blob: d9cfb782cb81a87fc5ef4f1734df1a308a55539b [file] [log] [blame]
Avi Kivity6aa8b732006-12-10 02:21:36 -08001/*
2 * Kernel-based Virtual Machine driver for Linux
3 *
4 * This module enables machines with Intel VT-x extensions to run virtual
5 * machines without emulation or binary translation.
6 *
7 * Copyright (C) 2006 Qumranet, Inc.
Nicolas Kaiser9611c182010-10-06 14:23:22 +02008 * Copyright 2010 Red Hat, Inc. and/or its affiliates.
Avi Kivity6aa8b732006-12-10 02:21:36 -08009 *
10 * Authors:
11 * Avi Kivity <avi@qumranet.com>
12 * Yaniv Kamay <yaniv@qumranet.com>
13 *
14 * This work is licensed under the terms of the GNU GPL, version 2. See
15 * the COPYING file in the top-level directory.
16 *
17 */
18
Hollis Blancharde2174022007-12-03 15:30:24 -060019#include "iodev.h"
Avi Kivity6aa8b732006-12-10 02:21:36 -080020
Avi Kivityedf88412007-12-16 11:02:48 +020021#include <linux/kvm_host.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080022#include <linux/kvm.h>
23#include <linux/module.h>
24#include <linux/errno.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080025#include <linux/percpu.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080026#include <linux/mm.h>
27#include <linux/miscdevice.h>
28#include <linux/vmalloc.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080029#include <linux/reboot.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080030#include <linux/debugfs.h>
31#include <linux/highmem.h>
32#include <linux/file.h>
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +010033#include <linux/syscore_ops.h>
Avi Kivity774c47f2007-02-12 00:54:47 -080034#include <linux/cpu.h>
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +040035#include <linux/sched.h>
Avi Kivityd9e368d2007-06-07 19:18:30 +030036#include <linux/cpumask.h>
37#include <linux/smp.h>
Avi Kivityd6d28162007-06-28 08:38:16 -040038#include <linux/anon_inodes.h>
Avi Kivity04d2cc72007-09-10 18:10:54 +030039#include <linux/profile.h>
Anthony Liguori7aa81cc2007-09-17 14:57:50 -050040#include <linux/kvm_para.h>
Izik Eidus6fc138d2007-10-09 19:20:39 +020041#include <linux/pagemap.h>
Anthony Liguori8d4e1282007-10-18 09:59:34 -050042#include <linux/mman.h>
Anthony Liguori35149e22008-04-02 14:46:56 -050043#include <linux/swap.h>
Sheng Yange56d5322009-03-12 21:45:39 +080044#include <linux/bitops.h>
Marcelo Tosatti547de292009-05-07 17:55:13 -030045#include <linux/spinlock.h>
Arnd Bergmann6ff58942009-10-22 14:19:27 +020046#include <linux/compat.h>
Marcelo Tosattibc6678a2009-12-23 14:35:21 -020047#include <linux/srcu.h>
Joerg Roedel8f0b1ab2010-01-28 12:37:56 +010048#include <linux/hugetlb.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090049#include <linux/slab.h>
Sasha Levin743eeb02011-07-27 16:00:48 +030050#include <linux/sort.h>
51#include <linux/bsearch.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080052
Avi Kivitye4956062007-06-28 14:15:57 -040053#include <asm/processor.h>
Avi Kivitye4956062007-06-28 14:15:57 -040054#include <asm/io.h>
55#include <asm/uaccess.h>
Izik Eidus3e021bf2007-11-19 11:16:57 +020056#include <asm/pgtable.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080057
Laurent Vivier5f94c172008-05-30 16:05:54 +020058#include "coalesced_mmio.h"
Gleb Natapovaf585b92010-10-14 11:22:46 +020059#include "async_pf.h"
Laurent Vivier5f94c172008-05-30 16:05:54 +020060
Marcelo Tosatti229456f2009-06-17 09:22:14 -030061#define CREATE_TRACE_POINTS
62#include <trace/events/kvm.h>
63
Avi Kivity6aa8b732006-12-10 02:21:36 -080064MODULE_AUTHOR("Qumranet");
65MODULE_LICENSE("GPL");
66
Marcelo Tosattifa40a822009-06-04 15:08:24 -030067/*
68 * Ordering of locks:
69 *
Sheng Yangfae3a352009-12-15 10:28:07 +080070 * kvm->lock --> kvm->slots_lock --> kvm->irq_lock
Marcelo Tosattifa40a822009-06-04 15:08:24 -030071 */
72
Jan Kiszkae935b832011-02-08 12:55:33 +010073DEFINE_RAW_SPINLOCK(kvm_lock);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +080074LIST_HEAD(vm_list);
Avi Kivity133de902007-02-12 00:54:44 -080075
Rusty Russell7f59f492008-12-07 21:25:45 +103076static cpumask_var_t cpus_hardware_enabled;
Alexander Graf10474ae2009-09-15 11:37:46 +020077static int kvm_usage_count = 0;
78static atomic_t hardware_enable_failed;
Avi Kivity1b6c0162007-05-24 13:03:52 +030079
Rusty Russellc16f8622007-07-30 21:12:19 +100080struct kmem_cache *kvm_vcpu_cache;
81EXPORT_SYMBOL_GPL(kvm_vcpu_cache);
Avi Kivity1165f5f2007-04-19 17:27:43 +030082
Avi Kivity15ad7142007-07-11 18:17:21 +030083static __read_mostly struct preempt_ops kvm_preempt_ops;
84
Hollis Blanchard76f7c872008-04-15 16:05:42 -050085struct dentry *kvm_debugfs_dir;
Avi Kivity6aa8b732006-12-10 02:21:36 -080086
Avi Kivitybccf2152007-02-21 18:04:26 +020087static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl,
88 unsigned long arg);
Alexander Graf1dda6062011-06-08 02:45:37 +020089#ifdef CONFIG_COMPAT
90static long kvm_vcpu_compat_ioctl(struct file *file, unsigned int ioctl,
91 unsigned long arg);
92#endif
Alexander Graf10474ae2009-09-15 11:37:46 +020093static int hardware_enable_all(void);
94static void hardware_disable_all(void);
Avi Kivitybccf2152007-02-21 18:04:26 +020095
Marcelo Tosattie93f8a02009-12-23 14:35:24 -020096static void kvm_io_bus_destroy(struct kvm_io_bus *bus);
97
Avi Kivityb7c41452010-12-02 17:52:50 +020098bool kvm_rebooting;
99EXPORT_SYMBOL_GPL(kvm_rebooting);
Avi Kivity4ecac3f2008-05-13 13:23:38 +0300100
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300101static bool largepages_enabled = true;
102
Gleb Natapovfa7bff82010-07-07 20:16:44 +0300103static struct page *hwpoison_page;
104static pfn_t hwpoison_pfn;
Huang Yingbf998152010-05-31 14:28:19 +0800105
Xiao Guangrongfce92dc2011-07-12 03:28:54 +0800106struct page *fault_page;
107pfn_t fault_pfn;
Gleb Natapovedba23e2010-07-07 20:16:45 +0300108
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +0800109inline int kvm_is_mmio_pfn(pfn_t pfn)
Ben-Ami Yassourcbff90a2008-07-28 19:26:24 +0300110{
Joerg Roedelfc5659c2009-02-18 14:08:58 +0100111 if (pfn_valid(pfn)) {
Andrea Arcangeli22e5c472011-01-13 15:47:20 -0800112 int reserved;
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800113 struct page *tail = pfn_to_page(pfn);
Andrea Arcangeli22e5c472011-01-13 15:47:20 -0800114 struct page *head = compound_trans_head(tail);
115 reserved = PageReserved(head);
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800116 if (head != tail) {
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800117 /*
Andrea Arcangeli22e5c472011-01-13 15:47:20 -0800118 * "head" is not a dangling pointer
119 * (compound_trans_head takes care of that)
120 * but the hugepage may have been splitted
121 * from under us (and we may not hold a
122 * reference count on the head page so it can
123 * be reused before we run PageReferenced), so
124 * we've to check PageTail before returning
125 * what we just read.
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800126 */
Andrea Arcangeli22e5c472011-01-13 15:47:20 -0800127 smp_rmb();
128 if (PageTail(tail))
129 return reserved;
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800130 }
131 return PageReserved(tail);
Joerg Roedelfc5659c2009-02-18 14:08:58 +0100132 }
Ben-Ami Yassourcbff90a2008-07-28 19:26:24 +0300133
134 return true;
135}
136
Avi Kivity6aa8b732006-12-10 02:21:36 -0800137/*
138 * Switches to specified vcpu, until a matching vcpu_put()
139 */
Carsten Otte313a3dc2007-10-11 19:16:52 +0200140void vcpu_load(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800141{
Avi Kivity15ad7142007-07-11 18:17:21 +0300142 int cpu;
143
Avi Kivitybccf2152007-02-21 18:04:26 +0200144 mutex_lock(&vcpu->mutex);
Rik van Riel34bb10b2011-02-01 09:52:41 -0500145 if (unlikely(vcpu->pid != current->pids[PIDTYPE_PID].pid)) {
146 /* The thread running this VCPU changed. */
147 struct pid *oldpid = vcpu->pid;
148 struct pid *newpid = get_task_pid(current, PIDTYPE_PID);
149 rcu_assign_pointer(vcpu->pid, newpid);
150 synchronize_rcu();
151 put_pid(oldpid);
152 }
Avi Kivity15ad7142007-07-11 18:17:21 +0300153 cpu = get_cpu();
154 preempt_notifier_register(&vcpu->preempt_notifier);
Carsten Otte313a3dc2007-10-11 19:16:52 +0200155 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +0300156 put_cpu();
Avi Kivitybccf2152007-02-21 18:04:26 +0200157}
158
Carsten Otte313a3dc2007-10-11 19:16:52 +0200159void vcpu_put(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800160{
Avi Kivity15ad7142007-07-11 18:17:21 +0300161 preempt_disable();
Carsten Otte313a3dc2007-10-11 19:16:52 +0200162 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +0300163 preempt_notifier_unregister(&vcpu->preempt_notifier);
164 preempt_enable();
Avi Kivity6aa8b732006-12-10 02:21:36 -0800165 mutex_unlock(&vcpu->mutex);
166}
167
Avi Kivityd9e368d2007-06-07 19:18:30 +0300168static void ack_flush(void *_completed)
169{
Avi Kivityd9e368d2007-06-07 19:18:30 +0300170}
171
Rusty Russell49846892008-12-08 20:26:24 +1030172static bool make_all_cpus_request(struct kvm *kvm, unsigned int req)
Avi Kivityd9e368d2007-06-07 19:18:30 +0300173{
Avi Kivity597a5f52008-07-20 14:24:22 +0300174 int i, cpu, me;
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030175 cpumask_var_t cpus;
176 bool called = true;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300177 struct kvm_vcpu *vcpu;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300178
Li Zefan79f55992009-06-15 14:58:26 +0800179 zalloc_cpumask_var(&cpus, GFP_ATOMIC);
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030180
Xiao Guangrong3cba4132011-01-12 15:41:22 +0800181 me = get_cpu();
Gleb Natapov988a2ca2009-06-09 15:56:29 +0300182 kvm_for_each_vcpu(i, vcpu, kvm) {
Xiao Guangrong3cba4132011-01-12 15:41:22 +0800183 kvm_make_request(req, vcpu);
Avi Kivityd9e368d2007-06-07 19:18:30 +0300184 cpu = vcpu->cpu;
Xiao Guangrong6b7e2d02011-01-12 15:40:31 +0800185
186 /* Set ->requests bit before we read ->mode */
187 smp_mb();
188
189 if (cpus != NULL && cpu != -1 && cpu != me &&
190 kvm_vcpu_exiting_guest_mode(vcpu) != OUTSIDE_GUEST_MODE)
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030191 cpumask_set_cpu(cpu, cpus);
Avi Kivityd9e368d2007-06-07 19:18:30 +0300192 }
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030193 if (unlikely(cpus == NULL))
194 smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1);
195 else if (!cpumask_empty(cpus))
196 smp_call_function_many(cpus, ack_flush, NULL, 1);
197 else
198 called = false;
Xiao Guangrong3cba4132011-01-12 15:41:22 +0800199 put_cpu();
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030200 free_cpumask_var(cpus);
Rusty Russell49846892008-12-08 20:26:24 +1030201 return called;
202}
203
204void kvm_flush_remote_tlbs(struct kvm *kvm)
205{
Xiao Guangronga4ee1ca2010-11-23 11:13:00 +0800206 int dirty_count = kvm->tlbs_dirty;
207
208 smp_mb();
Rusty Russell49846892008-12-08 20:26:24 +1030209 if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH))
210 ++kvm->stat.remote_tlb_flush;
Xiao Guangronga4ee1ca2010-11-23 11:13:00 +0800211 cmpxchg(&kvm->tlbs_dirty, dirty_count, 0);
Avi Kivityd9e368d2007-06-07 19:18:30 +0300212}
213
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500214void kvm_reload_remote_mmus(struct kvm *kvm)
215{
Rusty Russell49846892008-12-08 20:26:24 +1030216 make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD);
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500217}
218
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000219int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id)
220{
221 struct page *page;
222 int r;
223
224 mutex_init(&vcpu->mutex);
225 vcpu->cpu = -1;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000226 vcpu->kvm = kvm;
227 vcpu->vcpu_id = id;
Rik van Riel34bb10b2011-02-01 09:52:41 -0500228 vcpu->pid = NULL;
Eddie Dongb6958ce2007-07-18 12:15:21 +0300229 init_waitqueue_head(&vcpu->wq);
Gleb Natapovaf585b92010-10-14 11:22:46 +0200230 kvm_async_pf_vcpu_init(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000231
232 page = alloc_page(GFP_KERNEL | __GFP_ZERO);
233 if (!page) {
234 r = -ENOMEM;
235 goto fail;
236 }
237 vcpu->run = page_address(page);
238
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800239 r = kvm_arch_vcpu_init(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000240 if (r < 0)
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800241 goto fail_free_run;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000242 return 0;
243
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000244fail_free_run:
245 free_page((unsigned long)vcpu->run);
246fail:
Rusty Russell76fafa52007-10-08 10:50:48 +1000247 return r;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000248}
249EXPORT_SYMBOL_GPL(kvm_vcpu_init);
250
251void kvm_vcpu_uninit(struct kvm_vcpu *vcpu)
252{
Rik van Riel34bb10b2011-02-01 09:52:41 -0500253 put_pid(vcpu->pid);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800254 kvm_arch_vcpu_uninit(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000255 free_page((unsigned long)vcpu->run);
256}
257EXPORT_SYMBOL_GPL(kvm_vcpu_uninit);
258
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200259#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
260static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn)
261{
262 return container_of(mn, struct kvm, mmu_notifier);
263}
264
265static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn,
266 struct mm_struct *mm,
267 unsigned long address)
268{
269 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200270 int need_tlb_flush, idx;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200271
272 /*
273 * When ->invalidate_page runs, the linux pte has been zapped
274 * already but the page is still allocated until
275 * ->invalidate_page returns. So if we increase the sequence
276 * here the kvm page fault will notice if the spte can't be
277 * established because the page is going to be freed. If
278 * instead the kvm page fault establishes the spte before
279 * ->invalidate_page runs, kvm_unmap_hva will release it
280 * before returning.
281 *
282 * The sequence increase only need to be seen at spin_unlock
283 * time, and not at spin_lock time.
284 *
285 * Increasing the sequence after the spin_unlock would be
286 * unsafe because the kvm page fault could then establish the
287 * pte after kvm_unmap_hva returned, without noticing the page
288 * is going to be freed.
289 */
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200290 idx = srcu_read_lock(&kvm->srcu);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200291 spin_lock(&kvm->mmu_lock);
292 kvm->mmu_notifier_seq++;
Xiao Guangronga4ee1ca2010-11-23 11:13:00 +0800293 need_tlb_flush = kvm_unmap_hva(kvm, address) | kvm->tlbs_dirty;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200294 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200295 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200296
297 /* we've to flush the tlb before the pages can be freed */
298 if (need_tlb_flush)
299 kvm_flush_remote_tlbs(kvm);
300
301}
302
Izik Eidus3da0dd42009-09-23 21:47:18 +0300303static void kvm_mmu_notifier_change_pte(struct mmu_notifier *mn,
304 struct mm_struct *mm,
305 unsigned long address,
306 pte_t pte)
307{
308 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200309 int idx;
Izik Eidus3da0dd42009-09-23 21:47:18 +0300310
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200311 idx = srcu_read_lock(&kvm->srcu);
Izik Eidus3da0dd42009-09-23 21:47:18 +0300312 spin_lock(&kvm->mmu_lock);
313 kvm->mmu_notifier_seq++;
314 kvm_set_spte_hva(kvm, address, pte);
315 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200316 srcu_read_unlock(&kvm->srcu, idx);
Izik Eidus3da0dd42009-09-23 21:47:18 +0300317}
318
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200319static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn,
320 struct mm_struct *mm,
321 unsigned long start,
322 unsigned long end)
323{
324 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200325 int need_tlb_flush = 0, idx;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200326
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200327 idx = srcu_read_lock(&kvm->srcu);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200328 spin_lock(&kvm->mmu_lock);
329 /*
330 * The count increase must become visible at unlock time as no
331 * spte can be established without taking the mmu_lock and
332 * count is also read inside the mmu_lock critical section.
333 */
334 kvm->mmu_notifier_count++;
335 for (; start < end; start += PAGE_SIZE)
336 need_tlb_flush |= kvm_unmap_hva(kvm, start);
Xiao Guangronga4ee1ca2010-11-23 11:13:00 +0800337 need_tlb_flush |= kvm->tlbs_dirty;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200338 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200339 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200340
341 /* we've to flush the tlb before the pages can be freed */
342 if (need_tlb_flush)
343 kvm_flush_remote_tlbs(kvm);
344}
345
346static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
347 struct mm_struct *mm,
348 unsigned long start,
349 unsigned long end)
350{
351 struct kvm *kvm = mmu_notifier_to_kvm(mn);
352
353 spin_lock(&kvm->mmu_lock);
354 /*
355 * This sequence increase will notify the kvm page fault that
356 * the page that is going to be mapped in the spte could have
357 * been freed.
358 */
359 kvm->mmu_notifier_seq++;
360 /*
361 * The above sequence increase must be visible before the
362 * below count decrease but both values are read by the kvm
363 * page fault under mmu_lock spinlock so we don't need to add
364 * a smb_wmb() here in between the two.
365 */
366 kvm->mmu_notifier_count--;
367 spin_unlock(&kvm->mmu_lock);
368
369 BUG_ON(kvm->mmu_notifier_count < 0);
370}
371
372static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
373 struct mm_struct *mm,
374 unsigned long address)
375{
376 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200377 int young, idx;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200378
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200379 idx = srcu_read_lock(&kvm->srcu);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200380 spin_lock(&kvm->mmu_lock);
381 young = kvm_age_hva(kvm, address);
382 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200383 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200384
385 if (young)
386 kvm_flush_remote_tlbs(kvm);
387
388 return young;
389}
390
Andrea Arcangeli8ee53822011-01-13 15:47:10 -0800391static int kvm_mmu_notifier_test_young(struct mmu_notifier *mn,
392 struct mm_struct *mm,
393 unsigned long address)
394{
395 struct kvm *kvm = mmu_notifier_to_kvm(mn);
396 int young, idx;
397
398 idx = srcu_read_lock(&kvm->srcu);
399 spin_lock(&kvm->mmu_lock);
400 young = kvm_test_age_hva(kvm, address);
401 spin_unlock(&kvm->mmu_lock);
402 srcu_read_unlock(&kvm->srcu, idx);
403
404 return young;
405}
406
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100407static void kvm_mmu_notifier_release(struct mmu_notifier *mn,
408 struct mm_struct *mm)
409{
410 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Lai Jiangshaneda2bed2010-04-20 14:29:29 +0800411 int idx;
412
413 idx = srcu_read_lock(&kvm->srcu);
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100414 kvm_arch_flush_shadow(kvm);
Lai Jiangshaneda2bed2010-04-20 14:29:29 +0800415 srcu_read_unlock(&kvm->srcu, idx);
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100416}
417
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200418static const struct mmu_notifier_ops kvm_mmu_notifier_ops = {
419 .invalidate_page = kvm_mmu_notifier_invalidate_page,
420 .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start,
421 .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end,
422 .clear_flush_young = kvm_mmu_notifier_clear_flush_young,
Andrea Arcangeli8ee53822011-01-13 15:47:10 -0800423 .test_young = kvm_mmu_notifier_test_young,
Izik Eidus3da0dd42009-09-23 21:47:18 +0300424 .change_pte = kvm_mmu_notifier_change_pte,
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100425 .release = kvm_mmu_notifier_release,
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200426};
Avi Kivity4c07b0a2009-12-20 14:54:04 +0200427
428static int kvm_init_mmu_notifier(struct kvm *kvm)
429{
430 kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops;
431 return mmu_notifier_register(&kvm->mmu_notifier, current->mm);
432}
433
434#else /* !(CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER) */
435
436static int kvm_init_mmu_notifier(struct kvm *kvm)
437{
438 return 0;
439}
440
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200441#endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */
442
Avi Kivityf17abe92007-02-21 19:28:04 +0200443static struct kvm *kvm_create_vm(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800444{
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100445 int r, i;
446 struct kvm *kvm = kvm_arch_alloc_vm();
Avi Kivity6aa8b732006-12-10 02:21:36 -0800447
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100448 if (!kvm)
449 return ERR_PTR(-ENOMEM);
450
451 r = kvm_arch_init_vm(kvm);
452 if (r)
453 goto out_err_nodisable;
Alexander Graf10474ae2009-09-15 11:37:46 +0200454
455 r = hardware_enable_all();
456 if (r)
457 goto out_err_nodisable;
458
Avi Kivity75858a82009-01-04 17:10:50 +0200459#ifdef CONFIG_HAVE_KVM_IRQCHIP
460 INIT_HLIST_HEAD(&kvm->mask_notifier_list);
Gleb Natapov136bdfe2009-08-24 11:54:23 +0300461 INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list);
Avi Kivity75858a82009-01-04 17:10:50 +0200462#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -0800463
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200464 r = -ENOMEM;
465 kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
466 if (!kvm->memslots)
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100467 goto out_err_nosrcu;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200468 if (init_srcu_struct(&kvm->srcu))
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100469 goto out_err_nosrcu;
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200470 for (i = 0; i < KVM_NR_BUSES; i++) {
471 kvm->buses[i] = kzalloc(sizeof(struct kvm_io_bus),
472 GFP_KERNEL);
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100473 if (!kvm->buses[i])
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200474 goto out_err;
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200475 }
Mike Waychison74b5c5b2011-06-03 13:04:53 -0700476
OGAWA Hirofumi85722cd2011-05-11 09:28:28 +0900477 spin_lock_init(&kvm->mmu_lock);
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200478 kvm->mm = current->mm;
479 atomic_inc(&kvm->mm->mm_count);
Gregory Haskinsd34e6b12009-07-07 17:08:49 -0400480 kvm_eventfd_init(kvm);
Shaohua Li11ec2802007-07-23 14:51:37 +0800481 mutex_init(&kvm->lock);
Marcelo Tosatti60eead72009-06-04 15:08:23 -0300482 mutex_init(&kvm->irq_lock);
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200483 mutex_init(&kvm->slots_lock);
Izik Eidusd39f13b2008-03-30 16:01:25 +0300484 atomic_set(&kvm->users_count, 1);
Mike Waychison74b5c5b2011-06-03 13:04:53 -0700485
486 r = kvm_init_mmu_notifier(kvm);
487 if (r)
488 goto out_err;
489
Jan Kiszkae935b832011-02-08 12:55:33 +0100490 raw_spin_lock(&kvm_lock);
Rusty Russell5e58cfe2007-07-23 17:08:21 +1000491 list_add(&kvm->vm_list, &vm_list);
Jan Kiszkae935b832011-02-08 12:55:33 +0100492 raw_spin_unlock(&kvm_lock);
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100493
Avi Kivityf17abe92007-02-21 19:28:04 +0200494 return kvm;
Alexander Graf10474ae2009-09-15 11:37:46 +0200495
496out_err:
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100497 cleanup_srcu_struct(&kvm->srcu);
498out_err_nosrcu:
Alexander Graf10474ae2009-09-15 11:37:46 +0200499 hardware_disable_all();
500out_err_nodisable:
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200501 for (i = 0; i < KVM_NR_BUSES; i++)
502 kfree(kvm->buses[i]);
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200503 kfree(kvm->memslots);
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100504 kvm_arch_free_vm(kvm);
Alexander Graf10474ae2009-09-15 11:37:46 +0200505 return ERR_PTR(r);
Avi Kivityf17abe92007-02-21 19:28:04 +0200506}
507
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900508static void kvm_destroy_dirty_bitmap(struct kvm_memory_slot *memslot)
509{
510 if (!memslot->dirty_bitmap)
511 return;
512
Takuya Yoshikawa6f9e5c172010-11-01 14:36:09 +0900513 if (2 * kvm_dirty_bitmap_bytes(memslot) > PAGE_SIZE)
514 vfree(memslot->dirty_bitmap_head);
515 else
516 kfree(memslot->dirty_bitmap_head);
517
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900518 memslot->dirty_bitmap = NULL;
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900519 memslot->dirty_bitmap_head = NULL;
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900520}
521
Avi Kivity6aa8b732006-12-10 02:21:36 -0800522/*
523 * Free any memory in @free but not in @dont.
524 */
525static void kvm_free_physmem_slot(struct kvm_memory_slot *free,
526 struct kvm_memory_slot *dont)
527{
Joerg Roedelec04b262009-06-19 15:16:23 +0200528 int i;
529
Izik Eidus290fc382007-09-27 14:11:22 +0200530 if (!dont || free->rmap != dont->rmap)
531 vfree(free->rmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800532
533 if (!dont || free->dirty_bitmap != dont->dirty_bitmap)
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900534 kvm_destroy_dirty_bitmap(free);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800535
Joerg Roedelec04b262009-06-19 15:16:23 +0200536
537 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) {
538 if (!dont || free->lpage_info[i] != dont->lpage_info[i]) {
539 vfree(free->lpage_info[i]);
540 free->lpage_info[i] = NULL;
541 }
542 }
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300543
Avi Kivity6aa8b732006-12-10 02:21:36 -0800544 free->npages = 0;
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500545 free->rmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800546}
547
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800548void kvm_free_physmem(struct kvm *kvm)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800549{
550 int i;
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200551 struct kvm_memslots *slots = kvm->memslots;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800552
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200553 for (i = 0; i < slots->nmemslots; ++i)
554 kvm_free_physmem_slot(&slots->memslots[i], NULL);
555
556 kfree(kvm->memslots);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800557}
558
Avi Kivityf17abe92007-02-21 19:28:04 +0200559static void kvm_destroy_vm(struct kvm *kvm)
560{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200561 int i;
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200562 struct mm_struct *mm = kvm->mm;
563
Sheng Yangad8ba2c2009-01-06 10:03:02 +0800564 kvm_arch_sync_events(kvm);
Jan Kiszkae935b832011-02-08 12:55:33 +0100565 raw_spin_lock(&kvm_lock);
Avi Kivity133de902007-02-12 00:54:44 -0800566 list_del(&kvm->vm_list);
Jan Kiszkae935b832011-02-08 12:55:33 +0100567 raw_spin_unlock(&kvm_lock);
Avi Kivity399ec802008-11-19 13:58:46 +0200568 kvm_free_irq_routing(kvm);
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200569 for (i = 0; i < KVM_NR_BUSES; i++)
570 kvm_io_bus_destroy(kvm->buses[i]);
Avi Kivity980da6c2009-12-20 15:13:43 +0200571 kvm_coalesced_mmio_free(kvm);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200572#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
573 mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
Gleb Natapovf00be0c2009-03-19 12:20:36 +0200574#else
575 kvm_arch_flush_shadow(kvm);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200576#endif
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800577 kvm_arch_destroy_vm(kvm);
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100578 kvm_free_physmem(kvm);
579 cleanup_srcu_struct(&kvm->srcu);
580 kvm_arch_free_vm(kvm);
Alexander Graf10474ae2009-09-15 11:37:46 +0200581 hardware_disable_all();
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200582 mmdrop(mm);
Avi Kivityf17abe92007-02-21 19:28:04 +0200583}
584
Izik Eidusd39f13b2008-03-30 16:01:25 +0300585void kvm_get_kvm(struct kvm *kvm)
586{
587 atomic_inc(&kvm->users_count);
588}
589EXPORT_SYMBOL_GPL(kvm_get_kvm);
590
591void kvm_put_kvm(struct kvm *kvm)
592{
593 if (atomic_dec_and_test(&kvm->users_count))
594 kvm_destroy_vm(kvm);
595}
596EXPORT_SYMBOL_GPL(kvm_put_kvm);
597
598
Avi Kivityf17abe92007-02-21 19:28:04 +0200599static int kvm_vm_release(struct inode *inode, struct file *filp)
600{
601 struct kvm *kvm = filp->private_data;
602
Gregory Haskins721eecb2009-05-20 10:30:49 -0400603 kvm_irqfd_release(kvm);
604
Izik Eidusd39f13b2008-03-30 16:01:25 +0300605 kvm_put_kvm(kvm);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800606 return 0;
607}
608
Heiko Carstensd48ead82011-01-17 21:21:08 +0100609#ifndef CONFIG_S390
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900610/*
611 * Allocation size is twice as large as the actual dirty bitmap size.
612 * This makes it possible to do double buffering: see x86's
613 * kvm_vm_ioctl_get_dirty_log().
614 */
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900615static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot)
616{
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900617 unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot);
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900618
Takuya Yoshikawa6f9e5c172010-11-01 14:36:09 +0900619 if (dirty_bytes > PAGE_SIZE)
620 memslot->dirty_bitmap = vzalloc(dirty_bytes);
621 else
622 memslot->dirty_bitmap = kzalloc(dirty_bytes, GFP_KERNEL);
623
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900624 if (!memslot->dirty_bitmap)
625 return -ENOMEM;
626
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900627 memslot->dirty_bitmap_head = memslot->dirty_bitmap;
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900628 return 0;
629}
Heiko Carstensd48ead82011-01-17 21:21:08 +0100630#endif /* !CONFIG_S390 */
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900631
Avi Kivity6aa8b732006-12-10 02:21:36 -0800632/*
Avi Kivity6aa8b732006-12-10 02:21:36 -0800633 * Allocate some memory and give it an address in the guest physical address
634 * space.
635 *
636 * Discontiguous memory is allowed, mostly for framebuffers.
Sheng Yangf78e0e22007-10-29 09:40:42 +0800637 *
Marcelo Tosatti10589a42007-12-20 19:18:22 -0500638 * Must be called holding mmap_sem for write.
Avi Kivity6aa8b732006-12-10 02:21:36 -0800639 */
Sheng Yangf78e0e22007-10-29 09:40:42 +0800640int __kvm_set_memory_region(struct kvm *kvm,
641 struct kvm_userspace_memory_region *mem,
642 int user_alloc)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800643{
Avi Kivity8234b222010-12-27 12:08:45 +0200644 int r;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800645 gfn_t base_gfn;
Heiko Carstens28bcb112009-09-03 17:35:35 +0200646 unsigned long npages;
647 unsigned long i;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800648 struct kvm_memory_slot *memslot;
649 struct kvm_memory_slot old, new;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200650 struct kvm_memslots *slots, *old_memslots;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800651
652 r = -EINVAL;
653 /* General sanity checks */
654 if (mem->memory_size & (PAGE_SIZE - 1))
655 goto out;
656 if (mem->guest_phys_addr & (PAGE_SIZE - 1))
657 goto out;
Takuya Yoshikawafa3d3152011-05-07 16:35:38 +0900658 /* We can read the guest memory with __xxx_user() later on. */
659 if (user_alloc &&
660 ((mem->userspace_addr & (PAGE_SIZE - 1)) ||
Heiko Carstens9e3bb6b2011-05-24 07:51:27 +0200661 !access_ok(VERIFY_WRITE,
662 (void __user *)(unsigned long)mem->userspace_addr,
663 mem->memory_size)))
Hollis Blanchard78749802008-11-07 13:32:12 -0600664 goto out;
Izik Eiduse0d62c72007-10-24 23:57:46 +0200665 if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800666 goto out;
667 if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
668 goto out;
669
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200670 memslot = &kvm->memslots->memslots[mem->slot];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800671 base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
672 npages = mem->memory_size >> PAGE_SHIFT;
673
Takuya Yoshikawa660c22c2010-04-13 22:47:24 +0900674 r = -EINVAL;
675 if (npages > KVM_MEM_MAX_NR_PAGES)
676 goto out;
677
Avi Kivity6aa8b732006-12-10 02:21:36 -0800678 if (!npages)
679 mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES;
680
Avi Kivity6aa8b732006-12-10 02:21:36 -0800681 new = old = *memslot;
682
Avi Kivitye36d96f2010-06-21 10:56:36 +0300683 new.id = mem->slot;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800684 new.base_gfn = base_gfn;
685 new.npages = npages;
686 new.flags = mem->flags;
687
688 /* Disallow changing a memory slot's size. */
689 r = -EINVAL;
690 if (npages && old.npages && npages != old.npages)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800691 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800692
693 /* Check for overlaps */
694 r = -EEXIST;
695 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200696 struct kvm_memory_slot *s = &kvm->memslots->memslots[i];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800697
Jan Kiszka4cd481f2009-04-13 11:59:32 +0200698 if (s == memslot || !s->npages)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800699 continue;
700 if (!((base_gfn + npages <= s->base_gfn) ||
701 (base_gfn >= s->base_gfn + s->npages)))
Sheng Yangf78e0e22007-10-29 09:40:42 +0800702 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800703 }
Avi Kivity6aa8b732006-12-10 02:21:36 -0800704
Avi Kivity6aa8b732006-12-10 02:21:36 -0800705 /* Free page dirty bitmap if unneeded */
706 if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
Al Viro8b6d44c2007-02-09 16:38:40 +0000707 new.dirty_bitmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800708
709 r = -ENOMEM;
710
711 /* Allocate if a slot is being created */
Carsten Otteeff01142008-06-27 15:05:31 +0200712#ifndef CONFIG_S390
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500713 if (npages && !new.rmap) {
Takuya Yoshikawa26535032010-11-02 10:49:34 +0900714 new.rmap = vzalloc(npages * sizeof(*new.rmap));
Izik Eidus290fc382007-09-27 14:11:22 +0200715
716 if (!new.rmap)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800717 goto out_free;
Izik Eidus290fc382007-09-27 14:11:22 +0200718
Izik Eidus80b14b52007-10-25 11:54:04 +0200719 new.user_alloc = user_alloc;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200720 new.userspace_addr = mem->userspace_addr;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800721 }
Joerg Roedelec04b262009-06-19 15:16:23 +0200722 if (!npages)
723 goto skip_lpage;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300724
Joerg Roedelec04b262009-06-19 15:16:23 +0200725 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) {
Heiko Carstens28bcb112009-09-03 17:35:35 +0200726 unsigned long ugfn;
727 unsigned long j;
728 int lpages;
Joerg Roedelec04b262009-06-19 15:16:23 +0200729 int level = i + 2;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300730
Joerg Roedelec04b262009-06-19 15:16:23 +0200731 /* Avoid unused variable warning if no large pages */
732 (void)level;
733
734 if (new.lpage_info[i])
735 continue;
736
Joerg Roedel82855412010-07-01 16:00:11 +0200737 lpages = 1 + ((base_gfn + npages - 1)
738 >> KVM_HPAGE_GFN_SHIFT(level));
739 lpages -= base_gfn >> KVM_HPAGE_GFN_SHIFT(level);
Joerg Roedelec04b262009-06-19 15:16:23 +0200740
Takuya Yoshikawa26535032010-11-02 10:49:34 +0900741 new.lpage_info[i] = vzalloc(lpages * sizeof(*new.lpage_info[i]));
Joerg Roedelec04b262009-06-19 15:16:23 +0200742
743 if (!new.lpage_info[i])
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300744 goto out_free;
745
Joerg Roedel82855412010-07-01 16:00:11 +0200746 if (base_gfn & (KVM_PAGES_PER_HPAGE(level) - 1))
Joerg Roedelec04b262009-06-19 15:16:23 +0200747 new.lpage_info[i][0].write_count = 1;
Joerg Roedel82855412010-07-01 16:00:11 +0200748 if ((base_gfn+npages) & (KVM_PAGES_PER_HPAGE(level) - 1))
Joerg Roedelec04b262009-06-19 15:16:23 +0200749 new.lpage_info[i][lpages - 1].write_count = 1;
Avi Kivityac045272009-06-08 15:52:39 +0300750 ugfn = new.userspace_addr >> PAGE_SHIFT;
751 /*
752 * If the gfn and userspace address are not aligned wrt each
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300753 * other, or if explicitly asked to, disable large page
754 * support for this slot
Avi Kivityac045272009-06-08 15:52:39 +0300755 */
Joerg Roedelec04b262009-06-19 15:16:23 +0200756 if ((base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1) ||
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300757 !largepages_enabled)
Joerg Roedelec04b262009-06-19 15:16:23 +0200758 for (j = 0; j < lpages; ++j)
759 new.lpage_info[i][j].write_count = 1;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300760 }
Avi Kivity6aa8b732006-12-10 02:21:36 -0800761
Joerg Roedelec04b262009-06-19 15:16:23 +0200762skip_lpage:
763
Avi Kivity6aa8b732006-12-10 02:21:36 -0800764 /* Allocate page dirty bitmap if needed */
765 if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900766 if (kvm_create_dirty_bitmap(&new) < 0)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800767 goto out_free;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200768 /* destroy any largepage mappings for dirty tracking */
Avi Kivity6aa8b732006-12-10 02:21:36 -0800769 }
Christian Borntraeger3eea8432009-06-23 17:24:06 +0200770#else /* not defined CONFIG_S390 */
771 new.user_alloc = user_alloc;
772 if (user_alloc)
773 new.userspace_addr = mem->userspace_addr;
Carsten Otteeff01142008-06-27 15:05:31 +0200774#endif /* not defined CONFIG_S390 */
Avi Kivity6aa8b732006-12-10 02:21:36 -0800775
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200776 if (!npages) {
777 r = -ENOMEM;
778 slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
779 if (!slots)
780 goto out_free;
781 memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots));
782 if (mem->slot >= slots->nmemslots)
783 slots->nmemslots = mem->slot + 1;
Gleb Natapov49c77542010-10-18 15:22:23 +0200784 slots->generation++;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200785 slots->memslots[mem->slot].flags |= KVM_MEMSLOT_INVALID;
786
787 old_memslots = kvm->memslots;
788 rcu_assign_pointer(kvm->memslots, slots);
789 synchronize_srcu_expedited(&kvm->srcu);
790 /* From this point no new shadow pages pointing to a deleted
791 * memslot will be created.
792 *
793 * validation of sp->gfn happens in:
794 * - gfn_to_hva (kvm_read_guest, gfn_to_pfn)
795 * - kvm_is_visible_gfn (mmu_check_roots)
796 */
Marcelo Tosatti34d4cb82008-07-10 20:49:31 -0300797 kvm_arch_flush_shadow(kvm);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200798 kfree(old_memslots);
799 }
Marcelo Tosatti34d4cb82008-07-10 20:49:31 -0300800
Marcelo Tosattif7784b82009-12-23 14:35:18 -0200801 r = kvm_arch_prepare_memory_region(kvm, &new, old, mem, user_alloc);
802 if (r)
803 goto out_free;
804
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200805 /* map the pages in iommu page table */
806 if (npages) {
807 r = kvm_iommu_map_pages(kvm, &new);
808 if (r)
809 goto out_free;
810 }
Andrea Arcangeli604b38a2008-07-25 16:32:03 +0200811
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200812 r = -ENOMEM;
813 slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
814 if (!slots)
815 goto out_free;
816 memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots));
817 if (mem->slot >= slots->nmemslots)
818 slots->nmemslots = mem->slot + 1;
Gleb Natapov49c77542010-10-18 15:22:23 +0200819 slots->generation++;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200820
821 /* actual memory is freed via old in kvm_free_physmem_slot below */
822 if (!npages) {
823 new.rmap = NULL;
824 new.dirty_bitmap = NULL;
825 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i)
826 new.lpage_info[i] = NULL;
827 }
828
829 slots->memslots[mem->slot] = new;
830 old_memslots = kvm->memslots;
831 rcu_assign_pointer(kvm->memslots, slots);
832 synchronize_srcu_expedited(&kvm->srcu);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800833
Marcelo Tosattif7784b82009-12-23 14:35:18 -0200834 kvm_arch_commit_memory_region(kvm, mem, old, user_alloc);
Zhang Xiantao3ad82a72007-11-20 13:11:38 +0800835
Xiao Guangrongce88dec2011-07-12 03:33:44 +0800836 /*
837 * If the new memory slot is created, we need to clear all
838 * mmio sptes.
839 */
840 if (npages && old.base_gfn != mem->guest_phys_addr >> PAGE_SHIFT)
841 kvm_arch_flush_shadow(kvm);
842
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200843 kvm_free_physmem_slot(&old, &new);
844 kfree(old_memslots);
845
Avi Kivity6aa8b732006-12-10 02:21:36 -0800846 return 0;
847
Sheng Yangf78e0e22007-10-29 09:40:42 +0800848out_free:
Avi Kivity6aa8b732006-12-10 02:21:36 -0800849 kvm_free_physmem_slot(&new, &old);
850out:
851 return r;
Izik Eidus210c7c42007-10-24 23:52:57 +0200852
853}
Sheng Yangf78e0e22007-10-29 09:40:42 +0800854EXPORT_SYMBOL_GPL(__kvm_set_memory_region);
855
856int kvm_set_memory_region(struct kvm *kvm,
857 struct kvm_userspace_memory_region *mem,
858 int user_alloc)
859{
860 int r;
861
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200862 mutex_lock(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +0800863 r = __kvm_set_memory_region(kvm, mem, user_alloc);
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200864 mutex_unlock(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +0800865 return r;
866}
Izik Eidus210c7c42007-10-24 23:52:57 +0200867EXPORT_SYMBOL_GPL(kvm_set_memory_region);
868
Carsten Otte1fe779f2007-10-29 16:08:35 +0100869int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
870 struct
871 kvm_userspace_memory_region *mem,
872 int user_alloc)
Izik Eidus210c7c42007-10-24 23:52:57 +0200873{
Izik Eiduse0d62c72007-10-24 23:57:46 +0200874 if (mem->slot >= KVM_MEMORY_SLOTS)
875 return -EINVAL;
Izik Eidus210c7c42007-10-24 23:52:57 +0200876 return kvm_set_memory_region(kvm, mem, user_alloc);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800877}
878
Zhang Xiantao5bb064d2007-11-18 20:29:43 +0800879int kvm_get_dirty_log(struct kvm *kvm,
880 struct kvm_dirty_log *log, int *is_dirty)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800881{
882 struct kvm_memory_slot *memslot;
883 int r, i;
Takuya Yoshikawa87bf6e72010-04-12 19:35:35 +0900884 unsigned long n;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800885 unsigned long any = 0;
886
Avi Kivity6aa8b732006-12-10 02:21:36 -0800887 r = -EINVAL;
888 if (log->slot >= KVM_MEMORY_SLOTS)
889 goto out;
890
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200891 memslot = &kvm->memslots->memslots[log->slot];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800892 r = -ENOENT;
893 if (!memslot->dirty_bitmap)
894 goto out;
895
Takuya Yoshikawa87bf6e72010-04-12 19:35:35 +0900896 n = kvm_dirty_bitmap_bytes(memslot);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800897
Uri Lublincd1a4a92007-02-22 16:43:09 +0200898 for (i = 0; !any && i < n/sizeof(long); ++i)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800899 any = memslot->dirty_bitmap[i];
900
901 r = -EFAULT;
902 if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
903 goto out;
904
Zhang Xiantao5bb064d2007-11-18 20:29:43 +0800905 if (any)
906 *is_dirty = 1;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800907
908 r = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800909out:
Avi Kivity6aa8b732006-12-10 02:21:36 -0800910 return r;
911}
912
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300913void kvm_disable_largepages(void)
914{
915 largepages_enabled = false;
916}
917EXPORT_SYMBOL_GPL(kvm_disable_largepages);
918
Izik Eiduscea7bb22007-10-17 19:17:48 +0200919int is_error_page(struct page *page)
920{
Gleb Natapovedba23e2010-07-07 20:16:45 +0300921 return page == bad_page || page == hwpoison_page || page == fault_page;
Izik Eiduscea7bb22007-10-17 19:17:48 +0200922}
923EXPORT_SYMBOL_GPL(is_error_page);
924
Anthony Liguori35149e22008-04-02 14:46:56 -0500925int is_error_pfn(pfn_t pfn)
926{
Gleb Natapovedba23e2010-07-07 20:16:45 +0300927 return pfn == bad_pfn || pfn == hwpoison_pfn || pfn == fault_pfn;
Anthony Liguori35149e22008-04-02 14:46:56 -0500928}
929EXPORT_SYMBOL_GPL(is_error_pfn);
930
Huang Yingbf998152010-05-31 14:28:19 +0800931int is_hwpoison_pfn(pfn_t pfn)
932{
933 return pfn == hwpoison_pfn;
934}
935EXPORT_SYMBOL_GPL(is_hwpoison_pfn);
936
Gleb Natapovedba23e2010-07-07 20:16:45 +0300937int is_fault_pfn(pfn_t pfn)
938{
939 return pfn == fault_pfn;
940}
941EXPORT_SYMBOL_GPL(is_fault_pfn);
942
Xiao Guangrongfce92dc2011-07-12 03:28:54 +0800943int is_noslot_pfn(pfn_t pfn)
944{
945 return pfn == bad_pfn;
946}
947EXPORT_SYMBOL_GPL(is_noslot_pfn);
948
949int is_invalid_pfn(pfn_t pfn)
950{
951 return pfn == hwpoison_pfn || pfn == fault_pfn;
952}
953EXPORT_SYMBOL_GPL(is_invalid_pfn);
954
Izik Eidusf9d46eb2007-11-11 22:02:22 +0200955static inline unsigned long bad_hva(void)
956{
957 return PAGE_OFFSET;
958}
959
960int kvm_is_error_hva(unsigned long addr)
961{
962 return addr == bad_hva();
963}
964EXPORT_SYMBOL_GPL(kvm_is_error_hva);
965
Gleb Natapov49c77542010-10-18 15:22:23 +0200966static struct kvm_memory_slot *__gfn_to_memslot(struct kvm_memslots *slots,
967 gfn_t gfn)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800968{
969 int i;
970
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200971 for (i = 0; i < slots->nmemslots; ++i) {
972 struct kvm_memory_slot *memslot = &slots->memslots[i];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800973
974 if (gfn >= memslot->base_gfn
975 && gfn < memslot->base_gfn + memslot->npages)
976 return memslot;
977 }
Al Viro8b6d44c2007-02-09 16:38:40 +0000978 return NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800979}
Gleb Natapov49c77542010-10-18 15:22:23 +0200980
981struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn)
982{
983 return __gfn_to_memslot(kvm_memslots(kvm), gfn);
984}
Avi Kivitya1f4d392010-06-21 11:44:20 +0300985EXPORT_SYMBOL_GPL(gfn_to_memslot);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800986
Izik Eiduse0d62c72007-10-24 23:57:46 +0200987int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
988{
989 int i;
Lai Jiangshan90d83dc2010-04-19 17:41:23 +0800990 struct kvm_memslots *slots = kvm_memslots(kvm);
Izik Eiduse0d62c72007-10-24 23:57:46 +0200991
Izik Eiduse0d62c72007-10-24 23:57:46 +0200992 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200993 struct kvm_memory_slot *memslot = &slots->memslots[i];
Izik Eiduse0d62c72007-10-24 23:57:46 +0200994
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200995 if (memslot->flags & KVM_MEMSLOT_INVALID)
996 continue;
997
Izik Eiduse0d62c72007-10-24 23:57:46 +0200998 if (gfn >= memslot->base_gfn
999 && gfn < memslot->base_gfn + memslot->npages)
1000 return 1;
1001 }
1002 return 0;
1003}
1004EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);
1005
Joerg Roedel8f0b1ab2010-01-28 12:37:56 +01001006unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn)
1007{
1008 struct vm_area_struct *vma;
1009 unsigned long addr, size;
1010
1011 size = PAGE_SIZE;
1012
1013 addr = gfn_to_hva(kvm, gfn);
1014 if (kvm_is_error_hva(addr))
1015 return PAGE_SIZE;
1016
1017 down_read(&current->mm->mmap_sem);
1018 vma = find_vma(current->mm, addr);
1019 if (!vma)
1020 goto out;
1021
1022 size = vma_kernel_pagesize(vma);
1023
1024out:
1025 up_read(&current->mm->mmap_sem);
1026
1027 return size;
1028}
1029
Gleb Natapov49c77542010-10-18 15:22:23 +02001030static unsigned long gfn_to_hva_many(struct kvm_memory_slot *slot, gfn_t gfn,
Xiao Guangrong48987782010-08-22 19:11:43 +08001031 gfn_t *nr_pages)
Izik Eidus539cb662007-11-11 22:05:04 +02001032{
Marcelo Tosattibc6678a2009-12-23 14:35:21 -02001033 if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
Izik Eidus539cb662007-11-11 22:05:04 +02001034 return bad_hva();
Xiao Guangrong48987782010-08-22 19:11:43 +08001035
1036 if (nr_pages)
1037 *nr_pages = slot->npages - (gfn - slot->base_gfn);
1038
Takuya Yoshikawaf5c98032010-02-25 11:33:19 +09001039 return gfn_to_hva_memslot(slot, gfn);
Izik Eidus539cb662007-11-11 22:05:04 +02001040}
Xiao Guangrong48987782010-08-22 19:11:43 +08001041
1042unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
1043{
Gleb Natapov49c77542010-10-18 15:22:23 +02001044 return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL);
Xiao Guangrong48987782010-08-22 19:11:43 +08001045}
Sheng Yang0d150292008-04-25 21:44:50 +08001046EXPORT_SYMBOL_GPL(gfn_to_hva);
Izik Eidus539cb662007-11-11 22:05:04 +02001047
Gleb Natapov80300892010-10-19 18:13:41 +02001048static pfn_t get_fault_pfn(void)
1049{
1050 get_page(fault_page);
1051 return fault_pfn;
1052}
1053
Gleb Natapov0857b9e2011-02-01 13:21:47 +02001054int get_user_page_nowait(struct task_struct *tsk, struct mm_struct *mm,
1055 unsigned long start, int write, struct page **page)
1056{
1057 int flags = FOLL_TOUCH | FOLL_NOWAIT | FOLL_HWPOISON | FOLL_GET;
1058
1059 if (write)
1060 flags |= FOLL_WRITE;
1061
1062 return __get_user_pages(tsk, mm, start, 1, flags, page, NULL, NULL);
1063}
1064
Huang Yingfafc3db2011-01-30 11:15:49 +08001065static inline int check_user_page_hwpoison(unsigned long addr)
1066{
1067 int rc, flags = FOLL_TOUCH | FOLL_HWPOISON | FOLL_WRITE;
1068
1069 rc = __get_user_pages(current, current->mm, addr, 1,
1070 flags, NULL, NULL, NULL);
1071 return rc == -EHWPOISON;
1072}
1073
Gleb Natapovaf585b92010-10-14 11:22:46 +02001074static pfn_t hva_to_pfn(struct kvm *kvm, unsigned long addr, bool atomic,
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001075 bool *async, bool write_fault, bool *writable)
Avi Kivity954bbbc22007-03-30 14:02:32 +03001076{
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001077 struct page *page[1];
Gleb Natapovaf585b92010-10-14 11:22:46 +02001078 int npages = 0;
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001079 pfn_t pfn;
Avi Kivity954bbbc22007-03-30 14:02:32 +03001080
Gleb Natapovaf585b92010-10-14 11:22:46 +02001081 /* we can do it either atomically or asynchronously, not both */
1082 BUG_ON(atomic && async);
1083
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001084 BUG_ON(!write_fault && !writable);
1085
1086 if (writable)
1087 *writable = true;
1088
Gleb Natapovaf585b92010-10-14 11:22:46 +02001089 if (atomic || async)
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001090 npages = __get_user_pages_fast(addr, 1, 1, page);
Gleb Natapovaf585b92010-10-14 11:22:46 +02001091
1092 if (unlikely(npages != 1) && !atomic) {
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001093 might_sleep();
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001094
1095 if (writable)
1096 *writable = write_fault;
1097
Gleb Natapov0857b9e2011-02-01 13:21:47 +02001098 if (async) {
1099 down_read(&current->mm->mmap_sem);
1100 npages = get_user_page_nowait(current, current->mm,
1101 addr, write_fault, page);
1102 up_read(&current->mm->mmap_sem);
1103 } else
1104 npages = get_user_pages_fast(addr, 1, write_fault,
1105 page);
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001106
1107 /* map read fault as writable if possible */
1108 if (unlikely(!write_fault) && npages == 1) {
1109 struct page *wpage[1];
1110
1111 npages = __get_user_pages_fast(addr, 1, 1, wpage);
1112 if (npages == 1) {
1113 *writable = true;
1114 put_page(page[0]);
1115 page[0] = wpage[0];
1116 }
1117 npages = 1;
1118 }
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001119 }
Izik Eidus539cb662007-11-11 22:05:04 +02001120
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001121 if (unlikely(npages != 1)) {
1122 struct vm_area_struct *vma;
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001123
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001124 if (atomic)
Gleb Natapov80300892010-10-19 18:13:41 +02001125 return get_fault_pfn();
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001126
Huang Yingbbeb3402010-06-22 14:23:11 +08001127 down_read(&current->mm->mmap_sem);
Gleb Natapov0857b9e2011-02-01 13:21:47 +02001128 if (npages == -EHWPOISON ||
1129 (!async && check_user_page_hwpoison(addr))) {
Huang Yingbbeb3402010-06-22 14:23:11 +08001130 up_read(&current->mm->mmap_sem);
Huang Yingbf998152010-05-31 14:28:19 +08001131 get_page(hwpoison_page);
1132 return page_to_pfn(hwpoison_page);
1133 }
1134
Gleb Natapov80300892010-10-19 18:13:41 +02001135 vma = find_vma_intersection(current->mm, addr, addr+1);
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001136
Gleb Natapov80300892010-10-19 18:13:41 +02001137 if (vma == NULL)
1138 pfn = get_fault_pfn();
1139 else if ((vma->vm_flags & VM_PFNMAP)) {
1140 pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) +
1141 vma->vm_pgoff;
1142 BUG_ON(!kvm_is_mmio_pfn(pfn));
1143 } else {
1144 if (async && (vma->vm_flags & VM_WRITE))
Gleb Natapovaf585b92010-10-14 11:22:46 +02001145 *async = true;
Gleb Natapov80300892010-10-19 18:13:41 +02001146 pfn = get_fault_pfn();
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001147 }
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001148 up_read(&current->mm->mmap_sem);
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001149 } else
1150 pfn = page_to_pfn(page[0]);
1151
1152 return pfn;
Anthony Liguori35149e22008-04-02 14:46:56 -05001153}
1154
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001155pfn_t hva_to_pfn_atomic(struct kvm *kvm, unsigned long addr)
1156{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001157 return hva_to_pfn(kvm, addr, true, NULL, true, NULL);
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001158}
1159EXPORT_SYMBOL_GPL(hva_to_pfn_atomic);
1160
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001161static pfn_t __gfn_to_pfn(struct kvm *kvm, gfn_t gfn, bool atomic, bool *async,
1162 bool write_fault, bool *writable)
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001163{
1164 unsigned long addr;
1165
Gleb Natapovaf585b92010-10-14 11:22:46 +02001166 if (async)
1167 *async = false;
1168
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001169 addr = gfn_to_hva(kvm, gfn);
1170 if (kvm_is_error_hva(addr)) {
1171 get_page(bad_page);
1172 return page_to_pfn(bad_page);
1173 }
1174
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001175 return hva_to_pfn(kvm, addr, atomic, async, write_fault, writable);
Xiao Guangrong365fb3f2010-08-28 19:24:13 +08001176}
1177
1178pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn)
1179{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001180 return __gfn_to_pfn(kvm, gfn, true, NULL, true, NULL);
Xiao Guangrong365fb3f2010-08-28 19:24:13 +08001181}
1182EXPORT_SYMBOL_GPL(gfn_to_pfn_atomic);
1183
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001184pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
1185 bool write_fault, bool *writable)
Gleb Natapovaf585b92010-10-14 11:22:46 +02001186{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001187 return __gfn_to_pfn(kvm, gfn, false, async, write_fault, writable);
Gleb Natapovaf585b92010-10-14 11:22:46 +02001188}
1189EXPORT_SYMBOL_GPL(gfn_to_pfn_async);
1190
Xiao Guangrong365fb3f2010-08-28 19:24:13 +08001191pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
1192{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001193 return __gfn_to_pfn(kvm, gfn, false, NULL, true, NULL);
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001194}
Anthony Liguori35149e22008-04-02 14:46:56 -05001195EXPORT_SYMBOL_GPL(gfn_to_pfn);
1196
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001197pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1198 bool *writable)
1199{
1200 return __gfn_to_pfn(kvm, gfn, false, NULL, write_fault, writable);
1201}
1202EXPORT_SYMBOL_GPL(gfn_to_pfn_prot);
1203
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001204pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
1205 struct kvm_memory_slot *slot, gfn_t gfn)
1206{
1207 unsigned long addr = gfn_to_hva_memslot(slot, gfn);
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001208 return hva_to_pfn(kvm, addr, false, NULL, true, NULL);
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001209}
1210
Xiao Guangrong48987782010-08-22 19:11:43 +08001211int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
1212 int nr_pages)
1213{
1214 unsigned long addr;
1215 gfn_t entry;
1216
Gleb Natapov49c77542010-10-18 15:22:23 +02001217 addr = gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, &entry);
Xiao Guangrong48987782010-08-22 19:11:43 +08001218 if (kvm_is_error_hva(addr))
1219 return -1;
1220
1221 if (entry < nr_pages)
1222 return 0;
1223
1224 return __get_user_pages_fast(addr, nr_pages, 1, pages);
1225}
1226EXPORT_SYMBOL_GPL(gfn_to_page_many_atomic);
1227
Anthony Liguori35149e22008-04-02 14:46:56 -05001228struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
1229{
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001230 pfn_t pfn;
1231
1232 pfn = gfn_to_pfn(kvm, gfn);
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001233 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001234 return pfn_to_page(pfn);
1235
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001236 WARN_ON(kvm_is_mmio_pfn(pfn));
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001237
1238 get_page(bad_page);
1239 return bad_page;
Avi Kivity954bbbc22007-03-30 14:02:32 +03001240}
Anthony Liguoriaab61cc2007-10-29 15:15:20 -05001241
Avi Kivity954bbbc22007-03-30 14:02:32 +03001242EXPORT_SYMBOL_GPL(gfn_to_page);
1243
Izik Eidusb4231d62007-11-20 11:49:33 +02001244void kvm_release_page_clean(struct page *page)
1245{
Anthony Liguori35149e22008-04-02 14:46:56 -05001246 kvm_release_pfn_clean(page_to_pfn(page));
Izik Eidusb4231d62007-11-20 11:49:33 +02001247}
1248EXPORT_SYMBOL_GPL(kvm_release_page_clean);
1249
Anthony Liguori35149e22008-04-02 14:46:56 -05001250void kvm_release_pfn_clean(pfn_t pfn)
1251{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001252 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001253 put_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001254}
1255EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);
1256
Izik Eidusb4231d62007-11-20 11:49:33 +02001257void kvm_release_page_dirty(struct page *page)
Izik Eidus8a7ae052007-10-18 11:09:33 +02001258{
Anthony Liguori35149e22008-04-02 14:46:56 -05001259 kvm_release_pfn_dirty(page_to_pfn(page));
Izik Eidus8a7ae052007-10-18 11:09:33 +02001260}
Izik Eidusb4231d62007-11-20 11:49:33 +02001261EXPORT_SYMBOL_GPL(kvm_release_page_dirty);
Izik Eidus8a7ae052007-10-18 11:09:33 +02001262
Anthony Liguori35149e22008-04-02 14:46:56 -05001263void kvm_release_pfn_dirty(pfn_t pfn)
1264{
1265 kvm_set_pfn_dirty(pfn);
1266 kvm_release_pfn_clean(pfn);
1267}
1268EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
1269
1270void kvm_set_page_dirty(struct page *page)
1271{
1272 kvm_set_pfn_dirty(page_to_pfn(page));
1273}
1274EXPORT_SYMBOL_GPL(kvm_set_page_dirty);
1275
1276void kvm_set_pfn_dirty(pfn_t pfn)
1277{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001278 if (!kvm_is_mmio_pfn(pfn)) {
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001279 struct page *page = pfn_to_page(pfn);
1280 if (!PageReserved(page))
1281 SetPageDirty(page);
1282 }
Anthony Liguori35149e22008-04-02 14:46:56 -05001283}
1284EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);
1285
1286void kvm_set_pfn_accessed(pfn_t pfn)
1287{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001288 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001289 mark_page_accessed(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001290}
1291EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);
1292
1293void kvm_get_pfn(pfn_t pfn)
1294{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001295 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001296 get_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001297}
1298EXPORT_SYMBOL_GPL(kvm_get_pfn);
1299
Izik Eidus195aefd2007-10-01 22:14:18 +02001300static int next_segment(unsigned long len, int offset)
1301{
1302 if (len > PAGE_SIZE - offset)
1303 return PAGE_SIZE - offset;
1304 else
1305 return len;
1306}
1307
1308int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
1309 int len)
1310{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001311 int r;
1312 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001313
Izik Eiduse0506bc2007-11-11 22:10:22 +02001314 addr = gfn_to_hva(kvm, gfn);
1315 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001316 return -EFAULT;
Takuya Yoshikawafa3d3152011-05-07 16:35:38 +09001317 r = __copy_from_user(data, (void __user *)addr + offset, len);
Izik Eiduse0506bc2007-11-11 22:10:22 +02001318 if (r)
1319 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001320 return 0;
1321}
1322EXPORT_SYMBOL_GPL(kvm_read_guest_page);
1323
1324int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len)
1325{
1326 gfn_t gfn = gpa >> PAGE_SHIFT;
1327 int seg;
1328 int offset = offset_in_page(gpa);
1329 int ret;
1330
1331 while ((seg = next_segment(len, offset)) != 0) {
1332 ret = kvm_read_guest_page(kvm, gfn, data, offset, seg);
1333 if (ret < 0)
1334 return ret;
1335 offset = 0;
1336 len -= seg;
1337 data += seg;
1338 ++gfn;
1339 }
1340 return 0;
1341}
1342EXPORT_SYMBOL_GPL(kvm_read_guest);
1343
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001344int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
1345 unsigned long len)
1346{
1347 int r;
1348 unsigned long addr;
1349 gfn_t gfn = gpa >> PAGE_SHIFT;
1350 int offset = offset_in_page(gpa);
1351
1352 addr = gfn_to_hva(kvm, gfn);
1353 if (kvm_is_error_hva(addr))
1354 return -EFAULT;
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001355 pagefault_disable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001356 r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001357 pagefault_enable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001358 if (r)
1359 return -EFAULT;
1360 return 0;
1361}
1362EXPORT_SYMBOL(kvm_read_guest_atomic);
1363
Izik Eidus195aefd2007-10-01 22:14:18 +02001364int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
1365 int offset, int len)
1366{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001367 int r;
1368 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001369
Izik Eiduse0506bc2007-11-11 22:10:22 +02001370 addr = gfn_to_hva(kvm, gfn);
1371 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001372 return -EFAULT;
Xiao Guangrong8b0cedf2011-05-15 23:22:04 +08001373 r = __copy_to_user((void __user *)addr + offset, data, len);
Izik Eiduse0506bc2007-11-11 22:10:22 +02001374 if (r)
1375 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001376 mark_page_dirty(kvm, gfn);
1377 return 0;
1378}
1379EXPORT_SYMBOL_GPL(kvm_write_guest_page);
1380
1381int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
1382 unsigned long len)
1383{
1384 gfn_t gfn = gpa >> PAGE_SHIFT;
1385 int seg;
1386 int offset = offset_in_page(gpa);
1387 int ret;
1388
1389 while ((seg = next_segment(len, offset)) != 0) {
1390 ret = kvm_write_guest_page(kvm, gfn, data, offset, seg);
1391 if (ret < 0)
1392 return ret;
1393 offset = 0;
1394 len -= seg;
1395 data += seg;
1396 ++gfn;
1397 }
1398 return 0;
1399}
1400
Gleb Natapov49c77542010-10-18 15:22:23 +02001401int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1402 gpa_t gpa)
1403{
1404 struct kvm_memslots *slots = kvm_memslots(kvm);
1405 int offset = offset_in_page(gpa);
1406 gfn_t gfn = gpa >> PAGE_SHIFT;
1407
1408 ghc->gpa = gpa;
1409 ghc->generation = slots->generation;
1410 ghc->memslot = __gfn_to_memslot(slots, gfn);
1411 ghc->hva = gfn_to_hva_many(ghc->memslot, gfn, NULL);
1412 if (!kvm_is_error_hva(ghc->hva))
1413 ghc->hva += offset;
1414 else
1415 return -EFAULT;
1416
1417 return 0;
1418}
1419EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init);
1420
1421int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1422 void *data, unsigned long len)
1423{
1424 struct kvm_memslots *slots = kvm_memslots(kvm);
1425 int r;
1426
1427 if (slots->generation != ghc->generation)
1428 kvm_gfn_to_hva_cache_init(kvm, ghc, ghc->gpa);
1429
1430 if (kvm_is_error_hva(ghc->hva))
1431 return -EFAULT;
1432
Xiao Guangrong8b0cedf2011-05-15 23:22:04 +08001433 r = __copy_to_user((void __user *)ghc->hva, data, len);
Gleb Natapov49c77542010-10-18 15:22:23 +02001434 if (r)
1435 return -EFAULT;
1436 mark_page_dirty_in_slot(kvm, ghc->memslot, ghc->gpa >> PAGE_SHIFT);
1437
1438 return 0;
1439}
1440EXPORT_SYMBOL_GPL(kvm_write_guest_cached);
1441
Gleb Natapove03b6442011-07-11 15:28:11 -04001442int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1443 void *data, unsigned long len)
1444{
1445 struct kvm_memslots *slots = kvm_memslots(kvm);
1446 int r;
1447
1448 if (slots->generation != ghc->generation)
1449 kvm_gfn_to_hva_cache_init(kvm, ghc, ghc->gpa);
1450
1451 if (kvm_is_error_hva(ghc->hva))
1452 return -EFAULT;
1453
1454 r = __copy_from_user(data, (void __user *)ghc->hva, len);
1455 if (r)
1456 return -EFAULT;
1457
1458 return 0;
1459}
1460EXPORT_SYMBOL_GPL(kvm_read_guest_cached);
1461
Izik Eidus195aefd2007-10-01 22:14:18 +02001462int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
1463{
Heiko Carstens3bcc8a82010-10-27 17:21:21 +02001464 return kvm_write_guest_page(kvm, gfn, (const void *) empty_zero_page,
1465 offset, len);
Izik Eidus195aefd2007-10-01 22:14:18 +02001466}
1467EXPORT_SYMBOL_GPL(kvm_clear_guest_page);
1468
1469int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len)
1470{
1471 gfn_t gfn = gpa >> PAGE_SHIFT;
1472 int seg;
1473 int offset = offset_in_page(gpa);
1474 int ret;
1475
1476 while ((seg = next_segment(len, offset)) != 0) {
1477 ret = kvm_clear_guest_page(kvm, gfn, offset, seg);
1478 if (ret < 0)
1479 return ret;
1480 offset = 0;
1481 len -= seg;
1482 ++gfn;
1483 }
1484 return 0;
1485}
1486EXPORT_SYMBOL_GPL(kvm_clear_guest);
1487
Gleb Natapov49c77542010-10-18 15:22:23 +02001488void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
1489 gfn_t gfn)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001490{
Rusty Russell7e9d6192007-07-31 20:41:14 +10001491 if (memslot && memslot->dirty_bitmap) {
1492 unsigned long rel_gfn = gfn - memslot->base_gfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001493
Akinobu Mitacd7e48c2011-03-23 16:42:04 -07001494 __set_bit_le(rel_gfn, memslot->dirty_bitmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001495 }
1496}
1497
Gleb Natapov49c77542010-10-18 15:22:23 +02001498void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
1499{
1500 struct kvm_memory_slot *memslot;
1501
1502 memslot = gfn_to_memslot(kvm, gfn);
1503 mark_page_dirty_in_slot(kvm, memslot, gfn);
1504}
1505
Eddie Dongb6958ce2007-07-18 12:15:21 +03001506/*
1507 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
1508 */
Hollis Blanchard8776e512007-10-31 17:24:24 -05001509void kvm_vcpu_block(struct kvm_vcpu *vcpu)
Eddie Dongb6958ce2007-07-18 12:15:21 +03001510{
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001511 DEFINE_WAIT(wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001512
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001513 for (;;) {
1514 prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001515
Gleb Natapova1b37102009-07-09 15:33:52 +03001516 if (kvm_arch_vcpu_runnable(vcpu)) {
Avi Kivitya8eeb042010-05-10 12:34:53 +03001517 kvm_make_request(KVM_REQ_UNHALT, vcpu);
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001518 break;
Marcelo Tosattid7690172008-09-08 15:23:48 -03001519 }
Gleb Natapov09cec752009-03-23 15:11:44 +02001520 if (kvm_cpu_has_pending_timer(vcpu))
1521 break;
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001522 if (signal_pending(current))
1523 break;
1524
Eddie Dongb6958ce2007-07-18 12:15:21 +03001525 schedule();
Eddie Dongb6958ce2007-07-18 12:15:21 +03001526 }
1527
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001528 finish_wait(&vcpu->wq, &wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001529}
1530
Avi Kivity6aa8b732006-12-10 02:21:36 -08001531void kvm_resched(struct kvm_vcpu *vcpu)
1532{
Yaozu Dong3fca0362007-04-25 16:49:19 +03001533 if (!need_resched())
1534 return;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001535 cond_resched();
Avi Kivity6aa8b732006-12-10 02:21:36 -08001536}
1537EXPORT_SYMBOL_GPL(kvm_resched);
1538
Rik van Riel217ece62011-02-01 09:53:28 -05001539void kvm_vcpu_on_spin(struct kvm_vcpu *me)
Zhai, Edwind255f4f2009-10-09 18:03:20 +08001540{
Rik van Riel217ece62011-02-01 09:53:28 -05001541 struct kvm *kvm = me->kvm;
1542 struct kvm_vcpu *vcpu;
1543 int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
1544 int yielded = 0;
1545 int pass;
1546 int i;
Zhai, Edwind255f4f2009-10-09 18:03:20 +08001547
Rik van Riel217ece62011-02-01 09:53:28 -05001548 /*
1549 * We boost the priority of a VCPU that is runnable but not
1550 * currently running, because it got preempted by something
1551 * else and called schedule in __vcpu_run. Hopefully that
1552 * VCPU is holding the lock that we need and will release it.
1553 * We approximate round-robin by starting at the last boosted VCPU.
1554 */
1555 for (pass = 0; pass < 2 && !yielded; pass++) {
1556 kvm_for_each_vcpu(i, vcpu, kvm) {
1557 struct task_struct *task = NULL;
1558 struct pid *pid;
1559 if (!pass && i < last_boosted_vcpu) {
1560 i = last_boosted_vcpu;
1561 continue;
1562 } else if (pass && i > last_boosted_vcpu)
1563 break;
1564 if (vcpu == me)
1565 continue;
1566 if (waitqueue_active(&vcpu->wq))
1567 continue;
1568 rcu_read_lock();
1569 pid = rcu_dereference(vcpu->pid);
1570 if (pid)
1571 task = get_pid_task(vcpu->pid, PIDTYPE_PID);
1572 rcu_read_unlock();
1573 if (!task)
1574 continue;
1575 if (task->flags & PF_VCPU) {
1576 put_task_struct(task);
1577 continue;
1578 }
1579 if (yield_to(task, 1)) {
1580 put_task_struct(task);
1581 kvm->last_boosted_vcpu = i;
1582 yielded = 1;
1583 break;
1584 }
1585 put_task_struct(task);
1586 }
1587 }
Zhai, Edwind255f4f2009-10-09 18:03:20 +08001588}
1589EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin);
1590
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001591static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001592{
1593 struct kvm_vcpu *vcpu = vma->vm_file->private_data;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001594 struct page *page;
1595
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001596 if (vmf->pgoff == 0)
Avi Kivity039576c2007-03-20 12:46:50 +02001597 page = virt_to_page(vcpu->run);
Avi Kivity09566762008-01-23 18:14:23 +02001598#ifdef CONFIG_X86
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001599 else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
Zhang Xiantaoad312c72007-12-13 23:50:52 +08001600 page = virt_to_page(vcpu->arch.pio_data);
Avi Kivity09566762008-01-23 18:14:23 +02001601#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02001602#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1603 else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
1604 page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
1605#endif
Avi Kivity039576c2007-03-20 12:46:50 +02001606 else
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001607 return VM_FAULT_SIGBUS;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001608 get_page(page);
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001609 vmf->page = page;
1610 return 0;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001611}
1612
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +04001613static const struct vm_operations_struct kvm_vcpu_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001614 .fault = kvm_vcpu_fault,
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001615};
1616
1617static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma)
1618{
1619 vma->vm_ops = &kvm_vcpu_vm_ops;
1620 return 0;
1621}
1622
Avi Kivitybccf2152007-02-21 18:04:26 +02001623static int kvm_vcpu_release(struct inode *inode, struct file *filp)
1624{
1625 struct kvm_vcpu *vcpu = filp->private_data;
1626
Al Viro66c0b392008-04-19 20:33:56 +01001627 kvm_put_kvm(vcpu->kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001628 return 0;
1629}
1630
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01001631static struct file_operations kvm_vcpu_fops = {
Avi Kivitybccf2152007-02-21 18:04:26 +02001632 .release = kvm_vcpu_release,
1633 .unlocked_ioctl = kvm_vcpu_ioctl,
Alexander Graf1dda6062011-06-08 02:45:37 +02001634#ifdef CONFIG_COMPAT
1635 .compat_ioctl = kvm_vcpu_compat_ioctl,
1636#endif
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001637 .mmap = kvm_vcpu_mmap,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001638 .llseek = noop_llseek,
Avi Kivitybccf2152007-02-21 18:04:26 +02001639};
1640
1641/*
1642 * Allocates an inode for the vcpu.
1643 */
1644static int create_vcpu_fd(struct kvm_vcpu *vcpu)
1645{
Roland Dreier628ff7c2009-12-18 09:41:24 -08001646 return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, O_RDWR);
Avi Kivitybccf2152007-02-21 18:04:26 +02001647}
1648
Avi Kivityc5ea7662007-02-20 18:41:05 +02001649/*
1650 * Creates some virtual cpus. Good luck creating more than one.
1651 */
Gleb Natapov73880c82009-06-09 15:56:28 +03001652static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
Avi Kivityc5ea7662007-02-20 18:41:05 +02001653{
1654 int r;
Gleb Natapov988a2ca2009-06-09 15:56:29 +03001655 struct kvm_vcpu *vcpu, *v;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001656
Gleb Natapov73880c82009-06-09 15:56:28 +03001657 vcpu = kvm_arch_vcpu_create(kvm, id);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001658 if (IS_ERR(vcpu))
1659 return PTR_ERR(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001660
Avi Kivity15ad7142007-07-11 18:17:21 +03001661 preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);
1662
Avi Kivity26e52152007-11-20 15:30:24 +02001663 r = kvm_arch_vcpu_setup(vcpu);
1664 if (r)
Jan Kiszkad7805922011-05-23 10:33:05 +02001665 goto vcpu_destroy;
Avi Kivity26e52152007-11-20 15:30:24 +02001666
Shaohua Li11ec2802007-07-23 14:51:37 +08001667 mutex_lock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03001668 if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) {
1669 r = -EINVAL;
Jan Kiszkad7805922011-05-23 10:33:05 +02001670 goto unlock_vcpu_destroy;
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001671 }
Gleb Natapov73880c82009-06-09 15:56:28 +03001672
Gleb Natapov988a2ca2009-06-09 15:56:29 +03001673 kvm_for_each_vcpu(r, v, kvm)
1674 if (v->vcpu_id == id) {
Gleb Natapov73880c82009-06-09 15:56:28 +03001675 r = -EEXIST;
Jan Kiszkad7805922011-05-23 10:33:05 +02001676 goto unlock_vcpu_destroy;
Gleb Natapov73880c82009-06-09 15:56:28 +03001677 }
1678
1679 BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001680
1681 /* Now it's all set up, let userspace reach it */
Al Viro66c0b392008-04-19 20:33:56 +01001682 kvm_get_kvm(kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001683 r = create_vcpu_fd(vcpu);
Gleb Natapov73880c82009-06-09 15:56:28 +03001684 if (r < 0) {
1685 kvm_put_kvm(kvm);
Jan Kiszkad7805922011-05-23 10:33:05 +02001686 goto unlock_vcpu_destroy;
Gleb Natapov73880c82009-06-09 15:56:28 +03001687 }
1688
1689 kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu;
1690 smp_wmb();
1691 atomic_inc(&kvm->online_vcpus);
1692
1693#ifdef CONFIG_KVM_APIC_ARCHITECTURE
1694 if (kvm->bsp_vcpu_id == id)
1695 kvm->bsp_vcpu = vcpu;
1696#endif
1697 mutex_unlock(&kvm->lock);
Avi Kivitybccf2152007-02-21 18:04:26 +02001698 return r;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001699
Jan Kiszkad7805922011-05-23 10:33:05 +02001700unlock_vcpu_destroy:
Glauber Costa7d8fece2008-09-17 23:16:59 -03001701 mutex_unlock(&kvm->lock);
Jan Kiszkad7805922011-05-23 10:33:05 +02001702vcpu_destroy:
Hollis Blanchardd40ccc62007-11-19 14:04:43 -06001703 kvm_arch_vcpu_destroy(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001704 return r;
1705}
1706
Avi Kivity1961d272007-03-05 19:46:05 +02001707static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset)
1708{
1709 if (sigset) {
1710 sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP));
1711 vcpu->sigset_active = 1;
1712 vcpu->sigset = *sigset;
1713 } else
1714 vcpu->sigset_active = 0;
1715 return 0;
1716}
1717
Avi Kivitybccf2152007-02-21 18:04:26 +02001718static long kvm_vcpu_ioctl(struct file *filp,
1719 unsigned int ioctl, unsigned long arg)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001720{
Avi Kivitybccf2152007-02-21 18:04:26 +02001721 struct kvm_vcpu *vcpu = filp->private_data;
Al Viro2f366982007-02-09 16:38:35 +00001722 void __user *argp = (void __user *)arg;
Carsten Otte313a3dc2007-10-11 19:16:52 +02001723 int r;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001724 struct kvm_fpu *fpu = NULL;
1725 struct kvm_sregs *kvm_sregs = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001726
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001727 if (vcpu->kvm->mm != current->mm)
1728 return -EIO;
Avi Kivity2122ff52010-05-13 11:25:04 +03001729
1730#if defined(CONFIG_S390) || defined(CONFIG_PPC)
1731 /*
1732 * Special cases: vcpu ioctls that are asynchronous to vcpu execution,
1733 * so vcpu_load() would break it.
1734 */
1735 if (ioctl == KVM_S390_INTERRUPT || ioctl == KVM_INTERRUPT)
1736 return kvm_arch_vcpu_ioctl(filp, ioctl, arg);
1737#endif
1738
1739
1740 vcpu_load(vcpu);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001741 switch (ioctl) {
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001742 case KVM_RUN:
Avi Kivityf0fe5102007-03-07 13:11:17 +02001743 r = -EINVAL;
1744 if (arg)
1745 goto out;
Hollis Blanchardb6c7a5d2007-11-01 14:16:10 -05001746 r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
Gleb Natapov64be5002010-10-24 16:49:08 +02001747 trace_kvm_userspace_exit(vcpu->run->exit_reason, r);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001748 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001749 case KVM_GET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001750 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001751
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001752 r = -ENOMEM;
1753 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
1754 if (!kvm_regs)
1755 goto out;
1756 r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001757 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001758 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001759 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001760 if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
1761 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001762 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001763out_free1:
1764 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001765 break;
1766 }
1767 case KVM_SET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001768 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001769
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001770 r = -ENOMEM;
1771 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
1772 if (!kvm_regs)
1773 goto out;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001774 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001775 if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs)))
1776 goto out_free2;
1777 r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001778 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001779 goto out_free2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001780 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001781out_free2:
1782 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001783 break;
1784 }
1785 case KVM_GET_SREGS: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001786 kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL);
1787 r = -ENOMEM;
1788 if (!kvm_sregs)
1789 goto out;
1790 r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001791 if (r)
1792 goto out;
1793 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001794 if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001795 goto out;
1796 r = 0;
1797 break;
1798 }
1799 case KVM_SET_SREGS: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001800 kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL);
1801 r = -ENOMEM;
1802 if (!kvm_sregs)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001803 goto out;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001804 r = -EFAULT;
1805 if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs)))
1806 goto out;
1807 r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001808 if (r)
1809 goto out;
1810 r = 0;
1811 break;
1812 }
Marcelo Tosatti62d9f0d2008-04-11 13:24:45 -03001813 case KVM_GET_MP_STATE: {
1814 struct kvm_mp_state mp_state;
1815
1816 r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state);
1817 if (r)
1818 goto out;
1819 r = -EFAULT;
1820 if (copy_to_user(argp, &mp_state, sizeof mp_state))
1821 goto out;
1822 r = 0;
1823 break;
1824 }
1825 case KVM_SET_MP_STATE: {
1826 struct kvm_mp_state mp_state;
1827
1828 r = -EFAULT;
1829 if (copy_from_user(&mp_state, argp, sizeof mp_state))
1830 goto out;
1831 r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
1832 if (r)
1833 goto out;
1834 r = 0;
1835 break;
1836 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08001837 case KVM_TRANSLATE: {
1838 struct kvm_translation tr;
1839
1840 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001841 if (copy_from_user(&tr, argp, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001842 goto out;
Zhang Xiantao8b006792007-11-16 13:05:55 +08001843 r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001844 if (r)
1845 goto out;
1846 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001847 if (copy_to_user(argp, &tr, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001848 goto out;
1849 r = 0;
1850 break;
1851 }
Jan Kiszkad0bfb942008-12-15 13:52:10 +01001852 case KVM_SET_GUEST_DEBUG: {
1853 struct kvm_guest_debug dbg;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001854
1855 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001856 if (copy_from_user(&dbg, argp, sizeof dbg))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001857 goto out;
Jan Kiszkad0bfb942008-12-15 13:52:10 +01001858 r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001859 if (r)
1860 goto out;
1861 r = 0;
1862 break;
1863 }
Avi Kivity1961d272007-03-05 19:46:05 +02001864 case KVM_SET_SIGNAL_MASK: {
1865 struct kvm_signal_mask __user *sigmask_arg = argp;
1866 struct kvm_signal_mask kvm_sigmask;
1867 sigset_t sigset, *p;
1868
1869 p = NULL;
1870 if (argp) {
1871 r = -EFAULT;
1872 if (copy_from_user(&kvm_sigmask, argp,
1873 sizeof kvm_sigmask))
1874 goto out;
1875 r = -EINVAL;
1876 if (kvm_sigmask.len != sizeof sigset)
1877 goto out;
1878 r = -EFAULT;
1879 if (copy_from_user(&sigset, sigmask_arg->sigset,
1880 sizeof sigset))
1881 goto out;
1882 p = &sigset;
1883 }
Andi Kleen376d41f2010-06-10 13:10:47 +02001884 r = kvm_vcpu_ioctl_set_sigmask(vcpu, p);
Avi Kivity1961d272007-03-05 19:46:05 +02001885 break;
1886 }
Avi Kivityb8836732007-04-01 16:34:31 +03001887 case KVM_GET_FPU: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001888 fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL);
1889 r = -ENOMEM;
1890 if (!fpu)
1891 goto out;
1892 r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001893 if (r)
1894 goto out;
1895 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001896 if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
Avi Kivityb8836732007-04-01 16:34:31 +03001897 goto out;
1898 r = 0;
1899 break;
1900 }
1901 case KVM_SET_FPU: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001902 fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL);
1903 r = -ENOMEM;
1904 if (!fpu)
Avi Kivityb8836732007-04-01 16:34:31 +03001905 goto out;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001906 r = -EFAULT;
1907 if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu)))
1908 goto out;
1909 r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001910 if (r)
1911 goto out;
1912 r = 0;
1913 break;
1914 }
Avi Kivitybccf2152007-02-21 18:04:26 +02001915 default:
Carsten Otte313a3dc2007-10-11 19:16:52 +02001916 r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
Avi Kivitybccf2152007-02-21 18:04:26 +02001917 }
1918out:
Avi Kivity2122ff52010-05-13 11:25:04 +03001919 vcpu_put(vcpu);
Dave Hansenfa3795a2008-08-11 10:01:46 -07001920 kfree(fpu);
1921 kfree(kvm_sregs);
Avi Kivitybccf2152007-02-21 18:04:26 +02001922 return r;
1923}
1924
Alexander Graf1dda6062011-06-08 02:45:37 +02001925#ifdef CONFIG_COMPAT
1926static long kvm_vcpu_compat_ioctl(struct file *filp,
1927 unsigned int ioctl, unsigned long arg)
1928{
1929 struct kvm_vcpu *vcpu = filp->private_data;
1930 void __user *argp = compat_ptr(arg);
1931 int r;
1932
1933 if (vcpu->kvm->mm != current->mm)
1934 return -EIO;
1935
1936 switch (ioctl) {
1937 case KVM_SET_SIGNAL_MASK: {
1938 struct kvm_signal_mask __user *sigmask_arg = argp;
1939 struct kvm_signal_mask kvm_sigmask;
1940 compat_sigset_t csigset;
1941 sigset_t sigset;
1942
1943 if (argp) {
1944 r = -EFAULT;
1945 if (copy_from_user(&kvm_sigmask, argp,
1946 sizeof kvm_sigmask))
1947 goto out;
1948 r = -EINVAL;
1949 if (kvm_sigmask.len != sizeof csigset)
1950 goto out;
1951 r = -EFAULT;
1952 if (copy_from_user(&csigset, sigmask_arg->sigset,
1953 sizeof csigset))
1954 goto out;
1955 }
1956 sigset_from_compat(&sigset, &csigset);
1957 r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
1958 break;
1959 }
1960 default:
1961 r = kvm_vcpu_ioctl(filp, ioctl, arg);
1962 }
1963
1964out:
1965 return r;
1966}
1967#endif
1968
Avi Kivitybccf2152007-02-21 18:04:26 +02001969static long kvm_vm_ioctl(struct file *filp,
1970 unsigned int ioctl, unsigned long arg)
1971{
1972 struct kvm *kvm = filp->private_data;
1973 void __user *argp = (void __user *)arg;
Carsten Otte1fe779f2007-10-29 16:08:35 +01001974 int r;
Avi Kivitybccf2152007-02-21 18:04:26 +02001975
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001976 if (kvm->mm != current->mm)
1977 return -EIO;
Avi Kivitybccf2152007-02-21 18:04:26 +02001978 switch (ioctl) {
1979 case KVM_CREATE_VCPU:
1980 r = kvm_vm_ioctl_create_vcpu(kvm, arg);
1981 if (r < 0)
1982 goto out;
1983 break;
Izik Eidus6fc138d2007-10-09 19:20:39 +02001984 case KVM_SET_USER_MEMORY_REGION: {
1985 struct kvm_userspace_memory_region kvm_userspace_mem;
1986
1987 r = -EFAULT;
1988 if (copy_from_user(&kvm_userspace_mem, argp,
1989 sizeof kvm_userspace_mem))
1990 goto out;
1991
1992 r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001993 if (r)
1994 goto out;
1995 break;
1996 }
1997 case KVM_GET_DIRTY_LOG: {
1998 struct kvm_dirty_log log;
1999
2000 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00002001 if (copy_from_user(&log, argp, sizeof log))
Avi Kivity6aa8b732006-12-10 02:21:36 -08002002 goto out;
Avi Kivity2c6f5df2007-02-20 18:27:58 +02002003 r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002004 if (r)
2005 goto out;
2006 break;
2007 }
Laurent Vivier5f94c172008-05-30 16:05:54 +02002008#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
2009 case KVM_REGISTER_COALESCED_MMIO: {
2010 struct kvm_coalesced_mmio_zone zone;
2011 r = -EFAULT;
2012 if (copy_from_user(&zone, argp, sizeof zone))
2013 goto out;
Laurent Vivier5f94c172008-05-30 16:05:54 +02002014 r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
2015 if (r)
2016 goto out;
2017 r = 0;
2018 break;
2019 }
2020 case KVM_UNREGISTER_COALESCED_MMIO: {
2021 struct kvm_coalesced_mmio_zone zone;
2022 r = -EFAULT;
2023 if (copy_from_user(&zone, argp, sizeof zone))
2024 goto out;
Laurent Vivier5f94c172008-05-30 16:05:54 +02002025 r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
2026 if (r)
2027 goto out;
2028 r = 0;
2029 break;
2030 }
2031#endif
Gregory Haskins721eecb2009-05-20 10:30:49 -04002032 case KVM_IRQFD: {
2033 struct kvm_irqfd data;
2034
2035 r = -EFAULT;
2036 if (copy_from_user(&data, argp, sizeof data))
2037 goto out;
2038 r = kvm_irqfd(kvm, data.fd, data.gsi, data.flags);
2039 break;
2040 }
Gregory Haskinsd34e6b12009-07-07 17:08:49 -04002041 case KVM_IOEVENTFD: {
2042 struct kvm_ioeventfd data;
2043
2044 r = -EFAULT;
2045 if (copy_from_user(&data, argp, sizeof data))
2046 goto out;
2047 r = kvm_ioeventfd(kvm, &data);
2048 break;
2049 }
Gleb Natapov73880c82009-06-09 15:56:28 +03002050#ifdef CONFIG_KVM_APIC_ARCHITECTURE
2051 case KVM_SET_BOOT_CPU_ID:
2052 r = 0;
Marcelo Tosatti894a9c52009-06-23 15:05:14 -03002053 mutex_lock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03002054 if (atomic_read(&kvm->online_vcpus) != 0)
2055 r = -EBUSY;
2056 else
2057 kvm->bsp_vcpu_id = arg;
Marcelo Tosatti894a9c52009-06-23 15:05:14 -03002058 mutex_unlock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03002059 break;
2060#endif
Avi Kivityf17abe92007-02-21 19:28:04 +02002061 default:
Carsten Otte1fe779f2007-10-29 16:08:35 +01002062 r = kvm_arch_vm_ioctl(filp, ioctl, arg);
Avi Kivitybfd99ff2009-08-26 14:57:50 +03002063 if (r == -ENOTTY)
2064 r = kvm_vm_ioctl_assigned_device(kvm, ioctl, arg);
Avi Kivityf17abe92007-02-21 19:28:04 +02002065 }
2066out:
2067 return r;
2068}
2069
Arnd Bergmann6ff58942009-10-22 14:19:27 +02002070#ifdef CONFIG_COMPAT
2071struct compat_kvm_dirty_log {
2072 __u32 slot;
2073 __u32 padding1;
2074 union {
2075 compat_uptr_t dirty_bitmap; /* one bit per page */
2076 __u64 padding2;
2077 };
2078};
2079
2080static long kvm_vm_compat_ioctl(struct file *filp,
2081 unsigned int ioctl, unsigned long arg)
2082{
2083 struct kvm *kvm = filp->private_data;
2084 int r;
2085
2086 if (kvm->mm != current->mm)
2087 return -EIO;
2088 switch (ioctl) {
2089 case KVM_GET_DIRTY_LOG: {
2090 struct compat_kvm_dirty_log compat_log;
2091 struct kvm_dirty_log log;
2092
2093 r = -EFAULT;
2094 if (copy_from_user(&compat_log, (void __user *)arg,
2095 sizeof(compat_log)))
2096 goto out;
2097 log.slot = compat_log.slot;
2098 log.padding1 = compat_log.padding1;
2099 log.padding2 = compat_log.padding2;
2100 log.dirty_bitmap = compat_ptr(compat_log.dirty_bitmap);
2101
2102 r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
2103 if (r)
2104 goto out;
2105 break;
2106 }
2107 default:
2108 r = kvm_vm_ioctl(filp, ioctl, arg);
2109 }
2110
2111out:
2112 return r;
2113}
2114#endif
2115
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002116static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivityf17abe92007-02-21 19:28:04 +02002117{
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002118 struct page *page[1];
2119 unsigned long addr;
2120 int npages;
2121 gfn_t gfn = vmf->pgoff;
Avi Kivityf17abe92007-02-21 19:28:04 +02002122 struct kvm *kvm = vma->vm_file->private_data;
Avi Kivityf17abe92007-02-21 19:28:04 +02002123
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002124 addr = gfn_to_hva(kvm, gfn);
2125 if (kvm_is_error_hva(addr))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002126 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002127
2128 npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page,
2129 NULL);
2130 if (unlikely(npages != 1))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002131 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002132
2133 vmf->page = page[0];
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002134 return 0;
Avi Kivityf17abe92007-02-21 19:28:04 +02002135}
2136
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +04002137static const struct vm_operations_struct kvm_vm_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002138 .fault = kvm_vm_fault,
Avi Kivityf17abe92007-02-21 19:28:04 +02002139};
2140
2141static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma)
2142{
2143 vma->vm_ops = &kvm_vm_vm_ops;
2144 return 0;
2145}
2146
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01002147static struct file_operations kvm_vm_fops = {
Avi Kivityf17abe92007-02-21 19:28:04 +02002148 .release = kvm_vm_release,
2149 .unlocked_ioctl = kvm_vm_ioctl,
Arnd Bergmann6ff58942009-10-22 14:19:27 +02002150#ifdef CONFIG_COMPAT
2151 .compat_ioctl = kvm_vm_compat_ioctl,
2152#endif
Avi Kivityf17abe92007-02-21 19:28:04 +02002153 .mmap = kvm_vm_mmap,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002154 .llseek = noop_llseek,
Avi Kivityf17abe92007-02-21 19:28:04 +02002155};
2156
2157static int kvm_dev_ioctl_create_vm(void)
2158{
Heiko Carstensaac87632010-10-27 17:22:10 +02002159 int r;
Avi Kivityf17abe92007-02-21 19:28:04 +02002160 struct kvm *kvm;
2161
Avi Kivityf17abe92007-02-21 19:28:04 +02002162 kvm = kvm_create_vm();
Avi Kivityd6d28162007-06-28 08:38:16 -04002163 if (IS_ERR(kvm))
2164 return PTR_ERR(kvm);
Takuya Yoshikawa6ce5a092010-03-15 22:13:30 +09002165#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
2166 r = kvm_coalesced_mmio_init(kvm);
2167 if (r < 0) {
2168 kvm_put_kvm(kvm);
2169 return r;
2170 }
2171#endif
Heiko Carstensaac87632010-10-27 17:22:10 +02002172 r = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, O_RDWR);
2173 if (r < 0)
Al Viro66c0b392008-04-19 20:33:56 +01002174 kvm_put_kvm(kvm);
Avi Kivityf17abe92007-02-21 19:28:04 +02002175
Heiko Carstensaac87632010-10-27 17:22:10 +02002176 return r;
Avi Kivityf17abe92007-02-21 19:28:04 +02002177}
2178
Avi Kivity1a811b62008-12-08 18:25:27 +02002179static long kvm_dev_ioctl_check_extension_generic(long arg)
2180{
2181 switch (arg) {
Avi Kivityca9edae2008-12-08 18:29:29 +02002182 case KVM_CAP_USER_MEMORY:
Avi Kivity1a811b62008-12-08 18:25:27 +02002183 case KVM_CAP_DESTROY_MEMORY_REGION_WORKS:
Jan Kiszka4cd481f2009-04-13 11:59:32 +02002184 case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS:
Gleb Natapov73880c82009-06-09 15:56:28 +03002185#ifdef CONFIG_KVM_APIC_ARCHITECTURE
2186 case KVM_CAP_SET_BOOT_CPU_ID:
2187#endif
Avi Kivitya9c73992009-11-04 11:54:59 +02002188 case KVM_CAP_INTERNAL_ERROR_DATA:
Avi Kivity1a811b62008-12-08 18:25:27 +02002189 return 1;
Avi Kivity399ec802008-11-19 13:58:46 +02002190#ifdef CONFIG_HAVE_KVM_IRQCHIP
2191 case KVM_CAP_IRQ_ROUTING:
Sheng Yang36463142009-03-16 16:33:43 +08002192 return KVM_MAX_IRQ_ROUTES;
Avi Kivity399ec802008-11-19 13:58:46 +02002193#endif
Avi Kivity1a811b62008-12-08 18:25:27 +02002194 default:
2195 break;
2196 }
2197 return kvm_dev_ioctl_check_extension(arg);
2198}
2199
Avi Kivityf17abe92007-02-21 19:28:04 +02002200static long kvm_dev_ioctl(struct file *filp,
2201 unsigned int ioctl, unsigned long arg)
2202{
Avi Kivity07c45a32007-03-07 13:05:38 +02002203 long r = -EINVAL;
Avi Kivityf17abe92007-02-21 19:28:04 +02002204
2205 switch (ioctl) {
2206 case KVM_GET_API_VERSION:
Avi Kivityf0fe5102007-03-07 13:11:17 +02002207 r = -EINVAL;
2208 if (arg)
2209 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02002210 r = KVM_API_VERSION;
2211 break;
2212 case KVM_CREATE_VM:
Avi Kivityf0fe5102007-03-07 13:11:17 +02002213 r = -EINVAL;
2214 if (arg)
2215 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02002216 r = kvm_dev_ioctl_create_vm();
2217 break;
Zhang Xiantao018d00d2007-11-15 23:07:47 +08002218 case KVM_CHECK_EXTENSION:
Avi Kivity1a811b62008-12-08 18:25:27 +02002219 r = kvm_dev_ioctl_check_extension_generic(arg);
Avi Kivity5d308f42007-03-01 17:56:20 +02002220 break;
Avi Kivity07c45a32007-03-07 13:05:38 +02002221 case KVM_GET_VCPU_MMAP_SIZE:
2222 r = -EINVAL;
2223 if (arg)
2224 goto out;
Avi Kivityadb1ff42008-01-24 15:13:08 +02002225 r = PAGE_SIZE; /* struct kvm_run */
2226#ifdef CONFIG_X86
2227 r += PAGE_SIZE; /* pio data page */
2228#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02002229#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
2230 r += PAGE_SIZE; /* coalesced mmio ring page */
2231#endif
Avi Kivity07c45a32007-03-07 13:05:38 +02002232 break;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04002233 case KVM_TRACE_ENABLE:
2234 case KVM_TRACE_PAUSE:
2235 case KVM_TRACE_DISABLE:
Marcelo Tosatti2023a292009-06-18 11:47:28 -03002236 r = -EOPNOTSUPP;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04002237 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002238 default:
Carsten Otte043405e2007-10-10 17:16:19 +02002239 return kvm_arch_dev_ioctl(filp, ioctl, arg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002240 }
2241out:
2242 return r;
2243}
2244
Avi Kivity6aa8b732006-12-10 02:21:36 -08002245static struct file_operations kvm_chardev_ops = {
Avi Kivity6aa8b732006-12-10 02:21:36 -08002246 .unlocked_ioctl = kvm_dev_ioctl,
2247 .compat_ioctl = kvm_dev_ioctl,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002248 .llseek = noop_llseek,
Avi Kivity6aa8b732006-12-10 02:21:36 -08002249};
2250
2251static struct miscdevice kvm_dev = {
Avi Kivitybbe44322007-03-04 13:27:36 +02002252 KVM_MINOR,
Avi Kivity6aa8b732006-12-10 02:21:36 -08002253 "kvm",
2254 &kvm_chardev_ops,
2255};
2256
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002257static void hardware_enable_nolock(void *junk)
Avi Kivity1b6c0162007-05-24 13:03:52 +03002258{
2259 int cpu = raw_smp_processor_id();
Alexander Graf10474ae2009-09-15 11:37:46 +02002260 int r;
Avi Kivity1b6c0162007-05-24 13:03:52 +03002261
Rusty Russell7f59f492008-12-07 21:25:45 +10302262 if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03002263 return;
Alexander Graf10474ae2009-09-15 11:37:46 +02002264
Rusty Russell7f59f492008-12-07 21:25:45 +10302265 cpumask_set_cpu(cpu, cpus_hardware_enabled);
Alexander Graf10474ae2009-09-15 11:37:46 +02002266
2267 r = kvm_arch_hardware_enable(NULL);
2268
2269 if (r) {
2270 cpumask_clear_cpu(cpu, cpus_hardware_enabled);
2271 atomic_inc(&hardware_enable_failed);
2272 printk(KERN_INFO "kvm: enabling virtualization on "
2273 "CPU%d failed\n", cpu);
2274 }
Avi Kivity1b6c0162007-05-24 13:03:52 +03002275}
2276
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002277static void hardware_enable(void *junk)
2278{
Jan Kiszkae935b832011-02-08 12:55:33 +01002279 raw_spin_lock(&kvm_lock);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002280 hardware_enable_nolock(junk);
Jan Kiszkae935b832011-02-08 12:55:33 +01002281 raw_spin_unlock(&kvm_lock);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002282}
2283
2284static void hardware_disable_nolock(void *junk)
Avi Kivity1b6c0162007-05-24 13:03:52 +03002285{
2286 int cpu = raw_smp_processor_id();
2287
Rusty Russell7f59f492008-12-07 21:25:45 +10302288 if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03002289 return;
Rusty Russell7f59f492008-12-07 21:25:45 +10302290 cpumask_clear_cpu(cpu, cpus_hardware_enabled);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002291 kvm_arch_hardware_disable(NULL);
Avi Kivity1b6c0162007-05-24 13:03:52 +03002292}
2293
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002294static void hardware_disable(void *junk)
2295{
Jan Kiszkae935b832011-02-08 12:55:33 +01002296 raw_spin_lock(&kvm_lock);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002297 hardware_disable_nolock(junk);
Jan Kiszkae935b832011-02-08 12:55:33 +01002298 raw_spin_unlock(&kvm_lock);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002299}
2300
Alexander Graf10474ae2009-09-15 11:37:46 +02002301static void hardware_disable_all_nolock(void)
2302{
2303 BUG_ON(!kvm_usage_count);
2304
2305 kvm_usage_count--;
2306 if (!kvm_usage_count)
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002307 on_each_cpu(hardware_disable_nolock, NULL, 1);
Alexander Graf10474ae2009-09-15 11:37:46 +02002308}
2309
2310static void hardware_disable_all(void)
2311{
Jan Kiszkae935b832011-02-08 12:55:33 +01002312 raw_spin_lock(&kvm_lock);
Alexander Graf10474ae2009-09-15 11:37:46 +02002313 hardware_disable_all_nolock();
Jan Kiszkae935b832011-02-08 12:55:33 +01002314 raw_spin_unlock(&kvm_lock);
Alexander Graf10474ae2009-09-15 11:37:46 +02002315}
2316
2317static int hardware_enable_all(void)
2318{
2319 int r = 0;
2320
Jan Kiszkae935b832011-02-08 12:55:33 +01002321 raw_spin_lock(&kvm_lock);
Alexander Graf10474ae2009-09-15 11:37:46 +02002322
2323 kvm_usage_count++;
2324 if (kvm_usage_count == 1) {
2325 atomic_set(&hardware_enable_failed, 0);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002326 on_each_cpu(hardware_enable_nolock, NULL, 1);
Alexander Graf10474ae2009-09-15 11:37:46 +02002327
2328 if (atomic_read(&hardware_enable_failed)) {
2329 hardware_disable_all_nolock();
2330 r = -EBUSY;
2331 }
2332 }
2333
Jan Kiszkae935b832011-02-08 12:55:33 +01002334 raw_spin_unlock(&kvm_lock);
Alexander Graf10474ae2009-09-15 11:37:46 +02002335
2336 return r;
2337}
2338
Avi Kivity774c47f2007-02-12 00:54:47 -08002339static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val,
2340 void *v)
2341{
2342 int cpu = (long)v;
2343
Alexander Graf10474ae2009-09-15 11:37:46 +02002344 if (!kvm_usage_count)
2345 return NOTIFY_OK;
2346
Avi Kivity1a6f4d72007-11-11 18:37:32 +02002347 val &= ~CPU_TASKS_FROZEN;
Avi Kivity774c47f2007-02-12 00:54:47 -08002348 switch (val) {
Avi Kivitycec9ad22007-05-24 13:11:41 +03002349 case CPU_DYING:
Avi Kivity6ec8a852007-08-19 15:57:26 +03002350 printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n",
2351 cpu);
2352 hardware_disable(NULL);
2353 break;
Zachary Amsdenda908f22010-08-19 22:07:27 -10002354 case CPU_STARTING:
Jeremy Katz43934a32007-02-19 14:37:46 +02002355 printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n",
2356 cpu);
Zachary Amsdenda908f22010-08-19 22:07:27 -10002357 hardware_enable(NULL);
Avi Kivity774c47f2007-02-12 00:54:47 -08002358 break;
2359 }
2360 return NOTIFY_OK;
2361}
2362
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002363
Avi Kivityb7c41452010-12-02 17:52:50 +02002364asmlinkage void kvm_spurious_fault(void)
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002365{
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002366 /* Fault while not rebooting. We want the trace. */
2367 BUG();
2368}
Avi Kivityb7c41452010-12-02 17:52:50 +02002369EXPORT_SYMBOL_GPL(kvm_spurious_fault);
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002370
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002371static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
Mike Dayd77c26f2007-10-08 09:02:08 -04002372 void *v)
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002373{
Sheng Yang8e1c1812009-04-29 11:09:04 +08002374 /*
2375 * Some (well, at least mine) BIOSes hang on reboot if
2376 * in vmx root mode.
2377 *
2378 * And Intel TXT required VMX off for all cpu when system shutdown.
2379 */
2380 printk(KERN_INFO "kvm: exiting hardware virtualization\n");
2381 kvm_rebooting = true;
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002382 on_each_cpu(hardware_disable_nolock, NULL, 1);
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002383 return NOTIFY_OK;
2384}
2385
2386static struct notifier_block kvm_reboot_notifier = {
2387 .notifier_call = kvm_reboot,
2388 .priority = 0,
2389};
2390
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002391static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002392{
2393 int i;
2394
2395 for (i = 0; i < bus->dev_count; i++) {
Sasha Levin743eeb02011-07-27 16:00:48 +03002396 struct kvm_io_device *pos = bus->range[i].dev;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002397
2398 kvm_iodevice_destructor(pos);
2399 }
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002400 kfree(bus);
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002401}
2402
Sasha Levin743eeb02011-07-27 16:00:48 +03002403int kvm_io_bus_sort_cmp(const void *p1, const void *p2)
2404{
2405 const struct kvm_io_range *r1 = p1;
2406 const struct kvm_io_range *r2 = p2;
2407
2408 if (r1->addr < r2->addr)
2409 return -1;
2410 if (r1->addr + r1->len > r2->addr + r2->len)
2411 return 1;
2412 return 0;
2413}
2414
2415int kvm_io_bus_insert_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev,
2416 gpa_t addr, int len)
2417{
2418 if (bus->dev_count == NR_IOBUS_DEVS)
2419 return -ENOSPC;
2420
2421 bus->range[bus->dev_count++] = (struct kvm_io_range) {
2422 .addr = addr,
2423 .len = len,
2424 .dev = dev,
2425 };
2426
2427 sort(bus->range, bus->dev_count, sizeof(struct kvm_io_range),
2428 kvm_io_bus_sort_cmp, NULL);
2429
2430 return 0;
2431}
2432
2433int kvm_io_bus_get_first_dev(struct kvm_io_bus *bus,
2434 gpa_t addr, int len)
2435{
2436 struct kvm_io_range *range, key;
2437 int off;
2438
2439 key = (struct kvm_io_range) {
2440 .addr = addr,
2441 .len = len,
2442 };
2443
2444 range = bsearch(&key, bus->range, bus->dev_count,
2445 sizeof(struct kvm_io_range), kvm_io_bus_sort_cmp);
2446 if (range == NULL)
2447 return -ENOENT;
2448
2449 off = range - bus->range;
2450
2451 while (off > 0 && kvm_io_bus_sort_cmp(&key, &bus->range[off-1]) == 0)
2452 off--;
2453
2454 return off;
2455}
2456
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002457/* kvm_io_bus_write - called under kvm->slots_lock */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002458int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002459 int len, const void *val)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002460{
Sasha Levin743eeb02011-07-27 16:00:48 +03002461 int idx;
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002462 struct kvm_io_bus *bus;
Sasha Levin743eeb02011-07-27 16:00:48 +03002463 struct kvm_io_range range;
2464
2465 range = (struct kvm_io_range) {
2466 .addr = addr,
2467 .len = len,
2468 };
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002469
2470 bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu);
Sasha Levin743eeb02011-07-27 16:00:48 +03002471 idx = kvm_io_bus_get_first_dev(bus, addr, len);
2472 if (idx < 0)
2473 return -EOPNOTSUPP;
2474
2475 while (idx < bus->dev_count &&
2476 kvm_io_bus_sort_cmp(&range, &bus->range[idx]) == 0) {
2477 if (!kvm_iodevice_write(bus->range[idx].dev, addr, len, val))
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002478 return 0;
Sasha Levin743eeb02011-07-27 16:00:48 +03002479 idx++;
2480 }
2481
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002482 return -EOPNOTSUPP;
2483}
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002484
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002485/* kvm_io_bus_read - called under kvm->slots_lock */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002486int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
2487 int len, void *val)
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002488{
Sasha Levin743eeb02011-07-27 16:00:48 +03002489 int idx;
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002490 struct kvm_io_bus *bus;
Sasha Levin743eeb02011-07-27 16:00:48 +03002491 struct kvm_io_range range;
2492
2493 range = (struct kvm_io_range) {
2494 .addr = addr,
2495 .len = len,
2496 };
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002497
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002498 bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu);
Sasha Levin743eeb02011-07-27 16:00:48 +03002499 idx = kvm_io_bus_get_first_dev(bus, addr, len);
2500 if (idx < 0)
2501 return -EOPNOTSUPP;
2502
2503 while (idx < bus->dev_count &&
2504 kvm_io_bus_sort_cmp(&range, &bus->range[idx]) == 0) {
2505 if (!kvm_iodevice_read(bus->range[idx].dev, addr, len, val))
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002506 return 0;
Sasha Levin743eeb02011-07-27 16:00:48 +03002507 idx++;
2508 }
2509
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002510 return -EOPNOTSUPP;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002511}
2512
Marcelo Tosatti79fac952009-12-23 14:35:26 -02002513/* Caller must hold slots_lock. */
Sasha Levin743eeb02011-07-27 16:00:48 +03002514int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
2515 int len, struct kvm_io_device *dev)
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002516{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002517 struct kvm_io_bus *new_bus, *bus;
Gregory Haskins090b7af2009-07-07 17:08:44 -04002518
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002519 bus = kvm->buses[bus_idx];
Gregory Haskins090b7af2009-07-07 17:08:44 -04002520 if (bus->dev_count > NR_IOBUS_DEVS-1)
2521 return -ENOSPC;
2522
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002523 new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL);
2524 if (!new_bus)
2525 return -ENOMEM;
2526 memcpy(new_bus, bus, sizeof(struct kvm_io_bus));
Sasha Levin743eeb02011-07-27 16:00:48 +03002527 kvm_io_bus_insert_dev(new_bus, dev, addr, len);
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002528 rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
2529 synchronize_srcu_expedited(&kvm->srcu);
2530 kfree(bus);
Gregory Haskins090b7af2009-07-07 17:08:44 -04002531
2532 return 0;
2533}
2534
Marcelo Tosatti79fac952009-12-23 14:35:26 -02002535/* Caller must hold slots_lock. */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002536int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
2537 struct kvm_io_device *dev)
Gregory Haskins090b7af2009-07-07 17:08:44 -04002538{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002539 int i, r;
2540 struct kvm_io_bus *new_bus, *bus;
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002541
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002542 new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL);
2543 if (!new_bus)
2544 return -ENOMEM;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002545
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002546 bus = kvm->buses[bus_idx];
2547 memcpy(new_bus, bus, sizeof(struct kvm_io_bus));
2548
2549 r = -ENOENT;
2550 for (i = 0; i < new_bus->dev_count; i++)
Sasha Levin743eeb02011-07-27 16:00:48 +03002551 if (new_bus->range[i].dev == dev) {
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002552 r = 0;
Sasha Levin743eeb02011-07-27 16:00:48 +03002553 new_bus->dev_count--;
2554 new_bus->range[i] = new_bus->range[new_bus->dev_count];
2555 sort(new_bus->range, new_bus->dev_count,
2556 sizeof(struct kvm_io_range),
2557 kvm_io_bus_sort_cmp, NULL);
Gregory Haskins090b7af2009-07-07 17:08:44 -04002558 break;
2559 }
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002560
2561 if (r) {
2562 kfree(new_bus);
2563 return r;
2564 }
2565
2566 rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
2567 synchronize_srcu_expedited(&kvm->srcu);
2568 kfree(bus);
2569 return r;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002570}
2571
Avi Kivity774c47f2007-02-12 00:54:47 -08002572static struct notifier_block kvm_cpu_notifier = {
2573 .notifier_call = kvm_cpu_hotplug,
Avi Kivity774c47f2007-02-12 00:54:47 -08002574};
2575
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002576static int vm_stat_get(void *_offset, u64 *val)
Avi Kivityba1389b2007-11-18 16:24:12 +02002577{
2578 unsigned offset = (long)_offset;
Avi Kivityba1389b2007-11-18 16:24:12 +02002579 struct kvm *kvm;
2580
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002581 *val = 0;
Jan Kiszkae935b832011-02-08 12:55:33 +01002582 raw_spin_lock(&kvm_lock);
Avi Kivityba1389b2007-11-18 16:24:12 +02002583 list_for_each_entry(kvm, &vm_list, vm_list)
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002584 *val += *(u32 *)((void *)kvm + offset);
Jan Kiszkae935b832011-02-08 12:55:33 +01002585 raw_spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002586 return 0;
Avi Kivityba1389b2007-11-18 16:24:12 +02002587}
2588
2589DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n");
2590
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002591static int vcpu_stat_get(void *_offset, u64 *val)
Avi Kivity1165f5f2007-04-19 17:27:43 +03002592{
2593 unsigned offset = (long)_offset;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002594 struct kvm *kvm;
2595 struct kvm_vcpu *vcpu;
2596 int i;
2597
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002598 *val = 0;
Jan Kiszkae935b832011-02-08 12:55:33 +01002599 raw_spin_lock(&kvm_lock);
Avi Kivity1165f5f2007-04-19 17:27:43 +03002600 list_for_each_entry(kvm, &vm_list, vm_list)
Gleb Natapov988a2ca2009-06-09 15:56:29 +03002601 kvm_for_each_vcpu(i, vcpu, kvm)
2602 *val += *(u32 *)((void *)vcpu + offset);
2603
Jan Kiszkae935b832011-02-08 12:55:33 +01002604 raw_spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002605 return 0;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002606}
2607
Avi Kivityba1389b2007-11-18 16:24:12 +02002608DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n");
2609
Alexey Dobriyan828c0952009-10-01 15:43:56 -07002610static const struct file_operations *stat_fops[] = {
Avi Kivityba1389b2007-11-18 16:24:12 +02002611 [KVM_STAT_VCPU] = &vcpu_stat_fops,
2612 [KVM_STAT_VM] = &vm_stat_fops,
2613};
Avi Kivity1165f5f2007-04-19 17:27:43 +03002614
Zhang Xiantaoa16b0432007-11-16 14:38:21 +08002615static void kvm_init_debug(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002616{
2617 struct kvm_stats_debugfs_item *p;
2618
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002619 kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002620 for (p = debugfs_entries; p->name; ++p)
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002621 p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir,
Avi Kivity1165f5f2007-04-19 17:27:43 +03002622 (void *)(long)p->offset,
Avi Kivityba1389b2007-11-18 16:24:12 +02002623 stat_fops[p->kind]);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002624}
2625
2626static void kvm_exit_debug(void)
2627{
2628 struct kvm_stats_debugfs_item *p;
2629
2630 for (p = debugfs_entries; p->name; ++p)
2631 debugfs_remove(p->dentry);
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002632 debugfs_remove(kvm_debugfs_dir);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002633}
2634
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002635static int kvm_suspend(void)
Avi Kivity59ae6c62007-02-12 00:54:48 -08002636{
Alexander Graf10474ae2009-09-15 11:37:46 +02002637 if (kvm_usage_count)
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002638 hardware_disable_nolock(NULL);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002639 return 0;
2640}
2641
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002642static void kvm_resume(void)
Avi Kivity59ae6c62007-02-12 00:54:48 -08002643{
Zachary Amsdenca84d1a2010-08-19 22:07:28 -10002644 if (kvm_usage_count) {
Jan Kiszkae935b832011-02-08 12:55:33 +01002645 WARN_ON(raw_spin_is_locked(&kvm_lock));
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002646 hardware_enable_nolock(NULL);
Zachary Amsdenca84d1a2010-08-19 22:07:28 -10002647 }
Avi Kivity59ae6c62007-02-12 00:54:48 -08002648}
2649
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002650static struct syscore_ops kvm_syscore_ops = {
Avi Kivity59ae6c62007-02-12 00:54:48 -08002651 .suspend = kvm_suspend,
2652 .resume = kvm_resume,
2653};
2654
Izik Eiduscea7bb22007-10-17 19:17:48 +02002655struct page *bad_page;
Anthony Liguori35149e22008-04-02 14:46:56 -05002656pfn_t bad_pfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002657
Avi Kivity15ad7142007-07-11 18:17:21 +03002658static inline
2659struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn)
2660{
2661 return container_of(pn, struct kvm_vcpu, preempt_notifier);
2662}
2663
2664static void kvm_sched_in(struct preempt_notifier *pn, int cpu)
2665{
2666 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2667
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002668 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002669}
2670
2671static void kvm_sched_out(struct preempt_notifier *pn,
2672 struct task_struct *next)
2673{
2674 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2675
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002676 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002677}
2678
Avi Kivity0ee75be2010-04-28 15:39:01 +03002679int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
Rusty Russellc16f8622007-07-30 21:12:19 +10002680 struct module *module)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002681{
2682 int r;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002683 int cpu;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002684
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002685 r = kvm_arch_init(opaque);
2686 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002687 goto out_fail;
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002688
2689 bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2690
2691 if (bad_page == NULL) {
2692 r = -ENOMEM;
2693 goto out;
2694 }
2695
Anthony Liguori35149e22008-04-02 14:46:56 -05002696 bad_pfn = page_to_pfn(bad_page);
2697
Huang Yingbf998152010-05-31 14:28:19 +08002698 hwpoison_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2699
2700 if (hwpoison_page == NULL) {
2701 r = -ENOMEM;
2702 goto out_free_0;
2703 }
2704
2705 hwpoison_pfn = page_to_pfn(hwpoison_page);
2706
Gleb Natapovedba23e2010-07-07 20:16:45 +03002707 fault_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2708
2709 if (fault_page == NULL) {
2710 r = -ENOMEM;
2711 goto out_free_0;
2712 }
2713
2714 fault_pfn = page_to_pfn(fault_page);
2715
Avi Kivity8437a612009-06-06 14:52:35 -07002716 if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
Rusty Russell7f59f492008-12-07 21:25:45 +10302717 r = -ENOMEM;
2718 goto out_free_0;
2719 }
2720
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002721 r = kvm_arch_hardware_setup();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002722 if (r < 0)
Rusty Russell7f59f492008-12-07 21:25:45 +10302723 goto out_free_0a;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002724
Yang, Sheng002c7f72007-07-31 14:23:01 +03002725 for_each_online_cpu(cpu) {
2726 smp_call_function_single(cpu,
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002727 kvm_arch_check_processor_compat,
Jens Axboe8691e5a2008-06-06 11:18:06 +02002728 &r, 1);
Yang, Sheng002c7f72007-07-31 14:23:01 +03002729 if (r < 0)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002730 goto out_free_1;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002731 }
2732
Avi Kivity774c47f2007-02-12 00:54:47 -08002733 r = register_cpu_notifier(&kvm_cpu_notifier);
2734 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002735 goto out_free_2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002736 register_reboot_notifier(&kvm_reboot_notifier);
2737
Rusty Russellc16f8622007-07-30 21:12:19 +10002738 /* A kmem cache lets us meet the alignment requirements of fx_save. */
Avi Kivity0ee75be2010-04-28 15:39:01 +03002739 if (!vcpu_align)
2740 vcpu_align = __alignof__(struct kvm_vcpu);
2741 kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, vcpu_align,
Joe Perches56919c52007-11-12 20:06:51 -08002742 0, NULL);
Rusty Russellc16f8622007-07-30 21:12:19 +10002743 if (!kvm_vcpu_cache) {
2744 r = -ENOMEM;
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002745 goto out_free_3;
Rusty Russellc16f8622007-07-30 21:12:19 +10002746 }
2747
Gleb Natapovaf585b92010-10-14 11:22:46 +02002748 r = kvm_async_pf_init();
2749 if (r)
2750 goto out_free;
2751
Avi Kivity6aa8b732006-12-10 02:21:36 -08002752 kvm_chardev_ops.owner = module;
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01002753 kvm_vm_fops.owner = module;
2754 kvm_vcpu_fops.owner = module;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002755
2756 r = misc_register(&kvm_dev);
2757 if (r) {
Mike Dayd77c26f2007-10-08 09:02:08 -04002758 printk(KERN_ERR "kvm: misc device register failed\n");
Gleb Natapovaf585b92010-10-14 11:22:46 +02002759 goto out_unreg;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002760 }
2761
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002762 register_syscore_ops(&kvm_syscore_ops);
2763
Avi Kivity15ad7142007-07-11 18:17:21 +03002764 kvm_preempt_ops.sched_in = kvm_sched_in;
2765 kvm_preempt_ops.sched_out = kvm_sched_out;
2766
Darrick J. Wong0ea4ed82009-10-14 16:21:00 -07002767 kvm_init_debug();
2768
Avi Kivityc7addb92007-09-16 18:58:32 +02002769 return 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002770
Gleb Natapovaf585b92010-10-14 11:22:46 +02002771out_unreg:
2772 kvm_async_pf_deinit();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002773out_free:
Rusty Russellc16f8622007-07-30 21:12:19 +10002774 kmem_cache_destroy(kvm_vcpu_cache);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002775out_free_3:
Avi Kivity6aa8b732006-12-10 02:21:36 -08002776 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity774c47f2007-02-12 00:54:47 -08002777 unregister_cpu_notifier(&kvm_cpu_notifier);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002778out_free_2:
Zhang Xiantaod23087842007-11-29 15:35:39 +08002779out_free_1:
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002780 kvm_arch_hardware_unsetup();
Rusty Russell7f59f492008-12-07 21:25:45 +10302781out_free_0a:
2782 free_cpumask_var(cpus_hardware_enabled);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002783out_free_0:
Gleb Natapovedba23e2010-07-07 20:16:45 +03002784 if (fault_page)
2785 __free_page(fault_page);
Huang Yingbf998152010-05-31 14:28:19 +08002786 if (hwpoison_page)
2787 __free_page(hwpoison_page);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002788 __free_page(bad_page);
Avi Kivityca45aaa2007-03-01 19:21:03 +02002789out:
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002790 kvm_arch_exit();
Zhang Xiantaod23087842007-11-29 15:35:39 +08002791out_fail:
Avi Kivity6aa8b732006-12-10 02:21:36 -08002792 return r;
2793}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002794EXPORT_SYMBOL_GPL(kvm_init);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002795
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002796void kvm_exit(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002797{
Darrick J. Wong0ea4ed82009-10-14 16:21:00 -07002798 kvm_exit_debug();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002799 misc_deregister(&kvm_dev);
Rusty Russellc16f8622007-07-30 21:12:19 +10002800 kmem_cache_destroy(kvm_vcpu_cache);
Gleb Natapovaf585b92010-10-14 11:22:46 +02002801 kvm_async_pf_deinit();
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002802 unregister_syscore_ops(&kvm_syscore_ops);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002803 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002804 unregister_cpu_notifier(&kvm_cpu_notifier);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002805 on_each_cpu(hardware_disable_nolock, NULL, 1);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002806 kvm_arch_hardware_unsetup();
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002807 kvm_arch_exit();
Rusty Russell7f59f492008-12-07 21:25:45 +10302808 free_cpumask_var(cpus_hardware_enabled);
Huang Yingbf998152010-05-31 14:28:19 +08002809 __free_page(hwpoison_page);
Izik Eiduscea7bb22007-10-17 19:17:48 +02002810 __free_page(bad_page);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002811}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002812EXPORT_SYMBOL_GPL(kvm_exit);