xref: /openbmc/qemu/include/sysemu/kvm.h (revision 93977402)
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 extern bool kvm_msi_use_devid;
57 
58 #if defined CONFIG_KVM || !defined NEED_CPU_H
59 #define kvm_enabled()           (kvm_allowed)
60 /**
61  * kvm_irqchip_in_kernel:
62  *
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.
70  */
71 #define kvm_irqchip_in_kernel() (kvm_kernel_irqchip)
72 
73 /**
74  * kvm_irqchip_is_split:
75  *
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.
80  */
81 #define kvm_irqchip_is_split() (kvm_split_irqchip)
82 
83 /**
84  * kvm_async_interrupts_enabled:
85  *
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).
90  */
91 #define kvm_async_interrupts_enabled() (kvm_async_interrupts_allowed)
92 
93 /**
94  * kvm_halt_in_kernel
95  *
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.
98  */
99 #define kvm_halt_in_kernel() (kvm_halt_in_kernel_allowed)
100 
101 /**
102  * kvm_eventfds_enabled:
103  *
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).
107  */
108 #define kvm_eventfds_enabled() (kvm_eventfds_allowed)
109 
110 /**
111  * kvm_irqfds_enabled:
112  *
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).
116  */
117 #define kvm_irqfds_enabled() (kvm_irqfds_allowed)
118 
119 /**
120  * kvm_resamplefds_enabled:
121  *
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).
125  */
126 #define kvm_resamplefds_enabled() (kvm_resamplefds_allowed)
127 
128 /**
129  * kvm_msi_via_irqfd_enabled:
130  *
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.
134  */
135 #define kvm_msi_via_irqfd_enabled() (kvm_msi_via_irqfd_allowed)
136 
137 /**
138  * kvm_gsi_routing_enabled:
139  *
140  * Returns: true if GSI routing is enabled (ie the kernel supports
141  * it and we're running in a configuration that permits it).
142  */
143 #define kvm_gsi_routing_enabled() (kvm_gsi_routing_allowed)
144 
145 /**
146  * kvm_gsi_direct_mapping:
147  *
148  * Returns: true if GSI direct mapping is enabled.
149  */
150 #define kvm_gsi_direct_mapping() (kvm_gsi_direct_mapping)
151 
152 /**
153  * kvm_readonly_mem_enabled:
154  *
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).
157  */
158 #define kvm_readonly_mem_enabled() (kvm_readonly_mem_allowed)
159 
160 /**
161  * kvm_direct_msi_enabled:
162  *
163  * Returns: true if KVM allows direct MSI injection.
164  */
165 #define kvm_direct_msi_enabled() (kvm_direct_msi_allowed)
166 
167 /**
168  * kvm_ioeventfd_any_length_enabled:
169  * Returns: true if KVM allows any length io eventfd.
170  */
171 #define kvm_ioeventfd_any_length_enabled() (kvm_ioeventfd_any_length_allowed)
172 
173 /**
174  * kvm_msi_devid_required:
175  * Returns: true if KVM requires a device id to be provided while
176  * defining an MSI routing entry.
177  */
178 #define kvm_msi_devid_required() (kvm_msi_use_devid)
179 
180 #else
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)
196 #endif
197 
198 struct kvm_run;
199 struct kvm_lapic_state;
200 struct kvm_irq_routing_entry;
201 
202 typedef struct KVMCapabilityInfo {
203     const char *name;
204     int value;
205 } KVMCapabilityInfo;
206 
207 #define KVM_CAP_INFO(CAP) { "KVM_CAP_" stringify(CAP), KVM_CAP_##CAP }
208 #define KVM_CAP_LAST_INFO { NULL, 0 }
209 
210 struct KVMState;
211 typedef struct KVMState KVMState;
212 extern KVMState *kvm_state;
213 
214 /* external API */
215 
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);
225 
226 int kvm_init_vcpu(CPUState *cpu);
227 int kvm_cpu_exec(CPUState *cpu);
228 int kvm_destroy_vcpu(CPUState *cpu);
229 
230 #ifdef NEED_CPU_H
231 #include "cpu.h"
232 
233 void kvm_flush_coalesced_mmio_buffer(void);
234 
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);
241 #ifndef _WIN32
242 int kvm_set_signal_mask(CPUState *cpu, const sigset_t *sigset);
243 #endif
244 
245 int kvm_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
246 int kvm_on_sigbus(int code, void *addr);
247 
248 /* interface with exec.c */
249 
250 void phys_mem_set_alloc(void *(*alloc)(size_t, uint64_t *align));
251 
252 /* internal API */
253 
254 int kvm_ioctl(KVMState *s, int type, ...);
255 
256 int kvm_vm_ioctl(KVMState *s, int type, ...);
257 
258 int kvm_vcpu_ioctl(CPUState *cpu, int type, ...);
259 
260 /**
261  * kvm_device_ioctl - call an ioctl on a kvm device
262  * @fd: The KVM device file descriptor as returned from KVM_CREATE_DEVICE
263  * @type: The device-ctrl ioctl number
264  *
265  * Returns: -errno on error, nonnegative on success
266  */
267 int kvm_device_ioctl(int fd, int type, ...);
268 
269 /**
270  * kvm_vm_check_attr - check for existence of a specific vm attribute
271  * @s: The KVMState pointer
272  * @group: the group
273  * @attr: the attribute of that group to query for
274  *
275  * Returns: 1 if the attribute exists
276  *          0 if the attribute either does not exist or if the vm device
277  *            interface is unavailable
278  */
279 int kvm_vm_check_attr(KVMState *s, uint32_t group, uint64_t attr);
280 
281 /**
282  * kvm_device_check_attr - check for existence of a specific device attribute
283  * @fd: The device file descriptor
284  * @group: the group
285  * @attr: the attribute of that group to query for
286  *
287  * Returns: 1 if the attribute exists
288  *          0 if the attribute either does not exist or if the vm device
289  *            interface is unavailable
290  */
291 int kvm_device_check_attr(int fd, uint32_t group, uint64_t attr);
292 
293 /**
294  * kvm_device_access - set or get value of a specific vm attribute
295  * @fd: The device file descriptor
296  * @group: the group
297  * @attr: the attribute of that group to set or get
298  * @val: pointer to a storage area for the value
299  * @write: true for set and false for get operation
300  *
301  * This function is not allowed to fail. Use kvm_device_check_attr()
302  * in order to check for the availability of optional attributes.
303  */
304 void kvm_device_access(int fd, int group, uint64_t attr,
305                        void *val, bool write);
306 
307 /**
308  * kvm_create_device - create a KVM device for the device control API
309  * @KVMState: The KVMState pointer
310  * @type: The KVM device type (see Documentation/virtual/kvm/devices in the
311  *        kernel source)
312  * @test: If true, only test if device can be created, but don't actually
313  *        create the device.
314  *
315  * Returns: -errno on error, nonnegative on success: @test ? 0 : device fd;
316  */
317 int kvm_create_device(KVMState *s, uint64_t type, bool test);
318 
319 /**
320  * kvm_device_supported - probe whether KVM supports specific device
321  *
322  * @vmfd: The fd handler for VM
323  * @type: type of device
324  *
325  * @return: true if supported, otherwise false.
326  */
327 bool kvm_device_supported(int vmfd, uint64_t type);
328 
329 /* Arch specific hooks */
330 
331 extern const KVMCapabilityInfo kvm_arch_required_capabilities[];
332 
333 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run);
334 MemTxAttrs kvm_arch_post_run(CPUState *cpu, struct kvm_run *run);
335 
336 int kvm_arch_handle_exit(CPUState *cpu, struct kvm_run *run);
337 
338 int kvm_arch_process_async_events(CPUState *cpu);
339 
340 int kvm_arch_get_registers(CPUState *cpu);
341 
342 /* state subset only touched by the VCPU itself during runtime */
343 #define KVM_PUT_RUNTIME_STATE   1
344 /* state subset modified during VCPU reset */
345 #define KVM_PUT_RESET_STATE     2
346 /* full state set, modified during initialization or on vmload */
347 #define KVM_PUT_FULL_STATE      3
348 
349 int kvm_arch_put_registers(CPUState *cpu, int level);
350 
351 int kvm_arch_init(MachineState *ms, KVMState *s);
352 
353 int kvm_arch_init_vcpu(CPUState *cpu);
354 
355 bool kvm_vcpu_id_is_valid(int vcpu_id);
356 
357 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */
358 unsigned long kvm_arch_vcpu_id(CPUState *cpu);
359 
360 int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
361 int kvm_arch_on_sigbus(int code, void *addr);
362 
363 void kvm_arch_init_irq_routing(KVMState *s);
364 
365 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry *route,
366                              uint64_t address, uint32_t data, PCIDevice *dev);
367 
368 /* Notify arch about newly added MSI routes */
369 int kvm_arch_add_msi_route_post(struct kvm_irq_routing_entry *route,
370                                 int vector, PCIDevice *dev);
371 /* Notify arch about released MSI routes */
372 int kvm_arch_release_virq_post(int virq);
373 
374 int kvm_arch_msi_data_to_gsi(uint32_t data);
375 
376 int kvm_set_irq(KVMState *s, int irq, int level);
377 int kvm_irqchip_send_msi(KVMState *s, MSIMessage msg);
378 
379 void kvm_irqchip_add_irq_route(KVMState *s, int gsi, int irqchip, int pin);
380 
381 void kvm_get_apic_state(DeviceState *d, struct kvm_lapic_state *kapic);
382 
383 struct kvm_guest_debug;
384 struct kvm_debug_exit_arch;
385 
386 struct kvm_sw_breakpoint {
387     target_ulong pc;
388     target_ulong saved_insn;
389     int use_count;
390     QTAILQ_ENTRY(kvm_sw_breakpoint) entry;
391 };
392 
393 QTAILQ_HEAD(kvm_sw_breakpoint_head, kvm_sw_breakpoint);
394 
395 struct kvm_sw_breakpoint *kvm_find_sw_breakpoint(CPUState *cpu,
396                                                  target_ulong pc);
397 
398 int kvm_sw_breakpoints_active(CPUState *cpu);
399 
400 int kvm_arch_insert_sw_breakpoint(CPUState *cpu,
401                                   struct kvm_sw_breakpoint *bp);
402 int kvm_arch_remove_sw_breakpoint(CPUState *cpu,
403                                   struct kvm_sw_breakpoint *bp);
404 int kvm_arch_insert_hw_breakpoint(target_ulong addr,
405                                   target_ulong len, int type);
406 int kvm_arch_remove_hw_breakpoint(target_ulong addr,
407                                   target_ulong len, int type);
408 void kvm_arch_remove_all_hw_breakpoints(void);
409 
410 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg);
411 
412 bool kvm_arch_stop_on_emulation_error(CPUState *cpu);
413 
414 int kvm_check_extension(KVMState *s, unsigned int extension);
415 
416 int kvm_vm_check_extension(KVMState *s, unsigned int extension);
417 
418 #define kvm_vm_enable_cap(s, capability, cap_flags, ...)             \
419     ({                                                               \
420         struct kvm_enable_cap cap = {                                \
421             .cap = capability,                                       \
422             .flags = cap_flags,                                      \
423         };                                                           \
424         uint64_t args_tmp[] = { __VA_ARGS__ };                       \
425         int i;                                                       \
426         for (i = 0; i < (int)ARRAY_SIZE(args_tmp) &&                 \
427                      i < ARRAY_SIZE(cap.args); i++) {                \
428             cap.args[i] = args_tmp[i];                               \
429         }                                                            \
430         kvm_vm_ioctl(s, KVM_ENABLE_CAP, &cap);                       \
431     })
432 
433 #define kvm_vcpu_enable_cap(cpu, capability, cap_flags, ...)         \
434     ({                                                               \
435         struct kvm_enable_cap cap = {                                \
436             .cap = capability,                                       \
437             .flags = cap_flags,                                      \
438         };                                                           \
439         uint64_t args_tmp[] = { __VA_ARGS__ };                       \
440         int i;                                                       \
441         for (i = 0; i < (int)ARRAY_SIZE(args_tmp) &&                 \
442                      i < ARRAY_SIZE(cap.args); i++) {                \
443             cap.args[i] = args_tmp[i];                               \
444         }                                                            \
445         kvm_vcpu_ioctl(cpu, KVM_ENABLE_CAP, &cap);                   \
446     })
447 
448 uint32_t kvm_arch_get_supported_cpuid(KVMState *env, uint32_t function,
449                                       uint32_t index, int reg);
450 
451 void kvm_set_sigmask_len(KVMState *s, unsigned int sigmask_len);
452 
453 #if !defined(CONFIG_USER_ONLY)
454 int kvm_physical_memory_addr_from_host(KVMState *s, void *ram_addr,
455                                        hwaddr *phys_addr);
456 #endif
457 
458 #endif /* NEED_CPU_H */
459 
460 void kvm_cpu_synchronize_state(CPUState *cpu);
461 void kvm_cpu_synchronize_post_reset(CPUState *cpu);
462 void kvm_cpu_synchronize_post_init(CPUState *cpu);
463 
464 /* generic hooks - to be moved/refactored once there are more users */
465 
466 static inline void cpu_synchronize_state(CPUState *cpu)
467 {
468     if (kvm_enabled()) {
469         kvm_cpu_synchronize_state(cpu);
470     }
471 }
472 
473 static inline void cpu_synchronize_post_reset(CPUState *cpu)
474 {
475     if (kvm_enabled()) {
476         kvm_cpu_synchronize_post_reset(cpu);
477     }
478 }
479 
480 static inline void cpu_synchronize_post_init(CPUState *cpu)
481 {
482     if (kvm_enabled()) {
483         kvm_cpu_synchronize_post_init(cpu);
484     }
485 }
486 
487 /**
488  * kvm_irqchip_add_msi_route - Add MSI route for specific vector
489  * @s:      KVM state
490  * @vector: which vector to add. This can be either MSI/MSIX
491  *          vector. The function will automatically detect whether
492  *          MSI/MSIX is enabled, and fetch corresponding MSI
493  *          message.
494  * @dev:    Owner PCI device to add the route. If @dev is specified
495  *          as @NULL, an empty MSI message will be inited.
496  * @return: virq (>=0) when success, errno (<0) when failed.
497  */
498 int kvm_irqchip_add_msi_route(KVMState *s, int vector, PCIDevice *dev);
499 int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg,
500                                  PCIDevice *dev);
501 void kvm_irqchip_commit_routes(KVMState *s);
502 void kvm_irqchip_release_virq(KVMState *s, int virq);
503 
504 int kvm_irqchip_add_adapter_route(KVMState *s, AdapterInfo *adapter);
505 int kvm_irqchip_add_hv_sint_route(KVMState *s, uint32_t vcpu, uint32_t sint);
506 
507 int kvm_irqchip_add_irqfd_notifier_gsi(KVMState *s, EventNotifier *n,
508                                        EventNotifier *rn, int virq);
509 int kvm_irqchip_remove_irqfd_notifier_gsi(KVMState *s, EventNotifier *n,
510                                           int virq);
511 int kvm_irqchip_add_irqfd_notifier(KVMState *s, EventNotifier *n,
512                                    EventNotifier *rn, qemu_irq irq);
513 int kvm_irqchip_remove_irqfd_notifier(KVMState *s, EventNotifier *n,
514                                       qemu_irq irq);
515 void kvm_irqchip_set_qemuirq_gsi(KVMState *s, qemu_irq irq, int gsi);
516 void kvm_pc_gsi_handler(void *opaque, int n, int level);
517 void kvm_pc_setup_irq_routing(bool pci_enabled);
518 void kvm_init_irq_routing(KVMState *s);
519 
520 /**
521  * kvm_arch_irqchip_create:
522  * @KVMState: The KVMState pointer
523  * @MachineState: The MachineState pointer
524  *
525  * Allow architectures to create an in-kernel irq chip themselves.
526  *
527  * Returns: < 0: error
528  *            0: irq chip was not created
529  *          > 0: irq chip was created
530  */
531 int kvm_arch_irqchip_create(MachineState *ms, KVMState *s);
532 
533 /**
534  * kvm_set_one_reg - set a register value in KVM via KVM_SET_ONE_REG ioctl
535  * @id: The register ID
536  * @source: The pointer to the value to be set. It must point to a variable
537  *          of the correct type/size for the register being accessed.
538  *
539  * Returns: 0 on success, or a negative errno on failure.
540  */
541 int kvm_set_one_reg(CPUState *cs, uint64_t id, void *source);
542 
543 /**
544  * kvm_get_one_reg - get a register value from KVM via KVM_GET_ONE_REG ioctl
545  * @id: The register ID
546  * @target: The pointer where the value is to be stored. It must point to a
547  *          variable 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_get_one_reg(CPUState *cs, uint64_t id, void *target);
552 int kvm_get_max_memslots(void);
553 #endif
554