blob: 897bff3b7df914a32f309de7e6b687fa7633cfe2 [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
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030741 if (alloc_cpumask_var(&cpus, GFP_ATOMIC))
742 cpumask_clear(cpus);
743
Marcelo Tosatti84261922009-06-17 10:53:47 -0300744 spin_lock(&kvm->requests_lock);
Jan Kiszkae601e3b2009-07-20 11:30:12 +0200745 me = smp_processor_id();
Gleb Natapov988a2ca2009-06-09 15:56:29 +0300746 kvm_for_each_vcpu(i, vcpu, kvm) {
Rusty Russell49846892008-12-08 20:26:24 +1030747 if (test_and_set_bit(req, &vcpu->requests))
Avi Kivityd9e368d2007-06-07 19:18:30 +0300748 continue;
749 cpu = vcpu->cpu;
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030750 if (cpus != NULL && cpu != -1 && cpu != me)
751 cpumask_set_cpu(cpu, cpus);
Avi Kivityd9e368d2007-06-07 19:18:30 +0300752 }
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030753 if (unlikely(cpus == NULL))
754 smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1);
755 else if (!cpumask_empty(cpus))
756 smp_call_function_many(cpus, ack_flush, NULL, 1);
757 else
758 called = false;
Marcelo Tosatti84261922009-06-17 10:53:47 -0300759 spin_unlock(&kvm->requests_lock);
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030760 free_cpumask_var(cpus);
Rusty Russell49846892008-12-08 20:26:24 +1030761 return called;
762}
763
764void kvm_flush_remote_tlbs(struct kvm *kvm)
765{
766 if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH))
767 ++kvm->stat.remote_tlb_flush;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300768}
769
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500770void kvm_reload_remote_mmus(struct kvm *kvm)
771{
Rusty Russell49846892008-12-08 20:26:24 +1030772 make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD);
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500773}
774
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000775int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id)
776{
777 struct page *page;
778 int r;
779
780 mutex_init(&vcpu->mutex);
781 vcpu->cpu = -1;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000782 vcpu->kvm = kvm;
783 vcpu->vcpu_id = id;
Eddie Dongb6958ce2007-07-18 12:15:21 +0300784 init_waitqueue_head(&vcpu->wq);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000785
786 page = alloc_page(GFP_KERNEL | __GFP_ZERO);
787 if (!page) {
788 r = -ENOMEM;
789 goto fail;
790 }
791 vcpu->run = page_address(page);
792
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800793 r = kvm_arch_vcpu_init(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000794 if (r < 0)
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800795 goto fail_free_run;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000796 return 0;
797
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000798fail_free_run:
799 free_page((unsigned long)vcpu->run);
800fail:
Rusty Russell76fafa52007-10-08 10:50:48 +1000801 return r;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000802}
803EXPORT_SYMBOL_GPL(kvm_vcpu_init);
804
805void kvm_vcpu_uninit(struct kvm_vcpu *vcpu)
806{
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800807 kvm_arch_vcpu_uninit(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000808 free_page((unsigned long)vcpu->run);
809}
810EXPORT_SYMBOL_GPL(kvm_vcpu_uninit);
811
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200812#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
813static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn)
814{
815 return container_of(mn, struct kvm, mmu_notifier);
816}
817
818static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn,
819 struct mm_struct *mm,
820 unsigned long address)
821{
822 struct kvm *kvm = mmu_notifier_to_kvm(mn);
823 int need_tlb_flush;
824
825 /*
826 * When ->invalidate_page runs, the linux pte has been zapped
827 * already but the page is still allocated until
828 * ->invalidate_page returns. So if we increase the sequence
829 * here the kvm page fault will notice if the spte can't be
830 * established because the page is going to be freed. If
831 * instead the kvm page fault establishes the spte before
832 * ->invalidate_page runs, kvm_unmap_hva will release it
833 * before returning.
834 *
835 * The sequence increase only need to be seen at spin_unlock
836 * time, and not at spin_lock time.
837 *
838 * Increasing the sequence after the spin_unlock would be
839 * unsafe because the kvm page fault could then establish the
840 * pte after kvm_unmap_hva returned, without noticing the page
841 * is going to be freed.
842 */
843 spin_lock(&kvm->mmu_lock);
844 kvm->mmu_notifier_seq++;
845 need_tlb_flush = kvm_unmap_hva(kvm, address);
846 spin_unlock(&kvm->mmu_lock);
847
848 /* we've to flush the tlb before the pages can be freed */
849 if (need_tlb_flush)
850 kvm_flush_remote_tlbs(kvm);
851
852}
853
854static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn,
855 struct mm_struct *mm,
856 unsigned long start,
857 unsigned long end)
858{
859 struct kvm *kvm = mmu_notifier_to_kvm(mn);
860 int need_tlb_flush = 0;
861
862 spin_lock(&kvm->mmu_lock);
863 /*
864 * The count increase must become visible at unlock time as no
865 * spte can be established without taking the mmu_lock and
866 * count is also read inside the mmu_lock critical section.
867 */
868 kvm->mmu_notifier_count++;
869 for (; start < end; start += PAGE_SIZE)
870 need_tlb_flush |= kvm_unmap_hva(kvm, start);
871 spin_unlock(&kvm->mmu_lock);
872
873 /* we've to flush the tlb before the pages can be freed */
874 if (need_tlb_flush)
875 kvm_flush_remote_tlbs(kvm);
876}
877
878static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
879 struct mm_struct *mm,
880 unsigned long start,
881 unsigned long end)
882{
883 struct kvm *kvm = mmu_notifier_to_kvm(mn);
884
885 spin_lock(&kvm->mmu_lock);
886 /*
887 * This sequence increase will notify the kvm page fault that
888 * the page that is going to be mapped in the spte could have
889 * been freed.
890 */
891 kvm->mmu_notifier_seq++;
892 /*
893 * The above sequence increase must be visible before the
894 * below count decrease but both values are read by the kvm
895 * page fault under mmu_lock spinlock so we don't need to add
896 * a smb_wmb() here in between the two.
897 */
898 kvm->mmu_notifier_count--;
899 spin_unlock(&kvm->mmu_lock);
900
901 BUG_ON(kvm->mmu_notifier_count < 0);
902}
903
904static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
905 struct mm_struct *mm,
906 unsigned long address)
907{
908 struct kvm *kvm = mmu_notifier_to_kvm(mn);
909 int young;
910
911 spin_lock(&kvm->mmu_lock);
912 young = kvm_age_hva(kvm, address);
913 spin_unlock(&kvm->mmu_lock);
914
915 if (young)
916 kvm_flush_remote_tlbs(kvm);
917
918 return young;
919}
920
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100921static void kvm_mmu_notifier_release(struct mmu_notifier *mn,
922 struct mm_struct *mm)
923{
924 struct kvm *kvm = mmu_notifier_to_kvm(mn);
925 kvm_arch_flush_shadow(kvm);
926}
927
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200928static const struct mmu_notifier_ops kvm_mmu_notifier_ops = {
929 .invalidate_page = kvm_mmu_notifier_invalidate_page,
930 .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start,
931 .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end,
932 .clear_flush_young = kvm_mmu_notifier_clear_flush_young,
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100933 .release = kvm_mmu_notifier_release,
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200934};
935#endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */
936
Avi Kivityf17abe92007-02-21 19:28:04 +0200937static struct kvm *kvm_create_vm(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800938{
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800939 struct kvm *kvm = kvm_arch_create_vm();
Laurent Vivier5f94c172008-05-30 16:05:54 +0200940#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
941 struct page *page;
942#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -0800943
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800944 if (IS_ERR(kvm))
945 goto out;
Avi Kivity75858a82009-01-04 17:10:50 +0200946#ifdef CONFIG_HAVE_KVM_IRQCHIP
Avi Kivity399ec802008-11-19 13:58:46 +0200947 INIT_LIST_HEAD(&kvm->irq_routing);
Avi Kivity75858a82009-01-04 17:10:50 +0200948 INIT_HLIST_HEAD(&kvm->mask_notifier_list);
949#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -0800950
Laurent Vivier5f94c172008-05-30 16:05:54 +0200951#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
952 page = alloc_page(GFP_KERNEL | __GFP_ZERO);
953 if (!page) {
954 kfree(kvm);
955 return ERR_PTR(-ENOMEM);
956 }
957 kvm->coalesced_mmio_ring =
958 (struct kvm_coalesced_mmio_ring *)page_address(page);
959#endif
960
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200961#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
962 {
963 int err;
964 kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops;
965 err = mmu_notifier_register(&kvm->mmu_notifier, current->mm);
966 if (err) {
967#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
968 put_page(page);
969#endif
970 kfree(kvm);
971 return ERR_PTR(err);
972 }
973 }
974#endif
975
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200976 kvm->mm = current->mm;
977 atomic_inc(&kvm->mm->mm_count);
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -0500978 spin_lock_init(&kvm->mmu_lock);
Marcelo Tosatti84261922009-06-17 10:53:47 -0300979 spin_lock_init(&kvm->requests_lock);
Eddie Dong74906342007-06-19 18:05:03 +0300980 kvm_io_bus_init(&kvm->pio_bus);
Gregory Haskinsd34e6b12009-07-07 17:08:49 -0400981 kvm_eventfd_init(kvm);
Shaohua Li11ec2802007-07-23 14:51:37 +0800982 mutex_init(&kvm->lock);
Marcelo Tosatti60eead72009-06-04 15:08:23 -0300983 mutex_init(&kvm->irq_lock);
Gregory Haskins2eeb2e92007-05-31 14:08:53 -0400984 kvm_io_bus_init(&kvm->mmio_bus);
Izik Eidus72dc67a2008-02-10 18:04:15 +0200985 init_rwsem(&kvm->slots_lock);
Izik Eidusd39f13b2008-03-30 16:01:25 +0300986 atomic_set(&kvm->users_count, 1);
Rusty Russell5e58cfe2007-07-23 17:08:21 +1000987 spin_lock(&kvm_lock);
988 list_add(&kvm->vm_list, &vm_list);
989 spin_unlock(&kvm_lock);
Laurent Vivier5f94c172008-05-30 16:05:54 +0200990#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
991 kvm_coalesced_mmio_init(kvm);
992#endif
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800993out:
Avi Kivityf17abe92007-02-21 19:28:04 +0200994 return kvm;
995}
996
Avi Kivity6aa8b732006-12-10 02:21:36 -0800997/*
998 * Free any memory in @free but not in @dont.
999 */
1000static void kvm_free_physmem_slot(struct kvm_memory_slot *free,
1001 struct kvm_memory_slot *dont)
1002{
Joerg Roedelec04b262009-06-19 15:16:23 +02001003 int i;
1004
Izik Eidus290fc382007-09-27 14:11:22 +02001005 if (!dont || free->rmap != dont->rmap)
1006 vfree(free->rmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001007
1008 if (!dont || free->dirty_bitmap != dont->dirty_bitmap)
1009 vfree(free->dirty_bitmap);
1010
Joerg Roedelec04b262009-06-19 15:16:23 +02001011
1012 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) {
1013 if (!dont || free->lpage_info[i] != dont->lpage_info[i]) {
1014 vfree(free->lpage_info[i]);
1015 free->lpage_info[i] = NULL;
1016 }
1017 }
Marcelo Tosatti05da4552008-02-23 11:44:30 -03001018
Avi Kivity6aa8b732006-12-10 02:21:36 -08001019 free->npages = 0;
Al Viro8b6d44c2007-02-09 16:38:40 +00001020 free->dirty_bitmap = NULL;
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001021 free->rmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001022}
1023
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +08001024void kvm_free_physmem(struct kvm *kvm)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001025{
1026 int i;
1027
1028 for (i = 0; i < kvm->nmemslots; ++i)
Al Viro8b6d44c2007-02-09 16:38:40 +00001029 kvm_free_physmem_slot(&kvm->memslots[i], NULL);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001030}
1031
Avi Kivityf17abe92007-02-21 19:28:04 +02001032static void kvm_destroy_vm(struct kvm *kvm)
1033{
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001034 struct mm_struct *mm = kvm->mm;
1035
Sheng Yangad8ba2c2009-01-06 10:03:02 +08001036 kvm_arch_sync_events(kvm);
Avi Kivity133de902007-02-12 00:54:44 -08001037 spin_lock(&kvm_lock);
1038 list_del(&kvm->vm_list);
1039 spin_unlock(&kvm_lock);
Avi Kivity399ec802008-11-19 13:58:46 +02001040 kvm_free_irq_routing(kvm);
Eddie Dong74906342007-06-19 18:05:03 +03001041 kvm_io_bus_destroy(&kvm->pio_bus);
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04001042 kvm_io_bus_destroy(&kvm->mmio_bus);
Laurent Vivier5f94c172008-05-30 16:05:54 +02001043#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1044 if (kvm->coalesced_mmio_ring != NULL)
1045 free_page((unsigned long)kvm->coalesced_mmio_ring);
1046#endif
Andrea Arcangelie930bff2008-07-25 16:24:52 +02001047#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1048 mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
Gleb Natapovf00be0c2009-03-19 12:20:36 +02001049#else
1050 kvm_arch_flush_shadow(kvm);
Andrea Arcangelie930bff2008-07-25 16:24:52 +02001051#endif
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +08001052 kvm_arch_destroy_vm(kvm);
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001053 mmdrop(mm);
Avi Kivityf17abe92007-02-21 19:28:04 +02001054}
1055
Izik Eidusd39f13b2008-03-30 16:01:25 +03001056void kvm_get_kvm(struct kvm *kvm)
1057{
1058 atomic_inc(&kvm->users_count);
1059}
1060EXPORT_SYMBOL_GPL(kvm_get_kvm);
1061
1062void kvm_put_kvm(struct kvm *kvm)
1063{
1064 if (atomic_dec_and_test(&kvm->users_count))
1065 kvm_destroy_vm(kvm);
1066}
1067EXPORT_SYMBOL_GPL(kvm_put_kvm);
1068
1069
Avi Kivityf17abe92007-02-21 19:28:04 +02001070static int kvm_vm_release(struct inode *inode, struct file *filp)
1071{
1072 struct kvm *kvm = filp->private_data;
1073
Gregory Haskins721eecb2009-05-20 10:30:49 -04001074 kvm_irqfd_release(kvm);
1075
Izik Eidusd39f13b2008-03-30 16:01:25 +03001076 kvm_put_kvm(kvm);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001077 return 0;
1078}
1079
Avi Kivity6aa8b732006-12-10 02:21:36 -08001080/*
Avi Kivity6aa8b732006-12-10 02:21:36 -08001081 * Allocate some memory and give it an address in the guest physical address
1082 * space.
1083 *
1084 * Discontiguous memory is allowed, mostly for framebuffers.
Sheng Yangf78e0e22007-10-29 09:40:42 +08001085 *
Marcelo Tosatti10589a42007-12-20 19:18:22 -05001086 * Must be called holding mmap_sem for write.
Avi Kivity6aa8b732006-12-10 02:21:36 -08001087 */
Sheng Yangf78e0e22007-10-29 09:40:42 +08001088int __kvm_set_memory_region(struct kvm *kvm,
1089 struct kvm_userspace_memory_region *mem,
1090 int user_alloc)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001091{
1092 int r;
1093 gfn_t base_gfn;
Heiko Carstens28bcb112009-09-03 17:35:35 +02001094 unsigned long npages;
1095 unsigned long i;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001096 struct kvm_memory_slot *memslot;
1097 struct kvm_memory_slot old, new;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001098
1099 r = -EINVAL;
1100 /* General sanity checks */
1101 if (mem->memory_size & (PAGE_SIZE - 1))
1102 goto out;
1103 if (mem->guest_phys_addr & (PAGE_SIZE - 1))
1104 goto out;
Sheng Yange7cacd42008-11-11 15:30:40 +08001105 if (user_alloc && (mem->userspace_addr & (PAGE_SIZE - 1)))
Hollis Blanchard78749802008-11-07 13:32:12 -06001106 goto out;
Izik Eiduse0d62c72007-10-24 23:57:46 +02001107 if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001108 goto out;
1109 if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
1110 goto out;
1111
1112 memslot = &kvm->memslots[mem->slot];
1113 base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
1114 npages = mem->memory_size >> PAGE_SHIFT;
1115
1116 if (!npages)
1117 mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES;
1118
Avi Kivity6aa8b732006-12-10 02:21:36 -08001119 new = old = *memslot;
1120
1121 new.base_gfn = base_gfn;
1122 new.npages = npages;
1123 new.flags = mem->flags;
1124
1125 /* Disallow changing a memory slot's size. */
1126 r = -EINVAL;
1127 if (npages && old.npages && npages != old.npages)
Sheng Yangf78e0e22007-10-29 09:40:42 +08001128 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001129
1130 /* Check for overlaps */
1131 r = -EEXIST;
1132 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
1133 struct kvm_memory_slot *s = &kvm->memslots[i];
1134
Jan Kiszka4cd481f2009-04-13 11:59:32 +02001135 if (s == memslot || !s->npages)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001136 continue;
1137 if (!((base_gfn + npages <= s->base_gfn) ||
1138 (base_gfn >= s->base_gfn + s->npages)))
Sheng Yangf78e0e22007-10-29 09:40:42 +08001139 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001140 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08001141
Avi Kivity6aa8b732006-12-10 02:21:36 -08001142 /* Free page dirty bitmap if unneeded */
1143 if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
Al Viro8b6d44c2007-02-09 16:38:40 +00001144 new.dirty_bitmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001145
1146 r = -ENOMEM;
1147
1148 /* Allocate if a slot is being created */
Carsten Otteeff01142008-06-27 15:05:31 +02001149#ifndef CONFIG_S390
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001150 if (npages && !new.rmap) {
Mike Dayd77c26f2007-10-08 09:02:08 -04001151 new.rmap = vmalloc(npages * sizeof(struct page *));
Izik Eidus290fc382007-09-27 14:11:22 +02001152
1153 if (!new.rmap)
Sheng Yangf78e0e22007-10-29 09:40:42 +08001154 goto out_free;
Izik Eidus290fc382007-09-27 14:11:22 +02001155
Izik Eidus290fc382007-09-27 14:11:22 +02001156 memset(new.rmap, 0, npages * sizeof(*new.rmap));
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001157
Izik Eidus80b14b52007-10-25 11:54:04 +02001158 new.user_alloc = user_alloc;
Andrea Arcangeli604b38a2008-07-25 16:32:03 +02001159 /*
1160 * hva_to_rmmap() serialzies with the mmu_lock and to be
1161 * safe it has to ignore memslots with !user_alloc &&
1162 * !userspace_addr.
1163 */
1164 if (user_alloc)
1165 new.userspace_addr = mem->userspace_addr;
1166 else
1167 new.userspace_addr = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001168 }
Joerg Roedelec04b262009-06-19 15:16:23 +02001169 if (!npages)
1170 goto skip_lpage;
Marcelo Tosatti05da4552008-02-23 11:44:30 -03001171
Joerg Roedelec04b262009-06-19 15:16:23 +02001172 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) {
Heiko Carstens28bcb112009-09-03 17:35:35 +02001173 unsigned long ugfn;
1174 unsigned long j;
1175 int lpages;
Joerg Roedelec04b262009-06-19 15:16:23 +02001176 int level = i + 2;
Marcelo Tosatti05da4552008-02-23 11:44:30 -03001177
Joerg Roedelec04b262009-06-19 15:16:23 +02001178 /* Avoid unused variable warning if no large pages */
1179 (void)level;
1180
1181 if (new.lpage_info[i])
1182 continue;
1183
1184 lpages = 1 + (base_gfn + npages - 1) /
1185 KVM_PAGES_PER_HPAGE(level);
1186 lpages -= base_gfn / KVM_PAGES_PER_HPAGE(level);
1187
1188 new.lpage_info[i] = vmalloc(lpages * sizeof(*new.lpage_info[i]));
1189
1190 if (!new.lpage_info[i])
Marcelo Tosatti05da4552008-02-23 11:44:30 -03001191 goto out_free;
1192
Joerg Roedelec04b262009-06-19 15:16:23 +02001193 memset(new.lpage_info[i], 0,
1194 lpages * sizeof(*new.lpage_info[i]));
Marcelo Tosatti05da4552008-02-23 11:44:30 -03001195
Joerg Roedelec04b262009-06-19 15:16:23 +02001196 if (base_gfn % KVM_PAGES_PER_HPAGE(level))
1197 new.lpage_info[i][0].write_count = 1;
1198 if ((base_gfn+npages) % KVM_PAGES_PER_HPAGE(level))
1199 new.lpage_info[i][lpages - 1].write_count = 1;
Avi Kivityac045272009-06-08 15:52:39 +03001200 ugfn = new.userspace_addr >> PAGE_SHIFT;
1201 /*
1202 * If the gfn and userspace address are not aligned wrt each
Marcelo Tosatti54dee992009-06-11 12:07:44 -03001203 * other, or if explicitly asked to, disable large page
1204 * support for this slot
Avi Kivityac045272009-06-08 15:52:39 +03001205 */
Joerg Roedelec04b262009-06-19 15:16:23 +02001206 if ((base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1) ||
Marcelo Tosatti54dee992009-06-11 12:07:44 -03001207 !largepages_enabled)
Joerg Roedelec04b262009-06-19 15:16:23 +02001208 for (j = 0; j < lpages; ++j)
1209 new.lpage_info[i][j].write_count = 1;
Marcelo Tosatti05da4552008-02-23 11:44:30 -03001210 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08001211
Joerg Roedelec04b262009-06-19 15:16:23 +02001212skip_lpage:
1213
Avi Kivity6aa8b732006-12-10 02:21:36 -08001214 /* Allocate page dirty bitmap if needed */
1215 if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
1216 unsigned dirty_bytes = ALIGN(npages, BITS_PER_LONG) / 8;
1217
1218 new.dirty_bitmap = vmalloc(dirty_bytes);
1219 if (!new.dirty_bitmap)
Sheng Yangf78e0e22007-10-29 09:40:42 +08001220 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001221 memset(new.dirty_bitmap, 0, dirty_bytes);
Izik Eiduse2445842009-06-10 19:23:24 +03001222 if (old.npages)
1223 kvm_arch_flush_shadow(kvm);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001224 }
Christian Borntraeger3eea8432009-06-23 17:24:06 +02001225#else /* not defined CONFIG_S390 */
1226 new.user_alloc = user_alloc;
1227 if (user_alloc)
1228 new.userspace_addr = mem->userspace_addr;
Carsten Otteeff01142008-06-27 15:05:31 +02001229#endif /* not defined CONFIG_S390 */
Avi Kivity6aa8b732006-12-10 02:21:36 -08001230
Marcelo Tosatti34d4cb82008-07-10 20:49:31 -03001231 if (!npages)
1232 kvm_arch_flush_shadow(kvm);
1233
Andrea Arcangeli604b38a2008-07-25 16:32:03 +02001234 spin_lock(&kvm->mmu_lock);
1235 if (mem->slot >= kvm->nmemslots)
1236 kvm->nmemslots = mem->slot + 1;
1237
Avi Kivity6aa8b732006-12-10 02:21:36 -08001238 *memslot = new;
Andrea Arcangeli604b38a2008-07-25 16:32:03 +02001239 spin_unlock(&kvm->mmu_lock);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001240
Zhang Xiantao0de10342007-11-20 16:25:04 +08001241 r = kvm_arch_set_memory_region(kvm, mem, old, user_alloc);
1242 if (r) {
Andrea Arcangeli604b38a2008-07-25 16:32:03 +02001243 spin_lock(&kvm->mmu_lock);
Zhang Xiantao0de10342007-11-20 16:25:04 +08001244 *memslot = old;
Andrea Arcangeli604b38a2008-07-25 16:32:03 +02001245 spin_unlock(&kvm->mmu_lock);
Zhang Xiantao0de10342007-11-20 16:25:04 +08001246 goto out_free;
Zhang Xiantao3ad82a72007-11-20 13:11:38 +08001247 }
1248
Glauber Costa6f897242008-12-03 13:40:51 -02001249 kvm_free_physmem_slot(&old, npages ? &new : NULL);
1250 /* Slot deletion case: we have to update the current slot */
Marcelo Tosattib43b1902009-05-12 18:55:44 -03001251 spin_lock(&kvm->mmu_lock);
Glauber Costa6f897242008-12-03 13:40:51 -02001252 if (!npages)
1253 *memslot = old;
Marcelo Tosattib43b1902009-05-12 18:55:44 -03001254 spin_unlock(&kvm->mmu_lock);
Xiantao Zhang8a98f662008-10-06 13:47:38 +08001255#ifdef CONFIG_DMAR
Ben-Ami Yassour62c476c2008-09-14 03:48:28 +03001256 /* map the pages in iommu page table */
1257 r = kvm_iommu_map_pages(kvm, base_gfn, npages);
1258 if (r)
1259 goto out;
Xiantao Zhang8a98f662008-10-06 13:47:38 +08001260#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -08001261 return 0;
1262
Sheng Yangf78e0e22007-10-29 09:40:42 +08001263out_free:
Avi Kivity6aa8b732006-12-10 02:21:36 -08001264 kvm_free_physmem_slot(&new, &old);
1265out:
1266 return r;
Izik Eidus210c7c42007-10-24 23:52:57 +02001267
1268}
Sheng Yangf78e0e22007-10-29 09:40:42 +08001269EXPORT_SYMBOL_GPL(__kvm_set_memory_region);
1270
1271int kvm_set_memory_region(struct kvm *kvm,
1272 struct kvm_userspace_memory_region *mem,
1273 int user_alloc)
1274{
1275 int r;
1276
Izik Eidus72dc67a2008-02-10 18:04:15 +02001277 down_write(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +08001278 r = __kvm_set_memory_region(kvm, mem, user_alloc);
Izik Eidus72dc67a2008-02-10 18:04:15 +02001279 up_write(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +08001280 return r;
1281}
Izik Eidus210c7c42007-10-24 23:52:57 +02001282EXPORT_SYMBOL_GPL(kvm_set_memory_region);
1283
Carsten Otte1fe779f2007-10-29 16:08:35 +01001284int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
1285 struct
1286 kvm_userspace_memory_region *mem,
1287 int user_alloc)
Izik Eidus210c7c42007-10-24 23:52:57 +02001288{
Izik Eiduse0d62c72007-10-24 23:57:46 +02001289 if (mem->slot >= KVM_MEMORY_SLOTS)
1290 return -EINVAL;
Izik Eidus210c7c42007-10-24 23:52:57 +02001291 return kvm_set_memory_region(kvm, mem, user_alloc);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001292}
1293
Zhang Xiantao5bb064d2007-11-18 20:29:43 +08001294int kvm_get_dirty_log(struct kvm *kvm,
1295 struct kvm_dirty_log *log, int *is_dirty)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001296{
1297 struct kvm_memory_slot *memslot;
1298 int r, i;
1299 int n;
1300 unsigned long any = 0;
1301
Avi Kivity6aa8b732006-12-10 02:21:36 -08001302 r = -EINVAL;
1303 if (log->slot >= KVM_MEMORY_SLOTS)
1304 goto out;
1305
1306 memslot = &kvm->memslots[log->slot];
1307 r = -ENOENT;
1308 if (!memslot->dirty_bitmap)
1309 goto out;
1310
Uri Lublincd1a4a92007-02-22 16:43:09 +02001311 n = ALIGN(memslot->npages, BITS_PER_LONG) / 8;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001312
Uri Lublincd1a4a92007-02-22 16:43:09 +02001313 for (i = 0; !any && i < n/sizeof(long); ++i)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001314 any = memslot->dirty_bitmap[i];
1315
1316 r = -EFAULT;
1317 if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
1318 goto out;
1319
Zhang Xiantao5bb064d2007-11-18 20:29:43 +08001320 if (any)
1321 *is_dirty = 1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001322
1323 r = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001324out:
Avi Kivity6aa8b732006-12-10 02:21:36 -08001325 return r;
1326}
1327
Marcelo Tosatti54dee992009-06-11 12:07:44 -03001328void kvm_disable_largepages(void)
1329{
1330 largepages_enabled = false;
1331}
1332EXPORT_SYMBOL_GPL(kvm_disable_largepages);
1333
Izik Eiduscea7bb22007-10-17 19:17:48 +02001334int is_error_page(struct page *page)
1335{
1336 return page == bad_page;
1337}
1338EXPORT_SYMBOL_GPL(is_error_page);
1339
Anthony Liguori35149e22008-04-02 14:46:56 -05001340int is_error_pfn(pfn_t pfn)
1341{
1342 return pfn == bad_pfn;
1343}
1344EXPORT_SYMBOL_GPL(is_error_pfn);
1345
Izik Eidusf9d46eb2007-11-11 22:02:22 +02001346static inline unsigned long bad_hva(void)
1347{
1348 return PAGE_OFFSET;
1349}
1350
1351int kvm_is_error_hva(unsigned long addr)
1352{
1353 return addr == bad_hva();
1354}
1355EXPORT_SYMBOL_GPL(kvm_is_error_hva);
1356
Izik Eidus28430992008-10-03 17:40:32 +03001357struct kvm_memory_slot *gfn_to_memslot_unaliased(struct kvm *kvm, gfn_t gfn)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001358{
1359 int i;
1360
1361 for (i = 0; i < kvm->nmemslots; ++i) {
1362 struct kvm_memory_slot *memslot = &kvm->memslots[i];
1363
1364 if (gfn >= memslot->base_gfn
1365 && gfn < memslot->base_gfn + memslot->npages)
1366 return memslot;
1367 }
Al Viro8b6d44c2007-02-09 16:38:40 +00001368 return NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001369}
Izik Eidus28430992008-10-03 17:40:32 +03001370EXPORT_SYMBOL_GPL(gfn_to_memslot_unaliased);
Avi Kivitye8207542007-03-30 16:54:30 +03001371
1372struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn)
1373{
1374 gfn = unalias_gfn(kvm, gfn);
Izik Eidus28430992008-10-03 17:40:32 +03001375 return gfn_to_memslot_unaliased(kvm, gfn);
Avi Kivitye8207542007-03-30 16:54:30 +03001376}
Avi Kivity6aa8b732006-12-10 02:21:36 -08001377
Izik Eiduse0d62c72007-10-24 23:57:46 +02001378int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
1379{
1380 int i;
1381
1382 gfn = unalias_gfn(kvm, gfn);
1383 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
1384 struct kvm_memory_slot *memslot = &kvm->memslots[i];
1385
1386 if (gfn >= memslot->base_gfn
1387 && gfn < memslot->base_gfn + memslot->npages)
1388 return 1;
1389 }
1390 return 0;
1391}
1392EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);
1393
Marcelo Tosatti05da4552008-02-23 11:44:30 -03001394unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
Izik Eidus539cb662007-11-11 22:05:04 +02001395{
1396 struct kvm_memory_slot *slot;
1397
1398 gfn = unalias_gfn(kvm, gfn);
Izik Eidus28430992008-10-03 17:40:32 +03001399 slot = gfn_to_memslot_unaliased(kvm, gfn);
Izik Eidus539cb662007-11-11 22:05:04 +02001400 if (!slot)
1401 return bad_hva();
1402 return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE);
1403}
Sheng Yang0d150292008-04-25 21:44:50 +08001404EXPORT_SYMBOL_GPL(gfn_to_hva);
Izik Eidus539cb662007-11-11 22:05:04 +02001405
Anthony Liguori35149e22008-04-02 14:46:56 -05001406pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
Avi Kivity954bbbc2007-03-30 14:02:32 +03001407{
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001408 struct page *page[1];
Izik Eidus539cb662007-11-11 22:05:04 +02001409 unsigned long addr;
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001410 int npages;
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001411 pfn_t pfn;
Avi Kivity954bbbc2007-03-30 14:02:32 +03001412
Avi Kivity60395222007-10-21 11:03:36 +02001413 might_sleep();
1414
Izik Eidus539cb662007-11-11 22:05:04 +02001415 addr = gfn_to_hva(kvm, gfn);
1416 if (kvm_is_error_hva(addr)) {
Izik Eidus8a7ae052007-10-18 11:09:33 +02001417 get_page(bad_page);
Anthony Liguori35149e22008-04-02 14:46:56 -05001418 return page_to_pfn(bad_page);
Izik Eidus8a7ae052007-10-18 11:09:33 +02001419 }
Izik Eidus8a7ae052007-10-18 11:09:33 +02001420
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001421 npages = get_user_pages_fast(addr, 1, 1, page);
Izik Eidus539cb662007-11-11 22:05:04 +02001422
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001423 if (unlikely(npages != 1)) {
1424 struct vm_area_struct *vma;
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001425
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001426 down_read(&current->mm->mmap_sem);
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001427 vma = find_vma(current->mm, addr);
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001428
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001429 if (vma == NULL || addr < vma->vm_start ||
1430 !(vma->vm_flags & VM_PFNMAP)) {
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001431 up_read(&current->mm->mmap_sem);
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001432 get_page(bad_page);
1433 return page_to_pfn(bad_page);
1434 }
1435
1436 pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001437 up_read(&current->mm->mmap_sem);
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001438 BUG_ON(!kvm_is_mmio_pfn(pfn));
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001439 } else
1440 pfn = page_to_pfn(page[0]);
1441
1442 return pfn;
Anthony Liguori35149e22008-04-02 14:46:56 -05001443}
1444
1445EXPORT_SYMBOL_GPL(gfn_to_pfn);
1446
1447struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
1448{
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001449 pfn_t pfn;
1450
1451 pfn = gfn_to_pfn(kvm, gfn);
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001452 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001453 return pfn_to_page(pfn);
1454
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001455 WARN_ON(kvm_is_mmio_pfn(pfn));
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001456
1457 get_page(bad_page);
1458 return bad_page;
Avi Kivity954bbbc2007-03-30 14:02:32 +03001459}
Anthony Liguoriaab61cc2007-10-29 15:15:20 -05001460
Avi Kivity954bbbc2007-03-30 14:02:32 +03001461EXPORT_SYMBOL_GPL(gfn_to_page);
1462
Izik Eidusb4231d62007-11-20 11:49:33 +02001463void kvm_release_page_clean(struct page *page)
1464{
Anthony Liguori35149e22008-04-02 14:46:56 -05001465 kvm_release_pfn_clean(page_to_pfn(page));
Izik Eidusb4231d62007-11-20 11:49:33 +02001466}
1467EXPORT_SYMBOL_GPL(kvm_release_page_clean);
1468
Anthony Liguori35149e22008-04-02 14:46:56 -05001469void kvm_release_pfn_clean(pfn_t pfn)
1470{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001471 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001472 put_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001473}
1474EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);
1475
Izik Eidusb4231d62007-11-20 11:49:33 +02001476void kvm_release_page_dirty(struct page *page)
Izik Eidus8a7ae052007-10-18 11:09:33 +02001477{
Anthony Liguori35149e22008-04-02 14:46:56 -05001478 kvm_release_pfn_dirty(page_to_pfn(page));
Izik Eidus8a7ae052007-10-18 11:09:33 +02001479}
Izik Eidusb4231d62007-11-20 11:49:33 +02001480EXPORT_SYMBOL_GPL(kvm_release_page_dirty);
Izik Eidus8a7ae052007-10-18 11:09:33 +02001481
Anthony Liguori35149e22008-04-02 14:46:56 -05001482void kvm_release_pfn_dirty(pfn_t pfn)
1483{
1484 kvm_set_pfn_dirty(pfn);
1485 kvm_release_pfn_clean(pfn);
1486}
1487EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
1488
1489void kvm_set_page_dirty(struct page *page)
1490{
1491 kvm_set_pfn_dirty(page_to_pfn(page));
1492}
1493EXPORT_SYMBOL_GPL(kvm_set_page_dirty);
1494
1495void kvm_set_pfn_dirty(pfn_t pfn)
1496{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001497 if (!kvm_is_mmio_pfn(pfn)) {
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001498 struct page *page = pfn_to_page(pfn);
1499 if (!PageReserved(page))
1500 SetPageDirty(page);
1501 }
Anthony Liguori35149e22008-04-02 14:46:56 -05001502}
1503EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);
1504
1505void kvm_set_pfn_accessed(pfn_t pfn)
1506{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001507 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001508 mark_page_accessed(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001509}
1510EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);
1511
1512void kvm_get_pfn(pfn_t pfn)
1513{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001514 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001515 get_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001516}
1517EXPORT_SYMBOL_GPL(kvm_get_pfn);
1518
Izik Eidus195aefd2007-10-01 22:14:18 +02001519static int next_segment(unsigned long len, int offset)
1520{
1521 if (len > PAGE_SIZE - offset)
1522 return PAGE_SIZE - offset;
1523 else
1524 return len;
1525}
1526
1527int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
1528 int len)
1529{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001530 int r;
1531 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001532
Izik Eiduse0506bc2007-11-11 22:10:22 +02001533 addr = gfn_to_hva(kvm, gfn);
1534 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001535 return -EFAULT;
Izik Eiduse0506bc2007-11-11 22:10:22 +02001536 r = copy_from_user(data, (void __user *)addr + offset, len);
1537 if (r)
1538 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001539 return 0;
1540}
1541EXPORT_SYMBOL_GPL(kvm_read_guest_page);
1542
1543int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len)
1544{
1545 gfn_t gfn = gpa >> PAGE_SHIFT;
1546 int seg;
1547 int offset = offset_in_page(gpa);
1548 int ret;
1549
1550 while ((seg = next_segment(len, offset)) != 0) {
1551 ret = kvm_read_guest_page(kvm, gfn, data, offset, seg);
1552 if (ret < 0)
1553 return ret;
1554 offset = 0;
1555 len -= seg;
1556 data += seg;
1557 ++gfn;
1558 }
1559 return 0;
1560}
1561EXPORT_SYMBOL_GPL(kvm_read_guest);
1562
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001563int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
1564 unsigned long len)
1565{
1566 int r;
1567 unsigned long addr;
1568 gfn_t gfn = gpa >> PAGE_SHIFT;
1569 int offset = offset_in_page(gpa);
1570
1571 addr = gfn_to_hva(kvm, gfn);
1572 if (kvm_is_error_hva(addr))
1573 return -EFAULT;
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001574 pagefault_disable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001575 r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001576 pagefault_enable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001577 if (r)
1578 return -EFAULT;
1579 return 0;
1580}
1581EXPORT_SYMBOL(kvm_read_guest_atomic);
1582
Izik Eidus195aefd2007-10-01 22:14:18 +02001583int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
1584 int offset, int len)
1585{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001586 int r;
1587 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001588
Izik Eiduse0506bc2007-11-11 22:10:22 +02001589 addr = gfn_to_hva(kvm, gfn);
1590 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001591 return -EFAULT;
Izik Eiduse0506bc2007-11-11 22:10:22 +02001592 r = copy_to_user((void __user *)addr + offset, data, len);
1593 if (r)
1594 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001595 mark_page_dirty(kvm, gfn);
1596 return 0;
1597}
1598EXPORT_SYMBOL_GPL(kvm_write_guest_page);
1599
1600int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
1601 unsigned long len)
1602{
1603 gfn_t gfn = gpa >> PAGE_SHIFT;
1604 int seg;
1605 int offset = offset_in_page(gpa);
1606 int ret;
1607
1608 while ((seg = next_segment(len, offset)) != 0) {
1609 ret = kvm_write_guest_page(kvm, gfn, data, offset, seg);
1610 if (ret < 0)
1611 return ret;
1612 offset = 0;
1613 len -= seg;
1614 data += seg;
1615 ++gfn;
1616 }
1617 return 0;
1618}
1619
1620int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
1621{
Izik Eidus3e021bf2007-11-19 11:16:57 +02001622 return kvm_write_guest_page(kvm, gfn, empty_zero_page, offset, len);
Izik Eidus195aefd2007-10-01 22:14:18 +02001623}
1624EXPORT_SYMBOL_GPL(kvm_clear_guest_page);
1625
1626int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len)
1627{
1628 gfn_t gfn = gpa >> PAGE_SHIFT;
1629 int seg;
1630 int offset = offset_in_page(gpa);
1631 int ret;
1632
1633 while ((seg = next_segment(len, offset)) != 0) {
1634 ret = kvm_clear_guest_page(kvm, gfn, offset, seg);
1635 if (ret < 0)
1636 return ret;
1637 offset = 0;
1638 len -= seg;
1639 ++gfn;
1640 }
1641 return 0;
1642}
1643EXPORT_SYMBOL_GPL(kvm_clear_guest);
1644
Avi Kivity6aa8b732006-12-10 02:21:36 -08001645void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
1646{
Nguyen Anh Quynh31389942007-06-05 10:35:19 +03001647 struct kvm_memory_slot *memslot;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001648
Uri Lublin3b6fff12007-10-30 10:42:09 +02001649 gfn = unalias_gfn(kvm, gfn);
Izik Eidus28430992008-10-03 17:40:32 +03001650 memslot = gfn_to_memslot_unaliased(kvm, gfn);
Rusty Russell7e9d6192007-07-31 20:41:14 +10001651 if (memslot && memslot->dirty_bitmap) {
1652 unsigned long rel_gfn = gfn - memslot->base_gfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001653
Rusty Russell7e9d6192007-07-31 20:41:14 +10001654 /* avoid RMW */
1655 if (!test_bit(rel_gfn, memslot->dirty_bitmap))
1656 set_bit(rel_gfn, memslot->dirty_bitmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001657 }
1658}
1659
Eddie Dongb6958ce2007-07-18 12:15:21 +03001660/*
1661 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
1662 */
Hollis Blanchard8776e512007-10-31 17:24:24 -05001663void kvm_vcpu_block(struct kvm_vcpu *vcpu)
Eddie Dongb6958ce2007-07-18 12:15:21 +03001664{
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001665 DEFINE_WAIT(wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001666
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001667 for (;;) {
1668 prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001669
Gleb Natapova1b37102009-07-09 15:33:52 +03001670 if (kvm_arch_vcpu_runnable(vcpu)) {
Marcelo Tosattid7690172008-09-08 15:23:48 -03001671 set_bit(KVM_REQ_UNHALT, &vcpu->requests);
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001672 break;
Marcelo Tosattid7690172008-09-08 15:23:48 -03001673 }
Gleb Natapov09cec752009-03-23 15:11:44 +02001674 if (kvm_cpu_has_pending_timer(vcpu))
1675 break;
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001676 if (signal_pending(current))
1677 break;
1678
Eddie Dongb6958ce2007-07-18 12:15:21 +03001679 vcpu_put(vcpu);
1680 schedule();
1681 vcpu_load(vcpu);
1682 }
1683
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001684 finish_wait(&vcpu->wq, &wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001685}
1686
Avi Kivity6aa8b732006-12-10 02:21:36 -08001687void kvm_resched(struct kvm_vcpu *vcpu)
1688{
Yaozu Dong3fca0362007-04-25 16:49:19 +03001689 if (!need_resched())
1690 return;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001691 cond_resched();
Avi Kivity6aa8b732006-12-10 02:21:36 -08001692}
1693EXPORT_SYMBOL_GPL(kvm_resched);
1694
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001695static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001696{
1697 struct kvm_vcpu *vcpu = vma->vm_file->private_data;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001698 struct page *page;
1699
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001700 if (vmf->pgoff == 0)
Avi Kivity039576c2007-03-20 12:46:50 +02001701 page = virt_to_page(vcpu->run);
Avi Kivity09566762008-01-23 18:14:23 +02001702#ifdef CONFIG_X86
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001703 else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
Zhang Xiantaoad312c72007-12-13 23:50:52 +08001704 page = virt_to_page(vcpu->arch.pio_data);
Avi Kivity09566762008-01-23 18:14:23 +02001705#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02001706#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1707 else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
1708 page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
1709#endif
Avi Kivity039576c2007-03-20 12:46:50 +02001710 else
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001711 return VM_FAULT_SIGBUS;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001712 get_page(page);
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001713 vmf->page = page;
1714 return 0;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001715}
1716
1717static struct vm_operations_struct kvm_vcpu_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001718 .fault = kvm_vcpu_fault,
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001719};
1720
1721static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma)
1722{
1723 vma->vm_ops = &kvm_vcpu_vm_ops;
1724 return 0;
1725}
1726
Avi Kivitybccf2152007-02-21 18:04:26 +02001727static int kvm_vcpu_release(struct inode *inode, struct file *filp)
1728{
1729 struct kvm_vcpu *vcpu = filp->private_data;
1730
Al Viro66c0b392008-04-19 20:33:56 +01001731 kvm_put_kvm(vcpu->kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001732 return 0;
1733}
1734
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01001735static struct file_operations kvm_vcpu_fops = {
Avi Kivitybccf2152007-02-21 18:04:26 +02001736 .release = kvm_vcpu_release,
1737 .unlocked_ioctl = kvm_vcpu_ioctl,
1738 .compat_ioctl = kvm_vcpu_ioctl,
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001739 .mmap = kvm_vcpu_mmap,
Avi Kivitybccf2152007-02-21 18:04:26 +02001740};
1741
1742/*
1743 * Allocates an inode for the vcpu.
1744 */
1745static int create_vcpu_fd(struct kvm_vcpu *vcpu)
1746{
Gleb Natapov73880c82009-06-09 15:56:28 +03001747 return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, 0);
Avi Kivitybccf2152007-02-21 18:04:26 +02001748}
1749
Avi Kivityc5ea7662007-02-20 18:41:05 +02001750/*
1751 * Creates some virtual cpus. Good luck creating more than one.
1752 */
Gleb Natapov73880c82009-06-09 15:56:28 +03001753static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
Avi Kivityc5ea7662007-02-20 18:41:05 +02001754{
1755 int r;
Gleb Natapov988a2ca2009-06-09 15:56:29 +03001756 struct kvm_vcpu *vcpu, *v;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001757
Gleb Natapov73880c82009-06-09 15:56:28 +03001758 vcpu = kvm_arch_vcpu_create(kvm, id);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001759 if (IS_ERR(vcpu))
1760 return PTR_ERR(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001761
Avi Kivity15ad7142007-07-11 18:17:21 +03001762 preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);
1763
Avi Kivity26e52152007-11-20 15:30:24 +02001764 r = kvm_arch_vcpu_setup(vcpu);
1765 if (r)
Glauber Costa7d8fece2008-09-17 23:16:59 -03001766 return r;
Avi Kivity26e52152007-11-20 15:30:24 +02001767
Shaohua Li11ec2802007-07-23 14:51:37 +08001768 mutex_lock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03001769 if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) {
1770 r = -EINVAL;
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001771 goto vcpu_destroy;
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001772 }
Gleb Natapov73880c82009-06-09 15:56:28 +03001773
Gleb Natapov988a2ca2009-06-09 15:56:29 +03001774 kvm_for_each_vcpu(r, v, kvm)
1775 if (v->vcpu_id == id) {
Gleb Natapov73880c82009-06-09 15:56:28 +03001776 r = -EEXIST;
1777 goto vcpu_destroy;
1778 }
1779
1780 BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001781
1782 /* Now it's all set up, let userspace reach it */
Al Viro66c0b392008-04-19 20:33:56 +01001783 kvm_get_kvm(kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001784 r = create_vcpu_fd(vcpu);
Gleb Natapov73880c82009-06-09 15:56:28 +03001785 if (r < 0) {
1786 kvm_put_kvm(kvm);
1787 goto vcpu_destroy;
1788 }
1789
1790 kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu;
1791 smp_wmb();
1792 atomic_inc(&kvm->online_vcpus);
1793
1794#ifdef CONFIG_KVM_APIC_ARCHITECTURE
1795 if (kvm->bsp_vcpu_id == id)
1796 kvm->bsp_vcpu = vcpu;
1797#endif
1798 mutex_unlock(&kvm->lock);
Avi Kivitybccf2152007-02-21 18:04:26 +02001799 return r;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001800
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001801vcpu_destroy:
Glauber Costa7d8fece2008-09-17 23:16:59 -03001802 mutex_unlock(&kvm->lock);
Hollis Blanchardd40ccc62007-11-19 14:04:43 -06001803 kvm_arch_vcpu_destroy(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001804 return r;
1805}
1806
Avi Kivity1961d272007-03-05 19:46:05 +02001807static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset)
1808{
1809 if (sigset) {
1810 sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP));
1811 vcpu->sigset_active = 1;
1812 vcpu->sigset = *sigset;
1813 } else
1814 vcpu->sigset_active = 0;
1815 return 0;
1816}
1817
Sheng Yangc1e01512009-02-25 17:22:26 +08001818#ifdef __KVM_HAVE_MSIX
1819static int kvm_vm_ioctl_set_msix_nr(struct kvm *kvm,
1820 struct kvm_assigned_msix_nr *entry_nr)
1821{
1822 int r = 0;
1823 struct kvm_assigned_dev_kernel *adev;
1824
1825 mutex_lock(&kvm->lock);
1826
1827 adev = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head,
1828 entry_nr->assigned_dev_id);
1829 if (!adev) {
1830 r = -EINVAL;
1831 goto msix_nr_out;
1832 }
1833
1834 if (adev->entries_nr == 0) {
1835 adev->entries_nr = entry_nr->entry_nr;
1836 if (adev->entries_nr == 0 ||
1837 adev->entries_nr >= KVM_MAX_MSIX_PER_DEV) {
1838 r = -EINVAL;
1839 goto msix_nr_out;
1840 }
1841
1842 adev->host_msix_entries = kzalloc(sizeof(struct msix_entry) *
1843 entry_nr->entry_nr,
1844 GFP_KERNEL);
1845 if (!adev->host_msix_entries) {
1846 r = -ENOMEM;
1847 goto msix_nr_out;
1848 }
1849 adev->guest_msix_entries = kzalloc(
1850 sizeof(struct kvm_guest_msix_entry) *
1851 entry_nr->entry_nr, GFP_KERNEL);
1852 if (!adev->guest_msix_entries) {
1853 kfree(adev->host_msix_entries);
1854 r = -ENOMEM;
1855 goto msix_nr_out;
1856 }
1857 } else /* Not allowed set MSI-X number twice */
1858 r = -EINVAL;
1859msix_nr_out:
1860 mutex_unlock(&kvm->lock);
1861 return r;
1862}
1863
1864static int kvm_vm_ioctl_set_msix_entry(struct kvm *kvm,
1865 struct kvm_assigned_msix_entry *entry)
1866{
1867 int r = 0, i;
1868 struct kvm_assigned_dev_kernel *adev;
1869
1870 mutex_lock(&kvm->lock);
1871
1872 adev = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head,
1873 entry->assigned_dev_id);
1874
1875 if (!adev) {
1876 r = -EINVAL;
1877 goto msix_entry_out;
1878 }
1879
1880 for (i = 0; i < adev->entries_nr; i++)
1881 if (adev->guest_msix_entries[i].vector == 0 ||
1882 adev->guest_msix_entries[i].entry == entry->entry) {
1883 adev->guest_msix_entries[i].entry = entry->entry;
1884 adev->guest_msix_entries[i].vector = entry->gsi;
1885 adev->host_msix_entries[i].entry = entry->entry;
1886 break;
1887 }
1888 if (i == adev->entries_nr) {
1889 r = -ENOSPC;
1890 goto msix_entry_out;
1891 }
1892
1893msix_entry_out:
1894 mutex_unlock(&kvm->lock);
1895
1896 return r;
1897}
1898#endif
1899
Avi Kivitybccf2152007-02-21 18:04:26 +02001900static long kvm_vcpu_ioctl(struct file *filp,
1901 unsigned int ioctl, unsigned long arg)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001902{
Avi Kivitybccf2152007-02-21 18:04:26 +02001903 struct kvm_vcpu *vcpu = filp->private_data;
Al Viro2f3669872007-02-09 16:38:35 +00001904 void __user *argp = (void __user *)arg;
Carsten Otte313a3dc2007-10-11 19:16:52 +02001905 int r;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001906 struct kvm_fpu *fpu = NULL;
1907 struct kvm_sregs *kvm_sregs = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001908
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001909 if (vcpu->kvm->mm != current->mm)
1910 return -EIO;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001911 switch (ioctl) {
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001912 case KVM_RUN:
Avi Kivityf0fe5102007-03-07 13:11:17 +02001913 r = -EINVAL;
1914 if (arg)
1915 goto out;
Hollis Blanchardb6c7a5d2007-11-01 14:16:10 -05001916 r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001917 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001918 case KVM_GET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001919 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001920
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001921 r = -ENOMEM;
1922 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
1923 if (!kvm_regs)
1924 goto out;
1925 r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001926 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001927 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001928 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001929 if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
1930 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001931 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001932out_free1:
1933 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001934 break;
1935 }
1936 case KVM_SET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001937 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001938
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001939 r = -ENOMEM;
1940 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
1941 if (!kvm_regs)
1942 goto out;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001943 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001944 if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs)))
1945 goto out_free2;
1946 r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001947 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001948 goto out_free2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001949 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001950out_free2:
1951 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001952 break;
1953 }
1954 case KVM_GET_SREGS: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001955 kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL);
1956 r = -ENOMEM;
1957 if (!kvm_sregs)
1958 goto out;
1959 r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001960 if (r)
1961 goto out;
1962 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001963 if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001964 goto out;
1965 r = 0;
1966 break;
1967 }
1968 case KVM_SET_SREGS: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001969 kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL);
1970 r = -ENOMEM;
1971 if (!kvm_sregs)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001972 goto out;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001973 r = -EFAULT;
1974 if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs)))
1975 goto out;
1976 r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001977 if (r)
1978 goto out;
1979 r = 0;
1980 break;
1981 }
Marcelo Tosatti62d9f0d2008-04-11 13:24:45 -03001982 case KVM_GET_MP_STATE: {
1983 struct kvm_mp_state mp_state;
1984
1985 r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state);
1986 if (r)
1987 goto out;
1988 r = -EFAULT;
1989 if (copy_to_user(argp, &mp_state, sizeof mp_state))
1990 goto out;
1991 r = 0;
1992 break;
1993 }
1994 case KVM_SET_MP_STATE: {
1995 struct kvm_mp_state mp_state;
1996
1997 r = -EFAULT;
1998 if (copy_from_user(&mp_state, argp, sizeof mp_state))
1999 goto out;
2000 r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
2001 if (r)
2002 goto out;
2003 r = 0;
2004 break;
2005 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08002006 case KVM_TRANSLATE: {
2007 struct kvm_translation tr;
2008
2009 r = -EFAULT;
Al Viro2f3669872007-02-09 16:38:35 +00002010 if (copy_from_user(&tr, argp, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08002011 goto out;
Zhang Xiantao8b006792007-11-16 13:05:55 +08002012 r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002013 if (r)
2014 goto out;
2015 r = -EFAULT;
Al Viro2f3669872007-02-09 16:38:35 +00002016 if (copy_to_user(argp, &tr, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08002017 goto out;
2018 r = 0;
2019 break;
2020 }
Jan Kiszkad0bfb942008-12-15 13:52:10 +01002021 case KVM_SET_GUEST_DEBUG: {
2022 struct kvm_guest_debug dbg;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002023
2024 r = -EFAULT;
Al Viro2f3669872007-02-09 16:38:35 +00002025 if (copy_from_user(&dbg, argp, sizeof dbg))
Avi Kivity6aa8b732006-12-10 02:21:36 -08002026 goto out;
Jan Kiszkad0bfb942008-12-15 13:52:10 +01002027 r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002028 if (r)
2029 goto out;
2030 r = 0;
2031 break;
2032 }
Avi Kivity1961d272007-03-05 19:46:05 +02002033 case KVM_SET_SIGNAL_MASK: {
2034 struct kvm_signal_mask __user *sigmask_arg = argp;
2035 struct kvm_signal_mask kvm_sigmask;
2036 sigset_t sigset, *p;
2037
2038 p = NULL;
2039 if (argp) {
2040 r = -EFAULT;
2041 if (copy_from_user(&kvm_sigmask, argp,
2042 sizeof kvm_sigmask))
2043 goto out;
2044 r = -EINVAL;
2045 if (kvm_sigmask.len != sizeof sigset)
2046 goto out;
2047 r = -EFAULT;
2048 if (copy_from_user(&sigset, sigmask_arg->sigset,
2049 sizeof sigset))
2050 goto out;
2051 p = &sigset;
2052 }
2053 r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
2054 break;
2055 }
Avi Kivityb8836732007-04-01 16:34:31 +03002056 case KVM_GET_FPU: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07002057 fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL);
2058 r = -ENOMEM;
2059 if (!fpu)
2060 goto out;
2061 r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03002062 if (r)
2063 goto out;
2064 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07002065 if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
Avi Kivityb8836732007-04-01 16:34:31 +03002066 goto out;
2067 r = 0;
2068 break;
2069 }
2070 case KVM_SET_FPU: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07002071 fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL);
2072 r = -ENOMEM;
2073 if (!fpu)
Avi Kivityb8836732007-04-01 16:34:31 +03002074 goto out;
Dave Hansenfa3795a2008-08-11 10:01:46 -07002075 r = -EFAULT;
2076 if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu)))
2077 goto out;
2078 r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03002079 if (r)
2080 goto out;
2081 r = 0;
2082 break;
2083 }
Avi Kivitybccf2152007-02-21 18:04:26 +02002084 default:
Carsten Otte313a3dc2007-10-11 19:16:52 +02002085 r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
Avi Kivitybccf2152007-02-21 18:04:26 +02002086 }
2087out:
Dave Hansenfa3795a2008-08-11 10:01:46 -07002088 kfree(fpu);
2089 kfree(kvm_sregs);
Avi Kivitybccf2152007-02-21 18:04:26 +02002090 return r;
2091}
2092
2093static long kvm_vm_ioctl(struct file *filp,
2094 unsigned int ioctl, unsigned long arg)
2095{
2096 struct kvm *kvm = filp->private_data;
2097 void __user *argp = (void __user *)arg;
Carsten Otte1fe779f2007-10-29 16:08:35 +01002098 int r;
Avi Kivitybccf2152007-02-21 18:04:26 +02002099
Avi Kivity6d4e4c42007-11-21 16:41:05 +02002100 if (kvm->mm != current->mm)
2101 return -EIO;
Avi Kivitybccf2152007-02-21 18:04:26 +02002102 switch (ioctl) {
2103 case KVM_CREATE_VCPU:
2104 r = kvm_vm_ioctl_create_vcpu(kvm, arg);
2105 if (r < 0)
2106 goto out;
2107 break;
Izik Eidus6fc138d2007-10-09 19:20:39 +02002108 case KVM_SET_USER_MEMORY_REGION: {
2109 struct kvm_userspace_memory_region kvm_userspace_mem;
2110
2111 r = -EFAULT;
2112 if (copy_from_user(&kvm_userspace_mem, argp,
2113 sizeof kvm_userspace_mem))
2114 goto out;
2115
2116 r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002117 if (r)
2118 goto out;
2119 break;
2120 }
2121 case KVM_GET_DIRTY_LOG: {
2122 struct kvm_dirty_log log;
2123
2124 r = -EFAULT;
Al Viro2f3669872007-02-09 16:38:35 +00002125 if (copy_from_user(&log, argp, sizeof log))
Avi Kivity6aa8b732006-12-10 02:21:36 -08002126 goto out;
Avi Kivity2c6f5df2007-02-20 18:27:58 +02002127 r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002128 if (r)
2129 goto out;
2130 break;
2131 }
Laurent Vivier5f94c172008-05-30 16:05:54 +02002132#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
2133 case KVM_REGISTER_COALESCED_MMIO: {
2134 struct kvm_coalesced_mmio_zone zone;
2135 r = -EFAULT;
2136 if (copy_from_user(&zone, argp, sizeof zone))
2137 goto out;
2138 r = -ENXIO;
2139 r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
2140 if (r)
2141 goto out;
2142 r = 0;
2143 break;
2144 }
2145 case KVM_UNREGISTER_COALESCED_MMIO: {
2146 struct kvm_coalesced_mmio_zone zone;
2147 r = -EFAULT;
2148 if (copy_from_user(&zone, argp, sizeof zone))
2149 goto out;
2150 r = -ENXIO;
2151 r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
2152 if (r)
2153 goto out;
2154 r = 0;
2155 break;
2156 }
2157#endif
Xiantao Zhang8a98f662008-10-06 13:47:38 +08002158#ifdef KVM_CAP_DEVICE_ASSIGNMENT
2159 case KVM_ASSIGN_PCI_DEVICE: {
2160 struct kvm_assigned_pci_dev assigned_dev;
2161
2162 r = -EFAULT;
2163 if (copy_from_user(&assigned_dev, argp, sizeof assigned_dev))
2164 goto out;
2165 r = kvm_vm_ioctl_assign_device(kvm, &assigned_dev);
2166 if (r)
2167 goto out;
2168 break;
2169 }
2170 case KVM_ASSIGN_IRQ: {
Sheng Yange56d5322009-03-12 21:45:39 +08002171 r = -EOPNOTSUPP;
2172 break;
2173 }
2174#ifdef KVM_CAP_ASSIGN_DEV_IRQ
2175 case KVM_ASSIGN_DEV_IRQ: {
Xiantao Zhang8a98f662008-10-06 13:47:38 +08002176 struct kvm_assigned_irq assigned_irq;
2177
2178 r = -EFAULT;
2179 if (copy_from_user(&assigned_irq, argp, sizeof assigned_irq))
2180 goto out;
2181 r = kvm_vm_ioctl_assign_irq(kvm, &assigned_irq);
2182 if (r)
2183 goto out;
2184 break;
2185 }
Sheng Yange56d5322009-03-12 21:45:39 +08002186 case KVM_DEASSIGN_DEV_IRQ: {
2187 struct kvm_assigned_irq assigned_irq;
2188
2189 r = -EFAULT;
2190 if (copy_from_user(&assigned_irq, argp, sizeof assigned_irq))
2191 goto out;
2192 r = kvm_vm_ioctl_deassign_dev_irq(kvm, &assigned_irq);
2193 if (r)
2194 goto out;
2195 break;
2196 }
2197#endif
Xiantao Zhang8a98f662008-10-06 13:47:38 +08002198#endif
Weidong Han0a920352008-12-02 21:24:23 +08002199#ifdef KVM_CAP_DEVICE_DEASSIGNMENT
2200 case KVM_DEASSIGN_PCI_DEVICE: {
2201 struct kvm_assigned_pci_dev assigned_dev;
2202
2203 r = -EFAULT;
2204 if (copy_from_user(&assigned_dev, argp, sizeof assigned_dev))
2205 goto out;
2206 r = kvm_vm_ioctl_deassign_device(kvm, &assigned_dev);
2207 if (r)
2208 goto out;
2209 break;
2210 }
2211#endif
Avi Kivity399ec802008-11-19 13:58:46 +02002212#ifdef KVM_CAP_IRQ_ROUTING
2213 case KVM_SET_GSI_ROUTING: {
2214 struct kvm_irq_routing routing;
2215 struct kvm_irq_routing __user *urouting;
2216 struct kvm_irq_routing_entry *entries;
2217
2218 r = -EFAULT;
2219 if (copy_from_user(&routing, argp, sizeof(routing)))
2220 goto out;
2221 r = -EINVAL;
2222 if (routing.nr >= KVM_MAX_IRQ_ROUTES)
2223 goto out;
2224 if (routing.flags)
2225 goto out;
2226 r = -ENOMEM;
2227 entries = vmalloc(routing.nr * sizeof(*entries));
2228 if (!entries)
2229 goto out;
2230 r = -EFAULT;
2231 urouting = argp;
2232 if (copy_from_user(entries, urouting->entries,
2233 routing.nr * sizeof(*entries)))
2234 goto out_free_irq_routing;
2235 r = kvm_set_irq_routing(kvm, entries, routing.nr,
2236 routing.flags);
2237 out_free_irq_routing:
2238 vfree(entries);
2239 break;
2240 }
Avi Kivity6621fbc2009-08-10 17:00:52 +03002241#endif /* KVM_CAP_IRQ_ROUTING */
Sheng Yangc1e01512009-02-25 17:22:26 +08002242#ifdef __KVM_HAVE_MSIX
2243 case KVM_ASSIGN_SET_MSIX_NR: {
2244 struct kvm_assigned_msix_nr entry_nr;
2245 r = -EFAULT;
2246 if (copy_from_user(&entry_nr, argp, sizeof entry_nr))
2247 goto out;
2248 r = kvm_vm_ioctl_set_msix_nr(kvm, &entry_nr);
2249 if (r)
2250 goto out;
2251 break;
2252 }
2253 case KVM_ASSIGN_SET_MSIX_ENTRY: {
2254 struct kvm_assigned_msix_entry entry;
2255 r = -EFAULT;
2256 if (copy_from_user(&entry, argp, sizeof entry))
2257 goto out;
2258 r = kvm_vm_ioctl_set_msix_entry(kvm, &entry);
2259 if (r)
2260 goto out;
2261 break;
2262 }
Avi Kivity399ec802008-11-19 13:58:46 +02002263#endif
Gregory Haskins721eecb2009-05-20 10:30:49 -04002264 case KVM_IRQFD: {
2265 struct kvm_irqfd data;
2266
2267 r = -EFAULT;
2268 if (copy_from_user(&data, argp, sizeof data))
2269 goto out;
2270 r = kvm_irqfd(kvm, data.fd, data.gsi, data.flags);
2271 break;
2272 }
Gregory Haskinsd34e6b12009-07-07 17:08:49 -04002273 case KVM_IOEVENTFD: {
2274 struct kvm_ioeventfd data;
2275
2276 r = -EFAULT;
2277 if (copy_from_user(&data, argp, sizeof data))
2278 goto out;
2279 r = kvm_ioeventfd(kvm, &data);
2280 break;
2281 }
Gleb Natapov73880c82009-06-09 15:56:28 +03002282#ifdef CONFIG_KVM_APIC_ARCHITECTURE
2283 case KVM_SET_BOOT_CPU_ID:
2284 r = 0;
Marcelo Tosatti894a9c52009-06-23 15:05:14 -03002285 mutex_lock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03002286 if (atomic_read(&kvm->online_vcpus) != 0)
2287 r = -EBUSY;
2288 else
2289 kvm->bsp_vcpu_id = arg;
Marcelo Tosatti894a9c52009-06-23 15:05:14 -03002290 mutex_unlock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03002291 break;
2292#endif
Avi Kivityf17abe92007-02-21 19:28:04 +02002293 default:
Carsten Otte1fe779f2007-10-29 16:08:35 +01002294 r = kvm_arch_vm_ioctl(filp, ioctl, arg);
Avi Kivityf17abe92007-02-21 19:28:04 +02002295 }
2296out:
2297 return r;
2298}
2299
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002300static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivityf17abe92007-02-21 19:28:04 +02002301{
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002302 struct page *page[1];
2303 unsigned long addr;
2304 int npages;
2305 gfn_t gfn = vmf->pgoff;
Avi Kivityf17abe92007-02-21 19:28:04 +02002306 struct kvm *kvm = vma->vm_file->private_data;
Avi Kivityf17abe92007-02-21 19:28:04 +02002307
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002308 addr = gfn_to_hva(kvm, gfn);
2309 if (kvm_is_error_hva(addr))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002310 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002311
2312 npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page,
2313 NULL);
2314 if (unlikely(npages != 1))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002315 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002316
2317 vmf->page = page[0];
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002318 return 0;
Avi Kivityf17abe92007-02-21 19:28:04 +02002319}
2320
2321static struct vm_operations_struct kvm_vm_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002322 .fault = kvm_vm_fault,
Avi Kivityf17abe92007-02-21 19:28:04 +02002323};
2324
2325static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma)
2326{
2327 vma->vm_ops = &kvm_vm_vm_ops;
2328 return 0;
2329}
2330
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01002331static struct file_operations kvm_vm_fops = {
Avi Kivityf17abe92007-02-21 19:28:04 +02002332 .release = kvm_vm_release,
2333 .unlocked_ioctl = kvm_vm_ioctl,
2334 .compat_ioctl = kvm_vm_ioctl,
2335 .mmap = kvm_vm_mmap,
2336};
2337
2338static int kvm_dev_ioctl_create_vm(void)
2339{
Al Viro2030a422008-02-23 06:46:49 -05002340 int fd;
Avi Kivityf17abe92007-02-21 19:28:04 +02002341 struct kvm *kvm;
2342
Avi Kivityf17abe92007-02-21 19:28:04 +02002343 kvm = kvm_create_vm();
Avi Kivityd6d28162007-06-28 08:38:16 -04002344 if (IS_ERR(kvm))
2345 return PTR_ERR(kvm);
Ulrich Drepper7d9dbca2008-07-23 21:29:22 -07002346 fd = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, 0);
Al Viro2030a422008-02-23 06:46:49 -05002347 if (fd < 0)
Al Viro66c0b392008-04-19 20:33:56 +01002348 kvm_put_kvm(kvm);
Avi Kivityf17abe92007-02-21 19:28:04 +02002349
Avi Kivityf17abe92007-02-21 19:28:04 +02002350 return fd;
Avi Kivityf17abe92007-02-21 19:28:04 +02002351}
2352
Avi Kivity1a811b62008-12-08 18:25:27 +02002353static long kvm_dev_ioctl_check_extension_generic(long arg)
2354{
2355 switch (arg) {
Avi Kivityca9edae2008-12-08 18:29:29 +02002356 case KVM_CAP_USER_MEMORY:
Avi Kivity1a811b62008-12-08 18:25:27 +02002357 case KVM_CAP_DESTROY_MEMORY_REGION_WORKS:
Jan Kiszka4cd481f2009-04-13 11:59:32 +02002358 case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS:
Gleb Natapov73880c82009-06-09 15:56:28 +03002359#ifdef CONFIG_KVM_APIC_ARCHITECTURE
2360 case KVM_CAP_SET_BOOT_CPU_ID:
2361#endif
Avi Kivity1a811b62008-12-08 18:25:27 +02002362 return 1;
Avi Kivity399ec802008-11-19 13:58:46 +02002363#ifdef CONFIG_HAVE_KVM_IRQCHIP
2364 case KVM_CAP_IRQ_ROUTING:
Sheng Yang36463142009-03-16 16:33:43 +08002365 return KVM_MAX_IRQ_ROUTES;
Avi Kivity399ec802008-11-19 13:58:46 +02002366#endif
Avi Kivity1a811b62008-12-08 18:25:27 +02002367 default:
2368 break;
2369 }
2370 return kvm_dev_ioctl_check_extension(arg);
2371}
2372
Avi Kivityf17abe92007-02-21 19:28:04 +02002373static long kvm_dev_ioctl(struct file *filp,
2374 unsigned int ioctl, unsigned long arg)
2375{
Avi Kivity07c45a32007-03-07 13:05:38 +02002376 long r = -EINVAL;
Avi Kivityf17abe92007-02-21 19:28:04 +02002377
2378 switch (ioctl) {
2379 case KVM_GET_API_VERSION:
Avi Kivityf0fe5102007-03-07 13:11:17 +02002380 r = -EINVAL;
2381 if (arg)
2382 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02002383 r = KVM_API_VERSION;
2384 break;
2385 case KVM_CREATE_VM:
Avi Kivityf0fe5102007-03-07 13:11:17 +02002386 r = -EINVAL;
2387 if (arg)
2388 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02002389 r = kvm_dev_ioctl_create_vm();
2390 break;
Zhang Xiantao018d00d2007-11-15 23:07:47 +08002391 case KVM_CHECK_EXTENSION:
Avi Kivity1a811b62008-12-08 18:25:27 +02002392 r = kvm_dev_ioctl_check_extension_generic(arg);
Avi Kivity5d308f42007-03-01 17:56:20 +02002393 break;
Avi Kivity07c45a32007-03-07 13:05:38 +02002394 case KVM_GET_VCPU_MMAP_SIZE:
2395 r = -EINVAL;
2396 if (arg)
2397 goto out;
Avi Kivityadb1ff42008-01-24 15:13:08 +02002398 r = PAGE_SIZE; /* struct kvm_run */
2399#ifdef CONFIG_X86
2400 r += PAGE_SIZE; /* pio data page */
2401#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02002402#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
2403 r += PAGE_SIZE; /* coalesced mmio ring page */
2404#endif
Avi Kivity07c45a32007-03-07 13:05:38 +02002405 break;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04002406 case KVM_TRACE_ENABLE:
2407 case KVM_TRACE_PAUSE:
2408 case KVM_TRACE_DISABLE:
Marcelo Tosatti2023a292009-06-18 11:47:28 -03002409 r = -EOPNOTSUPP;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04002410 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002411 default:
Carsten Otte043405e2007-10-10 17:16:19 +02002412 return kvm_arch_dev_ioctl(filp, ioctl, arg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002413 }
2414out:
2415 return r;
2416}
2417
Avi Kivity6aa8b732006-12-10 02:21:36 -08002418static struct file_operations kvm_chardev_ops = {
Avi Kivity6aa8b732006-12-10 02:21:36 -08002419 .unlocked_ioctl = kvm_dev_ioctl,
2420 .compat_ioctl = kvm_dev_ioctl,
Avi Kivity6aa8b732006-12-10 02:21:36 -08002421};
2422
2423static struct miscdevice kvm_dev = {
Avi Kivitybbe44322007-03-04 13:27:36 +02002424 KVM_MINOR,
Avi Kivity6aa8b732006-12-10 02:21:36 -08002425 "kvm",
2426 &kvm_chardev_ops,
2427};
2428
Avi Kivity1b6c0162007-05-24 13:03:52 +03002429static void hardware_enable(void *junk)
2430{
2431 int cpu = raw_smp_processor_id();
2432
Rusty Russell7f59f492008-12-07 21:25:45 +10302433 if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03002434 return;
Rusty Russell7f59f492008-12-07 21:25:45 +10302435 cpumask_set_cpu(cpu, cpus_hardware_enabled);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002436 kvm_arch_hardware_enable(NULL);
Avi Kivity1b6c0162007-05-24 13:03:52 +03002437}
2438
2439static void hardware_disable(void *junk)
2440{
2441 int cpu = raw_smp_processor_id();
2442
Rusty Russell7f59f492008-12-07 21:25:45 +10302443 if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03002444 return;
Rusty Russell7f59f492008-12-07 21:25:45 +10302445 cpumask_clear_cpu(cpu, cpus_hardware_enabled);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002446 kvm_arch_hardware_disable(NULL);
Avi Kivity1b6c0162007-05-24 13:03:52 +03002447}
2448
Avi Kivity774c47f2007-02-12 00:54:47 -08002449static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val,
2450 void *v)
2451{
2452 int cpu = (long)v;
2453
Avi Kivity1a6f4d72007-11-11 18:37:32 +02002454 val &= ~CPU_TASKS_FROZEN;
Avi Kivity774c47f2007-02-12 00:54:47 -08002455 switch (val) {
Avi Kivitycec9ad22007-05-24 13:11:41 +03002456 case CPU_DYING:
Avi Kivity6ec8a8562007-08-19 15:57:26 +03002457 printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n",
2458 cpu);
2459 hardware_disable(NULL);
2460 break;
Avi Kivity774c47f2007-02-12 00:54:47 -08002461 case CPU_UP_CANCELED:
Jeremy Katz43934a32007-02-19 14:37:46 +02002462 printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n",
2463 cpu);
Jens Axboe8691e5a2008-06-06 11:18:06 +02002464 smp_call_function_single(cpu, hardware_disable, NULL, 1);
Avi Kivity774c47f2007-02-12 00:54:47 -08002465 break;
Jeremy Katz43934a32007-02-19 14:37:46 +02002466 case CPU_ONLINE:
2467 printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n",
2468 cpu);
Jens Axboe8691e5a2008-06-06 11:18:06 +02002469 smp_call_function_single(cpu, hardware_enable, NULL, 1);
Avi Kivity774c47f2007-02-12 00:54:47 -08002470 break;
2471 }
2472 return NOTIFY_OK;
2473}
2474
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002475
2476asmlinkage void kvm_handle_fault_on_reboot(void)
2477{
2478 if (kvm_rebooting)
2479 /* spin while reset goes on */
2480 while (true)
2481 ;
2482 /* Fault while not rebooting. We want the trace. */
2483 BUG();
2484}
2485EXPORT_SYMBOL_GPL(kvm_handle_fault_on_reboot);
2486
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002487static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
Mike Dayd77c26f2007-10-08 09:02:08 -04002488 void *v)
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002489{
Sheng Yang8e1c1812009-04-29 11:09:04 +08002490 /*
2491 * Some (well, at least mine) BIOSes hang on reboot if
2492 * in vmx root mode.
2493 *
2494 * And Intel TXT required VMX off for all cpu when system shutdown.
2495 */
2496 printk(KERN_INFO "kvm: exiting hardware virtualization\n");
2497 kvm_rebooting = true;
2498 on_each_cpu(hardware_disable, NULL, 1);
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002499 return NOTIFY_OK;
2500}
2501
2502static struct notifier_block kvm_reboot_notifier = {
2503 .notifier_call = kvm_reboot,
2504 .priority = 0,
2505};
2506
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002507void kvm_io_bus_init(struct kvm_io_bus *bus)
2508{
2509 memset(bus, 0, sizeof(*bus));
2510}
2511
2512void kvm_io_bus_destroy(struct kvm_io_bus *bus)
2513{
2514 int i;
2515
2516 for (i = 0; i < bus->dev_count; i++) {
2517 struct kvm_io_device *pos = bus->devs[i];
2518
2519 kvm_iodevice_destructor(pos);
2520 }
2521}
2522
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002523/* kvm_io_bus_write - called under kvm->slots_lock */
2524int kvm_io_bus_write(struct kvm_io_bus *bus, gpa_t addr,
2525 int len, const void *val)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002526{
2527 int i;
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002528 for (i = 0; i < bus->dev_count; i++)
2529 if (!kvm_iodevice_write(bus->devs[i], addr, len, val))
2530 return 0;
2531 return -EOPNOTSUPP;
2532}
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002533
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002534/* kvm_io_bus_read - called under kvm->slots_lock */
2535int kvm_io_bus_read(struct kvm_io_bus *bus, gpa_t addr, int len, void *val)
2536{
2537 int i;
2538 for (i = 0; i < bus->dev_count; i++)
2539 if (!kvm_iodevice_read(bus->devs[i], addr, len, val))
2540 return 0;
2541 return -EOPNOTSUPP;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002542}
2543
Gregory Haskins090b7af2009-07-07 17:08:44 -04002544int kvm_io_bus_register_dev(struct kvm *kvm, struct kvm_io_bus *bus,
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002545 struct kvm_io_device *dev)
2546{
Gregory Haskins090b7af2009-07-07 17:08:44 -04002547 int ret;
2548
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002549 down_write(&kvm->slots_lock);
Gregory Haskins090b7af2009-07-07 17:08:44 -04002550 ret = __kvm_io_bus_register_dev(bus, dev);
2551 up_write(&kvm->slots_lock);
2552
2553 return ret;
2554}
2555
2556/* An unlocked version. Caller must have write lock on slots_lock. */
2557int __kvm_io_bus_register_dev(struct kvm_io_bus *bus,
2558 struct kvm_io_device *dev)
2559{
2560 if (bus->dev_count > NR_IOBUS_DEVS-1)
2561 return -ENOSPC;
2562
2563 bus->devs[bus->dev_count++] = dev;
2564
2565 return 0;
2566}
2567
2568void kvm_io_bus_unregister_dev(struct kvm *kvm,
2569 struct kvm_io_bus *bus,
2570 struct kvm_io_device *dev)
2571{
2572 down_write(&kvm->slots_lock);
2573 __kvm_io_bus_unregister_dev(bus, dev);
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002574 up_write(&kvm->slots_lock);
2575}
2576
2577/* An unlocked version. Caller must have write lock on slots_lock. */
Gregory Haskins090b7af2009-07-07 17:08:44 -04002578void __kvm_io_bus_unregister_dev(struct kvm_io_bus *bus,
2579 struct kvm_io_device *dev)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002580{
Gregory Haskins090b7af2009-07-07 17:08:44 -04002581 int i;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002582
Gregory Haskins090b7af2009-07-07 17:08:44 -04002583 for (i = 0; i < bus->dev_count; i++)
2584 if (bus->devs[i] == dev) {
2585 bus->devs[i] = bus->devs[--bus->dev_count];
2586 break;
2587 }
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002588}
2589
Avi Kivity774c47f2007-02-12 00:54:47 -08002590static struct notifier_block kvm_cpu_notifier = {
2591 .notifier_call = kvm_cpu_hotplug,
2592 .priority = 20, /* must be > scheduler priority */
2593};
2594
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002595static int vm_stat_get(void *_offset, u64 *val)
Avi Kivityba1389b2007-11-18 16:24:12 +02002596{
2597 unsigned offset = (long)_offset;
Avi Kivityba1389b2007-11-18 16:24:12 +02002598 struct kvm *kvm;
2599
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002600 *val = 0;
Avi Kivityba1389b2007-11-18 16:24:12 +02002601 spin_lock(&kvm_lock);
2602 list_for_each_entry(kvm, &vm_list, vm_list)
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002603 *val += *(u32 *)((void *)kvm + offset);
Avi Kivityba1389b2007-11-18 16:24:12 +02002604 spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002605 return 0;
Avi Kivityba1389b2007-11-18 16:24:12 +02002606}
2607
2608DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n");
2609
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002610static int vcpu_stat_get(void *_offset, u64 *val)
Avi Kivity1165f5f2007-04-19 17:27:43 +03002611{
2612 unsigned offset = (long)_offset;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002613 struct kvm *kvm;
2614 struct kvm_vcpu *vcpu;
2615 int i;
2616
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002617 *val = 0;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002618 spin_lock(&kvm_lock);
2619 list_for_each_entry(kvm, &vm_list, vm_list)
Gleb Natapov988a2ca2009-06-09 15:56:29 +03002620 kvm_for_each_vcpu(i, vcpu, kvm)
2621 *val += *(u32 *)((void *)vcpu + offset);
2622
Avi Kivity1165f5f2007-04-19 17:27:43 +03002623 spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002624 return 0;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002625}
2626
Avi Kivityba1389b2007-11-18 16:24:12 +02002627DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n");
2628
2629static struct file_operations *stat_fops[] = {
2630 [KVM_STAT_VCPU] = &vcpu_stat_fops,
2631 [KVM_STAT_VM] = &vm_stat_fops,
2632};
Avi Kivity1165f5f2007-04-19 17:27:43 +03002633
Zhang Xiantaoa16b0432007-11-16 14:38:21 +08002634static void kvm_init_debug(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002635{
2636 struct kvm_stats_debugfs_item *p;
2637
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002638 kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002639 for (p = debugfs_entries; p->name; ++p)
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002640 p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir,
Avi Kivity1165f5f2007-04-19 17:27:43 +03002641 (void *)(long)p->offset,
Avi Kivityba1389b2007-11-18 16:24:12 +02002642 stat_fops[p->kind]);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002643}
2644
2645static void kvm_exit_debug(void)
2646{
2647 struct kvm_stats_debugfs_item *p;
2648
2649 for (p = debugfs_entries; p->name; ++p)
2650 debugfs_remove(p->dentry);
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002651 debugfs_remove(kvm_debugfs_dir);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002652}
2653
Avi Kivity59ae6c62007-02-12 00:54:48 -08002654static int kvm_suspend(struct sys_device *dev, pm_message_t state)
2655{
Avi Kivity4267c412007-05-24 13:09:41 +03002656 hardware_disable(NULL);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002657 return 0;
2658}
2659
2660static int kvm_resume(struct sys_device *dev)
2661{
Avi Kivity4267c412007-05-24 13:09:41 +03002662 hardware_enable(NULL);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002663 return 0;
2664}
2665
2666static struct sysdev_class kvm_sysdev_class = {
Kay Sieversaf5ca3f2007-12-20 02:09:39 +01002667 .name = "kvm",
Avi Kivity59ae6c62007-02-12 00:54:48 -08002668 .suspend = kvm_suspend,
2669 .resume = kvm_resume,
2670};
2671
2672static struct sys_device kvm_sysdev = {
2673 .id = 0,
2674 .cls = &kvm_sysdev_class,
2675};
2676
Izik Eiduscea7bb22007-10-17 19:17:48 +02002677struct page *bad_page;
Anthony Liguori35149e22008-04-02 14:46:56 -05002678pfn_t bad_pfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002679
Avi Kivity15ad7142007-07-11 18:17:21 +03002680static inline
2681struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn)
2682{
2683 return container_of(pn, struct kvm_vcpu, preempt_notifier);
2684}
2685
2686static void kvm_sched_in(struct preempt_notifier *pn, int cpu)
2687{
2688 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2689
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002690 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002691}
2692
2693static void kvm_sched_out(struct preempt_notifier *pn,
2694 struct task_struct *next)
2695{
2696 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2697
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002698 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002699}
2700
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002701int kvm_init(void *opaque, unsigned int vcpu_size,
Rusty Russellc16f8622007-07-30 21:12:19 +10002702 struct module *module)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002703{
2704 int r;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002705 int cpu;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002706
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002707 kvm_init_debug();
2708
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002709 r = kvm_arch_init(opaque);
2710 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002711 goto out_fail;
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002712
2713 bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2714
2715 if (bad_page == NULL) {
2716 r = -ENOMEM;
2717 goto out;
2718 }
2719
Anthony Liguori35149e22008-04-02 14:46:56 -05002720 bad_pfn = page_to_pfn(bad_page);
2721
Avi Kivity8437a612009-06-06 14:52:35 -07002722 if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
Rusty Russell7f59f492008-12-07 21:25:45 +10302723 r = -ENOMEM;
2724 goto out_free_0;
2725 }
2726
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002727 r = kvm_arch_hardware_setup();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002728 if (r < 0)
Rusty Russell7f59f492008-12-07 21:25:45 +10302729 goto out_free_0a;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002730
Yang, Sheng002c7f72007-07-31 14:23:01 +03002731 for_each_online_cpu(cpu) {
2732 smp_call_function_single(cpu,
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002733 kvm_arch_check_processor_compat,
Jens Axboe8691e5a2008-06-06 11:18:06 +02002734 &r, 1);
Yang, Sheng002c7f72007-07-31 14:23:01 +03002735 if (r < 0)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002736 goto out_free_1;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002737 }
2738
Jens Axboe15c8b6c2008-05-09 09:39:44 +02002739 on_each_cpu(hardware_enable, NULL, 1);
Avi Kivity774c47f2007-02-12 00:54:47 -08002740 r = register_cpu_notifier(&kvm_cpu_notifier);
2741 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002742 goto out_free_2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002743 register_reboot_notifier(&kvm_reboot_notifier);
2744
Avi Kivity59ae6c62007-02-12 00:54:48 -08002745 r = sysdev_class_register(&kvm_sysdev_class);
2746 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002747 goto out_free_3;
Avi Kivity59ae6c62007-02-12 00:54:48 -08002748
2749 r = sysdev_register(&kvm_sysdev);
2750 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002751 goto out_free_4;
Avi Kivity59ae6c62007-02-12 00:54:48 -08002752
Rusty Russellc16f8622007-07-30 21:12:19 +10002753 /* A kmem cache lets us meet the alignment requirements of fx_save. */
2754 kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size,
Joe Perches56919c52007-11-12 20:06:51 -08002755 __alignof__(struct kvm_vcpu),
2756 0, NULL);
Rusty Russellc16f8622007-07-30 21:12:19 +10002757 if (!kvm_vcpu_cache) {
2758 r = -ENOMEM;
Zhang Xiantaod23087842007-11-29 15:35:39 +08002759 goto out_free_5;
Rusty Russellc16f8622007-07-30 21:12:19 +10002760 }
2761
Avi Kivity6aa8b732006-12-10 02:21:36 -08002762 kvm_chardev_ops.owner = module;
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01002763 kvm_vm_fops.owner = module;
2764 kvm_vcpu_fops.owner = module;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002765
2766 r = misc_register(&kvm_dev);
2767 if (r) {
Mike Dayd77c26f2007-10-08 09:02:08 -04002768 printk(KERN_ERR "kvm: misc device register failed\n");
Avi Kivity6aa8b732006-12-10 02:21:36 -08002769 goto out_free;
2770 }
2771
Avi Kivity15ad7142007-07-11 18:17:21 +03002772 kvm_preempt_ops.sched_in = kvm_sched_in;
2773 kvm_preempt_ops.sched_out = kvm_sched_out;
2774
Avi Kivityc7addb92007-09-16 18:58:32 +02002775 return 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002776
2777out_free:
Rusty Russellc16f8622007-07-30 21:12:19 +10002778 kmem_cache_destroy(kvm_vcpu_cache);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002779out_free_5:
Avi Kivity59ae6c62007-02-12 00:54:48 -08002780 sysdev_unregister(&kvm_sysdev);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002781out_free_4:
Avi Kivity59ae6c62007-02-12 00:54:48 -08002782 sysdev_class_unregister(&kvm_sysdev_class);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002783out_free_3:
Avi Kivity6aa8b732006-12-10 02:21:36 -08002784 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity774c47f2007-02-12 00:54:47 -08002785 unregister_cpu_notifier(&kvm_cpu_notifier);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002786out_free_2:
Jens Axboe15c8b6c2008-05-09 09:39:44 +02002787 on_each_cpu(hardware_disable, NULL, 1);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002788out_free_1:
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002789 kvm_arch_hardware_unsetup();
Rusty Russell7f59f492008-12-07 21:25:45 +10302790out_free_0a:
2791 free_cpumask_var(cpus_hardware_enabled);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002792out_free_0:
2793 __free_page(bad_page);
Avi Kivityca45aaa2007-03-01 19:21:03 +02002794out:
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002795 kvm_arch_exit();
Zhang Xiantaod23087842007-11-29 15:35:39 +08002796out_fail:
Xiao Guangrongaed665f2009-08-03 17:10:06 +08002797 kvm_exit_debug();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002798 return r;
2799}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002800EXPORT_SYMBOL_GPL(kvm_init);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002801
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002802void kvm_exit(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002803{
Marcelo Tosatti229456f2009-06-17 09:22:14 -03002804 tracepoint_synchronize_unregister();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002805 misc_deregister(&kvm_dev);
Rusty Russellc16f8622007-07-30 21:12:19 +10002806 kmem_cache_destroy(kvm_vcpu_cache);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002807 sysdev_unregister(&kvm_sysdev);
2808 sysdev_class_unregister(&kvm_sysdev_class);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002809 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002810 unregister_cpu_notifier(&kvm_cpu_notifier);
Jens Axboe15c8b6c2008-05-09 09:39:44 +02002811 on_each_cpu(hardware_disable, NULL, 1);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002812 kvm_arch_hardware_unsetup();
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002813 kvm_arch_exit();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002814 kvm_exit_debug();
Rusty Russell7f59f492008-12-07 21:25:45 +10302815 free_cpumask_var(cpus_hardware_enabled);
Izik Eiduscea7bb22007-10-17 19:17:48 +02002816 __free_page(bad_page);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002817}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002818EXPORT_SYMBOL_GPL(kvm_exit);