*s
, EventNotifier
*n
,
490 int kvm_irqchip_add_irqfd_notifier(KVMState
*s
, EventNotifier
*n
,
491 EventNotifier
*rn
, qemu_irq irq
);
492 int kvm_irqchip_remove_irqfd_notifier(KVMState
*s
, EventNotifier
*n
,
494 void kvm_irqchip_set_qemuirq_gsi(KVMState
*s
, qemu_irq irq
, int gsi
);
495 void kvm_pc_gsi_handler(void *opaque
, int n
, int level
);
496 void kvm_pc_setup_irq_routing(bool pci_enabled
);
497 void kvm_init_irq_routing(KVMState
*s
);
500 * kvm_arch_irqchip_create:
501 * @KVMState: The KVMState pointer
502 * @MachineState: The MachineState pointer
504 * Allow architectures to create an in-kernel irq chip themselves.
506 * Returns: < 0: error
507 * 0: irq chip was not created
508 * > 0: irq chip was created
510 int kvm_arch_irqchip_create(MachineState
*ms
, KVMState
*s
);
513 * kvm_set_one_reg - set a register value in KVM via KVM_SET_ONE_REG ioctl
514 * @id: The register ID
515 * @source: The pointer to the value to be set. It must point to a variable
516 * of the correct type/size for the register being accessed.
518 * Returns: 0 on success, or a negative errno on failure.
520 int kvm_set_one_reg(CPUState
*cs
, uint64_t id
, void *source
);
523 * kvm_get_one_reg - get a register value from KVM via KVM_GET_ONE_REG ioctl
524 * @id: The register ID
525 * @target: The pointer where the value is to be stored. It must point to a
526 * variable of the correct type/size for the register being accessed.
528 * Returns: 0 on success, or a negative errno on failure.
530 int kvm_get_one_reg(CPUState
*cs
, uint64_t id
, void *target
);
531 struct ppc_radix_page_info
*kvm_get_radix_page_info(void);
532 int kvm_get_max_memslots(void);