blob: 5a0cd194dce00ee75b40b73d146ee478db1e4d38 [file] [log] [blame]
Avi Kivity6aa8b732006-12-10 02:21:36 -08001/*
2 * Kernel-based Virtual Machine driver for Linux
3 *
4 * This module enables machines with Intel VT-x extensions to run virtual
5 * machines without emulation or binary translation.
6 *
7 * Copyright (C) 2006 Qumranet, Inc.
8 *
9 * Authors:
10 * Avi Kivity <avi@qumranet.com>
11 * Yaniv Kamay <yaniv@qumranet.com>
12 *
13 * This work is licensed under the terms of the GNU GPL, version 2. See
14 * the COPYING file in the top-level directory.
15 *
16 */
17
Hollis Blancharde2174022007-12-03 15:30:24 -060018#include "iodev.h"
Avi Kivity6aa8b732006-12-10 02:21:36 -080019
Avi Kivityedf88412007-12-16 11:02:48 +020020#include <linux/kvm_host.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080021#include <linux/kvm.h>
22#include <linux/module.h>
23#include <linux/errno.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080024#include <linux/percpu.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080025#include <linux/mm.h>
26#include <linux/miscdevice.h>
27#include <linux/vmalloc.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080028#include <linux/reboot.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080029#include <linux/debugfs.h>
30#include <linux/highmem.h>
31#include <linux/file.h>
Avi Kivity59ae6c62007-02-12 00:54:48 -080032#include <linux/sysdev.h>
Avi Kivity774c47f2007-02-12 00:54:47 -080033#include <linux/cpu.h>
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +040034#include <linux/sched.h>
Avi Kivityd9e368d2007-06-07 19:18:30 +030035#include <linux/cpumask.h>
36#include <linux/smp.h>
Avi Kivityd6d28162007-06-28 08:38:16 -040037#include <linux/anon_inodes.h>
Avi Kivity04d2cc72007-09-10 18:10:54 +030038#include <linux/profile.h>
Anthony Liguori7aa81cc2007-09-17 14:57:50 -050039#include <linux/kvm_para.h>
Izik Eidus6fc138d2007-10-09 19:20:39 +020040#include <linux/pagemap.h>
Anthony Liguori8d4e1282007-10-18 09:59:34 -050041#include <linux/mman.h>
Anthony Liguori35149e22008-04-02 14:46:56 -050042#include <linux/swap.h>
Sheng Yange56d5322009-03-12 21:45:39 +080043#include <linux/bitops.h>
Marcelo Tosatti547de292009-05-07 17:55:13 -030044#include <linux/spinlock.h>
Arnd Bergmann6ff58942009-10-22 14:19:27 +020045#include <linux/compat.h>
Marcelo Tosattibc6678a2009-12-23 14:35:21 -020046#include <linux/srcu.h>
Joerg Roedel8f0b1ab2010-01-28 12:37:56 +010047#include <linux/hugetlb.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090048#include <linux/slab.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080049
Avi Kivitye4956062007-06-28 14:15:57 -040050#include <asm/processor.h>
Avi Kivitye4956062007-06-28 14:15:57 -040051#include <asm/io.h>
52#include <asm/uaccess.h>
Izik Eidus3e021bf2007-11-19 11:16:57 +020053#include <asm/pgtable.h>
Alexander Grafc8240bd2009-10-30 05:47:26 +000054#include <asm-generic/bitops/le.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080055
Laurent Vivier5f94c172008-05-30 16:05:54 +020056#include "coalesced_mmio.h"
Laurent Vivier5f94c172008-05-30 16:05:54 +020057
Marcelo Tosatti229456f2009-06-17 09:22:14 -030058#define CREATE_TRACE_POINTS
59#include <trace/events/kvm.h>
60
Avi Kivity6aa8b732006-12-10 02:21:36 -080061MODULE_AUTHOR("Qumranet");
62MODULE_LICENSE("GPL");
63
Marcelo Tosattifa40a822009-06-04 15:08:24 -030064/*
65 * Ordering of locks:
66 *
Sheng Yangfae3a352009-12-15 10:28:07 +080067 * kvm->lock --> kvm->slots_lock --> kvm->irq_lock
Marcelo Tosattifa40a822009-06-04 15:08:24 -030068 */
69
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +080070DEFINE_SPINLOCK(kvm_lock);
71LIST_HEAD(vm_list);
Avi Kivity133de902007-02-12 00:54:44 -080072
Rusty Russell7f59f492008-12-07 21:25:45 +103073static cpumask_var_t cpus_hardware_enabled;
Alexander Graf10474ae2009-09-15 11:37:46 +020074static int kvm_usage_count = 0;
75static atomic_t hardware_enable_failed;
Avi Kivity1b6c0162007-05-24 13:03:52 +030076
Rusty Russellc16f8622007-07-30 21:12:19 +100077struct kmem_cache *kvm_vcpu_cache;
78EXPORT_SYMBOL_GPL(kvm_vcpu_cache);
Avi Kivity1165f5f2007-04-19 17:27:43 +030079
Avi Kivity15ad7142007-07-11 18:17:21 +030080static __read_mostly struct preempt_ops kvm_preempt_ops;
81
Hollis Blanchard76f7c872008-04-15 16:05:42 -050082struct dentry *kvm_debugfs_dir;
Avi Kivity6aa8b732006-12-10 02:21:36 -080083
Avi Kivitybccf2152007-02-21 18:04:26 +020084static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl,
85 unsigned long arg);
Alexander Graf10474ae2009-09-15 11:37:46 +020086static int hardware_enable_all(void);
87static void hardware_disable_all(void);
Avi Kivitybccf2152007-02-21 18:04:26 +020088
Marcelo Tosattie93f8a02009-12-23 14:35:24 -020089static void kvm_io_bus_destroy(struct kvm_io_bus *bus);
90
Hannes Edere8ba5d32008-11-28 17:02:42 +010091static bool kvm_rebooting;
Avi Kivity4ecac3f2008-05-13 13:23:38 +030092
Marcelo Tosatti54dee992009-06-11 12:07:44 -030093static bool largepages_enabled = true;
94
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +080095inline int kvm_is_mmio_pfn(pfn_t pfn)
Ben-Ami Yassourcbff90a2008-07-28 19:26:24 +030096{
Joerg Roedelfc5659c2009-02-18 14:08:58 +010097 if (pfn_valid(pfn)) {
98 struct page *page = compound_head(pfn_to_page(pfn));
99 return PageReserved(page);
100 }
Ben-Ami Yassourcbff90a2008-07-28 19:26:24 +0300101
102 return true;
103}
104
Avi Kivity6aa8b732006-12-10 02:21:36 -0800105/*
106 * Switches to specified vcpu, until a matching vcpu_put()
107 */
Carsten Otte313a3dc2007-10-11 19:16:52 +0200108void vcpu_load(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800109{
Avi Kivity15ad7142007-07-11 18:17:21 +0300110 int cpu;
111
Avi Kivitybccf2152007-02-21 18:04:26 +0200112 mutex_lock(&vcpu->mutex);
Avi Kivity15ad7142007-07-11 18:17:21 +0300113 cpu = get_cpu();
114 preempt_notifier_register(&vcpu->preempt_notifier);
Carsten Otte313a3dc2007-10-11 19:16:52 +0200115 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +0300116 put_cpu();
Avi Kivitybccf2152007-02-21 18:04:26 +0200117}
118
Carsten Otte313a3dc2007-10-11 19:16:52 +0200119void vcpu_put(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800120{
Avi Kivity15ad7142007-07-11 18:17:21 +0300121 preempt_disable();
Carsten Otte313a3dc2007-10-11 19:16:52 +0200122 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +0300123 preempt_notifier_unregister(&vcpu->preempt_notifier);
124 preempt_enable();
Avi Kivity6aa8b732006-12-10 02:21:36 -0800125 mutex_unlock(&vcpu->mutex);
126}
127
Avi Kivityd9e368d2007-06-07 19:18:30 +0300128static void ack_flush(void *_completed)
129{
Avi Kivityd9e368d2007-06-07 19:18:30 +0300130}
131
Rusty Russell49846892008-12-08 20:26:24 +1030132static bool make_all_cpus_request(struct kvm *kvm, unsigned int req)
Avi Kivityd9e368d2007-06-07 19:18:30 +0300133{
Avi Kivity597a5f52008-07-20 14:24:22 +0300134 int i, cpu, me;
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030135 cpumask_var_t cpus;
136 bool called = true;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300137 struct kvm_vcpu *vcpu;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300138
Li Zefan79f55992009-06-15 14:58:26 +0800139 zalloc_cpumask_var(&cpus, GFP_ATOMIC);
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030140
Avi Kivity70e335e2010-02-18 11:25:22 +0200141 raw_spin_lock(&kvm->requests_lock);
Jan Kiszkae601e3b2009-07-20 11:30:12 +0200142 me = smp_processor_id();
Gleb Natapov988a2ca2009-06-09 15:56:29 +0300143 kvm_for_each_vcpu(i, vcpu, kvm) {
Rusty Russell49846892008-12-08 20:26:24 +1030144 if (test_and_set_bit(req, &vcpu->requests))
Avi Kivityd9e368d2007-06-07 19:18:30 +0300145 continue;
146 cpu = vcpu->cpu;
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030147 if (cpus != NULL && cpu != -1 && cpu != me)
148 cpumask_set_cpu(cpu, cpus);
Avi Kivityd9e368d2007-06-07 19:18:30 +0300149 }
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030150 if (unlikely(cpus == NULL))
151 smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1);
152 else if (!cpumask_empty(cpus))
153 smp_call_function_many(cpus, ack_flush, NULL, 1);
154 else
155 called = false;
Avi Kivity70e335e2010-02-18 11:25:22 +0200156 raw_spin_unlock(&kvm->requests_lock);
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030157 free_cpumask_var(cpus);
Rusty Russell49846892008-12-08 20:26:24 +1030158 return called;
159}
160
161void kvm_flush_remote_tlbs(struct kvm *kvm)
162{
163 if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH))
164 ++kvm->stat.remote_tlb_flush;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300165}
166
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500167void kvm_reload_remote_mmus(struct kvm *kvm)
168{
Rusty Russell49846892008-12-08 20:26:24 +1030169 make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD);
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500170}
171
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000172int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id)
173{
174 struct page *page;
175 int r;
176
177 mutex_init(&vcpu->mutex);
178 vcpu->cpu = -1;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000179 vcpu->kvm = kvm;
180 vcpu->vcpu_id = id;
Eddie Dongb6958ce2007-07-18 12:15:21 +0300181 init_waitqueue_head(&vcpu->wq);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000182
183 page = alloc_page(GFP_KERNEL | __GFP_ZERO);
184 if (!page) {
185 r = -ENOMEM;
186 goto fail;
187 }
188 vcpu->run = page_address(page);
189
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800190 r = kvm_arch_vcpu_init(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000191 if (r < 0)
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800192 goto fail_free_run;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000193 return 0;
194
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000195fail_free_run:
196 free_page((unsigned long)vcpu->run);
197fail:
Rusty Russell76fafa52007-10-08 10:50:48 +1000198 return r;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000199}
200EXPORT_SYMBOL_GPL(kvm_vcpu_init);
201
202void kvm_vcpu_uninit(struct kvm_vcpu *vcpu)
203{
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800204 kvm_arch_vcpu_uninit(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000205 free_page((unsigned long)vcpu->run);
206}
207EXPORT_SYMBOL_GPL(kvm_vcpu_uninit);
208
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200209#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
210static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn)
211{
212 return container_of(mn, struct kvm, mmu_notifier);
213}
214
215static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn,
216 struct mm_struct *mm,
217 unsigned long address)
218{
219 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200220 int need_tlb_flush, idx;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200221
222 /*
223 * When ->invalidate_page runs, the linux pte has been zapped
224 * already but the page is still allocated until
225 * ->invalidate_page returns. So if we increase the sequence
226 * here the kvm page fault will notice if the spte can't be
227 * established because the page is going to be freed. If
228 * instead the kvm page fault establishes the spte before
229 * ->invalidate_page runs, kvm_unmap_hva will release it
230 * before returning.
231 *
232 * The sequence increase only need to be seen at spin_unlock
233 * time, and not at spin_lock time.
234 *
235 * Increasing the sequence after the spin_unlock would be
236 * unsafe because the kvm page fault could then establish the
237 * pte after kvm_unmap_hva returned, without noticing the page
238 * is going to be freed.
239 */
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200240 idx = srcu_read_lock(&kvm->srcu);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200241 spin_lock(&kvm->mmu_lock);
242 kvm->mmu_notifier_seq++;
243 need_tlb_flush = kvm_unmap_hva(kvm, address);
244 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200245 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200246
247 /* we've to flush the tlb before the pages can be freed */
248 if (need_tlb_flush)
249 kvm_flush_remote_tlbs(kvm);
250
251}
252
Izik Eidus3da0dd42009-09-23 21:47:18 +0300253static void kvm_mmu_notifier_change_pte(struct mmu_notifier *mn,
254 struct mm_struct *mm,
255 unsigned long address,
256 pte_t pte)
257{
258 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200259 int idx;
Izik Eidus3da0dd42009-09-23 21:47:18 +0300260
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200261 idx = srcu_read_lock(&kvm->srcu);
Izik Eidus3da0dd42009-09-23 21:47:18 +0300262 spin_lock(&kvm->mmu_lock);
263 kvm->mmu_notifier_seq++;
264 kvm_set_spte_hva(kvm, address, pte);
265 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200266 srcu_read_unlock(&kvm->srcu, idx);
Izik Eidus3da0dd42009-09-23 21:47:18 +0300267}
268
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200269static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn,
270 struct mm_struct *mm,
271 unsigned long start,
272 unsigned long end)
273{
274 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200275 int need_tlb_flush = 0, idx;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200276
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200277 idx = srcu_read_lock(&kvm->srcu);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200278 spin_lock(&kvm->mmu_lock);
279 /*
280 * The count increase must become visible at unlock time as no
281 * spte can be established without taking the mmu_lock and
282 * count is also read inside the mmu_lock critical section.
283 */
284 kvm->mmu_notifier_count++;
285 for (; start < end; start += PAGE_SIZE)
286 need_tlb_flush |= kvm_unmap_hva(kvm, start);
287 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200288 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200289
290 /* we've to flush the tlb before the pages can be freed */
291 if (need_tlb_flush)
292 kvm_flush_remote_tlbs(kvm);
293}
294
295static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
296 struct mm_struct *mm,
297 unsigned long start,
298 unsigned long end)
299{
300 struct kvm *kvm = mmu_notifier_to_kvm(mn);
301
302 spin_lock(&kvm->mmu_lock);
303 /*
304 * This sequence increase will notify the kvm page fault that
305 * the page that is going to be mapped in the spte could have
306 * been freed.
307 */
308 kvm->mmu_notifier_seq++;
309 /*
310 * The above sequence increase must be visible before the
311 * below count decrease but both values are read by the kvm
312 * page fault under mmu_lock spinlock so we don't need to add
313 * a smb_wmb() here in between the two.
314 */
315 kvm->mmu_notifier_count--;
316 spin_unlock(&kvm->mmu_lock);
317
318 BUG_ON(kvm->mmu_notifier_count < 0);
319}
320
321static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
322 struct mm_struct *mm,
323 unsigned long address)
324{
325 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200326 int young, idx;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200327
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200328 idx = srcu_read_lock(&kvm->srcu);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200329 spin_lock(&kvm->mmu_lock);
330 young = kvm_age_hva(kvm, address);
331 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200332 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200333
334 if (young)
335 kvm_flush_remote_tlbs(kvm);
336
337 return young;
338}
339
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100340static void kvm_mmu_notifier_release(struct mmu_notifier *mn,
341 struct mm_struct *mm)
342{
343 struct kvm *kvm = mmu_notifier_to_kvm(mn);
344 kvm_arch_flush_shadow(kvm);
345}
346
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200347static const struct mmu_notifier_ops kvm_mmu_notifier_ops = {
348 .invalidate_page = kvm_mmu_notifier_invalidate_page,
349 .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start,
350 .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end,
351 .clear_flush_young = kvm_mmu_notifier_clear_flush_young,
Izik Eidus3da0dd42009-09-23 21:47:18 +0300352 .change_pte = kvm_mmu_notifier_change_pte,
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100353 .release = kvm_mmu_notifier_release,
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200354};
Avi Kivity4c07b0a2009-12-20 14:54:04 +0200355
356static int kvm_init_mmu_notifier(struct kvm *kvm)
357{
358 kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops;
359 return mmu_notifier_register(&kvm->mmu_notifier, current->mm);
360}
361
362#else /* !(CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER) */
363
364static int kvm_init_mmu_notifier(struct kvm *kvm)
365{
366 return 0;
367}
368
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200369#endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */
370
Avi Kivityf17abe92007-02-21 19:28:04 +0200371static struct kvm *kvm_create_vm(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800372{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200373 int r = 0, i;
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800374 struct kvm *kvm = kvm_arch_create_vm();
Avi Kivity6aa8b732006-12-10 02:21:36 -0800375
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800376 if (IS_ERR(kvm))
377 goto out;
Alexander Graf10474ae2009-09-15 11:37:46 +0200378
379 r = hardware_enable_all();
380 if (r)
381 goto out_err_nodisable;
382
Avi Kivity75858a82009-01-04 17:10:50 +0200383#ifdef CONFIG_HAVE_KVM_IRQCHIP
384 INIT_HLIST_HEAD(&kvm->mask_notifier_list);
Gleb Natapov136bdfe2009-08-24 11:54:23 +0300385 INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list);
Avi Kivity75858a82009-01-04 17:10:50 +0200386#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -0800387
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200388 r = -ENOMEM;
389 kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
390 if (!kvm->memslots)
391 goto out_err;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200392 if (init_srcu_struct(&kvm->srcu))
393 goto out_err;
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200394 for (i = 0; i < KVM_NR_BUSES; i++) {
395 kvm->buses[i] = kzalloc(sizeof(struct kvm_io_bus),
396 GFP_KERNEL);
397 if (!kvm->buses[i]) {
398 cleanup_srcu_struct(&kvm->srcu);
399 goto out_err;
400 }
401 }
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200402
Avi Kivity4c07b0a2009-12-20 14:54:04 +0200403 r = kvm_init_mmu_notifier(kvm);
Avi Kivity283d0c62009-12-20 14:25:19 +0200404 if (r) {
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200405 cleanup_srcu_struct(&kvm->srcu);
Avi Kivity283d0c62009-12-20 14:25:19 +0200406 goto out_err;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200407 }
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200408
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200409 kvm->mm = current->mm;
410 atomic_inc(&kvm->mm->mm_count);
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -0500411 spin_lock_init(&kvm->mmu_lock);
Avi Kivity70e335e2010-02-18 11:25:22 +0200412 raw_spin_lock_init(&kvm->requests_lock);
Gregory Haskinsd34e6b12009-07-07 17:08:49 -0400413 kvm_eventfd_init(kvm);
Shaohua Li11ec2802007-07-23 14:51:37 +0800414 mutex_init(&kvm->lock);
Marcelo Tosatti60eead72009-06-04 15:08:23 -0300415 mutex_init(&kvm->irq_lock);
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200416 mutex_init(&kvm->slots_lock);
Izik Eidusd39f13b2008-03-30 16:01:25 +0300417 atomic_set(&kvm->users_count, 1);
Rusty Russell5e58cfe2007-07-23 17:08:21 +1000418 spin_lock(&kvm_lock);
419 list_add(&kvm->vm_list, &vm_list);
420 spin_unlock(&kvm_lock);
Laurent Vivier5f94c172008-05-30 16:05:54 +0200421#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
422 kvm_coalesced_mmio_init(kvm);
423#endif
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800424out:
Avi Kivityf17abe92007-02-21 19:28:04 +0200425 return kvm;
Alexander Graf10474ae2009-09-15 11:37:46 +0200426
427out_err:
428 hardware_disable_all();
429out_err_nodisable:
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200430 for (i = 0; i < KVM_NR_BUSES; i++)
431 kfree(kvm->buses[i]);
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200432 kfree(kvm->memslots);
Alexander Graf10474ae2009-09-15 11:37:46 +0200433 kfree(kvm);
434 return ERR_PTR(r);
Avi Kivityf17abe92007-02-21 19:28:04 +0200435}
436
Avi Kivity6aa8b732006-12-10 02:21:36 -0800437/*
438 * Free any memory in @free but not in @dont.
439 */
440static void kvm_free_physmem_slot(struct kvm_memory_slot *free,
441 struct kvm_memory_slot *dont)
442{
Joerg Roedelec04b262009-06-19 15:16:23 +0200443 int i;
444
Izik Eidus290fc382007-09-27 14:11:22 +0200445 if (!dont || free->rmap != dont->rmap)
446 vfree(free->rmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800447
448 if (!dont || free->dirty_bitmap != dont->dirty_bitmap)
449 vfree(free->dirty_bitmap);
450
Joerg Roedelec04b262009-06-19 15:16:23 +0200451
452 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) {
453 if (!dont || free->lpage_info[i] != dont->lpage_info[i]) {
454 vfree(free->lpage_info[i]);
455 free->lpage_info[i] = NULL;
456 }
457 }
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300458
Avi Kivity6aa8b732006-12-10 02:21:36 -0800459 free->npages = 0;
Al Viro8b6d44c2007-02-09 16:38:40 +0000460 free->dirty_bitmap = NULL;
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500461 free->rmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800462}
463
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800464void kvm_free_physmem(struct kvm *kvm)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800465{
466 int i;
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200467 struct kvm_memslots *slots = kvm->memslots;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800468
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200469 for (i = 0; i < slots->nmemslots; ++i)
470 kvm_free_physmem_slot(&slots->memslots[i], NULL);
471
472 kfree(kvm->memslots);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800473}
474
Avi Kivityf17abe92007-02-21 19:28:04 +0200475static void kvm_destroy_vm(struct kvm *kvm)
476{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200477 int i;
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200478 struct mm_struct *mm = kvm->mm;
479
Sheng Yangad8ba2c2009-01-06 10:03:02 +0800480 kvm_arch_sync_events(kvm);
Avi Kivity133de902007-02-12 00:54:44 -0800481 spin_lock(&kvm_lock);
482 list_del(&kvm->vm_list);
483 spin_unlock(&kvm_lock);
Avi Kivity399ec802008-11-19 13:58:46 +0200484 kvm_free_irq_routing(kvm);
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200485 for (i = 0; i < KVM_NR_BUSES; i++)
486 kvm_io_bus_destroy(kvm->buses[i]);
Avi Kivity980da6c2009-12-20 15:13:43 +0200487 kvm_coalesced_mmio_free(kvm);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200488#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
489 mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
Gleb Natapovf00be0c2009-03-19 12:20:36 +0200490#else
491 kvm_arch_flush_shadow(kvm);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200492#endif
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800493 kvm_arch_destroy_vm(kvm);
Alexander Graf10474ae2009-09-15 11:37:46 +0200494 hardware_disable_all();
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200495 mmdrop(mm);
Avi Kivityf17abe92007-02-21 19:28:04 +0200496}
497
Izik Eidusd39f13b2008-03-30 16:01:25 +0300498void kvm_get_kvm(struct kvm *kvm)
499{
500 atomic_inc(&kvm->users_count);
501}
502EXPORT_SYMBOL_GPL(kvm_get_kvm);
503
504void kvm_put_kvm(struct kvm *kvm)
505{
506 if (atomic_dec_and_test(&kvm->users_count))
507 kvm_destroy_vm(kvm);
508}
509EXPORT_SYMBOL_GPL(kvm_put_kvm);
510
511
Avi Kivityf17abe92007-02-21 19:28:04 +0200512static int kvm_vm_release(struct inode *inode, struct file *filp)
513{
514 struct kvm *kvm = filp->private_data;
515
Gregory Haskins721eecb2009-05-20 10:30:49 -0400516 kvm_irqfd_release(kvm);
517
Izik Eidusd39f13b2008-03-30 16:01:25 +0300518 kvm_put_kvm(kvm);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800519 return 0;
520}
521
Avi Kivity6aa8b732006-12-10 02:21:36 -0800522/*
Avi Kivity6aa8b732006-12-10 02:21:36 -0800523 * Allocate some memory and give it an address in the guest physical address
524 * space.
525 *
526 * Discontiguous memory is allowed, mostly for framebuffers.
Sheng Yangf78e0e22007-10-29 09:40:42 +0800527 *
Marcelo Tosatti10589a42007-12-20 19:18:22 -0500528 * Must be called holding mmap_sem for write.
Avi Kivity6aa8b732006-12-10 02:21:36 -0800529 */
Sheng Yangf78e0e22007-10-29 09:40:42 +0800530int __kvm_set_memory_region(struct kvm *kvm,
531 struct kvm_userspace_memory_region *mem,
532 int user_alloc)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800533{
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200534 int r, flush_shadow = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800535 gfn_t base_gfn;
Heiko Carstens28bcb112009-09-03 17:35:35 +0200536 unsigned long npages;
537 unsigned long i;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800538 struct kvm_memory_slot *memslot;
539 struct kvm_memory_slot old, new;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200540 struct kvm_memslots *slots, *old_memslots;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800541
542 r = -EINVAL;
543 /* General sanity checks */
544 if (mem->memory_size & (PAGE_SIZE - 1))
545 goto out;
546 if (mem->guest_phys_addr & (PAGE_SIZE - 1))
547 goto out;
Sheng Yange7cacd42008-11-11 15:30:40 +0800548 if (user_alloc && (mem->userspace_addr & (PAGE_SIZE - 1)))
Hollis Blanchard78749802008-11-07 13:32:12 -0600549 goto out;
Izik Eiduse0d62c72007-10-24 23:57:46 +0200550 if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800551 goto out;
552 if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
553 goto out;
554
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200555 memslot = &kvm->memslots->memslots[mem->slot];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800556 base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
557 npages = mem->memory_size >> PAGE_SHIFT;
558
559 if (!npages)
560 mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES;
561
Avi Kivity6aa8b732006-12-10 02:21:36 -0800562 new = old = *memslot;
563
564 new.base_gfn = base_gfn;
565 new.npages = npages;
566 new.flags = mem->flags;
567
568 /* Disallow changing a memory slot's size. */
569 r = -EINVAL;
570 if (npages && old.npages && npages != old.npages)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800571 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800572
573 /* Check for overlaps */
574 r = -EEXIST;
575 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200576 struct kvm_memory_slot *s = &kvm->memslots->memslots[i];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800577
Jan Kiszka4cd481f2009-04-13 11:59:32 +0200578 if (s == memslot || !s->npages)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800579 continue;
580 if (!((base_gfn + npages <= s->base_gfn) ||
581 (base_gfn >= s->base_gfn + s->npages)))
Sheng Yangf78e0e22007-10-29 09:40:42 +0800582 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800583 }
Avi Kivity6aa8b732006-12-10 02:21:36 -0800584
Avi Kivity6aa8b732006-12-10 02:21:36 -0800585 /* Free page dirty bitmap if unneeded */
586 if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
Al Viro8b6d44c2007-02-09 16:38:40 +0000587 new.dirty_bitmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800588
589 r = -ENOMEM;
590
591 /* Allocate if a slot is being created */
Carsten Otteeff01142008-06-27 15:05:31 +0200592#ifndef CONFIG_S390
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500593 if (npages && !new.rmap) {
Mike Dayd77c26f2007-10-08 09:02:08 -0400594 new.rmap = vmalloc(npages * sizeof(struct page *));
Izik Eidus290fc382007-09-27 14:11:22 +0200595
596 if (!new.rmap)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800597 goto out_free;
Izik Eidus290fc382007-09-27 14:11:22 +0200598
Izik Eidus290fc382007-09-27 14:11:22 +0200599 memset(new.rmap, 0, npages * sizeof(*new.rmap));
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500600
Izik Eidus80b14b52007-10-25 11:54:04 +0200601 new.user_alloc = user_alloc;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200602 new.userspace_addr = mem->userspace_addr;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800603 }
Joerg Roedelec04b262009-06-19 15:16:23 +0200604 if (!npages)
605 goto skip_lpage;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300606
Joerg Roedelec04b262009-06-19 15:16:23 +0200607 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) {
Heiko Carstens28bcb112009-09-03 17:35:35 +0200608 unsigned long ugfn;
609 unsigned long j;
610 int lpages;
Joerg Roedelec04b262009-06-19 15:16:23 +0200611 int level = i + 2;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300612
Joerg Roedelec04b262009-06-19 15:16:23 +0200613 /* Avoid unused variable warning if no large pages */
614 (void)level;
615
616 if (new.lpage_info[i])
617 continue;
618
619 lpages = 1 + (base_gfn + npages - 1) /
620 KVM_PAGES_PER_HPAGE(level);
621 lpages -= base_gfn / KVM_PAGES_PER_HPAGE(level);
622
623 new.lpage_info[i] = vmalloc(lpages * sizeof(*new.lpage_info[i]));
624
625 if (!new.lpage_info[i])
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300626 goto out_free;
627
Joerg Roedelec04b262009-06-19 15:16:23 +0200628 memset(new.lpage_info[i], 0,
629 lpages * sizeof(*new.lpage_info[i]));
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300630
Joerg Roedelec04b262009-06-19 15:16:23 +0200631 if (base_gfn % KVM_PAGES_PER_HPAGE(level))
632 new.lpage_info[i][0].write_count = 1;
633 if ((base_gfn+npages) % KVM_PAGES_PER_HPAGE(level))
634 new.lpage_info[i][lpages - 1].write_count = 1;
Avi Kivityac045272009-06-08 15:52:39 +0300635 ugfn = new.userspace_addr >> PAGE_SHIFT;
636 /*
637 * If the gfn and userspace address are not aligned wrt each
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300638 * other, or if explicitly asked to, disable large page
639 * support for this slot
Avi Kivityac045272009-06-08 15:52:39 +0300640 */
Joerg Roedelec04b262009-06-19 15:16:23 +0200641 if ((base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1) ||
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300642 !largepages_enabled)
Joerg Roedelec04b262009-06-19 15:16:23 +0200643 for (j = 0; j < lpages; ++j)
644 new.lpage_info[i][j].write_count = 1;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300645 }
Avi Kivity6aa8b732006-12-10 02:21:36 -0800646
Joerg Roedelec04b262009-06-19 15:16:23 +0200647skip_lpage:
648
Avi Kivity6aa8b732006-12-10 02:21:36 -0800649 /* Allocate page dirty bitmap if needed */
650 if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
651 unsigned dirty_bytes = ALIGN(npages, BITS_PER_LONG) / 8;
652
653 new.dirty_bitmap = vmalloc(dirty_bytes);
654 if (!new.dirty_bitmap)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800655 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800656 memset(new.dirty_bitmap, 0, dirty_bytes);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200657 /* destroy any largepage mappings for dirty tracking */
Izik Eiduse2445842009-06-10 19:23:24 +0300658 if (old.npages)
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200659 flush_shadow = 1;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800660 }
Christian Borntraeger3eea8432009-06-23 17:24:06 +0200661#else /* not defined CONFIG_S390 */
662 new.user_alloc = user_alloc;
663 if (user_alloc)
664 new.userspace_addr = mem->userspace_addr;
Carsten Otteeff01142008-06-27 15:05:31 +0200665#endif /* not defined CONFIG_S390 */
Avi Kivity6aa8b732006-12-10 02:21:36 -0800666
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200667 if (!npages) {
668 r = -ENOMEM;
669 slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
670 if (!slots)
671 goto out_free;
672 memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots));
673 if (mem->slot >= slots->nmemslots)
674 slots->nmemslots = mem->slot + 1;
675 slots->memslots[mem->slot].flags |= KVM_MEMSLOT_INVALID;
676
677 old_memslots = kvm->memslots;
678 rcu_assign_pointer(kvm->memslots, slots);
679 synchronize_srcu_expedited(&kvm->srcu);
680 /* From this point no new shadow pages pointing to a deleted
681 * memslot will be created.
682 *
683 * validation of sp->gfn happens in:
684 * - gfn_to_hva (kvm_read_guest, gfn_to_pfn)
685 * - kvm_is_visible_gfn (mmu_check_roots)
686 */
Marcelo Tosatti34d4cb82008-07-10 20:49:31 -0300687 kvm_arch_flush_shadow(kvm);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200688 kfree(old_memslots);
689 }
Marcelo Tosatti34d4cb82008-07-10 20:49:31 -0300690
Marcelo Tosattif7784b82009-12-23 14:35:18 -0200691 r = kvm_arch_prepare_memory_region(kvm, &new, old, mem, user_alloc);
692 if (r)
693 goto out_free;
694
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200695#ifdef CONFIG_DMAR
696 /* map the pages in iommu page table */
697 if (npages) {
698 r = kvm_iommu_map_pages(kvm, &new);
699 if (r)
700 goto out_free;
701 }
702#endif
Andrea Arcangeli604b38a2008-07-25 16:32:03 +0200703
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200704 r = -ENOMEM;
705 slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
706 if (!slots)
707 goto out_free;
708 memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots));
709 if (mem->slot >= slots->nmemslots)
710 slots->nmemslots = mem->slot + 1;
711
712 /* actual memory is freed via old in kvm_free_physmem_slot below */
713 if (!npages) {
714 new.rmap = NULL;
715 new.dirty_bitmap = NULL;
716 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i)
717 new.lpage_info[i] = NULL;
718 }
719
720 slots->memslots[mem->slot] = new;
721 old_memslots = kvm->memslots;
722 rcu_assign_pointer(kvm->memslots, slots);
723 synchronize_srcu_expedited(&kvm->srcu);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800724
Marcelo Tosattif7784b82009-12-23 14:35:18 -0200725 kvm_arch_commit_memory_region(kvm, mem, old, user_alloc);
Zhang Xiantao3ad82a72007-11-20 13:11:38 +0800726
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200727 kvm_free_physmem_slot(&old, &new);
728 kfree(old_memslots);
729
730 if (flush_shadow)
731 kvm_arch_flush_shadow(kvm);
732
Avi Kivity6aa8b732006-12-10 02:21:36 -0800733 return 0;
734
Sheng Yangf78e0e22007-10-29 09:40:42 +0800735out_free:
Avi Kivity6aa8b732006-12-10 02:21:36 -0800736 kvm_free_physmem_slot(&new, &old);
737out:
738 return r;
Izik Eidus210c7c42007-10-24 23:52:57 +0200739
740}
Sheng Yangf78e0e22007-10-29 09:40:42 +0800741EXPORT_SYMBOL_GPL(__kvm_set_memory_region);
742
743int kvm_set_memory_region(struct kvm *kvm,
744 struct kvm_userspace_memory_region *mem,
745 int user_alloc)
746{
747 int r;
748
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200749 mutex_lock(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +0800750 r = __kvm_set_memory_region(kvm, mem, user_alloc);
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200751 mutex_unlock(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +0800752 return r;
753}
Izik Eidus210c7c42007-10-24 23:52:57 +0200754EXPORT_SYMBOL_GPL(kvm_set_memory_region);
755
Carsten Otte1fe779f2007-10-29 16:08:35 +0100756int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
757 struct
758 kvm_userspace_memory_region *mem,
759 int user_alloc)
Izik Eidus210c7c42007-10-24 23:52:57 +0200760{
Izik Eiduse0d62c72007-10-24 23:57:46 +0200761 if (mem->slot >= KVM_MEMORY_SLOTS)
762 return -EINVAL;
Izik Eidus210c7c42007-10-24 23:52:57 +0200763 return kvm_set_memory_region(kvm, mem, user_alloc);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800764}
765
Zhang Xiantao5bb064d2007-11-18 20:29:43 +0800766int kvm_get_dirty_log(struct kvm *kvm,
767 struct kvm_dirty_log *log, int *is_dirty)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800768{
769 struct kvm_memory_slot *memslot;
770 int r, i;
771 int n;
772 unsigned long any = 0;
773
Avi Kivity6aa8b732006-12-10 02:21:36 -0800774 r = -EINVAL;
775 if (log->slot >= KVM_MEMORY_SLOTS)
776 goto out;
777
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200778 memslot = &kvm->memslots->memslots[log->slot];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800779 r = -ENOENT;
780 if (!memslot->dirty_bitmap)
781 goto out;
782
Uri Lublincd1a4a92007-02-22 16:43:09 +0200783 n = ALIGN(memslot->npages, BITS_PER_LONG) / 8;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800784
Uri Lublincd1a4a92007-02-22 16:43:09 +0200785 for (i = 0; !any && i < n/sizeof(long); ++i)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800786 any = memslot->dirty_bitmap[i];
787
788 r = -EFAULT;
789 if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
790 goto out;
791
Zhang Xiantao5bb064d2007-11-18 20:29:43 +0800792 if (any)
793 *is_dirty = 1;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800794
795 r = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800796out:
Avi Kivity6aa8b732006-12-10 02:21:36 -0800797 return r;
798}
799
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300800void kvm_disable_largepages(void)
801{
802 largepages_enabled = false;
803}
804EXPORT_SYMBOL_GPL(kvm_disable_largepages);
805
Izik Eiduscea7bb22007-10-17 19:17:48 +0200806int is_error_page(struct page *page)
807{
808 return page == bad_page;
809}
810EXPORT_SYMBOL_GPL(is_error_page);
811
Anthony Liguori35149e22008-04-02 14:46:56 -0500812int is_error_pfn(pfn_t pfn)
813{
814 return pfn == bad_pfn;
815}
816EXPORT_SYMBOL_GPL(is_error_pfn);
817
Izik Eidusf9d46eb2007-11-11 22:02:22 +0200818static inline unsigned long bad_hva(void)
819{
820 return PAGE_OFFSET;
821}
822
823int kvm_is_error_hva(unsigned long addr)
824{
825 return addr == bad_hva();
826}
827EXPORT_SYMBOL_GPL(kvm_is_error_hva);
828
Izik Eidus28430992008-10-03 17:40:32 +0300829struct kvm_memory_slot *gfn_to_memslot_unaliased(struct kvm *kvm, gfn_t gfn)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800830{
831 int i;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200832 struct kvm_memslots *slots = rcu_dereference(kvm->memslots);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800833
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200834 for (i = 0; i < slots->nmemslots; ++i) {
835 struct kvm_memory_slot *memslot = &slots->memslots[i];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800836
837 if (gfn >= memslot->base_gfn
838 && gfn < memslot->base_gfn + memslot->npages)
839 return memslot;
840 }
Al Viro8b6d44c2007-02-09 16:38:40 +0000841 return NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800842}
Izik Eidus28430992008-10-03 17:40:32 +0300843EXPORT_SYMBOL_GPL(gfn_to_memslot_unaliased);
Avi Kivitye8207542007-03-30 16:54:30 +0300844
845struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn)
846{
847 gfn = unalias_gfn(kvm, gfn);
Izik Eidus28430992008-10-03 17:40:32 +0300848 return gfn_to_memslot_unaliased(kvm, gfn);
Avi Kivitye8207542007-03-30 16:54:30 +0300849}
Avi Kivity6aa8b732006-12-10 02:21:36 -0800850
Izik Eiduse0d62c72007-10-24 23:57:46 +0200851int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
852{
853 int i;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200854 struct kvm_memslots *slots = rcu_dereference(kvm->memslots);
Izik Eiduse0d62c72007-10-24 23:57:46 +0200855
Marcelo Tosattia983fb22009-12-23 14:35:23 -0200856 gfn = unalias_gfn_instantiation(kvm, gfn);
Izik Eiduse0d62c72007-10-24 23:57:46 +0200857 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200858 struct kvm_memory_slot *memslot = &slots->memslots[i];
Izik Eiduse0d62c72007-10-24 23:57:46 +0200859
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200860 if (memslot->flags & KVM_MEMSLOT_INVALID)
861 continue;
862
Izik Eiduse0d62c72007-10-24 23:57:46 +0200863 if (gfn >= memslot->base_gfn
864 && gfn < memslot->base_gfn + memslot->npages)
865 return 1;
866 }
867 return 0;
868}
869EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);
870
Joerg Roedel8f0b1ab2010-01-28 12:37:56 +0100871unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn)
872{
873 struct vm_area_struct *vma;
874 unsigned long addr, size;
875
876 size = PAGE_SIZE;
877
878 addr = gfn_to_hva(kvm, gfn);
879 if (kvm_is_error_hva(addr))
880 return PAGE_SIZE;
881
882 down_read(&current->mm->mmap_sem);
883 vma = find_vma(current->mm, addr);
884 if (!vma)
885 goto out;
886
887 size = vma_kernel_pagesize(vma);
888
889out:
890 up_read(&current->mm->mmap_sem);
891
892 return size;
893}
894
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200895int memslot_id(struct kvm *kvm, gfn_t gfn)
896{
897 int i;
898 struct kvm_memslots *slots = rcu_dereference(kvm->memslots);
899 struct kvm_memory_slot *memslot = NULL;
900
901 gfn = unalias_gfn(kvm, gfn);
902 for (i = 0; i < slots->nmemslots; ++i) {
903 memslot = &slots->memslots[i];
904
905 if (gfn >= memslot->base_gfn
906 && gfn < memslot->base_gfn + memslot->npages)
907 break;
908 }
909
910 return memslot - slots->memslots;
911}
912
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300913unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
Izik Eidus539cb662007-11-11 22:05:04 +0200914{
915 struct kvm_memory_slot *slot;
916
Marcelo Tosattia983fb22009-12-23 14:35:23 -0200917 gfn = unalias_gfn_instantiation(kvm, gfn);
Izik Eidus28430992008-10-03 17:40:32 +0300918 slot = gfn_to_memslot_unaliased(kvm, gfn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200919 if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
Izik Eidus539cb662007-11-11 22:05:04 +0200920 return bad_hva();
921 return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE);
922}
Sheng Yang0d150292008-04-25 21:44:50 +0800923EXPORT_SYMBOL_GPL(gfn_to_hva);
Izik Eidus539cb662007-11-11 22:05:04 +0200924
Marcelo Tosatti506f0d62009-12-23 14:35:19 -0200925static pfn_t hva_to_pfn(struct kvm *kvm, unsigned long addr)
Avi Kivity954bbbc2007-03-30 14:02:32 +0300926{
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500927 struct page *page[1];
928 int npages;
Anthony Liguori2e2e3732008-04-30 15:37:07 -0500929 pfn_t pfn;
Avi Kivity954bbbc2007-03-30 14:02:32 +0300930
Avi Kivity60395222007-10-21 11:03:36 +0200931 might_sleep();
932
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -0300933 npages = get_user_pages_fast(addr, 1, 1, page);
Izik Eidus539cb662007-11-11 22:05:04 +0200934
Anthony Liguori2e2e3732008-04-30 15:37:07 -0500935 if (unlikely(npages != 1)) {
936 struct vm_area_struct *vma;
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500937
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -0300938 down_read(&current->mm->mmap_sem);
Anthony Liguori2e2e3732008-04-30 15:37:07 -0500939 vma = find_vma(current->mm, addr);
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -0300940
Anthony Liguori2e2e3732008-04-30 15:37:07 -0500941 if (vma == NULL || addr < vma->vm_start ||
942 !(vma->vm_flags & VM_PFNMAP)) {
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -0300943 up_read(&current->mm->mmap_sem);
Anthony Liguori2e2e3732008-04-30 15:37:07 -0500944 get_page(bad_page);
945 return page_to_pfn(bad_page);
946 }
947
948 pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -0300949 up_read(&current->mm->mmap_sem);
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +0800950 BUG_ON(!kvm_is_mmio_pfn(pfn));
Anthony Liguori2e2e3732008-04-30 15:37:07 -0500951 } else
952 pfn = page_to_pfn(page[0]);
953
954 return pfn;
Anthony Liguori35149e22008-04-02 14:46:56 -0500955}
956
Marcelo Tosatti506f0d62009-12-23 14:35:19 -0200957pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
958{
959 unsigned long addr;
960
961 addr = gfn_to_hva(kvm, gfn);
962 if (kvm_is_error_hva(addr)) {
963 get_page(bad_page);
964 return page_to_pfn(bad_page);
965 }
966
967 return hva_to_pfn(kvm, addr);
968}
Anthony Liguori35149e22008-04-02 14:46:56 -0500969EXPORT_SYMBOL_GPL(gfn_to_pfn);
970
Marcelo Tosatti506f0d62009-12-23 14:35:19 -0200971static unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
972{
973 return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE);
974}
975
976pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
977 struct kvm_memory_slot *slot, gfn_t gfn)
978{
979 unsigned long addr = gfn_to_hva_memslot(slot, gfn);
980 return hva_to_pfn(kvm, addr);
981}
982
Anthony Liguori35149e22008-04-02 14:46:56 -0500983struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
984{
Anthony Liguori2e2e3732008-04-30 15:37:07 -0500985 pfn_t pfn;
986
987 pfn = gfn_to_pfn(kvm, gfn);
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +0800988 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -0500989 return pfn_to_page(pfn);
990
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +0800991 WARN_ON(kvm_is_mmio_pfn(pfn));
Anthony Liguori2e2e3732008-04-30 15:37:07 -0500992
993 get_page(bad_page);
994 return bad_page;
Avi Kivity954bbbc2007-03-30 14:02:32 +0300995}
Anthony Liguoriaab61cc2007-10-29 15:15:20 -0500996
Avi Kivity954bbbc2007-03-30 14:02:32 +0300997EXPORT_SYMBOL_GPL(gfn_to_page);
998
Izik Eidusb4231d62007-11-20 11:49:33 +0200999void kvm_release_page_clean(struct page *page)
1000{
Anthony Liguori35149e22008-04-02 14:46:56 -05001001 kvm_release_pfn_clean(page_to_pfn(page));
Izik Eidusb4231d62007-11-20 11:49:33 +02001002}
1003EXPORT_SYMBOL_GPL(kvm_release_page_clean);
1004
Anthony Liguori35149e22008-04-02 14:46:56 -05001005void kvm_release_pfn_clean(pfn_t pfn)
1006{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001007 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001008 put_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001009}
1010EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);
1011
Izik Eidusb4231d62007-11-20 11:49:33 +02001012void kvm_release_page_dirty(struct page *page)
Izik Eidus8a7ae052007-10-18 11:09:33 +02001013{
Anthony Liguori35149e22008-04-02 14:46:56 -05001014 kvm_release_pfn_dirty(page_to_pfn(page));
Izik Eidus8a7ae052007-10-18 11:09:33 +02001015}
Izik Eidusb4231d62007-11-20 11:49:33 +02001016EXPORT_SYMBOL_GPL(kvm_release_page_dirty);
Izik Eidus8a7ae052007-10-18 11:09:33 +02001017
Anthony Liguori35149e22008-04-02 14:46:56 -05001018void kvm_release_pfn_dirty(pfn_t pfn)
1019{
1020 kvm_set_pfn_dirty(pfn);
1021 kvm_release_pfn_clean(pfn);
1022}
1023EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
1024
1025void kvm_set_page_dirty(struct page *page)
1026{
1027 kvm_set_pfn_dirty(page_to_pfn(page));
1028}
1029EXPORT_SYMBOL_GPL(kvm_set_page_dirty);
1030
1031void kvm_set_pfn_dirty(pfn_t pfn)
1032{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001033 if (!kvm_is_mmio_pfn(pfn)) {
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001034 struct page *page = pfn_to_page(pfn);
1035 if (!PageReserved(page))
1036 SetPageDirty(page);
1037 }
Anthony Liguori35149e22008-04-02 14:46:56 -05001038}
1039EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);
1040
1041void kvm_set_pfn_accessed(pfn_t pfn)
1042{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001043 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001044 mark_page_accessed(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001045}
1046EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);
1047
1048void kvm_get_pfn(pfn_t pfn)
1049{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001050 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001051 get_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001052}
1053EXPORT_SYMBOL_GPL(kvm_get_pfn);
1054
Izik Eidus195aefd2007-10-01 22:14:18 +02001055static int next_segment(unsigned long len, int offset)
1056{
1057 if (len > PAGE_SIZE - offset)
1058 return PAGE_SIZE - offset;
1059 else
1060 return len;
1061}
1062
1063int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
1064 int len)
1065{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001066 int r;
1067 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001068
Izik Eiduse0506bc2007-11-11 22:10:22 +02001069 addr = gfn_to_hva(kvm, gfn);
1070 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001071 return -EFAULT;
Izik Eiduse0506bc2007-11-11 22:10:22 +02001072 r = copy_from_user(data, (void __user *)addr + offset, len);
1073 if (r)
1074 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001075 return 0;
1076}
1077EXPORT_SYMBOL_GPL(kvm_read_guest_page);
1078
1079int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len)
1080{
1081 gfn_t gfn = gpa >> PAGE_SHIFT;
1082 int seg;
1083 int offset = offset_in_page(gpa);
1084 int ret;
1085
1086 while ((seg = next_segment(len, offset)) != 0) {
1087 ret = kvm_read_guest_page(kvm, gfn, data, offset, seg);
1088 if (ret < 0)
1089 return ret;
1090 offset = 0;
1091 len -= seg;
1092 data += seg;
1093 ++gfn;
1094 }
1095 return 0;
1096}
1097EXPORT_SYMBOL_GPL(kvm_read_guest);
1098
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001099int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
1100 unsigned long len)
1101{
1102 int r;
1103 unsigned long addr;
1104 gfn_t gfn = gpa >> PAGE_SHIFT;
1105 int offset = offset_in_page(gpa);
1106
1107 addr = gfn_to_hva(kvm, gfn);
1108 if (kvm_is_error_hva(addr))
1109 return -EFAULT;
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001110 pagefault_disable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001111 r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001112 pagefault_enable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001113 if (r)
1114 return -EFAULT;
1115 return 0;
1116}
1117EXPORT_SYMBOL(kvm_read_guest_atomic);
1118
Izik Eidus195aefd2007-10-01 22:14:18 +02001119int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
1120 int offset, int len)
1121{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001122 int r;
1123 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001124
Izik Eiduse0506bc2007-11-11 22:10:22 +02001125 addr = gfn_to_hva(kvm, gfn);
1126 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001127 return -EFAULT;
Izik Eiduse0506bc2007-11-11 22:10:22 +02001128 r = copy_to_user((void __user *)addr + offset, data, len);
1129 if (r)
1130 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001131 mark_page_dirty(kvm, gfn);
1132 return 0;
1133}
1134EXPORT_SYMBOL_GPL(kvm_write_guest_page);
1135
1136int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
1137 unsigned long len)
1138{
1139 gfn_t gfn = gpa >> PAGE_SHIFT;
1140 int seg;
1141 int offset = offset_in_page(gpa);
1142 int ret;
1143
1144 while ((seg = next_segment(len, offset)) != 0) {
1145 ret = kvm_write_guest_page(kvm, gfn, data, offset, seg);
1146 if (ret < 0)
1147 return ret;
1148 offset = 0;
1149 len -= seg;
1150 data += seg;
1151 ++gfn;
1152 }
1153 return 0;
1154}
1155
1156int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
1157{
Izik Eidus3e021bf2007-11-19 11:16:57 +02001158 return kvm_write_guest_page(kvm, gfn, empty_zero_page, offset, len);
Izik Eidus195aefd2007-10-01 22:14:18 +02001159}
1160EXPORT_SYMBOL_GPL(kvm_clear_guest_page);
1161
1162int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len)
1163{
1164 gfn_t gfn = gpa >> PAGE_SHIFT;
1165 int seg;
1166 int offset = offset_in_page(gpa);
1167 int ret;
1168
1169 while ((seg = next_segment(len, offset)) != 0) {
1170 ret = kvm_clear_guest_page(kvm, gfn, offset, seg);
1171 if (ret < 0)
1172 return ret;
1173 offset = 0;
1174 len -= seg;
1175 ++gfn;
1176 }
1177 return 0;
1178}
1179EXPORT_SYMBOL_GPL(kvm_clear_guest);
1180
Avi Kivity6aa8b732006-12-10 02:21:36 -08001181void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
1182{
Nguyen Anh Quynh31389942007-06-05 10:35:19 +03001183 struct kvm_memory_slot *memslot;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001184
Uri Lublin3b6fff12007-10-30 10:42:09 +02001185 gfn = unalias_gfn(kvm, gfn);
Izik Eidus28430992008-10-03 17:40:32 +03001186 memslot = gfn_to_memslot_unaliased(kvm, gfn);
Rusty Russell7e9d6192007-07-31 20:41:14 +10001187 if (memslot && memslot->dirty_bitmap) {
1188 unsigned long rel_gfn = gfn - memslot->base_gfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001189
Rusty Russell7e9d6192007-07-31 20:41:14 +10001190 /* avoid RMW */
Alexander Grafc8240bd2009-10-30 05:47:26 +00001191 if (!generic_test_le_bit(rel_gfn, memslot->dirty_bitmap))
1192 generic___set_le_bit(rel_gfn, memslot->dirty_bitmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001193 }
1194}
1195
Eddie Dongb6958ce2007-07-18 12:15:21 +03001196/*
1197 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
1198 */
Hollis Blanchard8776e512007-10-31 17:24:24 -05001199void kvm_vcpu_block(struct kvm_vcpu *vcpu)
Eddie Dongb6958ce2007-07-18 12:15:21 +03001200{
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001201 DEFINE_WAIT(wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001202
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001203 for (;;) {
1204 prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001205
Gleb Natapova1b37102009-07-09 15:33:52 +03001206 if (kvm_arch_vcpu_runnable(vcpu)) {
Marcelo Tosattid7690172008-09-08 15:23:48 -03001207 set_bit(KVM_REQ_UNHALT, &vcpu->requests);
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001208 break;
Marcelo Tosattid7690172008-09-08 15:23:48 -03001209 }
Gleb Natapov09cec752009-03-23 15:11:44 +02001210 if (kvm_cpu_has_pending_timer(vcpu))
1211 break;
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001212 if (signal_pending(current))
1213 break;
1214
Eddie Dongb6958ce2007-07-18 12:15:21 +03001215 schedule();
Eddie Dongb6958ce2007-07-18 12:15:21 +03001216 }
1217
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001218 finish_wait(&vcpu->wq, &wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001219}
1220
Avi Kivity6aa8b732006-12-10 02:21:36 -08001221void kvm_resched(struct kvm_vcpu *vcpu)
1222{
Yaozu Dong3fca0362007-04-25 16:49:19 +03001223 if (!need_resched())
1224 return;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001225 cond_resched();
Avi Kivity6aa8b732006-12-10 02:21:36 -08001226}
1227EXPORT_SYMBOL_GPL(kvm_resched);
1228
Zhai, Edwind255f4f2009-10-09 18:03:20 +08001229void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu)
1230{
1231 ktime_t expires;
1232 DEFINE_WAIT(wait);
1233
1234 prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
1235
1236 /* Sleep for 100 us, and hope lock-holder got scheduled */
1237 expires = ktime_add_ns(ktime_get(), 100000UL);
1238 schedule_hrtimeout(&expires, HRTIMER_MODE_ABS);
1239
1240 finish_wait(&vcpu->wq, &wait);
1241}
1242EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin);
1243
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001244static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001245{
1246 struct kvm_vcpu *vcpu = vma->vm_file->private_data;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001247 struct page *page;
1248
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001249 if (vmf->pgoff == 0)
Avi Kivity039576c2007-03-20 12:46:50 +02001250 page = virt_to_page(vcpu->run);
Avi Kivity09566762008-01-23 18:14:23 +02001251#ifdef CONFIG_X86
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001252 else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
Zhang Xiantaoad312c72007-12-13 23:50:52 +08001253 page = virt_to_page(vcpu->arch.pio_data);
Avi Kivity09566762008-01-23 18:14:23 +02001254#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02001255#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1256 else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
1257 page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
1258#endif
Avi Kivity039576c2007-03-20 12:46:50 +02001259 else
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001260 return VM_FAULT_SIGBUS;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001261 get_page(page);
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001262 vmf->page = page;
1263 return 0;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001264}
1265
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +04001266static const struct vm_operations_struct kvm_vcpu_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001267 .fault = kvm_vcpu_fault,
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001268};
1269
1270static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma)
1271{
1272 vma->vm_ops = &kvm_vcpu_vm_ops;
1273 return 0;
1274}
1275
Avi Kivitybccf2152007-02-21 18:04:26 +02001276static int kvm_vcpu_release(struct inode *inode, struct file *filp)
1277{
1278 struct kvm_vcpu *vcpu = filp->private_data;
1279
Al Viro66c0b392008-04-19 20:33:56 +01001280 kvm_put_kvm(vcpu->kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001281 return 0;
1282}
1283
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01001284static struct file_operations kvm_vcpu_fops = {
Avi Kivitybccf2152007-02-21 18:04:26 +02001285 .release = kvm_vcpu_release,
1286 .unlocked_ioctl = kvm_vcpu_ioctl,
1287 .compat_ioctl = kvm_vcpu_ioctl,
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001288 .mmap = kvm_vcpu_mmap,
Avi Kivitybccf2152007-02-21 18:04:26 +02001289};
1290
1291/*
1292 * Allocates an inode for the vcpu.
1293 */
1294static int create_vcpu_fd(struct kvm_vcpu *vcpu)
1295{
Roland Dreier628ff7c2009-12-18 09:41:24 -08001296 return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, O_RDWR);
Avi Kivitybccf2152007-02-21 18:04:26 +02001297}
1298
Avi Kivityc5ea7662007-02-20 18:41:05 +02001299/*
1300 * Creates some virtual cpus. Good luck creating more than one.
1301 */
Gleb Natapov73880c82009-06-09 15:56:28 +03001302static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
Avi Kivityc5ea7662007-02-20 18:41:05 +02001303{
1304 int r;
Gleb Natapov988a2ca2009-06-09 15:56:29 +03001305 struct kvm_vcpu *vcpu, *v;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001306
Gleb Natapov73880c82009-06-09 15:56:28 +03001307 vcpu = kvm_arch_vcpu_create(kvm, id);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001308 if (IS_ERR(vcpu))
1309 return PTR_ERR(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001310
Avi Kivity15ad7142007-07-11 18:17:21 +03001311 preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);
1312
Avi Kivity26e52152007-11-20 15:30:24 +02001313 r = kvm_arch_vcpu_setup(vcpu);
1314 if (r)
Glauber Costa7d8fece2008-09-17 23:16:59 -03001315 return r;
Avi Kivity26e52152007-11-20 15:30:24 +02001316
Shaohua Li11ec2802007-07-23 14:51:37 +08001317 mutex_lock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03001318 if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) {
1319 r = -EINVAL;
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001320 goto vcpu_destroy;
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001321 }
Gleb Natapov73880c82009-06-09 15:56:28 +03001322
Gleb Natapov988a2ca2009-06-09 15:56:29 +03001323 kvm_for_each_vcpu(r, v, kvm)
1324 if (v->vcpu_id == id) {
Gleb Natapov73880c82009-06-09 15:56:28 +03001325 r = -EEXIST;
1326 goto vcpu_destroy;
1327 }
1328
1329 BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001330
1331 /* Now it's all set up, let userspace reach it */
Al Viro66c0b392008-04-19 20:33:56 +01001332 kvm_get_kvm(kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001333 r = create_vcpu_fd(vcpu);
Gleb Natapov73880c82009-06-09 15:56:28 +03001334 if (r < 0) {
1335 kvm_put_kvm(kvm);
1336 goto vcpu_destroy;
1337 }
1338
1339 kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu;
1340 smp_wmb();
1341 atomic_inc(&kvm->online_vcpus);
1342
1343#ifdef CONFIG_KVM_APIC_ARCHITECTURE
1344 if (kvm->bsp_vcpu_id == id)
1345 kvm->bsp_vcpu = vcpu;
1346#endif
1347 mutex_unlock(&kvm->lock);
Avi Kivitybccf2152007-02-21 18:04:26 +02001348 return r;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001349
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001350vcpu_destroy:
Glauber Costa7d8fece2008-09-17 23:16:59 -03001351 mutex_unlock(&kvm->lock);
Hollis Blanchardd40ccc62007-11-19 14:04:43 -06001352 kvm_arch_vcpu_destroy(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001353 return r;
1354}
1355
Avi Kivity1961d272007-03-05 19:46:05 +02001356static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset)
1357{
1358 if (sigset) {
1359 sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP));
1360 vcpu->sigset_active = 1;
1361 vcpu->sigset = *sigset;
1362 } else
1363 vcpu->sigset_active = 0;
1364 return 0;
1365}
1366
Avi Kivitybccf2152007-02-21 18:04:26 +02001367static long kvm_vcpu_ioctl(struct file *filp,
1368 unsigned int ioctl, unsigned long arg)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001369{
Avi Kivitybccf2152007-02-21 18:04:26 +02001370 struct kvm_vcpu *vcpu = filp->private_data;
Al Viro2f3669872007-02-09 16:38:35 +00001371 void __user *argp = (void __user *)arg;
Carsten Otte313a3dc2007-10-11 19:16:52 +02001372 int r;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001373 struct kvm_fpu *fpu = NULL;
1374 struct kvm_sregs *kvm_sregs = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001375
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001376 if (vcpu->kvm->mm != current->mm)
1377 return -EIO;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001378 switch (ioctl) {
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001379 case KVM_RUN:
Avi Kivityf0fe5102007-03-07 13:11:17 +02001380 r = -EINVAL;
1381 if (arg)
1382 goto out;
Hollis Blanchardb6c7a5d2007-11-01 14:16:10 -05001383 r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001384 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001385 case KVM_GET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001386 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001387
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001388 r = -ENOMEM;
1389 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
1390 if (!kvm_regs)
1391 goto out;
1392 r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001393 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001394 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001395 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001396 if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
1397 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001398 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001399out_free1:
1400 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001401 break;
1402 }
1403 case KVM_SET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001404 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001405
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001406 r = -ENOMEM;
1407 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
1408 if (!kvm_regs)
1409 goto out;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001410 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001411 if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs)))
1412 goto out_free2;
1413 r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001414 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001415 goto out_free2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001416 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001417out_free2:
1418 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001419 break;
1420 }
1421 case KVM_GET_SREGS: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001422 kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL);
1423 r = -ENOMEM;
1424 if (!kvm_sregs)
1425 goto out;
1426 r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001427 if (r)
1428 goto out;
1429 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001430 if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001431 goto out;
1432 r = 0;
1433 break;
1434 }
1435 case KVM_SET_SREGS: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001436 kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL);
1437 r = -ENOMEM;
1438 if (!kvm_sregs)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001439 goto out;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001440 r = -EFAULT;
1441 if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs)))
1442 goto out;
1443 r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001444 if (r)
1445 goto out;
1446 r = 0;
1447 break;
1448 }
Marcelo Tosatti62d9f0d2008-04-11 13:24:45 -03001449 case KVM_GET_MP_STATE: {
1450 struct kvm_mp_state mp_state;
1451
1452 r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state);
1453 if (r)
1454 goto out;
1455 r = -EFAULT;
1456 if (copy_to_user(argp, &mp_state, sizeof mp_state))
1457 goto out;
1458 r = 0;
1459 break;
1460 }
1461 case KVM_SET_MP_STATE: {
1462 struct kvm_mp_state mp_state;
1463
1464 r = -EFAULT;
1465 if (copy_from_user(&mp_state, argp, sizeof mp_state))
1466 goto out;
1467 r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
1468 if (r)
1469 goto out;
1470 r = 0;
1471 break;
1472 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08001473 case KVM_TRANSLATE: {
1474 struct kvm_translation tr;
1475
1476 r = -EFAULT;
Al Viro2f3669872007-02-09 16:38:35 +00001477 if (copy_from_user(&tr, argp, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001478 goto out;
Zhang Xiantao8b006792007-11-16 13:05:55 +08001479 r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001480 if (r)
1481 goto out;
1482 r = -EFAULT;
Al Viro2f3669872007-02-09 16:38:35 +00001483 if (copy_to_user(argp, &tr, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001484 goto out;
1485 r = 0;
1486 break;
1487 }
Jan Kiszkad0bfb942008-12-15 13:52:10 +01001488 case KVM_SET_GUEST_DEBUG: {
1489 struct kvm_guest_debug dbg;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001490
1491 r = -EFAULT;
Al Viro2f3669872007-02-09 16:38:35 +00001492 if (copy_from_user(&dbg, argp, sizeof dbg))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001493 goto out;
Jan Kiszkad0bfb942008-12-15 13:52:10 +01001494 r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001495 if (r)
1496 goto out;
1497 r = 0;
1498 break;
1499 }
Avi Kivity1961d272007-03-05 19:46:05 +02001500 case KVM_SET_SIGNAL_MASK: {
1501 struct kvm_signal_mask __user *sigmask_arg = argp;
1502 struct kvm_signal_mask kvm_sigmask;
1503 sigset_t sigset, *p;
1504
1505 p = NULL;
1506 if (argp) {
1507 r = -EFAULT;
1508 if (copy_from_user(&kvm_sigmask, argp,
1509 sizeof kvm_sigmask))
1510 goto out;
1511 r = -EINVAL;
1512 if (kvm_sigmask.len != sizeof sigset)
1513 goto out;
1514 r = -EFAULT;
1515 if (copy_from_user(&sigset, sigmask_arg->sigset,
1516 sizeof sigset))
1517 goto out;
1518 p = &sigset;
1519 }
1520 r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
1521 break;
1522 }
Avi Kivityb8836732007-04-01 16:34:31 +03001523 case KVM_GET_FPU: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001524 fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL);
1525 r = -ENOMEM;
1526 if (!fpu)
1527 goto out;
1528 r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001529 if (r)
1530 goto out;
1531 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001532 if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
Avi Kivityb8836732007-04-01 16:34:31 +03001533 goto out;
1534 r = 0;
1535 break;
1536 }
1537 case KVM_SET_FPU: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001538 fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL);
1539 r = -ENOMEM;
1540 if (!fpu)
Avi Kivityb8836732007-04-01 16:34:31 +03001541 goto out;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001542 r = -EFAULT;
1543 if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu)))
1544 goto out;
1545 r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001546 if (r)
1547 goto out;
1548 r = 0;
1549 break;
1550 }
Avi Kivitybccf2152007-02-21 18:04:26 +02001551 default:
Carsten Otte313a3dc2007-10-11 19:16:52 +02001552 r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
Avi Kivitybccf2152007-02-21 18:04:26 +02001553 }
1554out:
Dave Hansenfa3795a2008-08-11 10:01:46 -07001555 kfree(fpu);
1556 kfree(kvm_sregs);
Avi Kivitybccf2152007-02-21 18:04:26 +02001557 return r;
1558}
1559
1560static long kvm_vm_ioctl(struct file *filp,
1561 unsigned int ioctl, unsigned long arg)
1562{
1563 struct kvm *kvm = filp->private_data;
1564 void __user *argp = (void __user *)arg;
Carsten Otte1fe779f2007-10-29 16:08:35 +01001565 int r;
Avi Kivitybccf2152007-02-21 18:04:26 +02001566
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001567 if (kvm->mm != current->mm)
1568 return -EIO;
Avi Kivitybccf2152007-02-21 18:04:26 +02001569 switch (ioctl) {
1570 case KVM_CREATE_VCPU:
1571 r = kvm_vm_ioctl_create_vcpu(kvm, arg);
1572 if (r < 0)
1573 goto out;
1574 break;
Izik Eidus6fc138d2007-10-09 19:20:39 +02001575 case KVM_SET_USER_MEMORY_REGION: {
1576 struct kvm_userspace_memory_region kvm_userspace_mem;
1577
1578 r = -EFAULT;
1579 if (copy_from_user(&kvm_userspace_mem, argp,
1580 sizeof kvm_userspace_mem))
1581 goto out;
1582
1583 r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001584 if (r)
1585 goto out;
1586 break;
1587 }
1588 case KVM_GET_DIRTY_LOG: {
1589 struct kvm_dirty_log log;
1590
1591 r = -EFAULT;
Al Viro2f3669872007-02-09 16:38:35 +00001592 if (copy_from_user(&log, argp, sizeof log))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001593 goto out;
Avi Kivity2c6f5df2007-02-20 18:27:58 +02001594 r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001595 if (r)
1596 goto out;
1597 break;
1598 }
Laurent Vivier5f94c172008-05-30 16:05:54 +02001599#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1600 case KVM_REGISTER_COALESCED_MMIO: {
1601 struct kvm_coalesced_mmio_zone zone;
1602 r = -EFAULT;
1603 if (copy_from_user(&zone, argp, sizeof zone))
1604 goto out;
1605 r = -ENXIO;
1606 r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
1607 if (r)
1608 goto out;
1609 r = 0;
1610 break;
1611 }
1612 case KVM_UNREGISTER_COALESCED_MMIO: {
1613 struct kvm_coalesced_mmio_zone zone;
1614 r = -EFAULT;
1615 if (copy_from_user(&zone, argp, sizeof zone))
1616 goto out;
1617 r = -ENXIO;
1618 r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
1619 if (r)
1620 goto out;
1621 r = 0;
1622 break;
1623 }
1624#endif
Gregory Haskins721eecb2009-05-20 10:30:49 -04001625 case KVM_IRQFD: {
1626 struct kvm_irqfd data;
1627
1628 r = -EFAULT;
1629 if (copy_from_user(&data, argp, sizeof data))
1630 goto out;
1631 r = kvm_irqfd(kvm, data.fd, data.gsi, data.flags);
1632 break;
1633 }
Gregory Haskinsd34e6b12009-07-07 17:08:49 -04001634 case KVM_IOEVENTFD: {
1635 struct kvm_ioeventfd data;
1636
1637 r = -EFAULT;
1638 if (copy_from_user(&data, argp, sizeof data))
1639 goto out;
1640 r = kvm_ioeventfd(kvm, &data);
1641 break;
1642 }
Gleb Natapov73880c82009-06-09 15:56:28 +03001643#ifdef CONFIG_KVM_APIC_ARCHITECTURE
1644 case KVM_SET_BOOT_CPU_ID:
1645 r = 0;
Marcelo Tosatti894a9c52009-06-23 15:05:14 -03001646 mutex_lock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03001647 if (atomic_read(&kvm->online_vcpus) != 0)
1648 r = -EBUSY;
1649 else
1650 kvm->bsp_vcpu_id = arg;
Marcelo Tosatti894a9c52009-06-23 15:05:14 -03001651 mutex_unlock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03001652 break;
1653#endif
Avi Kivityf17abe92007-02-21 19:28:04 +02001654 default:
Carsten Otte1fe779f2007-10-29 16:08:35 +01001655 r = kvm_arch_vm_ioctl(filp, ioctl, arg);
Avi Kivitybfd99ff2009-08-26 14:57:50 +03001656 if (r == -ENOTTY)
1657 r = kvm_vm_ioctl_assigned_device(kvm, ioctl, arg);
Avi Kivityf17abe92007-02-21 19:28:04 +02001658 }
1659out:
1660 return r;
1661}
1662
Arnd Bergmann6ff58942009-10-22 14:19:27 +02001663#ifdef CONFIG_COMPAT
1664struct compat_kvm_dirty_log {
1665 __u32 slot;
1666 __u32 padding1;
1667 union {
1668 compat_uptr_t dirty_bitmap; /* one bit per page */
1669 __u64 padding2;
1670 };
1671};
1672
1673static long kvm_vm_compat_ioctl(struct file *filp,
1674 unsigned int ioctl, unsigned long arg)
1675{
1676 struct kvm *kvm = filp->private_data;
1677 int r;
1678
1679 if (kvm->mm != current->mm)
1680 return -EIO;
1681 switch (ioctl) {
1682 case KVM_GET_DIRTY_LOG: {
1683 struct compat_kvm_dirty_log compat_log;
1684 struct kvm_dirty_log log;
1685
1686 r = -EFAULT;
1687 if (copy_from_user(&compat_log, (void __user *)arg,
1688 sizeof(compat_log)))
1689 goto out;
1690 log.slot = compat_log.slot;
1691 log.padding1 = compat_log.padding1;
1692 log.padding2 = compat_log.padding2;
1693 log.dirty_bitmap = compat_ptr(compat_log.dirty_bitmap);
1694
1695 r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
1696 if (r)
1697 goto out;
1698 break;
1699 }
1700 default:
1701 r = kvm_vm_ioctl(filp, ioctl, arg);
1702 }
1703
1704out:
1705 return r;
1706}
1707#endif
1708
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001709static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivityf17abe92007-02-21 19:28:04 +02001710{
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03001711 struct page *page[1];
1712 unsigned long addr;
1713 int npages;
1714 gfn_t gfn = vmf->pgoff;
Avi Kivityf17abe92007-02-21 19:28:04 +02001715 struct kvm *kvm = vma->vm_file->private_data;
Avi Kivityf17abe92007-02-21 19:28:04 +02001716
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03001717 addr = gfn_to_hva(kvm, gfn);
1718 if (kvm_is_error_hva(addr))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001719 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03001720
1721 npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page,
1722 NULL);
1723 if (unlikely(npages != 1))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001724 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03001725
1726 vmf->page = page[0];
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001727 return 0;
Avi Kivityf17abe92007-02-21 19:28:04 +02001728}
1729
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +04001730static const struct vm_operations_struct kvm_vm_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001731 .fault = kvm_vm_fault,
Avi Kivityf17abe92007-02-21 19:28:04 +02001732};
1733
1734static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma)
1735{
1736 vma->vm_ops = &kvm_vm_vm_ops;
1737 return 0;
1738}
1739
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01001740static struct file_operations kvm_vm_fops = {
Avi Kivityf17abe92007-02-21 19:28:04 +02001741 .release = kvm_vm_release,
1742 .unlocked_ioctl = kvm_vm_ioctl,
Arnd Bergmann6ff58942009-10-22 14:19:27 +02001743#ifdef CONFIG_COMPAT
1744 .compat_ioctl = kvm_vm_compat_ioctl,
1745#endif
Avi Kivityf17abe92007-02-21 19:28:04 +02001746 .mmap = kvm_vm_mmap,
1747};
1748
1749static int kvm_dev_ioctl_create_vm(void)
1750{
Al Viro2030a422008-02-23 06:46:49 -05001751 int fd;
Avi Kivityf17abe92007-02-21 19:28:04 +02001752 struct kvm *kvm;
1753
Avi Kivityf17abe92007-02-21 19:28:04 +02001754 kvm = kvm_create_vm();
Avi Kivityd6d28162007-06-28 08:38:16 -04001755 if (IS_ERR(kvm))
1756 return PTR_ERR(kvm);
Roland Dreier628ff7c2009-12-18 09:41:24 -08001757 fd = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, O_RDWR);
Al Viro2030a422008-02-23 06:46:49 -05001758 if (fd < 0)
Al Viro66c0b392008-04-19 20:33:56 +01001759 kvm_put_kvm(kvm);
Avi Kivityf17abe92007-02-21 19:28:04 +02001760
Avi Kivityf17abe92007-02-21 19:28:04 +02001761 return fd;
Avi Kivityf17abe92007-02-21 19:28:04 +02001762}
1763
Avi Kivity1a811b62008-12-08 18:25:27 +02001764static long kvm_dev_ioctl_check_extension_generic(long arg)
1765{
1766 switch (arg) {
Avi Kivityca9edae2008-12-08 18:29:29 +02001767 case KVM_CAP_USER_MEMORY:
Avi Kivity1a811b62008-12-08 18:25:27 +02001768 case KVM_CAP_DESTROY_MEMORY_REGION_WORKS:
Jan Kiszka4cd481f2009-04-13 11:59:32 +02001769 case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS:
Gleb Natapov73880c82009-06-09 15:56:28 +03001770#ifdef CONFIG_KVM_APIC_ARCHITECTURE
1771 case KVM_CAP_SET_BOOT_CPU_ID:
1772#endif
Avi Kivitya9c73992009-11-04 11:54:59 +02001773 case KVM_CAP_INTERNAL_ERROR_DATA:
Avi Kivity1a811b62008-12-08 18:25:27 +02001774 return 1;
Avi Kivity399ec802008-11-19 13:58:46 +02001775#ifdef CONFIG_HAVE_KVM_IRQCHIP
1776 case KVM_CAP_IRQ_ROUTING:
Sheng Yang36463142009-03-16 16:33:43 +08001777 return KVM_MAX_IRQ_ROUTES;
Avi Kivity399ec802008-11-19 13:58:46 +02001778#endif
Avi Kivity1a811b62008-12-08 18:25:27 +02001779 default:
1780 break;
1781 }
1782 return kvm_dev_ioctl_check_extension(arg);
1783}
1784
Avi Kivityf17abe92007-02-21 19:28:04 +02001785static long kvm_dev_ioctl(struct file *filp,
1786 unsigned int ioctl, unsigned long arg)
1787{
Avi Kivity07c45a32007-03-07 13:05:38 +02001788 long r = -EINVAL;
Avi Kivityf17abe92007-02-21 19:28:04 +02001789
1790 switch (ioctl) {
1791 case KVM_GET_API_VERSION:
Avi Kivityf0fe5102007-03-07 13:11:17 +02001792 r = -EINVAL;
1793 if (arg)
1794 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02001795 r = KVM_API_VERSION;
1796 break;
1797 case KVM_CREATE_VM:
Avi Kivityf0fe5102007-03-07 13:11:17 +02001798 r = -EINVAL;
1799 if (arg)
1800 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02001801 r = kvm_dev_ioctl_create_vm();
1802 break;
Zhang Xiantao018d00d2007-11-15 23:07:47 +08001803 case KVM_CHECK_EXTENSION:
Avi Kivity1a811b62008-12-08 18:25:27 +02001804 r = kvm_dev_ioctl_check_extension_generic(arg);
Avi Kivity5d308f42007-03-01 17:56:20 +02001805 break;
Avi Kivity07c45a32007-03-07 13:05:38 +02001806 case KVM_GET_VCPU_MMAP_SIZE:
1807 r = -EINVAL;
1808 if (arg)
1809 goto out;
Avi Kivityadb1ff42008-01-24 15:13:08 +02001810 r = PAGE_SIZE; /* struct kvm_run */
1811#ifdef CONFIG_X86
1812 r += PAGE_SIZE; /* pio data page */
1813#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02001814#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1815 r += PAGE_SIZE; /* coalesced mmio ring page */
1816#endif
Avi Kivity07c45a32007-03-07 13:05:38 +02001817 break;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04001818 case KVM_TRACE_ENABLE:
1819 case KVM_TRACE_PAUSE:
1820 case KVM_TRACE_DISABLE:
Marcelo Tosatti2023a292009-06-18 11:47:28 -03001821 r = -EOPNOTSUPP;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04001822 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001823 default:
Carsten Otte043405e2007-10-10 17:16:19 +02001824 return kvm_arch_dev_ioctl(filp, ioctl, arg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001825 }
1826out:
1827 return r;
1828}
1829
Avi Kivity6aa8b732006-12-10 02:21:36 -08001830static struct file_operations kvm_chardev_ops = {
Avi Kivity6aa8b732006-12-10 02:21:36 -08001831 .unlocked_ioctl = kvm_dev_ioctl,
1832 .compat_ioctl = kvm_dev_ioctl,
Avi Kivity6aa8b732006-12-10 02:21:36 -08001833};
1834
1835static struct miscdevice kvm_dev = {
Avi Kivitybbe44322007-03-04 13:27:36 +02001836 KVM_MINOR,
Avi Kivity6aa8b732006-12-10 02:21:36 -08001837 "kvm",
1838 &kvm_chardev_ops,
1839};
1840
Avi Kivity1b6c0162007-05-24 13:03:52 +03001841static void hardware_enable(void *junk)
1842{
1843 int cpu = raw_smp_processor_id();
Alexander Graf10474ae2009-09-15 11:37:46 +02001844 int r;
Avi Kivity1b6c0162007-05-24 13:03:52 +03001845
Rusty Russell7f59f492008-12-07 21:25:45 +10301846 if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03001847 return;
Alexander Graf10474ae2009-09-15 11:37:46 +02001848
Rusty Russell7f59f492008-12-07 21:25:45 +10301849 cpumask_set_cpu(cpu, cpus_hardware_enabled);
Alexander Graf10474ae2009-09-15 11:37:46 +02001850
1851 r = kvm_arch_hardware_enable(NULL);
1852
1853 if (r) {
1854 cpumask_clear_cpu(cpu, cpus_hardware_enabled);
1855 atomic_inc(&hardware_enable_failed);
1856 printk(KERN_INFO "kvm: enabling virtualization on "
1857 "CPU%d failed\n", cpu);
1858 }
Avi Kivity1b6c0162007-05-24 13:03:52 +03001859}
1860
1861static void hardware_disable(void *junk)
1862{
1863 int cpu = raw_smp_processor_id();
1864
Rusty Russell7f59f492008-12-07 21:25:45 +10301865 if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03001866 return;
Rusty Russell7f59f492008-12-07 21:25:45 +10301867 cpumask_clear_cpu(cpu, cpus_hardware_enabled);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001868 kvm_arch_hardware_disable(NULL);
Avi Kivity1b6c0162007-05-24 13:03:52 +03001869}
1870
Alexander Graf10474ae2009-09-15 11:37:46 +02001871static void hardware_disable_all_nolock(void)
1872{
1873 BUG_ON(!kvm_usage_count);
1874
1875 kvm_usage_count--;
1876 if (!kvm_usage_count)
1877 on_each_cpu(hardware_disable, NULL, 1);
1878}
1879
1880static void hardware_disable_all(void)
1881{
1882 spin_lock(&kvm_lock);
1883 hardware_disable_all_nolock();
1884 spin_unlock(&kvm_lock);
1885}
1886
1887static int hardware_enable_all(void)
1888{
1889 int r = 0;
1890
1891 spin_lock(&kvm_lock);
1892
1893 kvm_usage_count++;
1894 if (kvm_usage_count == 1) {
1895 atomic_set(&hardware_enable_failed, 0);
1896 on_each_cpu(hardware_enable, NULL, 1);
1897
1898 if (atomic_read(&hardware_enable_failed)) {
1899 hardware_disable_all_nolock();
1900 r = -EBUSY;
1901 }
1902 }
1903
1904 spin_unlock(&kvm_lock);
1905
1906 return r;
1907}
1908
Avi Kivity774c47f2007-02-12 00:54:47 -08001909static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val,
1910 void *v)
1911{
1912 int cpu = (long)v;
1913
Alexander Graf10474ae2009-09-15 11:37:46 +02001914 if (!kvm_usage_count)
1915 return NOTIFY_OK;
1916
Avi Kivity1a6f4d72007-11-11 18:37:32 +02001917 val &= ~CPU_TASKS_FROZEN;
Avi Kivity774c47f2007-02-12 00:54:47 -08001918 switch (val) {
Avi Kivitycec9ad22007-05-24 13:11:41 +03001919 case CPU_DYING:
Avi Kivity6ec8a8562007-08-19 15:57:26 +03001920 printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n",
1921 cpu);
1922 hardware_disable(NULL);
1923 break;
Avi Kivity774c47f2007-02-12 00:54:47 -08001924 case CPU_UP_CANCELED:
Jeremy Katz43934a32007-02-19 14:37:46 +02001925 printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n",
1926 cpu);
Jens Axboe8691e5a2008-06-06 11:18:06 +02001927 smp_call_function_single(cpu, hardware_disable, NULL, 1);
Avi Kivity774c47f2007-02-12 00:54:47 -08001928 break;
Jeremy Katz43934a32007-02-19 14:37:46 +02001929 case CPU_ONLINE:
1930 printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n",
1931 cpu);
Jens Axboe8691e5a2008-06-06 11:18:06 +02001932 smp_call_function_single(cpu, hardware_enable, NULL, 1);
Avi Kivity774c47f2007-02-12 00:54:47 -08001933 break;
1934 }
1935 return NOTIFY_OK;
1936}
1937
Avi Kivity4ecac3f2008-05-13 13:23:38 +03001938
1939asmlinkage void kvm_handle_fault_on_reboot(void)
1940{
1941 if (kvm_rebooting)
1942 /* spin while reset goes on */
1943 while (true)
1944 ;
1945 /* Fault while not rebooting. We want the trace. */
1946 BUG();
1947}
1948EXPORT_SYMBOL_GPL(kvm_handle_fault_on_reboot);
1949
Rusty Russell9a2b85c2007-07-17 23:17:55 +10001950static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
Mike Dayd77c26f2007-10-08 09:02:08 -04001951 void *v)
Rusty Russell9a2b85c2007-07-17 23:17:55 +10001952{
Sheng Yang8e1c1812009-04-29 11:09:04 +08001953 /*
1954 * Some (well, at least mine) BIOSes hang on reboot if
1955 * in vmx root mode.
1956 *
1957 * And Intel TXT required VMX off for all cpu when system shutdown.
1958 */
1959 printk(KERN_INFO "kvm: exiting hardware virtualization\n");
1960 kvm_rebooting = true;
1961 on_each_cpu(hardware_disable, NULL, 1);
Rusty Russell9a2b85c2007-07-17 23:17:55 +10001962 return NOTIFY_OK;
1963}
1964
1965static struct notifier_block kvm_reboot_notifier = {
1966 .notifier_call = kvm_reboot,
1967 .priority = 0,
1968};
1969
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02001970static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04001971{
1972 int i;
1973
1974 for (i = 0; i < bus->dev_count; i++) {
1975 struct kvm_io_device *pos = bus->devs[i];
1976
1977 kvm_iodevice_destructor(pos);
1978 }
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02001979 kfree(bus);
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04001980}
1981
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03001982/* kvm_io_bus_write - called under kvm->slots_lock */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02001983int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03001984 int len, const void *val)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04001985{
1986 int i;
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02001987 struct kvm_io_bus *bus = rcu_dereference(kvm->buses[bus_idx]);
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03001988 for (i = 0; i < bus->dev_count; i++)
1989 if (!kvm_iodevice_write(bus->devs[i], addr, len, val))
1990 return 0;
1991 return -EOPNOTSUPP;
1992}
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04001993
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03001994/* kvm_io_bus_read - called under kvm->slots_lock */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02001995int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
1996 int len, void *val)
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03001997{
1998 int i;
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02001999 struct kvm_io_bus *bus = rcu_dereference(kvm->buses[bus_idx]);
2000
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002001 for (i = 0; i < bus->dev_count; i++)
2002 if (!kvm_iodevice_read(bus->devs[i], addr, len, val))
2003 return 0;
2004 return -EOPNOTSUPP;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002005}
2006
Marcelo Tosatti79fac952009-12-23 14:35:26 -02002007/* Caller must hold slots_lock. */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002008int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx,
2009 struct kvm_io_device *dev)
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002010{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002011 struct kvm_io_bus *new_bus, *bus;
Gregory Haskins090b7af2009-07-07 17:08:44 -04002012
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002013 bus = kvm->buses[bus_idx];
Gregory Haskins090b7af2009-07-07 17:08:44 -04002014 if (bus->dev_count > NR_IOBUS_DEVS-1)
2015 return -ENOSPC;
2016
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002017 new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL);
2018 if (!new_bus)
2019 return -ENOMEM;
2020 memcpy(new_bus, bus, sizeof(struct kvm_io_bus));
2021 new_bus->devs[new_bus->dev_count++] = dev;
2022 rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
2023 synchronize_srcu_expedited(&kvm->srcu);
2024 kfree(bus);
Gregory Haskins090b7af2009-07-07 17:08:44 -04002025
2026 return 0;
2027}
2028
Marcelo Tosatti79fac952009-12-23 14:35:26 -02002029/* Caller must hold slots_lock. */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002030int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
2031 struct kvm_io_device *dev)
Gregory Haskins090b7af2009-07-07 17:08:44 -04002032{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002033 int i, r;
2034 struct kvm_io_bus *new_bus, *bus;
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002035
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002036 new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL);
2037 if (!new_bus)
2038 return -ENOMEM;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002039
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002040 bus = kvm->buses[bus_idx];
2041 memcpy(new_bus, bus, sizeof(struct kvm_io_bus));
2042
2043 r = -ENOENT;
2044 for (i = 0; i < new_bus->dev_count; i++)
2045 if (new_bus->devs[i] == dev) {
2046 r = 0;
2047 new_bus->devs[i] = new_bus->devs[--new_bus->dev_count];
Gregory Haskins090b7af2009-07-07 17:08:44 -04002048 break;
2049 }
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002050
2051 if (r) {
2052 kfree(new_bus);
2053 return r;
2054 }
2055
2056 rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
2057 synchronize_srcu_expedited(&kvm->srcu);
2058 kfree(bus);
2059 return r;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002060}
2061
Avi Kivity774c47f2007-02-12 00:54:47 -08002062static struct notifier_block kvm_cpu_notifier = {
2063 .notifier_call = kvm_cpu_hotplug,
2064 .priority = 20, /* must be > scheduler priority */
2065};
2066
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002067static int vm_stat_get(void *_offset, u64 *val)
Avi Kivityba1389b2007-11-18 16:24:12 +02002068{
2069 unsigned offset = (long)_offset;
Avi Kivityba1389b2007-11-18 16:24:12 +02002070 struct kvm *kvm;
2071
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002072 *val = 0;
Avi Kivityba1389b2007-11-18 16:24:12 +02002073 spin_lock(&kvm_lock);
2074 list_for_each_entry(kvm, &vm_list, vm_list)
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002075 *val += *(u32 *)((void *)kvm + offset);
Avi Kivityba1389b2007-11-18 16:24:12 +02002076 spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002077 return 0;
Avi Kivityba1389b2007-11-18 16:24:12 +02002078}
2079
2080DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n");
2081
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002082static int vcpu_stat_get(void *_offset, u64 *val)
Avi Kivity1165f5f2007-04-19 17:27:43 +03002083{
2084 unsigned offset = (long)_offset;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002085 struct kvm *kvm;
2086 struct kvm_vcpu *vcpu;
2087 int i;
2088
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002089 *val = 0;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002090 spin_lock(&kvm_lock);
2091 list_for_each_entry(kvm, &vm_list, vm_list)
Gleb Natapov988a2ca2009-06-09 15:56:29 +03002092 kvm_for_each_vcpu(i, vcpu, kvm)
2093 *val += *(u32 *)((void *)vcpu + offset);
2094
Avi Kivity1165f5f2007-04-19 17:27:43 +03002095 spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002096 return 0;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002097}
2098
Avi Kivityba1389b2007-11-18 16:24:12 +02002099DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n");
2100
Alexey Dobriyan828c0952009-10-01 15:43:56 -07002101static const struct file_operations *stat_fops[] = {
Avi Kivityba1389b2007-11-18 16:24:12 +02002102 [KVM_STAT_VCPU] = &vcpu_stat_fops,
2103 [KVM_STAT_VM] = &vm_stat_fops,
2104};
Avi Kivity1165f5f2007-04-19 17:27:43 +03002105
Zhang Xiantaoa16b0432007-11-16 14:38:21 +08002106static void kvm_init_debug(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002107{
2108 struct kvm_stats_debugfs_item *p;
2109
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002110 kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002111 for (p = debugfs_entries; p->name; ++p)
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002112 p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir,
Avi Kivity1165f5f2007-04-19 17:27:43 +03002113 (void *)(long)p->offset,
Avi Kivityba1389b2007-11-18 16:24:12 +02002114 stat_fops[p->kind]);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002115}
2116
2117static void kvm_exit_debug(void)
2118{
2119 struct kvm_stats_debugfs_item *p;
2120
2121 for (p = debugfs_entries; p->name; ++p)
2122 debugfs_remove(p->dentry);
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002123 debugfs_remove(kvm_debugfs_dir);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002124}
2125
Avi Kivity59ae6c62007-02-12 00:54:48 -08002126static int kvm_suspend(struct sys_device *dev, pm_message_t state)
2127{
Alexander Graf10474ae2009-09-15 11:37:46 +02002128 if (kvm_usage_count)
2129 hardware_disable(NULL);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002130 return 0;
2131}
2132
2133static int kvm_resume(struct sys_device *dev)
2134{
Alexander Graf10474ae2009-09-15 11:37:46 +02002135 if (kvm_usage_count)
2136 hardware_enable(NULL);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002137 return 0;
2138}
2139
2140static struct sysdev_class kvm_sysdev_class = {
Kay Sieversaf5ca3f2007-12-20 02:09:39 +01002141 .name = "kvm",
Avi Kivity59ae6c62007-02-12 00:54:48 -08002142 .suspend = kvm_suspend,
2143 .resume = kvm_resume,
2144};
2145
2146static struct sys_device kvm_sysdev = {
2147 .id = 0,
2148 .cls = &kvm_sysdev_class,
2149};
2150
Izik Eiduscea7bb22007-10-17 19:17:48 +02002151struct page *bad_page;
Anthony Liguori35149e22008-04-02 14:46:56 -05002152pfn_t bad_pfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002153
Avi Kivity15ad7142007-07-11 18:17:21 +03002154static inline
2155struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn)
2156{
2157 return container_of(pn, struct kvm_vcpu, preempt_notifier);
2158}
2159
2160static void kvm_sched_in(struct preempt_notifier *pn, int cpu)
2161{
2162 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2163
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002164 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002165}
2166
2167static void kvm_sched_out(struct preempt_notifier *pn,
2168 struct task_struct *next)
2169{
2170 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2171
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002172 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002173}
2174
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002175int kvm_init(void *opaque, unsigned int vcpu_size,
Rusty Russellc16f8622007-07-30 21:12:19 +10002176 struct module *module)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002177{
2178 int r;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002179 int cpu;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002180
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002181 r = kvm_arch_init(opaque);
2182 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002183 goto out_fail;
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002184
2185 bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2186
2187 if (bad_page == NULL) {
2188 r = -ENOMEM;
2189 goto out;
2190 }
2191
Anthony Liguori35149e22008-04-02 14:46:56 -05002192 bad_pfn = page_to_pfn(bad_page);
2193
Avi Kivity8437a612009-06-06 14:52:35 -07002194 if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
Rusty Russell7f59f492008-12-07 21:25:45 +10302195 r = -ENOMEM;
2196 goto out_free_0;
2197 }
2198
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002199 r = kvm_arch_hardware_setup();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002200 if (r < 0)
Rusty Russell7f59f492008-12-07 21:25:45 +10302201 goto out_free_0a;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002202
Yang, Sheng002c7f72007-07-31 14:23:01 +03002203 for_each_online_cpu(cpu) {
2204 smp_call_function_single(cpu,
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002205 kvm_arch_check_processor_compat,
Jens Axboe8691e5a2008-06-06 11:18:06 +02002206 &r, 1);
Yang, Sheng002c7f72007-07-31 14:23:01 +03002207 if (r < 0)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002208 goto out_free_1;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002209 }
2210
Avi Kivity774c47f2007-02-12 00:54:47 -08002211 r = register_cpu_notifier(&kvm_cpu_notifier);
2212 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002213 goto out_free_2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002214 register_reboot_notifier(&kvm_reboot_notifier);
2215
Avi Kivity59ae6c62007-02-12 00:54:48 -08002216 r = sysdev_class_register(&kvm_sysdev_class);
2217 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002218 goto out_free_3;
Avi Kivity59ae6c62007-02-12 00:54:48 -08002219
2220 r = sysdev_register(&kvm_sysdev);
2221 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002222 goto out_free_4;
Avi Kivity59ae6c62007-02-12 00:54:48 -08002223
Rusty Russellc16f8622007-07-30 21:12:19 +10002224 /* A kmem cache lets us meet the alignment requirements of fx_save. */
2225 kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size,
Joe Perches56919c52007-11-12 20:06:51 -08002226 __alignof__(struct kvm_vcpu),
2227 0, NULL);
Rusty Russellc16f8622007-07-30 21:12:19 +10002228 if (!kvm_vcpu_cache) {
2229 r = -ENOMEM;
Zhang Xiantaod23087842007-11-29 15:35:39 +08002230 goto out_free_5;
Rusty Russellc16f8622007-07-30 21:12:19 +10002231 }
2232
Avi Kivity6aa8b732006-12-10 02:21:36 -08002233 kvm_chardev_ops.owner = module;
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01002234 kvm_vm_fops.owner = module;
2235 kvm_vcpu_fops.owner = module;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002236
2237 r = misc_register(&kvm_dev);
2238 if (r) {
Mike Dayd77c26f2007-10-08 09:02:08 -04002239 printk(KERN_ERR "kvm: misc device register failed\n");
Avi Kivity6aa8b732006-12-10 02:21:36 -08002240 goto out_free;
2241 }
2242
Avi Kivity15ad7142007-07-11 18:17:21 +03002243 kvm_preempt_ops.sched_in = kvm_sched_in;
2244 kvm_preempt_ops.sched_out = kvm_sched_out;
2245
Darrick J. Wong0ea4ed82009-10-14 16:21:00 -07002246 kvm_init_debug();
2247
Avi Kivityc7addb92007-09-16 18:58:32 +02002248 return 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002249
2250out_free:
Rusty Russellc16f8622007-07-30 21:12:19 +10002251 kmem_cache_destroy(kvm_vcpu_cache);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002252out_free_5:
Avi Kivity59ae6c62007-02-12 00:54:48 -08002253 sysdev_unregister(&kvm_sysdev);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002254out_free_4:
Avi Kivity59ae6c62007-02-12 00:54:48 -08002255 sysdev_class_unregister(&kvm_sysdev_class);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002256out_free_3:
Avi Kivity6aa8b732006-12-10 02:21:36 -08002257 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity774c47f2007-02-12 00:54:47 -08002258 unregister_cpu_notifier(&kvm_cpu_notifier);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002259out_free_2:
Zhang Xiantaod23087842007-11-29 15:35:39 +08002260out_free_1:
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002261 kvm_arch_hardware_unsetup();
Rusty Russell7f59f492008-12-07 21:25:45 +10302262out_free_0a:
2263 free_cpumask_var(cpus_hardware_enabled);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002264out_free_0:
2265 __free_page(bad_page);
Avi Kivityca45aaa2007-03-01 19:21:03 +02002266out:
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002267 kvm_arch_exit();
Zhang Xiantaod23087842007-11-29 15:35:39 +08002268out_fail:
Avi Kivity6aa8b732006-12-10 02:21:36 -08002269 return r;
2270}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002271EXPORT_SYMBOL_GPL(kvm_init);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002272
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002273void kvm_exit(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002274{
Marcelo Tosatti229456f2009-06-17 09:22:14 -03002275 tracepoint_synchronize_unregister();
Darrick J. Wong0ea4ed82009-10-14 16:21:00 -07002276 kvm_exit_debug();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002277 misc_deregister(&kvm_dev);
Rusty Russellc16f8622007-07-30 21:12:19 +10002278 kmem_cache_destroy(kvm_vcpu_cache);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002279 sysdev_unregister(&kvm_sysdev);
2280 sysdev_class_unregister(&kvm_sysdev_class);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002281 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002282 unregister_cpu_notifier(&kvm_cpu_notifier);
Jens Axboe15c8b6c2008-05-09 09:39:44 +02002283 on_each_cpu(hardware_disable, NULL, 1);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002284 kvm_arch_hardware_unsetup();
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002285 kvm_arch_exit();
Rusty Russell7f59f492008-12-07 21:25:45 +10302286 free_cpumask_var(cpus_hardware_enabled);
Izik Eiduscea7bb22007-10-17 19:17:48 +02002287 __free_page(bad_page);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002288}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002289EXPORT_SYMBOL_GPL(kvm_exit);