Commit | Line | Data |
---|---|---|
af585b92 GN |
1 | /* |
2 | * kvm asynchronous fault support | |
3 | * | |
4 | * Copyright 2010 Red Hat, Inc. | |
5 | * | |
6 | * Author: | |
7 | * Gleb Natapov <gleb@redhat.com> | |
8 | * | |
9 | * This file is free software; you can redistribute it and/or modify | |
10 | * it under the terms of version 2 of the GNU General Public License | |
11 | * as published by the Free Software Foundation. | |
12 | * | |
13 | * This program is distributed in the hope that it will be useful, | |
14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
16 | * GNU General Public License for more details. | |
17 | * | |
18 | * You should have received a copy of the GNU General Public License | |
19 | * along with this program; if not, write to the Free Software Foundation, | |
20 | * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA. | |
21 | */ | |
22 | ||
23 | #include <linux/kvm_host.h> | |
24 | #include <linux/slab.h> | |
25 | #include <linux/module.h> | |
26 | #include <linux/mmu_context.h> | |
27 | ||
28 | #include "async_pf.h" | |
29 | #include <trace/events/kvm.h> | |
30 | ||
e0ead41a DD |
31 | static inline void kvm_async_page_present_sync(struct kvm_vcpu *vcpu, |
32 | struct kvm_async_pf *work) | |
33 | { | |
34 | #ifdef CONFIG_KVM_ASYNC_PF_SYNC | |
35 | kvm_arch_async_page_present(vcpu, work); | |
36 | #endif | |
37 | } | |
38 | static inline void kvm_async_page_present_async(struct kvm_vcpu *vcpu, | |
39 | struct kvm_async_pf *work) | |
40 | { | |
41 | #ifndef CONFIG_KVM_ASYNC_PF_SYNC | |
42 | kvm_arch_async_page_present(vcpu, work); | |
43 | #endif | |
44 | } | |
45 | ||
af585b92 GN |
46 | static struct kmem_cache *async_pf_cache; |
47 | ||
48 | int kvm_async_pf_init(void) | |
49 | { | |
50 | async_pf_cache = KMEM_CACHE(kvm_async_pf, 0); | |
51 | ||
52 | if (!async_pf_cache) | |
53 | return -ENOMEM; | |
54 | ||
55 | return 0; | |
56 | } | |
57 | ||
58 | void kvm_async_pf_deinit(void) | |
59 | { | |
60 | if (async_pf_cache) | |
61 | kmem_cache_destroy(async_pf_cache); | |
62 | async_pf_cache = NULL; | |
63 | } | |
64 | ||
65 | void kvm_async_pf_vcpu_init(struct kvm_vcpu *vcpu) | |
66 | { | |
67 | INIT_LIST_HEAD(&vcpu->async_pf.done); | |
68 | INIT_LIST_HEAD(&vcpu->async_pf.queue); | |
69 | spin_lock_init(&vcpu->async_pf.lock); | |
70 | } | |
71 | ||
72 | static void async_pf_execute(struct work_struct *work) | |
73 | { | |
af585b92 GN |
74 | struct kvm_async_pf *apf = |
75 | container_of(work, struct kvm_async_pf, work); | |
76 | struct mm_struct *mm = apf->mm; | |
77 | struct kvm_vcpu *vcpu = apf->vcpu; | |
78 | unsigned long addr = apf->addr; | |
79 | gva_t gva = apf->gva; | |
80 | ||
81 | might_sleep(); | |
82 | ||
af585b92 | 83 | down_read(&mm->mmap_sem); |
f2e10669 | 84 | get_user_pages(current, mm, addr, 1, 1, 0, NULL, NULL); |
af585b92 | 85 | up_read(&mm->mmap_sem); |
e0ead41a | 86 | kvm_async_page_present_sync(vcpu, apf); |
af585b92 GN |
87 | |
88 | spin_lock(&vcpu->async_pf.lock); | |
89 | list_add_tail(&apf->link, &vcpu->async_pf.done); | |
af585b92 GN |
90 | spin_unlock(&vcpu->async_pf.lock); |
91 | ||
92 | /* | |
93 | * apf may be freed by kvm_check_async_pf_completion() after | |
94 | * this point | |
95 | */ | |
96 | ||
f2e10669 | 97 | trace_kvm_async_pf_completed(addr, gva); |
af585b92 GN |
98 | |
99 | if (waitqueue_active(&vcpu->wq)) | |
100 | wake_up_interruptible(&vcpu->wq); | |
101 | ||
102 | mmdrop(mm); | |
103 | kvm_put_kvm(vcpu->kvm); | |
104 | } | |
105 | ||
106 | void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu) | |
107 | { | |
108 | /* cancel outstanding work queue item */ | |
109 | while (!list_empty(&vcpu->async_pf.queue)) { | |
110 | struct kvm_async_pf *work = | |
111 | list_entry(vcpu->async_pf.queue.next, | |
112 | typeof(*work), queue); | |
af585b92 | 113 | list_del(&work->queue); |
9f2ceda4 DD |
114 | |
115 | #ifdef CONFIG_KVM_ASYNC_PF_SYNC | |
116 | flush_work(&work->work); | |
117 | #else | |
98fda169 | 118 | if (cancel_work_sync(&work->work)) { |
28b441e2 RK |
119 | mmdrop(work->mm); |
120 | kvm_put_kvm(vcpu->kvm); /* == work->vcpu->kvm */ | |
af585b92 | 121 | kmem_cache_free(async_pf_cache, work); |
28b441e2 | 122 | } |
9f2ceda4 | 123 | #endif |
af585b92 GN |
124 | } |
125 | ||
126 | spin_lock(&vcpu->async_pf.lock); | |
127 | while (!list_empty(&vcpu->async_pf.done)) { | |
128 | struct kvm_async_pf *work = | |
129 | list_entry(vcpu->async_pf.done.next, | |
130 | typeof(*work), link); | |
131 | list_del(&work->link); | |
af585b92 GN |
132 | kmem_cache_free(async_pf_cache, work); |
133 | } | |
134 | spin_unlock(&vcpu->async_pf.lock); | |
135 | ||
136 | vcpu->async_pf.queued = 0; | |
137 | } | |
138 | ||
139 | void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu) | |
140 | { | |
141 | struct kvm_async_pf *work; | |
142 | ||
15096ffc XG |
143 | while (!list_empty_careful(&vcpu->async_pf.done) && |
144 | kvm_arch_can_inject_async_page_present(vcpu)) { | |
145 | spin_lock(&vcpu->async_pf.lock); | |
146 | work = list_first_entry(&vcpu->async_pf.done, typeof(*work), | |
147 | link); | |
148 | list_del(&work->link); | |
149 | spin_unlock(&vcpu->async_pf.lock); | |
af585b92 | 150 | |
f2e10669 | 151 | kvm_arch_async_page_ready(vcpu, work); |
1179ba53 | 152 | kvm_async_page_present_async(vcpu, work); |
af585b92 | 153 | |
15096ffc XG |
154 | list_del(&work->queue); |
155 | vcpu->async_pf.queued--; | |
15096ffc XG |
156 | kmem_cache_free(async_pf_cache, work); |
157 | } | |
af585b92 GN |
158 | } |
159 | ||
e0ead41a | 160 | int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva, |
af585b92 GN |
161 | struct kvm_arch_async_pf *arch) |
162 | { | |
163 | struct kvm_async_pf *work; | |
164 | ||
165 | if (vcpu->async_pf.queued >= ASYNC_PF_PER_VCPU) | |
166 | return 0; | |
167 | ||
168 | /* setup delayed work */ | |
169 | ||
170 | /* | |
171 | * do alloc nowait since if we are going to sleep anyway we | |
172 | * may as well sleep faulting in page | |
173 | */ | |
174 | work = kmem_cache_zalloc(async_pf_cache, GFP_NOWAIT); | |
175 | if (!work) | |
176 | return 0; | |
177 | ||
f2e10669 | 178 | work->wakeup_all = false; |
af585b92 GN |
179 | work->vcpu = vcpu; |
180 | work->gva = gva; | |
e0ead41a | 181 | work->addr = hva; |
af585b92 GN |
182 | work->arch = *arch; |
183 | work->mm = current->mm; | |
184 | atomic_inc(&work->mm->mm_count); | |
185 | kvm_get_kvm(work->vcpu->kvm); | |
186 | ||
187 | /* this can't really happen otherwise gfn_to_pfn_async | |
188 | would succeed */ | |
189 | if (unlikely(kvm_is_error_hva(work->addr))) | |
190 | goto retry_sync; | |
191 | ||
192 | INIT_WORK(&work->work, async_pf_execute); | |
193 | if (!schedule_work(&work->work)) | |
194 | goto retry_sync; | |
195 | ||
196 | list_add_tail(&work->queue, &vcpu->async_pf.queue); | |
197 | vcpu->async_pf.queued++; | |
198 | kvm_arch_async_page_not_present(vcpu, work); | |
199 | return 1; | |
200 | retry_sync: | |
201 | kvm_put_kvm(work->vcpu->kvm); | |
202 | mmdrop(work->mm); | |
203 | kmem_cache_free(async_pf_cache, work); | |
204 | return 0; | |
205 | } | |
344d9588 GN |
206 | |
207 | int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu) | |
208 | { | |
209 | struct kvm_async_pf *work; | |
210 | ||
64f638c7 | 211 | if (!list_empty_careful(&vcpu->async_pf.done)) |
344d9588 GN |
212 | return 0; |
213 | ||
214 | work = kmem_cache_zalloc(async_pf_cache, GFP_ATOMIC); | |
215 | if (!work) | |
216 | return -ENOMEM; | |
217 | ||
f2e10669 | 218 | work->wakeup_all = true; |
344d9588 GN |
219 | INIT_LIST_HEAD(&work->queue); /* for list_del to work */ |
220 | ||
64f638c7 | 221 | spin_lock(&vcpu->async_pf.lock); |
344d9588 | 222 | list_add_tail(&work->link, &vcpu->async_pf.done); |
64f638c7 XG |
223 | spin_unlock(&vcpu->async_pf.lock); |
224 | ||
344d9588 GN |
225 | vcpu->async_pf.queued++; |
226 | return 0; | |
227 | } |