blob: a91f980077d843ca319dfe5d7b3b95e252e02bc3 [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.
Nicolas Kaiser9611c182010-10-06 14:23:22 +02008 * Copyright 2010 Red Hat, Inc. and/or its affiliates.
Avi Kivity6aa8b732006-12-10 02:21:36 -08009 *
10 * Authors:
11 * Avi Kivity <avi@qumranet.com>
12 * Yaniv Kamay <yaniv@qumranet.com>
13 *
14 * This work is licensed under the terms of the GNU GPL, version 2. See
15 * the COPYING file in the top-level directory.
16 *
17 */
18
Hollis Blancharde2174022007-12-03 15:30:24 -060019#include "iodev.h"
Avi Kivity6aa8b732006-12-10 02:21:36 -080020
Avi Kivityedf88412007-12-16 11:02:48 +020021#include <linux/kvm_host.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080022#include <linux/kvm.h>
23#include <linux/module.h>
24#include <linux/errno.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080025#include <linux/percpu.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080026#include <linux/mm.h>
27#include <linux/miscdevice.h>
28#include <linux/vmalloc.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080029#include <linux/reboot.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080030#include <linux/debugfs.h>
31#include <linux/highmem.h>
32#include <linux/file.h>
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +010033#include <linux/syscore_ops.h>
Avi Kivity774c47f2007-02-12 00:54:47 -080034#include <linux/cpu.h>
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +040035#include <linux/sched.h>
Avi Kivityd9e368d2007-06-07 19:18:30 +030036#include <linux/cpumask.h>
37#include <linux/smp.h>
Avi Kivityd6d28162007-06-28 08:38:16 -040038#include <linux/anon_inodes.h>
Avi Kivity04d2cc72007-09-10 18:10:54 +030039#include <linux/profile.h>
Anthony Liguori7aa81cc2007-09-17 14:57:50 -050040#include <linux/kvm_para.h>
Izik Eidus6fc138d2007-10-09 19:20:39 +020041#include <linux/pagemap.h>
Anthony Liguori8d4e1282007-10-18 09:59:34 -050042#include <linux/mman.h>
Anthony Liguori35149e22008-04-02 14:46:56 -050043#include <linux/swap.h>
Sheng Yange56d5322009-03-12 21:45:39 +080044#include <linux/bitops.h>
Marcelo Tosatti547de292009-05-07 17:55:13 -030045#include <linux/spinlock.h>
Arnd Bergmann6ff58942009-10-22 14:19:27 +020046#include <linux/compat.h>
Marcelo Tosattibc6678a2009-12-23 14:35:21 -020047#include <linux/srcu.h>
Joerg Roedel8f0b1ab2010-01-28 12:37:56 +010048#include <linux/hugetlb.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090049#include <linux/slab.h>
Sasha Levin743eeb02011-07-27 16:00:48 +030050#include <linux/sort.h>
51#include <linux/bsearch.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080052
Avi Kivitye4956062007-06-28 14:15:57 -040053#include <asm/processor.h>
Avi Kivitye4956062007-06-28 14:15:57 -040054#include <asm/io.h>
55#include <asm/uaccess.h>
Izik Eidus3e021bf2007-11-19 11:16:57 +020056#include <asm/pgtable.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080057
Laurent Vivier5f94c172008-05-30 16:05:54 +020058#include "coalesced_mmio.h"
Gleb Natapovaf585b92010-10-14 11:22:46 +020059#include "async_pf.h"
Laurent Vivier5f94c172008-05-30 16:05:54 +020060
Marcelo Tosatti229456f2009-06-17 09:22:14 -030061#define CREATE_TRACE_POINTS
62#include <trace/events/kvm.h>
63
Avi Kivity6aa8b732006-12-10 02:21:36 -080064MODULE_AUTHOR("Qumranet");
65MODULE_LICENSE("GPL");
66
Marcelo Tosattifa40a822009-06-04 15:08:24 -030067/*
68 * Ordering of locks:
69 *
Sheng Yangfae3a352009-12-15 10:28:07 +080070 * kvm->lock --> kvm->slots_lock --> kvm->irq_lock
Marcelo Tosattifa40a822009-06-04 15:08:24 -030071 */
72
Jan Kiszkae935b832011-02-08 12:55:33 +010073DEFINE_RAW_SPINLOCK(kvm_lock);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +080074LIST_HEAD(vm_list);
Avi Kivity133de902007-02-12 00:54:44 -080075
Rusty Russell7f59f492008-12-07 21:25:45 +103076static cpumask_var_t cpus_hardware_enabled;
Alexander Graf10474ae2009-09-15 11:37:46 +020077static int kvm_usage_count = 0;
78static atomic_t hardware_enable_failed;
Avi Kivity1b6c0162007-05-24 13:03:52 +030079
Rusty Russellc16f8622007-07-30 21:12:19 +100080struct kmem_cache *kvm_vcpu_cache;
81EXPORT_SYMBOL_GPL(kvm_vcpu_cache);
Avi Kivity1165f5f2007-04-19 17:27:43 +030082
Avi Kivity15ad7142007-07-11 18:17:21 +030083static __read_mostly struct preempt_ops kvm_preempt_ops;
84
Hollis Blanchard76f7c872008-04-15 16:05:42 -050085struct dentry *kvm_debugfs_dir;
Avi Kivity6aa8b732006-12-10 02:21:36 -080086
Avi Kivitybccf2152007-02-21 18:04:26 +020087static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl,
88 unsigned long arg);
Alexander Graf1dda6062011-06-08 02:45:37 +020089#ifdef CONFIG_COMPAT
90static long kvm_vcpu_compat_ioctl(struct file *file, unsigned int ioctl,
91 unsigned long arg);
92#endif
Alexander Graf10474ae2009-09-15 11:37:46 +020093static int hardware_enable_all(void);
94static void hardware_disable_all(void);
Avi Kivitybccf2152007-02-21 18:04:26 +020095
Marcelo Tosattie93f8a02009-12-23 14:35:24 -020096static void kvm_io_bus_destroy(struct kvm_io_bus *bus);
97
Avi Kivityb7c41452010-12-02 17:52:50 +020098bool kvm_rebooting;
99EXPORT_SYMBOL_GPL(kvm_rebooting);
Avi Kivity4ecac3f2008-05-13 13:23:38 +0300100
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300101static bool largepages_enabled = true;
102
Gleb Natapovfa7bff82010-07-07 20:16:44 +0300103static struct page *hwpoison_page;
104static pfn_t hwpoison_pfn;
Huang Yingbf998152010-05-31 14:28:19 +0800105
Xiao Guangrongfce92dc2011-07-12 03:28:54 +0800106struct page *fault_page;
107pfn_t fault_pfn;
Gleb Natapovedba23e2010-07-07 20:16:45 +0300108
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +0800109inline int kvm_is_mmio_pfn(pfn_t pfn)
Ben-Ami Yassourcbff90a2008-07-28 19:26:24 +0300110{
Joerg Roedelfc5659c2009-02-18 14:08:58 +0100111 if (pfn_valid(pfn)) {
Andrea Arcangeli22e5c472011-01-13 15:47:20 -0800112 int reserved;
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800113 struct page *tail = pfn_to_page(pfn);
Andrea Arcangeli22e5c472011-01-13 15:47:20 -0800114 struct page *head = compound_trans_head(tail);
115 reserved = PageReserved(head);
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800116 if (head != tail) {
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800117 /*
Andrea Arcangeli22e5c472011-01-13 15:47:20 -0800118 * "head" is not a dangling pointer
119 * (compound_trans_head takes care of that)
120 * but the hugepage may have been splitted
121 * from under us (and we may not hold a
122 * reference count on the head page so it can
123 * be reused before we run PageReferenced), so
124 * we've to check PageTail before returning
125 * what we just read.
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800126 */
Andrea Arcangeli22e5c472011-01-13 15:47:20 -0800127 smp_rmb();
128 if (PageTail(tail))
129 return reserved;
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800130 }
131 return PageReserved(tail);
Joerg Roedelfc5659c2009-02-18 14:08:58 +0100132 }
Ben-Ami Yassourcbff90a2008-07-28 19:26:24 +0300133
134 return true;
135}
136
Avi Kivity6aa8b732006-12-10 02:21:36 -0800137/*
138 * Switches to specified vcpu, until a matching vcpu_put()
139 */
Carsten Otte313a3dc2007-10-11 19:16:52 +0200140void vcpu_load(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800141{
Avi Kivity15ad7142007-07-11 18:17:21 +0300142 int cpu;
143
Avi Kivitybccf2152007-02-21 18:04:26 +0200144 mutex_lock(&vcpu->mutex);
Rik van Riel34bb10b2011-02-01 09:52:41 -0500145 if (unlikely(vcpu->pid != current->pids[PIDTYPE_PID].pid)) {
146 /* The thread running this VCPU changed. */
147 struct pid *oldpid = vcpu->pid;
148 struct pid *newpid = get_task_pid(current, PIDTYPE_PID);
149 rcu_assign_pointer(vcpu->pid, newpid);
150 synchronize_rcu();
151 put_pid(oldpid);
152 }
Avi Kivity15ad7142007-07-11 18:17:21 +0300153 cpu = get_cpu();
154 preempt_notifier_register(&vcpu->preempt_notifier);
Carsten Otte313a3dc2007-10-11 19:16:52 +0200155 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +0300156 put_cpu();
Avi Kivitybccf2152007-02-21 18:04:26 +0200157}
158
Carsten Otte313a3dc2007-10-11 19:16:52 +0200159void vcpu_put(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800160{
Avi Kivity15ad7142007-07-11 18:17:21 +0300161 preempt_disable();
Carsten Otte313a3dc2007-10-11 19:16:52 +0200162 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +0300163 preempt_notifier_unregister(&vcpu->preempt_notifier);
164 preempt_enable();
Avi Kivity6aa8b732006-12-10 02:21:36 -0800165 mutex_unlock(&vcpu->mutex);
166}
167
Avi Kivityd9e368d2007-06-07 19:18:30 +0300168static void ack_flush(void *_completed)
169{
Avi Kivityd9e368d2007-06-07 19:18:30 +0300170}
171
Rusty Russell49846892008-12-08 20:26:24 +1030172static bool make_all_cpus_request(struct kvm *kvm, unsigned int req)
Avi Kivityd9e368d2007-06-07 19:18:30 +0300173{
Avi Kivity597a5f52008-07-20 14:24:22 +0300174 int i, cpu, me;
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030175 cpumask_var_t cpus;
176 bool called = true;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300177 struct kvm_vcpu *vcpu;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300178
Li Zefan79f55992009-06-15 14:58:26 +0800179 zalloc_cpumask_var(&cpus, GFP_ATOMIC);
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030180
Xiao Guangrong3cba4132011-01-12 15:41:22 +0800181 me = get_cpu();
Gleb Natapov988a2ca2009-06-09 15:56:29 +0300182 kvm_for_each_vcpu(i, vcpu, kvm) {
Xiao Guangrong3cba4132011-01-12 15:41:22 +0800183 kvm_make_request(req, vcpu);
Avi Kivityd9e368d2007-06-07 19:18:30 +0300184 cpu = vcpu->cpu;
Xiao Guangrong6b7e2d02011-01-12 15:40:31 +0800185
186 /* Set ->requests bit before we read ->mode */
187 smp_mb();
188
189 if (cpus != NULL && cpu != -1 && cpu != me &&
190 kvm_vcpu_exiting_guest_mode(vcpu) != OUTSIDE_GUEST_MODE)
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030191 cpumask_set_cpu(cpu, cpus);
Avi Kivityd9e368d2007-06-07 19:18:30 +0300192 }
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030193 if (unlikely(cpus == NULL))
194 smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1);
195 else if (!cpumask_empty(cpus))
196 smp_call_function_many(cpus, ack_flush, NULL, 1);
197 else
198 called = false;
Xiao Guangrong3cba4132011-01-12 15:41:22 +0800199 put_cpu();
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030200 free_cpumask_var(cpus);
Rusty Russell49846892008-12-08 20:26:24 +1030201 return called;
202}
203
204void kvm_flush_remote_tlbs(struct kvm *kvm)
205{
Xiao Guangronga4ee1ca2010-11-23 11:13:00 +0800206 int dirty_count = kvm->tlbs_dirty;
207
208 smp_mb();
Rusty Russell49846892008-12-08 20:26:24 +1030209 if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH))
210 ++kvm->stat.remote_tlb_flush;
Xiao Guangronga4ee1ca2010-11-23 11:13:00 +0800211 cmpxchg(&kvm->tlbs_dirty, dirty_count, 0);
Avi Kivityd9e368d2007-06-07 19:18:30 +0300212}
213
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500214void kvm_reload_remote_mmus(struct kvm *kvm)
215{
Rusty Russell49846892008-12-08 20:26:24 +1030216 make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD);
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500217}
218
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000219int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id)
220{
221 struct page *page;
222 int r;
223
224 mutex_init(&vcpu->mutex);
225 vcpu->cpu = -1;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000226 vcpu->kvm = kvm;
227 vcpu->vcpu_id = id;
Rik van Riel34bb10b2011-02-01 09:52:41 -0500228 vcpu->pid = NULL;
Eddie Dongb6958ce2007-07-18 12:15:21 +0300229 init_waitqueue_head(&vcpu->wq);
Gleb Natapovaf585b92010-10-14 11:22:46 +0200230 kvm_async_pf_vcpu_init(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000231
232 page = alloc_page(GFP_KERNEL | __GFP_ZERO);
233 if (!page) {
234 r = -ENOMEM;
235 goto fail;
236 }
237 vcpu->run = page_address(page);
238
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800239 r = kvm_arch_vcpu_init(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000240 if (r < 0)
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800241 goto fail_free_run;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000242 return 0;
243
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000244fail_free_run:
245 free_page((unsigned long)vcpu->run);
246fail:
Rusty Russell76fafa52007-10-08 10:50:48 +1000247 return r;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000248}
249EXPORT_SYMBOL_GPL(kvm_vcpu_init);
250
251void kvm_vcpu_uninit(struct kvm_vcpu *vcpu)
252{
Rik van Riel34bb10b2011-02-01 09:52:41 -0500253 put_pid(vcpu->pid);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800254 kvm_arch_vcpu_uninit(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000255 free_page((unsigned long)vcpu->run);
256}
257EXPORT_SYMBOL_GPL(kvm_vcpu_uninit);
258
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200259#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
260static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn)
261{
262 return container_of(mn, struct kvm, mmu_notifier);
263}
264
265static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn,
266 struct mm_struct *mm,
267 unsigned long address)
268{
269 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200270 int need_tlb_flush, idx;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200271
272 /*
273 * When ->invalidate_page runs, the linux pte has been zapped
274 * already but the page is still allocated until
275 * ->invalidate_page returns. So if we increase the sequence
276 * here the kvm page fault will notice if the spte can't be
277 * established because the page is going to be freed. If
278 * instead the kvm page fault establishes the spte before
279 * ->invalidate_page runs, kvm_unmap_hva will release it
280 * before returning.
281 *
282 * The sequence increase only need to be seen at spin_unlock
283 * time, and not at spin_lock time.
284 *
285 * Increasing the sequence after the spin_unlock would be
286 * unsafe because the kvm page fault could then establish the
287 * pte after kvm_unmap_hva returned, without noticing the page
288 * is going to be freed.
289 */
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200290 idx = srcu_read_lock(&kvm->srcu);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200291 spin_lock(&kvm->mmu_lock);
292 kvm->mmu_notifier_seq++;
Xiao Guangronga4ee1ca2010-11-23 11:13:00 +0800293 need_tlb_flush = kvm_unmap_hva(kvm, address) | kvm->tlbs_dirty;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200294 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200295 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200296
297 /* we've to flush the tlb before the pages can be freed */
298 if (need_tlb_flush)
299 kvm_flush_remote_tlbs(kvm);
300
301}
302
Izik Eidus3da0dd42009-09-23 21:47:18 +0300303static void kvm_mmu_notifier_change_pte(struct mmu_notifier *mn,
304 struct mm_struct *mm,
305 unsigned long address,
306 pte_t pte)
307{
308 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200309 int idx;
Izik Eidus3da0dd42009-09-23 21:47:18 +0300310
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200311 idx = srcu_read_lock(&kvm->srcu);
Izik Eidus3da0dd42009-09-23 21:47:18 +0300312 spin_lock(&kvm->mmu_lock);
313 kvm->mmu_notifier_seq++;
314 kvm_set_spte_hva(kvm, address, pte);
315 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200316 srcu_read_unlock(&kvm->srcu, idx);
Izik Eidus3da0dd42009-09-23 21:47:18 +0300317}
318
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200319static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn,
320 struct mm_struct *mm,
321 unsigned long start,
322 unsigned long end)
323{
324 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200325 int need_tlb_flush = 0, idx;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200326
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200327 idx = srcu_read_lock(&kvm->srcu);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200328 spin_lock(&kvm->mmu_lock);
329 /*
330 * The count increase must become visible at unlock time as no
331 * spte can be established without taking the mmu_lock and
332 * count is also read inside the mmu_lock critical section.
333 */
334 kvm->mmu_notifier_count++;
335 for (; start < end; start += PAGE_SIZE)
336 need_tlb_flush |= kvm_unmap_hva(kvm, start);
Xiao Guangronga4ee1ca2010-11-23 11:13:00 +0800337 need_tlb_flush |= kvm->tlbs_dirty;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200338 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200339 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200340
341 /* we've to flush the tlb before the pages can be freed */
342 if (need_tlb_flush)
343 kvm_flush_remote_tlbs(kvm);
344}
345
346static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
347 struct mm_struct *mm,
348 unsigned long start,
349 unsigned long end)
350{
351 struct kvm *kvm = mmu_notifier_to_kvm(mn);
352
353 spin_lock(&kvm->mmu_lock);
354 /*
355 * This sequence increase will notify the kvm page fault that
356 * the page that is going to be mapped in the spte could have
357 * been freed.
358 */
359 kvm->mmu_notifier_seq++;
360 /*
361 * The above sequence increase must be visible before the
362 * below count decrease but both values are read by the kvm
363 * page fault under mmu_lock spinlock so we don't need to add
364 * a smb_wmb() here in between the two.
365 */
366 kvm->mmu_notifier_count--;
367 spin_unlock(&kvm->mmu_lock);
368
369 BUG_ON(kvm->mmu_notifier_count < 0);
370}
371
372static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
373 struct mm_struct *mm,
374 unsigned long address)
375{
376 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200377 int young, idx;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200378
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200379 idx = srcu_read_lock(&kvm->srcu);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200380 spin_lock(&kvm->mmu_lock);
381 young = kvm_age_hva(kvm, address);
382 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200383 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200384
385 if (young)
386 kvm_flush_remote_tlbs(kvm);
387
388 return young;
389}
390
Andrea Arcangeli8ee53822011-01-13 15:47:10 -0800391static int kvm_mmu_notifier_test_young(struct mmu_notifier *mn,
392 struct mm_struct *mm,
393 unsigned long address)
394{
395 struct kvm *kvm = mmu_notifier_to_kvm(mn);
396 int young, idx;
397
398 idx = srcu_read_lock(&kvm->srcu);
399 spin_lock(&kvm->mmu_lock);
400 young = kvm_test_age_hva(kvm, address);
401 spin_unlock(&kvm->mmu_lock);
402 srcu_read_unlock(&kvm->srcu, idx);
403
404 return young;
405}
406
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100407static void kvm_mmu_notifier_release(struct mmu_notifier *mn,
408 struct mm_struct *mm)
409{
410 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Lai Jiangshaneda2bed2010-04-20 14:29:29 +0800411 int idx;
412
413 idx = srcu_read_lock(&kvm->srcu);
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100414 kvm_arch_flush_shadow(kvm);
Lai Jiangshaneda2bed2010-04-20 14:29:29 +0800415 srcu_read_unlock(&kvm->srcu, idx);
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100416}
417
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200418static const struct mmu_notifier_ops kvm_mmu_notifier_ops = {
419 .invalidate_page = kvm_mmu_notifier_invalidate_page,
420 .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start,
421 .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end,
422 .clear_flush_young = kvm_mmu_notifier_clear_flush_young,
Andrea Arcangeli8ee53822011-01-13 15:47:10 -0800423 .test_young = kvm_mmu_notifier_test_young,
Izik Eidus3da0dd42009-09-23 21:47:18 +0300424 .change_pte = kvm_mmu_notifier_change_pte,
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100425 .release = kvm_mmu_notifier_release,
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200426};
Avi Kivity4c07b0a2009-12-20 14:54:04 +0200427
428static int kvm_init_mmu_notifier(struct kvm *kvm)
429{
430 kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops;
431 return mmu_notifier_register(&kvm->mmu_notifier, current->mm);
432}
433
434#else /* !(CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER) */
435
436static int kvm_init_mmu_notifier(struct kvm *kvm)
437{
438 return 0;
439}
440
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200441#endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */
442
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800443static void kvm_init_memslots_id(struct kvm *kvm)
444{
445 int i;
446 struct kvm_memslots *slots = kvm->memslots;
447
448 for (i = 0; i < KVM_MEM_SLOTS_NUM; i++)
Xiao Guangrongf85e2cb2011-11-24 17:41:54 +0800449 slots->id_to_index[i] = slots->memslots[i].id = i;
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800450}
451
Avi Kivityf17abe92007-02-21 19:28:04 +0200452static struct kvm *kvm_create_vm(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800453{
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100454 int r, i;
455 struct kvm *kvm = kvm_arch_alloc_vm();
Avi Kivity6aa8b732006-12-10 02:21:36 -0800456
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100457 if (!kvm)
458 return ERR_PTR(-ENOMEM);
459
460 r = kvm_arch_init_vm(kvm);
461 if (r)
462 goto out_err_nodisable;
Alexander Graf10474ae2009-09-15 11:37:46 +0200463
464 r = hardware_enable_all();
465 if (r)
466 goto out_err_nodisable;
467
Avi Kivity75858a82009-01-04 17:10:50 +0200468#ifdef CONFIG_HAVE_KVM_IRQCHIP
469 INIT_HLIST_HEAD(&kvm->mask_notifier_list);
Gleb Natapov136bdfe2009-08-24 11:54:23 +0300470 INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list);
Avi Kivity75858a82009-01-04 17:10:50 +0200471#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -0800472
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200473 r = -ENOMEM;
474 kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
475 if (!kvm->memslots)
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100476 goto out_err_nosrcu;
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800477 kvm_init_memslots_id(kvm);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200478 if (init_srcu_struct(&kvm->srcu))
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100479 goto out_err_nosrcu;
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200480 for (i = 0; i < KVM_NR_BUSES; i++) {
481 kvm->buses[i] = kzalloc(sizeof(struct kvm_io_bus),
482 GFP_KERNEL);
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100483 if (!kvm->buses[i])
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200484 goto out_err;
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200485 }
Mike Waychison74b5c5b2011-06-03 13:04:53 -0700486
OGAWA Hirofumi85722cd2011-05-11 09:28:28 +0900487 spin_lock_init(&kvm->mmu_lock);
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200488 kvm->mm = current->mm;
489 atomic_inc(&kvm->mm->mm_count);
Gregory Haskinsd34e6b12009-07-07 17:08:49 -0400490 kvm_eventfd_init(kvm);
Shaohua Li11ec2802007-07-23 14:51:37 +0800491 mutex_init(&kvm->lock);
Marcelo Tosatti60eead72009-06-04 15:08:23 -0300492 mutex_init(&kvm->irq_lock);
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200493 mutex_init(&kvm->slots_lock);
Izik Eidusd39f13b2008-03-30 16:01:25 +0300494 atomic_set(&kvm->users_count, 1);
Mike Waychison74b5c5b2011-06-03 13:04:53 -0700495
496 r = kvm_init_mmu_notifier(kvm);
497 if (r)
498 goto out_err;
499
Jan Kiszkae935b832011-02-08 12:55:33 +0100500 raw_spin_lock(&kvm_lock);
Rusty Russell5e58cfe2007-07-23 17:08:21 +1000501 list_add(&kvm->vm_list, &vm_list);
Jan Kiszkae935b832011-02-08 12:55:33 +0100502 raw_spin_unlock(&kvm_lock);
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100503
Avi Kivityf17abe92007-02-21 19:28:04 +0200504 return kvm;
Alexander Graf10474ae2009-09-15 11:37:46 +0200505
506out_err:
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100507 cleanup_srcu_struct(&kvm->srcu);
508out_err_nosrcu:
Alexander Graf10474ae2009-09-15 11:37:46 +0200509 hardware_disable_all();
510out_err_nodisable:
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200511 for (i = 0; i < KVM_NR_BUSES; i++)
512 kfree(kvm->buses[i]);
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200513 kfree(kvm->memslots);
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100514 kvm_arch_free_vm(kvm);
Alexander Graf10474ae2009-09-15 11:37:46 +0200515 return ERR_PTR(r);
Avi Kivityf17abe92007-02-21 19:28:04 +0200516}
517
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900518static void kvm_destroy_dirty_bitmap(struct kvm_memory_slot *memslot)
519{
520 if (!memslot->dirty_bitmap)
521 return;
522
Takuya Yoshikawa6f9e5c172010-11-01 14:36:09 +0900523 if (2 * kvm_dirty_bitmap_bytes(memslot) > PAGE_SIZE)
524 vfree(memslot->dirty_bitmap_head);
525 else
526 kfree(memslot->dirty_bitmap_head);
527
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900528 memslot->dirty_bitmap = NULL;
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900529 memslot->dirty_bitmap_head = NULL;
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900530}
531
Avi Kivity6aa8b732006-12-10 02:21:36 -0800532/*
533 * Free any memory in @free but not in @dont.
534 */
535static void kvm_free_physmem_slot(struct kvm_memory_slot *free,
536 struct kvm_memory_slot *dont)
537{
Joerg Roedelec04b262009-06-19 15:16:23 +0200538 int i;
539
Izik Eidus290fc382007-09-27 14:11:22 +0200540 if (!dont || free->rmap != dont->rmap)
541 vfree(free->rmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800542
543 if (!dont || free->dirty_bitmap != dont->dirty_bitmap)
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900544 kvm_destroy_dirty_bitmap(free);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800545
Joerg Roedelec04b262009-06-19 15:16:23 +0200546
547 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) {
548 if (!dont || free->lpage_info[i] != dont->lpage_info[i]) {
549 vfree(free->lpage_info[i]);
550 free->lpage_info[i] = NULL;
551 }
552 }
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300553
Avi Kivity6aa8b732006-12-10 02:21:36 -0800554 free->npages = 0;
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500555 free->rmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800556}
557
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800558void kvm_free_physmem(struct kvm *kvm)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800559{
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200560 struct kvm_memslots *slots = kvm->memslots;
Xiao Guangrongbe6ba0f2011-11-24 17:39:18 +0800561 struct kvm_memory_slot *memslot;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800562
Xiao Guangrongbe6ba0f2011-11-24 17:39:18 +0800563 kvm_for_each_memslot(memslot, slots)
564 kvm_free_physmem_slot(memslot, NULL);
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200565
566 kfree(kvm->memslots);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800567}
568
Avi Kivityf17abe92007-02-21 19:28:04 +0200569static void kvm_destroy_vm(struct kvm *kvm)
570{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200571 int i;
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200572 struct mm_struct *mm = kvm->mm;
573
Sheng Yangad8ba2c2009-01-06 10:03:02 +0800574 kvm_arch_sync_events(kvm);
Jan Kiszkae935b832011-02-08 12:55:33 +0100575 raw_spin_lock(&kvm_lock);
Avi Kivity133de902007-02-12 00:54:44 -0800576 list_del(&kvm->vm_list);
Jan Kiszkae935b832011-02-08 12:55:33 +0100577 raw_spin_unlock(&kvm_lock);
Avi Kivity399ec802008-11-19 13:58:46 +0200578 kvm_free_irq_routing(kvm);
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200579 for (i = 0; i < KVM_NR_BUSES; i++)
580 kvm_io_bus_destroy(kvm->buses[i]);
Avi Kivity980da6c2009-12-20 15:13:43 +0200581 kvm_coalesced_mmio_free(kvm);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200582#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
583 mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
Gleb Natapovf00be0c2009-03-19 12:20:36 +0200584#else
585 kvm_arch_flush_shadow(kvm);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200586#endif
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800587 kvm_arch_destroy_vm(kvm);
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100588 kvm_free_physmem(kvm);
589 cleanup_srcu_struct(&kvm->srcu);
590 kvm_arch_free_vm(kvm);
Alexander Graf10474ae2009-09-15 11:37:46 +0200591 hardware_disable_all();
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200592 mmdrop(mm);
Avi Kivityf17abe92007-02-21 19:28:04 +0200593}
594
Izik Eidusd39f13b2008-03-30 16:01:25 +0300595void kvm_get_kvm(struct kvm *kvm)
596{
597 atomic_inc(&kvm->users_count);
598}
599EXPORT_SYMBOL_GPL(kvm_get_kvm);
600
601void kvm_put_kvm(struct kvm *kvm)
602{
603 if (atomic_dec_and_test(&kvm->users_count))
604 kvm_destroy_vm(kvm);
605}
606EXPORT_SYMBOL_GPL(kvm_put_kvm);
607
608
Avi Kivityf17abe92007-02-21 19:28:04 +0200609static int kvm_vm_release(struct inode *inode, struct file *filp)
610{
611 struct kvm *kvm = filp->private_data;
612
Gregory Haskins721eecb2009-05-20 10:30:49 -0400613 kvm_irqfd_release(kvm);
614
Izik Eidusd39f13b2008-03-30 16:01:25 +0300615 kvm_put_kvm(kvm);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800616 return 0;
617}
618
Heiko Carstensd48ead82011-01-17 21:21:08 +0100619#ifndef CONFIG_S390
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900620/*
621 * Allocation size is twice as large as the actual dirty bitmap size.
622 * This makes it possible to do double buffering: see x86's
623 * kvm_vm_ioctl_get_dirty_log().
624 */
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900625static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot)
626{
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900627 unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot);
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900628
Takuya Yoshikawa6f9e5c172010-11-01 14:36:09 +0900629 if (dirty_bytes > PAGE_SIZE)
630 memslot->dirty_bitmap = vzalloc(dirty_bytes);
631 else
632 memslot->dirty_bitmap = kzalloc(dirty_bytes, GFP_KERNEL);
633
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900634 if (!memslot->dirty_bitmap)
635 return -ENOMEM;
636
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900637 memslot->dirty_bitmap_head = memslot->dirty_bitmap;
Takuya Yoshikawa7850ac52011-11-14 18:23:34 +0900638 memslot->nr_dirty_pages = 0;
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900639 return 0;
640}
Heiko Carstensd48ead82011-01-17 21:21:08 +0100641#endif /* !CONFIG_S390 */
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900642
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800643static struct kvm_memory_slot *
644search_memslots(struct kvm_memslots *slots, gfn_t gfn)
645{
646 struct kvm_memory_slot *memslot;
647
648 kvm_for_each_memslot(memslot, slots)
649 if (gfn >= memslot->base_gfn &&
650 gfn < memslot->base_gfn + memslot->npages)
651 return memslot;
652
653 return NULL;
654}
655
656static int cmp_memslot(const void *slot1, const void *slot2)
657{
658 struct kvm_memory_slot *s1, *s2;
659
660 s1 = (struct kvm_memory_slot *)slot1;
661 s2 = (struct kvm_memory_slot *)slot2;
662
663 if (s1->npages < s2->npages)
664 return 1;
665 if (s1->npages > s2->npages)
666 return -1;
667
668 return 0;
669}
670
671/*
672 * Sort the memslots base on its size, so the larger slots
673 * will get better fit.
674 */
675static void sort_memslots(struct kvm_memslots *slots)
676{
Xiao Guangrongf85e2cb2011-11-24 17:41:54 +0800677 int i;
678
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800679 sort(slots->memslots, KVM_MEM_SLOTS_NUM,
680 sizeof(struct kvm_memory_slot), cmp_memslot, NULL);
Xiao Guangrongf85e2cb2011-11-24 17:41:54 +0800681
682 for (i = 0; i < KVM_MEM_SLOTS_NUM; i++)
683 slots->id_to_index[slots->memslots[i].id] = i;
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800684}
685
Xiao Guangrongbe593d62011-11-24 17:38:24 +0800686void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new)
687{
688 if (new) {
689 int id = new->id;
Xiao Guangrong28a37542011-11-24 19:04:35 +0800690 struct kvm_memory_slot *old = id_to_memslot(slots, id);
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800691 unsigned long npages = old->npages;
Xiao Guangrongbe593d62011-11-24 17:38:24 +0800692
Xiao Guangrong28a37542011-11-24 19:04:35 +0800693 *old = *new;
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800694 if (new->npages != npages)
695 sort_memslots(slots);
Xiao Guangrongbe593d62011-11-24 17:38:24 +0800696 }
697
698 slots->generation++;
699}
700
Avi Kivity6aa8b732006-12-10 02:21:36 -0800701/*
Avi Kivity6aa8b732006-12-10 02:21:36 -0800702 * Allocate some memory and give it an address in the guest physical address
703 * space.
704 *
705 * Discontiguous memory is allowed, mostly for framebuffers.
Sheng Yangf78e0e22007-10-29 09:40:42 +0800706 *
Marcelo Tosatti10589a42007-12-20 19:18:22 -0500707 * Must be called holding mmap_sem for write.
Avi Kivity6aa8b732006-12-10 02:21:36 -0800708 */
Sheng Yangf78e0e22007-10-29 09:40:42 +0800709int __kvm_set_memory_region(struct kvm *kvm,
710 struct kvm_userspace_memory_region *mem,
711 int user_alloc)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800712{
Avi Kivity8234b222010-12-27 12:08:45 +0200713 int r;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800714 gfn_t base_gfn;
Heiko Carstens28bcb112009-09-03 17:35:35 +0200715 unsigned long npages;
716 unsigned long i;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800717 struct kvm_memory_slot *memslot;
718 struct kvm_memory_slot old, new;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200719 struct kvm_memslots *slots, *old_memslots;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800720
721 r = -EINVAL;
722 /* General sanity checks */
723 if (mem->memory_size & (PAGE_SIZE - 1))
724 goto out;
725 if (mem->guest_phys_addr & (PAGE_SIZE - 1))
726 goto out;
Takuya Yoshikawafa3d3152011-05-07 16:35:38 +0900727 /* We can read the guest memory with __xxx_user() later on. */
728 if (user_alloc &&
729 ((mem->userspace_addr & (PAGE_SIZE - 1)) ||
Heiko Carstens9e3bb6b2011-05-24 07:51:27 +0200730 !access_ok(VERIFY_WRITE,
731 (void __user *)(unsigned long)mem->userspace_addr,
732 mem->memory_size)))
Hollis Blanchard78749802008-11-07 13:32:12 -0600733 goto out;
Xiao Guangrong93a5cef2011-11-24 17:37:48 +0800734 if (mem->slot >= KVM_MEM_SLOTS_NUM)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800735 goto out;
736 if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
737 goto out;
738
Xiao Guangrong28a37542011-11-24 19:04:35 +0800739 memslot = id_to_memslot(kvm->memslots, mem->slot);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800740 base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
741 npages = mem->memory_size >> PAGE_SHIFT;
742
Takuya Yoshikawa660c22c2010-04-13 22:47:24 +0900743 r = -EINVAL;
744 if (npages > KVM_MEM_MAX_NR_PAGES)
745 goto out;
746
Avi Kivity6aa8b732006-12-10 02:21:36 -0800747 if (!npages)
748 mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES;
749
Avi Kivity6aa8b732006-12-10 02:21:36 -0800750 new = old = *memslot;
751
Avi Kivitye36d96f2010-06-21 10:56:36 +0300752 new.id = mem->slot;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800753 new.base_gfn = base_gfn;
754 new.npages = npages;
755 new.flags = mem->flags;
756
757 /* Disallow changing a memory slot's size. */
758 r = -EINVAL;
759 if (npages && old.npages && npages != old.npages)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800760 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800761
762 /* Check for overlaps */
763 r = -EEXIST;
764 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200765 struct kvm_memory_slot *s = &kvm->memslots->memslots[i];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800766
Jan Kiszka4cd481f2009-04-13 11:59:32 +0200767 if (s == memslot || !s->npages)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800768 continue;
769 if (!((base_gfn + npages <= s->base_gfn) ||
770 (base_gfn >= s->base_gfn + s->npages)))
Sheng Yangf78e0e22007-10-29 09:40:42 +0800771 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800772 }
Avi Kivity6aa8b732006-12-10 02:21:36 -0800773
Avi Kivity6aa8b732006-12-10 02:21:36 -0800774 /* Free page dirty bitmap if unneeded */
775 if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
Al Viro8b6d44c2007-02-09 16:38:40 +0000776 new.dirty_bitmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800777
778 r = -ENOMEM;
779
780 /* Allocate if a slot is being created */
Carsten Otteeff01142008-06-27 15:05:31 +0200781#ifndef CONFIG_S390
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500782 if (npages && !new.rmap) {
Takuya Yoshikawa26535032010-11-02 10:49:34 +0900783 new.rmap = vzalloc(npages * sizeof(*new.rmap));
Izik Eidus290fc382007-09-27 14:11:22 +0200784
785 if (!new.rmap)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800786 goto out_free;
Izik Eidus290fc382007-09-27 14:11:22 +0200787
Izik Eidus80b14b52007-10-25 11:54:04 +0200788 new.user_alloc = user_alloc;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200789 new.userspace_addr = mem->userspace_addr;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800790 }
Joerg Roedelec04b262009-06-19 15:16:23 +0200791 if (!npages)
792 goto skip_lpage;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300793
Joerg Roedelec04b262009-06-19 15:16:23 +0200794 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) {
Heiko Carstens28bcb112009-09-03 17:35:35 +0200795 unsigned long ugfn;
796 unsigned long j;
797 int lpages;
Joerg Roedelec04b262009-06-19 15:16:23 +0200798 int level = i + 2;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300799
Joerg Roedelec04b262009-06-19 15:16:23 +0200800 /* Avoid unused variable warning if no large pages */
801 (void)level;
802
803 if (new.lpage_info[i])
804 continue;
805
Joerg Roedel82855412010-07-01 16:00:11 +0200806 lpages = 1 + ((base_gfn + npages - 1)
807 >> KVM_HPAGE_GFN_SHIFT(level));
808 lpages -= base_gfn >> KVM_HPAGE_GFN_SHIFT(level);
Joerg Roedelec04b262009-06-19 15:16:23 +0200809
Takuya Yoshikawa26535032010-11-02 10:49:34 +0900810 new.lpage_info[i] = vzalloc(lpages * sizeof(*new.lpage_info[i]));
Joerg Roedelec04b262009-06-19 15:16:23 +0200811
812 if (!new.lpage_info[i])
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300813 goto out_free;
814
Joerg Roedel82855412010-07-01 16:00:11 +0200815 if (base_gfn & (KVM_PAGES_PER_HPAGE(level) - 1))
Joerg Roedelec04b262009-06-19 15:16:23 +0200816 new.lpage_info[i][0].write_count = 1;
Joerg Roedel82855412010-07-01 16:00:11 +0200817 if ((base_gfn+npages) & (KVM_PAGES_PER_HPAGE(level) - 1))
Joerg Roedelec04b262009-06-19 15:16:23 +0200818 new.lpage_info[i][lpages - 1].write_count = 1;
Avi Kivityac045272009-06-08 15:52:39 +0300819 ugfn = new.userspace_addr >> PAGE_SHIFT;
820 /*
821 * If the gfn and userspace address are not aligned wrt each
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300822 * other, or if explicitly asked to, disable large page
823 * support for this slot
Avi Kivityac045272009-06-08 15:52:39 +0300824 */
Joerg Roedelec04b262009-06-19 15:16:23 +0200825 if ((base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1) ||
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300826 !largepages_enabled)
Joerg Roedelec04b262009-06-19 15:16:23 +0200827 for (j = 0; j < lpages; ++j)
828 new.lpage_info[i][j].write_count = 1;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300829 }
Avi Kivity6aa8b732006-12-10 02:21:36 -0800830
Joerg Roedelec04b262009-06-19 15:16:23 +0200831skip_lpage:
832
Avi Kivity6aa8b732006-12-10 02:21:36 -0800833 /* Allocate page dirty bitmap if needed */
834 if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900835 if (kvm_create_dirty_bitmap(&new) < 0)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800836 goto out_free;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200837 /* destroy any largepage mappings for dirty tracking */
Avi Kivity6aa8b732006-12-10 02:21:36 -0800838 }
Christian Borntraeger3eea8432009-06-23 17:24:06 +0200839#else /* not defined CONFIG_S390 */
840 new.user_alloc = user_alloc;
841 if (user_alloc)
842 new.userspace_addr = mem->userspace_addr;
Carsten Otteeff01142008-06-27 15:05:31 +0200843#endif /* not defined CONFIG_S390 */
Avi Kivity6aa8b732006-12-10 02:21:36 -0800844
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200845 if (!npages) {
Xiao Guangrong28a37542011-11-24 19:04:35 +0800846 struct kvm_memory_slot *slot;
847
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200848 r = -ENOMEM;
Thomas Meyer6da64fd2011-11-08 20:32:19 +0100849 slots = kmemdup(kvm->memslots, sizeof(struct kvm_memslots),
850 GFP_KERNEL);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200851 if (!slots)
852 goto out_free;
Xiao Guangrong28a37542011-11-24 19:04:35 +0800853 slot = id_to_memslot(slots, mem->slot);
854 slot->flags |= KVM_MEMSLOT_INVALID;
855
Xiao Guangrongbe593d62011-11-24 17:38:24 +0800856 update_memslots(slots, NULL);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200857
858 old_memslots = kvm->memslots;
859 rcu_assign_pointer(kvm->memslots, slots);
860 synchronize_srcu_expedited(&kvm->srcu);
861 /* From this point no new shadow pages pointing to a deleted
862 * memslot will be created.
863 *
864 * validation of sp->gfn happens in:
865 * - gfn_to_hva (kvm_read_guest, gfn_to_pfn)
866 * - kvm_is_visible_gfn (mmu_check_roots)
867 */
Marcelo Tosatti34d4cb82008-07-10 20:49:31 -0300868 kvm_arch_flush_shadow(kvm);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200869 kfree(old_memslots);
870 }
Marcelo Tosatti34d4cb82008-07-10 20:49:31 -0300871
Marcelo Tosattif7784b82009-12-23 14:35:18 -0200872 r = kvm_arch_prepare_memory_region(kvm, &new, old, mem, user_alloc);
873 if (r)
874 goto out_free;
875
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200876 /* map the pages in iommu page table */
877 if (npages) {
878 r = kvm_iommu_map_pages(kvm, &new);
879 if (r)
880 goto out_free;
881 }
Andrea Arcangeli604b38a2008-07-25 16:32:03 +0200882
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200883 r = -ENOMEM;
Thomas Meyer6da64fd2011-11-08 20:32:19 +0100884 slots = kmemdup(kvm->memslots, sizeof(struct kvm_memslots),
885 GFP_KERNEL);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200886 if (!slots)
887 goto out_free;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200888
889 /* actual memory is freed via old in kvm_free_physmem_slot below */
890 if (!npages) {
891 new.rmap = NULL;
892 new.dirty_bitmap = NULL;
893 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i)
894 new.lpage_info[i] = NULL;
895 }
896
Xiao Guangrongbe593d62011-11-24 17:38:24 +0800897 update_memslots(slots, &new);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200898 old_memslots = kvm->memslots;
899 rcu_assign_pointer(kvm->memslots, slots);
900 synchronize_srcu_expedited(&kvm->srcu);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800901
Marcelo Tosattif7784b82009-12-23 14:35:18 -0200902 kvm_arch_commit_memory_region(kvm, mem, old, user_alloc);
Zhang Xiantao3ad82a72007-11-20 13:11:38 +0800903
Xiao Guangrongce88dec2011-07-12 03:33:44 +0800904 /*
905 * If the new memory slot is created, we need to clear all
906 * mmio sptes.
907 */
908 if (npages && old.base_gfn != mem->guest_phys_addr >> PAGE_SHIFT)
909 kvm_arch_flush_shadow(kvm);
910
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200911 kvm_free_physmem_slot(&old, &new);
912 kfree(old_memslots);
913
Avi Kivity6aa8b732006-12-10 02:21:36 -0800914 return 0;
915
Sheng Yangf78e0e22007-10-29 09:40:42 +0800916out_free:
Avi Kivity6aa8b732006-12-10 02:21:36 -0800917 kvm_free_physmem_slot(&new, &old);
918out:
919 return r;
Izik Eidus210c7c42007-10-24 23:52:57 +0200920
921}
Sheng Yangf78e0e22007-10-29 09:40:42 +0800922EXPORT_SYMBOL_GPL(__kvm_set_memory_region);
923
924int kvm_set_memory_region(struct kvm *kvm,
925 struct kvm_userspace_memory_region *mem,
926 int user_alloc)
927{
928 int r;
929
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200930 mutex_lock(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +0800931 r = __kvm_set_memory_region(kvm, mem, user_alloc);
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200932 mutex_unlock(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +0800933 return r;
934}
Izik Eidus210c7c42007-10-24 23:52:57 +0200935EXPORT_SYMBOL_GPL(kvm_set_memory_region);
936
Carsten Otte1fe779f2007-10-29 16:08:35 +0100937int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
938 struct
939 kvm_userspace_memory_region *mem,
940 int user_alloc)
Izik Eidus210c7c42007-10-24 23:52:57 +0200941{
Izik Eiduse0d62c72007-10-24 23:57:46 +0200942 if (mem->slot >= KVM_MEMORY_SLOTS)
943 return -EINVAL;
Izik Eidus210c7c42007-10-24 23:52:57 +0200944 return kvm_set_memory_region(kvm, mem, user_alloc);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800945}
946
Zhang Xiantao5bb064d2007-11-18 20:29:43 +0800947int kvm_get_dirty_log(struct kvm *kvm,
948 struct kvm_dirty_log *log, int *is_dirty)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800949{
950 struct kvm_memory_slot *memslot;
951 int r, i;
Takuya Yoshikawa87bf6e72010-04-12 19:35:35 +0900952 unsigned long n;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800953 unsigned long any = 0;
954
Avi Kivity6aa8b732006-12-10 02:21:36 -0800955 r = -EINVAL;
956 if (log->slot >= KVM_MEMORY_SLOTS)
957 goto out;
958
Xiao Guangrong28a37542011-11-24 19:04:35 +0800959 memslot = id_to_memslot(kvm->memslots, log->slot);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800960 r = -ENOENT;
961 if (!memslot->dirty_bitmap)
962 goto out;
963
Takuya Yoshikawa87bf6e72010-04-12 19:35:35 +0900964 n = kvm_dirty_bitmap_bytes(memslot);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800965
Uri Lublincd1a4a92007-02-22 16:43:09 +0200966 for (i = 0; !any && i < n/sizeof(long); ++i)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800967 any = memslot->dirty_bitmap[i];
968
969 r = -EFAULT;
970 if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
971 goto out;
972
Zhang Xiantao5bb064d2007-11-18 20:29:43 +0800973 if (any)
974 *is_dirty = 1;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800975
976 r = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800977out:
Avi Kivity6aa8b732006-12-10 02:21:36 -0800978 return r;
979}
980
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300981void kvm_disable_largepages(void)
982{
983 largepages_enabled = false;
984}
985EXPORT_SYMBOL_GPL(kvm_disable_largepages);
986
Izik Eiduscea7bb22007-10-17 19:17:48 +0200987int is_error_page(struct page *page)
988{
Gleb Natapovedba23e2010-07-07 20:16:45 +0300989 return page == bad_page || page == hwpoison_page || page == fault_page;
Izik Eiduscea7bb22007-10-17 19:17:48 +0200990}
991EXPORT_SYMBOL_GPL(is_error_page);
992
Anthony Liguori35149e22008-04-02 14:46:56 -0500993int is_error_pfn(pfn_t pfn)
994{
Gleb Natapovedba23e2010-07-07 20:16:45 +0300995 return pfn == bad_pfn || pfn == hwpoison_pfn || pfn == fault_pfn;
Anthony Liguori35149e22008-04-02 14:46:56 -0500996}
997EXPORT_SYMBOL_GPL(is_error_pfn);
998
Huang Yingbf998152010-05-31 14:28:19 +0800999int is_hwpoison_pfn(pfn_t pfn)
1000{
1001 return pfn == hwpoison_pfn;
1002}
1003EXPORT_SYMBOL_GPL(is_hwpoison_pfn);
1004
Gleb Natapovedba23e2010-07-07 20:16:45 +03001005int is_fault_pfn(pfn_t pfn)
1006{
1007 return pfn == fault_pfn;
1008}
1009EXPORT_SYMBOL_GPL(is_fault_pfn);
1010
Xiao Guangrongfce92dc2011-07-12 03:28:54 +08001011int is_noslot_pfn(pfn_t pfn)
1012{
1013 return pfn == bad_pfn;
1014}
1015EXPORT_SYMBOL_GPL(is_noslot_pfn);
1016
1017int is_invalid_pfn(pfn_t pfn)
1018{
1019 return pfn == hwpoison_pfn || pfn == fault_pfn;
1020}
1021EXPORT_SYMBOL_GPL(is_invalid_pfn);
1022
Izik Eidusf9d46eb2007-11-11 22:02:22 +02001023static inline unsigned long bad_hva(void)
1024{
1025 return PAGE_OFFSET;
1026}
1027
1028int kvm_is_error_hva(unsigned long addr)
1029{
1030 return addr == bad_hva();
1031}
1032EXPORT_SYMBOL_GPL(kvm_is_error_hva);
1033
Gleb Natapov49c77542010-10-18 15:22:23 +02001034static struct kvm_memory_slot *__gfn_to_memslot(struct kvm_memslots *slots,
1035 gfn_t gfn)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001036{
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +08001037 return search_memslots(slots, gfn);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001038}
Gleb Natapov49c77542010-10-18 15:22:23 +02001039
1040struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn)
1041{
1042 return __gfn_to_memslot(kvm_memslots(kvm), gfn);
1043}
Avi Kivitya1f4d392010-06-21 11:44:20 +03001044EXPORT_SYMBOL_GPL(gfn_to_memslot);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001045
Izik Eiduse0d62c72007-10-24 23:57:46 +02001046int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
1047{
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +08001048 struct kvm_memory_slot *memslot = gfn_to_memslot(kvm, gfn);
Izik Eiduse0d62c72007-10-24 23:57:46 +02001049
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +08001050 if (!memslot || memslot->id >= KVM_MEMORY_SLOTS ||
1051 memslot->flags & KVM_MEMSLOT_INVALID)
1052 return 0;
Izik Eiduse0d62c72007-10-24 23:57:46 +02001053
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +08001054 return 1;
Izik Eiduse0d62c72007-10-24 23:57:46 +02001055}
1056EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);
1057
Joerg Roedel8f0b1ab2010-01-28 12:37:56 +01001058unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn)
1059{
1060 struct vm_area_struct *vma;
1061 unsigned long addr, size;
1062
1063 size = PAGE_SIZE;
1064
1065 addr = gfn_to_hva(kvm, gfn);
1066 if (kvm_is_error_hva(addr))
1067 return PAGE_SIZE;
1068
1069 down_read(&current->mm->mmap_sem);
1070 vma = find_vma(current->mm, addr);
1071 if (!vma)
1072 goto out;
1073
1074 size = vma_kernel_pagesize(vma);
1075
1076out:
1077 up_read(&current->mm->mmap_sem);
1078
1079 return size;
1080}
1081
Gleb Natapov49c77542010-10-18 15:22:23 +02001082static unsigned long gfn_to_hva_many(struct kvm_memory_slot *slot, gfn_t gfn,
Xiao Guangrong48987782010-08-22 19:11:43 +08001083 gfn_t *nr_pages)
Izik Eidus539cb662007-11-11 22:05:04 +02001084{
Marcelo Tosattibc6678a2009-12-23 14:35:21 -02001085 if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
Izik Eidus539cb662007-11-11 22:05:04 +02001086 return bad_hva();
Xiao Guangrong48987782010-08-22 19:11:43 +08001087
1088 if (nr_pages)
1089 *nr_pages = slot->npages - (gfn - slot->base_gfn);
1090
Takuya Yoshikawaf5c98032010-02-25 11:33:19 +09001091 return gfn_to_hva_memslot(slot, gfn);
Izik Eidus539cb662007-11-11 22:05:04 +02001092}
Xiao Guangrong48987782010-08-22 19:11:43 +08001093
1094unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
1095{
Gleb Natapov49c77542010-10-18 15:22:23 +02001096 return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL);
Xiao Guangrong48987782010-08-22 19:11:43 +08001097}
Sheng Yang0d150292008-04-25 21:44:50 +08001098EXPORT_SYMBOL_GPL(gfn_to_hva);
Izik Eidus539cb662007-11-11 22:05:04 +02001099
Gleb Natapov80300892010-10-19 18:13:41 +02001100static pfn_t get_fault_pfn(void)
1101{
1102 get_page(fault_page);
1103 return fault_pfn;
1104}
1105
Gleb Natapov0857b9e2011-02-01 13:21:47 +02001106int get_user_page_nowait(struct task_struct *tsk, struct mm_struct *mm,
1107 unsigned long start, int write, struct page **page)
1108{
1109 int flags = FOLL_TOUCH | FOLL_NOWAIT | FOLL_HWPOISON | FOLL_GET;
1110
1111 if (write)
1112 flags |= FOLL_WRITE;
1113
1114 return __get_user_pages(tsk, mm, start, 1, flags, page, NULL, NULL);
1115}
1116
Huang Yingfafc3db2011-01-30 11:15:49 +08001117static inline int check_user_page_hwpoison(unsigned long addr)
1118{
1119 int rc, flags = FOLL_TOUCH | FOLL_HWPOISON | FOLL_WRITE;
1120
1121 rc = __get_user_pages(current, current->mm, addr, 1,
1122 flags, NULL, NULL, NULL);
1123 return rc == -EHWPOISON;
1124}
1125
Gleb Natapovaf585b92010-10-14 11:22:46 +02001126static pfn_t hva_to_pfn(struct kvm *kvm, unsigned long addr, bool atomic,
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001127 bool *async, bool write_fault, bool *writable)
Avi Kivity954bbbc22007-03-30 14:02:32 +03001128{
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001129 struct page *page[1];
Gleb Natapovaf585b92010-10-14 11:22:46 +02001130 int npages = 0;
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001131 pfn_t pfn;
Avi Kivity954bbbc22007-03-30 14:02:32 +03001132
Gleb Natapovaf585b92010-10-14 11:22:46 +02001133 /* we can do it either atomically or asynchronously, not both */
1134 BUG_ON(atomic && async);
1135
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001136 BUG_ON(!write_fault && !writable);
1137
1138 if (writable)
1139 *writable = true;
1140
Gleb Natapovaf585b92010-10-14 11:22:46 +02001141 if (atomic || async)
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001142 npages = __get_user_pages_fast(addr, 1, 1, page);
Gleb Natapovaf585b92010-10-14 11:22:46 +02001143
1144 if (unlikely(npages != 1) && !atomic) {
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001145 might_sleep();
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001146
1147 if (writable)
1148 *writable = write_fault;
1149
Gleb Natapov0857b9e2011-02-01 13:21:47 +02001150 if (async) {
1151 down_read(&current->mm->mmap_sem);
1152 npages = get_user_page_nowait(current, current->mm,
1153 addr, write_fault, page);
1154 up_read(&current->mm->mmap_sem);
1155 } else
1156 npages = get_user_pages_fast(addr, 1, write_fault,
1157 page);
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001158
1159 /* map read fault as writable if possible */
1160 if (unlikely(!write_fault) && npages == 1) {
1161 struct page *wpage[1];
1162
1163 npages = __get_user_pages_fast(addr, 1, 1, wpage);
1164 if (npages == 1) {
1165 *writable = true;
1166 put_page(page[0]);
1167 page[0] = wpage[0];
1168 }
1169 npages = 1;
1170 }
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001171 }
Izik Eidus539cb662007-11-11 22:05:04 +02001172
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001173 if (unlikely(npages != 1)) {
1174 struct vm_area_struct *vma;
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001175
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001176 if (atomic)
Gleb Natapov80300892010-10-19 18:13:41 +02001177 return get_fault_pfn();
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001178
Huang Yingbbeb3402010-06-22 14:23:11 +08001179 down_read(&current->mm->mmap_sem);
Gleb Natapov0857b9e2011-02-01 13:21:47 +02001180 if (npages == -EHWPOISON ||
1181 (!async && check_user_page_hwpoison(addr))) {
Huang Yingbbeb3402010-06-22 14:23:11 +08001182 up_read(&current->mm->mmap_sem);
Huang Yingbf998152010-05-31 14:28:19 +08001183 get_page(hwpoison_page);
1184 return page_to_pfn(hwpoison_page);
1185 }
1186
Gleb Natapov80300892010-10-19 18:13:41 +02001187 vma = find_vma_intersection(current->mm, addr, addr+1);
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001188
Gleb Natapov80300892010-10-19 18:13:41 +02001189 if (vma == NULL)
1190 pfn = get_fault_pfn();
1191 else if ((vma->vm_flags & VM_PFNMAP)) {
1192 pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) +
1193 vma->vm_pgoff;
1194 BUG_ON(!kvm_is_mmio_pfn(pfn));
1195 } else {
1196 if (async && (vma->vm_flags & VM_WRITE))
Gleb Natapovaf585b92010-10-14 11:22:46 +02001197 *async = true;
Gleb Natapov80300892010-10-19 18:13:41 +02001198 pfn = get_fault_pfn();
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001199 }
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001200 up_read(&current->mm->mmap_sem);
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001201 } else
1202 pfn = page_to_pfn(page[0]);
1203
1204 return pfn;
Anthony Liguori35149e22008-04-02 14:46:56 -05001205}
1206
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001207pfn_t hva_to_pfn_atomic(struct kvm *kvm, unsigned long addr)
1208{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001209 return hva_to_pfn(kvm, addr, true, NULL, true, NULL);
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001210}
1211EXPORT_SYMBOL_GPL(hva_to_pfn_atomic);
1212
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001213static pfn_t __gfn_to_pfn(struct kvm *kvm, gfn_t gfn, bool atomic, bool *async,
1214 bool write_fault, bool *writable)
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001215{
1216 unsigned long addr;
1217
Gleb Natapovaf585b92010-10-14 11:22:46 +02001218 if (async)
1219 *async = false;
1220
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001221 addr = gfn_to_hva(kvm, gfn);
1222 if (kvm_is_error_hva(addr)) {
1223 get_page(bad_page);
1224 return page_to_pfn(bad_page);
1225 }
1226
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001227 return hva_to_pfn(kvm, addr, atomic, async, write_fault, writable);
Xiao Guangrong365fb3f2010-08-28 19:24:13 +08001228}
1229
1230pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn)
1231{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001232 return __gfn_to_pfn(kvm, gfn, true, NULL, true, NULL);
Xiao Guangrong365fb3f2010-08-28 19:24:13 +08001233}
1234EXPORT_SYMBOL_GPL(gfn_to_pfn_atomic);
1235
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001236pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
1237 bool write_fault, bool *writable)
Gleb Natapovaf585b92010-10-14 11:22:46 +02001238{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001239 return __gfn_to_pfn(kvm, gfn, false, async, write_fault, writable);
Gleb Natapovaf585b92010-10-14 11:22:46 +02001240}
1241EXPORT_SYMBOL_GPL(gfn_to_pfn_async);
1242
Xiao Guangrong365fb3f2010-08-28 19:24:13 +08001243pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
1244{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001245 return __gfn_to_pfn(kvm, gfn, false, NULL, true, NULL);
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001246}
Anthony Liguori35149e22008-04-02 14:46:56 -05001247EXPORT_SYMBOL_GPL(gfn_to_pfn);
1248
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001249pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1250 bool *writable)
1251{
1252 return __gfn_to_pfn(kvm, gfn, false, NULL, write_fault, writable);
1253}
1254EXPORT_SYMBOL_GPL(gfn_to_pfn_prot);
1255
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001256pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
1257 struct kvm_memory_slot *slot, gfn_t gfn)
1258{
1259 unsigned long addr = gfn_to_hva_memslot(slot, gfn);
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001260 return hva_to_pfn(kvm, addr, false, NULL, true, NULL);
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001261}
1262
Xiao Guangrong48987782010-08-22 19:11:43 +08001263int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
1264 int nr_pages)
1265{
1266 unsigned long addr;
1267 gfn_t entry;
1268
Gleb Natapov49c77542010-10-18 15:22:23 +02001269 addr = gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, &entry);
Xiao Guangrong48987782010-08-22 19:11:43 +08001270 if (kvm_is_error_hva(addr))
1271 return -1;
1272
1273 if (entry < nr_pages)
1274 return 0;
1275
1276 return __get_user_pages_fast(addr, nr_pages, 1, pages);
1277}
1278EXPORT_SYMBOL_GPL(gfn_to_page_many_atomic);
1279
Anthony Liguori35149e22008-04-02 14:46:56 -05001280struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
1281{
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001282 pfn_t pfn;
1283
1284 pfn = gfn_to_pfn(kvm, gfn);
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001285 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001286 return pfn_to_page(pfn);
1287
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001288 WARN_ON(kvm_is_mmio_pfn(pfn));
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001289
1290 get_page(bad_page);
1291 return bad_page;
Avi Kivity954bbbc22007-03-30 14:02:32 +03001292}
Anthony Liguoriaab61cc2007-10-29 15:15:20 -05001293
Avi Kivity954bbbc22007-03-30 14:02:32 +03001294EXPORT_SYMBOL_GPL(gfn_to_page);
1295
Izik Eidusb4231d62007-11-20 11:49:33 +02001296void kvm_release_page_clean(struct page *page)
1297{
Anthony Liguori35149e22008-04-02 14:46:56 -05001298 kvm_release_pfn_clean(page_to_pfn(page));
Izik Eidusb4231d62007-11-20 11:49:33 +02001299}
1300EXPORT_SYMBOL_GPL(kvm_release_page_clean);
1301
Anthony Liguori35149e22008-04-02 14:46:56 -05001302void kvm_release_pfn_clean(pfn_t pfn)
1303{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001304 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001305 put_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001306}
1307EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);
1308
Izik Eidusb4231d62007-11-20 11:49:33 +02001309void kvm_release_page_dirty(struct page *page)
Izik Eidus8a7ae052007-10-18 11:09:33 +02001310{
Anthony Liguori35149e22008-04-02 14:46:56 -05001311 kvm_release_pfn_dirty(page_to_pfn(page));
Izik Eidus8a7ae052007-10-18 11:09:33 +02001312}
Izik Eidusb4231d62007-11-20 11:49:33 +02001313EXPORT_SYMBOL_GPL(kvm_release_page_dirty);
Izik Eidus8a7ae052007-10-18 11:09:33 +02001314
Anthony Liguori35149e22008-04-02 14:46:56 -05001315void kvm_release_pfn_dirty(pfn_t pfn)
1316{
1317 kvm_set_pfn_dirty(pfn);
1318 kvm_release_pfn_clean(pfn);
1319}
1320EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
1321
1322void kvm_set_page_dirty(struct page *page)
1323{
1324 kvm_set_pfn_dirty(page_to_pfn(page));
1325}
1326EXPORT_SYMBOL_GPL(kvm_set_page_dirty);
1327
1328void kvm_set_pfn_dirty(pfn_t pfn)
1329{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001330 if (!kvm_is_mmio_pfn(pfn)) {
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001331 struct page *page = pfn_to_page(pfn);
1332 if (!PageReserved(page))
1333 SetPageDirty(page);
1334 }
Anthony Liguori35149e22008-04-02 14:46:56 -05001335}
1336EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);
1337
1338void kvm_set_pfn_accessed(pfn_t pfn)
1339{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001340 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001341 mark_page_accessed(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001342}
1343EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);
1344
1345void kvm_get_pfn(pfn_t pfn)
1346{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001347 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001348 get_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001349}
1350EXPORT_SYMBOL_GPL(kvm_get_pfn);
1351
Izik Eidus195aefd2007-10-01 22:14:18 +02001352static int next_segment(unsigned long len, int offset)
1353{
1354 if (len > PAGE_SIZE - offset)
1355 return PAGE_SIZE - offset;
1356 else
1357 return len;
1358}
1359
1360int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
1361 int len)
1362{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001363 int r;
1364 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001365
Izik Eiduse0506bc2007-11-11 22:10:22 +02001366 addr = gfn_to_hva(kvm, gfn);
1367 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001368 return -EFAULT;
Takuya Yoshikawafa3d3152011-05-07 16:35:38 +09001369 r = __copy_from_user(data, (void __user *)addr + offset, len);
Izik Eiduse0506bc2007-11-11 22:10:22 +02001370 if (r)
1371 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001372 return 0;
1373}
1374EXPORT_SYMBOL_GPL(kvm_read_guest_page);
1375
1376int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len)
1377{
1378 gfn_t gfn = gpa >> PAGE_SHIFT;
1379 int seg;
1380 int offset = offset_in_page(gpa);
1381 int ret;
1382
1383 while ((seg = next_segment(len, offset)) != 0) {
1384 ret = kvm_read_guest_page(kvm, gfn, data, offset, seg);
1385 if (ret < 0)
1386 return ret;
1387 offset = 0;
1388 len -= seg;
1389 data += seg;
1390 ++gfn;
1391 }
1392 return 0;
1393}
1394EXPORT_SYMBOL_GPL(kvm_read_guest);
1395
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001396int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
1397 unsigned long len)
1398{
1399 int r;
1400 unsigned long addr;
1401 gfn_t gfn = gpa >> PAGE_SHIFT;
1402 int offset = offset_in_page(gpa);
1403
1404 addr = gfn_to_hva(kvm, gfn);
1405 if (kvm_is_error_hva(addr))
1406 return -EFAULT;
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001407 pagefault_disable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001408 r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001409 pagefault_enable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001410 if (r)
1411 return -EFAULT;
1412 return 0;
1413}
1414EXPORT_SYMBOL(kvm_read_guest_atomic);
1415
Izik Eidus195aefd2007-10-01 22:14:18 +02001416int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
1417 int offset, int len)
1418{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001419 int r;
1420 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001421
Izik Eiduse0506bc2007-11-11 22:10:22 +02001422 addr = gfn_to_hva(kvm, gfn);
1423 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001424 return -EFAULT;
Xiao Guangrong8b0cedf2011-05-15 23:22:04 +08001425 r = __copy_to_user((void __user *)addr + offset, data, len);
Izik Eiduse0506bc2007-11-11 22:10:22 +02001426 if (r)
1427 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001428 mark_page_dirty(kvm, gfn);
1429 return 0;
1430}
1431EXPORT_SYMBOL_GPL(kvm_write_guest_page);
1432
1433int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
1434 unsigned long len)
1435{
1436 gfn_t gfn = gpa >> PAGE_SHIFT;
1437 int seg;
1438 int offset = offset_in_page(gpa);
1439 int ret;
1440
1441 while ((seg = next_segment(len, offset)) != 0) {
1442 ret = kvm_write_guest_page(kvm, gfn, data, offset, seg);
1443 if (ret < 0)
1444 return ret;
1445 offset = 0;
1446 len -= seg;
1447 data += seg;
1448 ++gfn;
1449 }
1450 return 0;
1451}
1452
Gleb Natapov49c77542010-10-18 15:22:23 +02001453int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1454 gpa_t gpa)
1455{
1456 struct kvm_memslots *slots = kvm_memslots(kvm);
1457 int offset = offset_in_page(gpa);
1458 gfn_t gfn = gpa >> PAGE_SHIFT;
1459
1460 ghc->gpa = gpa;
1461 ghc->generation = slots->generation;
1462 ghc->memslot = __gfn_to_memslot(slots, gfn);
1463 ghc->hva = gfn_to_hva_many(ghc->memslot, gfn, NULL);
1464 if (!kvm_is_error_hva(ghc->hva))
1465 ghc->hva += offset;
1466 else
1467 return -EFAULT;
1468
1469 return 0;
1470}
1471EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init);
1472
1473int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1474 void *data, unsigned long len)
1475{
1476 struct kvm_memslots *slots = kvm_memslots(kvm);
1477 int r;
1478
1479 if (slots->generation != ghc->generation)
1480 kvm_gfn_to_hva_cache_init(kvm, ghc, ghc->gpa);
1481
1482 if (kvm_is_error_hva(ghc->hva))
1483 return -EFAULT;
1484
Xiao Guangrong8b0cedf2011-05-15 23:22:04 +08001485 r = __copy_to_user((void __user *)ghc->hva, data, len);
Gleb Natapov49c77542010-10-18 15:22:23 +02001486 if (r)
1487 return -EFAULT;
1488 mark_page_dirty_in_slot(kvm, ghc->memslot, ghc->gpa >> PAGE_SHIFT);
1489
1490 return 0;
1491}
1492EXPORT_SYMBOL_GPL(kvm_write_guest_cached);
1493
Gleb Natapove03b6442011-07-11 15:28:11 -04001494int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1495 void *data, unsigned long len)
1496{
1497 struct kvm_memslots *slots = kvm_memslots(kvm);
1498 int r;
1499
1500 if (slots->generation != ghc->generation)
1501 kvm_gfn_to_hva_cache_init(kvm, ghc, ghc->gpa);
1502
1503 if (kvm_is_error_hva(ghc->hva))
1504 return -EFAULT;
1505
1506 r = __copy_from_user(data, (void __user *)ghc->hva, len);
1507 if (r)
1508 return -EFAULT;
1509
1510 return 0;
1511}
1512EXPORT_SYMBOL_GPL(kvm_read_guest_cached);
1513
Izik Eidus195aefd2007-10-01 22:14:18 +02001514int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
1515{
Heiko Carstens3bcc8a82010-10-27 17:21:21 +02001516 return kvm_write_guest_page(kvm, gfn, (const void *) empty_zero_page,
1517 offset, len);
Izik Eidus195aefd2007-10-01 22:14:18 +02001518}
1519EXPORT_SYMBOL_GPL(kvm_clear_guest_page);
1520
1521int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len)
1522{
1523 gfn_t gfn = gpa >> PAGE_SHIFT;
1524 int seg;
1525 int offset = offset_in_page(gpa);
1526 int ret;
1527
1528 while ((seg = next_segment(len, offset)) != 0) {
1529 ret = kvm_clear_guest_page(kvm, gfn, offset, seg);
1530 if (ret < 0)
1531 return ret;
1532 offset = 0;
1533 len -= seg;
1534 ++gfn;
1535 }
1536 return 0;
1537}
1538EXPORT_SYMBOL_GPL(kvm_clear_guest);
1539
Gleb Natapov49c77542010-10-18 15:22:23 +02001540void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
1541 gfn_t gfn)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001542{
Rusty Russell7e9d6192007-07-31 20:41:14 +10001543 if (memslot && memslot->dirty_bitmap) {
1544 unsigned long rel_gfn = gfn - memslot->base_gfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001545
Takuya Yoshikawa50e92b32012-01-04 15:06:43 +09001546 if (!test_and_set_bit_le(rel_gfn, memslot->dirty_bitmap))
Takuya Yoshikawa7850ac52011-11-14 18:23:34 +09001547 memslot->nr_dirty_pages++;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001548 }
1549}
1550
Gleb Natapov49c77542010-10-18 15:22:23 +02001551void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
1552{
1553 struct kvm_memory_slot *memslot;
1554
1555 memslot = gfn_to_memslot(kvm, gfn);
1556 mark_page_dirty_in_slot(kvm, memslot, gfn);
1557}
1558
Eddie Dongb6958ce2007-07-18 12:15:21 +03001559/*
1560 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
1561 */
Hollis Blanchard8776e512007-10-31 17:24:24 -05001562void kvm_vcpu_block(struct kvm_vcpu *vcpu)
Eddie Dongb6958ce2007-07-18 12:15:21 +03001563{
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001564 DEFINE_WAIT(wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001565
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001566 for (;;) {
1567 prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001568
Gleb Natapova1b37102009-07-09 15:33:52 +03001569 if (kvm_arch_vcpu_runnable(vcpu)) {
Avi Kivitya8eeb042010-05-10 12:34:53 +03001570 kvm_make_request(KVM_REQ_UNHALT, vcpu);
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001571 break;
Marcelo Tosattid7690172008-09-08 15:23:48 -03001572 }
Gleb Natapov09cec752009-03-23 15:11:44 +02001573 if (kvm_cpu_has_pending_timer(vcpu))
1574 break;
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001575 if (signal_pending(current))
1576 break;
1577
Eddie Dongb6958ce2007-07-18 12:15:21 +03001578 schedule();
Eddie Dongb6958ce2007-07-18 12:15:21 +03001579 }
1580
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001581 finish_wait(&vcpu->wq, &wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001582}
1583
Avi Kivity6aa8b732006-12-10 02:21:36 -08001584void kvm_resched(struct kvm_vcpu *vcpu)
1585{
Yaozu Dong3fca0362007-04-25 16:49:19 +03001586 if (!need_resched())
1587 return;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001588 cond_resched();
Avi Kivity6aa8b732006-12-10 02:21:36 -08001589}
1590EXPORT_SYMBOL_GPL(kvm_resched);
1591
Rik van Riel217ece62011-02-01 09:53:28 -05001592void kvm_vcpu_on_spin(struct kvm_vcpu *me)
Zhai, Edwind255f4f2009-10-09 18:03:20 +08001593{
Rik van Riel217ece62011-02-01 09:53:28 -05001594 struct kvm *kvm = me->kvm;
1595 struct kvm_vcpu *vcpu;
1596 int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
1597 int yielded = 0;
1598 int pass;
1599 int i;
Zhai, Edwind255f4f2009-10-09 18:03:20 +08001600
Rik van Riel217ece62011-02-01 09:53:28 -05001601 /*
1602 * We boost the priority of a VCPU that is runnable but not
1603 * currently running, because it got preempted by something
1604 * else and called schedule in __vcpu_run. Hopefully that
1605 * VCPU is holding the lock that we need and will release it.
1606 * We approximate round-robin by starting at the last boosted VCPU.
1607 */
1608 for (pass = 0; pass < 2 && !yielded; pass++) {
1609 kvm_for_each_vcpu(i, vcpu, kvm) {
1610 struct task_struct *task = NULL;
1611 struct pid *pid;
1612 if (!pass && i < last_boosted_vcpu) {
1613 i = last_boosted_vcpu;
1614 continue;
1615 } else if (pass && i > last_boosted_vcpu)
1616 break;
1617 if (vcpu == me)
1618 continue;
1619 if (waitqueue_active(&vcpu->wq))
1620 continue;
1621 rcu_read_lock();
1622 pid = rcu_dereference(vcpu->pid);
1623 if (pid)
1624 task = get_pid_task(vcpu->pid, PIDTYPE_PID);
1625 rcu_read_unlock();
1626 if (!task)
1627 continue;
1628 if (task->flags & PF_VCPU) {
1629 put_task_struct(task);
1630 continue;
1631 }
1632 if (yield_to(task, 1)) {
1633 put_task_struct(task);
1634 kvm->last_boosted_vcpu = i;
1635 yielded = 1;
1636 break;
1637 }
1638 put_task_struct(task);
1639 }
1640 }
Zhai, Edwind255f4f2009-10-09 18:03:20 +08001641}
1642EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin);
1643
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001644static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001645{
1646 struct kvm_vcpu *vcpu = vma->vm_file->private_data;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001647 struct page *page;
1648
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001649 if (vmf->pgoff == 0)
Avi Kivity039576c2007-03-20 12:46:50 +02001650 page = virt_to_page(vcpu->run);
Avi Kivity09566762008-01-23 18:14:23 +02001651#ifdef CONFIG_X86
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001652 else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
Zhang Xiantaoad312c72007-12-13 23:50:52 +08001653 page = virt_to_page(vcpu->arch.pio_data);
Avi Kivity09566762008-01-23 18:14:23 +02001654#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02001655#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1656 else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
1657 page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
1658#endif
Avi Kivity039576c2007-03-20 12:46:50 +02001659 else
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001660 return VM_FAULT_SIGBUS;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001661 get_page(page);
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001662 vmf->page = page;
1663 return 0;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001664}
1665
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +04001666static const struct vm_operations_struct kvm_vcpu_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001667 .fault = kvm_vcpu_fault,
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001668};
1669
1670static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma)
1671{
1672 vma->vm_ops = &kvm_vcpu_vm_ops;
1673 return 0;
1674}
1675
Avi Kivitybccf2152007-02-21 18:04:26 +02001676static int kvm_vcpu_release(struct inode *inode, struct file *filp)
1677{
1678 struct kvm_vcpu *vcpu = filp->private_data;
1679
Al Viro66c0b392008-04-19 20:33:56 +01001680 kvm_put_kvm(vcpu->kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001681 return 0;
1682}
1683
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01001684static struct file_operations kvm_vcpu_fops = {
Avi Kivitybccf2152007-02-21 18:04:26 +02001685 .release = kvm_vcpu_release,
1686 .unlocked_ioctl = kvm_vcpu_ioctl,
Alexander Graf1dda6062011-06-08 02:45:37 +02001687#ifdef CONFIG_COMPAT
1688 .compat_ioctl = kvm_vcpu_compat_ioctl,
1689#endif
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001690 .mmap = kvm_vcpu_mmap,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001691 .llseek = noop_llseek,
Avi Kivitybccf2152007-02-21 18:04:26 +02001692};
1693
1694/*
1695 * Allocates an inode for the vcpu.
1696 */
1697static int create_vcpu_fd(struct kvm_vcpu *vcpu)
1698{
Roland Dreier628ff7c2009-12-18 09:41:24 -08001699 return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, O_RDWR);
Avi Kivitybccf2152007-02-21 18:04:26 +02001700}
1701
Avi Kivityc5ea7662007-02-20 18:41:05 +02001702/*
1703 * Creates some virtual cpus. Good luck creating more than one.
1704 */
Gleb Natapov73880c82009-06-09 15:56:28 +03001705static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
Avi Kivityc5ea7662007-02-20 18:41:05 +02001706{
1707 int r;
Gleb Natapov988a2ca2009-06-09 15:56:29 +03001708 struct kvm_vcpu *vcpu, *v;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001709
Gleb Natapov73880c82009-06-09 15:56:28 +03001710 vcpu = kvm_arch_vcpu_create(kvm, id);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001711 if (IS_ERR(vcpu))
1712 return PTR_ERR(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001713
Avi Kivity15ad7142007-07-11 18:17:21 +03001714 preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);
1715
Avi Kivity26e52152007-11-20 15:30:24 +02001716 r = kvm_arch_vcpu_setup(vcpu);
1717 if (r)
Jan Kiszkad7805922011-05-23 10:33:05 +02001718 goto vcpu_destroy;
Avi Kivity26e52152007-11-20 15:30:24 +02001719
Shaohua Li11ec2802007-07-23 14:51:37 +08001720 mutex_lock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03001721 if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) {
1722 r = -EINVAL;
Jan Kiszkad7805922011-05-23 10:33:05 +02001723 goto unlock_vcpu_destroy;
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001724 }
Gleb Natapov73880c82009-06-09 15:56:28 +03001725
Gleb Natapov988a2ca2009-06-09 15:56:29 +03001726 kvm_for_each_vcpu(r, v, kvm)
1727 if (v->vcpu_id == id) {
Gleb Natapov73880c82009-06-09 15:56:28 +03001728 r = -EEXIST;
Jan Kiszkad7805922011-05-23 10:33:05 +02001729 goto unlock_vcpu_destroy;
Gleb Natapov73880c82009-06-09 15:56:28 +03001730 }
1731
1732 BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001733
1734 /* Now it's all set up, let userspace reach it */
Al Viro66c0b392008-04-19 20:33:56 +01001735 kvm_get_kvm(kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001736 r = create_vcpu_fd(vcpu);
Gleb Natapov73880c82009-06-09 15:56:28 +03001737 if (r < 0) {
1738 kvm_put_kvm(kvm);
Jan Kiszkad7805922011-05-23 10:33:05 +02001739 goto unlock_vcpu_destroy;
Gleb Natapov73880c82009-06-09 15:56:28 +03001740 }
1741
1742 kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu;
1743 smp_wmb();
1744 atomic_inc(&kvm->online_vcpus);
1745
Gleb Natapov73880c82009-06-09 15:56:28 +03001746 mutex_unlock(&kvm->lock);
Avi Kivitybccf2152007-02-21 18:04:26 +02001747 return r;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001748
Jan Kiszkad7805922011-05-23 10:33:05 +02001749unlock_vcpu_destroy:
Glauber Costa7d8fece2008-09-17 23:16:59 -03001750 mutex_unlock(&kvm->lock);
Jan Kiszkad7805922011-05-23 10:33:05 +02001751vcpu_destroy:
Hollis Blanchardd40ccc62007-11-19 14:04:43 -06001752 kvm_arch_vcpu_destroy(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001753 return r;
1754}
1755
Avi Kivity1961d272007-03-05 19:46:05 +02001756static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset)
1757{
1758 if (sigset) {
1759 sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP));
1760 vcpu->sigset_active = 1;
1761 vcpu->sigset = *sigset;
1762 } else
1763 vcpu->sigset_active = 0;
1764 return 0;
1765}
1766
Avi Kivitybccf2152007-02-21 18:04:26 +02001767static long kvm_vcpu_ioctl(struct file *filp,
1768 unsigned int ioctl, unsigned long arg)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001769{
Avi Kivitybccf2152007-02-21 18:04:26 +02001770 struct kvm_vcpu *vcpu = filp->private_data;
Al Viro2f366982007-02-09 16:38:35 +00001771 void __user *argp = (void __user *)arg;
Carsten Otte313a3dc2007-10-11 19:16:52 +02001772 int r;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001773 struct kvm_fpu *fpu = NULL;
1774 struct kvm_sregs *kvm_sregs = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001775
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001776 if (vcpu->kvm->mm != current->mm)
1777 return -EIO;
Avi Kivity2122ff52010-05-13 11:25:04 +03001778
1779#if defined(CONFIG_S390) || defined(CONFIG_PPC)
1780 /*
1781 * Special cases: vcpu ioctls that are asynchronous to vcpu execution,
1782 * so vcpu_load() would break it.
1783 */
1784 if (ioctl == KVM_S390_INTERRUPT || ioctl == KVM_INTERRUPT)
1785 return kvm_arch_vcpu_ioctl(filp, ioctl, arg);
1786#endif
1787
1788
1789 vcpu_load(vcpu);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001790 switch (ioctl) {
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001791 case KVM_RUN:
Avi Kivityf0fe5102007-03-07 13:11:17 +02001792 r = -EINVAL;
1793 if (arg)
1794 goto out;
Hollis Blanchardb6c7a5d2007-11-01 14:16:10 -05001795 r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
Gleb Natapov64be5002010-10-24 16:49:08 +02001796 trace_kvm_userspace_exit(vcpu->run->exit_reason, r);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001797 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001798 case KVM_GET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001799 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001800
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001801 r = -ENOMEM;
1802 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
1803 if (!kvm_regs)
1804 goto out;
1805 r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001806 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001807 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001808 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001809 if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
1810 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001811 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001812out_free1:
1813 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001814 break;
1815 }
1816 case KVM_SET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001817 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001818
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001819 r = -ENOMEM;
Sasha Levinff5c2c02011-12-04 19:36:29 +02001820 kvm_regs = memdup_user(argp, sizeof(*kvm_regs));
1821 if (IS_ERR(kvm_regs)) {
1822 r = PTR_ERR(kvm_regs);
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001823 goto out;
Sasha Levinff5c2c02011-12-04 19:36:29 +02001824 }
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001825 r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001826 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001827 goto out_free2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001828 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001829out_free2:
1830 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001831 break;
1832 }
1833 case KVM_GET_SREGS: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001834 kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL);
1835 r = -ENOMEM;
1836 if (!kvm_sregs)
1837 goto out;
1838 r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001839 if (r)
1840 goto out;
1841 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001842 if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001843 goto out;
1844 r = 0;
1845 break;
1846 }
1847 case KVM_SET_SREGS: {
Sasha Levinff5c2c02011-12-04 19:36:29 +02001848 kvm_sregs = memdup_user(argp, sizeof(*kvm_sregs));
1849 if (IS_ERR(kvm_sregs)) {
1850 r = PTR_ERR(kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001851 goto out;
Sasha Levinff5c2c02011-12-04 19:36:29 +02001852 }
Dave Hansenfa3795a2008-08-11 10:01:46 -07001853 r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001854 if (r)
1855 goto out;
1856 r = 0;
1857 break;
1858 }
Marcelo Tosatti62d9f0d2008-04-11 13:24:45 -03001859 case KVM_GET_MP_STATE: {
1860 struct kvm_mp_state mp_state;
1861
1862 r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state);
1863 if (r)
1864 goto out;
1865 r = -EFAULT;
1866 if (copy_to_user(argp, &mp_state, sizeof mp_state))
1867 goto out;
1868 r = 0;
1869 break;
1870 }
1871 case KVM_SET_MP_STATE: {
1872 struct kvm_mp_state mp_state;
1873
1874 r = -EFAULT;
1875 if (copy_from_user(&mp_state, argp, sizeof mp_state))
1876 goto out;
1877 r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
1878 if (r)
1879 goto out;
1880 r = 0;
1881 break;
1882 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08001883 case KVM_TRANSLATE: {
1884 struct kvm_translation tr;
1885
1886 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001887 if (copy_from_user(&tr, argp, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001888 goto out;
Zhang Xiantao8b006792007-11-16 13:05:55 +08001889 r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001890 if (r)
1891 goto out;
1892 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001893 if (copy_to_user(argp, &tr, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001894 goto out;
1895 r = 0;
1896 break;
1897 }
Jan Kiszkad0bfb942008-12-15 13:52:10 +01001898 case KVM_SET_GUEST_DEBUG: {
1899 struct kvm_guest_debug dbg;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001900
1901 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001902 if (copy_from_user(&dbg, argp, sizeof dbg))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001903 goto out;
Jan Kiszkad0bfb942008-12-15 13:52:10 +01001904 r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001905 if (r)
1906 goto out;
1907 r = 0;
1908 break;
1909 }
Avi Kivity1961d272007-03-05 19:46:05 +02001910 case KVM_SET_SIGNAL_MASK: {
1911 struct kvm_signal_mask __user *sigmask_arg = argp;
1912 struct kvm_signal_mask kvm_sigmask;
1913 sigset_t sigset, *p;
1914
1915 p = NULL;
1916 if (argp) {
1917 r = -EFAULT;
1918 if (copy_from_user(&kvm_sigmask, argp,
1919 sizeof kvm_sigmask))
1920 goto out;
1921 r = -EINVAL;
1922 if (kvm_sigmask.len != sizeof sigset)
1923 goto out;
1924 r = -EFAULT;
1925 if (copy_from_user(&sigset, sigmask_arg->sigset,
1926 sizeof sigset))
1927 goto out;
1928 p = &sigset;
1929 }
Andi Kleen376d41f2010-06-10 13:10:47 +02001930 r = kvm_vcpu_ioctl_set_sigmask(vcpu, p);
Avi Kivity1961d272007-03-05 19:46:05 +02001931 break;
1932 }
Avi Kivityb8836732007-04-01 16:34:31 +03001933 case KVM_GET_FPU: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001934 fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL);
1935 r = -ENOMEM;
1936 if (!fpu)
1937 goto out;
1938 r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001939 if (r)
1940 goto out;
1941 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001942 if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
Avi Kivityb8836732007-04-01 16:34:31 +03001943 goto out;
1944 r = 0;
1945 break;
1946 }
1947 case KVM_SET_FPU: {
Sasha Levinff5c2c02011-12-04 19:36:29 +02001948 fpu = memdup_user(argp, sizeof(*fpu));
1949 if (IS_ERR(fpu)) {
1950 r = PTR_ERR(fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001951 goto out;
Sasha Levinff5c2c02011-12-04 19:36:29 +02001952 }
Dave Hansenfa3795a2008-08-11 10:01:46 -07001953 r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001954 if (r)
1955 goto out;
1956 r = 0;
1957 break;
1958 }
Avi Kivitybccf2152007-02-21 18:04:26 +02001959 default:
Carsten Otte313a3dc2007-10-11 19:16:52 +02001960 r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
Avi Kivitybccf2152007-02-21 18:04:26 +02001961 }
1962out:
Avi Kivity2122ff52010-05-13 11:25:04 +03001963 vcpu_put(vcpu);
Dave Hansenfa3795a2008-08-11 10:01:46 -07001964 kfree(fpu);
1965 kfree(kvm_sregs);
Avi Kivitybccf2152007-02-21 18:04:26 +02001966 return r;
1967}
1968
Alexander Graf1dda6062011-06-08 02:45:37 +02001969#ifdef CONFIG_COMPAT
1970static long kvm_vcpu_compat_ioctl(struct file *filp,
1971 unsigned int ioctl, unsigned long arg)
1972{
1973 struct kvm_vcpu *vcpu = filp->private_data;
1974 void __user *argp = compat_ptr(arg);
1975 int r;
1976
1977 if (vcpu->kvm->mm != current->mm)
1978 return -EIO;
1979
1980 switch (ioctl) {
1981 case KVM_SET_SIGNAL_MASK: {
1982 struct kvm_signal_mask __user *sigmask_arg = argp;
1983 struct kvm_signal_mask kvm_sigmask;
1984 compat_sigset_t csigset;
1985 sigset_t sigset;
1986
1987 if (argp) {
1988 r = -EFAULT;
1989 if (copy_from_user(&kvm_sigmask, argp,
1990 sizeof kvm_sigmask))
1991 goto out;
1992 r = -EINVAL;
1993 if (kvm_sigmask.len != sizeof csigset)
1994 goto out;
1995 r = -EFAULT;
1996 if (copy_from_user(&csigset, sigmask_arg->sigset,
1997 sizeof csigset))
1998 goto out;
1999 }
2000 sigset_from_compat(&sigset, &csigset);
2001 r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
2002 break;
2003 }
2004 default:
2005 r = kvm_vcpu_ioctl(filp, ioctl, arg);
2006 }
2007
2008out:
2009 return r;
2010}
2011#endif
2012
Avi Kivitybccf2152007-02-21 18:04:26 +02002013static long kvm_vm_ioctl(struct file *filp,
2014 unsigned int ioctl, unsigned long arg)
2015{
2016 struct kvm *kvm = filp->private_data;
2017 void __user *argp = (void __user *)arg;
Carsten Otte1fe779f2007-10-29 16:08:35 +01002018 int r;
Avi Kivitybccf2152007-02-21 18:04:26 +02002019
Avi Kivity6d4e4c42007-11-21 16:41:05 +02002020 if (kvm->mm != current->mm)
2021 return -EIO;
Avi Kivitybccf2152007-02-21 18:04:26 +02002022 switch (ioctl) {
2023 case KVM_CREATE_VCPU:
2024 r = kvm_vm_ioctl_create_vcpu(kvm, arg);
2025 if (r < 0)
2026 goto out;
2027 break;
Izik Eidus6fc138d2007-10-09 19:20:39 +02002028 case KVM_SET_USER_MEMORY_REGION: {
2029 struct kvm_userspace_memory_region kvm_userspace_mem;
2030
2031 r = -EFAULT;
2032 if (copy_from_user(&kvm_userspace_mem, argp,
2033 sizeof kvm_userspace_mem))
2034 goto out;
2035
2036 r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002037 if (r)
2038 goto out;
2039 break;
2040 }
2041 case KVM_GET_DIRTY_LOG: {
2042 struct kvm_dirty_log log;
2043
2044 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00002045 if (copy_from_user(&log, argp, sizeof log))
Avi Kivity6aa8b732006-12-10 02:21:36 -08002046 goto out;
Avi Kivity2c6f5df2007-02-20 18:27:58 +02002047 r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002048 if (r)
2049 goto out;
2050 break;
2051 }
Laurent Vivier5f94c172008-05-30 16:05:54 +02002052#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
2053 case KVM_REGISTER_COALESCED_MMIO: {
2054 struct kvm_coalesced_mmio_zone zone;
2055 r = -EFAULT;
2056 if (copy_from_user(&zone, argp, sizeof zone))
2057 goto out;
Laurent Vivier5f94c172008-05-30 16:05:54 +02002058 r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
2059 if (r)
2060 goto out;
2061 r = 0;
2062 break;
2063 }
2064 case KVM_UNREGISTER_COALESCED_MMIO: {
2065 struct kvm_coalesced_mmio_zone zone;
2066 r = -EFAULT;
2067 if (copy_from_user(&zone, argp, sizeof zone))
2068 goto out;
Laurent Vivier5f94c172008-05-30 16:05:54 +02002069 r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
2070 if (r)
2071 goto out;
2072 r = 0;
2073 break;
2074 }
2075#endif
Gregory Haskins721eecb2009-05-20 10:30:49 -04002076 case KVM_IRQFD: {
2077 struct kvm_irqfd data;
2078
2079 r = -EFAULT;
2080 if (copy_from_user(&data, argp, sizeof data))
2081 goto out;
2082 r = kvm_irqfd(kvm, data.fd, data.gsi, data.flags);
2083 break;
2084 }
Gregory Haskinsd34e6b12009-07-07 17:08:49 -04002085 case KVM_IOEVENTFD: {
2086 struct kvm_ioeventfd data;
2087
2088 r = -EFAULT;
2089 if (copy_from_user(&data, argp, sizeof data))
2090 goto out;
2091 r = kvm_ioeventfd(kvm, &data);
2092 break;
2093 }
Gleb Natapov73880c82009-06-09 15:56:28 +03002094#ifdef CONFIG_KVM_APIC_ARCHITECTURE
2095 case KVM_SET_BOOT_CPU_ID:
2096 r = 0;
Marcelo Tosatti894a9c52009-06-23 15:05:14 -03002097 mutex_lock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03002098 if (atomic_read(&kvm->online_vcpus) != 0)
2099 r = -EBUSY;
2100 else
2101 kvm->bsp_vcpu_id = arg;
Marcelo Tosatti894a9c52009-06-23 15:05:14 -03002102 mutex_unlock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03002103 break;
2104#endif
Avi Kivityf17abe92007-02-21 19:28:04 +02002105 default:
Carsten Otte1fe779f2007-10-29 16:08:35 +01002106 r = kvm_arch_vm_ioctl(filp, ioctl, arg);
Avi Kivitybfd99ff2009-08-26 14:57:50 +03002107 if (r == -ENOTTY)
2108 r = kvm_vm_ioctl_assigned_device(kvm, ioctl, arg);
Avi Kivityf17abe92007-02-21 19:28:04 +02002109 }
2110out:
2111 return r;
2112}
2113
Arnd Bergmann6ff58942009-10-22 14:19:27 +02002114#ifdef CONFIG_COMPAT
2115struct compat_kvm_dirty_log {
2116 __u32 slot;
2117 __u32 padding1;
2118 union {
2119 compat_uptr_t dirty_bitmap; /* one bit per page */
2120 __u64 padding2;
2121 };
2122};
2123
2124static long kvm_vm_compat_ioctl(struct file *filp,
2125 unsigned int ioctl, unsigned long arg)
2126{
2127 struct kvm *kvm = filp->private_data;
2128 int r;
2129
2130 if (kvm->mm != current->mm)
2131 return -EIO;
2132 switch (ioctl) {
2133 case KVM_GET_DIRTY_LOG: {
2134 struct compat_kvm_dirty_log compat_log;
2135 struct kvm_dirty_log log;
2136
2137 r = -EFAULT;
2138 if (copy_from_user(&compat_log, (void __user *)arg,
2139 sizeof(compat_log)))
2140 goto out;
2141 log.slot = compat_log.slot;
2142 log.padding1 = compat_log.padding1;
2143 log.padding2 = compat_log.padding2;
2144 log.dirty_bitmap = compat_ptr(compat_log.dirty_bitmap);
2145
2146 r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
2147 if (r)
2148 goto out;
2149 break;
2150 }
2151 default:
2152 r = kvm_vm_ioctl(filp, ioctl, arg);
2153 }
2154
2155out:
2156 return r;
2157}
2158#endif
2159
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002160static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivityf17abe92007-02-21 19:28:04 +02002161{
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002162 struct page *page[1];
2163 unsigned long addr;
2164 int npages;
2165 gfn_t gfn = vmf->pgoff;
Avi Kivityf17abe92007-02-21 19:28:04 +02002166 struct kvm *kvm = vma->vm_file->private_data;
Avi Kivityf17abe92007-02-21 19:28:04 +02002167
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002168 addr = gfn_to_hva(kvm, gfn);
2169 if (kvm_is_error_hva(addr))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002170 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002171
2172 npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page,
2173 NULL);
2174 if (unlikely(npages != 1))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002175 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002176
2177 vmf->page = page[0];
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002178 return 0;
Avi Kivityf17abe92007-02-21 19:28:04 +02002179}
2180
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +04002181static const struct vm_operations_struct kvm_vm_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002182 .fault = kvm_vm_fault,
Avi Kivityf17abe92007-02-21 19:28:04 +02002183};
2184
2185static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma)
2186{
2187 vma->vm_ops = &kvm_vm_vm_ops;
2188 return 0;
2189}
2190
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01002191static struct file_operations kvm_vm_fops = {
Avi Kivityf17abe92007-02-21 19:28:04 +02002192 .release = kvm_vm_release,
2193 .unlocked_ioctl = kvm_vm_ioctl,
Arnd Bergmann6ff58942009-10-22 14:19:27 +02002194#ifdef CONFIG_COMPAT
2195 .compat_ioctl = kvm_vm_compat_ioctl,
2196#endif
Avi Kivityf17abe92007-02-21 19:28:04 +02002197 .mmap = kvm_vm_mmap,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002198 .llseek = noop_llseek,
Avi Kivityf17abe92007-02-21 19:28:04 +02002199};
2200
2201static int kvm_dev_ioctl_create_vm(void)
2202{
Heiko Carstensaac87632010-10-27 17:22:10 +02002203 int r;
Avi Kivityf17abe92007-02-21 19:28:04 +02002204 struct kvm *kvm;
2205
Avi Kivityf17abe92007-02-21 19:28:04 +02002206 kvm = kvm_create_vm();
Avi Kivityd6d28162007-06-28 08:38:16 -04002207 if (IS_ERR(kvm))
2208 return PTR_ERR(kvm);
Takuya Yoshikawa6ce5a092010-03-15 22:13:30 +09002209#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
2210 r = kvm_coalesced_mmio_init(kvm);
2211 if (r < 0) {
2212 kvm_put_kvm(kvm);
2213 return r;
2214 }
2215#endif
Heiko Carstensaac87632010-10-27 17:22:10 +02002216 r = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, O_RDWR);
2217 if (r < 0)
Al Viro66c0b392008-04-19 20:33:56 +01002218 kvm_put_kvm(kvm);
Avi Kivityf17abe92007-02-21 19:28:04 +02002219
Heiko Carstensaac87632010-10-27 17:22:10 +02002220 return r;
Avi Kivityf17abe92007-02-21 19:28:04 +02002221}
2222
Avi Kivity1a811b62008-12-08 18:25:27 +02002223static long kvm_dev_ioctl_check_extension_generic(long arg)
2224{
2225 switch (arg) {
Avi Kivityca9edae2008-12-08 18:29:29 +02002226 case KVM_CAP_USER_MEMORY:
Avi Kivity1a811b62008-12-08 18:25:27 +02002227 case KVM_CAP_DESTROY_MEMORY_REGION_WORKS:
Jan Kiszka4cd481f2009-04-13 11:59:32 +02002228 case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS:
Gleb Natapov73880c82009-06-09 15:56:28 +03002229#ifdef CONFIG_KVM_APIC_ARCHITECTURE
2230 case KVM_CAP_SET_BOOT_CPU_ID:
2231#endif
Avi Kivitya9c73992009-11-04 11:54:59 +02002232 case KVM_CAP_INTERNAL_ERROR_DATA:
Avi Kivity1a811b62008-12-08 18:25:27 +02002233 return 1;
Avi Kivity399ec802008-11-19 13:58:46 +02002234#ifdef CONFIG_HAVE_KVM_IRQCHIP
2235 case KVM_CAP_IRQ_ROUTING:
Sheng Yang36463142009-03-16 16:33:43 +08002236 return KVM_MAX_IRQ_ROUTES;
Avi Kivity399ec802008-11-19 13:58:46 +02002237#endif
Avi Kivity1a811b62008-12-08 18:25:27 +02002238 default:
2239 break;
2240 }
2241 return kvm_dev_ioctl_check_extension(arg);
2242}
2243
Avi Kivityf17abe92007-02-21 19:28:04 +02002244static long kvm_dev_ioctl(struct file *filp,
2245 unsigned int ioctl, unsigned long arg)
2246{
Avi Kivity07c45a32007-03-07 13:05:38 +02002247 long r = -EINVAL;
Avi Kivityf17abe92007-02-21 19:28:04 +02002248
2249 switch (ioctl) {
2250 case KVM_GET_API_VERSION:
Avi Kivityf0fe5102007-03-07 13:11:17 +02002251 r = -EINVAL;
2252 if (arg)
2253 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02002254 r = KVM_API_VERSION;
2255 break;
2256 case KVM_CREATE_VM:
Avi Kivityf0fe5102007-03-07 13:11:17 +02002257 r = -EINVAL;
2258 if (arg)
2259 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02002260 r = kvm_dev_ioctl_create_vm();
2261 break;
Zhang Xiantao018d00d2007-11-15 23:07:47 +08002262 case KVM_CHECK_EXTENSION:
Avi Kivity1a811b62008-12-08 18:25:27 +02002263 r = kvm_dev_ioctl_check_extension_generic(arg);
Avi Kivity5d308f42007-03-01 17:56:20 +02002264 break;
Avi Kivity07c45a32007-03-07 13:05:38 +02002265 case KVM_GET_VCPU_MMAP_SIZE:
2266 r = -EINVAL;
2267 if (arg)
2268 goto out;
Avi Kivityadb1ff42008-01-24 15:13:08 +02002269 r = PAGE_SIZE; /* struct kvm_run */
2270#ifdef CONFIG_X86
2271 r += PAGE_SIZE; /* pio data page */
2272#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02002273#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
2274 r += PAGE_SIZE; /* coalesced mmio ring page */
2275#endif
Avi Kivity07c45a32007-03-07 13:05:38 +02002276 break;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04002277 case KVM_TRACE_ENABLE:
2278 case KVM_TRACE_PAUSE:
2279 case KVM_TRACE_DISABLE:
Marcelo Tosatti2023a292009-06-18 11:47:28 -03002280 r = -EOPNOTSUPP;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04002281 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002282 default:
Carsten Otte043405e2007-10-10 17:16:19 +02002283 return kvm_arch_dev_ioctl(filp, ioctl, arg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002284 }
2285out:
2286 return r;
2287}
2288
Avi Kivity6aa8b732006-12-10 02:21:36 -08002289static struct file_operations kvm_chardev_ops = {
Avi Kivity6aa8b732006-12-10 02:21:36 -08002290 .unlocked_ioctl = kvm_dev_ioctl,
2291 .compat_ioctl = kvm_dev_ioctl,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002292 .llseek = noop_llseek,
Avi Kivity6aa8b732006-12-10 02:21:36 -08002293};
2294
2295static struct miscdevice kvm_dev = {
Avi Kivitybbe44322007-03-04 13:27:36 +02002296 KVM_MINOR,
Avi Kivity6aa8b732006-12-10 02:21:36 -08002297 "kvm",
2298 &kvm_chardev_ops,
2299};
2300
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002301static void hardware_enable_nolock(void *junk)
Avi Kivity1b6c0162007-05-24 13:03:52 +03002302{
2303 int cpu = raw_smp_processor_id();
Alexander Graf10474ae2009-09-15 11:37:46 +02002304 int r;
Avi Kivity1b6c0162007-05-24 13:03:52 +03002305
Rusty Russell7f59f492008-12-07 21:25:45 +10302306 if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03002307 return;
Alexander Graf10474ae2009-09-15 11:37:46 +02002308
Rusty Russell7f59f492008-12-07 21:25:45 +10302309 cpumask_set_cpu(cpu, cpus_hardware_enabled);
Alexander Graf10474ae2009-09-15 11:37:46 +02002310
2311 r = kvm_arch_hardware_enable(NULL);
2312
2313 if (r) {
2314 cpumask_clear_cpu(cpu, cpus_hardware_enabled);
2315 atomic_inc(&hardware_enable_failed);
2316 printk(KERN_INFO "kvm: enabling virtualization on "
2317 "CPU%d failed\n", cpu);
2318 }
Avi Kivity1b6c0162007-05-24 13:03:52 +03002319}
2320
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002321static void hardware_enable(void *junk)
2322{
Jan Kiszkae935b832011-02-08 12:55:33 +01002323 raw_spin_lock(&kvm_lock);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002324 hardware_enable_nolock(junk);
Jan Kiszkae935b832011-02-08 12:55:33 +01002325 raw_spin_unlock(&kvm_lock);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002326}
2327
2328static void hardware_disable_nolock(void *junk)
Avi Kivity1b6c0162007-05-24 13:03:52 +03002329{
2330 int cpu = raw_smp_processor_id();
2331
Rusty Russell7f59f492008-12-07 21:25:45 +10302332 if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03002333 return;
Rusty Russell7f59f492008-12-07 21:25:45 +10302334 cpumask_clear_cpu(cpu, cpus_hardware_enabled);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002335 kvm_arch_hardware_disable(NULL);
Avi Kivity1b6c0162007-05-24 13:03:52 +03002336}
2337
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002338static void hardware_disable(void *junk)
2339{
Jan Kiszkae935b832011-02-08 12:55:33 +01002340 raw_spin_lock(&kvm_lock);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002341 hardware_disable_nolock(junk);
Jan Kiszkae935b832011-02-08 12:55:33 +01002342 raw_spin_unlock(&kvm_lock);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002343}
2344
Alexander Graf10474ae2009-09-15 11:37:46 +02002345static void hardware_disable_all_nolock(void)
2346{
2347 BUG_ON(!kvm_usage_count);
2348
2349 kvm_usage_count--;
2350 if (!kvm_usage_count)
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002351 on_each_cpu(hardware_disable_nolock, NULL, 1);
Alexander Graf10474ae2009-09-15 11:37:46 +02002352}
2353
2354static void hardware_disable_all(void)
2355{
Jan Kiszkae935b832011-02-08 12:55:33 +01002356 raw_spin_lock(&kvm_lock);
Alexander Graf10474ae2009-09-15 11:37:46 +02002357 hardware_disable_all_nolock();
Jan Kiszkae935b832011-02-08 12:55:33 +01002358 raw_spin_unlock(&kvm_lock);
Alexander Graf10474ae2009-09-15 11:37:46 +02002359}
2360
2361static int hardware_enable_all(void)
2362{
2363 int r = 0;
2364
Jan Kiszkae935b832011-02-08 12:55:33 +01002365 raw_spin_lock(&kvm_lock);
Alexander Graf10474ae2009-09-15 11:37:46 +02002366
2367 kvm_usage_count++;
2368 if (kvm_usage_count == 1) {
2369 atomic_set(&hardware_enable_failed, 0);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002370 on_each_cpu(hardware_enable_nolock, NULL, 1);
Alexander Graf10474ae2009-09-15 11:37:46 +02002371
2372 if (atomic_read(&hardware_enable_failed)) {
2373 hardware_disable_all_nolock();
2374 r = -EBUSY;
2375 }
2376 }
2377
Jan Kiszkae935b832011-02-08 12:55:33 +01002378 raw_spin_unlock(&kvm_lock);
Alexander Graf10474ae2009-09-15 11:37:46 +02002379
2380 return r;
2381}
2382
Avi Kivity774c47f2007-02-12 00:54:47 -08002383static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val,
2384 void *v)
2385{
2386 int cpu = (long)v;
2387
Alexander Graf10474ae2009-09-15 11:37:46 +02002388 if (!kvm_usage_count)
2389 return NOTIFY_OK;
2390
Avi Kivity1a6f4d72007-11-11 18:37:32 +02002391 val &= ~CPU_TASKS_FROZEN;
Avi Kivity774c47f2007-02-12 00:54:47 -08002392 switch (val) {
Avi Kivitycec9ad22007-05-24 13:11:41 +03002393 case CPU_DYING:
Avi Kivity6ec8a852007-08-19 15:57:26 +03002394 printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n",
2395 cpu);
2396 hardware_disable(NULL);
2397 break;
Zachary Amsdenda908f22010-08-19 22:07:27 -10002398 case CPU_STARTING:
Jeremy Katz43934a32007-02-19 14:37:46 +02002399 printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n",
2400 cpu);
Zachary Amsdenda908f22010-08-19 22:07:27 -10002401 hardware_enable(NULL);
Avi Kivity774c47f2007-02-12 00:54:47 -08002402 break;
2403 }
2404 return NOTIFY_OK;
2405}
2406
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002407
Avi Kivityb7c41452010-12-02 17:52:50 +02002408asmlinkage void kvm_spurious_fault(void)
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002409{
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002410 /* Fault while not rebooting. We want the trace. */
2411 BUG();
2412}
Avi Kivityb7c41452010-12-02 17:52:50 +02002413EXPORT_SYMBOL_GPL(kvm_spurious_fault);
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002414
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002415static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
Mike Dayd77c26f2007-10-08 09:02:08 -04002416 void *v)
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002417{
Sheng Yang8e1c1812009-04-29 11:09:04 +08002418 /*
2419 * Some (well, at least mine) BIOSes hang on reboot if
2420 * in vmx root mode.
2421 *
2422 * And Intel TXT required VMX off for all cpu when system shutdown.
2423 */
2424 printk(KERN_INFO "kvm: exiting hardware virtualization\n");
2425 kvm_rebooting = true;
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002426 on_each_cpu(hardware_disable_nolock, NULL, 1);
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002427 return NOTIFY_OK;
2428}
2429
2430static struct notifier_block kvm_reboot_notifier = {
2431 .notifier_call = kvm_reboot,
2432 .priority = 0,
2433};
2434
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002435static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002436{
2437 int i;
2438
2439 for (i = 0; i < bus->dev_count; i++) {
Sasha Levin743eeb02011-07-27 16:00:48 +03002440 struct kvm_io_device *pos = bus->range[i].dev;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002441
2442 kvm_iodevice_destructor(pos);
2443 }
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002444 kfree(bus);
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002445}
2446
Sasha Levin743eeb02011-07-27 16:00:48 +03002447int kvm_io_bus_sort_cmp(const void *p1, const void *p2)
2448{
2449 const struct kvm_io_range *r1 = p1;
2450 const struct kvm_io_range *r2 = p2;
2451
2452 if (r1->addr < r2->addr)
2453 return -1;
2454 if (r1->addr + r1->len > r2->addr + r2->len)
2455 return 1;
2456 return 0;
2457}
2458
2459int kvm_io_bus_insert_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev,
2460 gpa_t addr, int len)
2461{
2462 if (bus->dev_count == NR_IOBUS_DEVS)
2463 return -ENOSPC;
2464
2465 bus->range[bus->dev_count++] = (struct kvm_io_range) {
2466 .addr = addr,
2467 .len = len,
2468 .dev = dev,
2469 };
2470
2471 sort(bus->range, bus->dev_count, sizeof(struct kvm_io_range),
2472 kvm_io_bus_sort_cmp, NULL);
2473
2474 return 0;
2475}
2476
2477int kvm_io_bus_get_first_dev(struct kvm_io_bus *bus,
2478 gpa_t addr, int len)
2479{
2480 struct kvm_io_range *range, key;
2481 int off;
2482
2483 key = (struct kvm_io_range) {
2484 .addr = addr,
2485 .len = len,
2486 };
2487
2488 range = bsearch(&key, bus->range, bus->dev_count,
2489 sizeof(struct kvm_io_range), kvm_io_bus_sort_cmp);
2490 if (range == NULL)
2491 return -ENOENT;
2492
2493 off = range - bus->range;
2494
2495 while (off > 0 && kvm_io_bus_sort_cmp(&key, &bus->range[off-1]) == 0)
2496 off--;
2497
2498 return off;
2499}
2500
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002501/* kvm_io_bus_write - called under kvm->slots_lock */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002502int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002503 int len, const void *val)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002504{
Sasha Levin743eeb02011-07-27 16:00:48 +03002505 int idx;
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002506 struct kvm_io_bus *bus;
Sasha Levin743eeb02011-07-27 16:00:48 +03002507 struct kvm_io_range range;
2508
2509 range = (struct kvm_io_range) {
2510 .addr = addr,
2511 .len = len,
2512 };
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002513
2514 bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu);
Sasha Levin743eeb02011-07-27 16:00:48 +03002515 idx = kvm_io_bus_get_first_dev(bus, addr, len);
2516 if (idx < 0)
2517 return -EOPNOTSUPP;
2518
2519 while (idx < bus->dev_count &&
2520 kvm_io_bus_sort_cmp(&range, &bus->range[idx]) == 0) {
2521 if (!kvm_iodevice_write(bus->range[idx].dev, addr, len, val))
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002522 return 0;
Sasha Levin743eeb02011-07-27 16:00:48 +03002523 idx++;
2524 }
2525
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002526 return -EOPNOTSUPP;
2527}
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002528
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002529/* kvm_io_bus_read - called under kvm->slots_lock */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002530int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
2531 int len, void *val)
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002532{
Sasha Levin743eeb02011-07-27 16:00:48 +03002533 int idx;
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002534 struct kvm_io_bus *bus;
Sasha Levin743eeb02011-07-27 16:00:48 +03002535 struct kvm_io_range range;
2536
2537 range = (struct kvm_io_range) {
2538 .addr = addr,
2539 .len = len,
2540 };
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002541
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002542 bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu);
Sasha Levin743eeb02011-07-27 16:00:48 +03002543 idx = kvm_io_bus_get_first_dev(bus, addr, len);
2544 if (idx < 0)
2545 return -EOPNOTSUPP;
2546
2547 while (idx < bus->dev_count &&
2548 kvm_io_bus_sort_cmp(&range, &bus->range[idx]) == 0) {
2549 if (!kvm_iodevice_read(bus->range[idx].dev, addr, len, val))
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002550 return 0;
Sasha Levin743eeb02011-07-27 16:00:48 +03002551 idx++;
2552 }
2553
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002554 return -EOPNOTSUPP;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002555}
2556
Marcelo Tosatti79fac952009-12-23 14:35:26 -02002557/* Caller must hold slots_lock. */
Sasha Levin743eeb02011-07-27 16:00:48 +03002558int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
2559 int len, struct kvm_io_device *dev)
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002560{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002561 struct kvm_io_bus *new_bus, *bus;
Gregory Haskins090b7af2009-07-07 17:08:44 -04002562
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002563 bus = kvm->buses[bus_idx];
Gregory Haskins090b7af2009-07-07 17:08:44 -04002564 if (bus->dev_count > NR_IOBUS_DEVS-1)
2565 return -ENOSPC;
2566
Thomas Meyer6da64fd2011-11-08 20:32:19 +01002567 new_bus = kmemdup(bus, sizeof(struct kvm_io_bus), GFP_KERNEL);
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002568 if (!new_bus)
2569 return -ENOMEM;
Sasha Levin743eeb02011-07-27 16:00:48 +03002570 kvm_io_bus_insert_dev(new_bus, dev, addr, len);
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002571 rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
2572 synchronize_srcu_expedited(&kvm->srcu);
2573 kfree(bus);
Gregory Haskins090b7af2009-07-07 17:08:44 -04002574
2575 return 0;
2576}
2577
Marcelo Tosatti79fac952009-12-23 14:35:26 -02002578/* Caller must hold slots_lock. */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002579int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
2580 struct kvm_io_device *dev)
Gregory Haskins090b7af2009-07-07 17:08:44 -04002581{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002582 int i, r;
2583 struct kvm_io_bus *new_bus, *bus;
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002584
Sasha Levincdfca7b2011-12-04 19:36:28 +02002585 bus = kvm->buses[bus_idx];
2586
2587 new_bus = kmemdup(bus, sizeof(*bus), GFP_KERNEL);
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002588 if (!new_bus)
2589 return -ENOMEM;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002590
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002591 r = -ENOENT;
2592 for (i = 0; i < new_bus->dev_count; i++)
Sasha Levin743eeb02011-07-27 16:00:48 +03002593 if (new_bus->range[i].dev == dev) {
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002594 r = 0;
Sasha Levin743eeb02011-07-27 16:00:48 +03002595 new_bus->dev_count--;
2596 new_bus->range[i] = new_bus->range[new_bus->dev_count];
2597 sort(new_bus->range, new_bus->dev_count,
2598 sizeof(struct kvm_io_range),
2599 kvm_io_bus_sort_cmp, NULL);
Gregory Haskins090b7af2009-07-07 17:08:44 -04002600 break;
2601 }
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002602
2603 if (r) {
2604 kfree(new_bus);
2605 return r;
2606 }
2607
2608 rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
2609 synchronize_srcu_expedited(&kvm->srcu);
2610 kfree(bus);
2611 return r;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002612}
2613
Avi Kivity774c47f2007-02-12 00:54:47 -08002614static struct notifier_block kvm_cpu_notifier = {
2615 .notifier_call = kvm_cpu_hotplug,
Avi Kivity774c47f2007-02-12 00:54:47 -08002616};
2617
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002618static int vm_stat_get(void *_offset, u64 *val)
Avi Kivityba1389b2007-11-18 16:24:12 +02002619{
2620 unsigned offset = (long)_offset;
Avi Kivityba1389b2007-11-18 16:24:12 +02002621 struct kvm *kvm;
2622
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002623 *val = 0;
Jan Kiszkae935b832011-02-08 12:55:33 +01002624 raw_spin_lock(&kvm_lock);
Avi Kivityba1389b2007-11-18 16:24:12 +02002625 list_for_each_entry(kvm, &vm_list, vm_list)
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002626 *val += *(u32 *)((void *)kvm + offset);
Jan Kiszkae935b832011-02-08 12:55:33 +01002627 raw_spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002628 return 0;
Avi Kivityba1389b2007-11-18 16:24:12 +02002629}
2630
2631DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n");
2632
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002633static int vcpu_stat_get(void *_offset, u64 *val)
Avi Kivity1165f5f2007-04-19 17:27:43 +03002634{
2635 unsigned offset = (long)_offset;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002636 struct kvm *kvm;
2637 struct kvm_vcpu *vcpu;
2638 int i;
2639
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002640 *val = 0;
Jan Kiszkae935b832011-02-08 12:55:33 +01002641 raw_spin_lock(&kvm_lock);
Avi Kivity1165f5f2007-04-19 17:27:43 +03002642 list_for_each_entry(kvm, &vm_list, vm_list)
Gleb Natapov988a2ca2009-06-09 15:56:29 +03002643 kvm_for_each_vcpu(i, vcpu, kvm)
2644 *val += *(u32 *)((void *)vcpu + offset);
2645
Jan Kiszkae935b832011-02-08 12:55:33 +01002646 raw_spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002647 return 0;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002648}
2649
Avi Kivityba1389b2007-11-18 16:24:12 +02002650DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n");
2651
Alexey Dobriyan828c0952009-10-01 15:43:56 -07002652static const struct file_operations *stat_fops[] = {
Avi Kivityba1389b2007-11-18 16:24:12 +02002653 [KVM_STAT_VCPU] = &vcpu_stat_fops,
2654 [KVM_STAT_VM] = &vm_stat_fops,
2655};
Avi Kivity1165f5f2007-04-19 17:27:43 +03002656
Hamo4f69b682011-12-15 14:23:16 +08002657static int kvm_init_debug(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002658{
Hamo4f69b682011-12-15 14:23:16 +08002659 int r = -EFAULT;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002660 struct kvm_stats_debugfs_item *p;
2661
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002662 kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
Hamo4f69b682011-12-15 14:23:16 +08002663 if (kvm_debugfs_dir == NULL)
2664 goto out;
2665
2666 for (p = debugfs_entries; p->name; ++p) {
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002667 p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir,
Avi Kivity1165f5f2007-04-19 17:27:43 +03002668 (void *)(long)p->offset,
Avi Kivityba1389b2007-11-18 16:24:12 +02002669 stat_fops[p->kind]);
Hamo4f69b682011-12-15 14:23:16 +08002670 if (p->dentry == NULL)
2671 goto out_dir;
2672 }
2673
2674 return 0;
2675
2676out_dir:
2677 debugfs_remove_recursive(kvm_debugfs_dir);
2678out:
2679 return r;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002680}
2681
2682static void kvm_exit_debug(void)
2683{
2684 struct kvm_stats_debugfs_item *p;
2685
2686 for (p = debugfs_entries; p->name; ++p)
2687 debugfs_remove(p->dentry);
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002688 debugfs_remove(kvm_debugfs_dir);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002689}
2690
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002691static int kvm_suspend(void)
Avi Kivity59ae6c62007-02-12 00:54:48 -08002692{
Alexander Graf10474ae2009-09-15 11:37:46 +02002693 if (kvm_usage_count)
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002694 hardware_disable_nolock(NULL);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002695 return 0;
2696}
2697
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002698static void kvm_resume(void)
Avi Kivity59ae6c62007-02-12 00:54:48 -08002699{
Zachary Amsdenca84d1a2010-08-19 22:07:28 -10002700 if (kvm_usage_count) {
Jan Kiszkae935b832011-02-08 12:55:33 +01002701 WARN_ON(raw_spin_is_locked(&kvm_lock));
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002702 hardware_enable_nolock(NULL);
Zachary Amsdenca84d1a2010-08-19 22:07:28 -10002703 }
Avi Kivity59ae6c62007-02-12 00:54:48 -08002704}
2705
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002706static struct syscore_ops kvm_syscore_ops = {
Avi Kivity59ae6c62007-02-12 00:54:48 -08002707 .suspend = kvm_suspend,
2708 .resume = kvm_resume,
2709};
2710
Izik Eiduscea7bb22007-10-17 19:17:48 +02002711struct page *bad_page;
Anthony Liguori35149e22008-04-02 14:46:56 -05002712pfn_t bad_pfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002713
Avi Kivity15ad7142007-07-11 18:17:21 +03002714static inline
2715struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn)
2716{
2717 return container_of(pn, struct kvm_vcpu, preempt_notifier);
2718}
2719
2720static void kvm_sched_in(struct preempt_notifier *pn, int cpu)
2721{
2722 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2723
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002724 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002725}
2726
2727static void kvm_sched_out(struct preempt_notifier *pn,
2728 struct task_struct *next)
2729{
2730 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2731
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002732 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002733}
2734
Avi Kivity0ee75be2010-04-28 15:39:01 +03002735int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
Rusty Russellc16f8622007-07-30 21:12:19 +10002736 struct module *module)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002737{
2738 int r;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002739 int cpu;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002740
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002741 r = kvm_arch_init(opaque);
2742 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002743 goto out_fail;
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002744
2745 bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2746
2747 if (bad_page == NULL) {
2748 r = -ENOMEM;
2749 goto out;
2750 }
2751
Anthony Liguori35149e22008-04-02 14:46:56 -05002752 bad_pfn = page_to_pfn(bad_page);
2753
Huang Yingbf998152010-05-31 14:28:19 +08002754 hwpoison_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2755
2756 if (hwpoison_page == NULL) {
2757 r = -ENOMEM;
2758 goto out_free_0;
2759 }
2760
2761 hwpoison_pfn = page_to_pfn(hwpoison_page);
2762
Gleb Natapovedba23e2010-07-07 20:16:45 +03002763 fault_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2764
2765 if (fault_page == NULL) {
2766 r = -ENOMEM;
2767 goto out_free_0;
2768 }
2769
2770 fault_pfn = page_to_pfn(fault_page);
2771
Avi Kivity8437a612009-06-06 14:52:35 -07002772 if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
Rusty Russell7f59f492008-12-07 21:25:45 +10302773 r = -ENOMEM;
2774 goto out_free_0;
2775 }
2776
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002777 r = kvm_arch_hardware_setup();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002778 if (r < 0)
Rusty Russell7f59f492008-12-07 21:25:45 +10302779 goto out_free_0a;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002780
Yang, Sheng002c7f72007-07-31 14:23:01 +03002781 for_each_online_cpu(cpu) {
2782 smp_call_function_single(cpu,
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002783 kvm_arch_check_processor_compat,
Jens Axboe8691e5a2008-06-06 11:18:06 +02002784 &r, 1);
Yang, Sheng002c7f72007-07-31 14:23:01 +03002785 if (r < 0)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002786 goto out_free_1;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002787 }
2788
Avi Kivity774c47f2007-02-12 00:54:47 -08002789 r = register_cpu_notifier(&kvm_cpu_notifier);
2790 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002791 goto out_free_2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002792 register_reboot_notifier(&kvm_reboot_notifier);
2793
Rusty Russellc16f8622007-07-30 21:12:19 +10002794 /* A kmem cache lets us meet the alignment requirements of fx_save. */
Avi Kivity0ee75be2010-04-28 15:39:01 +03002795 if (!vcpu_align)
2796 vcpu_align = __alignof__(struct kvm_vcpu);
2797 kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, vcpu_align,
Joe Perches56919c52007-11-12 20:06:51 -08002798 0, NULL);
Rusty Russellc16f8622007-07-30 21:12:19 +10002799 if (!kvm_vcpu_cache) {
2800 r = -ENOMEM;
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002801 goto out_free_3;
Rusty Russellc16f8622007-07-30 21:12:19 +10002802 }
2803
Gleb Natapovaf585b92010-10-14 11:22:46 +02002804 r = kvm_async_pf_init();
2805 if (r)
2806 goto out_free;
2807
Avi Kivity6aa8b732006-12-10 02:21:36 -08002808 kvm_chardev_ops.owner = module;
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01002809 kvm_vm_fops.owner = module;
2810 kvm_vcpu_fops.owner = module;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002811
2812 r = misc_register(&kvm_dev);
2813 if (r) {
Mike Dayd77c26f2007-10-08 09:02:08 -04002814 printk(KERN_ERR "kvm: misc device register failed\n");
Gleb Natapovaf585b92010-10-14 11:22:46 +02002815 goto out_unreg;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002816 }
2817
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002818 register_syscore_ops(&kvm_syscore_ops);
2819
Avi Kivity15ad7142007-07-11 18:17:21 +03002820 kvm_preempt_ops.sched_in = kvm_sched_in;
2821 kvm_preempt_ops.sched_out = kvm_sched_out;
2822
Hamo4f69b682011-12-15 14:23:16 +08002823 r = kvm_init_debug();
2824 if (r) {
2825 printk(KERN_ERR "kvm: create debugfs files failed\n");
2826 goto out_undebugfs;
2827 }
Darrick J. Wong0ea4ed82009-10-14 16:21:00 -07002828
Avi Kivityc7addb92007-09-16 18:58:32 +02002829 return 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002830
Hamo4f69b682011-12-15 14:23:16 +08002831out_undebugfs:
2832 unregister_syscore_ops(&kvm_syscore_ops);
Gleb Natapovaf585b92010-10-14 11:22:46 +02002833out_unreg:
2834 kvm_async_pf_deinit();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002835out_free:
Rusty Russellc16f8622007-07-30 21:12:19 +10002836 kmem_cache_destroy(kvm_vcpu_cache);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002837out_free_3:
Avi Kivity6aa8b732006-12-10 02:21:36 -08002838 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity774c47f2007-02-12 00:54:47 -08002839 unregister_cpu_notifier(&kvm_cpu_notifier);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002840out_free_2:
Zhang Xiantaod23087842007-11-29 15:35:39 +08002841out_free_1:
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002842 kvm_arch_hardware_unsetup();
Rusty Russell7f59f492008-12-07 21:25:45 +10302843out_free_0a:
2844 free_cpumask_var(cpus_hardware_enabled);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002845out_free_0:
Gleb Natapovedba23e2010-07-07 20:16:45 +03002846 if (fault_page)
2847 __free_page(fault_page);
Huang Yingbf998152010-05-31 14:28:19 +08002848 if (hwpoison_page)
2849 __free_page(hwpoison_page);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002850 __free_page(bad_page);
Avi Kivityca45aaa2007-03-01 19:21:03 +02002851out:
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002852 kvm_arch_exit();
Zhang Xiantaod23087842007-11-29 15:35:39 +08002853out_fail:
Avi Kivity6aa8b732006-12-10 02:21:36 -08002854 return r;
2855}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002856EXPORT_SYMBOL_GPL(kvm_init);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002857
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002858void kvm_exit(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002859{
Darrick J. Wong0ea4ed82009-10-14 16:21:00 -07002860 kvm_exit_debug();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002861 misc_deregister(&kvm_dev);
Rusty Russellc16f8622007-07-30 21:12:19 +10002862 kmem_cache_destroy(kvm_vcpu_cache);
Gleb Natapovaf585b92010-10-14 11:22:46 +02002863 kvm_async_pf_deinit();
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002864 unregister_syscore_ops(&kvm_syscore_ops);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002865 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002866 unregister_cpu_notifier(&kvm_cpu_notifier);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002867 on_each_cpu(hardware_disable_nolock, NULL, 1);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002868 kvm_arch_hardware_unsetup();
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002869 kvm_arch_exit();
Rusty Russell7f59f492008-12-07 21:25:45 +10302870 free_cpumask_var(cpus_hardware_enabled);
Huang Yingbf998152010-05-31 14:28:19 +08002871 __free_page(hwpoison_page);
Izik Eiduscea7bb22007-10-17 19:17:48 +02002872 __free_page(bad_page);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002873}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002874EXPORT_SYMBOL_GPL(kvm_exit);