blob: d617f69131d7667d3847c7e0fc040a717625bf38 [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{
Alex Shibec87d62012-03-04 13:13:38 +0800206 long dirty_count = kvm->tlbs_dirty;
Xiao Guangronga4ee1ca2010-11-23 11:13:00 +0800207
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);
Takuya Yoshikawa565f3be2012-02-10 15:28:31 +0900292
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200293 kvm->mmu_notifier_seq++;
Xiao Guangronga4ee1ca2010-11-23 11:13:00 +0800294 need_tlb_flush = kvm_unmap_hva(kvm, address) | kvm->tlbs_dirty;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200295 /* we've to flush the tlb before the pages can be freed */
296 if (need_tlb_flush)
297 kvm_flush_remote_tlbs(kvm);
298
Takuya Yoshikawa565f3be2012-02-10 15:28:31 +0900299 spin_unlock(&kvm->mmu_lock);
300 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200301}
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 /* we've to flush the tlb before the pages can be freed */
339 if (need_tlb_flush)
340 kvm_flush_remote_tlbs(kvm);
Takuya Yoshikawa565f3be2012-02-10 15:28:31 +0900341
342 spin_unlock(&kvm->mmu_lock);
343 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200344}
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++;
Paul Mackerrasa355aa52011-12-12 12:37:21 +0000360 smp_wmb();
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200361 /*
362 * The above sequence increase must be visible before the
Paul Mackerrasa355aa52011-12-12 12:37:21 +0000363 * below count decrease, which is ensured by the smp_wmb above
364 * in conjunction with the smp_rmb in mmu_notifier_retry().
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200365 */
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);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200381
Takuya Yoshikawa565f3be2012-02-10 15:28:31 +0900382 young = kvm_age_hva(kvm, address);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200383 if (young)
384 kvm_flush_remote_tlbs(kvm);
385
Takuya Yoshikawa565f3be2012-02-10 15:28:31 +0900386 spin_unlock(&kvm->mmu_lock);
387 srcu_read_unlock(&kvm->srcu, idx);
388
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200389 return young;
390}
391
Andrea Arcangeli8ee53822011-01-13 15:47:10 -0800392static int kvm_mmu_notifier_test_young(struct mmu_notifier *mn,
393 struct mm_struct *mm,
394 unsigned long address)
395{
396 struct kvm *kvm = mmu_notifier_to_kvm(mn);
397 int young, idx;
398
399 idx = srcu_read_lock(&kvm->srcu);
400 spin_lock(&kvm->mmu_lock);
401 young = kvm_test_age_hva(kvm, address);
402 spin_unlock(&kvm->mmu_lock);
403 srcu_read_unlock(&kvm->srcu, idx);
404
405 return young;
406}
407
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100408static void kvm_mmu_notifier_release(struct mmu_notifier *mn,
409 struct mm_struct *mm)
410{
411 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Lai Jiangshaneda2bed2010-04-20 14:29:29 +0800412 int idx;
413
414 idx = srcu_read_lock(&kvm->srcu);
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100415 kvm_arch_flush_shadow(kvm);
Lai Jiangshaneda2bed2010-04-20 14:29:29 +0800416 srcu_read_unlock(&kvm->srcu, idx);
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100417}
418
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200419static const struct mmu_notifier_ops kvm_mmu_notifier_ops = {
420 .invalidate_page = kvm_mmu_notifier_invalidate_page,
421 .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start,
422 .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end,
423 .clear_flush_young = kvm_mmu_notifier_clear_flush_young,
Andrea Arcangeli8ee53822011-01-13 15:47:10 -0800424 .test_young = kvm_mmu_notifier_test_young,
Izik Eidus3da0dd42009-09-23 21:47:18 +0300425 .change_pte = kvm_mmu_notifier_change_pte,
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100426 .release = kvm_mmu_notifier_release,
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200427};
Avi Kivity4c07b0a2009-12-20 14:54:04 +0200428
429static int kvm_init_mmu_notifier(struct kvm *kvm)
430{
431 kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops;
432 return mmu_notifier_register(&kvm->mmu_notifier, current->mm);
433}
434
435#else /* !(CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER) */
436
437static int kvm_init_mmu_notifier(struct kvm *kvm)
438{
439 return 0;
440}
441
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200442#endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */
443
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800444static void kvm_init_memslots_id(struct kvm *kvm)
445{
446 int i;
447 struct kvm_memslots *slots = kvm->memslots;
448
449 for (i = 0; i < KVM_MEM_SLOTS_NUM; i++)
Xiao Guangrongf85e2cb2011-11-24 17:41:54 +0800450 slots->id_to_index[i] = slots->memslots[i].id = i;
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800451}
452
Carsten Ottee08b9632012-01-04 10:25:20 +0100453static struct kvm *kvm_create_vm(unsigned long type)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800454{
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100455 int r, i;
456 struct kvm *kvm = kvm_arch_alloc_vm();
Avi Kivity6aa8b732006-12-10 02:21:36 -0800457
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100458 if (!kvm)
459 return ERR_PTR(-ENOMEM);
460
Carsten Ottee08b9632012-01-04 10:25:20 +0100461 r = kvm_arch_init_vm(kvm, type);
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100462 if (r)
463 goto out_err_nodisable;
Alexander Graf10474ae2009-09-15 11:37:46 +0200464
465 r = hardware_enable_all();
466 if (r)
467 goto out_err_nodisable;
468
Avi Kivity75858a82009-01-04 17:10:50 +0200469#ifdef CONFIG_HAVE_KVM_IRQCHIP
470 INIT_HLIST_HEAD(&kvm->mask_notifier_list);
Gleb Natapov136bdfe2009-08-24 11:54:23 +0300471 INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list);
Avi Kivity75858a82009-01-04 17:10:50 +0200472#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -0800473
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200474 r = -ENOMEM;
475 kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
476 if (!kvm->memslots)
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100477 goto out_err_nosrcu;
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800478 kvm_init_memslots_id(kvm);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200479 if (init_srcu_struct(&kvm->srcu))
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100480 goto out_err_nosrcu;
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200481 for (i = 0; i < KVM_NR_BUSES; i++) {
482 kvm->buses[i] = kzalloc(sizeof(struct kvm_io_bus),
483 GFP_KERNEL);
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100484 if (!kvm->buses[i])
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200485 goto out_err;
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200486 }
Mike Waychison74b5c5b2011-06-03 13:04:53 -0700487
OGAWA Hirofumi85722cd2011-05-11 09:28:28 +0900488 spin_lock_init(&kvm->mmu_lock);
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200489 kvm->mm = current->mm;
490 atomic_inc(&kvm->mm->mm_count);
Gregory Haskinsd34e6b12009-07-07 17:08:49 -0400491 kvm_eventfd_init(kvm);
Shaohua Li11ec2802007-07-23 14:51:37 +0800492 mutex_init(&kvm->lock);
Marcelo Tosatti60eead72009-06-04 15:08:23 -0300493 mutex_init(&kvm->irq_lock);
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200494 mutex_init(&kvm->slots_lock);
Izik Eidusd39f13b2008-03-30 16:01:25 +0300495 atomic_set(&kvm->users_count, 1);
Mike Waychison74b5c5b2011-06-03 13:04:53 -0700496
497 r = kvm_init_mmu_notifier(kvm);
498 if (r)
499 goto out_err;
500
Jan Kiszkae935b832011-02-08 12:55:33 +0100501 raw_spin_lock(&kvm_lock);
Rusty Russell5e58cfe2007-07-23 17:08:21 +1000502 list_add(&kvm->vm_list, &vm_list);
Jan Kiszkae935b832011-02-08 12:55:33 +0100503 raw_spin_unlock(&kvm_lock);
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100504
Avi Kivityf17abe92007-02-21 19:28:04 +0200505 return kvm;
Alexander Graf10474ae2009-09-15 11:37:46 +0200506
507out_err:
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100508 cleanup_srcu_struct(&kvm->srcu);
509out_err_nosrcu:
Alexander Graf10474ae2009-09-15 11:37:46 +0200510 hardware_disable_all();
511out_err_nodisable:
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200512 for (i = 0; i < KVM_NR_BUSES; i++)
513 kfree(kvm->buses[i]);
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200514 kfree(kvm->memslots);
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100515 kvm_arch_free_vm(kvm);
Alexander Graf10474ae2009-09-15 11:37:46 +0200516 return ERR_PTR(r);
Avi Kivityf17abe92007-02-21 19:28:04 +0200517}
518
Takuya Yoshikawa92eca8f2012-05-20 13:13:28 +0900519/*
520 * Avoid using vmalloc for a small buffer.
521 * Should not be used when the size is statically known.
522 */
Takuya Yoshikawac1a7b322012-05-20 13:15:07 +0900523void *kvm_kvzalloc(unsigned long size)
Takuya Yoshikawa92eca8f2012-05-20 13:13:28 +0900524{
525 if (size > PAGE_SIZE)
526 return vzalloc(size);
527 else
528 return kzalloc(size, GFP_KERNEL);
529}
530
Takuya Yoshikawac1a7b322012-05-20 13:15:07 +0900531void kvm_kvfree(const void *addr)
Takuya Yoshikawa92eca8f2012-05-20 13:13:28 +0900532{
533 if (is_vmalloc_addr(addr))
534 vfree(addr);
535 else
536 kfree(addr);
537}
538
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900539static void kvm_destroy_dirty_bitmap(struct kvm_memory_slot *memslot)
540{
541 if (!memslot->dirty_bitmap)
542 return;
543
Takuya Yoshikawa92eca8f2012-05-20 13:13:28 +0900544 kvm_kvfree(memslot->dirty_bitmap);
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900545 memslot->dirty_bitmap = NULL;
546}
547
Avi Kivity6aa8b732006-12-10 02:21:36 -0800548/*
549 * Free any memory in @free but not in @dont.
550 */
551static void kvm_free_physmem_slot(struct kvm_memory_slot *free,
552 struct kvm_memory_slot *dont)
553{
Izik Eidus290fc382007-09-27 14:11:22 +0200554 if (!dont || free->rmap != dont->rmap)
555 vfree(free->rmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800556
557 if (!dont || free->dirty_bitmap != dont->dirty_bitmap)
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900558 kvm_destroy_dirty_bitmap(free);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800559
Takuya Yoshikawadb3fe4e2012-02-08 13:02:18 +0900560 kvm_arch_free_memslot(free, dont);
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300561
Avi Kivity6aa8b732006-12-10 02:21:36 -0800562 free->npages = 0;
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500563 free->rmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800564}
565
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800566void kvm_free_physmem(struct kvm *kvm)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800567{
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200568 struct kvm_memslots *slots = kvm->memslots;
Xiao Guangrongbe6ba0f2011-11-24 17:39:18 +0800569 struct kvm_memory_slot *memslot;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800570
Xiao Guangrongbe6ba0f2011-11-24 17:39:18 +0800571 kvm_for_each_memslot(memslot, slots)
572 kvm_free_physmem_slot(memslot, NULL);
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200573
574 kfree(kvm->memslots);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800575}
576
Avi Kivityf17abe92007-02-21 19:28:04 +0200577static void kvm_destroy_vm(struct kvm *kvm)
578{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200579 int i;
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200580 struct mm_struct *mm = kvm->mm;
581
Sheng Yangad8ba2c2009-01-06 10:03:02 +0800582 kvm_arch_sync_events(kvm);
Jan Kiszkae935b832011-02-08 12:55:33 +0100583 raw_spin_lock(&kvm_lock);
Avi Kivity133de902007-02-12 00:54:44 -0800584 list_del(&kvm->vm_list);
Jan Kiszkae935b832011-02-08 12:55:33 +0100585 raw_spin_unlock(&kvm_lock);
Avi Kivity399ec802008-11-19 13:58:46 +0200586 kvm_free_irq_routing(kvm);
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200587 for (i = 0; i < KVM_NR_BUSES; i++)
588 kvm_io_bus_destroy(kvm->buses[i]);
Avi Kivity980da6c2009-12-20 15:13:43 +0200589 kvm_coalesced_mmio_free(kvm);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200590#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
591 mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
Gleb Natapovf00be0c2009-03-19 12:20:36 +0200592#else
593 kvm_arch_flush_shadow(kvm);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200594#endif
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800595 kvm_arch_destroy_vm(kvm);
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100596 kvm_free_physmem(kvm);
597 cleanup_srcu_struct(&kvm->srcu);
598 kvm_arch_free_vm(kvm);
Alexander Graf10474ae2009-09-15 11:37:46 +0200599 hardware_disable_all();
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200600 mmdrop(mm);
Avi Kivityf17abe92007-02-21 19:28:04 +0200601}
602
Izik Eidusd39f13b2008-03-30 16:01:25 +0300603void kvm_get_kvm(struct kvm *kvm)
604{
605 atomic_inc(&kvm->users_count);
606}
607EXPORT_SYMBOL_GPL(kvm_get_kvm);
608
609void kvm_put_kvm(struct kvm *kvm)
610{
611 if (atomic_dec_and_test(&kvm->users_count))
612 kvm_destroy_vm(kvm);
613}
614EXPORT_SYMBOL_GPL(kvm_put_kvm);
615
616
Avi Kivityf17abe92007-02-21 19:28:04 +0200617static int kvm_vm_release(struct inode *inode, struct file *filp)
618{
619 struct kvm *kvm = filp->private_data;
620
Gregory Haskins721eecbf2009-05-20 10:30:49 -0400621 kvm_irqfd_release(kvm);
622
Izik Eidusd39f13b2008-03-30 16:01:25 +0300623 kvm_put_kvm(kvm);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800624 return 0;
625}
626
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900627/*
628 * Allocation size is twice as large as the actual dirty bitmap size.
Takuya Yoshikawa93474b22012-03-01 19:34:45 +0900629 * See x86's kvm_vm_ioctl_get_dirty_log() why this is needed.
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900630 */
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900631static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot)
632{
Takuya Yoshikawa189a2f72012-02-08 13:01:09 +0900633#ifndef CONFIG_S390
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900634 unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot);
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900635
Takuya Yoshikawa92eca8f2012-05-20 13:13:28 +0900636 memslot->dirty_bitmap = kvm_kvzalloc(dirty_bytes);
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900637 if (!memslot->dirty_bitmap)
638 return -ENOMEM;
639
Takuya Yoshikawa189a2f72012-02-08 13:01:09 +0900640#endif /* !CONFIG_S390 */
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900641 return 0;
642}
643
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800644static int cmp_memslot(const void *slot1, const void *slot2)
645{
646 struct kvm_memory_slot *s1, *s2;
647
648 s1 = (struct kvm_memory_slot *)slot1;
649 s2 = (struct kvm_memory_slot *)slot2;
650
651 if (s1->npages < s2->npages)
652 return 1;
653 if (s1->npages > s2->npages)
654 return -1;
655
656 return 0;
657}
658
659/*
660 * Sort the memslots base on its size, so the larger slots
661 * will get better fit.
662 */
663static void sort_memslots(struct kvm_memslots *slots)
664{
Xiao Guangrongf85e2cb2011-11-24 17:41:54 +0800665 int i;
666
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800667 sort(slots->memslots, KVM_MEM_SLOTS_NUM,
668 sizeof(struct kvm_memory_slot), cmp_memslot, NULL);
Xiao Guangrongf85e2cb2011-11-24 17:41:54 +0800669
670 for (i = 0; i < KVM_MEM_SLOTS_NUM; i++)
671 slots->id_to_index[slots->memslots[i].id] = i;
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800672}
673
Xiao Guangrongbe593d62011-11-24 17:38:24 +0800674void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new)
675{
676 if (new) {
677 int id = new->id;
Xiao Guangrong28a37542011-11-24 19:04:35 +0800678 struct kvm_memory_slot *old = id_to_memslot(slots, id);
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800679 unsigned long npages = old->npages;
Xiao Guangrongbe593d62011-11-24 17:38:24 +0800680
Xiao Guangrong28a37542011-11-24 19:04:35 +0800681 *old = *new;
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800682 if (new->npages != npages)
683 sort_memslots(slots);
Xiao Guangrongbe593d62011-11-24 17:38:24 +0800684 }
685
686 slots->generation++;
687}
688
Avi Kivity6aa8b732006-12-10 02:21:36 -0800689/*
Avi Kivity6aa8b732006-12-10 02:21:36 -0800690 * Allocate some memory and give it an address in the guest physical address
691 * space.
692 *
693 * Discontiguous memory is allowed, mostly for framebuffers.
Sheng Yangf78e0e22007-10-29 09:40:42 +0800694 *
Marcelo Tosatti10589a42007-12-20 19:18:22 -0500695 * Must be called holding mmap_sem for write.
Avi Kivity6aa8b732006-12-10 02:21:36 -0800696 */
Sheng Yangf78e0e22007-10-29 09:40:42 +0800697int __kvm_set_memory_region(struct kvm *kvm,
698 struct kvm_userspace_memory_region *mem,
699 int user_alloc)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800700{
Avi Kivity8234b222010-12-27 12:08:45 +0200701 int r;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800702 gfn_t base_gfn;
Heiko Carstens28bcb112009-09-03 17:35:35 +0200703 unsigned long npages;
704 unsigned long i;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800705 struct kvm_memory_slot *memslot;
706 struct kvm_memory_slot old, new;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200707 struct kvm_memslots *slots, *old_memslots;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800708
709 r = -EINVAL;
710 /* General sanity checks */
711 if (mem->memory_size & (PAGE_SIZE - 1))
712 goto out;
713 if (mem->guest_phys_addr & (PAGE_SIZE - 1))
714 goto out;
Takuya Yoshikawafa3d3152011-05-07 16:35:38 +0900715 /* We can read the guest memory with __xxx_user() later on. */
716 if (user_alloc &&
717 ((mem->userspace_addr & (PAGE_SIZE - 1)) ||
Heiko Carstens9e3bb6b2011-05-24 07:51:27 +0200718 !access_ok(VERIFY_WRITE,
719 (void __user *)(unsigned long)mem->userspace_addr,
720 mem->memory_size)))
Hollis Blanchard78749802008-11-07 13:32:12 -0600721 goto out;
Xiao Guangrong93a5cef2011-11-24 17:37:48 +0800722 if (mem->slot >= KVM_MEM_SLOTS_NUM)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800723 goto out;
724 if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
725 goto out;
726
Xiao Guangrong28a37542011-11-24 19:04:35 +0800727 memslot = id_to_memslot(kvm->memslots, mem->slot);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800728 base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
729 npages = mem->memory_size >> PAGE_SHIFT;
730
Takuya Yoshikawa660c22c2010-04-13 22:47:24 +0900731 r = -EINVAL;
732 if (npages > KVM_MEM_MAX_NR_PAGES)
733 goto out;
734
Avi Kivity6aa8b732006-12-10 02:21:36 -0800735 if (!npages)
736 mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES;
737
Avi Kivity6aa8b732006-12-10 02:21:36 -0800738 new = old = *memslot;
739
Avi Kivitye36d96f2010-06-21 10:56:36 +0300740 new.id = mem->slot;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800741 new.base_gfn = base_gfn;
742 new.npages = npages;
743 new.flags = mem->flags;
744
745 /* Disallow changing a memory slot's size. */
746 r = -EINVAL;
747 if (npages && old.npages && npages != old.npages)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800748 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800749
750 /* Check for overlaps */
751 r = -EEXIST;
752 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200753 struct kvm_memory_slot *s = &kvm->memslots->memslots[i];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800754
Jan Kiszka4cd481f2009-04-13 11:59:32 +0200755 if (s == memslot || !s->npages)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800756 continue;
757 if (!((base_gfn + npages <= s->base_gfn) ||
758 (base_gfn >= s->base_gfn + s->npages)))
Sheng Yangf78e0e22007-10-29 09:40:42 +0800759 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800760 }
Avi Kivity6aa8b732006-12-10 02:21:36 -0800761
Avi Kivity6aa8b732006-12-10 02:21:36 -0800762 /* Free page dirty bitmap if unneeded */
763 if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
Al Viro8b6d44c2007-02-09 16:38:40 +0000764 new.dirty_bitmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800765
766 r = -ENOMEM;
767
768 /* Allocate if a slot is being created */
Takuya Yoshikawa189a2f72012-02-08 13:01:09 +0900769 if (npages && !old.npages) {
770 new.user_alloc = user_alloc;
771 new.userspace_addr = mem->userspace_addr;
Carsten Otteeff01142008-06-27 15:05:31 +0200772#ifndef CONFIG_S390
Takuya Yoshikawa26535032010-11-02 10:49:34 +0900773 new.rmap = vzalloc(npages * sizeof(*new.rmap));
Izik Eidus290fc382007-09-27 14:11:22 +0200774 if (!new.rmap)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800775 goto out_free;
Takuya Yoshikawa189a2f72012-02-08 13:01:09 +0900776#endif /* not defined CONFIG_S390 */
Takuya Yoshikawadb3fe4e2012-02-08 13:02:18 +0900777 if (kvm_arch_create_memslot(&new, npages))
778 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800779 }
Joerg Roedelec04b262009-06-19 15:16:23 +0200780
Avi Kivity6aa8b732006-12-10 02:21:36 -0800781 /* Allocate page dirty bitmap if needed */
782 if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900783 if (kvm_create_dirty_bitmap(&new) < 0)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800784 goto out_free;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200785 /* destroy any largepage mappings for dirty tracking */
Avi Kivity6aa8b732006-12-10 02:21:36 -0800786 }
787
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200788 if (!npages) {
Xiao Guangrong28a37542011-11-24 19:04:35 +0800789 struct kvm_memory_slot *slot;
790
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200791 r = -ENOMEM;
Thomas Meyer6da64fd2011-11-08 20:32:19 +0100792 slots = kmemdup(kvm->memslots, sizeof(struct kvm_memslots),
793 GFP_KERNEL);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200794 if (!slots)
795 goto out_free;
Xiao Guangrong28a37542011-11-24 19:04:35 +0800796 slot = id_to_memslot(slots, mem->slot);
797 slot->flags |= KVM_MEMSLOT_INVALID;
798
Xiao Guangrongbe593d62011-11-24 17:38:24 +0800799 update_memslots(slots, NULL);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200800
801 old_memslots = kvm->memslots;
802 rcu_assign_pointer(kvm->memslots, slots);
803 synchronize_srcu_expedited(&kvm->srcu);
804 /* From this point no new shadow pages pointing to a deleted
805 * memslot will be created.
806 *
807 * validation of sp->gfn happens in:
808 * - gfn_to_hva (kvm_read_guest, gfn_to_pfn)
809 * - kvm_is_visible_gfn (mmu_check_roots)
810 */
Marcelo Tosatti34d4cb82008-07-10 20:49:31 -0300811 kvm_arch_flush_shadow(kvm);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200812 kfree(old_memslots);
813 }
Marcelo Tosatti34d4cb82008-07-10 20:49:31 -0300814
Marcelo Tosattif7784b82009-12-23 14:35:18 -0200815 r = kvm_arch_prepare_memory_region(kvm, &new, old, mem, user_alloc);
816 if (r)
817 goto out_free;
818
Alex Williamson32f6daa2012-04-11 09:51:49 -0600819 /* map/unmap the pages in iommu page table */
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200820 if (npages) {
821 r = kvm_iommu_map_pages(kvm, &new);
822 if (r)
823 goto out_free;
Alex Williamson32f6daa2012-04-11 09:51:49 -0600824 } else
825 kvm_iommu_unmap_pages(kvm, &old);
Andrea Arcangeli604b38a2008-07-25 16:32:03 +0200826
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200827 r = -ENOMEM;
Thomas Meyer6da64fd2011-11-08 20:32:19 +0100828 slots = kmemdup(kvm->memslots, sizeof(struct kvm_memslots),
829 GFP_KERNEL);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200830 if (!slots)
831 goto out_free;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200832
833 /* actual memory is freed via old in kvm_free_physmem_slot below */
834 if (!npages) {
835 new.rmap = NULL;
836 new.dirty_bitmap = NULL;
Takuya Yoshikawadb3fe4e2012-02-08 13:02:18 +0900837 memset(&new.arch, 0, sizeof(new.arch));
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200838 }
839
Xiao Guangrongbe593d62011-11-24 17:38:24 +0800840 update_memslots(slots, &new);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200841 old_memslots = kvm->memslots;
842 rcu_assign_pointer(kvm->memslots, slots);
843 synchronize_srcu_expedited(&kvm->srcu);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800844
Marcelo Tosattif7784b82009-12-23 14:35:18 -0200845 kvm_arch_commit_memory_region(kvm, mem, old, user_alloc);
Zhang Xiantao3ad82a72007-11-20 13:11:38 +0800846
Xiao Guangrongce88dec2011-07-12 03:33:44 +0800847 /*
848 * If the new memory slot is created, we need to clear all
849 * mmio sptes.
850 */
851 if (npages && old.base_gfn != mem->guest_phys_addr >> PAGE_SHIFT)
852 kvm_arch_flush_shadow(kvm);
853
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200854 kvm_free_physmem_slot(&old, &new);
855 kfree(old_memslots);
856
Avi Kivity6aa8b732006-12-10 02:21:36 -0800857 return 0;
858
Sheng Yangf78e0e22007-10-29 09:40:42 +0800859out_free:
Avi Kivity6aa8b732006-12-10 02:21:36 -0800860 kvm_free_physmem_slot(&new, &old);
861out:
862 return r;
Izik Eidus210c7c42007-10-24 23:52:57 +0200863
864}
Sheng Yangf78e0e22007-10-29 09:40:42 +0800865EXPORT_SYMBOL_GPL(__kvm_set_memory_region);
866
867int kvm_set_memory_region(struct kvm *kvm,
868 struct kvm_userspace_memory_region *mem,
869 int user_alloc)
870{
871 int r;
872
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200873 mutex_lock(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +0800874 r = __kvm_set_memory_region(kvm, mem, user_alloc);
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200875 mutex_unlock(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +0800876 return r;
877}
Izik Eidus210c7c42007-10-24 23:52:57 +0200878EXPORT_SYMBOL_GPL(kvm_set_memory_region);
879
Carsten Otte1fe779f2007-10-29 16:08:35 +0100880int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
881 struct
882 kvm_userspace_memory_region *mem,
883 int user_alloc)
Izik Eidus210c7c42007-10-24 23:52:57 +0200884{
Izik Eiduse0d62c72007-10-24 23:57:46 +0200885 if (mem->slot >= KVM_MEMORY_SLOTS)
886 return -EINVAL;
Izik Eidus210c7c42007-10-24 23:52:57 +0200887 return kvm_set_memory_region(kvm, mem, user_alloc);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800888}
889
Zhang Xiantao5bb064d2007-11-18 20:29:43 +0800890int kvm_get_dirty_log(struct kvm *kvm,
891 struct kvm_dirty_log *log, int *is_dirty)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800892{
893 struct kvm_memory_slot *memslot;
894 int r, i;
Takuya Yoshikawa87bf6e72010-04-12 19:35:35 +0900895 unsigned long n;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800896 unsigned long any = 0;
897
Avi Kivity6aa8b732006-12-10 02:21:36 -0800898 r = -EINVAL;
899 if (log->slot >= KVM_MEMORY_SLOTS)
900 goto out;
901
Xiao Guangrong28a37542011-11-24 19:04:35 +0800902 memslot = id_to_memslot(kvm->memslots, log->slot);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800903 r = -ENOENT;
904 if (!memslot->dirty_bitmap)
905 goto out;
906
Takuya Yoshikawa87bf6e72010-04-12 19:35:35 +0900907 n = kvm_dirty_bitmap_bytes(memslot);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800908
Uri Lublincd1a4a92007-02-22 16:43:09 +0200909 for (i = 0; !any && i < n/sizeof(long); ++i)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800910 any = memslot->dirty_bitmap[i];
911
912 r = -EFAULT;
913 if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
914 goto out;
915
Zhang Xiantao5bb064d2007-11-18 20:29:43 +0800916 if (any)
917 *is_dirty = 1;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800918
919 r = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800920out:
Avi Kivity6aa8b732006-12-10 02:21:36 -0800921 return r;
922}
923
Takuya Yoshikawadb3fe4e2012-02-08 13:02:18 +0900924bool kvm_largepages_enabled(void)
925{
926 return largepages_enabled;
927}
928
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300929void kvm_disable_largepages(void)
930{
931 largepages_enabled = false;
932}
933EXPORT_SYMBOL_GPL(kvm_disable_largepages);
934
Izik Eiduscea7bb22007-10-17 19:17:48 +0200935int is_error_page(struct page *page)
936{
Gleb Natapovedba23e2010-07-07 20:16:45 +0300937 return page == bad_page || page == hwpoison_page || page == fault_page;
Izik Eiduscea7bb22007-10-17 19:17:48 +0200938}
939EXPORT_SYMBOL_GPL(is_error_page);
940
Anthony Liguori35149e22008-04-02 14:46:56 -0500941int is_error_pfn(pfn_t pfn)
942{
Gleb Natapovedba23e2010-07-07 20:16:45 +0300943 return pfn == bad_pfn || pfn == hwpoison_pfn || pfn == fault_pfn;
Anthony Liguori35149e22008-04-02 14:46:56 -0500944}
945EXPORT_SYMBOL_GPL(is_error_pfn);
946
Huang Yingbf998152010-05-31 14:28:19 +0800947int is_hwpoison_pfn(pfn_t pfn)
948{
949 return pfn == hwpoison_pfn;
950}
951EXPORT_SYMBOL_GPL(is_hwpoison_pfn);
952
Gleb Natapovedba23e2010-07-07 20:16:45 +0300953int is_fault_pfn(pfn_t pfn)
954{
955 return pfn == fault_pfn;
956}
957EXPORT_SYMBOL_GPL(is_fault_pfn);
958
Xiao Guangrongfce92dc2011-07-12 03:28:54 +0800959int is_noslot_pfn(pfn_t pfn)
960{
961 return pfn == bad_pfn;
962}
963EXPORT_SYMBOL_GPL(is_noslot_pfn);
964
965int is_invalid_pfn(pfn_t pfn)
966{
967 return pfn == hwpoison_pfn || pfn == fault_pfn;
968}
969EXPORT_SYMBOL_GPL(is_invalid_pfn);
970
Izik Eidusf9d46eb2007-11-11 22:02:22 +0200971static inline unsigned long bad_hva(void)
972{
973 return PAGE_OFFSET;
974}
975
976int kvm_is_error_hva(unsigned long addr)
977{
978 return addr == bad_hva();
979}
980EXPORT_SYMBOL_GPL(kvm_is_error_hva);
981
Gleb Natapov49c77542010-10-18 15:22:23 +0200982struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn)
983{
984 return __gfn_to_memslot(kvm_memslots(kvm), gfn);
985}
Avi Kivitya1f4d3952010-06-21 11:44:20 +0300986EXPORT_SYMBOL_GPL(gfn_to_memslot);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800987
Izik Eiduse0d62c72007-10-24 23:57:46 +0200988int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
989{
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800990 struct kvm_memory_slot *memslot = gfn_to_memslot(kvm, gfn);
Izik Eiduse0d62c72007-10-24 23:57:46 +0200991
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800992 if (!memslot || memslot->id >= KVM_MEMORY_SLOTS ||
993 memslot->flags & KVM_MEMSLOT_INVALID)
994 return 0;
Izik Eiduse0d62c72007-10-24 23:57:46 +0200995
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800996 return 1;
Izik Eiduse0d62c72007-10-24 23:57:46 +0200997}
998EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);
999
Joerg Roedel8f0b1ab2010-01-28 12:37:56 +01001000unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn)
1001{
1002 struct vm_area_struct *vma;
1003 unsigned long addr, size;
1004
1005 size = PAGE_SIZE;
1006
1007 addr = gfn_to_hva(kvm, gfn);
1008 if (kvm_is_error_hva(addr))
1009 return PAGE_SIZE;
1010
1011 down_read(&current->mm->mmap_sem);
1012 vma = find_vma(current->mm, addr);
1013 if (!vma)
1014 goto out;
1015
1016 size = vma_kernel_pagesize(vma);
1017
1018out:
1019 up_read(&current->mm->mmap_sem);
1020
1021 return size;
1022}
1023
Gleb Natapov49c77542010-10-18 15:22:23 +02001024static unsigned long gfn_to_hva_many(struct kvm_memory_slot *slot, gfn_t gfn,
Xiao Guangrong48987782010-08-22 19:11:43 +08001025 gfn_t *nr_pages)
Izik Eidus539cb662007-11-11 22:05:04 +02001026{
Marcelo Tosattibc6678a2009-12-23 14:35:21 -02001027 if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
Izik Eidus539cb662007-11-11 22:05:04 +02001028 return bad_hva();
Xiao Guangrong48987782010-08-22 19:11:43 +08001029
1030 if (nr_pages)
1031 *nr_pages = slot->npages - (gfn - slot->base_gfn);
1032
Takuya Yoshikawaf5c98032010-02-25 11:33:19 +09001033 return gfn_to_hva_memslot(slot, gfn);
Izik Eidus539cb662007-11-11 22:05:04 +02001034}
Xiao Guangrong48987782010-08-22 19:11:43 +08001035
1036unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
1037{
Gleb Natapov49c77542010-10-18 15:22:23 +02001038 return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL);
Xiao Guangrong48987782010-08-22 19:11:43 +08001039}
Sheng Yang0d150292008-04-25 21:44:50 +08001040EXPORT_SYMBOL_GPL(gfn_to_hva);
Izik Eidus539cb662007-11-11 22:05:04 +02001041
Gleb Natapov80300892010-10-19 18:13:41 +02001042static pfn_t get_fault_pfn(void)
1043{
1044 get_page(fault_page);
1045 return fault_pfn;
1046}
1047
Gleb Natapov0857b9e2011-02-01 13:21:47 +02001048int get_user_page_nowait(struct task_struct *tsk, struct mm_struct *mm,
1049 unsigned long start, int write, struct page **page)
1050{
1051 int flags = FOLL_TOUCH | FOLL_NOWAIT | FOLL_HWPOISON | FOLL_GET;
1052
1053 if (write)
1054 flags |= FOLL_WRITE;
1055
1056 return __get_user_pages(tsk, mm, start, 1, flags, page, NULL, NULL);
1057}
1058
Huang Yingfafc3db2011-01-30 11:15:49 +08001059static inline int check_user_page_hwpoison(unsigned long addr)
1060{
1061 int rc, flags = FOLL_TOUCH | FOLL_HWPOISON | FOLL_WRITE;
1062
1063 rc = __get_user_pages(current, current->mm, addr, 1,
1064 flags, NULL, NULL, NULL);
1065 return rc == -EHWPOISON;
1066}
1067
Gleb Natapovaf585b92010-10-14 11:22:46 +02001068static pfn_t hva_to_pfn(struct kvm *kvm, unsigned long addr, bool atomic,
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001069 bool *async, bool write_fault, bool *writable)
Avi Kivity954bbbc2007-03-30 14:02:32 +03001070{
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001071 struct page *page[1];
Gleb Natapovaf585b92010-10-14 11:22:46 +02001072 int npages = 0;
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001073 pfn_t pfn;
Avi Kivity954bbbc2007-03-30 14:02:32 +03001074
Gleb Natapovaf585b92010-10-14 11:22:46 +02001075 /* we can do it either atomically or asynchronously, not both */
1076 BUG_ON(atomic && async);
1077
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001078 BUG_ON(!write_fault && !writable);
1079
1080 if (writable)
1081 *writable = true;
1082
Gleb Natapovaf585b92010-10-14 11:22:46 +02001083 if (atomic || async)
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001084 npages = __get_user_pages_fast(addr, 1, 1, page);
Gleb Natapovaf585b92010-10-14 11:22:46 +02001085
1086 if (unlikely(npages != 1) && !atomic) {
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001087 might_sleep();
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001088
1089 if (writable)
1090 *writable = write_fault;
1091
Gleb Natapov0857b9e2011-02-01 13:21:47 +02001092 if (async) {
1093 down_read(&current->mm->mmap_sem);
1094 npages = get_user_page_nowait(current, current->mm,
1095 addr, write_fault, page);
1096 up_read(&current->mm->mmap_sem);
1097 } else
1098 npages = get_user_pages_fast(addr, 1, write_fault,
1099 page);
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001100
1101 /* map read fault as writable if possible */
1102 if (unlikely(!write_fault) && npages == 1) {
1103 struct page *wpage[1];
1104
1105 npages = __get_user_pages_fast(addr, 1, 1, wpage);
1106 if (npages == 1) {
1107 *writable = true;
1108 put_page(page[0]);
1109 page[0] = wpage[0];
1110 }
1111 npages = 1;
1112 }
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001113 }
Izik Eidus539cb662007-11-11 22:05:04 +02001114
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001115 if (unlikely(npages != 1)) {
1116 struct vm_area_struct *vma;
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001117
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001118 if (atomic)
Gleb Natapov80300892010-10-19 18:13:41 +02001119 return get_fault_pfn();
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001120
Huang Yingbbeb3402010-06-22 14:23:11 +08001121 down_read(&current->mm->mmap_sem);
Gleb Natapov0857b9e2011-02-01 13:21:47 +02001122 if (npages == -EHWPOISON ||
1123 (!async && check_user_page_hwpoison(addr))) {
Huang Yingbbeb3402010-06-22 14:23:11 +08001124 up_read(&current->mm->mmap_sem);
Huang Yingbf998152010-05-31 14:28:19 +08001125 get_page(hwpoison_page);
1126 return page_to_pfn(hwpoison_page);
1127 }
1128
Gleb Natapov80300892010-10-19 18:13:41 +02001129 vma = find_vma_intersection(current->mm, addr, addr+1);
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001130
Gleb Natapov80300892010-10-19 18:13:41 +02001131 if (vma == NULL)
1132 pfn = get_fault_pfn();
1133 else if ((vma->vm_flags & VM_PFNMAP)) {
1134 pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) +
1135 vma->vm_pgoff;
1136 BUG_ON(!kvm_is_mmio_pfn(pfn));
1137 } else {
1138 if (async && (vma->vm_flags & VM_WRITE))
Gleb Natapovaf585b92010-10-14 11:22:46 +02001139 *async = true;
Gleb Natapov80300892010-10-19 18:13:41 +02001140 pfn = get_fault_pfn();
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001141 }
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001142 up_read(&current->mm->mmap_sem);
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001143 } else
1144 pfn = page_to_pfn(page[0]);
1145
1146 return pfn;
Anthony Liguori35149e22008-04-02 14:46:56 -05001147}
1148
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001149pfn_t hva_to_pfn_atomic(struct kvm *kvm, unsigned long addr)
1150{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001151 return hva_to_pfn(kvm, addr, true, NULL, true, NULL);
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001152}
1153EXPORT_SYMBOL_GPL(hva_to_pfn_atomic);
1154
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001155static pfn_t __gfn_to_pfn(struct kvm *kvm, gfn_t gfn, bool atomic, bool *async,
1156 bool write_fault, bool *writable)
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001157{
1158 unsigned long addr;
1159
Gleb Natapovaf585b92010-10-14 11:22:46 +02001160 if (async)
1161 *async = false;
1162
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001163 addr = gfn_to_hva(kvm, gfn);
1164 if (kvm_is_error_hva(addr)) {
1165 get_page(bad_page);
1166 return page_to_pfn(bad_page);
1167 }
1168
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001169 return hva_to_pfn(kvm, addr, atomic, async, write_fault, writable);
Xiao Guangrong365fb3f2010-08-28 19:24:13 +08001170}
1171
1172pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn)
1173{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001174 return __gfn_to_pfn(kvm, gfn, true, NULL, true, NULL);
Xiao Guangrong365fb3f2010-08-28 19:24:13 +08001175}
1176EXPORT_SYMBOL_GPL(gfn_to_pfn_atomic);
1177
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001178pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
1179 bool write_fault, bool *writable)
Gleb Natapovaf585b92010-10-14 11:22:46 +02001180{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001181 return __gfn_to_pfn(kvm, gfn, false, async, write_fault, writable);
Gleb Natapovaf585b92010-10-14 11:22:46 +02001182}
1183EXPORT_SYMBOL_GPL(gfn_to_pfn_async);
1184
Xiao Guangrong365fb3f2010-08-28 19:24:13 +08001185pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
1186{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001187 return __gfn_to_pfn(kvm, gfn, false, NULL, true, NULL);
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001188}
Anthony Liguori35149e22008-04-02 14:46:56 -05001189EXPORT_SYMBOL_GPL(gfn_to_pfn);
1190
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001191pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1192 bool *writable)
1193{
1194 return __gfn_to_pfn(kvm, gfn, false, NULL, write_fault, writable);
1195}
1196EXPORT_SYMBOL_GPL(gfn_to_pfn_prot);
1197
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001198pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
1199 struct kvm_memory_slot *slot, gfn_t gfn)
1200{
1201 unsigned long addr = gfn_to_hva_memslot(slot, gfn);
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001202 return hva_to_pfn(kvm, addr, false, NULL, true, NULL);
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001203}
1204
Xiao Guangrong48987782010-08-22 19:11:43 +08001205int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
1206 int nr_pages)
1207{
1208 unsigned long addr;
1209 gfn_t entry;
1210
Gleb Natapov49c77542010-10-18 15:22:23 +02001211 addr = gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, &entry);
Xiao Guangrong48987782010-08-22 19:11:43 +08001212 if (kvm_is_error_hva(addr))
1213 return -1;
1214
1215 if (entry < nr_pages)
1216 return 0;
1217
1218 return __get_user_pages_fast(addr, nr_pages, 1, pages);
1219}
1220EXPORT_SYMBOL_GPL(gfn_to_page_many_atomic);
1221
Anthony Liguori35149e22008-04-02 14:46:56 -05001222struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
1223{
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001224 pfn_t pfn;
1225
1226 pfn = gfn_to_pfn(kvm, gfn);
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001227 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001228 return pfn_to_page(pfn);
1229
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001230 WARN_ON(kvm_is_mmio_pfn(pfn));
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001231
1232 get_page(bad_page);
1233 return bad_page;
Avi Kivity954bbbc2007-03-30 14:02:32 +03001234}
Anthony Liguoriaab61cc2007-10-29 15:15:20 -05001235
Avi Kivity954bbbc2007-03-30 14:02:32 +03001236EXPORT_SYMBOL_GPL(gfn_to_page);
1237
Izik Eidusb4231d62007-11-20 11:49:33 +02001238void kvm_release_page_clean(struct page *page)
1239{
Anthony Liguori35149e22008-04-02 14:46:56 -05001240 kvm_release_pfn_clean(page_to_pfn(page));
Izik Eidusb4231d62007-11-20 11:49:33 +02001241}
1242EXPORT_SYMBOL_GPL(kvm_release_page_clean);
1243
Anthony Liguori35149e22008-04-02 14:46:56 -05001244void kvm_release_pfn_clean(pfn_t pfn)
1245{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001246 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001247 put_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001248}
1249EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);
1250
Izik Eidusb4231d62007-11-20 11:49:33 +02001251void kvm_release_page_dirty(struct page *page)
Izik Eidus8a7ae052007-10-18 11:09:33 +02001252{
Anthony Liguori35149e22008-04-02 14:46:56 -05001253 kvm_release_pfn_dirty(page_to_pfn(page));
Izik Eidus8a7ae052007-10-18 11:09:33 +02001254}
Izik Eidusb4231d62007-11-20 11:49:33 +02001255EXPORT_SYMBOL_GPL(kvm_release_page_dirty);
Izik Eidus8a7ae052007-10-18 11:09:33 +02001256
Anthony Liguori35149e22008-04-02 14:46:56 -05001257void kvm_release_pfn_dirty(pfn_t pfn)
1258{
1259 kvm_set_pfn_dirty(pfn);
1260 kvm_release_pfn_clean(pfn);
1261}
1262EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
1263
1264void kvm_set_page_dirty(struct page *page)
1265{
1266 kvm_set_pfn_dirty(page_to_pfn(page));
1267}
1268EXPORT_SYMBOL_GPL(kvm_set_page_dirty);
1269
1270void kvm_set_pfn_dirty(pfn_t pfn)
1271{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001272 if (!kvm_is_mmio_pfn(pfn)) {
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001273 struct page *page = pfn_to_page(pfn);
1274 if (!PageReserved(page))
1275 SetPageDirty(page);
1276 }
Anthony Liguori35149e22008-04-02 14:46:56 -05001277}
1278EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);
1279
1280void kvm_set_pfn_accessed(pfn_t pfn)
1281{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001282 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001283 mark_page_accessed(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001284}
1285EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);
1286
1287void kvm_get_pfn(pfn_t pfn)
1288{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001289 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001290 get_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001291}
1292EXPORT_SYMBOL_GPL(kvm_get_pfn);
1293
Izik Eidus195aefd2007-10-01 22:14:18 +02001294static int next_segment(unsigned long len, int offset)
1295{
1296 if (len > PAGE_SIZE - offset)
1297 return PAGE_SIZE - offset;
1298 else
1299 return len;
1300}
1301
1302int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
1303 int len)
1304{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001305 int r;
1306 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001307
Izik Eiduse0506bc2007-11-11 22:10:22 +02001308 addr = gfn_to_hva(kvm, gfn);
1309 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001310 return -EFAULT;
Takuya Yoshikawafa3d3152011-05-07 16:35:38 +09001311 r = __copy_from_user(data, (void __user *)addr + offset, len);
Izik Eiduse0506bc2007-11-11 22:10:22 +02001312 if (r)
1313 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001314 return 0;
1315}
1316EXPORT_SYMBOL_GPL(kvm_read_guest_page);
1317
1318int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len)
1319{
1320 gfn_t gfn = gpa >> PAGE_SHIFT;
1321 int seg;
1322 int offset = offset_in_page(gpa);
1323 int ret;
1324
1325 while ((seg = next_segment(len, offset)) != 0) {
1326 ret = kvm_read_guest_page(kvm, gfn, data, offset, seg);
1327 if (ret < 0)
1328 return ret;
1329 offset = 0;
1330 len -= seg;
1331 data += seg;
1332 ++gfn;
1333 }
1334 return 0;
1335}
1336EXPORT_SYMBOL_GPL(kvm_read_guest);
1337
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001338int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
1339 unsigned long len)
1340{
1341 int r;
1342 unsigned long addr;
1343 gfn_t gfn = gpa >> PAGE_SHIFT;
1344 int offset = offset_in_page(gpa);
1345
1346 addr = gfn_to_hva(kvm, gfn);
1347 if (kvm_is_error_hva(addr))
1348 return -EFAULT;
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001349 pagefault_disable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001350 r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001351 pagefault_enable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001352 if (r)
1353 return -EFAULT;
1354 return 0;
1355}
1356EXPORT_SYMBOL(kvm_read_guest_atomic);
1357
Izik Eidus195aefd2007-10-01 22:14:18 +02001358int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
1359 int offset, int len)
1360{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001361 int r;
1362 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001363
Izik Eiduse0506bc2007-11-11 22:10:22 +02001364 addr = gfn_to_hva(kvm, gfn);
1365 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001366 return -EFAULT;
Xiao Guangrong8b0cedf2011-05-15 23:22:04 +08001367 r = __copy_to_user((void __user *)addr + offset, data, len);
Izik Eiduse0506bc2007-11-11 22:10:22 +02001368 if (r)
1369 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001370 mark_page_dirty(kvm, gfn);
1371 return 0;
1372}
1373EXPORT_SYMBOL_GPL(kvm_write_guest_page);
1374
1375int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
1376 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_write_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}
1394
Gleb Natapov49c77542010-10-18 15:22:23 +02001395int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1396 gpa_t gpa)
1397{
1398 struct kvm_memslots *slots = kvm_memslots(kvm);
1399 int offset = offset_in_page(gpa);
1400 gfn_t gfn = gpa >> PAGE_SHIFT;
1401
1402 ghc->gpa = gpa;
1403 ghc->generation = slots->generation;
Paul Mackerras9d4cba72012-01-12 20:09:51 +00001404 ghc->memslot = gfn_to_memslot(kvm, gfn);
Gleb Natapov49c77542010-10-18 15:22:23 +02001405 ghc->hva = gfn_to_hva_many(ghc->memslot, gfn, NULL);
1406 if (!kvm_is_error_hva(ghc->hva))
1407 ghc->hva += offset;
1408 else
1409 return -EFAULT;
1410
1411 return 0;
1412}
1413EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init);
1414
1415int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1416 void *data, unsigned long len)
1417{
1418 struct kvm_memslots *slots = kvm_memslots(kvm);
1419 int r;
1420
1421 if (slots->generation != ghc->generation)
1422 kvm_gfn_to_hva_cache_init(kvm, ghc, ghc->gpa);
1423
1424 if (kvm_is_error_hva(ghc->hva))
1425 return -EFAULT;
1426
Xiao Guangrong8b0cedf2011-05-15 23:22:04 +08001427 r = __copy_to_user((void __user *)ghc->hva, data, len);
Gleb Natapov49c77542010-10-18 15:22:23 +02001428 if (r)
1429 return -EFAULT;
1430 mark_page_dirty_in_slot(kvm, ghc->memslot, ghc->gpa >> PAGE_SHIFT);
1431
1432 return 0;
1433}
1434EXPORT_SYMBOL_GPL(kvm_write_guest_cached);
1435
Gleb Natapove03b6442011-07-11 15:28:11 -04001436int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1437 void *data, unsigned long len)
1438{
1439 struct kvm_memslots *slots = kvm_memslots(kvm);
1440 int r;
1441
1442 if (slots->generation != ghc->generation)
1443 kvm_gfn_to_hva_cache_init(kvm, ghc, ghc->gpa);
1444
1445 if (kvm_is_error_hva(ghc->hva))
1446 return -EFAULT;
1447
1448 r = __copy_from_user(data, (void __user *)ghc->hva, len);
1449 if (r)
1450 return -EFAULT;
1451
1452 return 0;
1453}
1454EXPORT_SYMBOL_GPL(kvm_read_guest_cached);
1455
Izik Eidus195aefd2007-10-01 22:14:18 +02001456int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
1457{
Heiko Carstens3bcc8a82010-10-27 17:21:21 +02001458 return kvm_write_guest_page(kvm, gfn, (const void *) empty_zero_page,
1459 offset, len);
Izik Eidus195aefd2007-10-01 22:14:18 +02001460}
1461EXPORT_SYMBOL_GPL(kvm_clear_guest_page);
1462
1463int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len)
1464{
1465 gfn_t gfn = gpa >> PAGE_SHIFT;
1466 int seg;
1467 int offset = offset_in_page(gpa);
1468 int ret;
1469
1470 while ((seg = next_segment(len, offset)) != 0) {
1471 ret = kvm_clear_guest_page(kvm, gfn, offset, seg);
1472 if (ret < 0)
1473 return ret;
1474 offset = 0;
1475 len -= seg;
1476 ++gfn;
1477 }
1478 return 0;
1479}
1480EXPORT_SYMBOL_GPL(kvm_clear_guest);
1481
Gleb Natapov49c77542010-10-18 15:22:23 +02001482void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
1483 gfn_t gfn)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001484{
Rusty Russell7e9d6192007-07-31 20:41:14 +10001485 if (memslot && memslot->dirty_bitmap) {
1486 unsigned long rel_gfn = gfn - memslot->base_gfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001487
Takuya Yoshikawa93474b22012-03-01 19:34:45 +09001488 /* TODO: introduce set_bit_le() and use it */
1489 test_and_set_bit_le(rel_gfn, memslot->dirty_bitmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001490 }
1491}
1492
Gleb Natapov49c77542010-10-18 15:22:23 +02001493void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
1494{
1495 struct kvm_memory_slot *memslot;
1496
1497 memslot = gfn_to_memslot(kvm, gfn);
1498 mark_page_dirty_in_slot(kvm, memslot, gfn);
1499}
1500
Eddie Dongb6958ce2007-07-18 12:15:21 +03001501/*
1502 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
1503 */
Hollis Blanchard8776e512007-10-31 17:24:24 -05001504void kvm_vcpu_block(struct kvm_vcpu *vcpu)
Eddie Dongb6958ce2007-07-18 12:15:21 +03001505{
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001506 DEFINE_WAIT(wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001507
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001508 for (;;) {
1509 prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001510
Gleb Natapova1b37102009-07-09 15:33:52 +03001511 if (kvm_arch_vcpu_runnable(vcpu)) {
Avi Kivitya8eeb042010-05-10 12:34:53 +03001512 kvm_make_request(KVM_REQ_UNHALT, vcpu);
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001513 break;
Marcelo Tosattid7690172008-09-08 15:23:48 -03001514 }
Gleb Natapov09cec752009-03-23 15:11:44 +02001515 if (kvm_cpu_has_pending_timer(vcpu))
1516 break;
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001517 if (signal_pending(current))
1518 break;
1519
Eddie Dongb6958ce2007-07-18 12:15:21 +03001520 schedule();
Eddie Dongb6958ce2007-07-18 12:15:21 +03001521 }
1522
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001523 finish_wait(&vcpu->wq, &wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001524}
1525
Marcelo Tosatti8c847802012-03-14 17:58:48 -03001526#ifndef CONFIG_S390
Christoffer Dallb6d33832012-03-08 16:44:24 -05001527/*
1528 * Kick a sleeping VCPU, or a guest VCPU in guest mode, into host kernel mode.
1529 */
1530void kvm_vcpu_kick(struct kvm_vcpu *vcpu)
1531{
1532 int me;
1533 int cpu = vcpu->cpu;
1534 wait_queue_head_t *wqp;
1535
1536 wqp = kvm_arch_vcpu_wq(vcpu);
1537 if (waitqueue_active(wqp)) {
1538 wake_up_interruptible(wqp);
1539 ++vcpu->stat.halt_wakeup;
1540 }
1541
1542 me = get_cpu();
1543 if (cpu != me && (unsigned)cpu < nr_cpu_ids && cpu_online(cpu))
1544 if (kvm_arch_vcpu_should_kick(vcpu))
1545 smp_send_reschedule(cpu);
1546 put_cpu();
1547}
Marcelo Tosatti8c847802012-03-14 17:58:48 -03001548#endif /* !CONFIG_S390 */
Christoffer Dallb6d33832012-03-08 16:44:24 -05001549
Avi Kivity6aa8b732006-12-10 02:21:36 -08001550void kvm_resched(struct kvm_vcpu *vcpu)
1551{
Yaozu Dong3fca0362007-04-25 16:49:19 +03001552 if (!need_resched())
1553 return;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001554 cond_resched();
Avi Kivity6aa8b732006-12-10 02:21:36 -08001555}
1556EXPORT_SYMBOL_GPL(kvm_resched);
1557
Konstantin Weitz41628d32012-04-25 15:30:38 +02001558bool kvm_vcpu_yield_to(struct kvm_vcpu *target)
1559{
1560 struct pid *pid;
1561 struct task_struct *task = NULL;
1562
1563 rcu_read_lock();
1564 pid = rcu_dereference(target->pid);
1565 if (pid)
1566 task = get_pid_task(target->pid, PIDTYPE_PID);
1567 rcu_read_unlock();
1568 if (!task)
1569 return false;
1570 if (task->flags & PF_VCPU) {
1571 put_task_struct(task);
1572 return false;
1573 }
1574 if (yield_to(task, 1)) {
1575 put_task_struct(task);
1576 return true;
1577 }
1578 put_task_struct(task);
1579 return false;
1580}
1581EXPORT_SYMBOL_GPL(kvm_vcpu_yield_to);
1582
Rik van Riel217ece62011-02-01 09:53:28 -05001583void kvm_vcpu_on_spin(struct kvm_vcpu *me)
Zhai, Edwind255f4f2009-10-09 18:03:20 +08001584{
Rik van Riel217ece62011-02-01 09:53:28 -05001585 struct kvm *kvm = me->kvm;
1586 struct kvm_vcpu *vcpu;
1587 int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
1588 int yielded = 0;
1589 int pass;
1590 int i;
Zhai, Edwind255f4f2009-10-09 18:03:20 +08001591
Rik van Riel217ece62011-02-01 09:53:28 -05001592 /*
1593 * We boost the priority of a VCPU that is runnable but not
1594 * currently running, because it got preempted by something
1595 * else and called schedule in __vcpu_run. Hopefully that
1596 * VCPU is holding the lock that we need and will release it.
1597 * We approximate round-robin by starting at the last boosted VCPU.
1598 */
1599 for (pass = 0; pass < 2 && !yielded; pass++) {
1600 kvm_for_each_vcpu(i, vcpu, kvm) {
Rik van Riel5cfc2aa2012-06-19 16:51:04 -04001601 if (!pass && i <= last_boosted_vcpu) {
Rik van Riel217ece62011-02-01 09:53:28 -05001602 i = last_boosted_vcpu;
1603 continue;
1604 } else if (pass && i > last_boosted_vcpu)
1605 break;
1606 if (vcpu == me)
1607 continue;
1608 if (waitqueue_active(&vcpu->wq))
1609 continue;
Konstantin Weitz41628d32012-04-25 15:30:38 +02001610 if (kvm_vcpu_yield_to(vcpu)) {
Rik van Riel217ece62011-02-01 09:53:28 -05001611 kvm->last_boosted_vcpu = i;
1612 yielded = 1;
1613 break;
1614 }
Rik van Riel217ece62011-02-01 09:53:28 -05001615 }
1616 }
Zhai, Edwind255f4f2009-10-09 18:03:20 +08001617}
1618EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin);
1619
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001620static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001621{
1622 struct kvm_vcpu *vcpu = vma->vm_file->private_data;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001623 struct page *page;
1624
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001625 if (vmf->pgoff == 0)
Avi Kivity039576c2007-03-20 12:46:50 +02001626 page = virt_to_page(vcpu->run);
Avi Kivity09566762008-01-23 18:14:23 +02001627#ifdef CONFIG_X86
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001628 else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
Zhang Xiantaoad312c72007-12-13 23:50:52 +08001629 page = virt_to_page(vcpu->arch.pio_data);
Avi Kivity09566762008-01-23 18:14:23 +02001630#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02001631#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1632 else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
1633 page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
1634#endif
Avi Kivity039576c2007-03-20 12:46:50 +02001635 else
Carsten Otte5b1c1492012-01-04 10:25:23 +01001636 return kvm_arch_vcpu_fault(vcpu, vmf);
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001637 get_page(page);
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001638 vmf->page = page;
1639 return 0;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001640}
1641
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +04001642static const struct vm_operations_struct kvm_vcpu_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001643 .fault = kvm_vcpu_fault,
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001644};
1645
1646static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma)
1647{
1648 vma->vm_ops = &kvm_vcpu_vm_ops;
1649 return 0;
1650}
1651
Avi Kivitybccf2152007-02-21 18:04:26 +02001652static int kvm_vcpu_release(struct inode *inode, struct file *filp)
1653{
1654 struct kvm_vcpu *vcpu = filp->private_data;
1655
Al Viro66c0b392008-04-19 20:33:56 +01001656 kvm_put_kvm(vcpu->kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001657 return 0;
1658}
1659
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01001660static struct file_operations kvm_vcpu_fops = {
Avi Kivitybccf2152007-02-21 18:04:26 +02001661 .release = kvm_vcpu_release,
1662 .unlocked_ioctl = kvm_vcpu_ioctl,
Alexander Graf1dda6062011-06-08 02:45:37 +02001663#ifdef CONFIG_COMPAT
1664 .compat_ioctl = kvm_vcpu_compat_ioctl,
1665#endif
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001666 .mmap = kvm_vcpu_mmap,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001667 .llseek = noop_llseek,
Avi Kivitybccf2152007-02-21 18:04:26 +02001668};
1669
1670/*
1671 * Allocates an inode for the vcpu.
1672 */
1673static int create_vcpu_fd(struct kvm_vcpu *vcpu)
1674{
Roland Dreier628ff7c2009-12-18 09:41:24 -08001675 return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, O_RDWR);
Avi Kivitybccf2152007-02-21 18:04:26 +02001676}
1677
Avi Kivityc5ea7662007-02-20 18:41:05 +02001678/*
1679 * Creates some virtual cpus. Good luck creating more than one.
1680 */
Gleb Natapov73880c82009-06-09 15:56:28 +03001681static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
Avi Kivityc5ea7662007-02-20 18:41:05 +02001682{
1683 int r;
Gleb Natapov988a2ca2009-06-09 15:56:29 +03001684 struct kvm_vcpu *vcpu, *v;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001685
Gleb Natapov73880c82009-06-09 15:56:28 +03001686 vcpu = kvm_arch_vcpu_create(kvm, id);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001687 if (IS_ERR(vcpu))
1688 return PTR_ERR(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001689
Avi Kivity15ad7142007-07-11 18:17:21 +03001690 preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);
1691
Avi Kivity26e52152007-11-20 15:30:24 +02001692 r = kvm_arch_vcpu_setup(vcpu);
1693 if (r)
Jan Kiszkad7805922011-05-23 10:33:05 +02001694 goto vcpu_destroy;
Avi Kivity26e52152007-11-20 15:30:24 +02001695
Shaohua Li11ec2802007-07-23 14:51:37 +08001696 mutex_lock(&kvm->lock);
Avi Kivity3e515702012-03-05 14:23:29 +02001697 if (!kvm_vcpu_compatible(vcpu)) {
1698 r = -EINVAL;
1699 goto unlock_vcpu_destroy;
1700 }
Gleb Natapov73880c82009-06-09 15:56:28 +03001701 if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) {
1702 r = -EINVAL;
Jan Kiszkad7805922011-05-23 10:33:05 +02001703 goto unlock_vcpu_destroy;
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001704 }
Gleb Natapov73880c82009-06-09 15:56:28 +03001705
Gleb Natapov988a2ca2009-06-09 15:56:29 +03001706 kvm_for_each_vcpu(r, v, kvm)
1707 if (v->vcpu_id == id) {
Gleb Natapov73880c82009-06-09 15:56:28 +03001708 r = -EEXIST;
Jan Kiszkad7805922011-05-23 10:33:05 +02001709 goto unlock_vcpu_destroy;
Gleb Natapov73880c82009-06-09 15:56:28 +03001710 }
1711
1712 BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001713
1714 /* Now it's all set up, let userspace reach it */
Al Viro66c0b392008-04-19 20:33:56 +01001715 kvm_get_kvm(kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001716 r = create_vcpu_fd(vcpu);
Gleb Natapov73880c82009-06-09 15:56:28 +03001717 if (r < 0) {
1718 kvm_put_kvm(kvm);
Jan Kiszkad7805922011-05-23 10:33:05 +02001719 goto unlock_vcpu_destroy;
Gleb Natapov73880c82009-06-09 15:56:28 +03001720 }
1721
1722 kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu;
1723 smp_wmb();
1724 atomic_inc(&kvm->online_vcpus);
1725
Gleb Natapov73880c82009-06-09 15:56:28 +03001726 mutex_unlock(&kvm->lock);
Avi Kivitybccf2152007-02-21 18:04:26 +02001727 return r;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001728
Jan Kiszkad7805922011-05-23 10:33:05 +02001729unlock_vcpu_destroy:
Glauber Costa7d8fece2008-09-17 23:16:59 -03001730 mutex_unlock(&kvm->lock);
Jan Kiszkad7805922011-05-23 10:33:05 +02001731vcpu_destroy:
Hollis Blanchardd40ccc62007-11-19 14:04:43 -06001732 kvm_arch_vcpu_destroy(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001733 return r;
1734}
1735
Avi Kivity1961d272007-03-05 19:46:05 +02001736static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset)
1737{
1738 if (sigset) {
1739 sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP));
1740 vcpu->sigset_active = 1;
1741 vcpu->sigset = *sigset;
1742 } else
1743 vcpu->sigset_active = 0;
1744 return 0;
1745}
1746
Avi Kivitybccf2152007-02-21 18:04:26 +02001747static long kvm_vcpu_ioctl(struct file *filp,
1748 unsigned int ioctl, unsigned long arg)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001749{
Avi Kivitybccf2152007-02-21 18:04:26 +02001750 struct kvm_vcpu *vcpu = filp->private_data;
Al Viro2f3669872007-02-09 16:38:35 +00001751 void __user *argp = (void __user *)arg;
Carsten Otte313a3dc2007-10-11 19:16:52 +02001752 int r;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001753 struct kvm_fpu *fpu = NULL;
1754 struct kvm_sregs *kvm_sregs = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001755
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001756 if (vcpu->kvm->mm != current->mm)
1757 return -EIO;
Avi Kivity2122ff52010-05-13 11:25:04 +03001758
1759#if defined(CONFIG_S390) || defined(CONFIG_PPC)
1760 /*
1761 * Special cases: vcpu ioctls that are asynchronous to vcpu execution,
1762 * so vcpu_load() would break it.
1763 */
1764 if (ioctl == KVM_S390_INTERRUPT || ioctl == KVM_INTERRUPT)
1765 return kvm_arch_vcpu_ioctl(filp, ioctl, arg);
1766#endif
1767
1768
1769 vcpu_load(vcpu);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001770 switch (ioctl) {
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001771 case KVM_RUN:
Avi Kivityf0fe5102007-03-07 13:11:17 +02001772 r = -EINVAL;
1773 if (arg)
1774 goto out;
Hollis Blanchardb6c7a5d2007-11-01 14:16:10 -05001775 r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
Gleb Natapov64be5002010-10-24 16:49:08 +02001776 trace_kvm_userspace_exit(vcpu->run->exit_reason, r);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001777 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001778 case KVM_GET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001779 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001780
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001781 r = -ENOMEM;
1782 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
1783 if (!kvm_regs)
1784 goto out;
1785 r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001786 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001787 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001788 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001789 if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
1790 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001791 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001792out_free1:
1793 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001794 break;
1795 }
1796 case KVM_SET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001797 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001798
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001799 r = -ENOMEM;
Sasha Levinff5c2c02011-12-04 19:36:29 +02001800 kvm_regs = memdup_user(argp, sizeof(*kvm_regs));
1801 if (IS_ERR(kvm_regs)) {
1802 r = PTR_ERR(kvm_regs);
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001803 goto out;
Sasha Levinff5c2c02011-12-04 19:36:29 +02001804 }
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001805 r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001806 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001807 goto out_free2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001808 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001809out_free2:
1810 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001811 break;
1812 }
1813 case KVM_GET_SREGS: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001814 kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL);
1815 r = -ENOMEM;
1816 if (!kvm_sregs)
1817 goto out;
1818 r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001819 if (r)
1820 goto out;
1821 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001822 if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001823 goto out;
1824 r = 0;
1825 break;
1826 }
1827 case KVM_SET_SREGS: {
Sasha Levinff5c2c02011-12-04 19:36:29 +02001828 kvm_sregs = memdup_user(argp, sizeof(*kvm_sregs));
1829 if (IS_ERR(kvm_sregs)) {
1830 r = PTR_ERR(kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001831 goto out;
Sasha Levinff5c2c02011-12-04 19:36:29 +02001832 }
Dave Hansenfa3795a2008-08-11 10:01:46 -07001833 r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001834 if (r)
1835 goto out;
1836 r = 0;
1837 break;
1838 }
Marcelo Tosatti62d9f0d2008-04-11 13:24:45 -03001839 case KVM_GET_MP_STATE: {
1840 struct kvm_mp_state mp_state;
1841
1842 r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state);
1843 if (r)
1844 goto out;
1845 r = -EFAULT;
1846 if (copy_to_user(argp, &mp_state, sizeof mp_state))
1847 goto out;
1848 r = 0;
1849 break;
1850 }
1851 case KVM_SET_MP_STATE: {
1852 struct kvm_mp_state mp_state;
1853
1854 r = -EFAULT;
1855 if (copy_from_user(&mp_state, argp, sizeof mp_state))
1856 goto out;
1857 r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
1858 if (r)
1859 goto out;
1860 r = 0;
1861 break;
1862 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08001863 case KVM_TRANSLATE: {
1864 struct kvm_translation tr;
1865
1866 r = -EFAULT;
Al Viro2f3669872007-02-09 16:38:35 +00001867 if (copy_from_user(&tr, argp, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001868 goto out;
Zhang Xiantao8b006792007-11-16 13:05:55 +08001869 r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001870 if (r)
1871 goto out;
1872 r = -EFAULT;
Al Viro2f3669872007-02-09 16:38:35 +00001873 if (copy_to_user(argp, &tr, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001874 goto out;
1875 r = 0;
1876 break;
1877 }
Jan Kiszkad0bfb942008-12-15 13:52:10 +01001878 case KVM_SET_GUEST_DEBUG: {
1879 struct kvm_guest_debug dbg;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001880
1881 r = -EFAULT;
Al Viro2f3669872007-02-09 16:38:35 +00001882 if (copy_from_user(&dbg, argp, sizeof dbg))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001883 goto out;
Jan Kiszkad0bfb942008-12-15 13:52:10 +01001884 r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001885 if (r)
1886 goto out;
1887 r = 0;
1888 break;
1889 }
Avi Kivity1961d272007-03-05 19:46:05 +02001890 case KVM_SET_SIGNAL_MASK: {
1891 struct kvm_signal_mask __user *sigmask_arg = argp;
1892 struct kvm_signal_mask kvm_sigmask;
1893 sigset_t sigset, *p;
1894
1895 p = NULL;
1896 if (argp) {
1897 r = -EFAULT;
1898 if (copy_from_user(&kvm_sigmask, argp,
1899 sizeof kvm_sigmask))
1900 goto out;
1901 r = -EINVAL;
1902 if (kvm_sigmask.len != sizeof sigset)
1903 goto out;
1904 r = -EFAULT;
1905 if (copy_from_user(&sigset, sigmask_arg->sigset,
1906 sizeof sigset))
1907 goto out;
1908 p = &sigset;
1909 }
Andi Kleen376d41f2010-06-10 13:10:47 +02001910 r = kvm_vcpu_ioctl_set_sigmask(vcpu, p);
Avi Kivity1961d272007-03-05 19:46:05 +02001911 break;
1912 }
Avi Kivityb8836732007-04-01 16:34:31 +03001913 case KVM_GET_FPU: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001914 fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL);
1915 r = -ENOMEM;
1916 if (!fpu)
1917 goto out;
1918 r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001919 if (r)
1920 goto out;
1921 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001922 if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
Avi Kivityb8836732007-04-01 16:34:31 +03001923 goto out;
1924 r = 0;
1925 break;
1926 }
1927 case KVM_SET_FPU: {
Sasha Levinff5c2c02011-12-04 19:36:29 +02001928 fpu = memdup_user(argp, sizeof(*fpu));
1929 if (IS_ERR(fpu)) {
1930 r = PTR_ERR(fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001931 goto out;
Sasha Levinff5c2c02011-12-04 19:36:29 +02001932 }
Dave Hansenfa3795a2008-08-11 10:01:46 -07001933 r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001934 if (r)
1935 goto out;
1936 r = 0;
1937 break;
1938 }
Avi Kivitybccf2152007-02-21 18:04:26 +02001939 default:
Carsten Otte313a3dc2007-10-11 19:16:52 +02001940 r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
Avi Kivitybccf2152007-02-21 18:04:26 +02001941 }
1942out:
Avi Kivity2122ff52010-05-13 11:25:04 +03001943 vcpu_put(vcpu);
Dave Hansenfa3795a2008-08-11 10:01:46 -07001944 kfree(fpu);
1945 kfree(kvm_sregs);
Avi Kivitybccf2152007-02-21 18:04:26 +02001946 return r;
1947}
1948
Alexander Graf1dda6062011-06-08 02:45:37 +02001949#ifdef CONFIG_COMPAT
1950static long kvm_vcpu_compat_ioctl(struct file *filp,
1951 unsigned int ioctl, unsigned long arg)
1952{
1953 struct kvm_vcpu *vcpu = filp->private_data;
1954 void __user *argp = compat_ptr(arg);
1955 int r;
1956
1957 if (vcpu->kvm->mm != current->mm)
1958 return -EIO;
1959
1960 switch (ioctl) {
1961 case KVM_SET_SIGNAL_MASK: {
1962 struct kvm_signal_mask __user *sigmask_arg = argp;
1963 struct kvm_signal_mask kvm_sigmask;
1964 compat_sigset_t csigset;
1965 sigset_t sigset;
1966
1967 if (argp) {
1968 r = -EFAULT;
1969 if (copy_from_user(&kvm_sigmask, argp,
1970 sizeof kvm_sigmask))
1971 goto out;
1972 r = -EINVAL;
1973 if (kvm_sigmask.len != sizeof csigset)
1974 goto out;
1975 r = -EFAULT;
1976 if (copy_from_user(&csigset, sigmask_arg->sigset,
1977 sizeof csigset))
1978 goto out;
Alan Cox760a9a32012-08-22 14:34:11 +01001979 sigset_from_compat(&sigset, &csigset);
1980 r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
1981 } else
1982 r = kvm_vcpu_ioctl_set_sigmask(vcpu, NULL);
Alexander Graf1dda6062011-06-08 02:45:37 +02001983 break;
1984 }
1985 default:
1986 r = kvm_vcpu_ioctl(filp, ioctl, arg);
1987 }
1988
1989out:
1990 return r;
1991}
1992#endif
1993
Avi Kivitybccf2152007-02-21 18:04:26 +02001994static long kvm_vm_ioctl(struct file *filp,
1995 unsigned int ioctl, unsigned long arg)
1996{
1997 struct kvm *kvm = filp->private_data;
1998 void __user *argp = (void __user *)arg;
Carsten Otte1fe779f2007-10-29 16:08:35 +01001999 int r;
Avi Kivitybccf2152007-02-21 18:04:26 +02002000
Avi Kivity6d4e4c42007-11-21 16:41:05 +02002001 if (kvm->mm != current->mm)
2002 return -EIO;
Avi Kivitybccf2152007-02-21 18:04:26 +02002003 switch (ioctl) {
2004 case KVM_CREATE_VCPU:
2005 r = kvm_vm_ioctl_create_vcpu(kvm, arg);
2006 if (r < 0)
2007 goto out;
2008 break;
Izik Eidus6fc138d2007-10-09 19:20:39 +02002009 case KVM_SET_USER_MEMORY_REGION: {
2010 struct kvm_userspace_memory_region kvm_userspace_mem;
2011
2012 r = -EFAULT;
2013 if (copy_from_user(&kvm_userspace_mem, argp,
2014 sizeof kvm_userspace_mem))
2015 goto out;
2016
2017 r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002018 if (r)
2019 goto out;
2020 break;
2021 }
2022 case KVM_GET_DIRTY_LOG: {
2023 struct kvm_dirty_log log;
2024
2025 r = -EFAULT;
Al Viro2f3669872007-02-09 16:38:35 +00002026 if (copy_from_user(&log, argp, sizeof log))
Avi Kivity6aa8b732006-12-10 02:21:36 -08002027 goto out;
Avi Kivity2c6f5df2007-02-20 18:27:58 +02002028 r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002029 if (r)
2030 goto out;
2031 break;
2032 }
Laurent Vivier5f94c172008-05-30 16:05:54 +02002033#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
2034 case KVM_REGISTER_COALESCED_MMIO: {
2035 struct kvm_coalesced_mmio_zone zone;
2036 r = -EFAULT;
2037 if (copy_from_user(&zone, argp, sizeof zone))
2038 goto out;
Laurent Vivier5f94c172008-05-30 16:05:54 +02002039 r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
2040 if (r)
2041 goto out;
2042 r = 0;
2043 break;
2044 }
2045 case KVM_UNREGISTER_COALESCED_MMIO: {
2046 struct kvm_coalesced_mmio_zone zone;
2047 r = -EFAULT;
2048 if (copy_from_user(&zone, argp, sizeof zone))
2049 goto out;
Laurent Vivier5f94c172008-05-30 16:05:54 +02002050 r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
2051 if (r)
2052 goto out;
2053 r = 0;
2054 break;
2055 }
2056#endif
Gregory Haskins721eecbf2009-05-20 10:30:49 -04002057 case KVM_IRQFD: {
2058 struct kvm_irqfd data;
2059
2060 r = -EFAULT;
2061 if (copy_from_user(&data, argp, sizeof data))
2062 goto out;
Alex Williamsond4db2932012-06-29 09:56:08 -06002063 r = kvm_irqfd(kvm, &data);
Gregory Haskins721eecbf2009-05-20 10:30:49 -04002064 break;
2065 }
Gregory Haskinsd34e6b12009-07-07 17:08:49 -04002066 case KVM_IOEVENTFD: {
2067 struct kvm_ioeventfd data;
2068
2069 r = -EFAULT;
2070 if (copy_from_user(&data, argp, sizeof data))
2071 goto out;
2072 r = kvm_ioeventfd(kvm, &data);
2073 break;
2074 }
Gleb Natapov73880c82009-06-09 15:56:28 +03002075#ifdef CONFIG_KVM_APIC_ARCHITECTURE
2076 case KVM_SET_BOOT_CPU_ID:
2077 r = 0;
Marcelo Tosatti894a9c52009-06-23 15:05:14 -03002078 mutex_lock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03002079 if (atomic_read(&kvm->online_vcpus) != 0)
2080 r = -EBUSY;
2081 else
2082 kvm->bsp_vcpu_id = arg;
Marcelo Tosatti894a9c52009-06-23 15:05:14 -03002083 mutex_unlock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03002084 break;
2085#endif
Jan Kiszka07975ad2012-03-29 21:14:12 +02002086#ifdef CONFIG_HAVE_KVM_MSI
2087 case KVM_SIGNAL_MSI: {
2088 struct kvm_msi msi;
2089
2090 r = -EFAULT;
2091 if (copy_from_user(&msi, argp, sizeof msi))
2092 goto out;
2093 r = kvm_send_userspace_msi(kvm, &msi);
2094 break;
2095 }
2096#endif
Avi Kivityf17abe92007-02-21 19:28:04 +02002097 default:
Carsten Otte1fe779f2007-10-29 16:08:35 +01002098 r = kvm_arch_vm_ioctl(filp, ioctl, arg);
Avi Kivitybfd99ff2009-08-26 14:57:50 +03002099 if (r == -ENOTTY)
2100 r = kvm_vm_ioctl_assigned_device(kvm, ioctl, arg);
Avi Kivityf17abe92007-02-21 19:28:04 +02002101 }
2102out:
2103 return r;
2104}
2105
Arnd Bergmann6ff58942009-10-22 14:19:27 +02002106#ifdef CONFIG_COMPAT
2107struct compat_kvm_dirty_log {
2108 __u32 slot;
2109 __u32 padding1;
2110 union {
2111 compat_uptr_t dirty_bitmap; /* one bit per page */
2112 __u64 padding2;
2113 };
2114};
2115
2116static long kvm_vm_compat_ioctl(struct file *filp,
2117 unsigned int ioctl, unsigned long arg)
2118{
2119 struct kvm *kvm = filp->private_data;
2120 int r;
2121
2122 if (kvm->mm != current->mm)
2123 return -EIO;
2124 switch (ioctl) {
2125 case KVM_GET_DIRTY_LOG: {
2126 struct compat_kvm_dirty_log compat_log;
2127 struct kvm_dirty_log log;
2128
2129 r = -EFAULT;
2130 if (copy_from_user(&compat_log, (void __user *)arg,
2131 sizeof(compat_log)))
2132 goto out;
2133 log.slot = compat_log.slot;
2134 log.padding1 = compat_log.padding1;
2135 log.padding2 = compat_log.padding2;
2136 log.dirty_bitmap = compat_ptr(compat_log.dirty_bitmap);
2137
2138 r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
2139 if (r)
2140 goto out;
2141 break;
2142 }
2143 default:
2144 r = kvm_vm_ioctl(filp, ioctl, arg);
2145 }
2146
2147out:
2148 return r;
2149}
2150#endif
2151
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002152static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivityf17abe92007-02-21 19:28:04 +02002153{
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002154 struct page *page[1];
2155 unsigned long addr;
2156 int npages;
2157 gfn_t gfn = vmf->pgoff;
Avi Kivityf17abe92007-02-21 19:28:04 +02002158 struct kvm *kvm = vma->vm_file->private_data;
Avi Kivityf17abe92007-02-21 19:28:04 +02002159
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002160 addr = gfn_to_hva(kvm, gfn);
2161 if (kvm_is_error_hva(addr))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002162 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002163
2164 npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page,
2165 NULL);
2166 if (unlikely(npages != 1))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002167 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002168
2169 vmf->page = page[0];
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002170 return 0;
Avi Kivityf17abe92007-02-21 19:28:04 +02002171}
2172
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +04002173static const struct vm_operations_struct kvm_vm_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002174 .fault = kvm_vm_fault,
Avi Kivityf17abe92007-02-21 19:28:04 +02002175};
2176
2177static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma)
2178{
2179 vma->vm_ops = &kvm_vm_vm_ops;
2180 return 0;
2181}
2182
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01002183static struct file_operations kvm_vm_fops = {
Avi Kivityf17abe92007-02-21 19:28:04 +02002184 .release = kvm_vm_release,
2185 .unlocked_ioctl = kvm_vm_ioctl,
Arnd Bergmann6ff58942009-10-22 14:19:27 +02002186#ifdef CONFIG_COMPAT
2187 .compat_ioctl = kvm_vm_compat_ioctl,
2188#endif
Avi Kivityf17abe92007-02-21 19:28:04 +02002189 .mmap = kvm_vm_mmap,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002190 .llseek = noop_llseek,
Avi Kivityf17abe92007-02-21 19:28:04 +02002191};
2192
Carsten Ottee08b9632012-01-04 10:25:20 +01002193static int kvm_dev_ioctl_create_vm(unsigned long type)
Avi Kivityf17abe92007-02-21 19:28:04 +02002194{
Heiko Carstensaac87632010-10-27 17:22:10 +02002195 int r;
Avi Kivityf17abe92007-02-21 19:28:04 +02002196 struct kvm *kvm;
2197
Carsten Ottee08b9632012-01-04 10:25:20 +01002198 kvm = kvm_create_vm(type);
Avi Kivityd6d28162007-06-28 08:38:16 -04002199 if (IS_ERR(kvm))
2200 return PTR_ERR(kvm);
Takuya Yoshikawa6ce5a092010-03-15 22:13:30 +09002201#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
2202 r = kvm_coalesced_mmio_init(kvm);
2203 if (r < 0) {
2204 kvm_put_kvm(kvm);
2205 return r;
2206 }
2207#endif
Heiko Carstensaac87632010-10-27 17:22:10 +02002208 r = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, O_RDWR);
2209 if (r < 0)
Al Viro66c0b392008-04-19 20:33:56 +01002210 kvm_put_kvm(kvm);
Avi Kivityf17abe92007-02-21 19:28:04 +02002211
Heiko Carstensaac87632010-10-27 17:22:10 +02002212 return r;
Avi Kivityf17abe92007-02-21 19:28:04 +02002213}
2214
Avi Kivity1a811b62008-12-08 18:25:27 +02002215static long kvm_dev_ioctl_check_extension_generic(long arg)
2216{
2217 switch (arg) {
Avi Kivityca9edae2008-12-08 18:29:29 +02002218 case KVM_CAP_USER_MEMORY:
Avi Kivity1a811b62008-12-08 18:25:27 +02002219 case KVM_CAP_DESTROY_MEMORY_REGION_WORKS:
Jan Kiszka4cd481f2009-04-13 11:59:32 +02002220 case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS:
Gleb Natapov73880c82009-06-09 15:56:28 +03002221#ifdef CONFIG_KVM_APIC_ARCHITECTURE
2222 case KVM_CAP_SET_BOOT_CPU_ID:
2223#endif
Avi Kivitya9c73992009-11-04 11:54:59 +02002224 case KVM_CAP_INTERNAL_ERROR_DATA:
Jan Kiszka07975ad2012-03-29 21:14:12 +02002225#ifdef CONFIG_HAVE_KVM_MSI
2226 case KVM_CAP_SIGNAL_MSI:
2227#endif
Avi Kivity1a811b62008-12-08 18:25:27 +02002228 return 1;
Marc Zyngier9900b4b2012-06-15 15:07:02 -04002229#ifdef KVM_CAP_IRQ_ROUTING
Avi Kivity399ec802008-11-19 13:58:46 +02002230 case KVM_CAP_IRQ_ROUTING:
Sheng Yang36463142009-03-16 16:33:43 +08002231 return KVM_MAX_IRQ_ROUTES;
Avi Kivity399ec802008-11-19 13:58:46 +02002232#endif
Avi Kivity1a811b62008-12-08 18:25:27 +02002233 default:
2234 break;
2235 }
2236 return kvm_dev_ioctl_check_extension(arg);
2237}
2238
Avi Kivityf17abe92007-02-21 19:28:04 +02002239static long kvm_dev_ioctl(struct file *filp,
2240 unsigned int ioctl, unsigned long arg)
2241{
Avi Kivity07c45a32007-03-07 13:05:38 +02002242 long r = -EINVAL;
Avi Kivityf17abe92007-02-21 19:28:04 +02002243
2244 switch (ioctl) {
2245 case KVM_GET_API_VERSION:
Avi Kivityf0fe5102007-03-07 13:11:17 +02002246 r = -EINVAL;
2247 if (arg)
2248 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02002249 r = KVM_API_VERSION;
2250 break;
2251 case KVM_CREATE_VM:
Carsten Ottee08b9632012-01-04 10:25:20 +01002252 r = kvm_dev_ioctl_create_vm(arg);
Avi Kivityf17abe92007-02-21 19:28:04 +02002253 break;
Zhang Xiantao018d00d2007-11-15 23:07:47 +08002254 case KVM_CHECK_EXTENSION:
Avi Kivity1a811b62008-12-08 18:25:27 +02002255 r = kvm_dev_ioctl_check_extension_generic(arg);
Avi Kivity5d308f42007-03-01 17:56:20 +02002256 break;
Avi Kivity07c45a32007-03-07 13:05:38 +02002257 case KVM_GET_VCPU_MMAP_SIZE:
2258 r = -EINVAL;
2259 if (arg)
2260 goto out;
Avi Kivityadb1ff42008-01-24 15:13:08 +02002261 r = PAGE_SIZE; /* struct kvm_run */
2262#ifdef CONFIG_X86
2263 r += PAGE_SIZE; /* pio data page */
2264#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02002265#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
2266 r += PAGE_SIZE; /* coalesced mmio ring page */
2267#endif
Avi Kivity07c45a32007-03-07 13:05:38 +02002268 break;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04002269 case KVM_TRACE_ENABLE:
2270 case KVM_TRACE_PAUSE:
2271 case KVM_TRACE_DISABLE:
Marcelo Tosatti2023a292009-06-18 11:47:28 -03002272 r = -EOPNOTSUPP;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04002273 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002274 default:
Carsten Otte043405e2007-10-10 17:16:19 +02002275 return kvm_arch_dev_ioctl(filp, ioctl, arg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002276 }
2277out:
2278 return r;
2279}
2280
Avi Kivity6aa8b732006-12-10 02:21:36 -08002281static struct file_operations kvm_chardev_ops = {
Avi Kivity6aa8b732006-12-10 02:21:36 -08002282 .unlocked_ioctl = kvm_dev_ioctl,
2283 .compat_ioctl = kvm_dev_ioctl,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002284 .llseek = noop_llseek,
Avi Kivity6aa8b732006-12-10 02:21:36 -08002285};
2286
2287static struct miscdevice kvm_dev = {
Avi Kivitybbe44322007-03-04 13:27:36 +02002288 KVM_MINOR,
Avi Kivity6aa8b732006-12-10 02:21:36 -08002289 "kvm",
2290 &kvm_chardev_ops,
2291};
2292
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002293static void hardware_enable_nolock(void *junk)
Avi Kivity1b6c0162007-05-24 13:03:52 +03002294{
2295 int cpu = raw_smp_processor_id();
Alexander Graf10474ae2009-09-15 11:37:46 +02002296 int r;
Avi Kivity1b6c0162007-05-24 13:03:52 +03002297
Rusty Russell7f59f492008-12-07 21:25:45 +10302298 if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03002299 return;
Alexander Graf10474ae2009-09-15 11:37:46 +02002300
Rusty Russell7f59f492008-12-07 21:25:45 +10302301 cpumask_set_cpu(cpu, cpus_hardware_enabled);
Alexander Graf10474ae2009-09-15 11:37:46 +02002302
2303 r = kvm_arch_hardware_enable(NULL);
2304
2305 if (r) {
2306 cpumask_clear_cpu(cpu, cpus_hardware_enabled);
2307 atomic_inc(&hardware_enable_failed);
2308 printk(KERN_INFO "kvm: enabling virtualization on "
2309 "CPU%d failed\n", cpu);
2310 }
Avi Kivity1b6c0162007-05-24 13:03:52 +03002311}
2312
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002313static void hardware_enable(void *junk)
2314{
Jan Kiszkae935b832011-02-08 12:55:33 +01002315 raw_spin_lock(&kvm_lock);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002316 hardware_enable_nolock(junk);
Jan Kiszkae935b832011-02-08 12:55:33 +01002317 raw_spin_unlock(&kvm_lock);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002318}
2319
2320static void hardware_disable_nolock(void *junk)
Avi Kivity1b6c0162007-05-24 13:03:52 +03002321{
2322 int cpu = raw_smp_processor_id();
2323
Rusty Russell7f59f492008-12-07 21:25:45 +10302324 if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03002325 return;
Rusty Russell7f59f492008-12-07 21:25:45 +10302326 cpumask_clear_cpu(cpu, cpus_hardware_enabled);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002327 kvm_arch_hardware_disable(NULL);
Avi Kivity1b6c0162007-05-24 13:03:52 +03002328}
2329
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002330static void hardware_disable(void *junk)
2331{
Jan Kiszkae935b832011-02-08 12:55:33 +01002332 raw_spin_lock(&kvm_lock);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002333 hardware_disable_nolock(junk);
Jan Kiszkae935b832011-02-08 12:55:33 +01002334 raw_spin_unlock(&kvm_lock);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002335}
2336
Alexander Graf10474ae2009-09-15 11:37:46 +02002337static void hardware_disable_all_nolock(void)
2338{
2339 BUG_ON(!kvm_usage_count);
2340
2341 kvm_usage_count--;
2342 if (!kvm_usage_count)
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002343 on_each_cpu(hardware_disable_nolock, NULL, 1);
Alexander Graf10474ae2009-09-15 11:37:46 +02002344}
2345
2346static void hardware_disable_all(void)
2347{
Jan Kiszkae935b832011-02-08 12:55:33 +01002348 raw_spin_lock(&kvm_lock);
Alexander Graf10474ae2009-09-15 11:37:46 +02002349 hardware_disable_all_nolock();
Jan Kiszkae935b832011-02-08 12:55:33 +01002350 raw_spin_unlock(&kvm_lock);
Alexander Graf10474ae2009-09-15 11:37:46 +02002351}
2352
2353static int hardware_enable_all(void)
2354{
2355 int r = 0;
2356
Jan Kiszkae935b832011-02-08 12:55:33 +01002357 raw_spin_lock(&kvm_lock);
Alexander Graf10474ae2009-09-15 11:37:46 +02002358
2359 kvm_usage_count++;
2360 if (kvm_usage_count == 1) {
2361 atomic_set(&hardware_enable_failed, 0);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002362 on_each_cpu(hardware_enable_nolock, NULL, 1);
Alexander Graf10474ae2009-09-15 11:37:46 +02002363
2364 if (atomic_read(&hardware_enable_failed)) {
2365 hardware_disable_all_nolock();
2366 r = -EBUSY;
2367 }
2368 }
2369
Jan Kiszkae935b832011-02-08 12:55:33 +01002370 raw_spin_unlock(&kvm_lock);
Alexander Graf10474ae2009-09-15 11:37:46 +02002371
2372 return r;
2373}
2374
Avi Kivity774c47f2007-02-12 00:54:47 -08002375static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val,
2376 void *v)
2377{
2378 int cpu = (long)v;
2379
Alexander Graf10474ae2009-09-15 11:37:46 +02002380 if (!kvm_usage_count)
2381 return NOTIFY_OK;
2382
Avi Kivity1a6f4d72007-11-11 18:37:32 +02002383 val &= ~CPU_TASKS_FROZEN;
Avi Kivity774c47f2007-02-12 00:54:47 -08002384 switch (val) {
Avi Kivitycec9ad22007-05-24 13:11:41 +03002385 case CPU_DYING:
Avi Kivity6ec8a852007-08-19 15:57:26 +03002386 printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n",
2387 cpu);
2388 hardware_disable(NULL);
2389 break;
Zachary Amsdenda908f22010-08-19 22:07:27 -10002390 case CPU_STARTING:
Jeremy Katz43934a32007-02-19 14:37:46 +02002391 printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n",
2392 cpu);
Zachary Amsdenda908f22010-08-19 22:07:27 -10002393 hardware_enable(NULL);
Avi Kivity774c47f2007-02-12 00:54:47 -08002394 break;
2395 }
2396 return NOTIFY_OK;
2397}
2398
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002399
Avi Kivityb7c41452010-12-02 17:52:50 +02002400asmlinkage void kvm_spurious_fault(void)
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002401{
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002402 /* Fault while not rebooting. We want the trace. */
2403 BUG();
2404}
Avi Kivityb7c41452010-12-02 17:52:50 +02002405EXPORT_SYMBOL_GPL(kvm_spurious_fault);
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002406
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002407static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
Mike Dayd77c26f2007-10-08 09:02:08 -04002408 void *v)
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002409{
Sheng Yang8e1c1812009-04-29 11:09:04 +08002410 /*
2411 * Some (well, at least mine) BIOSes hang on reboot if
2412 * in vmx root mode.
2413 *
2414 * And Intel TXT required VMX off for all cpu when system shutdown.
2415 */
2416 printk(KERN_INFO "kvm: exiting hardware virtualization\n");
2417 kvm_rebooting = true;
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002418 on_each_cpu(hardware_disable_nolock, NULL, 1);
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002419 return NOTIFY_OK;
2420}
2421
2422static struct notifier_block kvm_reboot_notifier = {
2423 .notifier_call = kvm_reboot,
2424 .priority = 0,
2425};
2426
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002427static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002428{
2429 int i;
2430
2431 for (i = 0; i < bus->dev_count; i++) {
Sasha Levin743eeb02011-07-27 16:00:48 +03002432 struct kvm_io_device *pos = bus->range[i].dev;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002433
2434 kvm_iodevice_destructor(pos);
2435 }
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002436 kfree(bus);
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002437}
2438
Sasha Levin743eeb02011-07-27 16:00:48 +03002439int kvm_io_bus_sort_cmp(const void *p1, const void *p2)
2440{
2441 const struct kvm_io_range *r1 = p1;
2442 const struct kvm_io_range *r2 = p2;
2443
2444 if (r1->addr < r2->addr)
2445 return -1;
2446 if (r1->addr + r1->len > r2->addr + r2->len)
2447 return 1;
2448 return 0;
2449}
2450
2451int kvm_io_bus_insert_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev,
2452 gpa_t addr, int len)
2453{
Sasha Levin743eeb02011-07-27 16:00:48 +03002454 bus->range[bus->dev_count++] = (struct kvm_io_range) {
2455 .addr = addr,
2456 .len = len,
2457 .dev = dev,
2458 };
2459
2460 sort(bus->range, bus->dev_count, sizeof(struct kvm_io_range),
2461 kvm_io_bus_sort_cmp, NULL);
2462
2463 return 0;
2464}
2465
2466int kvm_io_bus_get_first_dev(struct kvm_io_bus *bus,
2467 gpa_t addr, int len)
2468{
2469 struct kvm_io_range *range, key;
2470 int off;
2471
2472 key = (struct kvm_io_range) {
2473 .addr = addr,
2474 .len = len,
2475 };
2476
2477 range = bsearch(&key, bus->range, bus->dev_count,
2478 sizeof(struct kvm_io_range), kvm_io_bus_sort_cmp);
2479 if (range == NULL)
2480 return -ENOENT;
2481
2482 off = range - bus->range;
2483
2484 while (off > 0 && kvm_io_bus_sort_cmp(&key, &bus->range[off-1]) == 0)
2485 off--;
2486
2487 return off;
2488}
2489
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002490/* kvm_io_bus_write - called under kvm->slots_lock */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002491int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002492 int len, const void *val)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002493{
Sasha Levin743eeb02011-07-27 16:00:48 +03002494 int idx;
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002495 struct kvm_io_bus *bus;
Sasha Levin743eeb02011-07-27 16:00:48 +03002496 struct kvm_io_range range;
2497
2498 range = (struct kvm_io_range) {
2499 .addr = addr,
2500 .len = len,
2501 };
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002502
2503 bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu);
Sasha Levin743eeb02011-07-27 16:00:48 +03002504 idx = kvm_io_bus_get_first_dev(bus, addr, len);
2505 if (idx < 0)
2506 return -EOPNOTSUPP;
2507
2508 while (idx < bus->dev_count &&
2509 kvm_io_bus_sort_cmp(&range, &bus->range[idx]) == 0) {
2510 if (!kvm_iodevice_write(bus->range[idx].dev, addr, len, val))
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002511 return 0;
Sasha Levin743eeb02011-07-27 16:00:48 +03002512 idx++;
2513 }
2514
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002515 return -EOPNOTSUPP;
2516}
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002517
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002518/* kvm_io_bus_read - called under kvm->slots_lock */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002519int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
2520 int len, void *val)
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002521{
Sasha Levin743eeb02011-07-27 16:00:48 +03002522 int idx;
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002523 struct kvm_io_bus *bus;
Sasha Levin743eeb02011-07-27 16:00:48 +03002524 struct kvm_io_range range;
2525
2526 range = (struct kvm_io_range) {
2527 .addr = addr,
2528 .len = len,
2529 };
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002530
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002531 bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu);
Sasha Levin743eeb02011-07-27 16:00:48 +03002532 idx = kvm_io_bus_get_first_dev(bus, addr, len);
2533 if (idx < 0)
2534 return -EOPNOTSUPP;
2535
2536 while (idx < bus->dev_count &&
2537 kvm_io_bus_sort_cmp(&range, &bus->range[idx]) == 0) {
2538 if (!kvm_iodevice_read(bus->range[idx].dev, addr, len, val))
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002539 return 0;
Sasha Levin743eeb02011-07-27 16:00:48 +03002540 idx++;
2541 }
2542
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002543 return -EOPNOTSUPP;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002544}
2545
Marcelo Tosatti79fac952009-12-23 14:35:26 -02002546/* Caller must hold slots_lock. */
Sasha Levin743eeb02011-07-27 16:00:48 +03002547int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
2548 int len, struct kvm_io_device *dev)
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002549{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002550 struct kvm_io_bus *new_bus, *bus;
Gregory Haskins090b7af2009-07-07 17:08:44 -04002551
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002552 bus = kvm->buses[bus_idx];
Amos Konga13007162012-03-09 12:17:32 +08002553 if (bus->dev_count > NR_IOBUS_DEVS - 1)
Gregory Haskins090b7af2009-07-07 17:08:44 -04002554 return -ENOSPC;
2555
Amos Konga13007162012-03-09 12:17:32 +08002556 new_bus = kzalloc(sizeof(*bus) + ((bus->dev_count + 1) *
2557 sizeof(struct kvm_io_range)), GFP_KERNEL);
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002558 if (!new_bus)
2559 return -ENOMEM;
Amos Konga13007162012-03-09 12:17:32 +08002560 memcpy(new_bus, bus, sizeof(*bus) + (bus->dev_count *
2561 sizeof(struct kvm_io_range)));
Sasha Levin743eeb02011-07-27 16:00:48 +03002562 kvm_io_bus_insert_dev(new_bus, dev, addr, len);
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002563 rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
2564 synchronize_srcu_expedited(&kvm->srcu);
2565 kfree(bus);
Gregory Haskins090b7af2009-07-07 17:08:44 -04002566
2567 return 0;
2568}
2569
Marcelo Tosatti79fac952009-12-23 14:35:26 -02002570/* Caller must hold slots_lock. */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002571int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
2572 struct kvm_io_device *dev)
Gregory Haskins090b7af2009-07-07 17:08:44 -04002573{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002574 int i, r;
2575 struct kvm_io_bus *new_bus, *bus;
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002576
Sasha Levincdfca7b2011-12-04 19:36:28 +02002577 bus = kvm->buses[bus_idx];
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002578 r = -ENOENT;
Amos Konga13007162012-03-09 12:17:32 +08002579 for (i = 0; i < bus->dev_count; i++)
2580 if (bus->range[i].dev == dev) {
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002581 r = 0;
Gregory Haskins090b7af2009-07-07 17:08:44 -04002582 break;
2583 }
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002584
Amos Konga13007162012-03-09 12:17:32 +08002585 if (r)
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002586 return r;
Amos Konga13007162012-03-09 12:17:32 +08002587
2588 new_bus = kzalloc(sizeof(*bus) + ((bus->dev_count - 1) *
2589 sizeof(struct kvm_io_range)), GFP_KERNEL);
2590 if (!new_bus)
2591 return -ENOMEM;
2592
2593 memcpy(new_bus, bus, sizeof(*bus) + i * sizeof(struct kvm_io_range));
2594 new_bus->dev_count--;
2595 memcpy(new_bus->range + i, bus->range + i + 1,
2596 (new_bus->dev_count - i) * sizeof(struct kvm_io_range));
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002597
2598 rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
2599 synchronize_srcu_expedited(&kvm->srcu);
2600 kfree(bus);
2601 return r;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002602}
2603
Avi Kivity774c47f2007-02-12 00:54:47 -08002604static struct notifier_block kvm_cpu_notifier = {
2605 .notifier_call = kvm_cpu_hotplug,
Avi Kivity774c47f2007-02-12 00:54:47 -08002606};
2607
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002608static int vm_stat_get(void *_offset, u64 *val)
Avi Kivityba1389b2007-11-18 16:24:12 +02002609{
2610 unsigned offset = (long)_offset;
Avi Kivityba1389b2007-11-18 16:24:12 +02002611 struct kvm *kvm;
2612
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002613 *val = 0;
Jan Kiszkae935b832011-02-08 12:55:33 +01002614 raw_spin_lock(&kvm_lock);
Avi Kivityba1389b2007-11-18 16:24:12 +02002615 list_for_each_entry(kvm, &vm_list, vm_list)
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002616 *val += *(u32 *)((void *)kvm + offset);
Jan Kiszkae935b832011-02-08 12:55:33 +01002617 raw_spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002618 return 0;
Avi Kivityba1389b2007-11-18 16:24:12 +02002619}
2620
2621DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n");
2622
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002623static int vcpu_stat_get(void *_offset, u64 *val)
Avi Kivity1165f5f2007-04-19 17:27:43 +03002624{
2625 unsigned offset = (long)_offset;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002626 struct kvm *kvm;
2627 struct kvm_vcpu *vcpu;
2628 int i;
2629
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002630 *val = 0;
Jan Kiszkae935b832011-02-08 12:55:33 +01002631 raw_spin_lock(&kvm_lock);
Avi Kivity1165f5f2007-04-19 17:27:43 +03002632 list_for_each_entry(kvm, &vm_list, vm_list)
Gleb Natapov988a2ca2009-06-09 15:56:29 +03002633 kvm_for_each_vcpu(i, vcpu, kvm)
2634 *val += *(u32 *)((void *)vcpu + offset);
2635
Jan Kiszkae935b832011-02-08 12:55:33 +01002636 raw_spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002637 return 0;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002638}
2639
Avi Kivityba1389b2007-11-18 16:24:12 +02002640DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n");
2641
Alexey Dobriyan828c0952009-10-01 15:43:56 -07002642static const struct file_operations *stat_fops[] = {
Avi Kivityba1389b2007-11-18 16:24:12 +02002643 [KVM_STAT_VCPU] = &vcpu_stat_fops,
2644 [KVM_STAT_VM] = &vm_stat_fops,
2645};
Avi Kivity1165f5f2007-04-19 17:27:43 +03002646
Hamo4f69b682011-12-15 14:23:16 +08002647static int kvm_init_debug(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002648{
Hamo4f69b682011-12-15 14:23:16 +08002649 int r = -EFAULT;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002650 struct kvm_stats_debugfs_item *p;
2651
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002652 kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
Hamo4f69b682011-12-15 14:23:16 +08002653 if (kvm_debugfs_dir == NULL)
2654 goto out;
2655
2656 for (p = debugfs_entries; p->name; ++p) {
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002657 p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir,
Avi Kivity1165f5f2007-04-19 17:27:43 +03002658 (void *)(long)p->offset,
Avi Kivityba1389b2007-11-18 16:24:12 +02002659 stat_fops[p->kind]);
Hamo4f69b682011-12-15 14:23:16 +08002660 if (p->dentry == NULL)
2661 goto out_dir;
2662 }
2663
2664 return 0;
2665
2666out_dir:
2667 debugfs_remove_recursive(kvm_debugfs_dir);
2668out:
2669 return r;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002670}
2671
2672static void kvm_exit_debug(void)
2673{
2674 struct kvm_stats_debugfs_item *p;
2675
2676 for (p = debugfs_entries; p->name; ++p)
2677 debugfs_remove(p->dentry);
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002678 debugfs_remove(kvm_debugfs_dir);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002679}
2680
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002681static int kvm_suspend(void)
Avi Kivity59ae6c62007-02-12 00:54:48 -08002682{
Alexander Graf10474ae2009-09-15 11:37:46 +02002683 if (kvm_usage_count)
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002684 hardware_disable_nolock(NULL);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002685 return 0;
2686}
2687
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002688static void kvm_resume(void)
Avi Kivity59ae6c62007-02-12 00:54:48 -08002689{
Zachary Amsdenca84d1a2010-08-19 22:07:28 -10002690 if (kvm_usage_count) {
Jan Kiszkae935b832011-02-08 12:55:33 +01002691 WARN_ON(raw_spin_is_locked(&kvm_lock));
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002692 hardware_enable_nolock(NULL);
Zachary Amsdenca84d1a2010-08-19 22:07:28 -10002693 }
Avi Kivity59ae6c62007-02-12 00:54:48 -08002694}
2695
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002696static struct syscore_ops kvm_syscore_ops = {
Avi Kivity59ae6c62007-02-12 00:54:48 -08002697 .suspend = kvm_suspend,
2698 .resume = kvm_resume,
2699};
2700
Izik Eiduscea7bb22007-10-17 19:17:48 +02002701struct page *bad_page;
Anthony Liguori35149e22008-04-02 14:46:56 -05002702pfn_t bad_pfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002703
Avi Kivity15ad7142007-07-11 18:17:21 +03002704static inline
2705struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn)
2706{
2707 return container_of(pn, struct kvm_vcpu, preempt_notifier);
2708}
2709
2710static void kvm_sched_in(struct preempt_notifier *pn, int cpu)
2711{
2712 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2713
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002714 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002715}
2716
2717static void kvm_sched_out(struct preempt_notifier *pn,
2718 struct task_struct *next)
2719{
2720 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2721
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002722 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002723}
2724
Avi Kivity0ee75be2010-04-28 15:39:01 +03002725int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
Rusty Russellc16f8622007-07-30 21:12:19 +10002726 struct module *module)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002727{
2728 int r;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002729 int cpu;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002730
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002731 r = kvm_arch_init(opaque);
2732 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002733 goto out_fail;
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002734
2735 bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2736
2737 if (bad_page == NULL) {
2738 r = -ENOMEM;
2739 goto out;
2740 }
2741
Anthony Liguori35149e22008-04-02 14:46:56 -05002742 bad_pfn = page_to_pfn(bad_page);
2743
Huang Yingbf998152010-05-31 14:28:19 +08002744 hwpoison_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2745
2746 if (hwpoison_page == NULL) {
2747 r = -ENOMEM;
2748 goto out_free_0;
2749 }
2750
2751 hwpoison_pfn = page_to_pfn(hwpoison_page);
2752
Gleb Natapovedba23e2010-07-07 20:16:45 +03002753 fault_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2754
2755 if (fault_page == NULL) {
2756 r = -ENOMEM;
2757 goto out_free_0;
2758 }
2759
2760 fault_pfn = page_to_pfn(fault_page);
2761
Avi Kivity8437a612009-06-06 14:52:35 -07002762 if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
Rusty Russell7f59f492008-12-07 21:25:45 +10302763 r = -ENOMEM;
2764 goto out_free_0;
2765 }
2766
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002767 r = kvm_arch_hardware_setup();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002768 if (r < 0)
Rusty Russell7f59f492008-12-07 21:25:45 +10302769 goto out_free_0a;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002770
Yang, Sheng002c7f72007-07-31 14:23:01 +03002771 for_each_online_cpu(cpu) {
2772 smp_call_function_single(cpu,
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002773 kvm_arch_check_processor_compat,
Jens Axboe8691e5a2008-06-06 11:18:06 +02002774 &r, 1);
Yang, Sheng002c7f72007-07-31 14:23:01 +03002775 if (r < 0)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002776 goto out_free_1;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002777 }
2778
Avi Kivity774c47f2007-02-12 00:54:47 -08002779 r = register_cpu_notifier(&kvm_cpu_notifier);
2780 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002781 goto out_free_2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002782 register_reboot_notifier(&kvm_reboot_notifier);
2783
Rusty Russellc16f8622007-07-30 21:12:19 +10002784 /* A kmem cache lets us meet the alignment requirements of fx_save. */
Avi Kivity0ee75be2010-04-28 15:39:01 +03002785 if (!vcpu_align)
2786 vcpu_align = __alignof__(struct kvm_vcpu);
2787 kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, vcpu_align,
Joe Perches56919c52007-11-12 20:06:51 -08002788 0, NULL);
Rusty Russellc16f8622007-07-30 21:12:19 +10002789 if (!kvm_vcpu_cache) {
2790 r = -ENOMEM;
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002791 goto out_free_3;
Rusty Russellc16f8622007-07-30 21:12:19 +10002792 }
2793
Gleb Natapovaf585b92010-10-14 11:22:46 +02002794 r = kvm_async_pf_init();
2795 if (r)
2796 goto out_free;
2797
Avi Kivity6aa8b732006-12-10 02:21:36 -08002798 kvm_chardev_ops.owner = module;
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01002799 kvm_vm_fops.owner = module;
2800 kvm_vcpu_fops.owner = module;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002801
2802 r = misc_register(&kvm_dev);
2803 if (r) {
Mike Dayd77c26f2007-10-08 09:02:08 -04002804 printk(KERN_ERR "kvm: misc device register failed\n");
Gleb Natapovaf585b92010-10-14 11:22:46 +02002805 goto out_unreg;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002806 }
2807
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002808 register_syscore_ops(&kvm_syscore_ops);
2809
Avi Kivity15ad7142007-07-11 18:17:21 +03002810 kvm_preempt_ops.sched_in = kvm_sched_in;
2811 kvm_preempt_ops.sched_out = kvm_sched_out;
2812
Hamo4f69b682011-12-15 14:23:16 +08002813 r = kvm_init_debug();
2814 if (r) {
2815 printk(KERN_ERR "kvm: create debugfs files failed\n");
2816 goto out_undebugfs;
2817 }
Darrick J. Wong0ea4ed82009-10-14 16:21:00 -07002818
Avi Kivityc7addb92007-09-16 18:58:32 +02002819 return 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002820
Hamo4f69b682011-12-15 14:23:16 +08002821out_undebugfs:
2822 unregister_syscore_ops(&kvm_syscore_ops);
Gleb Natapovaf585b92010-10-14 11:22:46 +02002823out_unreg:
2824 kvm_async_pf_deinit();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002825out_free:
Rusty Russellc16f8622007-07-30 21:12:19 +10002826 kmem_cache_destroy(kvm_vcpu_cache);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002827out_free_3:
Avi Kivity6aa8b732006-12-10 02:21:36 -08002828 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity774c47f2007-02-12 00:54:47 -08002829 unregister_cpu_notifier(&kvm_cpu_notifier);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002830out_free_2:
Zhang Xiantaod23087842007-11-29 15:35:39 +08002831out_free_1:
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002832 kvm_arch_hardware_unsetup();
Rusty Russell7f59f492008-12-07 21:25:45 +10302833out_free_0a:
2834 free_cpumask_var(cpus_hardware_enabled);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002835out_free_0:
Gleb Natapovedba23e2010-07-07 20:16:45 +03002836 if (fault_page)
2837 __free_page(fault_page);
Huang Yingbf998152010-05-31 14:28:19 +08002838 if (hwpoison_page)
2839 __free_page(hwpoison_page);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002840 __free_page(bad_page);
Avi Kivityca45aaa2007-03-01 19:21:03 +02002841out:
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002842 kvm_arch_exit();
Zhang Xiantaod23087842007-11-29 15:35:39 +08002843out_fail:
Avi Kivity6aa8b732006-12-10 02:21:36 -08002844 return r;
2845}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002846EXPORT_SYMBOL_GPL(kvm_init);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002847
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002848void kvm_exit(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002849{
Darrick J. Wong0ea4ed82009-10-14 16:21:00 -07002850 kvm_exit_debug();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002851 misc_deregister(&kvm_dev);
Rusty Russellc16f8622007-07-30 21:12:19 +10002852 kmem_cache_destroy(kvm_vcpu_cache);
Gleb Natapovaf585b92010-10-14 11:22:46 +02002853 kvm_async_pf_deinit();
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002854 unregister_syscore_ops(&kvm_syscore_ops);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002855 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002856 unregister_cpu_notifier(&kvm_cpu_notifier);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002857 on_each_cpu(hardware_disable_nolock, NULL, 1);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002858 kvm_arch_hardware_unsetup();
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002859 kvm_arch_exit();
Rusty Russell7f59f492008-12-07 21:25:45 +10302860 free_cpumask_var(cpus_hardware_enabled);
Xiao Guangrongf4119302012-07-03 14:31:01 +08002861 __free_page(fault_page);
Huang Yingbf998152010-05-31 14:28:19 +08002862 __free_page(hwpoison_page);
Izik Eiduscea7bb22007-10-17 19:17:48 +02002863 __free_page(bad_page);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002864}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002865EXPORT_SYMBOL_GPL(kvm_exit);