xref: /openbmc/qemu/include/sysemu/kvm.h (revision e09484ef)
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 "qom/cpu.h"
19 #include "exec/memattrs.h"
20 #include "hw/irq.h"
21 
22 #ifdef CONFIG_KVM
23 #include <linux/kvm.h>
24 #include <linux/kvm_para.h>
25 #else
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.
29  */
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
40 #endif
41 
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 
57 #if defined CONFIG_KVM || !defined NEED_CPU_H
58 #define kvm_enabled()           (kvm_allowed)
59 /**
60  * kvm_irqchip_in_kernel:
61  *
62  * Returns: true if the user asked us to create an in-kernel
63  * irqchip via the "kernel_irqchip=on" machine option.
64  * What this actually means is architecture and machine model
65  * specific: on PC, for instance, it means that the LAPIC,
66  * IOAPIC and PIT are all in kernel. This function should never
67  * be used from generic target-independent code: use one of the
68  * following functions or some other specific check instead.
69  */
70 #define kvm_irqchip_in_kernel() (kvm_kernel_irqchip)
71 
72 /**
73  * kvm_irqchip_is_split:
74  *
75  * Returns: true if the user asked us to split the irqchip
76  * implementation between user and kernel space. The details are
77  * architecture and machine specific. On PC, it means that the PIC,
78  * IOAPIC, and PIT are in user space while the LAPIC is in the kernel.
79  */
80 #define kvm_irqchip_is_split() (kvm_split_irqchip)
81 
82 /**
83  * kvm_async_interrupts_enabled:
84  *
85  * Returns: true if we can deliver interrupts to KVM
86  * asynchronously (ie by ioctl from any thread at any time)
87  * rather than having to do interrupt delivery synchronously
88  * (where the vcpu must be stopped at a suitable point first).
89  */
90 #define kvm_async_interrupts_enabled() (kvm_async_interrupts_allowed)
91 
92 /**
93  * kvm_halt_in_kernel
94  *
95  * Returns: true if halted cpus should still get a KVM_RUN ioctl to run
96  * inside of kernel space. This only works if MP state is implemented.
97  */
98 #define kvm_halt_in_kernel() (kvm_halt_in_kernel_allowed)
99 
100 /**
101  * kvm_eventfds_enabled:
102  *
103  * Returns: true if we can use eventfds to receive notifications
104  * from a KVM CPU (ie the kernel supports eventds and we are running
105  * with a configuration where it is meaningful to use them).
106  */
107 #define kvm_eventfds_enabled() (kvm_eventfds_allowed)
108 
109 /**
110  * kvm_irqfds_enabled:
111  *
112  * Returns: true if we can use irqfds to inject interrupts into
113  * a KVM CPU (ie the kernel supports irqfds and we are running
114  * with a configuration where it is meaningful to use them).
115  */
116 #define kvm_irqfds_enabled() (kvm_irqfds_allowed)
117 
118 /**
119  * kvm_resamplefds_enabled:
120  *
121  * Returns: true if we can use resamplefds to inject interrupts into
122  * a KVM CPU (ie the kernel supports resamplefds and we are running
123  * with a configuration where it is meaningful to use them).
124  */
125 #define kvm_resamplefds_enabled() (kvm_resamplefds_allowed)
126 
127 /**
128  * kvm_msi_via_irqfd_enabled:
129  *
130  * Returns: true if we can route a PCI MSI (Message Signaled Interrupt)
131  * to a KVM CPU via an irqfd. This requires that the kernel supports
132  * this and that we're running in a configuration that permits it.
133  */
134 #define kvm_msi_via_irqfd_enabled() (kvm_msi_via_irqfd_allowed)
135 
136 /**
137  * kvm_gsi_routing_enabled:
138  *
139  * Returns: true if GSI routing is enabled (ie the kernel supports
140  * it and we're running in a configuration that permits it).
141  */
142 #define kvm_gsi_routing_enabled() (kvm_gsi_routing_allowed)
143 
144 /**
145  * kvm_gsi_direct_mapping:
146  *
147  * Returns: true if GSI direct mapping is enabled.
148  */
149 #define kvm_gsi_direct_mapping() (kvm_gsi_direct_mapping)
150 
151 /**
152  * kvm_readonly_mem_enabled:
153  *
154  * Returns: true if KVM readonly memory is enabled (ie the kernel
155  * supports it and we're running in a configuration that permits it).
156  */
157 #define kvm_readonly_mem_enabled() (kvm_readonly_mem_allowed)
158 
159 /**
160  * kvm_direct_msi_enabled:
161  *
162  * Returns: true if KVM allows direct MSI injection.
163  */
164 #define kvm_direct_msi_enabled() (kvm_direct_msi_allowed)
165 
166 /**
167  * kvm_ioeventfd_any_length_enabled:
168  * Returns: true if KVM allows any length io eventfd.
169  */
170 #define kvm_ioeventfd_any_length_enabled() (kvm_ioeventfd_any_length_allowed)
171 
172 #else
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 #endif
188 
189 struct kvm_run;
190 struct kvm_lapic_state;
191 struct kvm_irq_routing_entry;
192 
193 typedef struct KVMCapabilityInfo {
194     const char *name;
195     int value;
196 } KVMCapabilityInfo;
197 
198 #define KVM_CAP_INFO(CAP) { "KVM_CAP_" stringify(CAP), KVM_CAP_##CAP }
199 #define KVM_CAP_LAST_INFO { NULL, 0 }
200 
201 struct KVMState;
202 typedef struct KVMState KVMState;
203 extern KVMState *kvm_state;
204 
205 /* external API */
206 
207 bool kvm_has_free_slot(MachineState *ms);
208 int kvm_has_sync_mmu(void);
209 int kvm_has_vcpu_events(void);
210 int kvm_has_robust_singlestep(void);
211 int kvm_has_debugregs(void);
212 int kvm_has_pit_state2(void);
213 int kvm_has_many_ioeventfds(void);
214 int kvm_has_gsi_routing(void);
215 int kvm_has_intx_set_mask(void);
216 
217 int kvm_init_vcpu(CPUState *cpu);
218 int kvm_cpu_exec(CPUState *cpu);
219 int kvm_destroy_vcpu(CPUState *cpu);
220 
221 #ifdef NEED_CPU_H
222 #include "cpu.h"
223 
224 void kvm_setup_guest_memory(void *start, size_t size);
225 void kvm_flush_coalesced_mmio_buffer(void);
226 
227 int kvm_insert_breakpoint(CPUState *cpu, target_ulong addr,
228                           target_ulong len, int type);
229 int kvm_remove_breakpoint(CPUState *cpu, target_ulong addr,
230                           target_ulong len, int type);
231 void kvm_remove_all_breakpoints(CPUState *cpu);
232 int kvm_update_guest_debug(CPUState *cpu, unsigned long reinject_trap);
233 #ifndef _WIN32
234 int kvm_set_signal_mask(CPUState *cpu, const sigset_t *sigset);
235 #endif
236 
237 int kvm_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
238 int kvm_on_sigbus(int code, void *addr);
239 
240 /* interface with exec.c */
241 
242 void phys_mem_set_alloc(void *(*alloc)(size_t, uint64_t *align));
243 
244 /* internal API */
245 
246 int kvm_ioctl(KVMState *s, int type, ...);
247 
248 int kvm_vm_ioctl(KVMState *s, int type, ...);
249 
250 int kvm_vcpu_ioctl(CPUState *cpu, int type, ...);
251 
252 /**
253  * kvm_device_ioctl - call an ioctl on a kvm device
254  * @fd: The KVM device file descriptor as returned from KVM_CREATE_DEVICE
255  * @type: The device-ctrl ioctl number
256  *
257  * Returns: -errno on error, nonnegative on success
258  */
259 int kvm_device_ioctl(int fd, int type, ...);
260 
261 /**
262  * kvm_vm_check_attr - check for existence of a specific vm attribute
263  * @s: The KVMState pointer
264  * @group: the group
265  * @attr: the attribute of that group to query for
266  *
267  * Returns: 1 if the attribute exists
268  *          0 if the attribute either does not exist or if the vm device
269  *            interface is unavailable
270  */
271 int kvm_vm_check_attr(KVMState *s, uint32_t group, uint64_t attr);
272 
273 /**
274  * kvm_device_check_attr - check for existence of a specific device attribute
275  * @fd: The device file descriptor
276  * @group: the group
277  * @attr: the attribute of that group to query for
278  *
279  * Returns: 1 if the attribute exists
280  *          0 if the attribute either does not exist or if the vm device
281  *            interface is unavailable
282  */
283 int kvm_device_check_attr(int fd, uint32_t group, uint64_t attr);
284 
285 /**
286  * kvm_device_access - set or get value of a specific vm attribute
287  * @fd: The device file descriptor
288  * @group: the group
289  * @attr: the attribute of that group to set or get
290  * @val: pointer to a storage area for the value
291  * @write: true for set and false for get operation
292  *
293  * This function is not allowed to fail. Use kvm_device_check_attr()
294  * in order to check for the availability of optional attributes.
295  */
296 void kvm_device_access(int fd, int group, uint64_t attr,
297                        void *val, bool write);
298 
299 /**
300  * kvm_create_device - create a KVM device for the device control API
301  * @KVMState: The KVMState pointer
302  * @type: The KVM device type (see Documentation/virtual/kvm/devices in the
303  *        kernel source)
304  * @test: If true, only test if device can be created, but don't actually
305  *        create the device.
306  *
307  * Returns: -errno on error, nonnegative on success: @test ? 0 : device fd;
308  */
309 int kvm_create_device(KVMState *s, uint64_t type, bool test);
310 
311 /**
312  * kvm_device_supported - probe whether KVM supports specific device
313  *
314  * @vmfd: The fd handler for VM
315  * @type: type of device
316  *
317  * @return: true if supported, otherwise false.
318  */
319 bool kvm_device_supported(int vmfd, uint64_t type);
320 
321 /* Arch specific hooks */
322 
323 extern const KVMCapabilityInfo kvm_arch_required_capabilities[];
324 
325 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run);
326 MemTxAttrs kvm_arch_post_run(CPUState *cpu, struct kvm_run *run);
327 
328 int kvm_arch_handle_exit(CPUState *cpu, struct kvm_run *run);
329 
330 int kvm_arch_handle_ioapic_eoi(CPUState *cpu, struct kvm_run *run);
331 
332 int kvm_arch_process_async_events(CPUState *cpu);
333 
334 int kvm_arch_get_registers(CPUState *cpu);
335 
336 /* state subset only touched by the VCPU itself during runtime */
337 #define KVM_PUT_RUNTIME_STATE   1
338 /* state subset modified during VCPU reset */
339 #define KVM_PUT_RESET_STATE     2
340 /* full state set, modified during initialization or on vmload */
341 #define KVM_PUT_FULL_STATE      3
342 
343 int kvm_arch_put_registers(CPUState *cpu, int level);
344 
345 int kvm_arch_init(MachineState *ms, KVMState *s);
346 
347 int kvm_arch_init_vcpu(CPUState *cpu);
348 
349 bool kvm_vcpu_id_is_valid(int vcpu_id);
350 
351 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */
352 unsigned long kvm_arch_vcpu_id(CPUState *cpu);
353 
354 int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
355 int kvm_arch_on_sigbus(int code, void *addr);
356 
357 void kvm_arch_init_irq_routing(KVMState *s);
358 
359 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry *route,
360                              uint64_t address, uint32_t data, PCIDevice *dev);
361 
362 /* Notify arch about newly added MSI routes */
363 int kvm_arch_add_msi_route_post(struct kvm_irq_routing_entry *route,
364                                 int vector, PCIDevice *dev);
365 /* Notify arch about released MSI routes */
366 int kvm_arch_release_virq_post(int virq);
367 
368 int kvm_arch_msi_data_to_gsi(uint32_t data);
369 
370 int kvm_set_irq(KVMState *s, int irq, int level);
371 int kvm_irqchip_send_msi(KVMState *s, MSIMessage msg);
372 
373 void kvm_irqchip_add_irq_route(KVMState *s, int gsi, int irqchip, int pin);
374 
375 void kvm_put_apic_state(DeviceState *d, struct kvm_lapic_state *kapic);
376 void kvm_get_apic_state(DeviceState *d, struct kvm_lapic_state *kapic);
377 
378 struct kvm_guest_debug;
379 struct kvm_debug_exit_arch;
380 
381 struct kvm_sw_breakpoint {
382     target_ulong pc;
383     target_ulong saved_insn;
384     int use_count;
385     QTAILQ_ENTRY(kvm_sw_breakpoint) entry;
386 };
387 
388 QTAILQ_HEAD(kvm_sw_breakpoint_head, kvm_sw_breakpoint);
389 
390 struct kvm_sw_breakpoint *kvm_find_sw_breakpoint(CPUState *cpu,
391                                                  target_ulong pc);
392 
393 int kvm_sw_breakpoints_active(CPUState *cpu);
394 
395 int kvm_arch_insert_sw_breakpoint(CPUState *cpu,
396                                   struct kvm_sw_breakpoint *bp);
397 int kvm_arch_remove_sw_breakpoint(CPUState *cpu,
398                                   struct kvm_sw_breakpoint *bp);
399 int kvm_arch_insert_hw_breakpoint(target_ulong addr,
400                                   target_ulong len, int type);
401 int kvm_arch_remove_hw_breakpoint(target_ulong addr,
402                                   target_ulong len, int type);
403 void kvm_arch_remove_all_hw_breakpoints(void);
404 
405 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg);
406 
407 bool kvm_arch_stop_on_emulation_error(CPUState *cpu);
408 
409 int kvm_check_extension(KVMState *s, unsigned int extension);
410 
411 int kvm_vm_check_extension(KVMState *s, unsigned int extension);
412 
413 #define kvm_vm_enable_cap(s, capability, cap_flags, ...)             \
414     ({                                                               \
415         struct kvm_enable_cap cap = {                                \
416             .cap = capability,                                       \
417             .flags = cap_flags,                                      \
418         };                                                           \
419         uint64_t args_tmp[] = { __VA_ARGS__ };                       \
420         int i;                                                       \
421         for (i = 0; i < (int)ARRAY_SIZE(args_tmp) &&                 \
422                      i < ARRAY_SIZE(cap.args); i++) {                \
423             cap.args[i] = args_tmp[i];                               \
424         }                                                            \
425         kvm_vm_ioctl(s, KVM_ENABLE_CAP, &cap);                       \
426     })
427 
428 #define kvm_vcpu_enable_cap(cpu, capability, cap_flags, ...)         \
429     ({                                                               \
430         struct kvm_enable_cap cap = {                                \
431             .cap = capability,                                       \
432             .flags = cap_flags,                                      \
433         };                                                           \
434         uint64_t args_tmp[] = { __VA_ARGS__ };                       \
435         int i;                                                       \
436         for (i = 0; i < (int)ARRAY_SIZE(args_tmp) &&                 \
437                      i < ARRAY_SIZE(cap.args); i++) {                \
438             cap.args[i] = args_tmp[i];                               \
439         }                                                            \
440         kvm_vcpu_ioctl(cpu, KVM_ENABLE_CAP, &cap);                   \
441     })
442 
443 uint32_t kvm_arch_get_supported_cpuid(KVMState *env, uint32_t function,
444                                       uint32_t index, int reg);
445 
446 void kvm_set_sigmask_len(KVMState *s, unsigned int sigmask_len);
447 
448 #if !defined(CONFIG_USER_ONLY)
449 int kvm_physical_memory_addr_from_host(KVMState *s, void *ram_addr,
450                                        hwaddr *phys_addr);
451 #endif
452 
453 #endif /* NEED_CPU_H */
454 
455 void kvm_cpu_synchronize_state(CPUState *cpu);
456 void kvm_cpu_synchronize_post_reset(CPUState *cpu);
457 void kvm_cpu_synchronize_post_init(CPUState *cpu);
458 
459 /* generic hooks - to be moved/refactored once there are more users */
460 
461 static inline void cpu_synchronize_state(CPUState *cpu)
462 {
463     if (kvm_enabled()) {
464         kvm_cpu_synchronize_state(cpu);
465     }
466 }
467 
468 static inline void cpu_synchronize_post_reset(CPUState *cpu)
469 {
470     if (kvm_enabled()) {
471         kvm_cpu_synchronize_post_reset(cpu);
472     }
473 }
474 
475 static inline void cpu_synchronize_post_init(CPUState *cpu)
476 {
477     if (kvm_enabled()) {
478         kvm_cpu_synchronize_post_init(cpu);
479     }
480 }
481 
482 /**
483  * kvm_irqchip_add_msi_route - Add MSI route for specific vector
484  * @s:      KVM state
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
488  *          message.
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.
492  */
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,
495                                  PCIDevice *dev);
496 void kvm_irqchip_commit_routes(KVMState *s);
497 void kvm_irqchip_release_virq(KVMState *s, int virq);
498 
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);
501 
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,
505                                           int virq);
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,
509                                       qemu_irq irq);
510 void kvm_irqchip_set_qemuirq_gsi(KVMState *s, qemu_irq irq, int gsi);
511 void kvm_pc_gsi_handler(void *opaque, int n, int level);
512 void kvm_pc_setup_irq_routing(bool pci_enabled);
513 void kvm_init_irq_routing(KVMState *s);
514 
515 /**
516  * kvm_arch_irqchip_create:
517  * @KVMState: The KVMState pointer
518  * @MachineState: The MachineState pointer
519  *
520  * Allow architectures to create an in-kernel irq chip themselves.
521  *
522  * Returns: < 0: error
523  *            0: irq chip was not created
524  *          > 0: irq chip was created
525  */
526 int kvm_arch_irqchip_create(MachineState *ms, KVMState *s);
527 
528 /**
529  * kvm_set_one_reg - set a register value in KVM via KVM_SET_ONE_REG ioctl
530  * @id: The register ID
531  * @source: The pointer to the value to be set. It must point to a variable
532  *          of the correct type/size for the register being accessed.
533  *
534  * Returns: 0 on success, or a negative errno on failure.
535  */
536 int kvm_set_one_reg(CPUState *cs, uint64_t id, void *source);
537 
538 /**
539  * kvm_get_one_reg - get a register value from KVM via KVM_GET_ONE_REG ioctl
540  * @id: The register ID
541  * @target: The pointer where the value is to be stored. It must point to a
542  *          variable of the correct type/size for the register being accessed.
543  *
544  * Returns: 0 on success, or a negative errno on failure.
545  */
546 int kvm_get_one_reg(CPUState *cs, uint64_t id, void *target);
547 int kvm_get_max_memslots(void);
548 #endif
549