xref: /openbmc/qemu/include/sysemu/kvm.h (revision 9f2d175d)
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 NEED_CPU_H
23 # ifdef CONFIG_KVM
24 #  include <linux/kvm.h>
25 #  include <linux/kvm_para.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 typedef struct KVMState KVMState;
205 extern KVMState *kvm_state;
206 
207 /* external API */
208 
209 bool kvm_has_free_slot(MachineState *ms);
210 bool kvm_has_sync_mmu(void);
211 int kvm_has_vcpu_events(void);
212 int kvm_has_robust_singlestep(void);
213 int kvm_has_debugregs(void);
214 int kvm_has_pit_state2(void);
215 int kvm_has_many_ioeventfds(void);
216 int kvm_has_gsi_routing(void);
217 int kvm_has_intx_set_mask(void);
218 
219 int kvm_init_vcpu(CPUState *cpu);
220 int kvm_cpu_exec(CPUState *cpu);
221 int kvm_destroy_vcpu(CPUState *cpu);
222 
223 /**
224  * kvm_arm_supports_user_irq
225  *
226  * Not all KVM implementations support notifications for kernel generated
227  * interrupt events to user space. This function indicates whether the current
228  * KVM implementation does support them.
229  *
230  * Returns: true if KVM supports using kernel generated IRQs from user space
231  */
232 bool kvm_arm_supports_user_irq(void);
233 
234 #ifdef NEED_CPU_H
235 #include "cpu.h"
236 
237 void kvm_flush_coalesced_mmio_buffer(void);
238 
239 int kvm_insert_breakpoint(CPUState *cpu, target_ulong addr,
240                           target_ulong len, int type);
241 int kvm_remove_breakpoint(CPUState *cpu, target_ulong addr,
242                           target_ulong len, int type);
243 void kvm_remove_all_breakpoints(CPUState *cpu);
244 int kvm_update_guest_debug(CPUState *cpu, unsigned long reinject_trap);
245 
246 int kvm_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
247 int kvm_on_sigbus(int code, void *addr);
248 
249 /* interface with exec.c */
250 
251 void phys_mem_set_alloc(void *(*alloc)(size_t, uint64_t *align, bool shared));
252 
253 /* internal API */
254 
255 int kvm_ioctl(KVMState *s, int type, ...);
256 
257 int kvm_vm_ioctl(KVMState *s, int type, ...);
258 
259 int kvm_vcpu_ioctl(CPUState *cpu, int type, ...);
260 
261 /**
262  * kvm_device_ioctl - call an ioctl on a kvm device
263  * @fd: The KVM device file descriptor as returned from KVM_CREATE_DEVICE
264  * @type: The device-ctrl ioctl number
265  *
266  * Returns: -errno on error, nonnegative on success
267  */
268 int kvm_device_ioctl(int fd, int type, ...);
269 
270 /**
271  * kvm_vm_check_attr - check for existence of a specific vm attribute
272  * @s: The KVMState pointer
273  * @group: the group
274  * @attr: the attribute of that group to query for
275  *
276  * Returns: 1 if the attribute exists
277  *          0 if the attribute either does not exist or if the vm device
278  *            interface is unavailable
279  */
280 int kvm_vm_check_attr(KVMState *s, uint32_t group, uint64_t attr);
281 
282 /**
283  * kvm_device_check_attr - check for existence of a specific device attribute
284  * @fd: The device file descriptor
285  * @group: the group
286  * @attr: the attribute of that group to query for
287  *
288  * Returns: 1 if the attribute exists
289  *          0 if the attribute either does not exist or if the vm device
290  *            interface is unavailable
291  */
292 int kvm_device_check_attr(int fd, uint32_t group, uint64_t attr);
293 
294 /**
295  * kvm_device_access - set or get value of a specific vm attribute
296  * @fd: The device file descriptor
297  * @group: the group
298  * @attr: the attribute of that group to set or get
299  * @val: pointer to a storage area for the value
300  * @write: true for set and false for get operation
301  * @errp: error object handle
302  *
303  * Returns: 0 on success
304  *          < 0 on error
305  * Use kvm_device_check_attr() in order to check for the availability
306  * of optional attributes.
307  */
308 int kvm_device_access(int fd, int group, uint64_t attr,
309                       void *val, bool write, Error **errp);
310 
311 /**
312  * kvm_create_device - create a KVM device for the device control API
313  * @KVMState: The KVMState pointer
314  * @type: The KVM device type (see Documentation/virtual/kvm/devices in the
315  *        kernel source)
316  * @test: If true, only test if device can be created, but don't actually
317  *        create the device.
318  *
319  * Returns: -errno on error, nonnegative on success: @test ? 0 : device fd;
320  */
321 int kvm_create_device(KVMState *s, uint64_t type, bool test);
322 
323 /**
324  * kvm_device_supported - probe whether KVM supports specific device
325  *
326  * @vmfd: The fd handler for VM
327  * @type: type of device
328  *
329  * @return: true if supported, otherwise false.
330  */
331 bool kvm_device_supported(int vmfd, uint64_t type);
332 
333 /* Arch specific hooks */
334 
335 extern const KVMCapabilityInfo kvm_arch_required_capabilities[];
336 
337 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run);
338 MemTxAttrs kvm_arch_post_run(CPUState *cpu, struct kvm_run *run);
339 
340 int kvm_arch_handle_exit(CPUState *cpu, struct kvm_run *run);
341 
342 int kvm_arch_process_async_events(CPUState *cpu);
343 
344 int kvm_arch_get_registers(CPUState *cpu);
345 
346 /* state subset only touched by the VCPU itself during runtime */
347 #define KVM_PUT_RUNTIME_STATE   1
348 /* state subset modified during VCPU reset */
349 #define KVM_PUT_RESET_STATE     2
350 /* full state set, modified during initialization or on vmload */
351 #define KVM_PUT_FULL_STATE      3
352 
353 int kvm_arch_put_registers(CPUState *cpu, int level);
354 
355 int kvm_arch_init(MachineState *ms, KVMState *s);
356 
357 int kvm_arch_init_vcpu(CPUState *cpu);
358 
359 bool kvm_vcpu_id_is_valid(int vcpu_id);
360 
361 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */
362 unsigned long kvm_arch_vcpu_id(CPUState *cpu);
363 
364 #ifdef TARGET_I386
365 #define KVM_HAVE_MCE_INJECTION 1
366 void kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
367 #endif
368 
369 void kvm_arch_init_irq_routing(KVMState *s);
370 
371 int kvm_arch_fixup_msi_route(struct kvm_irq_routing_entry *route,
372                              uint64_t address, uint32_t data, PCIDevice *dev);
373 
374 /* Notify arch about newly added MSI routes */
375 int kvm_arch_add_msi_route_post(struct kvm_irq_routing_entry *route,
376                                 int vector, PCIDevice *dev);
377 /* Notify arch about released MSI routes */
378 int kvm_arch_release_virq_post(int virq);
379 
380 int kvm_arch_msi_data_to_gsi(uint32_t data);
381 
382 int kvm_set_irq(KVMState *s, int irq, int level);
383 int kvm_irqchip_send_msi(KVMState *s, MSIMessage msg);
384 
385 void kvm_irqchip_add_irq_route(KVMState *s, int gsi, int irqchip, int pin);
386 
387 void kvm_get_apic_state(DeviceState *d, struct kvm_lapic_state *kapic);
388 
389 struct kvm_guest_debug;
390 struct kvm_debug_exit_arch;
391 
392 struct kvm_sw_breakpoint {
393     target_ulong pc;
394     target_ulong saved_insn;
395     int use_count;
396     QTAILQ_ENTRY(kvm_sw_breakpoint) entry;
397 };
398 
399 QTAILQ_HEAD(kvm_sw_breakpoint_head, kvm_sw_breakpoint);
400 
401 struct kvm_sw_breakpoint *kvm_find_sw_breakpoint(CPUState *cpu,
402                                                  target_ulong pc);
403 
404 int kvm_sw_breakpoints_active(CPUState *cpu);
405 
406 int kvm_arch_insert_sw_breakpoint(CPUState *cpu,
407                                   struct kvm_sw_breakpoint *bp);
408 int kvm_arch_remove_sw_breakpoint(CPUState *cpu,
409                                   struct kvm_sw_breakpoint *bp);
410 int kvm_arch_insert_hw_breakpoint(target_ulong addr,
411                                   target_ulong len, int type);
412 int kvm_arch_remove_hw_breakpoint(target_ulong addr,
413                                   target_ulong len, int type);
414 void kvm_arch_remove_all_hw_breakpoints(void);
415 
416 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg);
417 
418 bool kvm_arch_stop_on_emulation_error(CPUState *cpu);
419 
420 int kvm_check_extension(KVMState *s, unsigned int extension);
421 
422 int kvm_vm_check_extension(KVMState *s, unsigned int extension);
423 
424 #define kvm_vm_enable_cap(s, capability, cap_flags, ...)             \
425     ({                                                               \
426         struct kvm_enable_cap cap = {                                \
427             .cap = capability,                                       \
428             .flags = cap_flags,                                      \
429         };                                                           \
430         uint64_t args_tmp[] = { __VA_ARGS__ };                       \
431         size_t n = MIN(ARRAY_SIZE(args_tmp), ARRAY_SIZE(cap.args));  \
432         memcpy(cap.args, args_tmp, n * sizeof(cap.args[0]));         \
433         kvm_vm_ioctl(s, KVM_ENABLE_CAP, &cap);                       \
434     })
435 
436 #define kvm_vcpu_enable_cap(cpu, capability, cap_flags, ...)         \
437     ({                                                               \
438         struct kvm_enable_cap cap = {                                \
439             .cap = capability,                                       \
440             .flags = cap_flags,                                      \
441         };                                                           \
442         uint64_t args_tmp[] = { __VA_ARGS__ };                       \
443         size_t n = MIN(ARRAY_SIZE(args_tmp), ARRAY_SIZE(cap.args));  \
444         memcpy(cap.args, args_tmp, n * sizeof(cap.args[0]));         \
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 void kvm_cpu_synchronize_pre_loadvm(CPUState *cpu);
464 
465 void kvm_init_cpu_signals(CPUState *cpu);
466 
467 /**
468  * kvm_irqchip_add_msi_route - Add MSI route for specific vector
469  * @s:      KVM state
470  * @vector: which vector to add. This can be either MSI/MSIX
471  *          vector. The function will automatically detect whether
472  *          MSI/MSIX is enabled, and fetch corresponding MSI
473  *          message.
474  * @dev:    Owner PCI device to add the route. If @dev is specified
475  *          as @NULL, an empty MSI message will be inited.
476  * @return: virq (>=0) when success, errno (<0) when failed.
477  */
478 int kvm_irqchip_add_msi_route(KVMState *s, int vector, PCIDevice *dev);
479 int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg,
480                                  PCIDevice *dev);
481 void kvm_irqchip_commit_routes(KVMState *s);
482 void kvm_irqchip_release_virq(KVMState *s, int virq);
483 
484 int kvm_irqchip_add_adapter_route(KVMState *s, AdapterInfo *adapter);
485 int kvm_irqchip_add_hv_sint_route(KVMState *s, uint32_t vcpu, uint32_t sint);
486 
487 int kvm_irqchip_add_irqfd_notifier_gsi(KVMState *s, EventNotifier *n,
488                                        EventNotifier *rn, int virq);
489 int kvm_irqchip_remove_irqfd_notifier_gsi(KVMState *s, EventNotifier *n,
490                                           int virq);
491 int kvm_irqchip_add_irqfd_notifier(KVMState *s, EventNotifier *n,
492                                    EventNotifier *rn, qemu_irq irq);
493 int kvm_irqchip_remove_irqfd_notifier(KVMState *s, EventNotifier *n,
494                                       qemu_irq irq);
495 void kvm_irqchip_set_qemuirq_gsi(KVMState *s, qemu_irq irq, int gsi);
496 void kvm_pc_gsi_handler(void *opaque, int n, int level);
497 void kvm_pc_setup_irq_routing(bool pci_enabled);
498 void kvm_init_irq_routing(KVMState *s);
499 
500 /**
501  * kvm_arch_irqchip_create:
502  * @KVMState: The KVMState pointer
503  * @MachineState: The MachineState pointer
504  *
505  * Allow architectures to create an in-kernel irq chip themselves.
506  *
507  * Returns: < 0: error
508  *            0: irq chip was not created
509  *          > 0: irq chip was created
510  */
511 int kvm_arch_irqchip_create(MachineState *ms, KVMState *s);
512 
513 /**
514  * kvm_set_one_reg - set a register value in KVM via KVM_SET_ONE_REG ioctl
515  * @id: The register ID
516  * @source: The pointer to the value to be set. It must point to a variable
517  *          of the correct type/size for the register being accessed.
518  *
519  * Returns: 0 on success, or a negative errno on failure.
520  */
521 int kvm_set_one_reg(CPUState *cs, uint64_t id, void *source);
522 
523 /**
524  * kvm_get_one_reg - get a register value from KVM via KVM_GET_ONE_REG ioctl
525  * @id: The register ID
526  * @target: The pointer where the value is to be stored. It must point to a
527  *          variable of the correct type/size for the register being accessed.
528  *
529  * Returns: 0 on success, or a negative errno on failure.
530  */
531 int kvm_get_one_reg(CPUState *cs, uint64_t id, void *target);
532 struct ppc_radix_page_info *kvm_get_radix_page_info(void);
533 int kvm_get_max_memslots(void);
534 #endif
535