1 /* SPDX-License-Identifier: GPL-2.0-only */
6 #include <linux/types.h>
7 #include <linux/hardirq.h>
8 #include <linux/list.h>
9 #include <linux/mutex.h>
10 #include <linux/spinlock.h>
11 #include <linux/signal.h>
12 #include <linux/sched.h>
13 #include <linux/sched/stat.h>
14 #include <linux/bug.h>
15 #include <linux/minmax.h>
17 #include <linux/mmu_notifier.h>
18 #include <linux/preempt.h>
19 #include <linux/msi.h>
20 #include <linux/slab.h>
21 #include <linux/vmalloc.h>
22 #include <linux/rcupdate.h>
23 #include <linux/ratelimit.h>
24 #include <linux/err.h>
25 #include <linux/irqflags.h>
26 #include <linux/context_tracking.h>
27 #include <linux/irqbypass.h>
28 #include <linux/rcuwait.h>
29 #include <linux/refcount.h>
30 #include <linux/nospec.h>
31 #include <linux/notifier.h>
32 #include <asm/signal.h>
34 #include <linux/kvm.h>
35 #include <linux/kvm_para.h>
37 #include <linux/kvm_types.h>
39 #include <asm/kvm_host.h>
40 #include <linux/kvm_dirty_ring.h>
42 #ifndef KVM_MAX_VCPU_ID
43 #define KVM_MAX_VCPU_ID KVM_MAX_VCPUS
47 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
48 * in kvm, other bits are visible for userspace which are defined in
49 * include/linux/kvm_h.
51 #define KVM_MEMSLOT_INVALID (1UL << 16)
54 * Bit 63 of the memslot generation number is an "update in-progress flag",
55 * e.g. is temporarily set for the duration of install_new_memslots().
56 * This flag effectively creates a unique generation number that is used to
57 * mark cached memslot data, e.g. MMIO accesses, as potentially being stale,
58 * i.e. may (or may not) have come from the previous memslots generation.
60 * This is necessary because the actual memslots update is not atomic with
61 * respect to the generation number update. Updating the generation number
62 * first would allow a vCPU to cache a spte from the old memslots using the
63 * new generation number, and updating the generation number after switching
64 * to the new memslots would allow cache hits using the old generation number
65 * to reference the defunct memslots.
67 * This mechanism is used to prevent getting hits in KVM's caches while a
68 * memslot update is in-progress, and to prevent cache hits *after* updating
69 * the actual generation number against accesses that were inserted into the
70 * cache *before* the memslots were updated.
72 #define KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS BIT_ULL(63)
74 /* Two fragments for cross MMIO pages. */
75 #define KVM_MAX_MMIO_FRAGMENTS 2
77 #ifndef KVM_ADDRESS_SPACE_NUM
78 #define KVM_ADDRESS_SPACE_NUM 1
82 * For the normal pfn, the highest 12 bits should be zero,
83 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
84 * mask bit 63 to indicate the noslot pfn.
86 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
87 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
88 #define KVM_PFN_NOSLOT (0x1ULL << 63)
90 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
91 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
92 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
95 * error pfns indicate that the gfn is in slot but faild to
96 * translate it to pfn on host.
98 static inline bool is_error_pfn(kvm_pfn_t pfn
)
100 return !!(pfn
& KVM_PFN_ERR_MASK
);
104 * error_noslot pfns indicate that the gfn can not be
105 * translated to pfn - it is not in slot or failed to
106 * translate it to pfn.
108 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn
)
110 return !!(pfn
& KVM_PFN_ERR_NOSLOT_MASK
);
113 /* noslot pfn indicates that the gfn is not in slot. */
114 static inline bool is_noslot_pfn(kvm_pfn_t pfn
)
116 return pfn
== KVM_PFN_NOSLOT
;
120 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
121 * provide own defines and kvm_is_error_hva
123 #ifndef KVM_HVA_ERR_BAD
125 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
126 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
128 static inline bool kvm_is_error_hva(unsigned long addr
)
130 return addr
>= PAGE_OFFSET
;
135 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
137 static inline bool is_error_page(struct page
*page
)
142 #define KVM_REQUEST_MASK GENMASK(7,0)
143 #define KVM_REQUEST_NO_WAKEUP BIT(8)
144 #define KVM_REQUEST_WAIT BIT(9)
146 * Architecture-independent vcpu->requests bit members
147 * Bits 4-7 are reserved for more arch-independent bits.
149 #define KVM_REQ_TLB_FLUSH (0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
150 #define KVM_REQ_MMU_RELOAD (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
151 #define KVM_REQ_UNBLOCK 2
152 #define KVM_REQ_UNHALT 3
153 #define KVM_REQUEST_ARCH_BASE 8
155 #define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
156 BUILD_BUG_ON((unsigned)(nr) >= (sizeof_field(struct kvm_vcpu, requests) * 8) - KVM_REQUEST_ARCH_BASE); \
157 (unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) | (flags)); \
159 #define KVM_ARCH_REQ(nr) KVM_ARCH_REQ_FLAGS(nr, 0)
161 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
162 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
164 extern struct mutex kvm_lock
;
165 extern struct list_head vm_list
;
167 struct kvm_io_range
{
170 struct kvm_io_device
*dev
;
173 #define NR_IOBUS_DEVS 1000
178 struct kvm_io_range range
[];
184 KVM_VIRTIO_CCW_NOTIFY_BUS
,
189 int kvm_io_bus_write(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
190 int len
, const void *val
);
191 int kvm_io_bus_write_cookie(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
,
192 gpa_t addr
, int len
, const void *val
, long cookie
);
193 int kvm_io_bus_read(struct kvm_vcpu
*vcpu
, enum kvm_bus bus_idx
, gpa_t addr
,
195 int kvm_io_bus_register_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
, gpa_t addr
,
196 int len
, struct kvm_io_device
*dev
);
197 int kvm_io_bus_unregister_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
198 struct kvm_io_device
*dev
);
199 struct kvm_io_device
*kvm_io_bus_get_dev(struct kvm
*kvm
, enum kvm_bus bus_idx
,
202 #ifdef CONFIG_KVM_ASYNC_PF
203 struct kvm_async_pf
{
204 struct work_struct work
;
205 struct list_head link
;
206 struct list_head queue
;
207 struct kvm_vcpu
*vcpu
;
208 struct mm_struct
*mm
;
211 struct kvm_arch_async_pf arch
;
213 bool notpresent_injected
;
216 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu
*vcpu
);
217 void kvm_check_async_pf_completion(struct kvm_vcpu
*vcpu
);
218 bool kvm_setup_async_pf(struct kvm_vcpu
*vcpu
, gpa_t cr2_or_gpa
,
219 unsigned long hva
, struct kvm_arch_async_pf
*arch
);
220 int kvm_async_pf_wakeup_all(struct kvm_vcpu
*vcpu
);
223 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
224 struct kvm_gfn_range
{
225 struct kvm_memory_slot
*slot
;
231 bool kvm_unmap_gfn_range(struct kvm
*kvm
, struct kvm_gfn_range
*range
);
232 bool kvm_age_gfn(struct kvm
*kvm
, struct kvm_gfn_range
*range
);
233 bool kvm_test_age_gfn(struct kvm
*kvm
, struct kvm_gfn_range
*range
);
234 bool kvm_set_spte_gfn(struct kvm
*kvm
, struct kvm_gfn_range
*range
);
241 READING_SHADOW_PAGE_TABLES
,
244 #define KVM_UNMAPPED_PAGE ((void *) 0x500 + POISON_POINTER_DELTA)
246 struct kvm_host_map
{
248 * Only valid if the 'pfn' is managed by the host kernel (i.e. There is
249 * a 'struct page' for it. When using mem= kernel parameter some memory
250 * can be used as guest memory but they are not managed by host
252 * If 'pfn' is not managed by the host kernel, this field is
253 * initialized to KVM_UNMAPPED_PAGE.
262 * Used to check if the mapping is valid or not. Never use 'kvm_host_map'
263 * directly to check for that.
265 static inline bool kvm_vcpu_mapped(struct kvm_host_map
*map
)
270 static inline bool kvm_vcpu_can_poll(ktime_t cur
, ktime_t stop
)
272 return single_task_running() && !need_resched() && ktime_before(cur
, stop
);
276 * Sometimes a large or cross-page mmio needs to be broken up into separate
277 * exits for userspace servicing.
279 struct kvm_mmio_fragment
{
287 #ifdef CONFIG_PREEMPT_NOTIFIERS
288 struct preempt_notifier preempt_notifier
;
291 int vcpu_id
; /* id given by userspace at creation */
292 int vcpu_idx
; /* index in kvm->vcpus array */
296 unsigned long guest_debug
;
299 struct list_head blocked_vcpu_list
;
305 struct pid __rcu
*pid
;
308 struct kvm_vcpu_stat stat
;
309 unsigned int halt_poll_ns
;
312 #ifdef CONFIG_HAS_IOMEM
314 int mmio_read_completed
;
316 int mmio_cur_fragment
;
317 int mmio_nr_fragments
;
318 struct kvm_mmio_fragment mmio_fragments
[KVM_MAX_MMIO_FRAGMENTS
];
321 #ifdef CONFIG_KVM_ASYNC_PF
324 struct list_head queue
;
325 struct list_head done
;
330 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
332 * Cpu relax intercept or pause loop exit optimization
333 * in_spin_loop: set when a vcpu does a pause loop exit
334 * or cpu relax intercepted.
335 * dy_eligible: indicates whether vcpu is eligible for directed yield.
344 struct kvm_vcpu_arch arch
;
345 struct kvm_dirty_ring dirty_ring
;
348 /* must be called with irqs disabled */
349 static __always_inline
void guest_enter_irqoff(void)
352 * This is running in ioctl context so its safe to assume that it's the
353 * stime pending cputime to flush.
355 instrumentation_begin();
356 vtime_account_guest_enter();
357 instrumentation_end();
360 * KVM does not hold any references to rcu protected data when it
361 * switches CPU into a guest mode. In fact switching to a guest mode
362 * is very similar to exiting to userspace from rcu point of view. In
363 * addition CPU may stay in a guest mode for quite a long time (up to
364 * one time slice). Lets treat guest mode as quiescent state, just like
365 * we do with user-mode execution.
367 if (!context_tracking_guest_enter()) {
368 instrumentation_begin();
369 rcu_virt_note_context_switch(smp_processor_id());
370 instrumentation_end();
374 static __always_inline
void guest_exit_irqoff(void)
376 context_tracking_guest_exit();
378 instrumentation_begin();
379 /* Flush the guest cputime we spent on the guest */
380 vtime_account_guest_exit();
381 instrumentation_end();
384 static inline void guest_exit(void)
388 local_irq_save(flags
);
390 local_irq_restore(flags
);
393 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu
*vcpu
)
396 * The memory barrier ensures a previous write to vcpu->requests cannot
397 * be reordered with the read of vcpu->mode. It pairs with the general
398 * memory barrier following the write of vcpu->mode in VCPU RUN.
400 smp_mb__before_atomic();
401 return cmpxchg(&vcpu
->mode
, IN_GUEST_MODE
, EXITING_GUEST_MODE
);
405 * Some of the bitops functions do not support too long bitmaps.
406 * This number must be determined not to exceed such limits.
408 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
410 struct kvm_memory_slot
{
412 unsigned long npages
;
413 unsigned long *dirty_bitmap
;
414 struct kvm_arch_memory_slot arch
;
415 unsigned long userspace_addr
;
421 static inline bool kvm_slot_dirty_track_enabled(struct kvm_memory_slot
*slot
)
423 return slot
->flags
& KVM_MEM_LOG_DIRTY_PAGES
;
426 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot
*memslot
)
428 return ALIGN(memslot
->npages
, BITS_PER_LONG
) / 8;
431 static inline unsigned long *kvm_second_dirty_bitmap(struct kvm_memory_slot
*memslot
)
433 unsigned long len
= kvm_dirty_bitmap_bytes(memslot
);
435 return memslot
->dirty_bitmap
+ len
/ sizeof(*memslot
->dirty_bitmap
);
438 #ifndef KVM_DIRTY_LOG_MANUAL_CAPS
439 #define KVM_DIRTY_LOG_MANUAL_CAPS KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE
442 struct kvm_s390_adapter_int
{
455 struct kvm_kernel_irq_routing_entry
{
458 int (*set
)(struct kvm_kernel_irq_routing_entry
*e
,
459 struct kvm
*kvm
, int irq_source_id
, int level
,
473 struct kvm_s390_adapter_int adapter
;
474 struct kvm_hv_sint hv_sint
;
476 struct hlist_node link
;
479 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
480 struct kvm_irq_routing_table
{
481 int chip
[KVM_NR_IRQCHIPS
][KVM_IRQCHIP_NUM_PINS
];
484 * Array indexed by gsi. Each entry contains list of irq chips
485 * the gsi is connected to.
487 struct hlist_head map
[];
491 #ifndef KVM_PRIVATE_MEM_SLOTS
492 #define KVM_PRIVATE_MEM_SLOTS 0
495 #define KVM_MEM_SLOTS_NUM SHRT_MAX
496 #define KVM_USER_MEM_SLOTS (KVM_MEM_SLOTS_NUM - KVM_PRIVATE_MEM_SLOTS)
498 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
499 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu
*vcpu
)
507 * memslots are not sorted by id anymore, please use id_to_memslot()
508 * to get the memslot by its id.
510 struct kvm_memslots
{
512 /* The mapping table from slot id to the index in memslots[]. */
513 short id_to_index
[KVM_MEM_SLOTS_NUM
];
516 struct kvm_memory_slot memslots
[];
520 #ifdef KVM_HAVE_MMU_RWLOCK
524 #endif /* KVM_HAVE_MMU_RWLOCK */
526 struct mutex slots_lock
;
529 * Protects the arch-specific fields of struct kvm_memory_slots in
530 * use by the VM. To be used under the slots_lock (above) or in a
531 * kvm->srcu critical section where acquiring the slots_lock would
532 * lead to deadlock with the synchronize_srcu in
533 * install_new_memslots.
535 struct mutex slots_arch_lock
;
536 struct mm_struct
*mm
; /* userspace tied to this vm */
537 struct kvm_memslots __rcu
*memslots
[KVM_ADDRESS_SPACE_NUM
];
538 struct kvm_vcpu
*vcpus
[KVM_MAX_VCPUS
];
541 * created_vcpus is protected by kvm->lock, and is incremented
542 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
543 * incremented after storing the kvm_vcpu pointer in vcpus,
544 * and is accessed atomically.
546 atomic_t online_vcpus
;
548 int last_boosted_vcpu
;
549 struct list_head vm_list
;
551 struct kvm_io_bus __rcu
*buses
[KVM_NR_BUSES
];
552 #ifdef CONFIG_HAVE_KVM_EVENTFD
555 struct list_head items
;
556 struct list_head resampler_list
;
557 struct mutex resampler_lock
;
559 struct list_head ioeventfds
;
561 struct kvm_vm_stat stat
;
562 struct kvm_arch arch
;
563 refcount_t users_count
;
564 #ifdef CONFIG_KVM_MMIO
565 struct kvm_coalesced_mmio_ring
*coalesced_mmio_ring
;
566 spinlock_t ring_lock
;
567 struct list_head coalesced_zones
;
570 struct mutex irq_lock
;
571 #ifdef CONFIG_HAVE_KVM_IRQCHIP
573 * Update side is protected by irq_lock.
575 struct kvm_irq_routing_table __rcu
*irq_routing
;
577 #ifdef CONFIG_HAVE_KVM_IRQFD
578 struct hlist_head irq_ack_notifier_list
;
581 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
582 struct mmu_notifier mmu_notifier
;
583 unsigned long mmu_notifier_seq
;
584 long mmu_notifier_count
;
585 unsigned long mmu_notifier_range_start
;
586 unsigned long mmu_notifier_range_end
;
589 struct list_head devices
;
590 u64 manual_dirty_log_protect
;
591 struct dentry
*debugfs_dentry
;
592 struct kvm_stat_data
**debugfs_stat_data
;
593 struct srcu_struct srcu
;
594 struct srcu_struct irq_srcu
;
596 unsigned int max_halt_poll_ns
;
599 #ifdef CONFIG_HAVE_KVM_PM_NOTIFIER
600 struct notifier_block pm_notifier
;
604 #define kvm_err(fmt, ...) \
605 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
606 #define kvm_info(fmt, ...) \
607 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
608 #define kvm_debug(fmt, ...) \
609 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
610 #define kvm_debug_ratelimited(fmt, ...) \
611 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
613 #define kvm_pr_unimpl(fmt, ...) \
614 pr_err_ratelimited("kvm [%i]: " fmt, \
615 task_tgid_nr(current), ## __VA_ARGS__)
617 /* The guest did something we don't support. */
618 #define vcpu_unimpl(vcpu, fmt, ...) \
619 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
620 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
622 #define vcpu_debug(vcpu, fmt, ...) \
623 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
624 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
625 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
627 #define vcpu_err(vcpu, fmt, ...) \
628 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
630 static inline bool kvm_dirty_log_manual_protect_and_init_set(struct kvm
*kvm
)
632 return !!(kvm
->manual_dirty_log_protect
& KVM_DIRTY_LOG_INITIALLY_SET
);
635 static inline struct kvm_io_bus
*kvm_get_bus(struct kvm
*kvm
, enum kvm_bus idx
)
637 return srcu_dereference_check(kvm
->buses
[idx
], &kvm
->srcu
,
638 lockdep_is_held(&kvm
->slots_lock
) ||
639 !refcount_read(&kvm
->users_count
));
642 static inline struct kvm_vcpu
*kvm_get_vcpu(struct kvm
*kvm
, int i
)
644 int num_vcpus
= atomic_read(&kvm
->online_vcpus
);
645 i
= array_index_nospec(i
, num_vcpus
);
647 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. */
649 return kvm
->vcpus
[i
];
652 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
654 idx < atomic_read(&kvm->online_vcpus) && \
655 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
658 static inline struct kvm_vcpu
*kvm_get_vcpu_by_id(struct kvm
*kvm
, int id
)
660 struct kvm_vcpu
*vcpu
= NULL
;
665 if (id
< KVM_MAX_VCPUS
)
666 vcpu
= kvm_get_vcpu(kvm
, id
);
667 if (vcpu
&& vcpu
->vcpu_id
== id
)
669 kvm_for_each_vcpu(i
, vcpu
, kvm
)
670 if (vcpu
->vcpu_id
== id
)
675 static inline int kvm_vcpu_get_idx(struct kvm_vcpu
*vcpu
)
677 return vcpu
->vcpu_idx
;
680 #define kvm_for_each_memslot(memslot, slots) \
681 for (memslot = &slots->memslots[0]; \
682 memslot < slots->memslots + slots->used_slots; memslot++) \
683 if (WARN_ON_ONCE(!memslot->npages)) { \
686 void kvm_vcpu_destroy(struct kvm_vcpu
*vcpu
);
688 void vcpu_load(struct kvm_vcpu
*vcpu
);
689 void vcpu_put(struct kvm_vcpu
*vcpu
);
691 #ifdef __KVM_HAVE_IOAPIC
692 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm
*kvm
);
693 void kvm_arch_post_irq_routing_update(struct kvm
*kvm
);
695 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm
*kvm
)
698 static inline void kvm_arch_post_irq_routing_update(struct kvm
*kvm
)
703 #ifdef CONFIG_HAVE_KVM_IRQFD
704 int kvm_irqfd_init(void);
705 void kvm_irqfd_exit(void);
707 static inline int kvm_irqfd_init(void)
712 static inline void kvm_irqfd_exit(void)
716 int kvm_init(void *opaque
, unsigned vcpu_size
, unsigned vcpu_align
,
717 struct module
*module
);
720 void kvm_get_kvm(struct kvm
*kvm
);
721 void kvm_put_kvm(struct kvm
*kvm
);
722 bool file_is_kvm(struct file
*file
);
723 void kvm_put_kvm_no_destroy(struct kvm
*kvm
);
725 static inline struct kvm_memslots
*__kvm_memslots(struct kvm
*kvm
, int as_id
)
727 as_id
= array_index_nospec(as_id
, KVM_ADDRESS_SPACE_NUM
);
728 return srcu_dereference_check(kvm
->memslots
[as_id
], &kvm
->srcu
,
729 lockdep_is_held(&kvm
->slots_lock
) ||
730 !refcount_read(&kvm
->users_count
));
733 static inline struct kvm_memslots
*kvm_memslots(struct kvm
*kvm
)
735 return __kvm_memslots(kvm
, 0);
738 static inline struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
)
740 int as_id
= kvm_arch_vcpu_memslots_id(vcpu
);
742 return __kvm_memslots(vcpu
->kvm
, as_id
);
746 struct kvm_memory_slot
*id_to_memslot(struct kvm_memslots
*slots
, int id
)
748 int index
= slots
->id_to_index
[id
];
749 struct kvm_memory_slot
*slot
;
754 slot
= &slots
->memslots
[index
];
756 WARN_ON(slot
->id
!= id
);
761 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
762 * - create a new memory slot
763 * - delete an existing memory slot
764 * - modify an existing memory slot
765 * -- move it in the guest physical memory space
766 * -- just change its flags
768 * Since flags can be changed by some of these operations, the following
769 * differentiation is the best we can do for __kvm_set_memory_region():
778 int kvm_set_memory_region(struct kvm
*kvm
,
779 const struct kvm_userspace_memory_region
*mem
);
780 int __kvm_set_memory_region(struct kvm
*kvm
,
781 const struct kvm_userspace_memory_region
*mem
);
782 void kvm_arch_free_memslot(struct kvm
*kvm
, struct kvm_memory_slot
*slot
);
783 void kvm_arch_memslots_updated(struct kvm
*kvm
, u64 gen
);
784 int kvm_arch_prepare_memory_region(struct kvm
*kvm
,
785 struct kvm_memory_slot
*memslot
,
786 const struct kvm_userspace_memory_region
*mem
,
787 enum kvm_mr_change change
);
788 void kvm_arch_commit_memory_region(struct kvm
*kvm
,
789 const struct kvm_userspace_memory_region
*mem
,
790 struct kvm_memory_slot
*old
,
791 const struct kvm_memory_slot
*new,
792 enum kvm_mr_change change
);
793 /* flush all memory translations */
794 void kvm_arch_flush_shadow_all(struct kvm
*kvm
);
795 /* flush memory translations pointing to 'slot' */
796 void kvm_arch_flush_shadow_memslot(struct kvm
*kvm
,
797 struct kvm_memory_slot
*slot
);
799 int gfn_to_page_many_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
,
800 struct page
**pages
, int nr_pages
);
802 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
803 unsigned long gfn_to_hva(struct kvm
*kvm
, gfn_t gfn
);
804 unsigned long gfn_to_hva_prot(struct kvm
*kvm
, gfn_t gfn
, bool *writable
);
805 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
806 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
808 void kvm_release_page_clean(struct page
*page
);
809 void kvm_release_page_dirty(struct page
*page
);
810 void kvm_set_page_accessed(struct page
*page
);
812 kvm_pfn_t
gfn_to_pfn(struct kvm
*kvm
, gfn_t gfn
);
813 kvm_pfn_t
gfn_to_pfn_prot(struct kvm
*kvm
, gfn_t gfn
, bool write_fault
,
815 kvm_pfn_t
gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
);
816 kvm_pfn_t
gfn_to_pfn_memslot_atomic(struct kvm_memory_slot
*slot
, gfn_t gfn
);
817 kvm_pfn_t
__gfn_to_pfn_memslot(struct kvm_memory_slot
*slot
, gfn_t gfn
,
818 bool atomic
, bool *async
, bool write_fault
,
819 bool *writable
, hva_t
*hva
);
821 void kvm_release_pfn_clean(kvm_pfn_t pfn
);
822 void kvm_release_pfn_dirty(kvm_pfn_t pfn
);
823 void kvm_set_pfn_dirty(kvm_pfn_t pfn
);
824 void kvm_set_pfn_accessed(kvm_pfn_t pfn
);
825 void kvm_get_pfn(kvm_pfn_t pfn
);
827 void kvm_release_pfn(kvm_pfn_t pfn
, bool dirty
, struct gfn_to_pfn_cache
*cache
);
828 int kvm_read_guest_page(struct kvm
*kvm
, gfn_t gfn
, void *data
, int offset
,
830 int kvm_read_guest(struct kvm
*kvm
, gpa_t gpa
, void *data
, unsigned long len
);
831 int kvm_read_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
832 void *data
, unsigned long len
);
833 int kvm_read_guest_offset_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
834 void *data
, unsigned int offset
,
836 int kvm_write_guest_page(struct kvm
*kvm
, gfn_t gfn
, const void *data
,
837 int offset
, int len
);
838 int kvm_write_guest(struct kvm
*kvm
, gpa_t gpa
, const void *data
,
840 int kvm_write_guest_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
841 void *data
, unsigned long len
);
842 int kvm_write_guest_offset_cached(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
843 void *data
, unsigned int offset
,
845 int kvm_gfn_to_hva_cache_init(struct kvm
*kvm
, struct gfn_to_hva_cache
*ghc
,
846 gpa_t gpa
, unsigned long len
);
848 #define __kvm_get_guest(kvm, gfn, offset, v) \
850 unsigned long __addr = gfn_to_hva(kvm, gfn); \
851 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
852 int __ret = -EFAULT; \
854 if (!kvm_is_error_hva(__addr)) \
855 __ret = get_user(v, __uaddr); \
859 #define kvm_get_guest(kvm, gpa, v) \
862 struct kvm *__kvm = kvm; \
864 __kvm_get_guest(__kvm, __gpa >> PAGE_SHIFT, \
865 offset_in_page(__gpa), v); \
868 #define __kvm_put_guest(kvm, gfn, offset, v) \
870 unsigned long __addr = gfn_to_hva(kvm, gfn); \
871 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
872 int __ret = -EFAULT; \
874 if (!kvm_is_error_hva(__addr)) \
875 __ret = put_user(v, __uaddr); \
877 mark_page_dirty(kvm, gfn); \
881 #define kvm_put_guest(kvm, gpa, v) \
884 struct kvm *__kvm = kvm; \
886 __kvm_put_guest(__kvm, __gpa >> PAGE_SHIFT, \
887 offset_in_page(__gpa), v); \
890 int kvm_clear_guest(struct kvm
*kvm
, gpa_t gpa
, unsigned long len
);
891 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
892 bool kvm_is_visible_gfn(struct kvm
*kvm
, gfn_t gfn
);
893 bool kvm_vcpu_is_visible_gfn(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
894 unsigned long kvm_host_page_size(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
895 void mark_page_dirty_in_slot(struct kvm
*kvm
, struct kvm_memory_slot
*memslot
, gfn_t gfn
);
896 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
898 struct kvm_memslots
*kvm_vcpu_memslots(struct kvm_vcpu
*vcpu
);
899 struct kvm_memory_slot
*kvm_vcpu_gfn_to_memslot(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
900 kvm_pfn_t
kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
901 kvm_pfn_t
kvm_vcpu_gfn_to_pfn(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
902 int kvm_vcpu_map(struct kvm_vcpu
*vcpu
, gpa_t gpa
, struct kvm_host_map
*map
);
903 int kvm_map_gfn(struct kvm_vcpu
*vcpu
, gfn_t gfn
, struct kvm_host_map
*map
,
904 struct gfn_to_pfn_cache
*cache
, bool atomic
);
905 struct page
*kvm_vcpu_gfn_to_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
906 void kvm_vcpu_unmap(struct kvm_vcpu
*vcpu
, struct kvm_host_map
*map
, bool dirty
);
907 int kvm_unmap_gfn(struct kvm_vcpu
*vcpu
, struct kvm_host_map
*map
,
908 struct gfn_to_pfn_cache
*cache
, bool dirty
, bool atomic
);
909 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
910 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu
*vcpu
, gfn_t gfn
, bool *writable
);
911 int kvm_vcpu_read_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, void *data
, int offset
,
913 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
915 int kvm_vcpu_read_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, void *data
,
917 int kvm_vcpu_write_guest_page(struct kvm_vcpu
*vcpu
, gfn_t gfn
, const void *data
,
918 int offset
, int len
);
919 int kvm_vcpu_write_guest(struct kvm_vcpu
*vcpu
, gpa_t gpa
, const void *data
,
921 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu
*vcpu
, gfn_t gfn
);
923 void kvm_sigset_activate(struct kvm_vcpu
*vcpu
);
924 void kvm_sigset_deactivate(struct kvm_vcpu
*vcpu
);
926 void kvm_vcpu_block(struct kvm_vcpu
*vcpu
);
927 void kvm_arch_vcpu_blocking(struct kvm_vcpu
*vcpu
);
928 void kvm_arch_vcpu_unblocking(struct kvm_vcpu
*vcpu
);
929 bool kvm_vcpu_wake_up(struct kvm_vcpu
*vcpu
);
930 void kvm_vcpu_kick(struct kvm_vcpu
*vcpu
);
931 int kvm_vcpu_yield_to(struct kvm_vcpu
*target
);
932 void kvm_vcpu_on_spin(struct kvm_vcpu
*vcpu
, bool usermode_vcpu_not_eligible
);
934 void kvm_flush_remote_tlbs(struct kvm
*kvm
);
935 void kvm_reload_remote_mmus(struct kvm
*kvm
);
937 #ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE
938 int kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache
*mc
, int min
);
939 int kvm_mmu_memory_cache_nr_free_objects(struct kvm_mmu_memory_cache
*mc
);
940 void kvm_mmu_free_memory_cache(struct kvm_mmu_memory_cache
*mc
);
941 void *kvm_mmu_memory_cache_alloc(struct kvm_mmu_memory_cache
*mc
);
944 bool kvm_make_vcpus_request_mask(struct kvm
*kvm
, unsigned int req
,
945 struct kvm_vcpu
*except
,
946 unsigned long *vcpu_bitmap
, cpumask_var_t tmp
);
947 bool kvm_make_all_cpus_request(struct kvm
*kvm
, unsigned int req
);
948 bool kvm_make_all_cpus_request_except(struct kvm
*kvm
, unsigned int req
,
949 struct kvm_vcpu
*except
);
950 bool kvm_make_cpus_request_mask(struct kvm
*kvm
, unsigned int req
,
951 unsigned long *vcpu_bitmap
);
953 long kvm_arch_dev_ioctl(struct file
*filp
,
954 unsigned int ioctl
, unsigned long arg
);
955 long kvm_arch_vcpu_ioctl(struct file
*filp
,
956 unsigned int ioctl
, unsigned long arg
);
957 vm_fault_t
kvm_arch_vcpu_fault(struct kvm_vcpu
*vcpu
, struct vm_fault
*vmf
);
959 int kvm_vm_ioctl_check_extension(struct kvm
*kvm
, long ext
);
961 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm
*kvm
,
962 struct kvm_memory_slot
*slot
,
965 void kvm_arch_sync_dirty_log(struct kvm
*kvm
, struct kvm_memory_slot
*memslot
);
967 #ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
968 void kvm_arch_flush_remote_tlbs_memslot(struct kvm
*kvm
,
969 const struct kvm_memory_slot
*memslot
);
970 #else /* !CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
971 int kvm_vm_ioctl_get_dirty_log(struct kvm
*kvm
, struct kvm_dirty_log
*log
);
972 int kvm_get_dirty_log(struct kvm
*kvm
, struct kvm_dirty_log
*log
,
973 int *is_dirty
, struct kvm_memory_slot
**memslot
);
976 int kvm_vm_ioctl_irq_line(struct kvm
*kvm
, struct kvm_irq_level
*irq_level
,
978 int kvm_vm_ioctl_enable_cap(struct kvm
*kvm
,
979 struct kvm_enable_cap
*cap
);
980 long kvm_arch_vm_ioctl(struct file
*filp
,
981 unsigned int ioctl
, unsigned long arg
);
983 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
984 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu
*vcpu
, struct kvm_fpu
*fpu
);
986 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu
*vcpu
,
987 struct kvm_translation
*tr
);
989 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
990 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu
*vcpu
, struct kvm_regs
*regs
);
991 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu
*vcpu
,
992 struct kvm_sregs
*sregs
);
993 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu
*vcpu
,
994 struct kvm_sregs
*sregs
);
995 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu
*vcpu
,
996 struct kvm_mp_state
*mp_state
);
997 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu
*vcpu
,
998 struct kvm_mp_state
*mp_state
);
999 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu
*vcpu
,
1000 struct kvm_guest_debug
*dbg
);
1001 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu
*vcpu
);
1003 int kvm_arch_init(void *opaque
);
1004 void kvm_arch_exit(void);
1006 void kvm_arch_sched_in(struct kvm_vcpu
*vcpu
, int cpu
);
1008 void kvm_arch_vcpu_load(struct kvm_vcpu
*vcpu
, int cpu
);
1009 void kvm_arch_vcpu_put(struct kvm_vcpu
*vcpu
);
1010 int kvm_arch_vcpu_precreate(struct kvm
*kvm
, unsigned int id
);
1011 int kvm_arch_vcpu_create(struct kvm_vcpu
*vcpu
);
1012 void kvm_arch_vcpu_postcreate(struct kvm_vcpu
*vcpu
);
1013 void kvm_arch_vcpu_destroy(struct kvm_vcpu
*vcpu
);
1015 #ifdef CONFIG_HAVE_KVM_PM_NOTIFIER
1016 int kvm_arch_pm_notifier(struct kvm
*kvm
, unsigned long state
);
1019 #ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
1020 void kvm_arch_create_vcpu_debugfs(struct kvm_vcpu
*vcpu
, struct dentry
*debugfs_dentry
);
1023 int kvm_arch_hardware_enable(void);
1024 void kvm_arch_hardware_disable(void);
1025 int kvm_arch_hardware_setup(void *opaque
);
1026 void kvm_arch_hardware_unsetup(void);
1027 int kvm_arch_check_processor_compat(void *opaque
);
1028 int kvm_arch_vcpu_runnable(struct kvm_vcpu
*vcpu
);
1029 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu
*vcpu
);
1030 int kvm_arch_vcpu_should_kick(struct kvm_vcpu
*vcpu
);
1031 bool kvm_arch_dy_runnable(struct kvm_vcpu
*vcpu
);
1032 bool kvm_arch_dy_has_pending_interrupt(struct kvm_vcpu
*vcpu
);
1033 int kvm_arch_post_init_vm(struct kvm
*kvm
);
1034 void kvm_arch_pre_destroy_vm(struct kvm
*kvm
);
1036 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
1038 * All architectures that want to use vzalloc currently also
1039 * need their own kvm_arch_alloc_vm implementation.
1041 static inline struct kvm
*kvm_arch_alloc_vm(void)
1043 return kzalloc(sizeof(struct kvm
), GFP_KERNEL
);
1046 static inline void kvm_arch_free_vm(struct kvm
*kvm
)
1052 #ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1053 static inline int kvm_arch_flush_remote_tlb(struct kvm
*kvm
)
1059 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
1060 void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
);
1061 void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
);
1062 bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
);
1064 static inline void kvm_arch_register_noncoherent_dma(struct kvm
*kvm
)
1068 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm
*kvm
)
1072 static inline bool kvm_arch_has_noncoherent_dma(struct kvm
*kvm
)
1077 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
1078 void kvm_arch_start_assignment(struct kvm
*kvm
);
1079 void kvm_arch_end_assignment(struct kvm
*kvm
);
1080 bool kvm_arch_has_assigned_device(struct kvm
*kvm
);
1082 static inline void kvm_arch_start_assignment(struct kvm
*kvm
)
1086 static inline void kvm_arch_end_assignment(struct kvm
*kvm
)
1090 static inline bool kvm_arch_has_assigned_device(struct kvm
*kvm
)
1096 static inline struct rcuwait
*kvm_arch_vcpu_get_wait(struct kvm_vcpu
*vcpu
)
1098 #ifdef __KVM_HAVE_ARCH_WQP
1099 return vcpu
->arch
.waitp
;
1105 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
1107 * returns true if the virtual interrupt controller is initialized and
1108 * ready to accept virtual IRQ. On some architectures the virtual interrupt
1109 * controller is dynamically instantiated and this is not always true.
1111 bool kvm_arch_intc_initialized(struct kvm
*kvm
);
1113 static inline bool kvm_arch_intc_initialized(struct kvm
*kvm
)
1119 int kvm_arch_init_vm(struct kvm
*kvm
, unsigned long type
);
1120 void kvm_arch_destroy_vm(struct kvm
*kvm
);
1121 void kvm_arch_sync_events(struct kvm
*kvm
);
1123 int kvm_cpu_has_pending_timer(struct kvm_vcpu
*vcpu
);
1125 bool kvm_is_reserved_pfn(kvm_pfn_t pfn
);
1126 bool kvm_is_zone_device_pfn(kvm_pfn_t pfn
);
1127 bool kvm_is_transparent_hugepage(kvm_pfn_t pfn
);
1129 struct kvm_irq_ack_notifier
{
1130 struct hlist_node link
;
1132 void (*irq_acked
)(struct kvm_irq_ack_notifier
*kian
);
1135 int kvm_irq_map_gsi(struct kvm
*kvm
,
1136 struct kvm_kernel_irq_routing_entry
*entries
, int gsi
);
1137 int kvm_irq_map_chip_pin(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
1139 int kvm_set_irq(struct kvm
*kvm
, int irq_source_id
, u32 irq
, int level
,
1141 int kvm_set_msi(struct kvm_kernel_irq_routing_entry
*irq_entry
, struct kvm
*kvm
,
1142 int irq_source_id
, int level
, bool line_status
);
1143 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry
*e
,
1144 struct kvm
*kvm
, int irq_source_id
,
1145 int level
, bool line_status
);
1146 bool kvm_irq_has_notifier(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
1147 void kvm_notify_acked_gsi(struct kvm
*kvm
, int gsi
);
1148 void kvm_notify_acked_irq(struct kvm
*kvm
, unsigned irqchip
, unsigned pin
);
1149 void kvm_register_irq_ack_notifier(struct kvm
*kvm
,
1150 struct kvm_irq_ack_notifier
*kian
);
1151 void kvm_unregister_irq_ack_notifier(struct kvm
*kvm
,
1152 struct kvm_irq_ack_notifier
*kian
);
1153 int kvm_request_irq_source_id(struct kvm
*kvm
);
1154 void kvm_free_irq_source_id(struct kvm
*kvm
, int irq_source_id
);
1155 bool kvm_arch_irqfd_allowed(struct kvm
*kvm
, struct kvm_irqfd
*args
);
1158 * search_memslots() and __gfn_to_memslot() are here because they are
1159 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
1160 * gfn_to_memslot() itself isn't here as an inline because that would
1161 * bloat other code too much.
1163 * IMPORTANT: Slots are sorted from highest GFN to lowest GFN!
1165 static inline struct kvm_memory_slot
*
1166 search_memslots(struct kvm_memslots
*slots
, gfn_t gfn
)
1168 int start
= 0, end
= slots
->used_slots
;
1169 int slot
= atomic_read(&slots
->lru_slot
);
1170 struct kvm_memory_slot
*memslots
= slots
->memslots
;
1172 if (unlikely(!slots
->used_slots
))
1175 if (gfn
>= memslots
[slot
].base_gfn
&&
1176 gfn
< memslots
[slot
].base_gfn
+ memslots
[slot
].npages
)
1177 return &memslots
[slot
];
1179 while (start
< end
) {
1180 slot
= start
+ (end
- start
) / 2;
1182 if (gfn
>= memslots
[slot
].base_gfn
)
1188 if (start
< slots
->used_slots
&& gfn
>= memslots
[start
].base_gfn
&&
1189 gfn
< memslots
[start
].base_gfn
+ memslots
[start
].npages
) {
1190 atomic_set(&slots
->lru_slot
, start
);
1191 return &memslots
[start
];
1197 static inline struct kvm_memory_slot
*
1198 __gfn_to_memslot(struct kvm_memslots
*slots
, gfn_t gfn
)
1200 return search_memslots(slots
, gfn
);
1203 static inline unsigned long
1204 __gfn_to_hva_memslot(const struct kvm_memory_slot
*slot
, gfn_t gfn
)
1207 * The index was checked originally in search_memslots. To avoid
1208 * that a malicious guest builds a Spectre gadget out of e.g. page
1209 * table walks, do not let the processor speculate loads outside
1210 * the guest's registered memslots.
1212 unsigned long offset
= gfn
- slot
->base_gfn
;
1213 offset
= array_index_nospec(offset
, slot
->npages
);
1214 return slot
->userspace_addr
+ offset
* PAGE_SIZE
;
1217 static inline int memslot_id(struct kvm
*kvm
, gfn_t gfn
)
1219 return gfn_to_memslot(kvm
, gfn
)->id
;
1223 hva_to_gfn_memslot(unsigned long hva
, struct kvm_memory_slot
*slot
)
1225 gfn_t gfn_offset
= (hva
- slot
->userspace_addr
) >> PAGE_SHIFT
;
1227 return slot
->base_gfn
+ gfn_offset
;
1230 static inline gpa_t
gfn_to_gpa(gfn_t gfn
)
1232 return (gpa_t
)gfn
<< PAGE_SHIFT
;
1235 static inline gfn_t
gpa_to_gfn(gpa_t gpa
)
1237 return (gfn_t
)(gpa
>> PAGE_SHIFT
);
1240 static inline hpa_t
pfn_to_hpa(kvm_pfn_t pfn
)
1242 return (hpa_t
)pfn
<< PAGE_SHIFT
;
1245 static inline struct page
*kvm_vcpu_gpa_to_page(struct kvm_vcpu
*vcpu
,
1248 return kvm_vcpu_gfn_to_page(vcpu
, gpa_to_gfn(gpa
));
1251 static inline bool kvm_is_error_gpa(struct kvm
*kvm
, gpa_t gpa
)
1253 unsigned long hva
= gfn_to_hva(kvm
, gpa_to_gfn(gpa
));
1255 return kvm_is_error_hva(hva
);
1258 enum kvm_stat_kind
{
1263 struct kvm_stat_data
{
1265 struct kvm_stats_debugfs_item
*dbgfs_item
;
1268 struct kvm_stats_debugfs_item
{
1271 enum kvm_stat_kind kind
;
1275 struct _kvm_stats_desc
{
1276 struct kvm_stats_desc desc
;
1277 char name
[KVM_STATS_NAME_SIZE
];
1280 #define KVM_DBGFS_GET_MODE(dbgfs_item) \
1281 ((dbgfs_item)->mode ? (dbgfs_item)->mode : 0644)
1283 #define VM_STAT(n, x, ...) \
1284 { n, offsetof(struct kvm, stat.x), KVM_STAT_VM, ## __VA_ARGS__ }
1285 #define VCPU_STAT(n, x, ...) \
1286 { n, offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU, ## __VA_ARGS__ }
1288 #define STATS_DESC_COMMON(type, unit, base, exp) \
1289 .flags = type | unit | base | \
1290 BUILD_BUG_ON_ZERO(type & ~KVM_STATS_TYPE_MASK) | \
1291 BUILD_BUG_ON_ZERO(unit & ~KVM_STATS_UNIT_MASK) | \
1292 BUILD_BUG_ON_ZERO(base & ~KVM_STATS_BASE_MASK), \
1296 #define VM_GENERIC_STATS_DESC(stat, type, unit, base, exp) \
1299 STATS_DESC_COMMON(type, unit, base, exp), \
1300 .offset = offsetof(struct kvm_vm_stat, generic.stat) \
1304 #define VCPU_GENERIC_STATS_DESC(stat, type, unit, base, exp) \
1307 STATS_DESC_COMMON(type, unit, base, exp), \
1308 .offset = offsetof(struct kvm_vcpu_stat, generic.stat) \
1312 #define VM_STATS_DESC(stat, type, unit, base, exp) \
1315 STATS_DESC_COMMON(type, unit, base, exp), \
1316 .offset = offsetof(struct kvm_vm_stat, stat) \
1320 #define VCPU_STATS_DESC(stat, type, unit, base, exp) \
1323 STATS_DESC_COMMON(type, unit, base, exp), \
1324 .offset = offsetof(struct kvm_vcpu_stat, stat) \
1328 /* SCOPE: VM, VM_GENERIC, VCPU, VCPU_GENERIC */
1329 #define STATS_DESC(SCOPE, stat, type, unit, base, exp) \
1330 SCOPE##_STATS_DESC(stat, type, unit, base, exp)
1332 #define STATS_DESC_CUMULATIVE(SCOPE, name, unit, base, exponent) \
1333 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_CUMULATIVE, unit, base, exponent)
1334 #define STATS_DESC_INSTANT(SCOPE, name, unit, base, exponent) \
1335 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_INSTANT, unit, base, exponent)
1336 #define STATS_DESC_PEAK(SCOPE, name, unit, base, exponent) \
1337 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_PEAK, unit, base, exponent)
1339 /* Cumulative counter, read/write */
1340 #define STATS_DESC_COUNTER(SCOPE, name) \
1341 STATS_DESC_CUMULATIVE(SCOPE, name, KVM_STATS_UNIT_NONE, \
1342 KVM_STATS_BASE_POW10, 0)
1343 /* Instantaneous counter, read only */
1344 #define STATS_DESC_ICOUNTER(SCOPE, name) \
1345 STATS_DESC_INSTANT(SCOPE, name, KVM_STATS_UNIT_NONE, \
1346 KVM_STATS_BASE_POW10, 0)
1347 /* Peak counter, read/write */
1348 #define STATS_DESC_PCOUNTER(SCOPE, name) \
1349 STATS_DESC_PEAK(SCOPE, name, KVM_STATS_UNIT_NONE, \
1350 KVM_STATS_BASE_POW10, 0)
1352 /* Cumulative time in nanosecond */
1353 #define STATS_DESC_TIME_NSEC(SCOPE, name) \
1354 STATS_DESC_CUMULATIVE(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
1355 KVM_STATS_BASE_POW10, -9)
1357 extern struct kvm_stats_debugfs_item debugfs_entries
[];
1358 extern struct dentry
*kvm_debugfs_dir
;
1359 ssize_t
kvm_stats_read(char *id
, const struct kvm_stats_header
*header
,
1360 const struct _kvm_stats_desc
*desc
,
1361 void *stats
, size_t size_stats
,
1362 char __user
*user_buffer
, size_t size
, loff_t
*offset
);
1364 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1365 static inline int mmu_notifier_retry(struct kvm
*kvm
, unsigned long mmu_seq
)
1367 if (unlikely(kvm
->mmu_notifier_count
))
1370 * Ensure the read of mmu_notifier_count happens before the read
1371 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1372 * mmu_notifier_invalidate_range_end to make sure that the caller
1373 * either sees the old (non-zero) value of mmu_notifier_count or
1374 * the new (incremented) value of mmu_notifier_seq.
1375 * PowerPC Book3s HV KVM calls this under a per-page lock
1376 * rather than under kvm->mmu_lock, for scalability, so
1377 * can't rely on kvm->mmu_lock to keep things ordered.
1380 if (kvm
->mmu_notifier_seq
!= mmu_seq
)
1385 static inline int mmu_notifier_retry_hva(struct kvm
*kvm
,
1386 unsigned long mmu_seq
,
1389 lockdep_assert_held(&kvm
->mmu_lock
);
1391 * If mmu_notifier_count is non-zero, then the range maintained by
1392 * kvm_mmu_notifier_invalidate_range_start contains all addresses that
1393 * might be being invalidated. Note that it may include some false
1394 * positives, due to shortcuts when handing concurrent invalidations.
1396 if (unlikely(kvm
->mmu_notifier_count
) &&
1397 hva
>= kvm
->mmu_notifier_range_start
&&
1398 hva
< kvm
->mmu_notifier_range_end
)
1400 if (kvm
->mmu_notifier_seq
!= mmu_seq
)
1406 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1408 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1410 bool kvm_arch_can_set_irq_routing(struct kvm
*kvm
);
1411 int kvm_set_irq_routing(struct kvm
*kvm
,
1412 const struct kvm_irq_routing_entry
*entries
,
1415 int kvm_set_routing_entry(struct kvm
*kvm
,
1416 struct kvm_kernel_irq_routing_entry
*e
,
1417 const struct kvm_irq_routing_entry
*ue
);
1418 void kvm_free_irq_routing(struct kvm
*kvm
);
1422 static inline void kvm_free_irq_routing(struct kvm
*kvm
) {}
1426 int kvm_send_userspace_msi(struct kvm
*kvm
, struct kvm_msi
*msi
);
1428 #ifdef CONFIG_HAVE_KVM_EVENTFD
1430 void kvm_eventfd_init(struct kvm
*kvm
);
1431 int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
);
1433 #ifdef CONFIG_HAVE_KVM_IRQFD
1434 int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
);
1435 void kvm_irqfd_release(struct kvm
*kvm
);
1436 void kvm_irq_routing_update(struct kvm
*);
1438 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1443 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1448 static inline void kvm_eventfd_init(struct kvm
*kvm
) {}
1450 static inline int kvm_irqfd(struct kvm
*kvm
, struct kvm_irqfd
*args
)
1455 static inline void kvm_irqfd_release(struct kvm
*kvm
) {}
1457 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1458 static inline void kvm_irq_routing_update(struct kvm
*kvm
)
1463 static inline int kvm_ioeventfd(struct kvm
*kvm
, struct kvm_ioeventfd
*args
)
1468 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1470 void kvm_arch_irq_routing_update(struct kvm
*kvm
);
1472 static inline void kvm_make_request(int req
, struct kvm_vcpu
*vcpu
)
1475 * Ensure the rest of the request is published to kvm_check_request's
1476 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1479 set_bit(req
& KVM_REQUEST_MASK
, (void *)&vcpu
->requests
);
1482 static inline bool kvm_request_pending(struct kvm_vcpu
*vcpu
)
1484 return READ_ONCE(vcpu
->requests
);
1487 static inline bool kvm_test_request(int req
, struct kvm_vcpu
*vcpu
)
1489 return test_bit(req
& KVM_REQUEST_MASK
, (void *)&vcpu
->requests
);
1492 static inline void kvm_clear_request(int req
, struct kvm_vcpu
*vcpu
)
1494 clear_bit(req
& KVM_REQUEST_MASK
, (void *)&vcpu
->requests
);
1497 static inline bool kvm_check_request(int req
, struct kvm_vcpu
*vcpu
)
1499 if (kvm_test_request(req
, vcpu
)) {
1500 kvm_clear_request(req
, vcpu
);
1503 * Ensure the rest of the request is visible to kvm_check_request's
1504 * caller. Paired with the smp_wmb in kvm_make_request.
1506 smp_mb__after_atomic();
1513 extern bool kvm_rebooting
;
1515 extern unsigned int halt_poll_ns
;
1516 extern unsigned int halt_poll_ns_grow
;
1517 extern unsigned int halt_poll_ns_grow_start
;
1518 extern unsigned int halt_poll_ns_shrink
;
1521 const struct kvm_device_ops
*ops
;
1524 struct list_head vm_node
;
1527 /* create, destroy, and name are mandatory */
1528 struct kvm_device_ops
{
1532 * create is called holding kvm->lock and any operations not suitable
1533 * to do while holding the lock should be deferred to init (see
1536 int (*create
)(struct kvm_device
*dev
, u32 type
);
1539 * init is called after create if create is successful and is called
1540 * outside of holding kvm->lock.
1542 void (*init
)(struct kvm_device
*dev
);
1545 * Destroy is responsible for freeing dev.
1547 * Destroy may be called before or after destructors are called
1548 * on emulated I/O regions, depending on whether a reference is
1549 * held by a vcpu or other kvm component that gets destroyed
1550 * after the emulated I/O.
1552 void (*destroy
)(struct kvm_device
*dev
);
1555 * Release is an alternative method to free the device. It is
1556 * called when the device file descriptor is closed. Once
1557 * release is called, the destroy method will not be called
1558 * anymore as the device is removed from the device list of
1559 * the VM. kvm->lock is held.
1561 void (*release
)(struct kvm_device
*dev
);
1563 int (*set_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1564 int (*get_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1565 int (*has_attr
)(struct kvm_device
*dev
, struct kvm_device_attr
*attr
);
1566 long (*ioctl
)(struct kvm_device
*dev
, unsigned int ioctl
,
1568 int (*mmap
)(struct kvm_device
*dev
, struct vm_area_struct
*vma
);
1571 void kvm_device_get(struct kvm_device
*dev
);
1572 void kvm_device_put(struct kvm_device
*dev
);
1573 struct kvm_device
*kvm_device_from_filp(struct file
*filp
);
1574 int kvm_register_device_ops(const struct kvm_device_ops
*ops
, u32 type
);
1575 void kvm_unregister_device_ops(u32 type
);
1577 extern struct kvm_device_ops kvm_mpic_ops
;
1578 extern struct kvm_device_ops kvm_arm_vgic_v2_ops
;
1579 extern struct kvm_device_ops kvm_arm_vgic_v3_ops
;
1581 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1583 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1585 vcpu
->spin_loop
.in_spin_loop
= val
;
1587 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1589 vcpu
->spin_loop
.dy_eligible
= val
;
1592 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1594 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu
*vcpu
, bool val
)
1598 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu
*vcpu
, bool val
)
1601 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1603 static inline bool kvm_is_visible_memslot(struct kvm_memory_slot
*memslot
)
1605 return (memslot
&& memslot
->id
< KVM_USER_MEM_SLOTS
&&
1606 !(memslot
->flags
& KVM_MEMSLOT_INVALID
));
1609 struct kvm_vcpu
*kvm_get_running_vcpu(void);
1610 struct kvm_vcpu
* __percpu
*kvm_get_running_vcpus(void);
1612 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1613 bool kvm_arch_has_irq_bypass(void);
1614 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer
*,
1615 struct irq_bypass_producer
*);
1616 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer
*,
1617 struct irq_bypass_producer
*);
1618 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer
*);
1619 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer
*);
1620 int kvm_arch_update_irqfd_routing(struct kvm
*kvm
, unsigned int host_irq
,
1621 uint32_t guest_irq
, bool set
);
1622 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1624 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1625 /* If we wakeup during the poll time, was it a sucessful poll? */
1626 static inline bool vcpu_valid_wakeup(struct kvm_vcpu
*vcpu
)
1628 return vcpu
->valid_wakeup
;
1632 static inline bool vcpu_valid_wakeup(struct kvm_vcpu
*vcpu
)
1636 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
1638 #ifdef CONFIG_HAVE_KVM_NO_POLL
1639 /* Callback that tells if we must not poll */
1640 bool kvm_arch_no_poll(struct kvm_vcpu
*vcpu
);
1642 static inline bool kvm_arch_no_poll(struct kvm_vcpu
*vcpu
)
1646 #endif /* CONFIG_HAVE_KVM_NO_POLL */
1648 #ifdef CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL
1649 long kvm_arch_vcpu_async_ioctl(struct file
*filp
,
1650 unsigned int ioctl
, unsigned long arg
);
1652 static inline long kvm_arch_vcpu_async_ioctl(struct file
*filp
,
1656 return -ENOIOCTLCMD
;
1658 #endif /* CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL */
1660 void kvm_arch_mmu_notifier_invalidate_range(struct kvm
*kvm
,
1661 unsigned long start
, unsigned long end
);
1663 #ifdef CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE
1664 int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu
*vcpu
);
1666 static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu
*vcpu
)
1670 #endif /* CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE */
1672 typedef int (*kvm_vm_thread_fn_t
)(struct kvm
*kvm
, uintptr_t data
);
1674 int kvm_vm_create_worker_thread(struct kvm
*kvm
, kvm_vm_thread_fn_t thread_fn
,
1675 uintptr_t data
, const char *name
,
1676 struct task_struct
**thread_ptr
);
1678 #ifdef CONFIG_KVM_XFER_TO_GUEST_WORK
1679 static inline void kvm_handle_signal_exit(struct kvm_vcpu
*vcpu
)
1681 vcpu
->run
->exit_reason
= KVM_EXIT_INTR
;
1682 vcpu
->stat
.signal_exits
++;
1684 #endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */
1687 * This defines how many reserved entries we want to keep before we
1688 * kick the vcpu to the userspace to avoid dirty ring full. This
1689 * value can be tuned to higher if e.g. PML is enabled on the host.
1691 #define KVM_DIRTY_RING_RSVD_ENTRIES 64
1693 /* Max number of entries allowed for each kvm dirty ring */
1694 #define KVM_DIRTY_RING_MAX_ENTRIES 65536