xref: /openbmc/qemu/include/sysemu/kvm.h (revision dd205025)
1 /*
2  * QEMU KVM support
3  *
4  * Copyright IBM, Corp. 2008
5  *
6  * Authors:
7  *  Anthony Liguori   <aliguori@us.ibm.com>
8  *
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.
11  *
12  */
13 
14 #ifndef QEMU_KVM_H
15 #define QEMU_KVM_H
16 
17 #include "qemu/queue.h"
18 #include "hw/core/cpu.h"
19 #include "exec/memattrs.h"
20 #include "sysemu/accel.h"
21 
22 #ifdef NEED_CPU_H
23 # ifdef CONFIG_KVM
24 #  include <linux/kvm.h>
25 #  define CONFIG_KVM_IS_POSSIBLE
26 # endif
27 #else
28 # define CONFIG_KVM_IS_POSSIBLE
29 #endif
30 
31 #ifdef CONFIG_KVM_IS_POSSIBLE
32 
33 extern bool kvm_allowed;
34 extern bool kvm_kernel_irqchip;
35 extern bool kvm_split_irqchip;
36 extern bool kvm_async_interrupts_allowed;
37 extern bool kvm_halt_in_kernel_allowed;
38 extern bool kvm_eventfds_allowed;
39 extern bool kvm_irqfds_allowed;
40 extern bool kvm_resamplefds_allowed;
41 extern bool kvm_msi_via_irqfd_allowed;
42 extern bool kvm_gsi_routing_allowed;
43 extern bool kvm_gsi_direct_mapping;
44 extern bool kvm_readonly_mem_allowed;
45 extern bool kvm_direct_msi_allowed;
46 extern bool kvm_ioeventfd_any_length_allowed;
47 extern bool kvm_msi_use_devid;
48 
49 #define kvm_enabled()           (kvm_allowed)
50 /**
51  * kvm_irqchip_in_kernel:
52  *
53  * Returns: true if the user asked us to create an in-kernel
54  * irqchip via the "kernel_irqchip=on" machine option.
55  * What this actually means is architecture and machine model
56  * specific: on PC, for instance, it means that the LAPIC,
57  * IOAPIC and PIT are all in kernel. This function should never
58  * be used from generic target-independent code: use one of the
59  * following functions or some other specific check instead.
60  */
61 #define kvm_irqchip_in_kernel() (kvm_kernel_irqchip)
62 
63 /**
64  * kvm_irqchip_is_split:
65  *
66  * Returns: true if the user asked us to split the irqchip
67  * implementation between user and kernel space. The details are
68  * architecture and machine specific. On PC, it means that the PIC,
69  * IOAPIC, and PIT are in user space while the LAPIC is in the kernel.
70  */
71 #define kvm_irqchip_is_split() (kvm_split_irqchip)
72 
73 /**
74  * kvm_async_interrupts_enabled:
75  *
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).
80  */
81 #define kvm_async_interrupts_enabled() (kvm_async_interrupts_allowed)
82 
83 /**
84  * kvm_halt_in_kernel
85  *
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.
88  */
89 #define kvm_halt_in_kernel() (kvm_halt_in_kernel_allowed)
90 
91 /**
92  * kvm_eventfds_enabled:
93  *
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).
97  */
98 #define kvm_eventfds_enabled() (kvm_eventfds_allowed)
99 
100 /**
101  * kvm_irqfds_enabled:
102  *
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).
106  */
107 #define kvm_irqfds_enabled() (kvm_irqfds_allowed)
108 
109 /**
110  * kvm_resamplefds_enabled:
111  *
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).
115  */
116 #define kvm_resamplefds_enabled() (kvm_resamplefds_allowed)
117 
118 /**
119  * kvm_msi_via_irqfd_enabled:
120  *
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.
124  */
125 #define kvm_msi_via_irqfd_enabled() (kvm_msi_via_irqfd_allowed)
126 
127 /**
128  * kvm_gsi_routing_enabled:
129  *
130  * Returns: true if GSI routing is enabled (ie the kernel supports
131  * it and we're running in a configuration that permits it).
132  */
133 #define kvm_gsi_routing_enabled() (kvm_gsi_routing_allowed)
134 
135 /**
136  * kvm_gsi_direct_mapping:
137  *
138  * Returns: true if GSI direct mapping is enabled.
139  */
140 #define kvm_gsi_direct_mapping() (kvm_gsi_direct_mapping)
141 
142 /**
143  * kvm_readonly_mem_enabled:
144  *
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).
147  */
148 #define kvm_readonly_mem_enabled() (kvm_readonly_mem_allowed)
149 
150 /**
151  * kvm_direct_msi_enabled:
152  *
153  * Returns: true if KVM allows direct MSI injection.
154  */
155 #define kvm_direct_msi_enabled() (kvm_direct_msi_allowed)
156 
157 /**
158  * kvm_ioeventfd_any_length_enabled:
159  * Returns: true if KVM allows any length io eventfd.
160  */
161 #define kvm_ioeventfd_any_length_enabled() (kvm_ioeventfd_any_length_allowed)
162 
163 /**
164  * kvm_msi_devid_required:
165  * Returns: true if KVM requires a device id to be provided while
166  * defining an MSI routing entry.
167  */
168 #define kvm_msi_devid_required() (kvm_msi_use_devid)
169 
170 #else
171 
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)
187 
188 #endif  /* CONFIG_KVM_IS_POSSIBLE */
189 
190 struct kvm_run;
191 struct kvm_lapic_state;
192 struct kvm_irq_routing_entry;
193 
194 typedef struct KVMCapabilityInfo {
195     const char *name;
196     int value;
197 } KVMCapabilityInfo;
198 
199 #define KVM_CAP_INFO(CAP) { "KVM_CAP_" stringify(CAP), KVM_CAP_##CAP }
200 #define KVM_CAP_LAST_INFO { NULL, 0 }
201 
202 struct KVMState;
203 
204 #define TYPE_KVM_ACCEL ACCEL_CLASS_NAME("kvm")
205 typedef struct KVMState KVMState;
206 #define KVM_STATE(obj) \
207     OBJECT_CHECK(KVMState, (obj), TYPE_KVM_ACCEL)
208 
209 extern KVMState *kvm_state;
210 typedef struct Notifier Notifier;
211 
212 /* external API */
213 
214 bool kvm_has_free_slot(MachineState *ms);
215 bool kvm_has_sync_mmu(void);
216 int kvm_has_vcpu_events(void);
217 int kvm_has_robust_singlestep(void);
218 int kvm_has_debugregs(void);
219 int kvm_max_nested_state_length(void);
220 int kvm_has_pit_state2(void);
221 int kvm_has_many_ioeventfds(void);
222 int kvm_has_gsi_routing(void);
223 int kvm_has_intx_set_mask(void);
224 
225 int kvm_init_vcpu(CPUState *cpu);
226 int kvm_cpu_exec(CPUState *cpu);
227 int kvm_destroy_vcpu(CPUState *cpu);
228 
229 /**
230  * kvm_arm_supports_user_irq
231  *
232  * Not all KVM implementations support notifications for kernel generated
233  * interrupt events to user space. This function indicates whether the current
234  * KVM implementation does support them.
235  *
236  * Returns: true if KVM supports using kernel generated IRQs from user space
237  */
238 bool kvm_arm_supports_user_irq(void);
239 
240 /**
241  * kvm_memcrypt_enabled - return boolean indicating whether memory encryption
242  *                        is enabled
243  * Returns: 1 memory encryption is enabled
244  *          0 memory encryption is disabled
245  */
246 bool kvm_memcrypt_enabled(void);
247 
248 /**
249  * kvm_memcrypt_encrypt_data: encrypt the memory range
250  *
251  * Return: 1 failed to encrypt the range
252  *         0 succesfully encrypted memory region
253  */
254 int kvm_memcrypt_encrypt_data(uint8_t *ptr, uint64_t len);
255 
256 
257 #ifdef NEED_CPU_H
258 #include "cpu.h"
259 
260 void kvm_flush_coalesced_mmio_buffer(void);
261 
262 int kvm_insert_breakpoint(CPUState *cpu, target_ulong addr,
263                           target_ulong len, int type);
264 int kvm_remove_breakpoint(CPUState *cpu, target_ulong addr,
265                           target_ulong len, int type);
266 void kvm_remove_all_breakpoints(CPUState *cpu);
267 int kvm_update_guest_debug(CPUState *cpu, unsigned long reinject_trap);
268 
269 int kvm_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
270 int kvm_on_sigbus(int code, void *addr);
271 
272 /* interface with exec.c */
273 
274 void phys_mem_set_alloc(void *(*alloc)(size_t, uint64_t *align, bool shared));
275 
276 /* internal API */
277 
278 int kvm_ioctl(KVMState *s, int type, ...);
279 
280 int kvm_vm_ioctl(KVMState *s, int type, ...);
281 
282 int kvm_vcpu_ioctl(CPUState *cpu, int type, ...);
283 
284 /**
285  * kvm_device_ioctl - call an ioctl on a kvm device
286  * @fd: The KVM device file descriptor as returned from KVM_CREATE_DEVICE
287  * @type: The device-ctrl ioctl number
288  *
289  * Returns: -errno on error, nonnegative on success
290  */
291 int kvm_device_ioctl(int fd, int type, ...);
292 
293 /**
294  * kvm_vm_check_attr - check for existence of a specific vm attribute
295  * @s: The KVMState pointer
296  * @group: the group
297  * @attr: the attribute of that group to query for
298  *
299  * Returns: 1 if the attribute exists
300  *          0 if the attribute either does not exist or if the vm device
301  *            interface is unavailable
302  */
303 int kvm_vm_check_attr(KVMState *s, uint32_t group, uint64_t attr);
304 
305 /**
306  * kvm_device_check_attr - check for existence of a specific device attribute
307  * @fd: The device file descriptor
308  * @group: the group
309  * @attr: the attribute of that group to query for
310  *
311  * Returns: 1 if the attribute exists
312  *          0 if the attribute either does not exist or if the vm device
313  *            interface is unavailable
314  */
315 int kvm_device_check_attr(int fd, uint32_t group, uint64_t attr);
316 
317 /**
318  * kvm_device_access - set or get value of a specific device attribute
319  * @fd: The device file descriptor
320  * @group: the group
321  * @attr: the attribute of that group to set or get
322  * @val: pointer to a storage area for the value
323  * @write: true for set and false for get operation
324  * @errp: error object handle
325  *
326  * Returns: 0 on success
327  *          < 0 on error
328  * Use kvm_device_check_attr() in order to check for the availability
329  * of optional attributes.
330  */
331 int kvm_device_access(int fd, int group, uint64_t attr,
332                       void *val, bool write, Error **errp);
333 
334 /**
335  * kvm_create_device - create a KVM device for the device control API
336  * @KVMState: The KVMState pointer
337  * @type: The KVM device type (see Documentation/virtual/kvm/devices in the
338  *        kernel source)
339  * @test: If true, only test if device can be created, but don't actually
340  *        create the device.
341  *
342  * Returns: -errno on error, nonnegative on success: @test ? 0 : device fd;
343  */
344 int kvm_create_device(KVMState *s, uint64_t type, bool test);
345 
346 /**
347  * kvm_device_supported - probe whether KVM supports specific device
348  *
349  * @vmfd: The fd handler for VM
350  * @type: type of device
351  *
352  * @return: true if supported, otherwise false.
353  */
354 bool kvm_device_supported(int vmfd, uint64_t type);
355 
356 /* Arch specific hooks */
357 
358 extern const KVMCapabilityInfo kvm_arch_required_capabilities[];
359 
360 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run);
361 MemTxAttrs kvm_arch_post_run(CPUState *cpu, struct kvm_run *run);
362 
363 int kvm_arch_handle_exit(CPUState *cpu, struct kvm_run *run);
364 
365 int kvm_arch_process_async_events(CPUState *cpu);
366 
367 int kvm_arch_get_registers(CPUState *cpu);
368 
369 /* state subset only touched by the VCPU itself during runtime */
370 #define KVM_PUT_RUNTIME_STATE   1
371 /* state subset modified during VCPU reset */
372 #define KVM_PUT_RESET_STATE     2
373 /* full state set, modified during initialization or on vmload */
374 #define KVM_PUT_FULL_STATE      3
375 
376 int kvm_arch_put_registers(CPUState *cpu, int level);
377 
378 int kvm_arch_init(MachineState *ms, KVMState *s);
379 
380 int kvm_arch_init_vcpu(CPUState *cpu);
381 int kvm_arch_destroy_vcpu(CPUState *cpu);
382 
383 bool kvm_vcpu_id_is_valid(int vcpu_id);
384 
385 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */
386 unsigned long kvm_arch_vcpu_id(CPUState *cpu);
387 
388 #ifdef KVM_HAVE_MCE_INJECTION
389 void kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
390 #endif
391 
392 void kvm_arch_init_irq_routing(KVMState *s);
393 
394 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry *route,
395                              uint64_t address, uint32_t data, PCIDevice *dev);
396 
397 /* Notify arch about newly added MSI routes */
398 int kvm_arch_add_msi_route_post(struct kvm_irq_routing_entry *route,
399                                 int vector, PCIDevice *dev);
400 /* Notify arch about released MSI routes */
401 int kvm_arch_release_virq_post(int virq);
402 
403 int kvm_arch_msi_data_to_gsi(uint32_t data);
404 
405 int kvm_set_irq(KVMState *s, int irq, int level);
406 int kvm_irqchip_send_msi(KVMState *s, MSIMessage msg);
407 
408 void kvm_irqchip_add_irq_route(KVMState *s, int gsi, int irqchip, int pin);
409 
410 void kvm_irqchip_add_change_notifier(Notifier *n);
411 void kvm_irqchip_remove_change_notifier(Notifier *n);
412 void kvm_irqchip_change_notify(void);
413 
414 void kvm_get_apic_state(DeviceState *d, struct kvm_lapic_state *kapic);
415 
416 struct kvm_guest_debug;
417 struct kvm_debug_exit_arch;
418 
419 struct kvm_sw_breakpoint {
420     target_ulong pc;
421     target_ulong saved_insn;
422     int use_count;
423     QTAILQ_ENTRY(kvm_sw_breakpoint) entry;
424 };
425 
426 struct kvm_sw_breakpoint *kvm_find_sw_breakpoint(CPUState *cpu,
427                                                  target_ulong pc);
428 
429 int kvm_sw_breakpoints_active(CPUState *cpu);
430 
431 int kvm_arch_insert_sw_breakpoint(CPUState *cpu,
432                                   struct kvm_sw_breakpoint *bp);
433 int kvm_arch_remove_sw_breakpoint(CPUState *cpu,
434                                   struct kvm_sw_breakpoint *bp);
435 int kvm_arch_insert_hw_breakpoint(target_ulong addr,
436                                   target_ulong len, int type);
437 int kvm_arch_remove_hw_breakpoint(target_ulong addr,
438                                   target_ulong len, int type);
439 void kvm_arch_remove_all_hw_breakpoints(void);
440 
441 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg);
442 
443 bool kvm_arch_stop_on_emulation_error(CPUState *cpu);
444 
445 int kvm_check_extension(KVMState *s, unsigned int extension);
446 
447 int kvm_vm_check_extension(KVMState *s, unsigned int extension);
448 
449 #define kvm_vm_enable_cap(s, capability, cap_flags, ...)             \
450     ({                                                               \
451         struct kvm_enable_cap cap = {                                \
452             .cap = capability,                                       \
453             .flags = cap_flags,                                      \
454         };                                                           \
455         uint64_t args_tmp[] = { __VA_ARGS__ };                       \
456         size_t n = MIN(ARRAY_SIZE(args_tmp), ARRAY_SIZE(cap.args));  \
457         memcpy(cap.args, args_tmp, n * sizeof(cap.args[0]));         \
458         kvm_vm_ioctl(s, KVM_ENABLE_CAP, &cap);                       \
459     })
460 
461 #define kvm_vcpu_enable_cap(cpu, capability, cap_flags, ...)         \
462     ({                                                               \
463         struct kvm_enable_cap cap = {                                \
464             .cap = capability,                                       \
465             .flags = cap_flags,                                      \
466         };                                                           \
467         uint64_t args_tmp[] = { __VA_ARGS__ };                       \
468         size_t n = MIN(ARRAY_SIZE(args_tmp), ARRAY_SIZE(cap.args));  \
469         memcpy(cap.args, args_tmp, n * sizeof(cap.args[0]));         \
470         kvm_vcpu_ioctl(cpu, KVM_ENABLE_CAP, &cap);                   \
471     })
472 
473 uint32_t kvm_arch_get_supported_cpuid(KVMState *env, uint32_t function,
474                                       uint32_t index, int reg);
475 uint64_t kvm_arch_get_supported_msr_feature(KVMState *s, uint32_t index);
476 
477 
478 void kvm_set_sigmask_len(KVMState *s, unsigned int sigmask_len);
479 
480 #if !defined(CONFIG_USER_ONLY)
481 int kvm_physical_memory_addr_from_host(KVMState *s, void *ram_addr,
482                                        hwaddr *phys_addr);
483 #endif
484 
485 #endif /* NEED_CPU_H */
486 
487 void kvm_cpu_synchronize_state(CPUState *cpu);
488 void kvm_cpu_synchronize_post_reset(CPUState *cpu);
489 void kvm_cpu_synchronize_post_init(CPUState *cpu);
490 void kvm_cpu_synchronize_pre_loadvm(CPUState *cpu);
491 
492 void kvm_init_cpu_signals(CPUState *cpu);
493 
494 /**
495  * kvm_irqchip_add_msi_route - Add MSI route for specific vector
496  * @s:      KVM state
497  * @vector: which vector to add. This can be either MSI/MSIX
498  *          vector. The function will automatically detect whether
499  *          MSI/MSIX is enabled, and fetch corresponding MSI
500  *          message.
501  * @dev:    Owner PCI device to add the route. If @dev is specified
502  *          as @NULL, an empty MSI message will be inited.
503  * @return: virq (>=0) when success, errno (<0) when failed.
504  */
505 int kvm_irqchip_add_msi_route(KVMState *s, int vector, PCIDevice *dev);
506 int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg,
507                                  PCIDevice *dev);
508 void kvm_irqchip_commit_routes(KVMState *s);
509 void kvm_irqchip_release_virq(KVMState *s, int virq);
510 
511 int kvm_irqchip_add_adapter_route(KVMState *s, AdapterInfo *adapter);
512 int kvm_irqchip_add_hv_sint_route(KVMState *s, uint32_t vcpu, uint32_t sint);
513 
514 int kvm_irqchip_add_irqfd_notifier_gsi(KVMState *s, EventNotifier *n,
515                                        EventNotifier *rn, int virq);
516 int kvm_irqchip_remove_irqfd_notifier_gsi(KVMState *s, EventNotifier *n,
517                                           int virq);
518 int kvm_irqchip_add_irqfd_notifier(KVMState *s, EventNotifier *n,
519                                    EventNotifier *rn, qemu_irq irq);
520 int kvm_irqchip_remove_irqfd_notifier(KVMState *s, EventNotifier *n,
521                                       qemu_irq irq);
522 void kvm_irqchip_set_qemuirq_gsi(KVMState *s, qemu_irq irq, int gsi);
523 void kvm_pc_setup_irq_routing(bool pci_enabled);
524 void kvm_init_irq_routing(KVMState *s);
525 
526 bool kvm_kernel_irqchip_allowed(void);
527 bool kvm_kernel_irqchip_required(void);
528 bool kvm_kernel_irqchip_split(void);
529 
530 /**
531  * kvm_arch_irqchip_create:
532  * @KVMState: The KVMState pointer
533  *
534  * Allow architectures to create an in-kernel irq chip themselves.
535  *
536  * Returns: < 0: error
537  *            0: irq chip was not created
538  *          > 0: irq chip was created
539  */
540 int kvm_arch_irqchip_create(KVMState *s);
541 
542 /**
543  * kvm_set_one_reg - set a register value in KVM via KVM_SET_ONE_REG ioctl
544  * @id: The register ID
545  * @source: The pointer to the value to be set. It must point to a variable
546  *          of the correct type/size for the register being accessed.
547  *
548  * Returns: 0 on success, or a negative errno on failure.
549  */
550 int kvm_set_one_reg(CPUState *cs, uint64_t id, void *source);
551 
552 /**
553  * kvm_get_one_reg - get a register value from KVM via KVM_GET_ONE_REG ioctl
554  * @id: The register ID
555  * @target: The pointer where the value is to be stored. It must point to a
556  *          variable of the correct type/size for the register being accessed.
557  *
558  * Returns: 0 on success, or a negative errno on failure.
559  */
560 int kvm_get_one_reg(CPUState *cs, uint64_t id, void *target);
561 struct ppc_radix_page_info *kvm_get_radix_page_info(void);
562 int kvm_get_max_memslots(void);
563 
564 /* Notify resamplefd for EOI of specific interrupts. */
565 void kvm_resample_fd_notify(int gsi);
566 
567 #endif
568