numa: add -numa node,memdev= option
[qemu.git] / include / sysemu / kvm.h
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 <errno.h>
18 #include "config-host.h"
19 #include "qemu/queue.h"
20 #include "qom/cpu.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_async_interrupts_allowed;
45 extern bool kvm_halt_in_kernel_allowed;
46 extern bool kvm_eventfds_allowed;
47 extern bool kvm_irqfds_allowed;
48 extern bool kvm_msi_via_irqfd_allowed;
49 extern bool kvm_gsi_routing_allowed;
50 extern bool kvm_gsi_direct_mapping;
51 extern bool kvm_readonly_mem_allowed;
52
53 #if defined CONFIG_KVM || !defined NEED_CPU_H
54 #define kvm_enabled() (kvm_allowed)
55 /**
56 * kvm_irqchip_in_kernel:
57 *
58 * Returns: true if the user asked us to create an in-kernel
59 * irqchip via the "kernel_irqchip=on" machine option.
60 * What this actually means is architecture and machine model
61 * specific: on PC, for instance, it means that the LAPIC,
62 * IOAPIC and PIT are all in kernel. This function should never
63 * be used from generic target-independent code: use one of the
64 * following functions or some other specific check instead.
65 */
66 #define kvm_irqchip_in_kernel() (kvm_kernel_irqchip)
67
68 /**
69 * kvm_async_interrupts_enabled:
70 *
71 * Returns: true if we can deliver interrupts to KVM
72 * asynchronously (ie by ioctl from any thread at any time)
73 * rather than having to do interrupt delivery synchronously
74 * (where the vcpu must be stopped at a suitable point first).
75 */
76 #define kvm_async_interrupts_enabled() (kvm_async_interrupts_allowed)
77
78 /**
79 * kvm_halt_in_kernel
80 *
81 * Returns: true if halted cpus should still get a KVM_RUN ioctl to run
82 * inside of kernel space. This only works if MP state is implemented.
83 */
84 #define kvm_halt_in_kernel() (kvm_halt_in_kernel_allowed)
85
86 /**
87 * kvm_eventfds_enabled:
88 *
89 * Returns: true if we can use eventfds to receive notifications
90 * from a KVM CPU (ie the kernel supports eventds and we are running
91 * with a configuration where it is meaningful to use them).
92 */
93 #define kvm_eventfds_enabled() (kvm_eventfds_allowed)
94
95 /**
96 * kvm_irqfds_enabled:
97 *
98 * Returns: true if we can use irqfds to inject interrupts into
99 * a KVM CPU (ie the kernel supports irqfds and we are running
100 * with a configuration where it is meaningful to use them).
101 */
102 #define kvm_irqfds_enabled() (kvm_irqfds_allowed)
103
104 /**
105 * kvm_msi_via_irqfd_enabled:
106 *
107 * Returns: true if we can route a PCI MSI (Message Signaled Interrupt)
108 * to a KVM CPU via an irqfd. This requires that the kernel supports
109 * this and that we're running in a configuration that permits it.
110 */
111 #define kvm_msi_via_irqfd_enabled() (kvm_msi_via_irqfd_allowed)
112
113 /**
114 * kvm_gsi_routing_enabled:
115 *
116 * Returns: true if GSI routing is enabled (ie the kernel supports
117 * it and we're running in a configuration that permits it).
118 */
119 #define kvm_gsi_routing_enabled() (kvm_gsi_routing_allowed)
120
121 /**
122 * kvm_gsi_direct_mapping:
123 *
124 * Returns: true if GSI direct mapping is enabled.
125 */
126 #define kvm_gsi_direct_mapping() (kvm_gsi_direct_mapping)
127
128 /**
129 * kvm_readonly_mem_enabled:
130 *
131 * Returns: true if KVM readonly memory is enabled (ie the kernel
132 * supports it and we're running in a configuration that permits it).
133 */
134 #define kvm_readonly_mem_enabled() (kvm_readonly_mem_allowed)
135
136 #else
137 #define kvm_enabled() (0)
138 #define kvm_irqchip_in_kernel() (false)
139 #define kvm_async_interrupts_enabled() (false)
140 #define kvm_halt_in_kernel() (false)
141 #define kvm_eventfds_enabled() (false)
142 #define kvm_irqfds_enabled() (false)
143 #define kvm_msi_via_irqfd_enabled() (false)
144 #define kvm_gsi_routing_allowed() (false)
145 #define kvm_gsi_direct_mapping() (false)
146 #define kvm_readonly_mem_enabled() (false)
147 #endif
148
149 struct kvm_run;
150 struct kvm_lapic_state;
151
152 typedef struct KVMCapabilityInfo {
153 const char *name;
154 int value;
155 } KVMCapabilityInfo;
156
157 #define KVM_CAP_INFO(CAP) { "KVM_CAP_" stringify(CAP), KVM_CAP_##CAP }
158 #define KVM_CAP_LAST_INFO { NULL, 0 }
159
160 struct KVMState;
161 typedef struct KVMState KVMState;
162 extern KVMState *kvm_state;
163
164 /* external API */
165
166 int kvm_init(MachineClass *mc);
167
168 int kvm_has_sync_mmu(void);
169 int kvm_has_vcpu_events(void);
170 int kvm_has_robust_singlestep(void);
171 int kvm_has_debugregs(void);
172 int kvm_has_xsave(void);
173 int kvm_has_xcrs(void);
174 int kvm_has_pit_state2(void);
175 int kvm_has_many_ioeventfds(void);
176 int kvm_has_gsi_routing(void);
177 int kvm_has_intx_set_mask(void);
178
179 int kvm_init_vcpu(CPUState *cpu);
180 int kvm_cpu_exec(CPUState *cpu);
181
182 #ifdef NEED_CPU_H
183
184 void kvm_setup_guest_memory(void *start, size_t size);
185 void kvm_flush_coalesced_mmio_buffer(void);
186
187 int kvm_insert_breakpoint(CPUState *cpu, target_ulong addr,
188 target_ulong len, int type);
189 int kvm_remove_breakpoint(CPUState *cpu, target_ulong addr,
190 target_ulong len, int type);
191 void kvm_remove_all_breakpoints(CPUState *cpu);
192 int kvm_update_guest_debug(CPUState *cpu, unsigned long reinject_trap);
193 #ifndef _WIN32
194 int kvm_set_signal_mask(CPUState *cpu, const sigset_t *sigset);
195 #endif
196
197 int kvm_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
198 int kvm_on_sigbus(int code, void *addr);
199
200 /* internal API */
201
202 int kvm_ioctl(KVMState *s, int type, ...);
203
204 int kvm_vm_ioctl(KVMState *s, int type, ...);
205
206 int kvm_vcpu_ioctl(CPUState *cpu, int type, ...);
207
208 /**
209 * kvm_device_ioctl - call an ioctl on a kvm device
210 * @fd: The KVM device file descriptor as returned from KVM_CREATE_DEVICE
211 * @type: The device-ctrl ioctl number
212 *
213 * Returns: -errno on error, nonnegative on success
214 */
215 int kvm_device_ioctl(int fd, int type, ...);
216
217 /**
218 * kvm_create_device - create a KVM device for the device control API
219 * @KVMState: The KVMState pointer
220 * @type: The KVM device type (see Documentation/virtual/kvm/devices in the
221 * kernel source)
222 * @test: If true, only test if device can be created, but don't actually
223 * create the device.
224 *
225 * Returns: -errno on error, nonnegative on success: @test ? 0 : device fd;
226 */
227 int kvm_create_device(KVMState *s, uint64_t type, bool test);
228
229
230 /* Arch specific hooks */
231
232 extern const KVMCapabilityInfo kvm_arch_required_capabilities[];
233
234 void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run);
235 void kvm_arch_post_run(CPUState *cpu, struct kvm_run *run);
236
237 int kvm_arch_handle_exit(CPUState *cpu, struct kvm_run *run);
238
239 int kvm_arch_process_async_events(CPUState *cpu);
240
241 int kvm_arch_get_registers(CPUState *cpu);
242
243 /* state subset only touched by the VCPU itself during runtime */
244 #define KVM_PUT_RUNTIME_STATE 1
245 /* state subset modified during VCPU reset */
246 #define KVM_PUT_RESET_STATE 2
247 /* full state set, modified during initialization or on vmload */
248 #define KVM_PUT_FULL_STATE 3
249
250 int kvm_arch_put_registers(CPUState *cpu, int level);
251
252 int kvm_arch_init(KVMState *s);
253
254 int kvm_arch_init_vcpu(CPUState *cpu);
255
256 /* Returns VCPU ID to be used on KVM_CREATE_VCPU ioctl() */
257 unsigned long kvm_arch_vcpu_id(CPUState *cpu);
258
259 int kvm_arch_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
260 int kvm_arch_on_sigbus(int code, void *addr);
261
262 void kvm_arch_init_irq_routing(KVMState *s);
263
264 int kvm_set_irq(KVMState *s, int irq, int level);
265 int kvm_irqchip_send_msi(KVMState *s, MSIMessage msg);
266
267 void kvm_irqchip_add_irq_route(KVMState *s, int gsi, int irqchip, int pin);
268 void kvm_irqchip_commit_routes(KVMState *s);
269
270 void kvm_put_apic_state(DeviceState *d, struct kvm_lapic_state *kapic);
271 void kvm_get_apic_state(DeviceState *d, struct kvm_lapic_state *kapic);
272
273 struct kvm_guest_debug;
274 struct kvm_debug_exit_arch;
275
276 struct kvm_sw_breakpoint {
277 target_ulong pc;
278 target_ulong saved_insn;
279 int use_count;
280 QTAILQ_ENTRY(kvm_sw_breakpoint) entry;
281 };
282
283 QTAILQ_HEAD(kvm_sw_breakpoint_head, kvm_sw_breakpoint);
284
285 struct kvm_sw_breakpoint *kvm_find_sw_breakpoint(CPUState *cpu,
286 target_ulong pc);
287
288 int kvm_sw_breakpoints_active(CPUState *cpu);
289
290 int kvm_arch_insert_sw_breakpoint(CPUState *cpu,
291 struct kvm_sw_breakpoint *bp);
292 int kvm_arch_remove_sw_breakpoint(CPUState *cpu,
293 struct kvm_sw_breakpoint *bp);
294 int kvm_arch_insert_hw_breakpoint(target_ulong addr,
295 target_ulong len, int type);
296 int kvm_arch_remove_hw_breakpoint(target_ulong addr,
297 target_ulong len, int type);
298 void kvm_arch_remove_all_hw_breakpoints(void);
299
300 void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg);
301
302 bool kvm_arch_stop_on_emulation_error(CPUState *cpu);
303
304 int kvm_check_extension(KVMState *s, unsigned int extension);
305
306 #define kvm_vm_enable_cap(s, capability, cap_flags, ...) \
307 ({ \
308 struct kvm_enable_cap cap = { \
309 .cap = capability, \
310 .flags = cap_flags, \
311 }; \
312 uint64_t args_tmp[] = { __VA_ARGS__ }; \
313 int i; \
314 for (i = 0; i < (int)ARRAY_SIZE(args_tmp) && \
315 i < ARRAY_SIZE(cap.args); i++) { \
316 cap.args[i] = args_tmp[i]; \
317 } \
318 kvm_vm_ioctl(s, KVM_ENABLE_CAP, &cap); \
319 })
320
321 #define kvm_vcpu_enable_cap(cpu, capability, cap_flags, ...) \
322 ({ \
323 struct kvm_enable_cap cap = { \
324 .cap = capability, \
325 .flags = cap_flags, \
326 }; \
327 uint64_t args_tmp[] = { __VA_ARGS__ }; \
328 int i; \
329 for (i = 0; i < (int)ARRAY_SIZE(args_tmp) && \
330 i < ARRAY_SIZE(cap.args); i++) { \
331 cap.args[i] = args_tmp[i]; \
332 } \
333 kvm_vcpu_ioctl(cpu, KVM_ENABLE_CAP, &cap); \
334 })
335
336 uint32_t kvm_arch_get_supported_cpuid(KVMState *env, uint32_t function,
337 uint32_t index, int reg);
338
339 #if !defined(CONFIG_USER_ONLY)
340 int kvm_physical_memory_addr_from_host(KVMState *s, void *ram_addr,
341 hwaddr *phys_addr);
342 #endif
343
344 #endif /* NEED_CPU_H */
345
346 void kvm_cpu_synchronize_state(CPUState *cpu);
347 void kvm_cpu_synchronize_post_reset(CPUState *cpu);
348 void kvm_cpu_synchronize_post_init(CPUState *cpu);
349
350 /* generic hooks - to be moved/refactored once there are more users */
351
352 static inline void cpu_synchronize_state(CPUState *cpu)
353 {
354 if (kvm_enabled()) {
355 kvm_cpu_synchronize_state(cpu);
356 }
357 }
358
359 static inline void cpu_synchronize_post_reset(CPUState *cpu)
360 {
361 if (kvm_enabled()) {
362 kvm_cpu_synchronize_post_reset(cpu);
363 }
364 }
365
366 static inline void cpu_synchronize_post_init(CPUState *cpu)
367 {
368 if (kvm_enabled()) {
369 kvm_cpu_synchronize_post_init(cpu);
370 }
371 }
372
373 int kvm_irqchip_add_msi_route(KVMState *s, MSIMessage msg);
374 int kvm_irqchip_update_msi_route(KVMState *s, int virq, MSIMessage msg);
375 void kvm_irqchip_release_virq(KVMState *s, int virq);
376
377 int kvm_irqchip_add_adapter_route(KVMState *s, AdapterInfo *adapter);
378
379 int kvm_irqchip_add_irqfd_notifier(KVMState *s, EventNotifier *n,
380 EventNotifier *rn, int virq);
381 int kvm_irqchip_remove_irqfd_notifier(KVMState *s, EventNotifier *n, int virq);
382 void kvm_pc_gsi_handler(void *opaque, int n, int level);
383 void kvm_pc_setup_irq_routing(bool pci_enabled);
384 void kvm_init_irq_routing(KVMState *s);
385
386 /**
387 * kvm_arch_irqchip_create:
388 * @KVMState: The KVMState pointer
389 *
390 * Allow architectures to create an in-kernel irq chip themselves.
391 *
392 * Returns: < 0: error
393 * 0: irq chip was not created
394 * > 0: irq chip was created
395 */
396 int kvm_arch_irqchip_create(KVMState *s);
397
398 /**
399 * kvm_set_one_reg - set a register value in KVM via KVM_SET_ONE_REG ioctl
400 * @id: The register ID
401 * @source: The pointer to the value to be set. It must point to a variable
402 * of the correct type/size for the register being accessed.
403 *
404 * Returns: 0 on success, or a negative errno on failure.
405 */
406 int kvm_set_one_reg(CPUState *cs, uint64_t id, void *source);
407
408 /**
409 * kvm_get_one_reg - get a register value from KVM via KVM_GET_ONE_REG ioctl
410 * @id: The register ID
411 * @target: The pointer where the value is to be stored. It must point to a
412 * variable of the correct type/size for the register being accessed.
413 *
414 * Returns: 0 on success, or a negative errno on failure.
415 */
416 int kvm_get_one_reg(CPUState *cs, uint64_t id, void *target);
417 #endif