xref: /openbmc/qemu/include/sysemu/kvm.h (revision e6546342)
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 #include "qom/object.h"
22 
23 #ifdef NEED_CPU_H
24 # ifdef CONFIG_KVM
25 #  include <linux/kvm.h>
26 #  define CONFIG_KVM_IS_POSSIBLE
27 # endif
28 #else
29 # define CONFIG_KVM_IS_POSSIBLE
30 #endif
31 
32 #ifdef CONFIG_KVM_IS_POSSIBLE
33 
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 
50 #define kvm_enabled()           (kvm_allowed)
51 /**
52  * kvm_irqchip_in_kernel:
53  *
54  * Returns: true if the user asked us to create an in-kernel
55  * irqchip via the "kernel_irqchip=on" machine option.
56  * What this actually means is architecture and machine model
57  * specific: on PC, for instance, it means that the LAPIC,
58  * IOAPIC and PIT are all in kernel. This function should never
59  * be used from generic target-independent code: use one of the
60  * following functions or some other specific check instead.
61  */
62 #define kvm_irqchip_in_kernel() (kvm_kernel_irqchip)
63 
64 /**
65  * kvm_irqchip_is_split:
66  *
67  * Returns: true if the user asked us to split the irqchip
68  * implementation between user and kernel space. The details are
69  * architecture and machine specific. On PC, it means that the PIC,
70  * IOAPIC, and PIT are in user space while the LAPIC is in the kernel.
71  */
72 #define kvm_irqchip_is_split() (kvm_split_irqchip)
73 
74 /**
75  * kvm_async_interrupts_enabled:
76  *
77  * Returns: true if we can deliver interrupts to KVM
78  * asynchronously (ie by ioctl from any thread at any time)
79  * rather than having to do interrupt delivery synchronously
80  * (where the vcpu must be stopped at a suitable point first).
81  */
82 #define kvm_async_interrupts_enabled() (kvm_async_interrupts_allowed)
83 
84 /**
85  * kvm_halt_in_kernel
86  *
87  * Returns: true if halted cpus should still get a KVM_RUN ioctl to run
88  * inside of kernel space. This only works if MP state is implemented.
89  */
90 #define kvm_halt_in_kernel() (kvm_halt_in_kernel_allowed)
91 
92 /**
93  * kvm_eventfds_enabled:
94  *
95  * Returns: true if we can use eventfds to receive notifications
96  * from a KVM CPU (ie the kernel supports eventds and we are running
97  * with a configuration where it is meaningful to use them).
98  */
99 #define kvm_eventfds_enabled() (kvm_eventfds_allowed)
100 
101 /**
102  * kvm_irqfds_enabled:
103  *
104  * Returns: true if we can use irqfds to inject interrupts into
105  * a KVM CPU (ie the kernel supports irqfds and we are running
106  * with a configuration where it is meaningful to use them).
107  */
108 #define kvm_irqfds_enabled() (kvm_irqfds_allowed)
109 
110 /**
111  * kvm_resamplefds_enabled:
112  *
113  * Returns: true if we can use resamplefds to inject interrupts into
114  * a KVM CPU (ie the kernel supports resamplefds and we are running
115  * with a configuration where it is meaningful to use them).
116  */
117 #define kvm_resamplefds_enabled() (kvm_resamplefds_allowed)
118 
119 /**
120  * kvm_msi_via_irqfd_enabled:
121  *
122  * Returns: true if we can route a PCI MSI (Message Signaled Interrupt)
123  * to a KVM CPU via an irqfd. This requires that the kernel supports
124  * this and that we're running in a configuration that permits it.
125  */
126 #define kvm_msi_via_irqfd_enabled() (kvm_msi_via_irqfd_allowed)
127 
128 /**
129  * kvm_gsi_routing_enabled:
130  *
131  * Returns: true if GSI routing is enabled (ie the kernel supports
132  * it and we're running in a configuration that permits it).
133  */
134 #define kvm_gsi_routing_enabled() (kvm_gsi_routing_allowed)
135 
136 /**
137  * kvm_gsi_direct_mapping:
138  *
139  * Returns: true if GSI direct mapping is enabled.
140  */
141 #define kvm_gsi_direct_mapping() (kvm_gsi_direct_mapping)
142 
143 /**
144  * kvm_readonly_mem_enabled:
145  *
146  * Returns: true if KVM readonly memory is enabled (ie the kernel
147  * supports it and we're running in a configuration that permits it).
148  */
149 #define kvm_readonly_mem_enabled() (kvm_readonly_mem_allowed)
150 
151 /**
152  * kvm_direct_msi_enabled:
153  *
154  * Returns: true if KVM allows direct MSI injection.
155  */
156 #define kvm_direct_msi_enabled() (kvm_direct_msi_allowed)
157 
158 /**
159  * kvm_ioeventfd_any_length_enabled:
160  * Returns: true if KVM allows any length io eventfd.
161  */
162 #define kvm_ioeventfd_any_length_enabled() (kvm_ioeventfd_any_length_allowed)
163 
164 /**
165  * kvm_msi_devid_required:
166  * Returns: true if KVM requires a device id to be provided while
167  * defining an MSI routing entry.
168  */
169 #define kvm_msi_devid_required() (kvm_msi_use_devid)
170 
171 #else
172 
173 #define kvm_enabled()           (0)
174 #define kvm_irqchip_in_kernel() (false)
175 #define kvm_irqchip_is_split() (false)
176 #define kvm_async_interrupts_enabled() (false)
177 #define kvm_halt_in_kernel() (false)
178 #define kvm_eventfds_enabled() (false)
179 #define kvm_irqfds_enabled() (false)
180 #define kvm_resamplefds_enabled() (false)
181 #define kvm_msi_via_irqfd_enabled() (false)
182 #define kvm_gsi_routing_allowed() (false)
183 #define kvm_gsi_direct_mapping() (false)
184 #define kvm_readonly_mem_enabled() (false)
185 #define kvm_direct_msi_enabled() (false)
186 #define kvm_ioeventfd_any_length_enabled() (false)
187 #define kvm_msi_devid_required() (false)
188 
189 #endif  /* CONFIG_KVM_IS_POSSIBLE */
190 
191 struct kvm_run;
192 struct kvm_lapic_state;
193 struct kvm_irq_routing_entry;
194 
195 typedef struct KVMCapabilityInfo {
196     const char *name;
197     int value;
198 } KVMCapabilityInfo;
199 
200 #define KVM_CAP_INFO(CAP) { "KVM_CAP_" stringify(CAP), KVM_CAP_##CAP }
201 #define KVM_CAP_LAST_INFO { NULL, 0 }
202 
203 struct KVMState;
204 
205 #define TYPE_KVM_ACCEL ACCEL_CLASS_NAME("kvm")
206 typedef struct KVMState KVMState;
207 DECLARE_INSTANCE_CHECKER(KVMState, KVM_STATE,
208                          TYPE_KVM_ACCEL)
209 
210 extern KVMState *kvm_state;
211 typedef struct Notifier Notifier;
212 
213 /* external API */
214 
215 bool kvm_has_free_slot(MachineState *ms);
216 bool kvm_has_sync_mmu(void);
217 int kvm_has_vcpu_events(void);
218 int kvm_has_robust_singlestep(void);
219 int kvm_has_debugregs(void);
220 int kvm_max_nested_state_length(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);
225 
226 int kvm_init_vcpu(CPUState *cpu);
227 int kvm_cpu_exec(CPUState *cpu);
228 int kvm_destroy_vcpu(CPUState *cpu);
229 
230 /**
231  * kvm_arm_supports_user_irq
232  *
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.
236  *
237  * Returns: true if KVM supports using kernel generated IRQs from user space
238  */
239 bool kvm_arm_supports_user_irq(void);
240 
241 /**
242  * kvm_memcrypt_enabled - return boolean indicating whether memory encryption
243  *                        is enabled
244  * Returns: 1 memory encryption is enabled
245  *          0 memory encryption is disabled
246  */
247 bool kvm_memcrypt_enabled(void);
248 
249 /**
250  * kvm_memcrypt_encrypt_data: encrypt the memory range
251  *
252  * Return: 1 failed to encrypt the range
253  *         0 succesfully encrypted memory region
254  */
255 int kvm_memcrypt_encrypt_data(uint8_t *ptr, uint64_t len);
256 
257 
258 #ifdef NEED_CPU_H
259 #include "cpu.h"
260 
261 void kvm_flush_coalesced_mmio_buffer(void);
262 
263 int kvm_insert_breakpoint(CPUState *cpu, target_ulong addr,
264                           target_ulong len, int type);
265 int kvm_remove_breakpoint(CPUState *cpu, target_ulong addr,
266                           target_ulong len, int type);
267 void kvm_remove_all_breakpoints(CPUState *cpu);
268 int kvm_update_guest_debug(CPUState *cpu, unsigned long reinject_trap);
269 
270 int kvm_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
271 int kvm_on_sigbus(int code, void *addr);
272 
273 /* interface with exec.c */
274 
275 void phys_mem_set_alloc(void *(*alloc)(size_t, uint64_t *align, bool shared));
276 
277 /* internal API */
278 
279 int kvm_ioctl(KVMState *s, int type, ...);
280 
281 int kvm_vm_ioctl(KVMState *s, int type, ...);
282 
283 int kvm_vcpu_ioctl(CPUState *cpu, int type, ...);
284 
285 /**
286  * kvm_device_ioctl - call an ioctl on a kvm device
287  * @fd: The KVM device file descriptor as returned from KVM_CREATE_DEVICE
288  * @type: The device-ctrl ioctl number
289  *
290  * Returns: -errno on error, nonnegative on success
291  */
292 int kvm_device_ioctl(int fd, int type, ...);
293 
294 /**
295  * kvm_vm_check_attr - check for existence of a specific vm attribute
296  * @s: The KVMState pointer
297  * @group: the group
298  * @attr: the attribute of that group to query for
299  *
300  * Returns: 1 if the attribute exists
301  *          0 if the attribute either does not exist or if the vm device
302  *            interface is unavailable
303  */
304 int kvm_vm_check_attr(KVMState *s, uint32_t group, uint64_t attr);
305 
306 /**
307  * kvm_device_check_attr - check for existence of a specific device attribute
308  * @fd: The device file descriptor
309  * @group: the group
310  * @attr: the attribute of that group to query for
311  *
312  * Returns: 1 if the attribute exists
313  *          0 if the attribute either does not exist or if the vm device
314  *            interface is unavailable
315  */
316 int kvm_device_check_attr(int fd, uint32_t group, uint64_t attr);
317 
318 /**
319  * kvm_device_access - set or get value of a specific device attribute
320  * @fd: The device file descriptor
321  * @group: the group
322  * @attr: the attribute of that group to set or get
323  * @val: pointer to a storage area for the value
324  * @write: true for set and false for get operation
325  * @errp: error object handle
326  *
327  * Returns: 0 on success
328  *          < 0 on error
329  * Use kvm_device_check_attr() in order to check for the availability
330  * of optional attributes.
331  */
332 int kvm_device_access(int fd, int group, uint64_t attr,
333                       void *val, bool write, Error **errp);
334 
335 /**
336  * kvm_create_device - create a KVM device for the device control API
337  * @KVMState: The KVMState pointer
338  * @type: The KVM device type (see Documentation/virtual/kvm/devices in the
339  *        kernel source)
340  * @test: If true, only test if device can be created, but don't actually
341  *        create the device.
342  *
343  * Returns: -errno on error, nonnegative on success: @test ? 0 : device fd;
344  */
345 int kvm_create_device(KVMState *s, uint64_t type, bool test);
346 
347 /**
348  * kvm_device_supported - probe whether KVM supports specific device
349  *
350  * @vmfd: The fd handler for VM
351  * @type: type of device
352  *
353  * @return: true if supported, otherwise false.
354  */
355 bool kvm_device_supported(int vmfd, uint64_t type);
356 
357 /* Arch specific hooks */
358 
359 extern const KVMCapabilityInfo kvm_arch_required_capabilities[];
360 
361 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run);
362 MemTxAttrs kvm_arch_post_run(CPUState *cpu, struct kvm_run *run);
363 
364 int kvm_arch_handle_exit(CPUState *cpu, struct kvm_run *run);
365 
366 int kvm_arch_process_async_events(CPUState *cpu);
367 
368 int kvm_arch_get_registers(CPUState *cpu);
369 
370 /* state subset only touched by the VCPU itself during runtime */
371 #define KVM_PUT_RUNTIME_STATE   1
372 /* state subset modified during VCPU reset */
373 #define KVM_PUT_RESET_STATE     2
374 /* full state set, modified during initialization or on vmload */
375 #define KVM_PUT_FULL_STATE      3
376 
377 int kvm_arch_put_registers(CPUState *cpu, int level);
378 
379 int kvm_arch_init(MachineState *ms, KVMState *s);
380 
381 int kvm_arch_init_vcpu(CPUState *cpu);
382 int kvm_arch_destroy_vcpu(CPUState *cpu);
383 
384 bool kvm_vcpu_id_is_valid(int vcpu_id);
385 
386 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */
387 unsigned long kvm_arch_vcpu_id(CPUState *cpu);
388 
389 #ifdef KVM_HAVE_MCE_INJECTION
390 void kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
391 #endif
392 
393 void kvm_arch_init_irq_routing(KVMState *s);
394 
395 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry *route,
396                              uint64_t address, uint32_t data, PCIDevice *dev);
397 
398 /* Notify arch about newly added MSI routes */
399 int kvm_arch_add_msi_route_post(struct kvm_irq_routing_entry *route,
400                                 int vector, PCIDevice *dev);
401 /* Notify arch about released MSI routes */
402 int kvm_arch_release_virq_post(int virq);
403 
404 int kvm_arch_msi_data_to_gsi(uint32_t data);
405 
406 int kvm_set_irq(KVMState *s, int irq, int level);
407 int kvm_irqchip_send_msi(KVMState *s, MSIMessage msg);
408 
409 void kvm_irqchip_add_irq_route(KVMState *s, int gsi, int irqchip, int pin);
410 
411 void kvm_irqchip_add_change_notifier(Notifier *n);
412 void kvm_irqchip_remove_change_notifier(Notifier *n);
413 void kvm_irqchip_change_notify(void);
414 
415 void kvm_get_apic_state(DeviceState *d, struct kvm_lapic_state *kapic);
416 
417 struct kvm_guest_debug;
418 struct kvm_debug_exit_arch;
419 
420 struct kvm_sw_breakpoint {
421     target_ulong pc;
422     target_ulong saved_insn;
423     int use_count;
424     QTAILQ_ENTRY(kvm_sw_breakpoint) entry;
425 };
426 
427 struct kvm_sw_breakpoint *kvm_find_sw_breakpoint(CPUState *cpu,
428                                                  target_ulong pc);
429 
430 int kvm_sw_breakpoints_active(CPUState *cpu);
431 
432 int kvm_arch_insert_sw_breakpoint(CPUState *cpu,
433                                   struct kvm_sw_breakpoint *bp);
434 int kvm_arch_remove_sw_breakpoint(CPUState *cpu,
435                                   struct kvm_sw_breakpoint *bp);
436 int kvm_arch_insert_hw_breakpoint(target_ulong addr,
437                                   target_ulong len, int type);
438 int kvm_arch_remove_hw_breakpoint(target_ulong addr,
439                                   target_ulong len, int type);
440 void kvm_arch_remove_all_hw_breakpoints(void);
441 
442 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg);
443 
444 bool kvm_arch_stop_on_emulation_error(CPUState *cpu);
445 
446 int kvm_check_extension(KVMState *s, unsigned int extension);
447 
448 int kvm_vm_check_extension(KVMState *s, unsigned int extension);
449 
450 #define kvm_vm_enable_cap(s, capability, cap_flags, ...)             \
451     ({                                                               \
452         struct kvm_enable_cap cap = {                                \
453             .cap = capability,                                       \
454             .flags = cap_flags,                                      \
455         };                                                           \
456         uint64_t args_tmp[] = { __VA_ARGS__ };                       \
457         size_t n = MIN(ARRAY_SIZE(args_tmp), ARRAY_SIZE(cap.args));  \
458         memcpy(cap.args, args_tmp, n * sizeof(cap.args[0]));         \
459         kvm_vm_ioctl(s, KVM_ENABLE_CAP, &cap);                       \
460     })
461 
462 #define kvm_vcpu_enable_cap(cpu, capability, cap_flags, ...)         \
463     ({                                                               \
464         struct kvm_enable_cap cap = {                                \
465             .cap = capability,                                       \
466             .flags = cap_flags,                                      \
467         };                                                           \
468         uint64_t args_tmp[] = { __VA_ARGS__ };                       \
469         size_t n = MIN(ARRAY_SIZE(args_tmp), ARRAY_SIZE(cap.args));  \
470         memcpy(cap.args, args_tmp, n * sizeof(cap.args[0]));         \
471         kvm_vcpu_ioctl(cpu, KVM_ENABLE_CAP, &cap);                   \
472     })
473 
474 uint32_t kvm_arch_get_supported_cpuid(KVMState *env, uint32_t function,
475                                       uint32_t index, int reg);
476 uint64_t kvm_arch_get_supported_msr_feature(KVMState *s, uint32_t index);
477 
478 
479 void kvm_set_sigmask_len(KVMState *s, unsigned int sigmask_len);
480 
481 #if !defined(CONFIG_USER_ONLY)
482 int kvm_physical_memory_addr_from_host(KVMState *s, void *ram_addr,
483                                        hwaddr *phys_addr);
484 #endif
485 
486 #endif /* NEED_CPU_H */
487 
488 void kvm_cpu_synchronize_state(CPUState *cpu);
489 void kvm_cpu_synchronize_post_reset(CPUState *cpu);
490 void kvm_cpu_synchronize_post_init(CPUState *cpu);
491 void kvm_cpu_synchronize_pre_loadvm(CPUState *cpu);
492 
493 void kvm_init_cpu_signals(CPUState *cpu);
494 
495 /**
496  * kvm_irqchip_add_msi_route - Add MSI route for specific vector
497  * @s:      KVM state
498  * @vector: which vector to add. This can be either MSI/MSIX
499  *          vector. The function will automatically detect whether
500  *          MSI/MSIX is enabled, and fetch corresponding MSI
501  *          message.
502  * @dev:    Owner PCI device to add the route. If @dev is specified
503  *          as @NULL, an empty MSI message will be inited.
504  * @return: virq (>=0) when success, errno (<0) when failed.
505  */
506 int kvm_irqchip_add_msi_route(KVMState *s, int vector, PCIDevice *dev);
507 int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg,
508                                  PCIDevice *dev);
509 void kvm_irqchip_commit_routes(KVMState *s);
510 void kvm_irqchip_release_virq(KVMState *s, int virq);
511 
512 int kvm_irqchip_add_adapter_route(KVMState *s, AdapterInfo *adapter);
513 int kvm_irqchip_add_hv_sint_route(KVMState *s, uint32_t vcpu, uint32_t sint);
514 
515 int kvm_irqchip_add_irqfd_notifier_gsi(KVMState *s, EventNotifier *n,
516                                        EventNotifier *rn, int virq);
517 int kvm_irqchip_remove_irqfd_notifier_gsi(KVMState *s, EventNotifier *n,
518                                           int virq);
519 int kvm_irqchip_add_irqfd_notifier(KVMState *s, EventNotifier *n,
520                                    EventNotifier *rn, qemu_irq irq);
521 int kvm_irqchip_remove_irqfd_notifier(KVMState *s, EventNotifier *n,
522                                       qemu_irq irq);
523 void kvm_irqchip_set_qemuirq_gsi(KVMState *s, qemu_irq irq, int gsi);
524 void kvm_pc_setup_irq_routing(bool pci_enabled);
525 void kvm_init_irq_routing(KVMState *s);
526 
527 bool kvm_kernel_irqchip_allowed(void);
528 bool kvm_kernel_irqchip_required(void);
529 bool kvm_kernel_irqchip_split(void);
530 
531 /**
532  * kvm_arch_irqchip_create:
533  * @KVMState: The KVMState pointer
534  *
535  * Allow architectures to create an in-kernel irq chip themselves.
536  *
537  * Returns: < 0: error
538  *            0: irq chip was not created
539  *          > 0: irq chip was created
540  */
541 int kvm_arch_irqchip_create(KVMState *s);
542 
543 /**
544  * kvm_set_one_reg - set a register value in KVM via KVM_SET_ONE_REG ioctl
545  * @id: The register ID
546  * @source: The pointer to the value to be set. It must point to a variable
547  *          of the correct type/size for the register being accessed.
548  *
549  * Returns: 0 on success, or a negative errno on failure.
550  */
551 int kvm_set_one_reg(CPUState *cs, uint64_t id, void *source);
552 
553 /**
554  * kvm_get_one_reg - get a register value from KVM via KVM_GET_ONE_REG ioctl
555  * @id: The register ID
556  * @target: The pointer where the value is to be stored. It must point to a
557  *          variable of the correct type/size for the register being accessed.
558  *
559  * Returns: 0 on success, or a negative errno on failure.
560  */
561 int kvm_get_one_reg(CPUState *cs, uint64_t id, void *target);
562 struct ppc_radix_page_info *kvm_get_radix_page_info(void);
563 int kvm_get_max_memslots(void);
564 
565 /* Notify resamplefd for EOI of specific interrupts. */
566 void kvm_resample_fd_notify(int gsi);
567 
568 #endif
569