blob: 034a798b0431e98209489f793b904bfda9cf463c [file] [log] [blame]
Avi Kivity6aa8b732006-12-10 02:21:36 -08001/*
2 * Kernel-based Virtual Machine driver for Linux
3 *
4 * This module enables machines with Intel VT-x extensions to run virtual
5 * machines without emulation or binary translation.
6 *
7 * Copyright (C) 2006 Qumranet, Inc.
8 *
9 * Authors:
10 * Avi Kivity <avi@qumranet.com>
11 * Yaniv Kamay <yaniv@qumranet.com>
12 *
13 * This work is licensed under the terms of the GNU GPL, version 2. See
14 * the COPYING file in the top-level directory.
15 *
16 */
17
Hollis Blancharde2174022007-12-03 15:30:24 -060018#include "iodev.h"
Avi Kivity6aa8b732006-12-10 02:21:36 -080019
Avi Kivityedf88412007-12-16 11:02:48 +020020#include <linux/kvm_host.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080021#include <linux/kvm.h>
22#include <linux/module.h>
23#include <linux/errno.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080024#include <linux/percpu.h>
25#include <linux/gfp.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>
Avi Kivity59ae6c62007-02-12 00:54:48 -080033#include <linux/sysdev.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>
Avi Kivity6aa8b732006-12-10 02:21:36 -080046
Avi Kivitye4956062007-06-28 14:15:57 -040047#include <asm/processor.h>
Avi Kivitye4956062007-06-28 14:15:57 -040048#include <asm/io.h>
49#include <asm/uaccess.h>
Izik Eidus3e021bf2007-11-19 11:16:57 +020050#include <asm/pgtable.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080051
Laurent Vivier5f94c172008-05-30 16:05:54 +020052#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
53#include "coalesced_mmio.h"
54#endif
55
Xiantao Zhang8a98f662008-10-06 13:47:38 +080056#ifdef KVM_CAP_DEVICE_ASSIGNMENT
57#include <linux/pci.h>
58#include <linux/interrupt.h>
59#include "irq.h"
60#endif
61
Marcelo Tosatti229456f2009-06-17 09:22:14 -030062#define CREATE_TRACE_POINTS
63#include <trace/events/kvm.h>
64
Avi Kivity6aa8b732006-12-10 02:21:36 -080065MODULE_AUTHOR("Qumranet");
66MODULE_LICENSE("GPL");
67
Marcelo Tosattifa40a822009-06-04 15:08:24 -030068/*
69 * Ordering of locks:
70 *
Michael S. Tsirkin22fc0292009-06-29 22:24:45 +030071 * kvm->slots_lock --> kvm->lock --> kvm->irq_lock
Marcelo Tosattifa40a822009-06-04 15:08:24 -030072 */
73
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +080074DEFINE_SPINLOCK(kvm_lock);
75LIST_HEAD(vm_list);
Avi Kivity133de902007-02-12 00:54:44 -080076
Rusty Russell7f59f492008-12-07 21:25:45 +103077static cpumask_var_t cpus_hardware_enabled;
Avi Kivity1b6c0162007-05-24 13:03:52 +030078
Rusty Russellc16f8622007-07-30 21:12:19 +100079struct kmem_cache *kvm_vcpu_cache;
80EXPORT_SYMBOL_GPL(kvm_vcpu_cache);
Avi Kivity1165f5f2007-04-19 17:27:43 +030081
Avi Kivity15ad7142007-07-11 18:17:21 +030082static __read_mostly struct preempt_ops kvm_preempt_ops;
83
Hollis Blanchard76f7c872008-04-15 16:05:42 -050084struct dentry *kvm_debugfs_dir;
Avi Kivity6aa8b732006-12-10 02:21:36 -080085
Avi Kivitybccf2152007-02-21 18:04:26 +020086static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl,
87 unsigned long arg);
88
Hannes Edere8ba5d32008-11-28 17:02:42 +010089static bool kvm_rebooting;
Avi Kivity4ecac3f2008-05-13 13:23:38 +030090
Marcelo Tosatti54dee992009-06-11 12:07:44 -030091static bool largepages_enabled = true;
92
Xiantao Zhang8a98f662008-10-06 13:47:38 +080093#ifdef KVM_CAP_DEVICE_ASSIGNMENT
94static struct kvm_assigned_dev_kernel *kvm_find_assigned_dev(struct list_head *head,
95 int assigned_dev_id)
96{
97 struct list_head *ptr;
98 struct kvm_assigned_dev_kernel *match;
99
100 list_for_each(ptr, head) {
101 match = list_entry(ptr, struct kvm_assigned_dev_kernel, list);
102 if (match->assigned_dev_id == assigned_dev_id)
103 return match;
104 }
105 return NULL;
106}
107
Sheng Yang2350bd12009-02-25 17:22:27 +0800108static int find_index_from_host_irq(struct kvm_assigned_dev_kernel
109 *assigned_dev, int irq)
110{
111 int i, index;
112 struct msix_entry *host_msix_entries;
113
114 host_msix_entries = assigned_dev->host_msix_entries;
115
116 index = -1;
117 for (i = 0; i < assigned_dev->entries_nr; i++)
118 if (irq == host_msix_entries[i].vector) {
119 index = i;
120 break;
121 }
122 if (index < 0) {
123 printk(KERN_WARNING "Fail to find correlated MSI-X entry!\n");
124 return 0;
125 }
126
127 return index;
128}
129
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800130static void kvm_assigned_dev_interrupt_work_handler(struct work_struct *work)
131{
132 struct kvm_assigned_dev_kernel *assigned_dev;
Sheng Yang2350bd12009-02-25 17:22:27 +0800133 struct kvm *kvm;
Sheng Yang968a6342009-04-30 10:58:42 +0800134 int i;
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800135
136 assigned_dev = container_of(work, struct kvm_assigned_dev_kernel,
137 interrupt_work);
Sheng Yang2350bd12009-02-25 17:22:27 +0800138 kvm = assigned_dev->kvm;
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800139
Marcelo Tosattifa40a822009-06-04 15:08:24 -0300140 mutex_lock(&kvm->irq_lock);
Marcelo Tosatti547de292009-05-07 17:55:13 -0300141 spin_lock_irq(&assigned_dev->assigned_dev_lock);
Sheng Yange56d5322009-03-12 21:45:39 +0800142 if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_HOST_MSIX) {
Sheng Yang2350bd12009-02-25 17:22:27 +0800143 struct kvm_guest_msix_entry *guest_entries =
144 assigned_dev->guest_msix_entries;
145 for (i = 0; i < assigned_dev->entries_nr; i++) {
146 if (!(guest_entries[i].flags &
147 KVM_ASSIGNED_MSIX_PENDING))
148 continue;
149 guest_entries[i].flags &= ~KVM_ASSIGNED_MSIX_PENDING;
150 kvm_set_irq(assigned_dev->kvm,
151 assigned_dev->irq_source_id,
152 guest_entries[i].vector, 1);
Sheng Yang2350bd12009-02-25 17:22:27 +0800153 }
Sheng Yang968a6342009-04-30 10:58:42 +0800154 } else
Sheng Yang2350bd12009-02-25 17:22:27 +0800155 kvm_set_irq(assigned_dev->kvm, assigned_dev->irq_source_id,
156 assigned_dev->guest_irq, 1);
Sheng Yang2350bd12009-02-25 17:22:27 +0800157
Marcelo Tosatti547de292009-05-07 17:55:13 -0300158 spin_unlock_irq(&assigned_dev->assigned_dev_lock);
Marcelo Tosattifa40a822009-06-04 15:08:24 -0300159 mutex_unlock(&assigned_dev->kvm->irq_lock);
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800160}
161
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800162static irqreturn_t kvm_assigned_dev_intr(int irq, void *dev_id)
163{
Marcelo Tosatti547de292009-05-07 17:55:13 -0300164 unsigned long flags;
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800165 struct kvm_assigned_dev_kernel *assigned_dev =
166 (struct kvm_assigned_dev_kernel *) dev_id;
167
Marcelo Tosatti547de292009-05-07 17:55:13 -0300168 spin_lock_irqsave(&assigned_dev->assigned_dev_lock, flags);
Sheng Yange56d5322009-03-12 21:45:39 +0800169 if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_HOST_MSIX) {
Sheng Yang2350bd12009-02-25 17:22:27 +0800170 int index = find_index_from_host_irq(assigned_dev, irq);
171 if (index < 0)
Marcelo Tosatti547de292009-05-07 17:55:13 -0300172 goto out;
Sheng Yang2350bd12009-02-25 17:22:27 +0800173 assigned_dev->guest_msix_entries[index].flags |=
174 KVM_ASSIGNED_MSIX_PENDING;
175 }
176
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800177 schedule_work(&assigned_dev->interrupt_work);
Mark McLoughlindefaf152008-12-02 12:16:33 +0000178
Sheng Yang968a6342009-04-30 10:58:42 +0800179 if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_GUEST_INTX) {
180 disable_irq_nosync(irq);
181 assigned_dev->host_irq_disabled = true;
182 }
Mark McLoughlindefaf152008-12-02 12:16:33 +0000183
Marcelo Tosatti547de292009-05-07 17:55:13 -0300184out:
185 spin_unlock_irqrestore(&assigned_dev->assigned_dev_lock, flags);
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800186 return IRQ_HANDLED;
187}
188
189/* Ack the irq line for an assigned device */
190static void kvm_assigned_dev_ack_irq(struct kvm_irq_ack_notifier *kian)
191{
192 struct kvm_assigned_dev_kernel *dev;
Marcelo Tosatti547de292009-05-07 17:55:13 -0300193 unsigned long flags;
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800194
195 if (kian->gsi == -1)
196 return;
197
198 dev = container_of(kian, struct kvm_assigned_dev_kernel,
199 ack_notifier);
Mark McLoughlindefaf152008-12-02 12:16:33 +0000200
Sheng Yang5550af42008-10-15 20:15:06 +0800201 kvm_set_irq(dev->kvm, dev->irq_source_id, dev->guest_irq, 0);
Mark McLoughlindefaf152008-12-02 12:16:33 +0000202
203 /* The guest irq may be shared so this ack may be
204 * from another device.
205 */
Marcelo Tosatti547de292009-05-07 17:55:13 -0300206 spin_lock_irqsave(&dev->assigned_dev_lock, flags);
Mark McLoughlindefaf152008-12-02 12:16:33 +0000207 if (dev->host_irq_disabled) {
208 enable_irq(dev->host_irq);
209 dev->host_irq_disabled = false;
210 }
Marcelo Tosatti547de292009-05-07 17:55:13 -0300211 spin_unlock_irqrestore(&dev->assigned_dev_lock, flags);
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800212}
213
Sheng Yange56d5322009-03-12 21:45:39 +0800214static void deassign_guest_irq(struct kvm *kvm,
215 struct kvm_assigned_dev_kernel *assigned_dev)
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800216{
Marcelo Tosattifa40a822009-06-04 15:08:24 -0300217 kvm_unregister_irq_ack_notifier(kvm, &assigned_dev->ack_notifier);
Sheng Yange56d5322009-03-12 21:45:39 +0800218 assigned_dev->ack_notifier.gsi = -1;
Mark McLoughlinf29b2672008-12-01 13:57:47 +0000219
220 if (assigned_dev->irq_source_id != -1)
221 kvm_free_irq_source_id(kvm, assigned_dev->irq_source_id);
222 assigned_dev->irq_source_id = -1;
Sheng Yange56d5322009-03-12 21:45:39 +0800223 assigned_dev->irq_requested_type &= ~(KVM_DEV_IRQ_GUEST_MASK);
224}
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800225
Sheng Yange56d5322009-03-12 21:45:39 +0800226/* The function implicit hold kvm->lock mutex due to cancel_work_sync() */
227static void deassign_host_irq(struct kvm *kvm,
228 struct kvm_assigned_dev_kernel *assigned_dev)
229{
Sheng Yangba4cef32009-01-06 10:03:03 +0800230 /*
231 * In kvm_free_device_irq, cancel_work_sync return true if:
232 * 1. work is scheduled, and then cancelled.
233 * 2. work callback is executed.
234 *
235 * The first one ensured that the irq is disabled and no more events
236 * would happen. But for the second one, the irq may be enabled (e.g.
237 * for MSI). So we disable irq here to prevent further events.
238 *
239 * Notice this maybe result in nested disable if the interrupt type is
240 * INTx, but it's OK for we are going to free it.
241 *
242 * If this function is a part of VM destroy, please ensure that till
243 * now, the kvm state is still legal for probably we also have to wait
244 * interrupt_work done.
245 */
Sheng Yange56d5322009-03-12 21:45:39 +0800246 if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_HOST_MSIX) {
Sheng Yangd510d6c2009-02-25 17:22:28 +0800247 int i;
248 for (i = 0; i < assigned_dev->entries_nr; i++)
249 disable_irq_nosync(assigned_dev->
250 host_msix_entries[i].vector);
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800251
Sheng Yangd510d6c2009-02-25 17:22:28 +0800252 cancel_work_sync(&assigned_dev->interrupt_work);
Mark McLoughlin4a643be2008-12-01 13:57:49 +0000253
Sheng Yangd510d6c2009-02-25 17:22:28 +0800254 for (i = 0; i < assigned_dev->entries_nr; i++)
255 free_irq(assigned_dev->host_msix_entries[i].vector,
256 (void *)assigned_dev);
257
258 assigned_dev->entries_nr = 0;
259 kfree(assigned_dev->host_msix_entries);
260 kfree(assigned_dev->guest_msix_entries);
261 pci_disable_msix(assigned_dev->dev);
262 } else {
263 /* Deal with MSI and INTx */
264 disable_irq_nosync(assigned_dev->host_irq);
265 cancel_work_sync(&assigned_dev->interrupt_work);
266
267 free_irq(assigned_dev->host_irq, (void *)assigned_dev);
268
Sheng Yange56d5322009-03-12 21:45:39 +0800269 if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_HOST_MSI)
Sheng Yangd510d6c2009-02-25 17:22:28 +0800270 pci_disable_msi(assigned_dev->dev);
271 }
Mark McLoughlin4a643be2008-12-01 13:57:49 +0000272
Sheng Yange56d5322009-03-12 21:45:39 +0800273 assigned_dev->irq_requested_type &= ~(KVM_DEV_IRQ_HOST_MASK);
Mark McLoughlin4a643be2008-12-01 13:57:49 +0000274}
275
Sheng Yange56d5322009-03-12 21:45:39 +0800276static int kvm_deassign_irq(struct kvm *kvm,
277 struct kvm_assigned_dev_kernel *assigned_dev,
278 unsigned long irq_requested_type)
279{
280 unsigned long guest_irq_type, host_irq_type;
281
282 if (!irqchip_in_kernel(kvm))
283 return -EINVAL;
284 /* no irq assignment to deassign */
285 if (!assigned_dev->irq_requested_type)
286 return -ENXIO;
287
288 host_irq_type = irq_requested_type & KVM_DEV_IRQ_HOST_MASK;
289 guest_irq_type = irq_requested_type & KVM_DEV_IRQ_GUEST_MASK;
290
291 if (host_irq_type)
292 deassign_host_irq(kvm, assigned_dev);
293 if (guest_irq_type)
294 deassign_guest_irq(kvm, assigned_dev);
295
296 return 0;
297}
298
299static void kvm_free_assigned_irq(struct kvm *kvm,
300 struct kvm_assigned_dev_kernel *assigned_dev)
301{
302 kvm_deassign_irq(kvm, assigned_dev, assigned_dev->irq_requested_type);
303}
Mark McLoughlin4a643be2008-12-01 13:57:49 +0000304
305static void kvm_free_assigned_device(struct kvm *kvm,
306 struct kvm_assigned_dev_kernel
307 *assigned_dev)
308{
309 kvm_free_assigned_irq(kvm, assigned_dev);
310
Sheng Yang6eb55812008-10-31 12:37:41 +0800311 pci_reset_function(assigned_dev->dev);
312
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800313 pci_release_regions(assigned_dev->dev);
314 pci_disable_device(assigned_dev->dev);
315 pci_dev_put(assigned_dev->dev);
316
317 list_del(&assigned_dev->list);
318 kfree(assigned_dev);
319}
320
321void kvm_free_all_assigned_devices(struct kvm *kvm)
322{
323 struct list_head *ptr, *ptr2;
324 struct kvm_assigned_dev_kernel *assigned_dev;
325
326 list_for_each_safe(ptr, ptr2, &kvm->arch.assigned_dev_head) {
327 assigned_dev = list_entry(ptr,
328 struct kvm_assigned_dev_kernel,
329 list);
330
331 kvm_free_assigned_device(kvm, assigned_dev);
332 }
333}
334
Sheng Yange56d5322009-03-12 21:45:39 +0800335static int assigned_device_enable_host_intx(struct kvm *kvm,
336 struct kvm_assigned_dev_kernel *dev)
Sheng Yang00e3ed32008-11-24 14:32:50 +0800337{
Sheng Yange56d5322009-03-12 21:45:39 +0800338 dev->host_irq = dev->dev->irq;
339 /* Even though this is PCI, we don't want to use shared
340 * interrupts. Sharing host devices with guest-assigned devices
341 * on the same interrupt line is not a happy situation: there
342 * are going to be long delays in accepting, acking, etc.
343 */
344 if (request_irq(dev->host_irq, kvm_assigned_dev_intr,
345 0, "kvm_assigned_intx_device", (void *)dev))
346 return -EIO;
Sheng Yang00e3ed32008-11-24 14:32:50 +0800347 return 0;
348}
349
Sheng Yange56d5322009-03-12 21:45:39 +0800350#ifdef __KVM_HAVE_MSI
351static int assigned_device_enable_host_msi(struct kvm *kvm,
352 struct kvm_assigned_dev_kernel *dev)
Sheng Yang6b9cc7f2008-11-24 14:32:56 +0800353{
354 int r;
355
Sheng Yange56d5322009-03-12 21:45:39 +0800356 if (!dev->dev->msi_enabled) {
357 r = pci_enable_msi(dev->dev);
358 if (r)
359 return r;
Sheng Yang5319c662008-11-24 14:32:57 +0800360 }
Sheng Yang6b9cc7f2008-11-24 14:32:56 +0800361
Sheng Yange56d5322009-03-12 21:45:39 +0800362 dev->host_irq = dev->dev->irq;
363 if (request_irq(dev->host_irq, kvm_assigned_dev_intr, 0,
364 "kvm_assigned_msi_device", (void *)dev)) {
365 pci_disable_msi(dev->dev);
366 return -EIO;
Sheng Yang6b9cc7f2008-11-24 14:32:56 +0800367 }
368
Sheng Yang6b9cc7f2008-11-24 14:32:56 +0800369 return 0;
370}
371#endif
372
Sheng Yangd510d6c2009-02-25 17:22:28 +0800373#ifdef __KVM_HAVE_MSIX
Sheng Yange56d5322009-03-12 21:45:39 +0800374static int assigned_device_enable_host_msix(struct kvm *kvm,
375 struct kvm_assigned_dev_kernel *dev)
Sheng Yangd510d6c2009-02-25 17:22:28 +0800376{
Sheng Yange56d5322009-03-12 21:45:39 +0800377 int i, r = -EINVAL;
Sheng Yangd510d6c2009-02-25 17:22:28 +0800378
Sheng Yange56d5322009-03-12 21:45:39 +0800379 /* host_msix_entries and guest_msix_entries should have been
380 * initialized */
381 if (dev->entries_nr == 0)
382 return r;
Sheng Yangd510d6c2009-02-25 17:22:28 +0800383
Sheng Yange56d5322009-03-12 21:45:39 +0800384 r = pci_enable_msix(dev->dev, dev->host_msix_entries, dev->entries_nr);
385 if (r)
386 return r;
Sheng Yangd510d6c2009-02-25 17:22:28 +0800387
Sheng Yange56d5322009-03-12 21:45:39 +0800388 for (i = 0; i < dev->entries_nr; i++) {
389 r = request_irq(dev->host_msix_entries[i].vector,
390 kvm_assigned_dev_intr, 0,
391 "kvm_assigned_msix_device",
392 (void *)dev);
393 /* FIXME: free requested_irq's on failure */
Sheng Yangd510d6c2009-02-25 17:22:28 +0800394 if (r)
395 return r;
Sheng Yangd510d6c2009-02-25 17:22:28 +0800396 }
397
Sheng Yange56d5322009-03-12 21:45:39 +0800398 return 0;
399}
Sheng Yangd510d6c2009-02-25 17:22:28 +0800400
Sheng Yange56d5322009-03-12 21:45:39 +0800401#endif
402
403static int assigned_device_enable_guest_intx(struct kvm *kvm,
404 struct kvm_assigned_dev_kernel *dev,
405 struct kvm_assigned_irq *irq)
406{
407 dev->guest_irq = irq->guest_irq;
408 dev->ack_notifier.gsi = irq->guest_irq;
409 return 0;
410}
411
412#ifdef __KVM_HAVE_MSI
413static int assigned_device_enable_guest_msi(struct kvm *kvm,
414 struct kvm_assigned_dev_kernel *dev,
415 struct kvm_assigned_irq *irq)
416{
417 dev->guest_irq = irq->guest_irq;
418 dev->ack_notifier.gsi = -1;
Sheng Yang968a6342009-04-30 10:58:42 +0800419 dev->host_irq_disabled = false;
Sheng Yange56d5322009-03-12 21:45:39 +0800420 return 0;
421}
422#endif
423#ifdef __KVM_HAVE_MSIX
424static int assigned_device_enable_guest_msix(struct kvm *kvm,
425 struct kvm_assigned_dev_kernel *dev,
426 struct kvm_assigned_irq *irq)
427{
428 dev->guest_irq = irq->guest_irq;
429 dev->ack_notifier.gsi = -1;
Sheng Yang968a6342009-04-30 10:58:42 +0800430 dev->host_irq_disabled = false;
Sheng Yangd510d6c2009-02-25 17:22:28 +0800431 return 0;
432}
433#endif
434
Sheng Yange56d5322009-03-12 21:45:39 +0800435static int assign_host_irq(struct kvm *kvm,
436 struct kvm_assigned_dev_kernel *dev,
437 __u32 host_irq_type)
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800438{
Sheng Yange56d5322009-03-12 21:45:39 +0800439 int r = -EEXIST;
440
441 if (dev->irq_requested_type & KVM_DEV_IRQ_HOST_MASK)
442 return r;
443
444 switch (host_irq_type) {
445 case KVM_DEV_IRQ_HOST_INTX:
446 r = assigned_device_enable_host_intx(kvm, dev);
447 break;
448#ifdef __KVM_HAVE_MSI
449 case KVM_DEV_IRQ_HOST_MSI:
450 r = assigned_device_enable_host_msi(kvm, dev);
451 break;
452#endif
453#ifdef __KVM_HAVE_MSIX
454 case KVM_DEV_IRQ_HOST_MSIX:
455 r = assigned_device_enable_host_msix(kvm, dev);
456 break;
457#endif
458 default:
459 r = -EINVAL;
460 }
461
462 if (!r)
463 dev->irq_requested_type |= host_irq_type;
464
465 return r;
466}
467
468static int assign_guest_irq(struct kvm *kvm,
469 struct kvm_assigned_dev_kernel *dev,
470 struct kvm_assigned_irq *irq,
471 unsigned long guest_irq_type)
472{
473 int id;
474 int r = -EEXIST;
475
476 if (dev->irq_requested_type & KVM_DEV_IRQ_GUEST_MASK)
477 return r;
478
479 id = kvm_request_irq_source_id(kvm);
480 if (id < 0)
481 return id;
482
483 dev->irq_source_id = id;
484
485 switch (guest_irq_type) {
486 case KVM_DEV_IRQ_GUEST_INTX:
487 r = assigned_device_enable_guest_intx(kvm, dev, irq);
488 break;
489#ifdef __KVM_HAVE_MSI
490 case KVM_DEV_IRQ_GUEST_MSI:
491 r = assigned_device_enable_guest_msi(kvm, dev, irq);
492 break;
493#endif
494#ifdef __KVM_HAVE_MSIX
495 case KVM_DEV_IRQ_GUEST_MSIX:
496 r = assigned_device_enable_guest_msix(kvm, dev, irq);
497 break;
498#endif
499 default:
500 r = -EINVAL;
501 }
502
503 if (!r) {
504 dev->irq_requested_type |= guest_irq_type;
505 kvm_register_irq_ack_notifier(kvm, &dev->ack_notifier);
506 } else
507 kvm_free_irq_source_id(kvm, dev->irq_source_id);
508
509 return r;
510}
511
512/* TODO Deal with KVM_DEV_IRQ_ASSIGNED_MASK_MSIX */
513static int kvm_vm_ioctl_assign_irq(struct kvm *kvm,
514 struct kvm_assigned_irq *assigned_irq)
515{
516 int r = -EINVAL;
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800517 struct kvm_assigned_dev_kernel *match;
Sheng Yange56d5322009-03-12 21:45:39 +0800518 unsigned long host_irq_type, guest_irq_type;
519
520 if (!capable(CAP_SYS_RAWIO))
521 return -EPERM;
522
523 if (!irqchip_in_kernel(kvm))
524 return r;
525
526 mutex_lock(&kvm->lock);
527 r = -ENODEV;
528 match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head,
529 assigned_irq->assigned_dev_id);
530 if (!match)
531 goto out;
532
533 host_irq_type = (assigned_irq->flags & KVM_DEV_IRQ_HOST_MASK);
534 guest_irq_type = (assigned_irq->flags & KVM_DEV_IRQ_GUEST_MASK);
535
536 r = -EINVAL;
537 /* can only assign one type at a time */
538 if (hweight_long(host_irq_type) > 1)
539 goto out;
540 if (hweight_long(guest_irq_type) > 1)
541 goto out;
542 if (host_irq_type == 0 && guest_irq_type == 0)
543 goto out;
544
545 r = 0;
546 if (host_irq_type)
547 r = assign_host_irq(kvm, match, host_irq_type);
548 if (r)
549 goto out;
550
551 if (guest_irq_type)
552 r = assign_guest_irq(kvm, match, assigned_irq, guest_irq_type);
553out:
554 mutex_unlock(&kvm->lock);
555 return r;
556}
557
558static int kvm_vm_ioctl_deassign_dev_irq(struct kvm *kvm,
559 struct kvm_assigned_irq
560 *assigned_irq)
561{
562 int r = -ENODEV;
563 struct kvm_assigned_dev_kernel *match;
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800564
565 mutex_lock(&kvm->lock);
566
567 match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head,
568 assigned_irq->assigned_dev_id);
Sheng Yange56d5322009-03-12 21:45:39 +0800569 if (!match)
570 goto out;
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800571
Sheng Yange56d5322009-03-12 21:45:39 +0800572 r = kvm_deassign_irq(kvm, match, assigned_irq->flags);
573out:
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800574 mutex_unlock(&kvm->lock);
575 return r;
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800576}
577
578static int kvm_vm_ioctl_assign_device(struct kvm *kvm,
579 struct kvm_assigned_pci_dev *assigned_dev)
580{
581 int r = 0;
582 struct kvm_assigned_dev_kernel *match;
583 struct pci_dev *dev;
584
Mark McLoughlin682edb42009-02-05 18:23:46 +0000585 down_read(&kvm->slots_lock);
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800586 mutex_lock(&kvm->lock);
587
588 match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head,
589 assigned_dev->assigned_dev_id);
590 if (match) {
591 /* device already assigned */
Sheng Yange56d5322009-03-12 21:45:39 +0800592 r = -EEXIST;
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800593 goto out;
594 }
595
596 match = kzalloc(sizeof(struct kvm_assigned_dev_kernel), GFP_KERNEL);
597 if (match == NULL) {
598 printk(KERN_INFO "%s: Couldn't allocate memory\n",
599 __func__);
600 r = -ENOMEM;
601 goto out;
602 }
603 dev = pci_get_bus_and_slot(assigned_dev->busnr,
604 assigned_dev->devfn);
605 if (!dev) {
606 printk(KERN_INFO "%s: host device not found\n", __func__);
607 r = -EINVAL;
608 goto out_free;
609 }
610 if (pci_enable_device(dev)) {
611 printk(KERN_INFO "%s: Could not enable PCI device\n", __func__);
612 r = -EBUSY;
613 goto out_put;
614 }
615 r = pci_request_regions(dev, "kvm_assigned_device");
616 if (r) {
617 printk(KERN_INFO "%s: Could not get access to device regions\n",
618 __func__);
619 goto out_disable;
620 }
Sheng Yang6eb55812008-10-31 12:37:41 +0800621
622 pci_reset_function(dev);
623
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800624 match->assigned_dev_id = assigned_dev->assigned_dev_id;
625 match->host_busnr = assigned_dev->busnr;
626 match->host_devfn = assigned_dev->devfn;
Weidong Hanb6535742008-12-08 23:29:53 +0800627 match->flags = assigned_dev->flags;
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800628 match->dev = dev;
Marcelo Tosatti547de292009-05-07 17:55:13 -0300629 spin_lock_init(&match->assigned_dev_lock);
Mark McLoughlinf29b2672008-12-01 13:57:47 +0000630 match->irq_source_id = -1;
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800631 match->kvm = kvm;
Sheng Yange56d5322009-03-12 21:45:39 +0800632 match->ack_notifier.irq_acked = kvm_assigned_dev_ack_irq;
633 INIT_WORK(&match->interrupt_work,
634 kvm_assigned_dev_interrupt_work_handler);
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800635
636 list_add(&match->list, &kvm->arch.assigned_dev_head);
637
638 if (assigned_dev->flags & KVM_DEV_ASSIGN_ENABLE_IOMMU) {
Joerg Roedel19de40a2008-12-03 14:43:34 +0100639 if (!kvm->arch.iommu_domain) {
Weidong Han260782b2008-12-02 21:03:39 +0800640 r = kvm_iommu_map_guest(kvm);
641 if (r)
642 goto out_list_del;
643 }
644 r = kvm_assign_device(kvm, match);
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800645 if (r)
646 goto out_list_del;
647 }
648
649out:
650 mutex_unlock(&kvm->lock);
Mark McLoughlin682edb42009-02-05 18:23:46 +0000651 up_read(&kvm->slots_lock);
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800652 return r;
653out_list_del:
654 list_del(&match->list);
655 pci_release_regions(dev);
656out_disable:
657 pci_disable_device(dev);
658out_put:
659 pci_dev_put(dev);
660out_free:
661 kfree(match);
662 mutex_unlock(&kvm->lock);
Mark McLoughlin682edb42009-02-05 18:23:46 +0000663 up_read(&kvm->slots_lock);
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800664 return r;
665}
666#endif
667
Weidong Han0a920352008-12-02 21:24:23 +0800668#ifdef KVM_CAP_DEVICE_DEASSIGNMENT
669static int kvm_vm_ioctl_deassign_device(struct kvm *kvm,
670 struct kvm_assigned_pci_dev *assigned_dev)
671{
672 int r = 0;
673 struct kvm_assigned_dev_kernel *match;
674
675 mutex_lock(&kvm->lock);
676
677 match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head,
678 assigned_dev->assigned_dev_id);
679 if (!match) {
680 printk(KERN_INFO "%s: device hasn't been assigned before, "
681 "so cannot be deassigned\n", __func__);
682 r = -EINVAL;
683 goto out;
684 }
685
Weidong Han4a906e42009-02-13 17:27:51 +0800686 if (match->flags & KVM_DEV_ASSIGN_ENABLE_IOMMU)
Weidong Han0a920352008-12-02 21:24:23 +0800687 kvm_deassign_device(kvm, match);
688
689 kvm_free_assigned_device(kvm, match);
690
691out:
692 mutex_unlock(&kvm->lock);
693 return r;
694}
695#endif
696
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +0800697inline int kvm_is_mmio_pfn(pfn_t pfn)
Ben-Ami Yassourcbff90a2008-07-28 19:26:24 +0300698{
Joerg Roedelfc5659c2009-02-18 14:08:58 +0100699 if (pfn_valid(pfn)) {
700 struct page *page = compound_head(pfn_to_page(pfn));
701 return PageReserved(page);
702 }
Ben-Ami Yassourcbff90a2008-07-28 19:26:24 +0300703
704 return true;
705}
706
Avi Kivity6aa8b732006-12-10 02:21:36 -0800707/*
708 * Switches to specified vcpu, until a matching vcpu_put()
709 */
Carsten Otte313a3dc2007-10-11 19:16:52 +0200710void vcpu_load(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800711{
Avi Kivity15ad7142007-07-11 18:17:21 +0300712 int cpu;
713
Avi Kivitybccf2152007-02-21 18:04:26 +0200714 mutex_lock(&vcpu->mutex);
Avi Kivity15ad7142007-07-11 18:17:21 +0300715 cpu = get_cpu();
716 preempt_notifier_register(&vcpu->preempt_notifier);
Carsten Otte313a3dc2007-10-11 19:16:52 +0200717 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +0300718 put_cpu();
Avi Kivitybccf2152007-02-21 18:04:26 +0200719}
720
Carsten Otte313a3dc2007-10-11 19:16:52 +0200721void vcpu_put(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800722{
Avi Kivity15ad7142007-07-11 18:17:21 +0300723 preempt_disable();
Carsten Otte313a3dc2007-10-11 19:16:52 +0200724 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +0300725 preempt_notifier_unregister(&vcpu->preempt_notifier);
726 preempt_enable();
Avi Kivity6aa8b732006-12-10 02:21:36 -0800727 mutex_unlock(&vcpu->mutex);
728}
729
Avi Kivityd9e368d2007-06-07 19:18:30 +0300730static void ack_flush(void *_completed)
731{
Avi Kivityd9e368d2007-06-07 19:18:30 +0300732}
733
Rusty Russell49846892008-12-08 20:26:24 +1030734static bool make_all_cpus_request(struct kvm *kvm, unsigned int req)
Avi Kivityd9e368d2007-06-07 19:18:30 +0300735{
Avi Kivity597a5f52008-07-20 14:24:22 +0300736 int i, cpu, me;
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030737 cpumask_var_t cpus;
738 bool called = true;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300739 struct kvm_vcpu *vcpu;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300740
Li Zefan79f55992009-06-15 14:58:26 +0800741 zalloc_cpumask_var(&cpus, GFP_ATOMIC);
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030742
Marcelo Tosatti84261922009-06-17 10:53:47 -0300743 spin_lock(&kvm->requests_lock);
Jan Kiszkae601e3b2009-07-20 11:30:12 +0200744 me = smp_processor_id();
Gleb Natapov988a2ca2009-06-09 15:56:29 +0300745 kvm_for_each_vcpu(i, vcpu, kvm) {
Rusty Russell49846892008-12-08 20:26:24 +1030746 if (test_and_set_bit(req, &vcpu->requests))
Avi Kivityd9e368d2007-06-07 19:18:30 +0300747 continue;
748 cpu = vcpu->cpu;
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030749 if (cpus != NULL && cpu != -1 && cpu != me)
750 cpumask_set_cpu(cpu, cpus);
Avi Kivityd9e368d2007-06-07 19:18:30 +0300751 }
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030752 if (unlikely(cpus == NULL))
753 smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1);
754 else if (!cpumask_empty(cpus))
755 smp_call_function_many(cpus, ack_flush, NULL, 1);
756 else
757 called = false;
Marcelo Tosatti84261922009-06-17 10:53:47 -0300758 spin_unlock(&kvm->requests_lock);
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030759 free_cpumask_var(cpus);
Rusty Russell49846892008-12-08 20:26:24 +1030760 return called;
761}
762
763void kvm_flush_remote_tlbs(struct kvm *kvm)
764{
765 if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH))
766 ++kvm->stat.remote_tlb_flush;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300767}
768
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500769void kvm_reload_remote_mmus(struct kvm *kvm)
770{
Rusty Russell49846892008-12-08 20:26:24 +1030771 make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD);
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500772}
773
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000774int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id)
775{
776 struct page *page;
777 int r;
778
779 mutex_init(&vcpu->mutex);
780 vcpu->cpu = -1;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000781 vcpu->kvm = kvm;
782 vcpu->vcpu_id = id;
Eddie Dongb6958ce2007-07-18 12:15:21 +0300783 init_waitqueue_head(&vcpu->wq);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000784
785 page = alloc_page(GFP_KERNEL | __GFP_ZERO);
786 if (!page) {
787 r = -ENOMEM;
788 goto fail;
789 }
790 vcpu->run = page_address(page);
791
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800792 r = kvm_arch_vcpu_init(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000793 if (r < 0)
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800794 goto fail_free_run;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000795 return 0;
796
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000797fail_free_run:
798 free_page((unsigned long)vcpu->run);
799fail:
Rusty Russell76fafa52007-10-08 10:50:48 +1000800 return r;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000801}
802EXPORT_SYMBOL_GPL(kvm_vcpu_init);
803
804void kvm_vcpu_uninit(struct kvm_vcpu *vcpu)
805{
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800806 kvm_arch_vcpu_uninit(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000807 free_page((unsigned long)vcpu->run);
808}
809EXPORT_SYMBOL_GPL(kvm_vcpu_uninit);
810
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200811#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
812static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn)
813{
814 return container_of(mn, struct kvm, mmu_notifier);
815}
816
817static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn,
818 struct mm_struct *mm,
819 unsigned long address)
820{
821 struct kvm *kvm = mmu_notifier_to_kvm(mn);
822 int need_tlb_flush;
823
824 /*
825 * When ->invalidate_page runs, the linux pte has been zapped
826 * already but the page is still allocated until
827 * ->invalidate_page returns. So if we increase the sequence
828 * here the kvm page fault will notice if the spte can't be
829 * established because the page is going to be freed. If
830 * instead the kvm page fault establishes the spte before
831 * ->invalidate_page runs, kvm_unmap_hva will release it
832 * before returning.
833 *
834 * The sequence increase only need to be seen at spin_unlock
835 * time, and not at spin_lock time.
836 *
837 * Increasing the sequence after the spin_unlock would be
838 * unsafe because the kvm page fault could then establish the
839 * pte after kvm_unmap_hva returned, without noticing the page
840 * is going to be freed.
841 */
842 spin_lock(&kvm->mmu_lock);
843 kvm->mmu_notifier_seq++;
844 need_tlb_flush = kvm_unmap_hva(kvm, address);
845 spin_unlock(&kvm->mmu_lock);
846
847 /* we've to flush the tlb before the pages can be freed */
848 if (need_tlb_flush)
849 kvm_flush_remote_tlbs(kvm);
850
851}
852
853static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn,
854 struct mm_struct *mm,
855 unsigned long start,
856 unsigned long end)
857{
858 struct kvm *kvm = mmu_notifier_to_kvm(mn);
859 int need_tlb_flush = 0;
860
861 spin_lock(&kvm->mmu_lock);
862 /*
863 * The count increase must become visible at unlock time as no
864 * spte can be established without taking the mmu_lock and
865 * count is also read inside the mmu_lock critical section.
866 */
867 kvm->mmu_notifier_count++;
868 for (; start < end; start += PAGE_SIZE)
869 need_tlb_flush |= kvm_unmap_hva(kvm, start);
870 spin_unlock(&kvm->mmu_lock);
871
872 /* we've to flush the tlb before the pages can be freed */
873 if (need_tlb_flush)
874 kvm_flush_remote_tlbs(kvm);
875}
876
877static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
878 struct mm_struct *mm,
879 unsigned long start,
880 unsigned long end)
881{
882 struct kvm *kvm = mmu_notifier_to_kvm(mn);
883
884 spin_lock(&kvm->mmu_lock);
885 /*
886 * This sequence increase will notify the kvm page fault that
887 * the page that is going to be mapped in the spte could have
888 * been freed.
889 */
890 kvm->mmu_notifier_seq++;
891 /*
892 * The above sequence increase must be visible before the
893 * below count decrease but both values are read by the kvm
894 * page fault under mmu_lock spinlock so we don't need to add
895 * a smb_wmb() here in between the two.
896 */
897 kvm->mmu_notifier_count--;
898 spin_unlock(&kvm->mmu_lock);
899
900 BUG_ON(kvm->mmu_notifier_count < 0);
901}
902
903static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
904 struct mm_struct *mm,
905 unsigned long address)
906{
907 struct kvm *kvm = mmu_notifier_to_kvm(mn);
908 int young;
909
910 spin_lock(&kvm->mmu_lock);
911 young = kvm_age_hva(kvm, address);
912 spin_unlock(&kvm->mmu_lock);
913
914 if (young)
915 kvm_flush_remote_tlbs(kvm);
916
917 return young;
918}
919
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100920static void kvm_mmu_notifier_release(struct mmu_notifier *mn,
921 struct mm_struct *mm)
922{
923 struct kvm *kvm = mmu_notifier_to_kvm(mn);
924 kvm_arch_flush_shadow(kvm);
925}
926
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200927static const struct mmu_notifier_ops kvm_mmu_notifier_ops = {
928 .invalidate_page = kvm_mmu_notifier_invalidate_page,
929 .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start,
930 .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end,
931 .clear_flush_young = kvm_mmu_notifier_clear_flush_young,
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100932 .release = kvm_mmu_notifier_release,
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200933};
934#endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */
935
Avi Kivityf17abe92007-02-21 19:28:04 +0200936static struct kvm *kvm_create_vm(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800937{
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800938 struct kvm *kvm = kvm_arch_create_vm();
Laurent Vivier5f94c172008-05-30 16:05:54 +0200939#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
940 struct page *page;
941#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -0800942
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800943 if (IS_ERR(kvm))
944 goto out;
Avi Kivity75858a82009-01-04 17:10:50 +0200945#ifdef CONFIG_HAVE_KVM_IRQCHIP
Avi Kivity399ec802008-11-19 13:58:46 +0200946 INIT_LIST_HEAD(&kvm->irq_routing);
Avi Kivity75858a82009-01-04 17:10:50 +0200947 INIT_HLIST_HEAD(&kvm->mask_notifier_list);
948#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -0800949
Laurent Vivier5f94c172008-05-30 16:05:54 +0200950#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
951 page = alloc_page(GFP_KERNEL | __GFP_ZERO);
952 if (!page) {
953 kfree(kvm);
954 return ERR_PTR(-ENOMEM);
955 }
956 kvm->coalesced_mmio_ring =
957 (struct kvm_coalesced_mmio_ring *)page_address(page);
958#endif
959
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200960#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
961 {
962 int err;
963 kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops;
964 err = mmu_notifier_register(&kvm->mmu_notifier, current->mm);
965 if (err) {
966#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
967 put_page(page);
968#endif
969 kfree(kvm);
970 return ERR_PTR(err);
971 }
972 }
973#endif
974
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200975 kvm->mm = current->mm;
976 atomic_inc(&kvm->mm->mm_count);
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -0500977 spin_lock_init(&kvm->mmu_lock);
Marcelo Tosatti84261922009-06-17 10:53:47 -0300978 spin_lock_init(&kvm->requests_lock);
Eddie Dong74906342007-06-19 18:05:03 +0300979 kvm_io_bus_init(&kvm->pio_bus);
Gregory Haskinsd34e6b12009-07-07 17:08:49 -0400980 kvm_eventfd_init(kvm);
Shaohua Li11ec2802007-07-23 14:51:37 +0800981 mutex_init(&kvm->lock);
Marcelo Tosatti60eead72009-06-04 15:08:23 -0300982 mutex_init(&kvm->irq_lock);
Gregory Haskins2eeb2e92007-05-31 14:08:53 -0400983 kvm_io_bus_init(&kvm->mmio_bus);
Izik Eidus72dc67a2008-02-10 18:04:15 +0200984 init_rwsem(&kvm->slots_lock);
Izik Eidusd39f13b2008-03-30 16:01:25 +0300985 atomic_set(&kvm->users_count, 1);
Rusty Russell5e58cfe2007-07-23 17:08:21 +1000986 spin_lock(&kvm_lock);
987 list_add(&kvm->vm_list, &vm_list);
988 spin_unlock(&kvm_lock);
Laurent Vivier5f94c172008-05-30 16:05:54 +0200989#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
990 kvm_coalesced_mmio_init(kvm);
991#endif
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800992out:
Avi Kivityf17abe92007-02-21 19:28:04 +0200993 return kvm;
994}
995
Avi Kivity6aa8b732006-12-10 02:21:36 -0800996/*
997 * Free any memory in @free but not in @dont.
998 */
999static void kvm_free_physmem_slot(struct kvm_memory_slot *free,
1000 struct kvm_memory_slot *dont)
1001{
Joerg Roedelec04b262009-06-19 15:16:23 +02001002 int i;
1003
Izik Eidus290fc382007-09-27 14:11:22 +02001004 if (!dont || free->rmap != dont->rmap)
1005 vfree(free->rmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001006
1007 if (!dont || free->dirty_bitmap != dont->dirty_bitmap)
1008 vfree(free->dirty_bitmap);
1009
Joerg Roedelec04b262009-06-19 15:16:23 +02001010
1011 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) {
1012 if (!dont || free->lpage_info[i] != dont->lpage_info[i]) {
1013 vfree(free->lpage_info[i]);
1014 free->lpage_info[i] = NULL;
1015 }
1016 }
Marcelo Tosatti05da4552008-02-23 11:44:30 -03001017
Avi Kivity6aa8b732006-12-10 02:21:36 -08001018 free->npages = 0;
Al Viro8b6d44c2007-02-09 16:38:40 +00001019 free->dirty_bitmap = NULL;
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001020 free->rmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001021}
1022
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +08001023void kvm_free_physmem(struct kvm *kvm)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001024{
1025 int i;
1026
1027 for (i = 0; i < kvm->nmemslots; ++i)
Al Viro8b6d44c2007-02-09 16:38:40 +00001028 kvm_free_physmem_slot(&kvm->memslots[i], NULL);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001029}
1030
Avi Kivityf17abe92007-02-21 19:28:04 +02001031static void kvm_destroy_vm(struct kvm *kvm)
1032{
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001033 struct mm_struct *mm = kvm->mm;
1034
Sheng Yangad8ba2c2009-01-06 10:03:02 +08001035 kvm_arch_sync_events(kvm);
Avi Kivity133de902007-02-12 00:54:44 -08001036 spin_lock(&kvm_lock);
1037 list_del(&kvm->vm_list);
1038 spin_unlock(&kvm_lock);
Avi Kivity399ec802008-11-19 13:58:46 +02001039 kvm_free_irq_routing(kvm);
Eddie Dong74906342007-06-19 18:05:03 +03001040 kvm_io_bus_destroy(&kvm->pio_bus);
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04001041 kvm_io_bus_destroy(&kvm->mmio_bus);
Laurent Vivier5f94c172008-05-30 16:05:54 +02001042#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1043 if (kvm->coalesced_mmio_ring != NULL)
1044 free_page((unsigned long)kvm->coalesced_mmio_ring);
1045#endif
Andrea Arcangelie930bff2008-07-25 16:24:52 +02001046#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1047 mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
Gleb Natapovf00be0c2009-03-19 12:20:36 +02001048#else
1049 kvm_arch_flush_shadow(kvm);
Andrea Arcangelie930bff2008-07-25 16:24:52 +02001050#endif
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +08001051 kvm_arch_destroy_vm(kvm);
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001052 mmdrop(mm);
Avi Kivityf17abe92007-02-21 19:28:04 +02001053}
1054
Izik Eidusd39f13b2008-03-30 16:01:25 +03001055void kvm_get_kvm(struct kvm *kvm)
1056{
1057 atomic_inc(&kvm->users_count);
1058}
1059EXPORT_SYMBOL_GPL(kvm_get_kvm);
1060
1061void kvm_put_kvm(struct kvm *kvm)
1062{
1063 if (atomic_dec_and_test(&kvm->users_count))
1064 kvm_destroy_vm(kvm);
1065}
1066EXPORT_SYMBOL_GPL(kvm_put_kvm);
1067
1068
Avi Kivityf17abe92007-02-21 19:28:04 +02001069static int kvm_vm_release(struct inode *inode, struct file *filp)
1070{
1071 struct kvm *kvm = filp->private_data;
1072
Gregory Haskins721eecb2009-05-20 10:30:49 -04001073 kvm_irqfd_release(kvm);
1074
Izik Eidusd39f13b2008-03-30 16:01:25 +03001075 kvm_put_kvm(kvm);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001076 return 0;
1077}
1078
Avi Kivity6aa8b732006-12-10 02:21:36 -08001079/*
Avi Kivity6aa8b732006-12-10 02:21:36 -08001080 * Allocate some memory and give it an address in the guest physical address
1081 * space.
1082 *
1083 * Discontiguous memory is allowed, mostly for framebuffers.
Sheng Yangf78e0e22007-10-29 09:40:42 +08001084 *
Marcelo Tosatti10589a42007-12-20 19:18:22 -05001085 * Must be called holding mmap_sem for write.
Avi Kivity6aa8b732006-12-10 02:21:36 -08001086 */
Sheng Yangf78e0e22007-10-29 09:40:42 +08001087int __kvm_set_memory_region(struct kvm *kvm,
1088 struct kvm_userspace_memory_region *mem,
1089 int user_alloc)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001090{
1091 int r;
1092 gfn_t base_gfn;
Heiko Carstens28bcb112009-09-03 17:35:35 +02001093 unsigned long npages;
1094 unsigned long i;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001095 struct kvm_memory_slot *memslot;
1096 struct kvm_memory_slot old, new;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001097
1098 r = -EINVAL;
1099 /* General sanity checks */
1100 if (mem->memory_size & (PAGE_SIZE - 1))
1101 goto out;
1102 if (mem->guest_phys_addr & (PAGE_SIZE - 1))
1103 goto out;
Sheng Yange7cacd42008-11-11 15:30:40 +08001104 if (user_alloc && (mem->userspace_addr & (PAGE_SIZE - 1)))
Hollis Blanchard78749802008-11-07 13:32:12 -06001105 goto out;
Izik Eiduse0d62c72007-10-24 23:57:46 +02001106 if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001107 goto out;
1108 if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
1109 goto out;
1110
1111 memslot = &kvm->memslots[mem->slot];
1112 base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
1113 npages = mem->memory_size >> PAGE_SHIFT;
1114
1115 if (!npages)
1116 mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES;
1117
Avi Kivity6aa8b732006-12-10 02:21:36 -08001118 new = old = *memslot;
1119
1120 new.base_gfn = base_gfn;
1121 new.npages = npages;
1122 new.flags = mem->flags;
1123
1124 /* Disallow changing a memory slot's size. */
1125 r = -EINVAL;
1126 if (npages && old.npages && npages != old.npages)
Sheng Yangf78e0e22007-10-29 09:40:42 +08001127 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001128
1129 /* Check for overlaps */
1130 r = -EEXIST;
1131 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
1132 struct kvm_memory_slot *s = &kvm->memslots[i];
1133
Jan Kiszka4cd481f2009-04-13 11:59:32 +02001134 if (s == memslot || !s->npages)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001135 continue;
1136 if (!((base_gfn + npages <= s->base_gfn) ||
1137 (base_gfn >= s->base_gfn + s->npages)))
Sheng Yangf78e0e22007-10-29 09:40:42 +08001138 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001139 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08001140
Avi Kivity6aa8b732006-12-10 02:21:36 -08001141 /* Free page dirty bitmap if unneeded */
1142 if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
Al Viro8b6d44c2007-02-09 16:38:40 +00001143 new.dirty_bitmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001144
1145 r = -ENOMEM;
1146
1147 /* Allocate if a slot is being created */
Carsten Otteeff01142008-06-27 15:05:31 +02001148#ifndef CONFIG_S390
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001149 if (npages && !new.rmap) {
Mike Dayd77c26f2007-10-08 09:02:08 -04001150 new.rmap = vmalloc(npages * sizeof(struct page *));
Izik Eidus290fc382007-09-27 14:11:22 +02001151
1152 if (!new.rmap)
Sheng Yangf78e0e22007-10-29 09:40:42 +08001153 goto out_free;
Izik Eidus290fc382007-09-27 14:11:22 +02001154
Izik Eidus290fc382007-09-27 14:11:22 +02001155 memset(new.rmap, 0, npages * sizeof(*new.rmap));
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001156
Izik Eidus80b14b52007-10-25 11:54:04 +02001157 new.user_alloc = user_alloc;
Andrea Arcangeli604b38a2008-07-25 16:32:03 +02001158 /*
1159 * hva_to_rmmap() serialzies with the mmu_lock and to be
1160 * safe it has to ignore memslots with !user_alloc &&
1161 * !userspace_addr.
1162 */
1163 if (user_alloc)
1164 new.userspace_addr = mem->userspace_addr;
1165 else
1166 new.userspace_addr = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001167 }
Joerg Roedelec04b262009-06-19 15:16:23 +02001168 if (!npages)
1169 goto skip_lpage;
Marcelo Tosatti05da4552008-02-23 11:44:30 -03001170
Joerg Roedelec04b262009-06-19 15:16:23 +02001171 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) {
Heiko Carstens28bcb112009-09-03 17:35:35 +02001172 unsigned long ugfn;
1173 unsigned long j;
1174 int lpages;
Joerg Roedelec04b262009-06-19 15:16:23 +02001175 int level = i + 2;
Marcelo Tosatti05da4552008-02-23 11:44:30 -03001176
Joerg Roedelec04b262009-06-19 15:16:23 +02001177 /* Avoid unused variable warning if no large pages */
1178 (void)level;
1179
1180 if (new.lpage_info[i])
1181 continue;
1182
1183 lpages = 1 + (base_gfn + npages - 1) /
1184 KVM_PAGES_PER_HPAGE(level);
1185 lpages -= base_gfn / KVM_PAGES_PER_HPAGE(level);
1186
1187 new.lpage_info[i] = vmalloc(lpages * sizeof(*new.lpage_info[i]));
1188
1189 if (!new.lpage_info[i])
Marcelo Tosatti05da4552008-02-23 11:44:30 -03001190 goto out_free;
1191
Joerg Roedelec04b262009-06-19 15:16:23 +02001192 memset(new.lpage_info[i], 0,
1193 lpages * sizeof(*new.lpage_info[i]));
Marcelo Tosatti05da4552008-02-23 11:44:30 -03001194
Joerg Roedelec04b262009-06-19 15:16:23 +02001195 if (base_gfn % KVM_PAGES_PER_HPAGE(level))
1196 new.lpage_info[i][0].write_count = 1;
1197 if ((base_gfn+npages) % KVM_PAGES_PER_HPAGE(level))
1198 new.lpage_info[i][lpages - 1].write_count = 1;
Avi Kivityac045272009-06-08 15:52:39 +03001199 ugfn = new.userspace_addr >> PAGE_SHIFT;
1200 /*
1201 * If the gfn and userspace address are not aligned wrt each
Marcelo Tosatti54dee992009-06-11 12:07:44 -03001202 * other, or if explicitly asked to, disable large page
1203 * support for this slot
Avi Kivityac045272009-06-08 15:52:39 +03001204 */
Joerg Roedelec04b262009-06-19 15:16:23 +02001205 if ((base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1) ||
Marcelo Tosatti54dee992009-06-11 12:07:44 -03001206 !largepages_enabled)
Joerg Roedelec04b262009-06-19 15:16:23 +02001207 for (j = 0; j < lpages; ++j)
1208 new.lpage_info[i][j].write_count = 1;
Marcelo Tosatti05da4552008-02-23 11:44:30 -03001209 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08001210
Joerg Roedelec04b262009-06-19 15:16:23 +02001211skip_lpage:
1212
Avi Kivity6aa8b732006-12-10 02:21:36 -08001213 /* Allocate page dirty bitmap if needed */
1214 if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
1215 unsigned dirty_bytes = ALIGN(npages, BITS_PER_LONG) / 8;
1216
1217 new.dirty_bitmap = vmalloc(dirty_bytes);
1218 if (!new.dirty_bitmap)
Sheng Yangf78e0e22007-10-29 09:40:42 +08001219 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001220 memset(new.dirty_bitmap, 0, dirty_bytes);
Izik Eiduse2445842009-06-10 19:23:24 +03001221 if (old.npages)
1222 kvm_arch_flush_shadow(kvm);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001223 }
Christian Borntraeger3eea8432009-06-23 17:24:06 +02001224#else /* not defined CONFIG_S390 */
1225 new.user_alloc = user_alloc;
1226 if (user_alloc)
1227 new.userspace_addr = mem->userspace_addr;
Carsten Otteeff01142008-06-27 15:05:31 +02001228#endif /* not defined CONFIG_S390 */
Avi Kivity6aa8b732006-12-10 02:21:36 -08001229
Marcelo Tosatti34d4cb82008-07-10 20:49:31 -03001230 if (!npages)
1231 kvm_arch_flush_shadow(kvm);
1232
Andrea Arcangeli604b38a2008-07-25 16:32:03 +02001233 spin_lock(&kvm->mmu_lock);
1234 if (mem->slot >= kvm->nmemslots)
1235 kvm->nmemslots = mem->slot + 1;
1236
Avi Kivity6aa8b732006-12-10 02:21:36 -08001237 *memslot = new;
Andrea Arcangeli604b38a2008-07-25 16:32:03 +02001238 spin_unlock(&kvm->mmu_lock);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001239
Zhang Xiantao0de10342007-11-20 16:25:04 +08001240 r = kvm_arch_set_memory_region(kvm, mem, old, user_alloc);
1241 if (r) {
Andrea Arcangeli604b38a2008-07-25 16:32:03 +02001242 spin_lock(&kvm->mmu_lock);
Zhang Xiantao0de10342007-11-20 16:25:04 +08001243 *memslot = old;
Andrea Arcangeli604b38a2008-07-25 16:32:03 +02001244 spin_unlock(&kvm->mmu_lock);
Zhang Xiantao0de10342007-11-20 16:25:04 +08001245 goto out_free;
Zhang Xiantao3ad82a72007-11-20 13:11:38 +08001246 }
1247
Glauber Costa6f897242008-12-03 13:40:51 -02001248 kvm_free_physmem_slot(&old, npages ? &new : NULL);
1249 /* Slot deletion case: we have to update the current slot */
Marcelo Tosattib43b1902009-05-12 18:55:44 -03001250 spin_lock(&kvm->mmu_lock);
Glauber Costa6f897242008-12-03 13:40:51 -02001251 if (!npages)
1252 *memslot = old;
Marcelo Tosattib43b1902009-05-12 18:55:44 -03001253 spin_unlock(&kvm->mmu_lock);
Xiantao Zhang8a98f662008-10-06 13:47:38 +08001254#ifdef CONFIG_DMAR
Ben-Ami Yassour62c476c2008-09-14 03:48:28 +03001255 /* map the pages in iommu page table */
1256 r = kvm_iommu_map_pages(kvm, base_gfn, npages);
1257 if (r)
1258 goto out;
Xiantao Zhang8a98f662008-10-06 13:47:38 +08001259#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -08001260 return 0;
1261
Sheng Yangf78e0e22007-10-29 09:40:42 +08001262out_free:
Avi Kivity6aa8b732006-12-10 02:21:36 -08001263 kvm_free_physmem_slot(&new, &old);
1264out:
1265 return r;
Izik Eidus210c7c42007-10-24 23:52:57 +02001266
1267}
Sheng Yangf78e0e22007-10-29 09:40:42 +08001268EXPORT_SYMBOL_GPL(__kvm_set_memory_region);
1269
1270int kvm_set_memory_region(struct kvm *kvm,
1271 struct kvm_userspace_memory_region *mem,
1272 int user_alloc)
1273{
1274 int r;
1275
Izik Eidus72dc67a2008-02-10 18:04:15 +02001276 down_write(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +08001277 r = __kvm_set_memory_region(kvm, mem, user_alloc);
Izik Eidus72dc67a2008-02-10 18:04:15 +02001278 up_write(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +08001279 return r;
1280}
Izik Eidus210c7c42007-10-24 23:52:57 +02001281EXPORT_SYMBOL_GPL(kvm_set_memory_region);
1282
Carsten Otte1fe779f2007-10-29 16:08:35 +01001283int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
1284 struct
1285 kvm_userspace_memory_region *mem,
1286 int user_alloc)
Izik Eidus210c7c42007-10-24 23:52:57 +02001287{
Izik Eiduse0d62c72007-10-24 23:57:46 +02001288 if (mem->slot >= KVM_MEMORY_SLOTS)
1289 return -EINVAL;
Izik Eidus210c7c42007-10-24 23:52:57 +02001290 return kvm_set_memory_region(kvm, mem, user_alloc);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001291}
1292
Zhang Xiantao5bb064d2007-11-18 20:29:43 +08001293int kvm_get_dirty_log(struct kvm *kvm,
1294 struct kvm_dirty_log *log, int *is_dirty)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001295{
1296 struct kvm_memory_slot *memslot;
1297 int r, i;
1298 int n;
1299 unsigned long any = 0;
1300
Avi Kivity6aa8b732006-12-10 02:21:36 -08001301 r = -EINVAL;
1302 if (log->slot >= KVM_MEMORY_SLOTS)
1303 goto out;
1304
1305 memslot = &kvm->memslots[log->slot];
1306 r = -ENOENT;
1307 if (!memslot->dirty_bitmap)
1308 goto out;
1309
Uri Lublincd1a4a92007-02-22 16:43:09 +02001310 n = ALIGN(memslot->npages, BITS_PER_LONG) / 8;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001311
Uri Lublincd1a4a92007-02-22 16:43:09 +02001312 for (i = 0; !any && i < n/sizeof(long); ++i)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001313 any = memslot->dirty_bitmap[i];
1314
1315 r = -EFAULT;
1316 if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
1317 goto out;
1318
Zhang Xiantao5bb064d2007-11-18 20:29:43 +08001319 if (any)
1320 *is_dirty = 1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001321
1322 r = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001323out:
Avi Kivity6aa8b732006-12-10 02:21:36 -08001324 return r;
1325}
1326
Marcelo Tosatti54dee992009-06-11 12:07:44 -03001327void kvm_disable_largepages(void)
1328{
1329 largepages_enabled = false;
1330}
1331EXPORT_SYMBOL_GPL(kvm_disable_largepages);
1332
Izik Eiduscea7bb22007-10-17 19:17:48 +02001333int is_error_page(struct page *page)
1334{
1335 return page == bad_page;
1336}
1337EXPORT_SYMBOL_GPL(is_error_page);
1338
Anthony Liguori35149e22008-04-02 14:46:56 -05001339int is_error_pfn(pfn_t pfn)
1340{
1341 return pfn == bad_pfn;
1342}
1343EXPORT_SYMBOL_GPL(is_error_pfn);
1344
Izik Eidusf9d46eb2007-11-11 22:02:22 +02001345static inline unsigned long bad_hva(void)
1346{
1347 return PAGE_OFFSET;
1348}
1349
1350int kvm_is_error_hva(unsigned long addr)
1351{
1352 return addr == bad_hva();
1353}
1354EXPORT_SYMBOL_GPL(kvm_is_error_hva);
1355
Izik Eidus28430992008-10-03 17:40:32 +03001356struct kvm_memory_slot *gfn_to_memslot_unaliased(struct kvm *kvm, gfn_t gfn)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001357{
1358 int i;
1359
1360 for (i = 0; i < kvm->nmemslots; ++i) {
1361 struct kvm_memory_slot *memslot = &kvm->memslots[i];
1362
1363 if (gfn >= memslot->base_gfn
1364 && gfn < memslot->base_gfn + memslot->npages)
1365 return memslot;
1366 }
Al Viro8b6d44c2007-02-09 16:38:40 +00001367 return NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001368}
Izik Eidus28430992008-10-03 17:40:32 +03001369EXPORT_SYMBOL_GPL(gfn_to_memslot_unaliased);
Avi Kivitye8207542007-03-30 16:54:30 +03001370
1371struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn)
1372{
1373 gfn = unalias_gfn(kvm, gfn);
Izik Eidus28430992008-10-03 17:40:32 +03001374 return gfn_to_memslot_unaliased(kvm, gfn);
Avi Kivitye8207542007-03-30 16:54:30 +03001375}
Avi Kivity6aa8b732006-12-10 02:21:36 -08001376
Izik Eiduse0d62c72007-10-24 23:57:46 +02001377int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
1378{
1379 int i;
1380
1381 gfn = unalias_gfn(kvm, gfn);
1382 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
1383 struct kvm_memory_slot *memslot = &kvm->memslots[i];
1384
1385 if (gfn >= memslot->base_gfn
1386 && gfn < memslot->base_gfn + memslot->npages)
1387 return 1;
1388 }
1389 return 0;
1390}
1391EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);
1392
Marcelo Tosatti05da4552008-02-23 11:44:30 -03001393unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
Izik Eidus539cb662007-11-11 22:05:04 +02001394{
1395 struct kvm_memory_slot *slot;
1396
1397 gfn = unalias_gfn(kvm, gfn);
Izik Eidus28430992008-10-03 17:40:32 +03001398 slot = gfn_to_memslot_unaliased(kvm, gfn);
Izik Eidus539cb662007-11-11 22:05:04 +02001399 if (!slot)
1400 return bad_hva();
1401 return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE);
1402}
Sheng Yang0d150292008-04-25 21:44:50 +08001403EXPORT_SYMBOL_GPL(gfn_to_hva);
Izik Eidus539cb662007-11-11 22:05:04 +02001404
Anthony Liguori35149e22008-04-02 14:46:56 -05001405pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
Avi Kivity954bbbc22007-03-30 14:02:32 +03001406{
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001407 struct page *page[1];
Izik Eidus539cb662007-11-11 22:05:04 +02001408 unsigned long addr;
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001409 int npages;
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001410 pfn_t pfn;
Avi Kivity954bbbc22007-03-30 14:02:32 +03001411
Avi Kivity60395222007-10-21 11:03:36 +02001412 might_sleep();
1413
Izik Eidus539cb662007-11-11 22:05:04 +02001414 addr = gfn_to_hva(kvm, gfn);
1415 if (kvm_is_error_hva(addr)) {
Izik Eidus8a7ae052007-10-18 11:09:33 +02001416 get_page(bad_page);
Anthony Liguori35149e22008-04-02 14:46:56 -05001417 return page_to_pfn(bad_page);
Izik Eidus8a7ae052007-10-18 11:09:33 +02001418 }
Izik Eidus8a7ae052007-10-18 11:09:33 +02001419
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001420 npages = get_user_pages_fast(addr, 1, 1, page);
Izik Eidus539cb662007-11-11 22:05:04 +02001421
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001422 if (unlikely(npages != 1)) {
1423 struct vm_area_struct *vma;
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001424
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001425 down_read(&current->mm->mmap_sem);
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001426 vma = find_vma(current->mm, addr);
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001427
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001428 if (vma == NULL || addr < vma->vm_start ||
1429 !(vma->vm_flags & VM_PFNMAP)) {
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001430 up_read(&current->mm->mmap_sem);
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001431 get_page(bad_page);
1432 return page_to_pfn(bad_page);
1433 }
1434
1435 pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001436 up_read(&current->mm->mmap_sem);
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001437 BUG_ON(!kvm_is_mmio_pfn(pfn));
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001438 } else
1439 pfn = page_to_pfn(page[0]);
1440
1441 return pfn;
Anthony Liguori35149e22008-04-02 14:46:56 -05001442}
1443
1444EXPORT_SYMBOL_GPL(gfn_to_pfn);
1445
1446struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
1447{
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001448 pfn_t pfn;
1449
1450 pfn = gfn_to_pfn(kvm, gfn);
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001451 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001452 return pfn_to_page(pfn);
1453
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001454 WARN_ON(kvm_is_mmio_pfn(pfn));
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001455
1456 get_page(bad_page);
1457 return bad_page;
Avi Kivity954bbbc22007-03-30 14:02:32 +03001458}
Anthony Liguoriaab61cc2007-10-29 15:15:20 -05001459
Avi Kivity954bbbc22007-03-30 14:02:32 +03001460EXPORT_SYMBOL_GPL(gfn_to_page);
1461
Izik Eidusb4231d62007-11-20 11:49:33 +02001462void kvm_release_page_clean(struct page *page)
1463{
Anthony Liguori35149e22008-04-02 14:46:56 -05001464 kvm_release_pfn_clean(page_to_pfn(page));
Izik Eidusb4231d62007-11-20 11:49:33 +02001465}
1466EXPORT_SYMBOL_GPL(kvm_release_page_clean);
1467
Anthony Liguori35149e22008-04-02 14:46:56 -05001468void kvm_release_pfn_clean(pfn_t pfn)
1469{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001470 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001471 put_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001472}
1473EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);
1474
Izik Eidusb4231d62007-11-20 11:49:33 +02001475void kvm_release_page_dirty(struct page *page)
Izik Eidus8a7ae052007-10-18 11:09:33 +02001476{
Anthony Liguori35149e22008-04-02 14:46:56 -05001477 kvm_release_pfn_dirty(page_to_pfn(page));
Izik Eidus8a7ae052007-10-18 11:09:33 +02001478}
Izik Eidusb4231d62007-11-20 11:49:33 +02001479EXPORT_SYMBOL_GPL(kvm_release_page_dirty);
Izik Eidus8a7ae052007-10-18 11:09:33 +02001480
Anthony Liguori35149e22008-04-02 14:46:56 -05001481void kvm_release_pfn_dirty(pfn_t pfn)
1482{
1483 kvm_set_pfn_dirty(pfn);
1484 kvm_release_pfn_clean(pfn);
1485}
1486EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
1487
1488void kvm_set_page_dirty(struct page *page)
1489{
1490 kvm_set_pfn_dirty(page_to_pfn(page));
1491}
1492EXPORT_SYMBOL_GPL(kvm_set_page_dirty);
1493
1494void kvm_set_pfn_dirty(pfn_t pfn)
1495{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001496 if (!kvm_is_mmio_pfn(pfn)) {
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001497 struct page *page = pfn_to_page(pfn);
1498 if (!PageReserved(page))
1499 SetPageDirty(page);
1500 }
Anthony Liguori35149e22008-04-02 14:46:56 -05001501}
1502EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);
1503
1504void kvm_set_pfn_accessed(pfn_t pfn)
1505{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001506 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001507 mark_page_accessed(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001508}
1509EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);
1510
1511void kvm_get_pfn(pfn_t pfn)
1512{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001513 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001514 get_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001515}
1516EXPORT_SYMBOL_GPL(kvm_get_pfn);
1517
Izik Eidus195aefd2007-10-01 22:14:18 +02001518static int next_segment(unsigned long len, int offset)
1519{
1520 if (len > PAGE_SIZE - offset)
1521 return PAGE_SIZE - offset;
1522 else
1523 return len;
1524}
1525
1526int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
1527 int len)
1528{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001529 int r;
1530 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001531
Izik Eiduse0506bc2007-11-11 22:10:22 +02001532 addr = gfn_to_hva(kvm, gfn);
1533 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001534 return -EFAULT;
Izik Eiduse0506bc2007-11-11 22:10:22 +02001535 r = copy_from_user(data, (void __user *)addr + offset, len);
1536 if (r)
1537 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001538 return 0;
1539}
1540EXPORT_SYMBOL_GPL(kvm_read_guest_page);
1541
1542int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len)
1543{
1544 gfn_t gfn = gpa >> PAGE_SHIFT;
1545 int seg;
1546 int offset = offset_in_page(gpa);
1547 int ret;
1548
1549 while ((seg = next_segment(len, offset)) != 0) {
1550 ret = kvm_read_guest_page(kvm, gfn, data, offset, seg);
1551 if (ret < 0)
1552 return ret;
1553 offset = 0;
1554 len -= seg;
1555 data += seg;
1556 ++gfn;
1557 }
1558 return 0;
1559}
1560EXPORT_SYMBOL_GPL(kvm_read_guest);
1561
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001562int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
1563 unsigned long len)
1564{
1565 int r;
1566 unsigned long addr;
1567 gfn_t gfn = gpa >> PAGE_SHIFT;
1568 int offset = offset_in_page(gpa);
1569
1570 addr = gfn_to_hva(kvm, gfn);
1571 if (kvm_is_error_hva(addr))
1572 return -EFAULT;
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001573 pagefault_disable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001574 r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001575 pagefault_enable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001576 if (r)
1577 return -EFAULT;
1578 return 0;
1579}
1580EXPORT_SYMBOL(kvm_read_guest_atomic);
1581
Izik Eidus195aefd2007-10-01 22:14:18 +02001582int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
1583 int offset, int len)
1584{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001585 int r;
1586 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001587
Izik Eiduse0506bc2007-11-11 22:10:22 +02001588 addr = gfn_to_hva(kvm, gfn);
1589 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001590 return -EFAULT;
Izik Eiduse0506bc2007-11-11 22:10:22 +02001591 r = copy_to_user((void __user *)addr + offset, data, len);
1592 if (r)
1593 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001594 mark_page_dirty(kvm, gfn);
1595 return 0;
1596}
1597EXPORT_SYMBOL_GPL(kvm_write_guest_page);
1598
1599int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
1600 unsigned long len)
1601{
1602 gfn_t gfn = gpa >> PAGE_SHIFT;
1603 int seg;
1604 int offset = offset_in_page(gpa);
1605 int ret;
1606
1607 while ((seg = next_segment(len, offset)) != 0) {
1608 ret = kvm_write_guest_page(kvm, gfn, data, offset, seg);
1609 if (ret < 0)
1610 return ret;
1611 offset = 0;
1612 len -= seg;
1613 data += seg;
1614 ++gfn;
1615 }
1616 return 0;
1617}
1618
1619int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
1620{
Izik Eidus3e021bf2007-11-19 11:16:57 +02001621 return kvm_write_guest_page(kvm, gfn, empty_zero_page, offset, len);
Izik Eidus195aefd2007-10-01 22:14:18 +02001622}
1623EXPORT_SYMBOL_GPL(kvm_clear_guest_page);
1624
1625int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len)
1626{
1627 gfn_t gfn = gpa >> PAGE_SHIFT;
1628 int seg;
1629 int offset = offset_in_page(gpa);
1630 int ret;
1631
1632 while ((seg = next_segment(len, offset)) != 0) {
1633 ret = kvm_clear_guest_page(kvm, gfn, offset, seg);
1634 if (ret < 0)
1635 return ret;
1636 offset = 0;
1637 len -= seg;
1638 ++gfn;
1639 }
1640 return 0;
1641}
1642EXPORT_SYMBOL_GPL(kvm_clear_guest);
1643
Avi Kivity6aa8b732006-12-10 02:21:36 -08001644void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
1645{
Nguyen Anh Quynh31389942007-06-05 10:35:19 +03001646 struct kvm_memory_slot *memslot;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001647
Uri Lublin3b6fff12007-10-30 10:42:09 +02001648 gfn = unalias_gfn(kvm, gfn);
Izik Eidus28430992008-10-03 17:40:32 +03001649 memslot = gfn_to_memslot_unaliased(kvm, gfn);
Rusty Russell7e9d6192007-07-31 20:41:14 +10001650 if (memslot && memslot->dirty_bitmap) {
1651 unsigned long rel_gfn = gfn - memslot->base_gfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001652
Rusty Russell7e9d6192007-07-31 20:41:14 +10001653 /* avoid RMW */
1654 if (!test_bit(rel_gfn, memslot->dirty_bitmap))
1655 set_bit(rel_gfn, memslot->dirty_bitmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001656 }
1657}
1658
Eddie Dongb6958ce2007-07-18 12:15:21 +03001659/*
1660 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
1661 */
Hollis Blanchard8776e512007-10-31 17:24:24 -05001662void kvm_vcpu_block(struct kvm_vcpu *vcpu)
Eddie Dongb6958ce2007-07-18 12:15:21 +03001663{
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001664 DEFINE_WAIT(wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001665
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001666 for (;;) {
1667 prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001668
Gleb Natapova1b37102009-07-09 15:33:52 +03001669 if (kvm_arch_vcpu_runnable(vcpu)) {
Marcelo Tosattid7690172008-09-08 15:23:48 -03001670 set_bit(KVM_REQ_UNHALT, &vcpu->requests);
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001671 break;
Marcelo Tosattid7690172008-09-08 15:23:48 -03001672 }
Gleb Natapov09cec752009-03-23 15:11:44 +02001673 if (kvm_cpu_has_pending_timer(vcpu))
1674 break;
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001675 if (signal_pending(current))
1676 break;
1677
Eddie Dongb6958ce2007-07-18 12:15:21 +03001678 vcpu_put(vcpu);
1679 schedule();
1680 vcpu_load(vcpu);
1681 }
1682
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001683 finish_wait(&vcpu->wq, &wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001684}
1685
Avi Kivity6aa8b732006-12-10 02:21:36 -08001686void kvm_resched(struct kvm_vcpu *vcpu)
1687{
Yaozu Dong3fca0362007-04-25 16:49:19 +03001688 if (!need_resched())
1689 return;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001690 cond_resched();
Avi Kivity6aa8b732006-12-10 02:21:36 -08001691}
1692EXPORT_SYMBOL_GPL(kvm_resched);
1693
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001694static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001695{
1696 struct kvm_vcpu *vcpu = vma->vm_file->private_data;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001697 struct page *page;
1698
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001699 if (vmf->pgoff == 0)
Avi Kivity039576c2007-03-20 12:46:50 +02001700 page = virt_to_page(vcpu->run);
Avi Kivity09566762008-01-23 18:14:23 +02001701#ifdef CONFIG_X86
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001702 else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
Zhang Xiantaoad312c72007-12-13 23:50:52 +08001703 page = virt_to_page(vcpu->arch.pio_data);
Avi Kivity09566762008-01-23 18:14:23 +02001704#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02001705#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1706 else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
1707 page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
1708#endif
Avi Kivity039576c2007-03-20 12:46:50 +02001709 else
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001710 return VM_FAULT_SIGBUS;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001711 get_page(page);
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001712 vmf->page = page;
1713 return 0;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001714}
1715
1716static struct vm_operations_struct kvm_vcpu_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001717 .fault = kvm_vcpu_fault,
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001718};
1719
1720static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma)
1721{
1722 vma->vm_ops = &kvm_vcpu_vm_ops;
1723 return 0;
1724}
1725
Avi Kivitybccf2152007-02-21 18:04:26 +02001726static int kvm_vcpu_release(struct inode *inode, struct file *filp)
1727{
1728 struct kvm_vcpu *vcpu = filp->private_data;
1729
Al Viro66c0b392008-04-19 20:33:56 +01001730 kvm_put_kvm(vcpu->kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001731 return 0;
1732}
1733
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01001734static struct file_operations kvm_vcpu_fops = {
Avi Kivitybccf2152007-02-21 18:04:26 +02001735 .release = kvm_vcpu_release,
1736 .unlocked_ioctl = kvm_vcpu_ioctl,
1737 .compat_ioctl = kvm_vcpu_ioctl,
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001738 .mmap = kvm_vcpu_mmap,
Avi Kivitybccf2152007-02-21 18:04:26 +02001739};
1740
1741/*
1742 * Allocates an inode for the vcpu.
1743 */
1744static int create_vcpu_fd(struct kvm_vcpu *vcpu)
1745{
Gleb Natapov73880c82009-06-09 15:56:28 +03001746 return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, 0);
Avi Kivitybccf2152007-02-21 18:04:26 +02001747}
1748
Avi Kivityc5ea7662007-02-20 18:41:05 +02001749/*
1750 * Creates some virtual cpus. Good luck creating more than one.
1751 */
Gleb Natapov73880c82009-06-09 15:56:28 +03001752static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
Avi Kivityc5ea7662007-02-20 18:41:05 +02001753{
1754 int r;
Gleb Natapov988a2ca2009-06-09 15:56:29 +03001755 struct kvm_vcpu *vcpu, *v;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001756
Gleb Natapov73880c82009-06-09 15:56:28 +03001757 vcpu = kvm_arch_vcpu_create(kvm, id);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001758 if (IS_ERR(vcpu))
1759 return PTR_ERR(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001760
Avi Kivity15ad7142007-07-11 18:17:21 +03001761 preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);
1762
Avi Kivity26e52152007-11-20 15:30:24 +02001763 r = kvm_arch_vcpu_setup(vcpu);
1764 if (r)
Glauber Costa7d8fece2008-09-17 23:16:59 -03001765 return r;
Avi Kivity26e52152007-11-20 15:30:24 +02001766
Shaohua Li11ec2802007-07-23 14:51:37 +08001767 mutex_lock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03001768 if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) {
1769 r = -EINVAL;
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001770 goto vcpu_destroy;
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001771 }
Gleb Natapov73880c82009-06-09 15:56:28 +03001772
Gleb Natapov988a2ca2009-06-09 15:56:29 +03001773 kvm_for_each_vcpu(r, v, kvm)
1774 if (v->vcpu_id == id) {
Gleb Natapov73880c82009-06-09 15:56:28 +03001775 r = -EEXIST;
1776 goto vcpu_destroy;
1777 }
1778
1779 BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001780
1781 /* Now it's all set up, let userspace reach it */
Al Viro66c0b392008-04-19 20:33:56 +01001782 kvm_get_kvm(kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001783 r = create_vcpu_fd(vcpu);
Gleb Natapov73880c82009-06-09 15:56:28 +03001784 if (r < 0) {
1785 kvm_put_kvm(kvm);
1786 goto vcpu_destroy;
1787 }
1788
1789 kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu;
1790 smp_wmb();
1791 atomic_inc(&kvm->online_vcpus);
1792
1793#ifdef CONFIG_KVM_APIC_ARCHITECTURE
1794 if (kvm->bsp_vcpu_id == id)
1795 kvm->bsp_vcpu = vcpu;
1796#endif
1797 mutex_unlock(&kvm->lock);
Avi Kivitybccf2152007-02-21 18:04:26 +02001798 return r;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001799
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001800vcpu_destroy:
Glauber Costa7d8fece2008-09-17 23:16:59 -03001801 mutex_unlock(&kvm->lock);
Hollis Blanchardd40ccc62007-11-19 14:04:43 -06001802 kvm_arch_vcpu_destroy(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001803 return r;
1804}
1805
Avi Kivity1961d272007-03-05 19:46:05 +02001806static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset)
1807{
1808 if (sigset) {
1809 sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP));
1810 vcpu->sigset_active = 1;
1811 vcpu->sigset = *sigset;
1812 } else
1813 vcpu->sigset_active = 0;
1814 return 0;
1815}
1816
Sheng Yangc1e01512009-02-25 17:22:26 +08001817#ifdef __KVM_HAVE_MSIX
1818static int kvm_vm_ioctl_set_msix_nr(struct kvm *kvm,
1819 struct kvm_assigned_msix_nr *entry_nr)
1820{
1821 int r = 0;
1822 struct kvm_assigned_dev_kernel *adev;
1823
1824 mutex_lock(&kvm->lock);
1825
1826 adev = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head,
1827 entry_nr->assigned_dev_id);
1828 if (!adev) {
1829 r = -EINVAL;
1830 goto msix_nr_out;
1831 }
1832
1833 if (adev->entries_nr == 0) {
1834 adev->entries_nr = entry_nr->entry_nr;
1835 if (adev->entries_nr == 0 ||
1836 adev->entries_nr >= KVM_MAX_MSIX_PER_DEV) {
1837 r = -EINVAL;
1838 goto msix_nr_out;
1839 }
1840
1841 adev->host_msix_entries = kzalloc(sizeof(struct msix_entry) *
1842 entry_nr->entry_nr,
1843 GFP_KERNEL);
1844 if (!adev->host_msix_entries) {
1845 r = -ENOMEM;
1846 goto msix_nr_out;
1847 }
1848 adev->guest_msix_entries = kzalloc(
1849 sizeof(struct kvm_guest_msix_entry) *
1850 entry_nr->entry_nr, GFP_KERNEL);
1851 if (!adev->guest_msix_entries) {
1852 kfree(adev->host_msix_entries);
1853 r = -ENOMEM;
1854 goto msix_nr_out;
1855 }
1856 } else /* Not allowed set MSI-X number twice */
1857 r = -EINVAL;
1858msix_nr_out:
1859 mutex_unlock(&kvm->lock);
1860 return r;
1861}
1862
1863static int kvm_vm_ioctl_set_msix_entry(struct kvm *kvm,
1864 struct kvm_assigned_msix_entry *entry)
1865{
1866 int r = 0, i;
1867 struct kvm_assigned_dev_kernel *adev;
1868
1869 mutex_lock(&kvm->lock);
1870
1871 adev = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head,
1872 entry->assigned_dev_id);
1873
1874 if (!adev) {
1875 r = -EINVAL;
1876 goto msix_entry_out;
1877 }
1878
1879 for (i = 0; i < adev->entries_nr; i++)
1880 if (adev->guest_msix_entries[i].vector == 0 ||
1881 adev->guest_msix_entries[i].entry == entry->entry) {
1882 adev->guest_msix_entries[i].entry = entry->entry;
1883 adev->guest_msix_entries[i].vector = entry->gsi;
1884 adev->host_msix_entries[i].entry = entry->entry;
1885 break;
1886 }
1887 if (i == adev->entries_nr) {
1888 r = -ENOSPC;
1889 goto msix_entry_out;
1890 }
1891
1892msix_entry_out:
1893 mutex_unlock(&kvm->lock);
1894
1895 return r;
1896}
1897#endif
1898
Avi Kivitybccf2152007-02-21 18:04:26 +02001899static long kvm_vcpu_ioctl(struct file *filp,
1900 unsigned int ioctl, unsigned long arg)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001901{
Avi Kivitybccf2152007-02-21 18:04:26 +02001902 struct kvm_vcpu *vcpu = filp->private_data;
Al Viro2f366982007-02-09 16:38:35 +00001903 void __user *argp = (void __user *)arg;
Carsten Otte313a3dc2007-10-11 19:16:52 +02001904 int r;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001905 struct kvm_fpu *fpu = NULL;
1906 struct kvm_sregs *kvm_sregs = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001907
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001908 if (vcpu->kvm->mm != current->mm)
1909 return -EIO;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001910 switch (ioctl) {
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001911 case KVM_RUN:
Avi Kivityf0fe5102007-03-07 13:11:17 +02001912 r = -EINVAL;
1913 if (arg)
1914 goto out;
Hollis Blanchardb6c7a5d2007-11-01 14:16:10 -05001915 r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001916 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001917 case KVM_GET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001918 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001919
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001920 r = -ENOMEM;
1921 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
1922 if (!kvm_regs)
1923 goto out;
1924 r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001925 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001926 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001927 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001928 if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
1929 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001930 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001931out_free1:
1932 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001933 break;
1934 }
1935 case KVM_SET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001936 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001937
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001938 r = -ENOMEM;
1939 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
1940 if (!kvm_regs)
1941 goto out;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001942 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001943 if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs)))
1944 goto out_free2;
1945 r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001946 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001947 goto out_free2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001948 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001949out_free2:
1950 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001951 break;
1952 }
1953 case KVM_GET_SREGS: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001954 kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL);
1955 r = -ENOMEM;
1956 if (!kvm_sregs)
1957 goto out;
1958 r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001959 if (r)
1960 goto out;
1961 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001962 if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001963 goto out;
1964 r = 0;
1965 break;
1966 }
1967 case KVM_SET_SREGS: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001968 kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL);
1969 r = -ENOMEM;
1970 if (!kvm_sregs)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001971 goto out;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001972 r = -EFAULT;
1973 if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs)))
1974 goto out;
1975 r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001976 if (r)
1977 goto out;
1978 r = 0;
1979 break;
1980 }
Marcelo Tosatti62d9f0d2008-04-11 13:24:45 -03001981 case KVM_GET_MP_STATE: {
1982 struct kvm_mp_state mp_state;
1983
1984 r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state);
1985 if (r)
1986 goto out;
1987 r = -EFAULT;
1988 if (copy_to_user(argp, &mp_state, sizeof mp_state))
1989 goto out;
1990 r = 0;
1991 break;
1992 }
1993 case KVM_SET_MP_STATE: {
1994 struct kvm_mp_state mp_state;
1995
1996 r = -EFAULT;
1997 if (copy_from_user(&mp_state, argp, sizeof mp_state))
1998 goto out;
1999 r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
2000 if (r)
2001 goto out;
2002 r = 0;
2003 break;
2004 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08002005 case KVM_TRANSLATE: {
2006 struct kvm_translation tr;
2007
2008 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00002009 if (copy_from_user(&tr, argp, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08002010 goto out;
Zhang Xiantao8b006792007-11-16 13:05:55 +08002011 r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002012 if (r)
2013 goto out;
2014 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00002015 if (copy_to_user(argp, &tr, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08002016 goto out;
2017 r = 0;
2018 break;
2019 }
Jan Kiszkad0bfb942008-12-15 13:52:10 +01002020 case KVM_SET_GUEST_DEBUG: {
2021 struct kvm_guest_debug dbg;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002022
2023 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00002024 if (copy_from_user(&dbg, argp, sizeof dbg))
Avi Kivity6aa8b732006-12-10 02:21:36 -08002025 goto out;
Jan Kiszkad0bfb942008-12-15 13:52:10 +01002026 r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002027 if (r)
2028 goto out;
2029 r = 0;
2030 break;
2031 }
Avi Kivity1961d272007-03-05 19:46:05 +02002032 case KVM_SET_SIGNAL_MASK: {
2033 struct kvm_signal_mask __user *sigmask_arg = argp;
2034 struct kvm_signal_mask kvm_sigmask;
2035 sigset_t sigset, *p;
2036
2037 p = NULL;
2038 if (argp) {
2039 r = -EFAULT;
2040 if (copy_from_user(&kvm_sigmask, argp,
2041 sizeof kvm_sigmask))
2042 goto out;
2043 r = -EINVAL;
2044 if (kvm_sigmask.len != sizeof sigset)
2045 goto out;
2046 r = -EFAULT;
2047 if (copy_from_user(&sigset, sigmask_arg->sigset,
2048 sizeof sigset))
2049 goto out;
2050 p = &sigset;
2051 }
2052 r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
2053 break;
2054 }
Avi Kivityb8836732007-04-01 16:34:31 +03002055 case KVM_GET_FPU: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07002056 fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL);
2057 r = -ENOMEM;
2058 if (!fpu)
2059 goto out;
2060 r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03002061 if (r)
2062 goto out;
2063 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07002064 if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
Avi Kivityb8836732007-04-01 16:34:31 +03002065 goto out;
2066 r = 0;
2067 break;
2068 }
2069 case KVM_SET_FPU: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07002070 fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL);
2071 r = -ENOMEM;
2072 if (!fpu)
Avi Kivityb8836732007-04-01 16:34:31 +03002073 goto out;
Dave Hansenfa3795a2008-08-11 10:01:46 -07002074 r = -EFAULT;
2075 if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu)))
2076 goto out;
2077 r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03002078 if (r)
2079 goto out;
2080 r = 0;
2081 break;
2082 }
Avi Kivitybccf2152007-02-21 18:04:26 +02002083 default:
Carsten Otte313a3dc2007-10-11 19:16:52 +02002084 r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
Avi Kivitybccf2152007-02-21 18:04:26 +02002085 }
2086out:
Dave Hansenfa3795a2008-08-11 10:01:46 -07002087 kfree(fpu);
2088 kfree(kvm_sregs);
Avi Kivitybccf2152007-02-21 18:04:26 +02002089 return r;
2090}
2091
2092static long kvm_vm_ioctl(struct file *filp,
2093 unsigned int ioctl, unsigned long arg)
2094{
2095 struct kvm *kvm = filp->private_data;
2096 void __user *argp = (void __user *)arg;
Carsten Otte1fe779f2007-10-29 16:08:35 +01002097 int r;
Avi Kivitybccf2152007-02-21 18:04:26 +02002098
Avi Kivity6d4e4c42007-11-21 16:41:05 +02002099 if (kvm->mm != current->mm)
2100 return -EIO;
Avi Kivitybccf2152007-02-21 18:04:26 +02002101 switch (ioctl) {
2102 case KVM_CREATE_VCPU:
2103 r = kvm_vm_ioctl_create_vcpu(kvm, arg);
2104 if (r < 0)
2105 goto out;
2106 break;
Izik Eidus6fc138d2007-10-09 19:20:39 +02002107 case KVM_SET_USER_MEMORY_REGION: {
2108 struct kvm_userspace_memory_region kvm_userspace_mem;
2109
2110 r = -EFAULT;
2111 if (copy_from_user(&kvm_userspace_mem, argp,
2112 sizeof kvm_userspace_mem))
2113 goto out;
2114
2115 r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002116 if (r)
2117 goto out;
2118 break;
2119 }
2120 case KVM_GET_DIRTY_LOG: {
2121 struct kvm_dirty_log log;
2122
2123 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00002124 if (copy_from_user(&log, argp, sizeof log))
Avi Kivity6aa8b732006-12-10 02:21:36 -08002125 goto out;
Avi Kivity2c6f5df2007-02-20 18:27:58 +02002126 r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002127 if (r)
2128 goto out;
2129 break;
2130 }
Laurent Vivier5f94c172008-05-30 16:05:54 +02002131#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
2132 case KVM_REGISTER_COALESCED_MMIO: {
2133 struct kvm_coalesced_mmio_zone zone;
2134 r = -EFAULT;
2135 if (copy_from_user(&zone, argp, sizeof zone))
2136 goto out;
2137 r = -ENXIO;
2138 r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
2139 if (r)
2140 goto out;
2141 r = 0;
2142 break;
2143 }
2144 case KVM_UNREGISTER_COALESCED_MMIO: {
2145 struct kvm_coalesced_mmio_zone zone;
2146 r = -EFAULT;
2147 if (copy_from_user(&zone, argp, sizeof zone))
2148 goto out;
2149 r = -ENXIO;
2150 r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
2151 if (r)
2152 goto out;
2153 r = 0;
2154 break;
2155 }
2156#endif
Xiantao Zhang8a98f662008-10-06 13:47:38 +08002157#ifdef KVM_CAP_DEVICE_ASSIGNMENT
2158 case KVM_ASSIGN_PCI_DEVICE: {
2159 struct kvm_assigned_pci_dev assigned_dev;
2160
2161 r = -EFAULT;
2162 if (copy_from_user(&assigned_dev, argp, sizeof assigned_dev))
2163 goto out;
2164 r = kvm_vm_ioctl_assign_device(kvm, &assigned_dev);
2165 if (r)
2166 goto out;
2167 break;
2168 }
2169 case KVM_ASSIGN_IRQ: {
Sheng Yange56d5322009-03-12 21:45:39 +08002170 r = -EOPNOTSUPP;
2171 break;
2172 }
2173#ifdef KVM_CAP_ASSIGN_DEV_IRQ
2174 case KVM_ASSIGN_DEV_IRQ: {
Xiantao Zhang8a98f662008-10-06 13:47:38 +08002175 struct kvm_assigned_irq assigned_irq;
2176
2177 r = -EFAULT;
2178 if (copy_from_user(&assigned_irq, argp, sizeof assigned_irq))
2179 goto out;
2180 r = kvm_vm_ioctl_assign_irq(kvm, &assigned_irq);
2181 if (r)
2182 goto out;
2183 break;
2184 }
Sheng Yange56d5322009-03-12 21:45:39 +08002185 case KVM_DEASSIGN_DEV_IRQ: {
2186 struct kvm_assigned_irq assigned_irq;
2187
2188 r = -EFAULT;
2189 if (copy_from_user(&assigned_irq, argp, sizeof assigned_irq))
2190 goto out;
2191 r = kvm_vm_ioctl_deassign_dev_irq(kvm, &assigned_irq);
2192 if (r)
2193 goto out;
2194 break;
2195 }
2196#endif
Xiantao Zhang8a98f662008-10-06 13:47:38 +08002197#endif
Weidong Han0a920352008-12-02 21:24:23 +08002198#ifdef KVM_CAP_DEVICE_DEASSIGNMENT
2199 case KVM_DEASSIGN_PCI_DEVICE: {
2200 struct kvm_assigned_pci_dev assigned_dev;
2201
2202 r = -EFAULT;
2203 if (copy_from_user(&assigned_dev, argp, sizeof assigned_dev))
2204 goto out;
2205 r = kvm_vm_ioctl_deassign_device(kvm, &assigned_dev);
2206 if (r)
2207 goto out;
2208 break;
2209 }
2210#endif
Avi Kivity399ec802008-11-19 13:58:46 +02002211#ifdef KVM_CAP_IRQ_ROUTING
2212 case KVM_SET_GSI_ROUTING: {
2213 struct kvm_irq_routing routing;
2214 struct kvm_irq_routing __user *urouting;
2215 struct kvm_irq_routing_entry *entries;
2216
2217 r = -EFAULT;
2218 if (copy_from_user(&routing, argp, sizeof(routing)))
2219 goto out;
2220 r = -EINVAL;
2221 if (routing.nr >= KVM_MAX_IRQ_ROUTES)
2222 goto out;
2223 if (routing.flags)
2224 goto out;
2225 r = -ENOMEM;
2226 entries = vmalloc(routing.nr * sizeof(*entries));
2227 if (!entries)
2228 goto out;
2229 r = -EFAULT;
2230 urouting = argp;
2231 if (copy_from_user(entries, urouting->entries,
2232 routing.nr * sizeof(*entries)))
2233 goto out_free_irq_routing;
2234 r = kvm_set_irq_routing(kvm, entries, routing.nr,
2235 routing.flags);
2236 out_free_irq_routing:
2237 vfree(entries);
2238 break;
2239 }
Avi Kivity6621fbc2009-08-10 17:00:52 +03002240#endif /* KVM_CAP_IRQ_ROUTING */
Sheng Yangc1e01512009-02-25 17:22:26 +08002241#ifdef __KVM_HAVE_MSIX
2242 case KVM_ASSIGN_SET_MSIX_NR: {
2243 struct kvm_assigned_msix_nr entry_nr;
2244 r = -EFAULT;
2245 if (copy_from_user(&entry_nr, argp, sizeof entry_nr))
2246 goto out;
2247 r = kvm_vm_ioctl_set_msix_nr(kvm, &entry_nr);
2248 if (r)
2249 goto out;
2250 break;
2251 }
2252 case KVM_ASSIGN_SET_MSIX_ENTRY: {
2253 struct kvm_assigned_msix_entry entry;
2254 r = -EFAULT;
2255 if (copy_from_user(&entry, argp, sizeof entry))
2256 goto out;
2257 r = kvm_vm_ioctl_set_msix_entry(kvm, &entry);
2258 if (r)
2259 goto out;
2260 break;
2261 }
Avi Kivity399ec802008-11-19 13:58:46 +02002262#endif
Gregory Haskins721eecb2009-05-20 10:30:49 -04002263 case KVM_IRQFD: {
2264 struct kvm_irqfd data;
2265
2266 r = -EFAULT;
2267 if (copy_from_user(&data, argp, sizeof data))
2268 goto out;
2269 r = kvm_irqfd(kvm, data.fd, data.gsi, data.flags);
2270 break;
2271 }
Gregory Haskinsd34e6b12009-07-07 17:08:49 -04002272 case KVM_IOEVENTFD: {
2273 struct kvm_ioeventfd data;
2274
2275 r = -EFAULT;
2276 if (copy_from_user(&data, argp, sizeof data))
2277 goto out;
2278 r = kvm_ioeventfd(kvm, &data);
2279 break;
2280 }
Gleb Natapov73880c82009-06-09 15:56:28 +03002281#ifdef CONFIG_KVM_APIC_ARCHITECTURE
2282 case KVM_SET_BOOT_CPU_ID:
2283 r = 0;
Marcelo Tosatti894a9c52009-06-23 15:05:14 -03002284 mutex_lock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03002285 if (atomic_read(&kvm->online_vcpus) != 0)
2286 r = -EBUSY;
2287 else
2288 kvm->bsp_vcpu_id = arg;
Marcelo Tosatti894a9c52009-06-23 15:05:14 -03002289 mutex_unlock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03002290 break;
2291#endif
Avi Kivityf17abe92007-02-21 19:28:04 +02002292 default:
Carsten Otte1fe779f2007-10-29 16:08:35 +01002293 r = kvm_arch_vm_ioctl(filp, ioctl, arg);
Avi Kivityf17abe92007-02-21 19:28:04 +02002294 }
2295out:
2296 return r;
2297}
2298
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002299static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivityf17abe92007-02-21 19:28:04 +02002300{
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002301 struct page *page[1];
2302 unsigned long addr;
2303 int npages;
2304 gfn_t gfn = vmf->pgoff;
Avi Kivityf17abe92007-02-21 19:28:04 +02002305 struct kvm *kvm = vma->vm_file->private_data;
Avi Kivityf17abe92007-02-21 19:28:04 +02002306
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002307 addr = gfn_to_hva(kvm, gfn);
2308 if (kvm_is_error_hva(addr))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002309 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002310
2311 npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page,
2312 NULL);
2313 if (unlikely(npages != 1))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002314 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002315
2316 vmf->page = page[0];
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002317 return 0;
Avi Kivityf17abe92007-02-21 19:28:04 +02002318}
2319
2320static struct vm_operations_struct kvm_vm_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002321 .fault = kvm_vm_fault,
Avi Kivityf17abe92007-02-21 19:28:04 +02002322};
2323
2324static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma)
2325{
2326 vma->vm_ops = &kvm_vm_vm_ops;
2327 return 0;
2328}
2329
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01002330static struct file_operations kvm_vm_fops = {
Avi Kivityf17abe92007-02-21 19:28:04 +02002331 .release = kvm_vm_release,
2332 .unlocked_ioctl = kvm_vm_ioctl,
2333 .compat_ioctl = kvm_vm_ioctl,
2334 .mmap = kvm_vm_mmap,
2335};
2336
2337static int kvm_dev_ioctl_create_vm(void)
2338{
Al Viro2030a422008-02-23 06:46:49 -05002339 int fd;
Avi Kivityf17abe92007-02-21 19:28:04 +02002340 struct kvm *kvm;
2341
Avi Kivityf17abe92007-02-21 19:28:04 +02002342 kvm = kvm_create_vm();
Avi Kivityd6d28162007-06-28 08:38:16 -04002343 if (IS_ERR(kvm))
2344 return PTR_ERR(kvm);
Ulrich Drepper7d9dbca2008-07-23 21:29:22 -07002345 fd = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, 0);
Al Viro2030a422008-02-23 06:46:49 -05002346 if (fd < 0)
Al Viro66c0b392008-04-19 20:33:56 +01002347 kvm_put_kvm(kvm);
Avi Kivityf17abe92007-02-21 19:28:04 +02002348
Avi Kivityf17abe92007-02-21 19:28:04 +02002349 return fd;
Avi Kivityf17abe92007-02-21 19:28:04 +02002350}
2351
Avi Kivity1a811b62008-12-08 18:25:27 +02002352static long kvm_dev_ioctl_check_extension_generic(long arg)
2353{
2354 switch (arg) {
Avi Kivityca9edae2008-12-08 18:29:29 +02002355 case KVM_CAP_USER_MEMORY:
Avi Kivity1a811b62008-12-08 18:25:27 +02002356 case KVM_CAP_DESTROY_MEMORY_REGION_WORKS:
Jan Kiszka4cd481f2009-04-13 11:59:32 +02002357 case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS:
Gleb Natapov73880c82009-06-09 15:56:28 +03002358#ifdef CONFIG_KVM_APIC_ARCHITECTURE
2359 case KVM_CAP_SET_BOOT_CPU_ID:
2360#endif
Avi Kivity1a811b62008-12-08 18:25:27 +02002361 return 1;
Avi Kivity399ec802008-11-19 13:58:46 +02002362#ifdef CONFIG_HAVE_KVM_IRQCHIP
2363 case KVM_CAP_IRQ_ROUTING:
Sheng Yang36463142009-03-16 16:33:43 +08002364 return KVM_MAX_IRQ_ROUTES;
Avi Kivity399ec802008-11-19 13:58:46 +02002365#endif
Avi Kivity1a811b62008-12-08 18:25:27 +02002366 default:
2367 break;
2368 }
2369 return kvm_dev_ioctl_check_extension(arg);
2370}
2371
Avi Kivityf17abe92007-02-21 19:28:04 +02002372static long kvm_dev_ioctl(struct file *filp,
2373 unsigned int ioctl, unsigned long arg)
2374{
Avi Kivity07c45a32007-03-07 13:05:38 +02002375 long r = -EINVAL;
Avi Kivityf17abe92007-02-21 19:28:04 +02002376
2377 switch (ioctl) {
2378 case KVM_GET_API_VERSION:
Avi Kivityf0fe5102007-03-07 13:11:17 +02002379 r = -EINVAL;
2380 if (arg)
2381 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02002382 r = KVM_API_VERSION;
2383 break;
2384 case KVM_CREATE_VM:
Avi Kivityf0fe5102007-03-07 13:11:17 +02002385 r = -EINVAL;
2386 if (arg)
2387 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02002388 r = kvm_dev_ioctl_create_vm();
2389 break;
Zhang Xiantao018d00d2007-11-15 23:07:47 +08002390 case KVM_CHECK_EXTENSION:
Avi Kivity1a811b62008-12-08 18:25:27 +02002391 r = kvm_dev_ioctl_check_extension_generic(arg);
Avi Kivity5d308f42007-03-01 17:56:20 +02002392 break;
Avi Kivity07c45a32007-03-07 13:05:38 +02002393 case KVM_GET_VCPU_MMAP_SIZE:
2394 r = -EINVAL;
2395 if (arg)
2396 goto out;
Avi Kivityadb1ff42008-01-24 15:13:08 +02002397 r = PAGE_SIZE; /* struct kvm_run */
2398#ifdef CONFIG_X86
2399 r += PAGE_SIZE; /* pio data page */
2400#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02002401#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
2402 r += PAGE_SIZE; /* coalesced mmio ring page */
2403#endif
Avi Kivity07c45a32007-03-07 13:05:38 +02002404 break;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04002405 case KVM_TRACE_ENABLE:
2406 case KVM_TRACE_PAUSE:
2407 case KVM_TRACE_DISABLE:
Marcelo Tosatti2023a292009-06-18 11:47:28 -03002408 r = -EOPNOTSUPP;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04002409 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002410 default:
Carsten Otte043405e2007-10-10 17:16:19 +02002411 return kvm_arch_dev_ioctl(filp, ioctl, arg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002412 }
2413out:
2414 return r;
2415}
2416
Avi Kivity6aa8b732006-12-10 02:21:36 -08002417static struct file_operations kvm_chardev_ops = {
Avi Kivity6aa8b732006-12-10 02:21:36 -08002418 .unlocked_ioctl = kvm_dev_ioctl,
2419 .compat_ioctl = kvm_dev_ioctl,
Avi Kivity6aa8b732006-12-10 02:21:36 -08002420};
2421
2422static struct miscdevice kvm_dev = {
Avi Kivitybbe44322007-03-04 13:27:36 +02002423 KVM_MINOR,
Avi Kivity6aa8b732006-12-10 02:21:36 -08002424 "kvm",
2425 &kvm_chardev_ops,
2426};
2427
Avi Kivity1b6c0162007-05-24 13:03:52 +03002428static void hardware_enable(void *junk)
2429{
2430 int cpu = raw_smp_processor_id();
2431
Rusty Russell7f59f492008-12-07 21:25:45 +10302432 if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03002433 return;
Rusty Russell7f59f492008-12-07 21:25:45 +10302434 cpumask_set_cpu(cpu, cpus_hardware_enabled);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002435 kvm_arch_hardware_enable(NULL);
Avi Kivity1b6c0162007-05-24 13:03:52 +03002436}
2437
2438static void hardware_disable(void *junk)
2439{
2440 int cpu = raw_smp_processor_id();
2441
Rusty Russell7f59f492008-12-07 21:25:45 +10302442 if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03002443 return;
Rusty Russell7f59f492008-12-07 21:25:45 +10302444 cpumask_clear_cpu(cpu, cpus_hardware_enabled);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002445 kvm_arch_hardware_disable(NULL);
Avi Kivity1b6c0162007-05-24 13:03:52 +03002446}
2447
Avi Kivity774c47f2007-02-12 00:54:47 -08002448static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val,
2449 void *v)
2450{
2451 int cpu = (long)v;
2452
Avi Kivity1a6f4d72007-11-11 18:37:32 +02002453 val &= ~CPU_TASKS_FROZEN;
Avi Kivity774c47f2007-02-12 00:54:47 -08002454 switch (val) {
Avi Kivitycec9ad22007-05-24 13:11:41 +03002455 case CPU_DYING:
Avi Kivity6ec8a852007-08-19 15:57:26 +03002456 printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n",
2457 cpu);
2458 hardware_disable(NULL);
2459 break;
Avi Kivity774c47f2007-02-12 00:54:47 -08002460 case CPU_UP_CANCELED:
Jeremy Katz43934a32007-02-19 14:37:46 +02002461 printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n",
2462 cpu);
Jens Axboe8691e5a2008-06-06 11:18:06 +02002463 smp_call_function_single(cpu, hardware_disable, NULL, 1);
Avi Kivity774c47f2007-02-12 00:54:47 -08002464 break;
Jeremy Katz43934a32007-02-19 14:37:46 +02002465 case CPU_ONLINE:
2466 printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n",
2467 cpu);
Jens Axboe8691e5a2008-06-06 11:18:06 +02002468 smp_call_function_single(cpu, hardware_enable, NULL, 1);
Avi Kivity774c47f2007-02-12 00:54:47 -08002469 break;
2470 }
2471 return NOTIFY_OK;
2472}
2473
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002474
2475asmlinkage void kvm_handle_fault_on_reboot(void)
2476{
2477 if (kvm_rebooting)
2478 /* spin while reset goes on */
2479 while (true)
2480 ;
2481 /* Fault while not rebooting. We want the trace. */
2482 BUG();
2483}
2484EXPORT_SYMBOL_GPL(kvm_handle_fault_on_reboot);
2485
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002486static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
Mike Dayd77c26f2007-10-08 09:02:08 -04002487 void *v)
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002488{
Sheng Yang8e1c1812009-04-29 11:09:04 +08002489 /*
2490 * Some (well, at least mine) BIOSes hang on reboot if
2491 * in vmx root mode.
2492 *
2493 * And Intel TXT required VMX off for all cpu when system shutdown.
2494 */
2495 printk(KERN_INFO "kvm: exiting hardware virtualization\n");
2496 kvm_rebooting = true;
2497 on_each_cpu(hardware_disable, NULL, 1);
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002498 return NOTIFY_OK;
2499}
2500
2501static struct notifier_block kvm_reboot_notifier = {
2502 .notifier_call = kvm_reboot,
2503 .priority = 0,
2504};
2505
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002506void kvm_io_bus_init(struct kvm_io_bus *bus)
2507{
2508 memset(bus, 0, sizeof(*bus));
2509}
2510
2511void kvm_io_bus_destroy(struct kvm_io_bus *bus)
2512{
2513 int i;
2514
2515 for (i = 0; i < bus->dev_count; i++) {
2516 struct kvm_io_device *pos = bus->devs[i];
2517
2518 kvm_iodevice_destructor(pos);
2519 }
2520}
2521
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002522/* kvm_io_bus_write - called under kvm->slots_lock */
2523int kvm_io_bus_write(struct kvm_io_bus *bus, gpa_t addr,
2524 int len, const void *val)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002525{
2526 int i;
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002527 for (i = 0; i < bus->dev_count; i++)
2528 if (!kvm_iodevice_write(bus->devs[i], addr, len, val))
2529 return 0;
2530 return -EOPNOTSUPP;
2531}
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002532
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002533/* kvm_io_bus_read - called under kvm->slots_lock */
2534int kvm_io_bus_read(struct kvm_io_bus *bus, gpa_t addr, int len, void *val)
2535{
2536 int i;
2537 for (i = 0; i < bus->dev_count; i++)
2538 if (!kvm_iodevice_read(bus->devs[i], addr, len, val))
2539 return 0;
2540 return -EOPNOTSUPP;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002541}
2542
Gregory Haskins090b7af2009-07-07 17:08:44 -04002543int kvm_io_bus_register_dev(struct kvm *kvm, struct kvm_io_bus *bus,
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002544 struct kvm_io_device *dev)
2545{
Gregory Haskins090b7af2009-07-07 17:08:44 -04002546 int ret;
2547
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002548 down_write(&kvm->slots_lock);
Gregory Haskins090b7af2009-07-07 17:08:44 -04002549 ret = __kvm_io_bus_register_dev(bus, dev);
2550 up_write(&kvm->slots_lock);
2551
2552 return ret;
2553}
2554
2555/* An unlocked version. Caller must have write lock on slots_lock. */
2556int __kvm_io_bus_register_dev(struct kvm_io_bus *bus,
2557 struct kvm_io_device *dev)
2558{
2559 if (bus->dev_count > NR_IOBUS_DEVS-1)
2560 return -ENOSPC;
2561
2562 bus->devs[bus->dev_count++] = dev;
2563
2564 return 0;
2565}
2566
2567void kvm_io_bus_unregister_dev(struct kvm *kvm,
2568 struct kvm_io_bus *bus,
2569 struct kvm_io_device *dev)
2570{
2571 down_write(&kvm->slots_lock);
2572 __kvm_io_bus_unregister_dev(bus, dev);
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002573 up_write(&kvm->slots_lock);
2574}
2575
2576/* An unlocked version. Caller must have write lock on slots_lock. */
Gregory Haskins090b7af2009-07-07 17:08:44 -04002577void __kvm_io_bus_unregister_dev(struct kvm_io_bus *bus,
2578 struct kvm_io_device *dev)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002579{
Gregory Haskins090b7af2009-07-07 17:08:44 -04002580 int i;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002581
Gregory Haskins090b7af2009-07-07 17:08:44 -04002582 for (i = 0; i < bus->dev_count; i++)
2583 if (bus->devs[i] == dev) {
2584 bus->devs[i] = bus->devs[--bus->dev_count];
2585 break;
2586 }
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002587}
2588
Avi Kivity774c47f2007-02-12 00:54:47 -08002589static struct notifier_block kvm_cpu_notifier = {
2590 .notifier_call = kvm_cpu_hotplug,
2591 .priority = 20, /* must be > scheduler priority */
2592};
2593
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002594static int vm_stat_get(void *_offset, u64 *val)
Avi Kivityba1389b2007-11-18 16:24:12 +02002595{
2596 unsigned offset = (long)_offset;
Avi Kivityba1389b2007-11-18 16:24:12 +02002597 struct kvm *kvm;
2598
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002599 *val = 0;
Avi Kivityba1389b2007-11-18 16:24:12 +02002600 spin_lock(&kvm_lock);
2601 list_for_each_entry(kvm, &vm_list, vm_list)
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002602 *val += *(u32 *)((void *)kvm + offset);
Avi Kivityba1389b2007-11-18 16:24:12 +02002603 spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002604 return 0;
Avi Kivityba1389b2007-11-18 16:24:12 +02002605}
2606
2607DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n");
2608
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002609static int vcpu_stat_get(void *_offset, u64 *val)
Avi Kivity1165f5f2007-04-19 17:27:43 +03002610{
2611 unsigned offset = (long)_offset;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002612 struct kvm *kvm;
2613 struct kvm_vcpu *vcpu;
2614 int i;
2615
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002616 *val = 0;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002617 spin_lock(&kvm_lock);
2618 list_for_each_entry(kvm, &vm_list, vm_list)
Gleb Natapov988a2ca2009-06-09 15:56:29 +03002619 kvm_for_each_vcpu(i, vcpu, kvm)
2620 *val += *(u32 *)((void *)vcpu + offset);
2621
Avi Kivity1165f5f2007-04-19 17:27:43 +03002622 spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002623 return 0;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002624}
2625
Avi Kivityba1389b2007-11-18 16:24:12 +02002626DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n");
2627
2628static struct file_operations *stat_fops[] = {
2629 [KVM_STAT_VCPU] = &vcpu_stat_fops,
2630 [KVM_STAT_VM] = &vm_stat_fops,
2631};
Avi Kivity1165f5f2007-04-19 17:27:43 +03002632
Zhang Xiantaoa16b0432007-11-16 14:38:21 +08002633static void kvm_init_debug(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002634{
2635 struct kvm_stats_debugfs_item *p;
2636
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002637 kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002638 for (p = debugfs_entries; p->name; ++p)
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002639 p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir,
Avi Kivity1165f5f2007-04-19 17:27:43 +03002640 (void *)(long)p->offset,
Avi Kivityba1389b2007-11-18 16:24:12 +02002641 stat_fops[p->kind]);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002642}
2643
2644static void kvm_exit_debug(void)
2645{
2646 struct kvm_stats_debugfs_item *p;
2647
2648 for (p = debugfs_entries; p->name; ++p)
2649 debugfs_remove(p->dentry);
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002650 debugfs_remove(kvm_debugfs_dir);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002651}
2652
Avi Kivity59ae6c62007-02-12 00:54:48 -08002653static int kvm_suspend(struct sys_device *dev, pm_message_t state)
2654{
Avi Kivity4267c412007-05-24 13:09:41 +03002655 hardware_disable(NULL);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002656 return 0;
2657}
2658
2659static int kvm_resume(struct sys_device *dev)
2660{
Avi Kivity4267c412007-05-24 13:09:41 +03002661 hardware_enable(NULL);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002662 return 0;
2663}
2664
2665static struct sysdev_class kvm_sysdev_class = {
Kay Sieversaf5ca3f2007-12-20 02:09:39 +01002666 .name = "kvm",
Avi Kivity59ae6c62007-02-12 00:54:48 -08002667 .suspend = kvm_suspend,
2668 .resume = kvm_resume,
2669};
2670
2671static struct sys_device kvm_sysdev = {
2672 .id = 0,
2673 .cls = &kvm_sysdev_class,
2674};
2675
Izik Eiduscea7bb22007-10-17 19:17:48 +02002676struct page *bad_page;
Anthony Liguori35149e22008-04-02 14:46:56 -05002677pfn_t bad_pfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002678
Avi Kivity15ad7142007-07-11 18:17:21 +03002679static inline
2680struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn)
2681{
2682 return container_of(pn, struct kvm_vcpu, preempt_notifier);
2683}
2684
2685static void kvm_sched_in(struct preempt_notifier *pn, int cpu)
2686{
2687 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2688
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002689 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002690}
2691
2692static void kvm_sched_out(struct preempt_notifier *pn,
2693 struct task_struct *next)
2694{
2695 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2696
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002697 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002698}
2699
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002700int kvm_init(void *opaque, unsigned int vcpu_size,
Rusty Russellc16f8622007-07-30 21:12:19 +10002701 struct module *module)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002702{
2703 int r;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002704 int cpu;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002705
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002706 kvm_init_debug();
2707
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002708 r = kvm_arch_init(opaque);
2709 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002710 goto out_fail;
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002711
2712 bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2713
2714 if (bad_page == NULL) {
2715 r = -ENOMEM;
2716 goto out;
2717 }
2718
Anthony Liguori35149e22008-04-02 14:46:56 -05002719 bad_pfn = page_to_pfn(bad_page);
2720
Avi Kivity8437a612009-06-06 14:52:35 -07002721 if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
Rusty Russell7f59f492008-12-07 21:25:45 +10302722 r = -ENOMEM;
2723 goto out_free_0;
2724 }
2725
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002726 r = kvm_arch_hardware_setup();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002727 if (r < 0)
Rusty Russell7f59f492008-12-07 21:25:45 +10302728 goto out_free_0a;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002729
Yang, Sheng002c7f72007-07-31 14:23:01 +03002730 for_each_online_cpu(cpu) {
2731 smp_call_function_single(cpu,
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002732 kvm_arch_check_processor_compat,
Jens Axboe8691e5a2008-06-06 11:18:06 +02002733 &r, 1);
Yang, Sheng002c7f72007-07-31 14:23:01 +03002734 if (r < 0)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002735 goto out_free_1;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002736 }
2737
Jens Axboe15c8b6c2008-05-09 09:39:44 +02002738 on_each_cpu(hardware_enable, NULL, 1);
Avi Kivity774c47f2007-02-12 00:54:47 -08002739 r = register_cpu_notifier(&kvm_cpu_notifier);
2740 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002741 goto out_free_2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002742 register_reboot_notifier(&kvm_reboot_notifier);
2743
Avi Kivity59ae6c62007-02-12 00:54:48 -08002744 r = sysdev_class_register(&kvm_sysdev_class);
2745 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002746 goto out_free_3;
Avi Kivity59ae6c62007-02-12 00:54:48 -08002747
2748 r = sysdev_register(&kvm_sysdev);
2749 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002750 goto out_free_4;
Avi Kivity59ae6c62007-02-12 00:54:48 -08002751
Rusty Russellc16f8622007-07-30 21:12:19 +10002752 /* A kmem cache lets us meet the alignment requirements of fx_save. */
2753 kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size,
Joe Perches56919c52007-11-12 20:06:51 -08002754 __alignof__(struct kvm_vcpu),
2755 0, NULL);
Rusty Russellc16f8622007-07-30 21:12:19 +10002756 if (!kvm_vcpu_cache) {
2757 r = -ENOMEM;
Zhang Xiantaod23087842007-11-29 15:35:39 +08002758 goto out_free_5;
Rusty Russellc16f8622007-07-30 21:12:19 +10002759 }
2760
Avi Kivity6aa8b732006-12-10 02:21:36 -08002761 kvm_chardev_ops.owner = module;
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01002762 kvm_vm_fops.owner = module;
2763 kvm_vcpu_fops.owner = module;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002764
2765 r = misc_register(&kvm_dev);
2766 if (r) {
Mike Dayd77c26f2007-10-08 09:02:08 -04002767 printk(KERN_ERR "kvm: misc device register failed\n");
Avi Kivity6aa8b732006-12-10 02:21:36 -08002768 goto out_free;
2769 }
2770
Avi Kivity15ad7142007-07-11 18:17:21 +03002771 kvm_preempt_ops.sched_in = kvm_sched_in;
2772 kvm_preempt_ops.sched_out = kvm_sched_out;
2773
Avi Kivityc7addb92007-09-16 18:58:32 +02002774 return 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002775
2776out_free:
Rusty Russellc16f8622007-07-30 21:12:19 +10002777 kmem_cache_destroy(kvm_vcpu_cache);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002778out_free_5:
Avi Kivity59ae6c62007-02-12 00:54:48 -08002779 sysdev_unregister(&kvm_sysdev);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002780out_free_4:
Avi Kivity59ae6c62007-02-12 00:54:48 -08002781 sysdev_class_unregister(&kvm_sysdev_class);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002782out_free_3:
Avi Kivity6aa8b732006-12-10 02:21:36 -08002783 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity774c47f2007-02-12 00:54:47 -08002784 unregister_cpu_notifier(&kvm_cpu_notifier);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002785out_free_2:
Jens Axboe15c8b6c2008-05-09 09:39:44 +02002786 on_each_cpu(hardware_disable, NULL, 1);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002787out_free_1:
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002788 kvm_arch_hardware_unsetup();
Rusty Russell7f59f492008-12-07 21:25:45 +10302789out_free_0a:
2790 free_cpumask_var(cpus_hardware_enabled);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002791out_free_0:
2792 __free_page(bad_page);
Avi Kivityca45aaa2007-03-01 19:21:03 +02002793out:
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002794 kvm_arch_exit();
Zhang Xiantaod23087842007-11-29 15:35:39 +08002795out_fail:
Xiao Guangrongaed665f2009-08-03 17:10:06 +08002796 kvm_exit_debug();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002797 return r;
2798}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002799EXPORT_SYMBOL_GPL(kvm_init);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002800
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002801void kvm_exit(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002802{
Marcelo Tosatti229456f2009-06-17 09:22:14 -03002803 tracepoint_synchronize_unregister();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002804 misc_deregister(&kvm_dev);
Rusty Russellc16f8622007-07-30 21:12:19 +10002805 kmem_cache_destroy(kvm_vcpu_cache);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002806 sysdev_unregister(&kvm_sysdev);
2807 sysdev_class_unregister(&kvm_sysdev_class);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002808 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002809 unregister_cpu_notifier(&kvm_cpu_notifier);
Jens Axboe15c8b6c2008-05-09 09:39:44 +02002810 on_each_cpu(hardware_disable, NULL, 1);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002811 kvm_arch_hardware_unsetup();
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002812 kvm_arch_exit();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002813 kvm_exit_debug();
Rusty Russell7f59f492008-12-07 21:25:45 +10302814 free_cpumask_var(cpus_hardware_enabled);
Izik Eiduscea7bb22007-10-17 19:17:48 +02002815 __free_page(bad_page);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002816}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002817EXPORT_SYMBOL_GPL(kvm_exit);