]>
Commit | Line | Data |
---|---|---|
20c8ccb1 | 1 | /* SPDX-License-Identifier: GPL-2.0-only */ |
edf88417 AK |
2 | #ifndef __KVM_HOST_H |
3 | #define __KVM_HOST_H | |
6aa8b732 | 4 | |
6aa8b732 AK |
5 | |
6 | #include <linux/types.h> | |
e56a7a28 | 7 | #include <linux/hardirq.h> |
6aa8b732 AK |
8 | #include <linux/list.h> |
9 | #include <linux/mutex.h> | |
10 | #include <linux/spinlock.h> | |
06ff0d37 MR |
11 | #include <linux/signal.h> |
12 | #include <linux/sched.h> | |
187f1882 | 13 | #include <linux/bug.h> |
4a42d848 | 14 | #include <linux/minmax.h> |
6aa8b732 | 15 | #include <linux/mm.h> |
b297e672 | 16 | #include <linux/mmu_notifier.h> |
15ad7146 | 17 | #include <linux/preempt.h> |
0937c48d | 18 | #include <linux/msi.h> |
d89f5eff | 19 | #include <linux/slab.h> |
d1e5b0e9 | 20 | #include <linux/vmalloc.h> |
bd2b53b2 | 21 | #include <linux/rcupdate.h> |
bd80158a | 22 | #include <linux/ratelimit.h> |
83f09228 | 23 | #include <linux/err.h> |
c11f11fc | 24 | #include <linux/irqflags.h> |
521921ba | 25 | #include <linux/context_tracking.h> |
1a02b270 | 26 | #include <linux/irqbypass.h> |
da4ad88c | 27 | #include <linux/rcuwait.h> |
e3736c3e | 28 | #include <linux/refcount.h> |
1d487e9b | 29 | #include <linux/nospec.h> |
e8edc6e0 | 30 | #include <asm/signal.h> |
6aa8b732 | 31 | |
6aa8b732 | 32 | #include <linux/kvm.h> |
102d8325 | 33 | #include <linux/kvm_para.h> |
6aa8b732 | 34 | |
edf88417 | 35 | #include <linux/kvm_types.h> |
d77a39d9 | 36 | |
edf88417 | 37 | #include <asm/kvm_host.h> |
fb04a1ed | 38 | #include <linux/kvm_dirty_ring.h> |
d657a98e | 39 | |
0b1b1dfd GK |
40 | #ifndef KVM_MAX_VCPU_ID |
41 | #define KVM_MAX_VCPU_ID KVM_MAX_VCPUS | |
42 | #endif | |
43 | ||
67b29204 XG |
44 | /* |
45 | * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used | |
46 | * in kvm, other bits are visible for userspace which are defined in | |
47 | * include/linux/kvm_h. | |
48 | */ | |
49 | #define KVM_MEMSLOT_INVALID (1UL << 16) | |
50 | ||
361209e0 | 51 | /* |
164bf7e5 | 52 | * Bit 63 of the memslot generation number is an "update in-progress flag", |
361209e0 SC |
53 | * e.g. is temporarily set for the duration of install_new_memslots(). |
54 | * This flag effectively creates a unique generation number that is used to | |
55 | * mark cached memslot data, e.g. MMIO accesses, as potentially being stale, | |
56 | * i.e. may (or may not) have come from the previous memslots generation. | |
57 | * | |
58 | * This is necessary because the actual memslots update is not atomic with | |
59 | * respect to the generation number update. Updating the generation number | |
60 | * first would allow a vCPU to cache a spte from the old memslots using the | |
61 | * new generation number, and updating the generation number after switching | |
62 | * to the new memslots would allow cache hits using the old generation number | |
63 | * to reference the defunct memslots. | |
64 | * | |
65 | * This mechanism is used to prevent getting hits in KVM's caches while a | |
66 | * memslot update is in-progress, and to prevent cache hits *after* updating | |
67 | * the actual generation number against accesses that were inserted into the | |
68 | * cache *before* the memslots were updated. | |
69 | */ | |
164bf7e5 | 70 | #define KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS BIT_ULL(63) |
361209e0 | 71 | |
87da7e66 XG |
72 | /* Two fragments for cross MMIO pages. */ |
73 | #define KVM_MAX_MMIO_FRAGMENTS 2 | |
f78146b0 | 74 | |
f481b069 PB |
75 | #ifndef KVM_ADDRESS_SPACE_NUM |
76 | #define KVM_ADDRESS_SPACE_NUM 1 | |
77 | #endif | |
78 | ||
9c5b1172 XG |
79 | /* |
80 | * For the normal pfn, the highest 12 bits should be zero, | |
81c52c56 XG |
81 | * so we can mask bit 62 ~ bit 52 to indicate the error pfn, |
82 | * mask bit 63 to indicate the noslot pfn. | |
9c5b1172 | 83 | */ |
81c52c56 XG |
84 | #define KVM_PFN_ERR_MASK (0x7ffULL << 52) |
85 | #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52) | |
86 | #define KVM_PFN_NOSLOT (0x1ULL << 63) | |
9c5b1172 XG |
87 | |
88 | #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK) | |
89 | #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1) | |
81c52c56 | 90 | #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2) |
6c8ee57b | 91 | |
81c52c56 XG |
92 | /* |
93 | * error pfns indicate that the gfn is in slot but faild to | |
94 | * translate it to pfn on host. | |
95 | */ | |
ba049e93 | 96 | static inline bool is_error_pfn(kvm_pfn_t pfn) |
83f09228 | 97 | { |
9c5b1172 | 98 | return !!(pfn & KVM_PFN_ERR_MASK); |
83f09228 XG |
99 | } |
100 | ||
81c52c56 XG |
101 | /* |
102 | * error_noslot pfns indicate that the gfn can not be | |
103 | * translated to pfn - it is not in slot or failed to | |
104 | * translate it to pfn. | |
105 | */ | |
ba049e93 | 106 | static inline bool is_error_noslot_pfn(kvm_pfn_t pfn) |
83f09228 | 107 | { |
81c52c56 | 108 | return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK); |
83f09228 XG |
109 | } |
110 | ||
81c52c56 | 111 | /* noslot pfn indicates that the gfn is not in slot. */ |
ba049e93 | 112 | static inline bool is_noslot_pfn(kvm_pfn_t pfn) |
83f09228 | 113 | { |
81c52c56 | 114 | return pfn == KVM_PFN_NOSLOT; |
83f09228 XG |
115 | } |
116 | ||
bf640876 DD |
117 | /* |
118 | * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390) | |
119 | * provide own defines and kvm_is_error_hva | |
120 | */ | |
121 | #ifndef KVM_HVA_ERR_BAD | |
122 | ||
7068d097 XG |
123 | #define KVM_HVA_ERR_BAD (PAGE_OFFSET) |
124 | #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE) | |
ca3a490c XG |
125 | |
126 | static inline bool kvm_is_error_hva(unsigned long addr) | |
127 | { | |
7068d097 | 128 | return addr >= PAGE_OFFSET; |
ca3a490c XG |
129 | } |
130 | ||
bf640876 DD |
131 | #endif |
132 | ||
6cede2e6 XG |
133 | #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT)) |
134 | ||
9c5b1172 | 135 | static inline bool is_error_page(struct page *page) |
6cede2e6 XG |
136 | { |
137 | return IS_ERR(page); | |
138 | } | |
139 | ||
930f7fd6 RK |
140 | #define KVM_REQUEST_MASK GENMASK(7,0) |
141 | #define KVM_REQUEST_NO_WAKEUP BIT(8) | |
7a97cec2 | 142 | #define KVM_REQUEST_WAIT BIT(9) |
d9e368d6 | 143 | /* |
2860c4b1 PB |
144 | * Architecture-independent vcpu->requests bit members |
145 | * Bits 4-7 are reserved for more arch-independent bits. | |
d9e368d6 | 146 | */ |
7a97cec2 PB |
147 | #define KVM_REQ_TLB_FLUSH (0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP) |
148 | #define KVM_REQ_MMU_RELOAD (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP) | |
149 | #define KVM_REQ_PENDING_TIMER 2 | |
150 | #define KVM_REQ_UNHALT 3 | |
2387149e AJ |
151 | #define KVM_REQUEST_ARCH_BASE 8 |
152 | ||
153 | #define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \ | |
c593642c | 154 | BUILD_BUG_ON((unsigned)(nr) >= (sizeof_field(struct kvm_vcpu, requests) * 8) - KVM_REQUEST_ARCH_BASE); \ |
2387149e AJ |
155 | (unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) | (flags)); \ |
156 | }) | |
157 | #define KVM_ARCH_REQ(nr) KVM_ARCH_REQ_FLAGS(nr, 0) | |
0cd31043 | 158 | |
7a84428a AW |
159 | #define KVM_USERSPACE_IRQ_SOURCE_ID 0 |
160 | #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1 | |
5550af4d | 161 | |
0d9ce162 | 162 | extern struct mutex kvm_lock; |
fc1b7492 GL |
163 | extern struct list_head vm_list; |
164 | ||
743eeb0b SL |
165 | struct kvm_io_range { |
166 | gpa_t addr; | |
167 | int len; | |
168 | struct kvm_io_device *dev; | |
169 | }; | |
170 | ||
786a9f88 | 171 | #define NR_IOBUS_DEVS 1000 |
a1300716 | 172 | |
2eeb2e94 | 173 | struct kvm_io_bus { |
6ea34c9b AK |
174 | int dev_count; |
175 | int ioeventfd_count; | |
a1300716 | 176 | struct kvm_io_range range[]; |
2eeb2e94 GH |
177 | }; |
178 | ||
e93f8a0f MT |
179 | enum kvm_bus { |
180 | KVM_MMIO_BUS, | |
181 | KVM_PIO_BUS, | |
060f0ce6 | 182 | KVM_VIRTIO_CCW_NOTIFY_BUS, |
68c3b4d1 | 183 | KVM_FAST_MMIO_BUS, |
e93f8a0f MT |
184 | KVM_NR_BUSES |
185 | }; | |
186 | ||
e32edf4f | 187 | int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr, |
e93f8a0f | 188 | int len, const void *val); |
e32edf4f NN |
189 | int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, |
190 | gpa_t addr, int len, const void *val, long cookie); | |
191 | int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr, | |
192 | int len, void *val); | |
743eeb0b SL |
193 | int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
194 | int len, struct kvm_io_device *dev); | |
5d3c4c79 SC |
195 | int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx, |
196 | struct kvm_io_device *dev); | |
8a39d006 AP |
197 | struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx, |
198 | gpa_t addr); | |
2eeb2e94 | 199 | |
af585b92 GN |
200 | #ifdef CONFIG_KVM_ASYNC_PF |
201 | struct kvm_async_pf { | |
202 | struct work_struct work; | |
203 | struct list_head link; | |
204 | struct list_head queue; | |
205 | struct kvm_vcpu *vcpu; | |
206 | struct mm_struct *mm; | |
736c291c | 207 | gpa_t cr2_or_gpa; |
af585b92 GN |
208 | unsigned long addr; |
209 | struct kvm_arch_async_pf arch; | |
f2e10669 | 210 | bool wakeup_all; |
2a18b7e7 | 211 | bool notpresent_injected; |
af585b92 GN |
212 | }; |
213 | ||
214 | void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu); | |
215 | void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu); | |
e8c22266 VK |
216 | bool kvm_setup_async_pf(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, |
217 | unsigned long hva, struct kvm_arch_async_pf *arch); | |
344d9588 | 218 | int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu); |
af585b92 GN |
219 | #endif |
220 | ||
5f7c292b | 221 | #ifdef KVM_ARCH_WANT_MMU_NOTIFIER |
3039bcc7 SC |
222 | struct kvm_gfn_range { |
223 | struct kvm_memory_slot *slot; | |
224 | gfn_t start; | |
225 | gfn_t end; | |
226 | pte_t pte; | |
227 | bool may_block; | |
228 | }; | |
229 | bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range); | |
230 | bool kvm_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range); | |
231 | bool kvm_test_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range); | |
232 | bool kvm_set_spte_gfn(struct kvm *kvm, struct kvm_gfn_range *range); | |
5f7c292b SC |
233 | #endif |
234 | ||
6b7e2d09 XG |
235 | enum { |
236 | OUTSIDE_GUEST_MODE, | |
237 | IN_GUEST_MODE, | |
c142786c AK |
238 | EXITING_GUEST_MODE, |
239 | READING_SHADOW_PAGE_TABLES, | |
6b7e2d09 XG |
240 | }; |
241 | ||
e45adf66 KA |
242 | #define KVM_UNMAPPED_PAGE ((void *) 0x500 + POISON_POINTER_DELTA) |
243 | ||
244 | struct kvm_host_map { | |
245 | /* | |
246 | * Only valid if the 'pfn' is managed by the host kernel (i.e. There is | |
247 | * a 'struct page' for it. When using mem= kernel parameter some memory | |
248 | * can be used as guest memory but they are not managed by host | |
249 | * kernel). | |
250 | * If 'pfn' is not managed by the host kernel, this field is | |
251 | * initialized to KVM_UNMAPPED_PAGE. | |
252 | */ | |
253 | struct page *page; | |
254 | void *hva; | |
255 | kvm_pfn_t pfn; | |
256 | kvm_pfn_t gfn; | |
257 | }; | |
258 | ||
259 | /* | |
260 | * Used to check if the mapping is valid or not. Never use 'kvm_host_map' | |
261 | * directly to check for that. | |
262 | */ | |
263 | static inline bool kvm_vcpu_mapped(struct kvm_host_map *map) | |
264 | { | |
265 | return !!map->hva; | |
266 | } | |
267 | ||
f78146b0 AK |
268 | /* |
269 | * Sometimes a large or cross-page mmio needs to be broken up into separate | |
270 | * exits for userspace servicing. | |
271 | */ | |
272 | struct kvm_mmio_fragment { | |
273 | gpa_t gpa; | |
274 | void *data; | |
275 | unsigned len; | |
276 | }; | |
277 | ||
d17fbbf7 ZX |
278 | struct kvm_vcpu { |
279 | struct kvm *kvm; | |
31bb117e | 280 | #ifdef CONFIG_PREEMPT_NOTIFIERS |
d17fbbf7 | 281 | struct preempt_notifier preempt_notifier; |
31bb117e | 282 | #endif |
6b7e2d09 | 283 | int cpu; |
8750e72a RK |
284 | int vcpu_id; /* id given by userspace at creation */ |
285 | int vcpu_idx; /* index in kvm->vcpus array */ | |
6b7e2d09 XG |
286 | int srcu_idx; |
287 | int mode; | |
86dafed5 | 288 | u64 requests; |
d0bfb940 | 289 | unsigned long guest_debug; |
6b7e2d09 | 290 | |
bf9f6ac8 FW |
291 | int pre_pcpu; |
292 | struct list_head blocked_vcpu_list; | |
293 | ||
6b7e2d09 XG |
294 | struct mutex mutex; |
295 | struct kvm_run *run; | |
f656ce01 | 296 | |
da4ad88c | 297 | struct rcuwait wait; |
0e4524a5 | 298 | struct pid __rcu *pid; |
d17fbbf7 ZX |
299 | int sigset_active; |
300 | sigset_t sigset; | |
301 | struct kvm_vcpu_stat stat; | |
19020f8a | 302 | unsigned int halt_poll_ns; |
3491caf2 | 303 | bool valid_wakeup; |
d17fbbf7 | 304 | |
34c16eec | 305 | #ifdef CONFIG_HAS_IOMEM |
d17fbbf7 ZX |
306 | int mmio_needed; |
307 | int mmio_read_completed; | |
308 | int mmio_is_write; | |
f78146b0 AK |
309 | int mmio_cur_fragment; |
310 | int mmio_nr_fragments; | |
311 | struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS]; | |
34c16eec | 312 | #endif |
1165f5fe | 313 | |
af585b92 GN |
314 | #ifdef CONFIG_KVM_ASYNC_PF |
315 | struct { | |
316 | u32 queued; | |
317 | struct list_head queue; | |
318 | struct list_head done; | |
319 | spinlock_t lock; | |
320 | } async_pf; | |
321 | #endif | |
322 | ||
4c088493 R |
323 | #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT |
324 | /* | |
325 | * Cpu relax intercept or pause loop exit optimization | |
326 | * in_spin_loop: set when a vcpu does a pause loop exit | |
327 | * or cpu relax intercepted. | |
328 | * dy_eligible: indicates whether vcpu is eligible for directed yield. | |
329 | */ | |
330 | struct { | |
331 | bool in_spin_loop; | |
332 | bool dy_eligible; | |
333 | } spin_loop; | |
334 | #endif | |
3a08a8f9 | 335 | bool preempted; |
d73eb57b | 336 | bool ready; |
d657a98e | 337 | struct kvm_vcpu_arch arch; |
fb04a1ed | 338 | struct kvm_dirty_ring dirty_ring; |
d657a98e ZX |
339 | }; |
340 | ||
1ca0016c SC |
341 | /* must be called with irqs disabled */ |
342 | static __always_inline void guest_enter_irqoff(void) | |
343 | { | |
344 | /* | |
345 | * This is running in ioctl context so its safe to assume that it's the | |
346 | * stime pending cputime to flush. | |
347 | */ | |
348 | instrumentation_begin(); | |
349 | vtime_account_guest_enter(); | |
350 | instrumentation_end(); | |
351 | ||
352 | /* | |
353 | * KVM does not hold any references to rcu protected data when it | |
354 | * switches CPU into a guest mode. In fact switching to a guest mode | |
355 | * is very similar to exiting to userspace from rcu point of view. In | |
356 | * addition CPU may stay in a guest mode for quite a long time (up to | |
357 | * one time slice). Lets treat guest mode as quiescent state, just like | |
358 | * we do with user-mode execution. | |
359 | */ | |
360 | if (!context_tracking_guest_enter()) { | |
361 | instrumentation_begin(); | |
362 | rcu_virt_note_context_switch(smp_processor_id()); | |
363 | instrumentation_end(); | |
364 | } | |
365 | } | |
366 | ||
367 | static __always_inline void guest_exit_irqoff(void) | |
368 | { | |
369 | context_tracking_guest_exit(); | |
370 | ||
371 | instrumentation_begin(); | |
372 | /* Flush the guest cputime we spent on the guest */ | |
373 | vtime_account_guest_exit(); | |
374 | instrumentation_end(); | |
375 | } | |
376 | ||
377 | static inline void guest_exit(void) | |
378 | { | |
379 | unsigned long flags; | |
380 | ||
381 | local_irq_save(flags); | |
382 | guest_exit_irqoff(); | |
383 | local_irq_restore(flags); | |
384 | } | |
385 | ||
6b7e2d09 XG |
386 | static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu) |
387 | { | |
cde9af6e AJ |
388 | /* |
389 | * The memory barrier ensures a previous write to vcpu->requests cannot | |
390 | * be reordered with the read of vcpu->mode. It pairs with the general | |
391 | * memory barrier following the write of vcpu->mode in VCPU RUN. | |
392 | */ | |
393 | smp_mb__before_atomic(); | |
6b7e2d09 XG |
394 | return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE); |
395 | } | |
396 | ||
660c22c4 TY |
397 | /* |
398 | * Some of the bitops functions do not support too long bitmaps. | |
399 | * This number must be determined not to exceed such limits. | |
400 | */ | |
401 | #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1) | |
402 | ||
6aa8b732 AK |
403 | struct kvm_memory_slot { |
404 | gfn_t base_gfn; | |
405 | unsigned long npages; | |
6aa8b732 | 406 | unsigned long *dirty_bitmap; |
db3fe4eb | 407 | struct kvm_arch_memory_slot arch; |
8a7ae055 | 408 | unsigned long userspace_addr; |
6104f472 | 409 | u32 flags; |
1e702d9a | 410 | short id; |
9e9eb226 | 411 | u16 as_id; |
6aa8b732 AK |
412 | }; |
413 | ||
044c59c4 PX |
414 | static inline bool kvm_slot_dirty_track_enabled(struct kvm_memory_slot *slot) |
415 | { | |
416 | return slot->flags & KVM_MEM_LOG_DIRTY_PAGES; | |
417 | } | |
418 | ||
87bf6e7d TY |
419 | static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot) |
420 | { | |
421 | return ALIGN(memslot->npages, BITS_PER_LONG) / 8; | |
422 | } | |
423 | ||
03133347 CI |
424 | static inline unsigned long *kvm_second_dirty_bitmap(struct kvm_memory_slot *memslot) |
425 | { | |
426 | unsigned long len = kvm_dirty_bitmap_bytes(memslot); | |
427 | ||
428 | return memslot->dirty_bitmap + len / sizeof(*memslot->dirty_bitmap); | |
429 | } | |
430 | ||
3c9bd400 JZ |
431 | #ifndef KVM_DIRTY_LOG_MANUAL_CAPS |
432 | #define KVM_DIRTY_LOG_MANUAL_CAPS KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE | |
433 | #endif | |
434 | ||
84223598 CH |
435 | struct kvm_s390_adapter_int { |
436 | u64 ind_addr; | |
437 | u64 summary_addr; | |
438 | u64 ind_offset; | |
439 | u32 summary_offset; | |
440 | u32 adapter_id; | |
441 | }; | |
442 | ||
5c919412 AS |
443 | struct kvm_hv_sint { |
444 | u32 vcpu; | |
445 | u32 sint; | |
446 | }; | |
447 | ||
399ec807 AK |
448 | struct kvm_kernel_irq_routing_entry { |
449 | u32 gsi; | |
5116d8f6 | 450 | u32 type; |
4925663a | 451 | int (*set)(struct kvm_kernel_irq_routing_entry *e, |
aa2fbe6d YZ |
452 | struct kvm *kvm, int irq_source_id, int level, |
453 | bool line_status); | |
399ec807 AK |
454 | union { |
455 | struct { | |
456 | unsigned irqchip; | |
457 | unsigned pin; | |
458 | } irqchip; | |
0455e72c EA |
459 | struct { |
460 | u32 address_lo; | |
461 | u32 address_hi; | |
462 | u32 data; | |
463 | u32 flags; | |
464 | u32 devid; | |
465 | } msi; | |
84223598 | 466 | struct kvm_s390_adapter_int adapter; |
5c919412 | 467 | struct kvm_hv_sint hv_sint; |
399ec807 | 468 | }; |
46e624b9 GN |
469 | struct hlist_node link; |
470 | }; | |
471 | ||
b053b2ae SR |
472 | #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING |
473 | struct kvm_irq_routing_table { | |
474 | int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS]; | |
475 | u32 nr_rt_entries; | |
476 | /* | |
477 | * Array indexed by gsi. Each entry contains list of irq chips | |
478 | * the gsi is connected to. | |
479 | */ | |
764e515f | 480 | struct hlist_head map[]; |
b053b2ae SR |
481 | }; |
482 | #endif | |
483 | ||
0743247f AW |
484 | #ifndef KVM_PRIVATE_MEM_SLOTS |
485 | #define KVM_PRIVATE_MEM_SLOTS 0 | |
486 | #endif | |
487 | ||
4fc096a9 VK |
488 | #define KVM_MEM_SLOTS_NUM SHRT_MAX |
489 | #define KVM_USER_MEM_SLOTS (KVM_MEM_SLOTS_NUM - KVM_PRIVATE_MEM_SLOTS) | |
93a5cef0 | 490 | |
f481b069 PB |
491 | #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE |
492 | static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu) | |
493 | { | |
494 | return 0; | |
495 | } | |
496 | #endif | |
497 | ||
bf3e05bc XG |
498 | /* |
499 | * Note: | |
500 | * memslots are not sorted by id anymore, please use id_to_memslot() | |
501 | * to get the memslot by its id. | |
502 | */ | |
46a26bf5 | 503 | struct kvm_memslots { |
49c7754c | 504 | u64 generation; |
f85e2cb5 | 505 | /* The mapping table from slot id to the index in memslots[]. */ |
1e702d9a | 506 | short id_to_index[KVM_MEM_SLOTS_NUM]; |
d4ae84a0 | 507 | atomic_t lru_slot; |
9c1a5d38 | 508 | int used_slots; |
36947254 | 509 | struct kvm_memory_slot memslots[]; |
46a26bf5 MT |
510 | }; |
511 | ||
6aa8b732 | 512 | struct kvm { |
531810ca BG |
513 | #ifdef KVM_HAVE_MMU_RWLOCK |
514 | rwlock_t mmu_lock; | |
515 | #else | |
aaee2c94 | 516 | spinlock_t mmu_lock; |
531810ca BG |
517 | #endif /* KVM_HAVE_MMU_RWLOCK */ |
518 | ||
79fac95e | 519 | struct mutex slots_lock; |
6d4e4c4f | 520 | struct mm_struct *mm; /* userspace tied to this vm */ |
a80cf7b5 | 521 | struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM]; |
fb3f0f51 | 522 | struct kvm_vcpu *vcpus[KVM_MAX_VCPUS]; |
6c7caebc PB |
523 | |
524 | /* | |
525 | * created_vcpus is protected by kvm->lock, and is incremented | |
526 | * at the beginning of KVM_CREATE_VCPU. online_vcpus is only | |
527 | * incremented after storing the kvm_vcpu pointer in vcpus, | |
528 | * and is accessed atomically. | |
529 | */ | |
73880c80 | 530 | atomic_t online_vcpus; |
6c7caebc | 531 | int created_vcpus; |
217ece61 | 532 | int last_boosted_vcpu; |
133de902 | 533 | struct list_head vm_list; |
60eead79 | 534 | struct mutex lock; |
4a12f951 | 535 | struct kvm_io_bus __rcu *buses[KVM_NR_BUSES]; |
721eecbf GH |
536 | #ifdef CONFIG_HAVE_KVM_EVENTFD |
537 | struct { | |
538 | spinlock_t lock; | |
539 | struct list_head items; | |
7a84428a AW |
540 | struct list_head resampler_list; |
541 | struct mutex resampler_lock; | |
721eecbf | 542 | } irqfds; |
d34e6b17 | 543 | struct list_head ioeventfds; |
721eecbf | 544 | #endif |
ba1389b7 | 545 | struct kvm_vm_stat stat; |
d69fb81f | 546 | struct kvm_arch arch; |
e3736c3e | 547 | refcount_t users_count; |
4b4357e0 | 548 | #ifdef CONFIG_KVM_MMIO |
5f94c174 | 549 | struct kvm_coalesced_mmio_ring *coalesced_mmio_ring; |
2b3c246a SL |
550 | spinlock_t ring_lock; |
551 | struct list_head coalesced_zones; | |
5f94c174 | 552 | #endif |
e930bffe | 553 | |
60eead79 | 554 | struct mutex irq_lock; |
75858a84 | 555 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
bd2b53b2 | 556 | /* |
9957c86d | 557 | * Update side is protected by irq_lock. |
bd2b53b2 | 558 | */ |
4b6a2872 | 559 | struct kvm_irq_routing_table __rcu *irq_routing; |
c77dcacb PB |
560 | #endif |
561 | #ifdef CONFIG_HAVE_KVM_IRQFD | |
136bdfee | 562 | struct hlist_head irq_ack_notifier_list; |
75858a84 AK |
563 | #endif |
564 | ||
36c1ed82 | 565 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
e930bffe AA |
566 | struct mmu_notifier mmu_notifier; |
567 | unsigned long mmu_notifier_seq; | |
568 | long mmu_notifier_count; | |
4a42d848 DS |
569 | unsigned long mmu_notifier_range_start; |
570 | unsigned long mmu_notifier_range_end; | |
e930bffe | 571 | #endif |
a086f6a1 | 572 | long tlbs_dirty; |
07f0a7bd | 573 | struct list_head devices; |
3c9bd400 | 574 | u64 manual_dirty_log_protect; |
536a6f88 JF |
575 | struct dentry *debugfs_dentry; |
576 | struct kvm_stat_data **debugfs_stat_data; | |
6ade8694 PM |
577 | struct srcu_struct srcu; |
578 | struct srcu_struct irq_srcu; | |
fdeaf7e3 | 579 | pid_t userspace_pid; |
acd05785 | 580 | unsigned int max_halt_poll_ns; |
fb04a1ed | 581 | u32 dirty_ring_size; |
6aa8b732 AK |
582 | }; |
583 | ||
a737f256 CD |
584 | #define kvm_err(fmt, ...) \ |
585 | pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__) | |
586 | #define kvm_info(fmt, ...) \ | |
587 | pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__) | |
588 | #define kvm_debug(fmt, ...) \ | |
589 | pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__) | |
ae0f5499 BD |
590 | #define kvm_debug_ratelimited(fmt, ...) \ |
591 | pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \ | |
592 | ## __VA_ARGS__) | |
a737f256 CD |
593 | #define kvm_pr_unimpl(fmt, ...) \ |
594 | pr_err_ratelimited("kvm [%i]: " fmt, \ | |
595 | task_tgid_nr(current), ## __VA_ARGS__) | |
f0242478 | 596 | |
a737f256 CD |
597 | /* The guest did something we don't support. */ |
598 | #define vcpu_unimpl(vcpu, fmt, ...) \ | |
671d9ab3 BP |
599 | kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \ |
600 | (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__) | |
6aa8b732 | 601 | |
ee86dbc6 AS |
602 | #define vcpu_debug(vcpu, fmt, ...) \ |
603 | kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__) | |
ae0f5499 BD |
604 | #define vcpu_debug_ratelimited(vcpu, fmt, ...) \ |
605 | kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \ | |
606 | ## __VA_ARGS__) | |
765eaa0f AS |
607 | #define vcpu_err(vcpu, fmt, ...) \ |
608 | kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__) | |
ee86dbc6 | 609 | |
3c9bd400 JZ |
610 | static inline bool kvm_dirty_log_manual_protect_and_init_set(struct kvm *kvm) |
611 | { | |
612 | return !!(kvm->manual_dirty_log_protect & KVM_DIRTY_LOG_INITIALLY_SET); | |
613 | } | |
614 | ||
4a12f951 CB |
615 | static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx) |
616 | { | |
617 | return srcu_dereference_check(kvm->buses[idx], &kvm->srcu, | |
3898da94 PB |
618 | lockdep_is_held(&kvm->slots_lock) || |
619 | !refcount_read(&kvm->users_count)); | |
4a12f951 CB |
620 | } |
621 | ||
988a2cae GN |
622 | static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i) |
623 | { | |
1d487e9b PB |
624 | int num_vcpus = atomic_read(&kvm->online_vcpus); |
625 | i = array_index_nospec(i, num_vcpus); | |
626 | ||
627 | /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. */ | |
988a2cae GN |
628 | smp_rmb(); |
629 | return kvm->vcpus[i]; | |
630 | } | |
631 | ||
632 | #define kvm_for_each_vcpu(idx, vcpup, kvm) \ | |
b42fc3cb JM |
633 | for (idx = 0; \ |
634 | idx < atomic_read(&kvm->online_vcpus) && \ | |
635 | (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \ | |
636 | idx++) | |
988a2cae | 637 | |
db27a7a3 DH |
638 | static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id) |
639 | { | |
9b9e3fc4 | 640 | struct kvm_vcpu *vcpu = NULL; |
db27a7a3 DH |
641 | int i; |
642 | ||
9b9e3fc4 | 643 | if (id < 0) |
c896939f | 644 | return NULL; |
9b9e3fc4 GK |
645 | if (id < KVM_MAX_VCPUS) |
646 | vcpu = kvm_get_vcpu(kvm, id); | |
c896939f DH |
647 | if (vcpu && vcpu->vcpu_id == id) |
648 | return vcpu; | |
db27a7a3 DH |
649 | kvm_for_each_vcpu(i, vcpu, kvm) |
650 | if (vcpu->vcpu_id == id) | |
651 | return vcpu; | |
652 | return NULL; | |
653 | } | |
654 | ||
497d72d8 CD |
655 | static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu) |
656 | { | |
8750e72a | 657 | return vcpu->vcpu_idx; |
497d72d8 CD |
658 | } |
659 | ||
0577d1ab SC |
660 | #define kvm_for_each_memslot(memslot, slots) \ |
661 | for (memslot = &slots->memslots[0]; \ | |
662 | memslot < slots->memslots + slots->used_slots; memslot++) \ | |
663 | if (WARN_ON_ONCE(!memslot->npages)) { \ | |
664 | } else | |
be6ba0f0 | 665 | |
4543bdc0 | 666 | void kvm_vcpu_destroy(struct kvm_vcpu *vcpu); |
fb3f0f51 | 667 | |
ec7660cc | 668 | void vcpu_load(struct kvm_vcpu *vcpu); |
313a3dc7 CO |
669 | void vcpu_put(struct kvm_vcpu *vcpu); |
670 | ||
6ef768fa | 671 | #ifdef __KVM_HAVE_IOAPIC |
993225ad | 672 | void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm); |
abdb080f | 673 | void kvm_arch_post_irq_routing_update(struct kvm *kvm); |
6ef768fa | 674 | #else |
993225ad | 675 | static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm) |
6ef768fa PB |
676 | { |
677 | } | |
abdb080f | 678 | static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm) |
b053b2ae SR |
679 | { |
680 | } | |
6ef768fa PB |
681 | #endif |
682 | ||
297e2105 | 683 | #ifdef CONFIG_HAVE_KVM_IRQFD |
a0f155e9 CH |
684 | int kvm_irqfd_init(void); |
685 | void kvm_irqfd_exit(void); | |
686 | #else | |
687 | static inline int kvm_irqfd_init(void) | |
688 | { | |
689 | return 0; | |
690 | } | |
691 | ||
692 | static inline void kvm_irqfd_exit(void) | |
693 | { | |
694 | } | |
695 | #endif | |
0ee75bea | 696 | int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align, |
c16f862d | 697 | struct module *module); |
cb498ea2 | 698 | void kvm_exit(void); |
6aa8b732 | 699 | |
d39f13b0 IE |
700 | void kvm_get_kvm(struct kvm *kvm); |
701 | void kvm_put_kvm(struct kvm *kvm); | |
54526d1f | 702 | bool file_is_kvm(struct file *file); |
149487bd | 703 | void kvm_put_kvm_no_destroy(struct kvm *kvm); |
d39f13b0 | 704 | |
f481b069 | 705 | static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id) |
90d83dc3 | 706 | { |
1d487e9b | 707 | as_id = array_index_nospec(as_id, KVM_ADDRESS_SPACE_NUM); |
7e988b10 | 708 | return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu, |
3898da94 PB |
709 | lockdep_is_held(&kvm->slots_lock) || |
710 | !refcount_read(&kvm->users_count)); | |
90d83dc3 LJ |
711 | } |
712 | ||
f481b069 PB |
713 | static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm) |
714 | { | |
715 | return __kvm_memslots(kvm, 0); | |
716 | } | |
717 | ||
8e73485c PB |
718 | static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu) |
719 | { | |
f481b069 PB |
720 | int as_id = kvm_arch_vcpu_memslots_id(vcpu); |
721 | ||
722 | return __kvm_memslots(vcpu->kvm, as_id); | |
8e73485c PB |
723 | } |
724 | ||
0577d1ab SC |
725 | static inline |
726 | struct kvm_memory_slot *id_to_memslot(struct kvm_memslots *slots, int id) | |
28a37544 | 727 | { |
f85e2cb5 XG |
728 | int index = slots->id_to_index[id]; |
729 | struct kvm_memory_slot *slot; | |
bf3e05bc | 730 | |
0577d1ab SC |
731 | if (index < 0) |
732 | return NULL; | |
733 | ||
f85e2cb5 | 734 | slot = &slots->memslots[index]; |
bf3e05bc | 735 | |
f85e2cb5 XG |
736 | WARN_ON(slot->id != id); |
737 | return slot; | |
28a37544 XG |
738 | } |
739 | ||
74d0727c TY |
740 | /* |
741 | * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations: | |
742 | * - create a new memory slot | |
743 | * - delete an existing memory slot | |
744 | * - modify an existing memory slot | |
745 | * -- move it in the guest physical memory space | |
746 | * -- just change its flags | |
747 | * | |
748 | * Since flags can be changed by some of these operations, the following | |
749 | * differentiation is the best we can do for __kvm_set_memory_region(): | |
750 | */ | |
751 | enum kvm_mr_change { | |
752 | KVM_MR_CREATE, | |
753 | KVM_MR_DELETE, | |
754 | KVM_MR_MOVE, | |
755 | KVM_MR_FLAGS_ONLY, | |
756 | }; | |
757 | ||
210c7c4d | 758 | int kvm_set_memory_region(struct kvm *kvm, |
09170a49 | 759 | const struct kvm_userspace_memory_region *mem); |
f78e0e2e | 760 | int __kvm_set_memory_region(struct kvm *kvm, |
09170a49 | 761 | const struct kvm_userspace_memory_region *mem); |
e96c81ee | 762 | void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot); |
15248258 | 763 | void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen); |
f7784b8e MT |
764 | int kvm_arch_prepare_memory_region(struct kvm *kvm, |
765 | struct kvm_memory_slot *memslot, | |
09170a49 | 766 | const struct kvm_userspace_memory_region *mem, |
7b6195a9 | 767 | enum kvm_mr_change change); |
f7784b8e | 768 | void kvm_arch_commit_memory_region(struct kvm *kvm, |
09170a49 | 769 | const struct kvm_userspace_memory_region *mem, |
9d4c197c | 770 | struct kvm_memory_slot *old, |
f36f3f28 | 771 | const struct kvm_memory_slot *new, |
8482644a | 772 | enum kvm_mr_change change); |
2df72e9b MT |
773 | /* flush all memory translations */ |
774 | void kvm_arch_flush_shadow_all(struct kvm *kvm); | |
775 | /* flush memory translations pointing to 'slot' */ | |
776 | void kvm_arch_flush_shadow_memslot(struct kvm *kvm, | |
777 | struct kvm_memory_slot *slot); | |
a983fb23 | 778 | |
d9ef13c2 PB |
779 | int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn, |
780 | struct page **pages, int nr_pages); | |
48987781 | 781 | |
954bbbc2 | 782 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn); |
05da4558 | 783 | unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn); |
ba6a3541 | 784 | unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable); |
4d8b81ab | 785 | unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn); |
64d83126 CD |
786 | unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn, |
787 | bool *writable); | |
b4231d61 IE |
788 | void kvm_release_page_clean(struct page *page); |
789 | void kvm_release_page_dirty(struct page *page); | |
35149e21 AL |
790 | void kvm_set_page_accessed(struct page *page); |
791 | ||
ba049e93 DW |
792 | kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn); |
793 | kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault, | |
612819c3 | 794 | bool *writable); |
ba049e93 DW |
795 | kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn); |
796 | kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn); | |
797 | kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn, | |
798 | bool atomic, bool *async, bool write_fault, | |
4a42d848 | 799 | bool *writable, hva_t *hva); |
037d92dc | 800 | |
ba049e93 | 801 | void kvm_release_pfn_clean(kvm_pfn_t pfn); |
f7a6509f | 802 | void kvm_release_pfn_dirty(kvm_pfn_t pfn); |
ba049e93 DW |
803 | void kvm_set_pfn_dirty(kvm_pfn_t pfn); |
804 | void kvm_set_pfn_accessed(kvm_pfn_t pfn); | |
805 | void kvm_get_pfn(kvm_pfn_t pfn); | |
35149e21 | 806 | |
91724814 | 807 | void kvm_release_pfn(kvm_pfn_t pfn, bool dirty, struct gfn_to_pfn_cache *cache); |
195aefde IE |
808 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, |
809 | int len); | |
810 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len); | |
4e335d9e PB |
811 | int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc, |
812 | void *data, unsigned long len); | |
0958f0ce VK |
813 | int kvm_read_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc, |
814 | void *data, unsigned int offset, | |
815 | unsigned long len); | |
195aefde IE |
816 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, |
817 | int offset, int len); | |
818 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
819 | unsigned long len); | |
4e335d9e PB |
820 | int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc, |
821 | void *data, unsigned long len); | |
822 | int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc, | |
7a86dab8 JM |
823 | void *data, unsigned int offset, |
824 | unsigned long len); | |
4e335d9e PB |
825 | int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc, |
826 | gpa_t gpa, unsigned long len); | |
cac0f1b7 | 827 | |
53f98558 AJ |
828 | #define __kvm_get_guest(kvm, gfn, offset, v) \ |
829 | ({ \ | |
830 | unsigned long __addr = gfn_to_hva(kvm, gfn); \ | |
831 | typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \ | |
832 | int __ret = -EFAULT; \ | |
833 | \ | |
834 | if (!kvm_is_error_hva(__addr)) \ | |
835 | __ret = get_user(v, __uaddr); \ | |
836 | __ret; \ | |
837 | }) | |
838 | ||
839 | #define kvm_get_guest(kvm, gpa, v) \ | |
840 | ({ \ | |
841 | gpa_t __gpa = gpa; \ | |
842 | struct kvm *__kvm = kvm; \ | |
843 | \ | |
844 | __kvm_get_guest(__kvm, __gpa >> PAGE_SHIFT, \ | |
845 | offset_in_page(__gpa), v); \ | |
846 | }) | |
847 | ||
4d2d4ce0 | 848 | #define __kvm_put_guest(kvm, gfn, offset, v) \ |
cac0f1b7 SP |
849 | ({ \ |
850 | unsigned long __addr = gfn_to_hva(kvm, gfn); \ | |
4d2d4ce0 | 851 | typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \ |
cac0f1b7 SP |
852 | int __ret = -EFAULT; \ |
853 | \ | |
854 | if (!kvm_is_error_hva(__addr)) \ | |
4d2d4ce0 | 855 | __ret = put_user(v, __uaddr); \ |
cac0f1b7 SP |
856 | if (!__ret) \ |
857 | mark_page_dirty(kvm, gfn); \ | |
858 | __ret; \ | |
859 | }) | |
860 | ||
4d2d4ce0 | 861 | #define kvm_put_guest(kvm, gpa, v) \ |
cac0f1b7 SP |
862 | ({ \ |
863 | gpa_t __gpa = gpa; \ | |
864 | struct kvm *__kvm = kvm; \ | |
4d2d4ce0 | 865 | \ |
cac0f1b7 | 866 | __kvm_put_guest(__kvm, __gpa >> PAGE_SHIFT, \ |
4d2d4ce0 | 867 | offset_in_page(__gpa), v); \ |
cac0f1b7 SP |
868 | }) |
869 | ||
195aefde | 870 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len); |
6aa8b732 | 871 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn); |
33e94154 | 872 | bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn); |
995decb6 | 873 | bool kvm_vcpu_is_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn); |
f9b84e19 | 874 | unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn); |
28bd726a | 875 | void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot, gfn_t gfn); |
6aa8b732 AK |
876 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn); |
877 | ||
8e73485c PB |
878 | struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu); |
879 | struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn); | |
ba049e93 DW |
880 | kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn); |
881 | kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn); | |
e45adf66 | 882 | int kvm_vcpu_map(struct kvm_vcpu *vcpu, gpa_t gpa, struct kvm_host_map *map); |
91724814 BO |
883 | int kvm_map_gfn(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map, |
884 | struct gfn_to_pfn_cache *cache, bool atomic); | |
8e73485c | 885 | struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn); |
e45adf66 | 886 | void kvm_vcpu_unmap(struct kvm_vcpu *vcpu, struct kvm_host_map *map, bool dirty); |
91724814 BO |
887 | int kvm_unmap_gfn(struct kvm_vcpu *vcpu, struct kvm_host_map *map, |
888 | struct gfn_to_pfn_cache *cache, bool dirty, bool atomic); | |
8e73485c PB |
889 | unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn); |
890 | unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable); | |
891 | int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset, | |
892 | int len); | |
893 | int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data, | |
894 | unsigned long len); | |
895 | int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data, | |
896 | unsigned long len); | |
897 | int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data, | |
898 | int offset, int len); | |
899 | int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data, | |
900 | unsigned long len); | |
901 | void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn); | |
902 | ||
20b7035c JS |
903 | void kvm_sigset_activate(struct kvm_vcpu *vcpu); |
904 | void kvm_sigset_deactivate(struct kvm_vcpu *vcpu); | |
905 | ||
8776e519 | 906 | void kvm_vcpu_block(struct kvm_vcpu *vcpu); |
3217f7c2 CD |
907 | void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu); |
908 | void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu); | |
178f02ff | 909 | bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu); |
b6d33834 | 910 | void kvm_vcpu_kick(struct kvm_vcpu *vcpu); |
fa93384f | 911 | int kvm_vcpu_yield_to(struct kvm_vcpu *target); |
199b5763 | 912 | void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool usermode_vcpu_not_eligible); |
a4ee1ca4 | 913 | |
d9e368d6 | 914 | void kvm_flush_remote_tlbs(struct kvm *kvm); |
2e53d63a | 915 | void kvm_reload_remote_mmus(struct kvm *kvm); |
7053df4e | 916 | |
6926f95a SC |
917 | #ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE |
918 | int kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache *mc, int min); | |
919 | int kvm_mmu_memory_cache_nr_free_objects(struct kvm_mmu_memory_cache *mc); | |
920 | void kvm_mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc); | |
921 | void *kvm_mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc); | |
922 | #endif | |
923 | ||
7053df4e | 924 | bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req, |
54163a34 | 925 | struct kvm_vcpu *except, |
7053df4e | 926 | unsigned long *vcpu_bitmap, cpumask_var_t tmp); |
445b8236 | 927 | bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req); |
54163a34 SS |
928 | bool kvm_make_all_cpus_request_except(struct kvm *kvm, unsigned int req, |
929 | struct kvm_vcpu *except); | |
7ee30bc1 NNL |
930 | bool kvm_make_cpus_request_mask(struct kvm *kvm, unsigned int req, |
931 | unsigned long *vcpu_bitmap); | |
6aa8b732 | 932 | |
043405e1 CO |
933 | long kvm_arch_dev_ioctl(struct file *filp, |
934 | unsigned int ioctl, unsigned long arg); | |
313a3dc7 CO |
935 | long kvm_arch_vcpu_ioctl(struct file *filp, |
936 | unsigned int ioctl, unsigned long arg); | |
1499fa80 | 937 | vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf); |
018d00d2 | 938 | |
784aa3d7 | 939 | int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext); |
018d00d2 | 940 | |
3b0f1d01 | 941 | void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm, |
ba0513b5 MS |
942 | struct kvm_memory_slot *slot, |
943 | gfn_t gfn_offset, | |
944 | unsigned long mask); | |
0dff0846 SC |
945 | void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot); |
946 | ||
947 | #ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT | |
948 | void kvm_arch_flush_remote_tlbs_memslot(struct kvm *kvm, | |
6c9dd6d2 | 949 | const struct kvm_memory_slot *memslot); |
0dff0846 SC |
950 | #else /* !CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */ |
951 | int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log); | |
952 | int kvm_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log, | |
2a49f61d | 953 | int *is_dirty, struct kvm_memory_slot **memslot); |
0dff0846 | 954 | #endif |
5bb064dc | 955 | |
aa2fbe6d YZ |
956 | int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level, |
957 | bool line_status); | |
e5d83c74 PB |
958 | int kvm_vm_ioctl_enable_cap(struct kvm *kvm, |
959 | struct kvm_enable_cap *cap); | |
1fe779f8 CO |
960 | long kvm_arch_vm_ioctl(struct file *filp, |
961 | unsigned int ioctl, unsigned long arg); | |
313a3dc7 | 962 | |
d0752060 HB |
963 | int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu); |
964 | int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu); | |
965 | ||
8b006791 ZX |
966 | int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu, |
967 | struct kvm_translation *tr); | |
968 | ||
b6c7a5dc HB |
969 | int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs); |
970 | int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs); | |
971 | int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu, | |
972 | struct kvm_sregs *sregs); | |
973 | int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu, | |
974 | struct kvm_sregs *sregs); | |
62d9f0db MT |
975 | int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu, |
976 | struct kvm_mp_state *mp_state); | |
977 | int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu, | |
978 | struct kvm_mp_state *mp_state); | |
d0bfb940 JK |
979 | int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu, |
980 | struct kvm_guest_debug *dbg); | |
1b94f6f8 | 981 | int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu); |
b6c7a5dc | 982 | |
f8c16bba ZX |
983 | int kvm_arch_init(void *opaque); |
984 | void kvm_arch_exit(void); | |
043405e1 | 985 | |
e790d9ef RK |
986 | void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu); |
987 | ||
e9b11c17 ZX |
988 | void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu); |
989 | void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu); | |
897cc38e | 990 | int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id); |
e529ef66 | 991 | int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu); |
31928aa5 | 992 | void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu); |
d40ccc62 | 993 | void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu); |
e9b11c17 | 994 | |
741cbbae | 995 | #ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS |
d56f5136 | 996 | void kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu, struct dentry *debugfs_dentry); |
741cbbae | 997 | #endif |
235539b4 | 998 | |
13a34e06 RK |
999 | int kvm_arch_hardware_enable(void); |
1000 | void kvm_arch_hardware_disable(void); | |
b9904085 | 1001 | int kvm_arch_hardware_setup(void *opaque); |
e9b11c17 | 1002 | void kvm_arch_hardware_unsetup(void); |
b9904085 | 1003 | int kvm_arch_check_processor_compat(void *opaque); |
1d737c8a | 1004 | int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu); |
199b5763 | 1005 | bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu); |
b6d33834 | 1006 | int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu); |
17e433b5 | 1007 | bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu); |
52acd22f | 1008 | bool kvm_arch_dy_has_pending_interrupt(struct kvm_vcpu *vcpu); |
d970a325 PB |
1009 | int kvm_arch_post_init_vm(struct kvm *kvm); |
1010 | void kvm_arch_pre_destroy_vm(struct kvm *kvm); | |
e9b11c17 | 1011 | |
d89f5eff | 1012 | #ifndef __KVM_HAVE_ARCH_VM_ALLOC |
d1e5b0e9 MO |
1013 | /* |
1014 | * All architectures that want to use vzalloc currently also | |
1015 | * need their own kvm_arch_alloc_vm implementation. | |
1016 | */ | |
d89f5eff JK |
1017 | static inline struct kvm *kvm_arch_alloc_vm(void) |
1018 | { | |
1019 | return kzalloc(sizeof(struct kvm), GFP_KERNEL); | |
1020 | } | |
1021 | ||
1022 | static inline void kvm_arch_free_vm(struct kvm *kvm) | |
1023 | { | |
1024 | kfree(kvm); | |
1025 | } | |
1026 | #endif | |
1027 | ||
b08660e5 TL |
1028 | #ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB |
1029 | static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm) | |
1030 | { | |
1031 | return -ENOTSUPP; | |
1032 | } | |
1033 | #endif | |
1034 | ||
e0f0bbc5 AW |
1035 | #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA |
1036 | void kvm_arch_register_noncoherent_dma(struct kvm *kvm); | |
1037 | void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm); | |
1038 | bool kvm_arch_has_noncoherent_dma(struct kvm *kvm); | |
1039 | #else | |
1040 | static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm) | |
1041 | { | |
1042 | } | |
1043 | ||
1044 | static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm) | |
1045 | { | |
1046 | } | |
1047 | ||
1048 | static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm) | |
1049 | { | |
1050 | return false; | |
1051 | } | |
1052 | #endif | |
5544eb9b PB |
1053 | #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE |
1054 | void kvm_arch_start_assignment(struct kvm *kvm); | |
1055 | void kvm_arch_end_assignment(struct kvm *kvm); | |
1056 | bool kvm_arch_has_assigned_device(struct kvm *kvm); | |
1057 | #else | |
1058 | static inline void kvm_arch_start_assignment(struct kvm *kvm) | |
1059 | { | |
1060 | } | |
1061 | ||
1062 | static inline void kvm_arch_end_assignment(struct kvm *kvm) | |
1063 | { | |
1064 | } | |
1065 | ||
1066 | static inline bool kvm_arch_has_assigned_device(struct kvm *kvm) | |
1067 | { | |
1068 | return false; | |
1069 | } | |
1070 | #endif | |
e0f0bbc5 | 1071 | |
da4ad88c | 1072 | static inline struct rcuwait *kvm_arch_vcpu_get_wait(struct kvm_vcpu *vcpu) |
b6d33834 | 1073 | { |
2246f8b5 | 1074 | #ifdef __KVM_HAVE_ARCH_WQP |
da4ad88c | 1075 | return vcpu->arch.waitp; |
2246f8b5 | 1076 | #else |
da4ad88c | 1077 | return &vcpu->wait; |
b6d33834 | 1078 | #endif |
2246f8b5 | 1079 | } |
b6d33834 | 1080 | |
01c94e64 EA |
1081 | #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED |
1082 | /* | |
1083 | * returns true if the virtual interrupt controller is initialized and | |
1084 | * ready to accept virtual IRQ. On some architectures the virtual interrupt | |
1085 | * controller is dynamically instantiated and this is not always true. | |
1086 | */ | |
1087 | bool kvm_arch_intc_initialized(struct kvm *kvm); | |
1088 | #else | |
1089 | static inline bool kvm_arch_intc_initialized(struct kvm *kvm) | |
1090 | { | |
1091 | return true; | |
1092 | } | |
1093 | #endif | |
1094 | ||
e08b9637 | 1095 | int kvm_arch_init_vm(struct kvm *kvm, unsigned long type); |
d19a9cd2 | 1096 | void kvm_arch_destroy_vm(struct kvm *kvm); |
ad8ba2cd | 1097 | void kvm_arch_sync_events(struct kvm *kvm); |
e9b11c17 | 1098 | |
3d80840d | 1099 | int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu); |
682c59a3 | 1100 | |
ba049e93 | 1101 | bool kvm_is_reserved_pfn(kvm_pfn_t pfn); |
a78986aa | 1102 | bool kvm_is_zone_device_pfn(kvm_pfn_t pfn); |
005ba37c | 1103 | bool kvm_is_transparent_hugepage(kvm_pfn_t pfn); |
c77fb9dc | 1104 | |
62c476c7 BAY |
1105 | struct kvm_irq_ack_notifier { |
1106 | struct hlist_node link; | |
1107 | unsigned gsi; | |
1108 | void (*irq_acked)(struct kvm_irq_ack_notifier *kian); | |
1109 | }; | |
1110 | ||
9957c86d PM |
1111 | int kvm_irq_map_gsi(struct kvm *kvm, |
1112 | struct kvm_kernel_irq_routing_entry *entries, int gsi); | |
1113 | int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin); | |
8ba918d4 | 1114 | |
aa2fbe6d YZ |
1115 | int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level, |
1116 | bool line_status); | |
bd2b53b2 | 1117 | int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm, |
aa2fbe6d | 1118 | int irq_source_id, int level, bool line_status); |
b97e6de9 PB |
1119 | int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e, |
1120 | struct kvm *kvm, int irq_source_id, | |
1121 | int level, bool line_status); | |
c7c9c56c | 1122 | bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin); |
ba1aefcd | 1123 | void kvm_notify_acked_gsi(struct kvm *kvm, int gsi); |
44882eed | 1124 | void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin); |
3de42dc0 XZ |
1125 | void kvm_register_irq_ack_notifier(struct kvm *kvm, |
1126 | struct kvm_irq_ack_notifier *kian); | |
fa40a821 MT |
1127 | void kvm_unregister_irq_ack_notifier(struct kvm *kvm, |
1128 | struct kvm_irq_ack_notifier *kian); | |
5550af4d SY |
1129 | int kvm_request_irq_source_id(struct kvm *kvm); |
1130 | void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id); | |
cdc238eb | 1131 | bool kvm_arch_irqfd_allowed(struct kvm *kvm, struct kvm_irqfd *args); |
62c476c7 | 1132 | |
9d4cba7f PM |
1133 | /* |
1134 | * search_memslots() and __gfn_to_memslot() are here because they are | |
1135 | * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c. | |
1136 | * gfn_to_memslot() itself isn't here as an inline because that would | |
1137 | * bloat other code too much. | |
0577d1ab SC |
1138 | * |
1139 | * IMPORTANT: Slots are sorted from highest GFN to lowest GFN! | |
9d4cba7f PM |
1140 | */ |
1141 | static inline struct kvm_memory_slot * | |
1142 | search_memslots(struct kvm_memslots *slots, gfn_t gfn) | |
1143 | { | |
9c1a5d38 | 1144 | int start = 0, end = slots->used_slots; |
d4ae84a0 | 1145 | int slot = atomic_read(&slots->lru_slot); |
9c1a5d38 IM |
1146 | struct kvm_memory_slot *memslots = slots->memslots; |
1147 | ||
0774a964 SC |
1148 | if (unlikely(!slots->used_slots)) |
1149 | return NULL; | |
1150 | ||
9c1a5d38 IM |
1151 | if (gfn >= memslots[slot].base_gfn && |
1152 | gfn < memslots[slot].base_gfn + memslots[slot].npages) | |
1153 | return &memslots[slot]; | |
1154 | ||
1155 | while (start < end) { | |
1156 | slot = start + (end - start) / 2; | |
1157 | ||
1158 | if (gfn >= memslots[slot].base_gfn) | |
1159 | end = slot; | |
1160 | else | |
1161 | start = slot + 1; | |
1162 | } | |
1163 | ||
b6467ab1 | 1164 | if (start < slots->used_slots && gfn >= memslots[start].base_gfn && |
9c1a5d38 IM |
1165 | gfn < memslots[start].base_gfn + memslots[start].npages) { |
1166 | atomic_set(&slots->lru_slot, start); | |
1167 | return &memslots[start]; | |
1168 | } | |
9d4cba7f PM |
1169 | |
1170 | return NULL; | |
1171 | } | |
1172 | ||
1173 | static inline struct kvm_memory_slot * | |
1174 | __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn) | |
1175 | { | |
1176 | return search_memslots(slots, gfn); | |
1177 | } | |
1178 | ||
66a03505 | 1179 | static inline unsigned long |
8ca6f063 | 1180 | __gfn_to_hva_memslot(const struct kvm_memory_slot *slot, gfn_t gfn) |
66a03505 GS |
1181 | { |
1182 | return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE; | |
1183 | } | |
1184 | ||
0ee8dcb8 XG |
1185 | static inline int memslot_id(struct kvm *kvm, gfn_t gfn) |
1186 | { | |
1187 | return gfn_to_memslot(kvm, gfn)->id; | |
1188 | } | |
1189 | ||
d19a748b TY |
1190 | static inline gfn_t |
1191 | hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot) | |
887c08ac | 1192 | { |
d19a748b TY |
1193 | gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT; |
1194 | ||
1195 | return slot->base_gfn + gfn_offset; | |
887c08ac XG |
1196 | } |
1197 | ||
1755fbcc AK |
1198 | static inline gpa_t gfn_to_gpa(gfn_t gfn) |
1199 | { | |
1200 | return (gpa_t)gfn << PAGE_SHIFT; | |
1201 | } | |
6aa8b732 | 1202 | |
c30a358d JR |
1203 | static inline gfn_t gpa_to_gfn(gpa_t gpa) |
1204 | { | |
1205 | return (gfn_t)(gpa >> PAGE_SHIFT); | |
1206 | } | |
1207 | ||
ba049e93 | 1208 | static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn) |
62c476c7 BAY |
1209 | { |
1210 | return (hpa_t)pfn << PAGE_SHIFT; | |
1211 | } | |
1212 | ||
5e2f30b7 DH |
1213 | static inline struct page *kvm_vcpu_gpa_to_page(struct kvm_vcpu *vcpu, |
1214 | gpa_t gpa) | |
1215 | { | |
1216 | return kvm_vcpu_gfn_to_page(vcpu, gpa_to_gfn(gpa)); | |
1217 | } | |
1218 | ||
dfeec843 HC |
1219 | static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa) |
1220 | { | |
1221 | unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa)); | |
1222 | ||
1223 | return kvm_is_error_hva(hva); | |
1224 | } | |
1225 | ||
ba1389b7 AK |
1226 | enum kvm_stat_kind { |
1227 | KVM_STAT_VM, | |
1228 | KVM_STAT_VCPU, | |
1229 | }; | |
1230 | ||
536a6f88 | 1231 | struct kvm_stat_data { |
536a6f88 | 1232 | struct kvm *kvm; |
09cbcef6 | 1233 | struct kvm_stats_debugfs_item *dbgfs_item; |
536a6f88 JF |
1234 | }; |
1235 | ||
417bc304 HB |
1236 | struct kvm_stats_debugfs_item { |
1237 | const char *name; | |
1238 | int offset; | |
ba1389b7 | 1239 | enum kvm_stat_kind kind; |
833b45de | 1240 | int mode; |
417bc304 | 1241 | }; |
09cbcef6 MP |
1242 | |
1243 | #define KVM_DBGFS_GET_MODE(dbgfs_item) \ | |
1244 | ((dbgfs_item)->mode ? (dbgfs_item)->mode : 0644) | |
1245 | ||
812756a8 EGE |
1246 | #define VM_STAT(n, x, ...) \ |
1247 | { n, offsetof(struct kvm, stat.x), KVM_STAT_VM, ## __VA_ARGS__ } | |
1248 | #define VCPU_STAT(n, x, ...) \ | |
1249 | { n, offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU, ## __VA_ARGS__ } | |
1250 | ||
417bc304 | 1251 | extern struct kvm_stats_debugfs_item debugfs_entries[]; |
76f7c879 | 1252 | extern struct dentry *kvm_debugfs_dir; |
d4c9ff2d | 1253 | |
36c1ed82 | 1254 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
8ca40a70 | 1255 | static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq) |
e930bffe | 1256 | { |
8ca40a70 | 1257 | if (unlikely(kvm->mmu_notifier_count)) |
e930bffe AA |
1258 | return 1; |
1259 | /* | |
a355aa54 PM |
1260 | * Ensure the read of mmu_notifier_count happens before the read |
1261 | * of mmu_notifier_seq. This interacts with the smp_wmb() in | |
1262 | * mmu_notifier_invalidate_range_end to make sure that the caller | |
1263 | * either sees the old (non-zero) value of mmu_notifier_count or | |
1264 | * the new (incremented) value of mmu_notifier_seq. | |
1265 | * PowerPC Book3s HV KVM calls this under a per-page lock | |
1266 | * rather than under kvm->mmu_lock, for scalability, so | |
1267 | * can't rely on kvm->mmu_lock to keep things ordered. | |
e930bffe | 1268 | */ |
a355aa54 | 1269 | smp_rmb(); |
8ca40a70 | 1270 | if (kvm->mmu_notifier_seq != mmu_seq) |
e930bffe AA |
1271 | return 1; |
1272 | return 0; | |
1273 | } | |
4a42d848 DS |
1274 | |
1275 | static inline int mmu_notifier_retry_hva(struct kvm *kvm, | |
1276 | unsigned long mmu_seq, | |
1277 | unsigned long hva) | |
1278 | { | |
1279 | lockdep_assert_held(&kvm->mmu_lock); | |
1280 | /* | |
1281 | * If mmu_notifier_count is non-zero, then the range maintained by | |
1282 | * kvm_mmu_notifier_invalidate_range_start contains all addresses that | |
1283 | * might be being invalidated. Note that it may include some false | |
1284 | * positives, due to shortcuts when handing concurrent invalidations. | |
1285 | */ | |
1286 | if (unlikely(kvm->mmu_notifier_count) && | |
1287 | hva >= kvm->mmu_notifier_range_start && | |
1288 | hva < kvm->mmu_notifier_range_end) | |
1289 | return 1; | |
1290 | if (kvm->mmu_notifier_seq != mmu_seq) | |
1291 | return 1; | |
1292 | return 0; | |
1293 | } | |
e930bffe AA |
1294 | #endif |
1295 | ||
a725d56a | 1296 | #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING |
399ec807 | 1297 | |
ddc9cfb7 | 1298 | #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */ |
399ec807 | 1299 | |
5c0aea0e | 1300 | bool kvm_arch_can_set_irq_routing(struct kvm *kvm); |
399ec807 AK |
1301 | int kvm_set_irq_routing(struct kvm *kvm, |
1302 | const struct kvm_irq_routing_entry *entries, | |
1303 | unsigned nr, | |
1304 | unsigned flags); | |
c63cf538 RK |
1305 | int kvm_set_routing_entry(struct kvm *kvm, |
1306 | struct kvm_kernel_irq_routing_entry *e, | |
e8cde093 | 1307 | const struct kvm_irq_routing_entry *ue); |
399ec807 AK |
1308 | void kvm_free_irq_routing(struct kvm *kvm); |
1309 | ||
1310 | #else | |
1311 | ||
1312 | static inline void kvm_free_irq_routing(struct kvm *kvm) {} | |
1313 | ||
1314 | #endif | |
1315 | ||
297e2105 PM |
1316 | int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi); |
1317 | ||
721eecbf GH |
1318 | #ifdef CONFIG_HAVE_KVM_EVENTFD |
1319 | ||
d34e6b17 | 1320 | void kvm_eventfd_init(struct kvm *kvm); |
914daba8 AG |
1321 | int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args); |
1322 | ||
297e2105 | 1323 | #ifdef CONFIG_HAVE_KVM_IRQFD |
d4db2935 | 1324 | int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args); |
721eecbf | 1325 | void kvm_irqfd_release(struct kvm *kvm); |
9957c86d | 1326 | void kvm_irq_routing_update(struct kvm *); |
914daba8 AG |
1327 | #else |
1328 | static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args) | |
1329 | { | |
1330 | return -EINVAL; | |
1331 | } | |
1332 | ||
1333 | static inline void kvm_irqfd_release(struct kvm *kvm) {} | |
1334 | #endif | |
721eecbf GH |
1335 | |
1336 | #else | |
1337 | ||
d34e6b17 | 1338 | static inline void kvm_eventfd_init(struct kvm *kvm) {} |
bd2b53b2 | 1339 | |
d4db2935 | 1340 | static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args) |
721eecbf GH |
1341 | { |
1342 | return -EINVAL; | |
1343 | } | |
1344 | ||
1345 | static inline void kvm_irqfd_release(struct kvm *kvm) {} | |
bd2b53b2 | 1346 | |
27923eb1 | 1347 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
9957c86d | 1348 | static inline void kvm_irq_routing_update(struct kvm *kvm) |
bd2b53b2 | 1349 | { |
bd2b53b2 | 1350 | } |
27923eb1 | 1351 | #endif |
bd2b53b2 | 1352 | |
d34e6b17 GH |
1353 | static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args) |
1354 | { | |
1355 | return -ENOSYS; | |
1356 | } | |
721eecbf GH |
1357 | |
1358 | #endif /* CONFIG_HAVE_KVM_EVENTFD */ | |
1359 | ||
07646749 SO |
1360 | void kvm_arch_irq_routing_update(struct kvm *kvm); |
1361 | ||
a8eeb04a AK |
1362 | static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu) |
1363 | { | |
2e4682ba PB |
1364 | /* |
1365 | * Ensure the rest of the request is published to kvm_check_request's | |
1366 | * caller. Paired with the smp_mb__after_atomic in kvm_check_request. | |
1367 | */ | |
1368 | smp_wmb(); | |
86dafed5 | 1369 | set_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests); |
a8eeb04a AK |
1370 | } |
1371 | ||
2fa6e1e1 RK |
1372 | static inline bool kvm_request_pending(struct kvm_vcpu *vcpu) |
1373 | { | |
1374 | return READ_ONCE(vcpu->requests); | |
1375 | } | |
1376 | ||
72875d8a RK |
1377 | static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu) |
1378 | { | |
86dafed5 | 1379 | return test_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests); |
72875d8a RK |
1380 | } |
1381 | ||
1382 | static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu) | |
1383 | { | |
86dafed5 | 1384 | clear_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests); |
72875d8a RK |
1385 | } |
1386 | ||
a8eeb04a AK |
1387 | static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu) |
1388 | { | |
72875d8a RK |
1389 | if (kvm_test_request(req, vcpu)) { |
1390 | kvm_clear_request(req, vcpu); | |
2e4682ba PB |
1391 | |
1392 | /* | |
1393 | * Ensure the rest of the request is visible to kvm_check_request's | |
1394 | * caller. Paired with the smp_wmb in kvm_make_request. | |
1395 | */ | |
1396 | smp_mb__after_atomic(); | |
0719837c AK |
1397 | return true; |
1398 | } else { | |
1399 | return false; | |
1400 | } | |
a8eeb04a AK |
1401 | } |
1402 | ||
8b415dcd GL |
1403 | extern bool kvm_rebooting; |
1404 | ||
ec76d819 SJS |
1405 | extern unsigned int halt_poll_ns; |
1406 | extern unsigned int halt_poll_ns_grow; | |
49113d36 | 1407 | extern unsigned int halt_poll_ns_grow_start; |
ec76d819 SJS |
1408 | extern unsigned int halt_poll_ns_shrink; |
1409 | ||
852b6d57 | 1410 | struct kvm_device { |
8538cb22 | 1411 | const struct kvm_device_ops *ops; |
852b6d57 | 1412 | struct kvm *kvm; |
852b6d57 | 1413 | void *private; |
07f0a7bd | 1414 | struct list_head vm_node; |
852b6d57 SW |
1415 | }; |
1416 | ||
1417 | /* create, destroy, and name are mandatory */ | |
1418 | struct kvm_device_ops { | |
1419 | const char *name; | |
a28ebea2 CD |
1420 | |
1421 | /* | |
1422 | * create is called holding kvm->lock and any operations not suitable | |
1423 | * to do while holding the lock should be deferred to init (see | |
1424 | * below). | |
1425 | */ | |
852b6d57 SW |
1426 | int (*create)(struct kvm_device *dev, u32 type); |
1427 | ||
023e9fdd CD |
1428 | /* |
1429 | * init is called after create if create is successful and is called | |
1430 | * outside of holding kvm->lock. | |
1431 | */ | |
1432 | void (*init)(struct kvm_device *dev); | |
1433 | ||
852b6d57 SW |
1434 | /* |
1435 | * Destroy is responsible for freeing dev. | |
1436 | * | |
1437 | * Destroy may be called before or after destructors are called | |
1438 | * on emulated I/O regions, depending on whether a reference is | |
1439 | * held by a vcpu or other kvm component that gets destroyed | |
1440 | * after the emulated I/O. | |
1441 | */ | |
1442 | void (*destroy)(struct kvm_device *dev); | |
1443 | ||
2bde9b3e CLG |
1444 | /* |
1445 | * Release is an alternative method to free the device. It is | |
1446 | * called when the device file descriptor is closed. Once | |
1447 | * release is called, the destroy method will not be called | |
1448 | * anymore as the device is removed from the device list of | |
1449 | * the VM. kvm->lock is held. | |
1450 | */ | |
1451 | void (*release)(struct kvm_device *dev); | |
1452 | ||
852b6d57 SW |
1453 | int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr); |
1454 | int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr); | |
1455 | int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr); | |
1456 | long (*ioctl)(struct kvm_device *dev, unsigned int ioctl, | |
1457 | unsigned long arg); | |
a1cd3f08 | 1458 | int (*mmap)(struct kvm_device *dev, struct vm_area_struct *vma); |
852b6d57 SW |
1459 | }; |
1460 | ||
1461 | void kvm_device_get(struct kvm_device *dev); | |
1462 | void kvm_device_put(struct kvm_device *dev); | |
1463 | struct kvm_device *kvm_device_from_filp(struct file *filp); | |
8538cb22 | 1464 | int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type); |
571ee1b6 | 1465 | void kvm_unregister_device_ops(u32 type); |
852b6d57 | 1466 | |
5df554ad | 1467 | extern struct kvm_device_ops kvm_mpic_ops; |
ea2f83a7 | 1468 | extern struct kvm_device_ops kvm_arm_vgic_v2_ops; |
a0675c25 | 1469 | extern struct kvm_device_ops kvm_arm_vgic_v3_ops; |
5df554ad | 1470 | |
4c088493 R |
1471 | #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT |
1472 | ||
1473 | static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val) | |
1474 | { | |
1475 | vcpu->spin_loop.in_spin_loop = val; | |
1476 | } | |
1477 | static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val) | |
1478 | { | |
1479 | vcpu->spin_loop.dy_eligible = val; | |
1480 | } | |
1481 | ||
1482 | #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */ | |
1483 | ||
1484 | static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val) | |
1485 | { | |
1486 | } | |
1487 | ||
1488 | static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val) | |
1489 | { | |
1490 | } | |
4c088493 | 1491 | #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */ |
1a02b270 | 1492 | |
c36b7150 PB |
1493 | static inline bool kvm_is_visible_memslot(struct kvm_memory_slot *memslot) |
1494 | { | |
1495 | return (memslot && memslot->id < KVM_USER_MEM_SLOTS && | |
1496 | !(memslot->flags & KVM_MEMSLOT_INVALID)); | |
1497 | } | |
1498 | ||
7495e22b | 1499 | struct kvm_vcpu *kvm_get_running_vcpu(void); |
fcd07f9a | 1500 | struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void); |
7495e22b | 1501 | |
1a02b270 | 1502 | #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS |
14717e20 | 1503 | bool kvm_arch_has_irq_bypass(void); |
1a02b270 EA |
1504 | int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *, |
1505 | struct irq_bypass_producer *); | |
1506 | void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *, | |
1507 | struct irq_bypass_producer *); | |
1508 | void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *); | |
1509 | void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *); | |
f70c20aa FW |
1510 | int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq, |
1511 | uint32_t guest_irq, bool set); | |
1a02b270 | 1512 | #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */ |
35181e86 | 1513 | |
3491caf2 CB |
1514 | #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS |
1515 | /* If we wakeup during the poll time, was it a sucessful poll? */ | |
1516 | static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu) | |
1517 | { | |
1518 | return vcpu->valid_wakeup; | |
1519 | } | |
1520 | ||
1521 | #else | |
1522 | static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu) | |
1523 | { | |
1524 | return true; | |
1525 | } | |
1526 | #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */ | |
1527 | ||
cdd6ad3a CB |
1528 | #ifdef CONFIG_HAVE_KVM_NO_POLL |
1529 | /* Callback that tells if we must not poll */ | |
1530 | bool kvm_arch_no_poll(struct kvm_vcpu *vcpu); | |
1531 | #else | |
1532 | static inline bool kvm_arch_no_poll(struct kvm_vcpu *vcpu) | |
1533 | { | |
1534 | return false; | |
1535 | } | |
1536 | #endif /* CONFIG_HAVE_KVM_NO_POLL */ | |
1537 | ||
5cb0944c PB |
1538 | #ifdef CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL |
1539 | long kvm_arch_vcpu_async_ioctl(struct file *filp, | |
1540 | unsigned int ioctl, unsigned long arg); | |
1541 | #else | |
1542 | static inline long kvm_arch_vcpu_async_ioctl(struct file *filp, | |
1543 | unsigned int ioctl, | |
1544 | unsigned long arg) | |
1545 | { | |
1546 | return -ENOIOCTLCMD; | |
1547 | } | |
1548 | #endif /* CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL */ | |
1549 | ||
e649b3f0 ET |
1550 | void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm, |
1551 | unsigned long start, unsigned long end); | |
f75e4924 | 1552 | |
bd2a6394 CD |
1553 | #ifdef CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE |
1554 | int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu); | |
1555 | #else | |
1556 | static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu) | |
1557 | { | |
1558 | return 0; | |
1559 | } | |
1560 | #endif /* CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE */ | |
1561 | ||
c57c8046 JS |
1562 | typedef int (*kvm_vm_thread_fn_t)(struct kvm *kvm, uintptr_t data); |
1563 | ||
1564 | int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn, | |
1565 | uintptr_t data, const char *name, | |
1566 | struct task_struct **thread_ptr); | |
1567 | ||
935ace2f TG |
1568 | #ifdef CONFIG_KVM_XFER_TO_GUEST_WORK |
1569 | static inline void kvm_handle_signal_exit(struct kvm_vcpu *vcpu) | |
1570 | { | |
1571 | vcpu->run->exit_reason = KVM_EXIT_INTR; | |
1572 | vcpu->stat.signal_exits++; | |
1573 | } | |
1574 | #endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */ | |
1575 | ||
fb04a1ed PX |
1576 | /* |
1577 | * This defines how many reserved entries we want to keep before we | |
1578 | * kick the vcpu to the userspace to avoid dirty ring full. This | |
1579 | * value can be tuned to higher if e.g. PML is enabled on the host. | |
1580 | */ | |
1581 | #define KVM_DIRTY_RING_RSVD_ENTRIES 64 | |
1582 | ||
1583 | /* Max number of entries allowed for each kvm dirty ring */ | |
1584 | #define KVM_DIRTY_RING_MAX_ENTRIES 65536 | |
1585 | ||
bfd99ff5 | 1586 | #endif |