blob: 605697e9c4dd50570b04cba68d80219450055828 [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>
Avi Kivity6aa8b732006-12-10 02:21:36 -080044
Avi Kivitye4956062007-06-28 14:15:57 -040045#include <asm/processor.h>
Avi Kivitye4956062007-06-28 14:15:57 -040046#include <asm/io.h>
47#include <asm/uaccess.h>
Izik Eidus3e021bf2007-11-19 11:16:57 +020048#include <asm/pgtable.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080049
Laurent Vivier5f94c172008-05-30 16:05:54 +020050#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
51#include "coalesced_mmio.h"
52#endif
53
Xiantao Zhang8a98f662008-10-06 13:47:38 +080054#ifdef KVM_CAP_DEVICE_ASSIGNMENT
55#include <linux/pci.h>
56#include <linux/interrupt.h>
57#include "irq.h"
58#endif
59
Avi Kivity6aa8b732006-12-10 02:21:36 -080060MODULE_AUTHOR("Qumranet");
61MODULE_LICENSE("GPL");
62
Sheng Yang5319c662008-11-24 14:32:57 +080063static int msi2intx = 1;
64module_param(msi2intx, bool, 0);
65
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +080066DEFINE_SPINLOCK(kvm_lock);
67LIST_HEAD(vm_list);
Avi Kivity133de902007-02-12 00:54:44 -080068
Rusty Russell7f59f492008-12-07 21:25:45 +103069static cpumask_var_t cpus_hardware_enabled;
Avi Kivity1b6c0162007-05-24 13:03:52 +030070
Rusty Russellc16f8622007-07-30 21:12:19 +100071struct kmem_cache *kvm_vcpu_cache;
72EXPORT_SYMBOL_GPL(kvm_vcpu_cache);
Avi Kivity1165f5f2007-04-19 17:27:43 +030073
Avi Kivity15ad7142007-07-11 18:17:21 +030074static __read_mostly struct preempt_ops kvm_preempt_ops;
75
Hollis Blanchard76f7c872008-04-15 16:05:42 -050076struct dentry *kvm_debugfs_dir;
Avi Kivity6aa8b732006-12-10 02:21:36 -080077
Avi Kivitybccf2152007-02-21 18:04:26 +020078static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl,
79 unsigned long arg);
80
Hannes Edere8ba5d32008-11-28 17:02:42 +010081static bool kvm_rebooting;
Avi Kivity4ecac3f2008-05-13 13:23:38 +030082
Xiantao Zhang8a98f662008-10-06 13:47:38 +080083#ifdef KVM_CAP_DEVICE_ASSIGNMENT
84static struct kvm_assigned_dev_kernel *kvm_find_assigned_dev(struct list_head *head,
85 int assigned_dev_id)
86{
87 struct list_head *ptr;
88 struct kvm_assigned_dev_kernel *match;
89
90 list_for_each(ptr, head) {
91 match = list_entry(ptr, struct kvm_assigned_dev_kernel, list);
92 if (match->assigned_dev_id == assigned_dev_id)
93 return match;
94 }
95 return NULL;
96}
97
98static void kvm_assigned_dev_interrupt_work_handler(struct work_struct *work)
99{
100 struct kvm_assigned_dev_kernel *assigned_dev;
101
102 assigned_dev = container_of(work, struct kvm_assigned_dev_kernel,
103 interrupt_work);
104
105 /* This is taken to safely inject irq inside the guest. When
106 * the interrupt injection (or the ioapic code) uses a
107 * finer-grained lock, update this
108 */
109 mutex_lock(&assigned_dev->kvm->lock);
Sheng Yang79950e12009-02-10 13:57:06 +0800110 kvm_set_irq(assigned_dev->kvm, assigned_dev->irq_source_id,
111 assigned_dev->guest_irq, 1);
112
113 if (assigned_dev->irq_requested_type & KVM_ASSIGNED_DEV_GUEST_MSI) {
Sheng Yang6b9cc7f2008-11-24 14:32:56 +0800114 enable_irq(assigned_dev->host_irq);
Mark McLoughlindefaf152008-12-02 12:16:33 +0000115 assigned_dev->host_irq_disabled = false;
Sheng Yang6b9cc7f2008-11-24 14:32:56 +0800116 }
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800117 mutex_unlock(&assigned_dev->kvm->lock);
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800118}
119
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800120static irqreturn_t kvm_assigned_dev_intr(int irq, void *dev_id)
121{
122 struct kvm_assigned_dev_kernel *assigned_dev =
123 (struct kvm_assigned_dev_kernel *) dev_id;
124
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800125 schedule_work(&assigned_dev->interrupt_work);
Mark McLoughlindefaf152008-12-02 12:16:33 +0000126
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800127 disable_irq_nosync(irq);
Mark McLoughlindefaf152008-12-02 12:16:33 +0000128 assigned_dev->host_irq_disabled = true;
129
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800130 return IRQ_HANDLED;
131}
132
133/* Ack the irq line for an assigned device */
134static void kvm_assigned_dev_ack_irq(struct kvm_irq_ack_notifier *kian)
135{
136 struct kvm_assigned_dev_kernel *dev;
137
138 if (kian->gsi == -1)
139 return;
140
141 dev = container_of(kian, struct kvm_assigned_dev_kernel,
142 ack_notifier);
Mark McLoughlindefaf152008-12-02 12:16:33 +0000143
Sheng Yang5550af42008-10-15 20:15:06 +0800144 kvm_set_irq(dev->kvm, dev->irq_source_id, dev->guest_irq, 0);
Mark McLoughlindefaf152008-12-02 12:16:33 +0000145
146 /* The guest irq may be shared so this ack may be
147 * from another device.
148 */
149 if (dev->host_irq_disabled) {
150 enable_irq(dev->host_irq);
151 dev->host_irq_disabled = false;
152 }
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800153}
154
Sheng Yangba4cef32009-01-06 10:03:03 +0800155/* The function implicit hold kvm->lock mutex due to cancel_work_sync() */
Mark McLoughlin4a643be2008-12-01 13:57:49 +0000156static void kvm_free_assigned_irq(struct kvm *kvm,
157 struct kvm_assigned_dev_kernel *assigned_dev)
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800158{
Mark McLoughlin4a643be2008-12-01 13:57:49 +0000159 if (!irqchip_in_kernel(kvm))
160 return;
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800161
Sheng Yange19e30e2008-10-20 16:07:10 +0800162 kvm_unregister_irq_ack_notifier(&assigned_dev->ack_notifier);
Mark McLoughlinf29b2672008-12-01 13:57:47 +0000163
164 if (assigned_dev->irq_source_id != -1)
165 kvm_free_irq_source_id(kvm, assigned_dev->irq_source_id);
166 assigned_dev->irq_source_id = -1;
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800167
Mark McLoughlin4a643be2008-12-01 13:57:49 +0000168 if (!assigned_dev->irq_requested_type)
169 return;
170
Sheng Yangba4cef32009-01-06 10:03:03 +0800171 /*
172 * In kvm_free_device_irq, cancel_work_sync return true if:
173 * 1. work is scheduled, and then cancelled.
174 * 2. work callback is executed.
175 *
176 * The first one ensured that the irq is disabled and no more events
177 * would happen. But for the second one, the irq may be enabled (e.g.
178 * for MSI). So we disable irq here to prevent further events.
179 *
180 * Notice this maybe result in nested disable if the interrupt type is
181 * INTx, but it's OK for we are going to free it.
182 *
183 * If this function is a part of VM destroy, please ensure that till
184 * now, the kvm state is still legal for probably we also have to wait
185 * interrupt_work done.
186 */
187 disable_irq_nosync(assigned_dev->host_irq);
188 cancel_work_sync(&assigned_dev->interrupt_work);
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800189
Mark McLoughlin4a643be2008-12-01 13:57:49 +0000190 free_irq(assigned_dev->host_irq, (void *)assigned_dev);
191
192 if (assigned_dev->irq_requested_type & KVM_ASSIGNED_DEV_HOST_MSI)
193 pci_disable_msi(assigned_dev->dev);
194
195 assigned_dev->irq_requested_type = 0;
196}
197
198
199static void kvm_free_assigned_device(struct kvm *kvm,
200 struct kvm_assigned_dev_kernel
201 *assigned_dev)
202{
203 kvm_free_assigned_irq(kvm, assigned_dev);
204
Sheng Yang6eb55812008-10-31 12:37:41 +0800205 pci_reset_function(assigned_dev->dev);
206
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800207 pci_release_regions(assigned_dev->dev);
208 pci_disable_device(assigned_dev->dev);
209 pci_dev_put(assigned_dev->dev);
210
211 list_del(&assigned_dev->list);
212 kfree(assigned_dev);
213}
214
215void kvm_free_all_assigned_devices(struct kvm *kvm)
216{
217 struct list_head *ptr, *ptr2;
218 struct kvm_assigned_dev_kernel *assigned_dev;
219
220 list_for_each_safe(ptr, ptr2, &kvm->arch.assigned_dev_head) {
221 assigned_dev = list_entry(ptr,
222 struct kvm_assigned_dev_kernel,
223 list);
224
225 kvm_free_assigned_device(kvm, assigned_dev);
226 }
227}
228
Sheng Yang00e3ed32008-11-24 14:32:50 +0800229static int assigned_device_update_intx(struct kvm *kvm,
230 struct kvm_assigned_dev_kernel *adev,
231 struct kvm_assigned_irq *airq)
232{
Sheng Yangfbac7812008-11-24 14:32:52 +0800233 adev->guest_irq = airq->guest_irq;
234 adev->ack_notifier.gsi = airq->guest_irq;
235
236 if (adev->irq_requested_type & KVM_ASSIGNED_DEV_HOST_INTX)
Sheng Yang00e3ed32008-11-24 14:32:50 +0800237 return 0;
Sheng Yang00e3ed32008-11-24 14:32:50 +0800238
239 if (irqchip_in_kernel(kvm)) {
Sheng Yang5319c662008-11-24 14:32:57 +0800240 if (!msi2intx &&
Sheng Yangd7cff1c2009-01-06 16:25:10 +0800241 (adev->irq_requested_type & KVM_ASSIGNED_DEV_HOST_MSI)) {
242 free_irq(adev->host_irq, (void *)adev);
Sheng Yang6b9cc7f2008-11-24 14:32:56 +0800243 pci_disable_msi(adev->dev);
244 }
245
Sheng Yang00e3ed32008-11-24 14:32:50 +0800246 if (!capable(CAP_SYS_RAWIO))
247 return -EPERM;
248
249 if (airq->host_irq)
250 adev->host_irq = airq->host_irq;
251 else
252 adev->host_irq = adev->dev->irq;
Sheng Yang00e3ed32008-11-24 14:32:50 +0800253
254 /* Even though this is PCI, we don't want to use shared
255 * interrupts. Sharing host devices with guest-assigned devices
256 * on the same interrupt line is not a happy situation: there
257 * are going to be long delays in accepting, acking, etc.
258 */
259 if (request_irq(adev->host_irq, kvm_assigned_dev_intr,
260 0, "kvm_assigned_intx_device", (void *)adev))
261 return -EIO;
262 }
263
Sheng Yang4f906c12008-11-24 14:32:51 +0800264 adev->irq_requested_type = KVM_ASSIGNED_DEV_GUEST_INTX |
265 KVM_ASSIGNED_DEV_HOST_INTX;
Sheng Yang00e3ed32008-11-24 14:32:50 +0800266 return 0;
267}
268
Sheng Yang6b9cc7f2008-11-24 14:32:56 +0800269#ifdef CONFIG_X86
270static int assigned_device_update_msi(struct kvm *kvm,
271 struct kvm_assigned_dev_kernel *adev,
272 struct kvm_assigned_irq *airq)
273{
274 int r;
275
Sheng Yang79950e12009-02-10 13:57:06 +0800276 adev->guest_irq = airq->guest_irq;
Sheng Yang5319c662008-11-24 14:32:57 +0800277 if (airq->flags & KVM_DEV_IRQ_ASSIGN_ENABLE_MSI) {
278 /* x86 don't care upper address of guest msi message addr */
279 adev->irq_requested_type |= KVM_ASSIGNED_DEV_GUEST_MSI;
280 adev->irq_requested_type &= ~KVM_ASSIGNED_DEV_GUEST_INTX;
Sheng Yang5319c662008-11-24 14:32:57 +0800281 adev->ack_notifier.gsi = -1;
282 } else if (msi2intx) {
283 adev->irq_requested_type |= KVM_ASSIGNED_DEV_GUEST_INTX;
284 adev->irq_requested_type &= ~KVM_ASSIGNED_DEV_GUEST_MSI;
Sheng Yang5319c662008-11-24 14:32:57 +0800285 adev->ack_notifier.gsi = airq->guest_irq;
Sheng Yang17071fe2009-01-06 16:25:11 +0800286 } else {
287 /*
288 * Guest require to disable device MSI, we disable MSI and
289 * re-enable INTx by default again. Notice it's only for
290 * non-msi2intx.
291 */
292 assigned_device_update_intx(kvm, adev, airq);
293 return 0;
Sheng Yang5319c662008-11-24 14:32:57 +0800294 }
Sheng Yang6b9cc7f2008-11-24 14:32:56 +0800295
296 if (adev->irq_requested_type & KVM_ASSIGNED_DEV_HOST_MSI)
297 return 0;
298
299 if (irqchip_in_kernel(kvm)) {
Sheng Yang5319c662008-11-24 14:32:57 +0800300 if (!msi2intx) {
301 if (adev->irq_requested_type &
302 KVM_ASSIGNED_DEV_HOST_INTX)
303 free_irq(adev->host_irq, (void *)adev);
Sheng Yang6b9cc7f2008-11-24 14:32:56 +0800304
Sheng Yang5319c662008-11-24 14:32:57 +0800305 r = pci_enable_msi(adev->dev);
306 if (r)
307 return r;
308 }
Sheng Yang6b9cc7f2008-11-24 14:32:56 +0800309
310 adev->host_irq = adev->dev->irq;
311 if (request_irq(adev->host_irq, kvm_assigned_dev_intr, 0,
312 "kvm_assigned_msi_device", (void *)adev))
313 return -EIO;
314 }
315
Sheng Yang5319c662008-11-24 14:32:57 +0800316 if (!msi2intx)
317 adev->irq_requested_type = KVM_ASSIGNED_DEV_GUEST_MSI;
318
319 adev->irq_requested_type |= KVM_ASSIGNED_DEV_HOST_MSI;
Sheng Yang6b9cc7f2008-11-24 14:32:56 +0800320 return 0;
321}
322#endif
323
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800324static int kvm_vm_ioctl_assign_irq(struct kvm *kvm,
325 struct kvm_assigned_irq
326 *assigned_irq)
327{
328 int r = 0;
329 struct kvm_assigned_dev_kernel *match;
Sheng Yang17071fe2009-01-06 16:25:11 +0800330 u32 current_flags = 0, changed_flags;
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800331
332 mutex_lock(&kvm->lock);
333
334 match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head,
335 assigned_irq->assigned_dev_id);
336 if (!match) {
337 mutex_unlock(&kvm->lock);
338 return -EINVAL;
339 }
340
Sheng Yang4f906c12008-11-24 14:32:51 +0800341 if (!match->irq_requested_type) {
Sheng Yang342ffb92008-11-24 14:32:49 +0800342 INIT_WORK(&match->interrupt_work,
343 kvm_assigned_dev_interrupt_work_handler);
344 if (irqchip_in_kernel(kvm)) {
345 /* Register ack nofitier */
346 match->ack_notifier.gsi = -1;
347 match->ack_notifier.irq_acked =
348 kvm_assigned_dev_ack_irq;
349 kvm_register_irq_ack_notifier(kvm,
350 &match->ack_notifier);
351
352 /* Request IRQ source ID */
353 r = kvm_request_irq_source_id(kvm);
354 if (r < 0)
355 goto out_release;
356 else
357 match->irq_source_id = r;
Sheng Yang5319c662008-11-24 14:32:57 +0800358
359#ifdef CONFIG_X86
360 /* Determine host device irq type, we can know the
361 * result from dev->msi_enabled */
362 if (msi2intx)
363 pci_enable_msi(match->dev);
364#endif
Sheng Yang342ffb92008-11-24 14:32:49 +0800365 }
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800366 }
367
Sheng Yang17071fe2009-01-06 16:25:11 +0800368 if ((match->irq_requested_type & KVM_ASSIGNED_DEV_HOST_MSI) &&
369 (match->irq_requested_type & KVM_ASSIGNED_DEV_GUEST_MSI))
370 current_flags |= KVM_DEV_IRQ_ASSIGN_ENABLE_MSI;
371
372 changed_flags = assigned_irq->flags ^ current_flags;
373
374 if ((changed_flags & KVM_DEV_IRQ_ASSIGN_MSI_ACTION) ||
Sheng Yang5319c662008-11-24 14:32:57 +0800375 (msi2intx && match->dev->msi_enabled)) {
Sheng Yang6b9cc7f2008-11-24 14:32:56 +0800376#ifdef CONFIG_X86
377 r = assigned_device_update_msi(kvm, match, assigned_irq);
378 if (r) {
379 printk(KERN_WARNING "kvm: failed to enable "
380 "MSI device!\n");
381 goto out_release;
382 }
383#else
384 r = -ENOTTY;
385#endif
386 } else if (assigned_irq->host_irq == 0 && match->dev->irq == 0) {
387 /* Host device IRQ 0 means don't support INTx */
Sheng Yang5319c662008-11-24 14:32:57 +0800388 if (!msi2intx) {
389 printk(KERN_WARNING
390 "kvm: wait device to enable MSI!\n");
391 r = 0;
392 } else {
393 printk(KERN_WARNING
394 "kvm: failed to enable MSI device!\n");
395 r = -ENOTTY;
396 goto out_release;
397 }
Sheng Yang6b9cc7f2008-11-24 14:32:56 +0800398 } else {
399 /* Non-sharing INTx mode */
400 r = assigned_device_update_intx(kvm, match, assigned_irq);
401 if (r) {
402 printk(KERN_WARNING "kvm: failed to enable "
403 "INTx device!\n");
404 goto out_release;
405 }
406 }
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800407
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800408 mutex_unlock(&kvm->lock);
409 return r;
410out_release:
411 mutex_unlock(&kvm->lock);
412 kvm_free_assigned_device(kvm, match);
413 return r;
414}
415
416static int kvm_vm_ioctl_assign_device(struct kvm *kvm,
417 struct kvm_assigned_pci_dev *assigned_dev)
418{
419 int r = 0;
420 struct kvm_assigned_dev_kernel *match;
421 struct pci_dev *dev;
422
Mark McLoughlin682edb42009-02-05 18:23:46 +0000423 down_read(&kvm->slots_lock);
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800424 mutex_lock(&kvm->lock);
425
426 match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head,
427 assigned_dev->assigned_dev_id);
428 if (match) {
429 /* device already assigned */
430 r = -EINVAL;
431 goto out;
432 }
433
434 match = kzalloc(sizeof(struct kvm_assigned_dev_kernel), GFP_KERNEL);
435 if (match == NULL) {
436 printk(KERN_INFO "%s: Couldn't allocate memory\n",
437 __func__);
438 r = -ENOMEM;
439 goto out;
440 }
441 dev = pci_get_bus_and_slot(assigned_dev->busnr,
442 assigned_dev->devfn);
443 if (!dev) {
444 printk(KERN_INFO "%s: host device not found\n", __func__);
445 r = -EINVAL;
446 goto out_free;
447 }
448 if (pci_enable_device(dev)) {
449 printk(KERN_INFO "%s: Could not enable PCI device\n", __func__);
450 r = -EBUSY;
451 goto out_put;
452 }
453 r = pci_request_regions(dev, "kvm_assigned_device");
454 if (r) {
455 printk(KERN_INFO "%s: Could not get access to device regions\n",
456 __func__);
457 goto out_disable;
458 }
Sheng Yang6eb55812008-10-31 12:37:41 +0800459
460 pci_reset_function(dev);
461
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800462 match->assigned_dev_id = assigned_dev->assigned_dev_id;
463 match->host_busnr = assigned_dev->busnr;
464 match->host_devfn = assigned_dev->devfn;
Weidong Hanb6535742008-12-08 23:29:53 +0800465 match->flags = assigned_dev->flags;
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800466 match->dev = dev;
Mark McLoughlinf29b2672008-12-01 13:57:47 +0000467 match->irq_source_id = -1;
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800468 match->kvm = kvm;
469
470 list_add(&match->list, &kvm->arch.assigned_dev_head);
471
472 if (assigned_dev->flags & KVM_DEV_ASSIGN_ENABLE_IOMMU) {
Joerg Roedel19de40a2008-12-03 14:43:34 +0100473 if (!kvm->arch.iommu_domain) {
Weidong Han260782b2008-12-02 21:03:39 +0800474 r = kvm_iommu_map_guest(kvm);
475 if (r)
476 goto out_list_del;
477 }
478 r = kvm_assign_device(kvm, match);
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800479 if (r)
480 goto out_list_del;
481 }
482
483out:
484 mutex_unlock(&kvm->lock);
Mark McLoughlin682edb42009-02-05 18:23:46 +0000485 up_read(&kvm->slots_lock);
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800486 return r;
487out_list_del:
488 list_del(&match->list);
489 pci_release_regions(dev);
490out_disable:
491 pci_disable_device(dev);
492out_put:
493 pci_dev_put(dev);
494out_free:
495 kfree(match);
496 mutex_unlock(&kvm->lock);
Mark McLoughlin682edb42009-02-05 18:23:46 +0000497 up_read(&kvm->slots_lock);
Xiantao Zhang8a98f662008-10-06 13:47:38 +0800498 return r;
499}
500#endif
501
Weidong Han0a920352008-12-02 21:24:23 +0800502#ifdef KVM_CAP_DEVICE_DEASSIGNMENT
503static int kvm_vm_ioctl_deassign_device(struct kvm *kvm,
504 struct kvm_assigned_pci_dev *assigned_dev)
505{
506 int r = 0;
507 struct kvm_assigned_dev_kernel *match;
508
509 mutex_lock(&kvm->lock);
510
511 match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head,
512 assigned_dev->assigned_dev_id);
513 if (!match) {
514 printk(KERN_INFO "%s: device hasn't been assigned before, "
515 "so cannot be deassigned\n", __func__);
516 r = -EINVAL;
517 goto out;
518 }
519
Weidong Han4a906e42009-02-13 17:27:51 +0800520 if (match->flags & KVM_DEV_ASSIGN_ENABLE_IOMMU)
Weidong Han0a920352008-12-02 21:24:23 +0800521 kvm_deassign_device(kvm, match);
522
523 kvm_free_assigned_device(kvm, match);
524
525out:
526 mutex_unlock(&kvm->lock);
527 return r;
528}
529#endif
530
James Morris5aacf0c2006-12-22 01:04:55 -0800531static inline int valid_vcpu(int n)
532{
533 return likely(n >= 0 && n < KVM_MAX_VCPUS);
534}
535
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +0800536inline int kvm_is_mmio_pfn(pfn_t pfn)
Ben-Ami Yassourcbff90a2008-07-28 19:26:24 +0300537{
Joerg Roedelfc5659c2009-02-18 14:08:58 +0100538 if (pfn_valid(pfn)) {
539 struct page *page = compound_head(pfn_to_page(pfn));
540 return PageReserved(page);
541 }
Ben-Ami Yassourcbff90a2008-07-28 19:26:24 +0300542
543 return true;
544}
545
Avi Kivity6aa8b732006-12-10 02:21:36 -0800546/*
547 * Switches to specified vcpu, until a matching vcpu_put()
548 */
Carsten Otte313a3dc2007-10-11 19:16:52 +0200549void vcpu_load(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800550{
Avi Kivity15ad7142007-07-11 18:17:21 +0300551 int cpu;
552
Avi Kivitybccf2152007-02-21 18:04:26 +0200553 mutex_lock(&vcpu->mutex);
Avi Kivity15ad7142007-07-11 18:17:21 +0300554 cpu = get_cpu();
555 preempt_notifier_register(&vcpu->preempt_notifier);
Carsten Otte313a3dc2007-10-11 19:16:52 +0200556 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +0300557 put_cpu();
Avi Kivitybccf2152007-02-21 18:04:26 +0200558}
559
Carsten Otte313a3dc2007-10-11 19:16:52 +0200560void vcpu_put(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800561{
Avi Kivity15ad7142007-07-11 18:17:21 +0300562 preempt_disable();
Carsten Otte313a3dc2007-10-11 19:16:52 +0200563 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +0300564 preempt_notifier_unregister(&vcpu->preempt_notifier);
565 preempt_enable();
Avi Kivity6aa8b732006-12-10 02:21:36 -0800566 mutex_unlock(&vcpu->mutex);
567}
568
Avi Kivityd9e368d2007-06-07 19:18:30 +0300569static void ack_flush(void *_completed)
570{
Avi Kivityd9e368d2007-06-07 19:18:30 +0300571}
572
Rusty Russell49846892008-12-08 20:26:24 +1030573static bool make_all_cpus_request(struct kvm *kvm, unsigned int req)
Avi Kivityd9e368d2007-06-07 19:18:30 +0300574{
Avi Kivity597a5f52008-07-20 14:24:22 +0300575 int i, cpu, me;
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030576 cpumask_var_t cpus;
577 bool called = true;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300578 struct kvm_vcpu *vcpu;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300579
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030580 if (alloc_cpumask_var(&cpus, GFP_ATOMIC))
581 cpumask_clear(cpus);
582
Avi Kivity597a5f52008-07-20 14:24:22 +0300583 me = get_cpu();
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000584 for (i = 0; i < KVM_MAX_VCPUS; ++i) {
585 vcpu = kvm->vcpus[i];
586 if (!vcpu)
587 continue;
Rusty Russell49846892008-12-08 20:26:24 +1030588 if (test_and_set_bit(req, &vcpu->requests))
Avi Kivityd9e368d2007-06-07 19:18:30 +0300589 continue;
590 cpu = vcpu->cpu;
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030591 if (cpus != NULL && cpu != -1 && cpu != me)
592 cpumask_set_cpu(cpu, cpus);
Avi Kivityd9e368d2007-06-07 19:18:30 +0300593 }
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030594 if (unlikely(cpus == NULL))
595 smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1);
596 else if (!cpumask_empty(cpus))
597 smp_call_function_many(cpus, ack_flush, NULL, 1);
598 else
599 called = false;
Avi Kivity597a5f52008-07-20 14:24:22 +0300600 put_cpu();
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030601 free_cpumask_var(cpus);
Rusty Russell49846892008-12-08 20:26:24 +1030602 return called;
603}
604
605void kvm_flush_remote_tlbs(struct kvm *kvm)
606{
607 if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH))
608 ++kvm->stat.remote_tlb_flush;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300609}
610
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500611void kvm_reload_remote_mmus(struct kvm *kvm)
612{
Rusty Russell49846892008-12-08 20:26:24 +1030613 make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD);
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500614}
615
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000616int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id)
617{
618 struct page *page;
619 int r;
620
621 mutex_init(&vcpu->mutex);
622 vcpu->cpu = -1;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000623 vcpu->kvm = kvm;
624 vcpu->vcpu_id = id;
Eddie Dongb6958ce2007-07-18 12:15:21 +0300625 init_waitqueue_head(&vcpu->wq);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000626
627 page = alloc_page(GFP_KERNEL | __GFP_ZERO);
628 if (!page) {
629 r = -ENOMEM;
630 goto fail;
631 }
632 vcpu->run = page_address(page);
633
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800634 r = kvm_arch_vcpu_init(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000635 if (r < 0)
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800636 goto fail_free_run;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000637 return 0;
638
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000639fail_free_run:
640 free_page((unsigned long)vcpu->run);
641fail:
Rusty Russell76fafa52007-10-08 10:50:48 +1000642 return r;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000643}
644EXPORT_SYMBOL_GPL(kvm_vcpu_init);
645
646void kvm_vcpu_uninit(struct kvm_vcpu *vcpu)
647{
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800648 kvm_arch_vcpu_uninit(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000649 free_page((unsigned long)vcpu->run);
650}
651EXPORT_SYMBOL_GPL(kvm_vcpu_uninit);
652
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200653#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
654static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn)
655{
656 return container_of(mn, struct kvm, mmu_notifier);
657}
658
659static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn,
660 struct mm_struct *mm,
661 unsigned long address)
662{
663 struct kvm *kvm = mmu_notifier_to_kvm(mn);
664 int need_tlb_flush;
665
666 /*
667 * When ->invalidate_page runs, the linux pte has been zapped
668 * already but the page is still allocated until
669 * ->invalidate_page returns. So if we increase the sequence
670 * here the kvm page fault will notice if the spte can't be
671 * established because the page is going to be freed. If
672 * instead the kvm page fault establishes the spte before
673 * ->invalidate_page runs, kvm_unmap_hva will release it
674 * before returning.
675 *
676 * The sequence increase only need to be seen at spin_unlock
677 * time, and not at spin_lock time.
678 *
679 * Increasing the sequence after the spin_unlock would be
680 * unsafe because the kvm page fault could then establish the
681 * pte after kvm_unmap_hva returned, without noticing the page
682 * is going to be freed.
683 */
684 spin_lock(&kvm->mmu_lock);
685 kvm->mmu_notifier_seq++;
686 need_tlb_flush = kvm_unmap_hva(kvm, address);
687 spin_unlock(&kvm->mmu_lock);
688
689 /* we've to flush the tlb before the pages can be freed */
690 if (need_tlb_flush)
691 kvm_flush_remote_tlbs(kvm);
692
693}
694
695static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn,
696 struct mm_struct *mm,
697 unsigned long start,
698 unsigned long end)
699{
700 struct kvm *kvm = mmu_notifier_to_kvm(mn);
701 int need_tlb_flush = 0;
702
703 spin_lock(&kvm->mmu_lock);
704 /*
705 * The count increase must become visible at unlock time as no
706 * spte can be established without taking the mmu_lock and
707 * count is also read inside the mmu_lock critical section.
708 */
709 kvm->mmu_notifier_count++;
710 for (; start < end; start += PAGE_SIZE)
711 need_tlb_flush |= kvm_unmap_hva(kvm, start);
712 spin_unlock(&kvm->mmu_lock);
713
714 /* we've to flush the tlb before the pages can be freed */
715 if (need_tlb_flush)
716 kvm_flush_remote_tlbs(kvm);
717}
718
719static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
720 struct mm_struct *mm,
721 unsigned long start,
722 unsigned long end)
723{
724 struct kvm *kvm = mmu_notifier_to_kvm(mn);
725
726 spin_lock(&kvm->mmu_lock);
727 /*
728 * This sequence increase will notify the kvm page fault that
729 * the page that is going to be mapped in the spte could have
730 * been freed.
731 */
732 kvm->mmu_notifier_seq++;
733 /*
734 * The above sequence increase must be visible before the
735 * below count decrease but both values are read by the kvm
736 * page fault under mmu_lock spinlock so we don't need to add
737 * a smb_wmb() here in between the two.
738 */
739 kvm->mmu_notifier_count--;
740 spin_unlock(&kvm->mmu_lock);
741
742 BUG_ON(kvm->mmu_notifier_count < 0);
743}
744
745static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
746 struct mm_struct *mm,
747 unsigned long address)
748{
749 struct kvm *kvm = mmu_notifier_to_kvm(mn);
750 int young;
751
752 spin_lock(&kvm->mmu_lock);
753 young = kvm_age_hva(kvm, address);
754 spin_unlock(&kvm->mmu_lock);
755
756 if (young)
757 kvm_flush_remote_tlbs(kvm);
758
759 return young;
760}
761
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100762static void kvm_mmu_notifier_release(struct mmu_notifier *mn,
763 struct mm_struct *mm)
764{
765 struct kvm *kvm = mmu_notifier_to_kvm(mn);
766 kvm_arch_flush_shadow(kvm);
767}
768
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200769static const struct mmu_notifier_ops kvm_mmu_notifier_ops = {
770 .invalidate_page = kvm_mmu_notifier_invalidate_page,
771 .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start,
772 .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end,
773 .clear_flush_young = kvm_mmu_notifier_clear_flush_young,
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100774 .release = kvm_mmu_notifier_release,
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200775};
776#endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */
777
Avi Kivityf17abe92007-02-21 19:28:04 +0200778static struct kvm *kvm_create_vm(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800779{
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800780 struct kvm *kvm = kvm_arch_create_vm();
Laurent Vivier5f94c172008-05-30 16:05:54 +0200781#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
782 struct page *page;
783#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -0800784
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800785 if (IS_ERR(kvm))
786 goto out;
Avi Kivity75858a82009-01-04 17:10:50 +0200787#ifdef CONFIG_HAVE_KVM_IRQCHIP
Avi Kivity399ec802008-11-19 13:58:46 +0200788 INIT_LIST_HEAD(&kvm->irq_routing);
Avi Kivity75858a82009-01-04 17:10:50 +0200789 INIT_HLIST_HEAD(&kvm->mask_notifier_list);
790#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -0800791
Laurent Vivier5f94c172008-05-30 16:05:54 +0200792#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
793 page = alloc_page(GFP_KERNEL | __GFP_ZERO);
794 if (!page) {
795 kfree(kvm);
796 return ERR_PTR(-ENOMEM);
797 }
798 kvm->coalesced_mmio_ring =
799 (struct kvm_coalesced_mmio_ring *)page_address(page);
800#endif
801
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200802#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
803 {
804 int err;
805 kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops;
806 err = mmu_notifier_register(&kvm->mmu_notifier, current->mm);
807 if (err) {
808#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
809 put_page(page);
810#endif
811 kfree(kvm);
812 return ERR_PTR(err);
813 }
814 }
815#endif
816
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200817 kvm->mm = current->mm;
818 atomic_inc(&kvm->mm->mm_count);
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -0500819 spin_lock_init(&kvm->mmu_lock);
Eddie Dong74906342007-06-19 18:05:03 +0300820 kvm_io_bus_init(&kvm->pio_bus);
Shaohua Li11ec2802007-07-23 14:51:37 +0800821 mutex_init(&kvm->lock);
Gregory Haskins2eeb2e92007-05-31 14:08:53 -0400822 kvm_io_bus_init(&kvm->mmio_bus);
Izik Eidus72dc67a2008-02-10 18:04:15 +0200823 init_rwsem(&kvm->slots_lock);
Izik Eidusd39f13b2008-03-30 16:01:25 +0300824 atomic_set(&kvm->users_count, 1);
Rusty Russell5e58cfe2007-07-23 17:08:21 +1000825 spin_lock(&kvm_lock);
826 list_add(&kvm->vm_list, &vm_list);
827 spin_unlock(&kvm_lock);
Laurent Vivier5f94c172008-05-30 16:05:54 +0200828#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
829 kvm_coalesced_mmio_init(kvm);
830#endif
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800831out:
Avi Kivityf17abe92007-02-21 19:28:04 +0200832 return kvm;
833}
834
Avi Kivity6aa8b732006-12-10 02:21:36 -0800835/*
836 * Free any memory in @free but not in @dont.
837 */
838static void kvm_free_physmem_slot(struct kvm_memory_slot *free,
839 struct kvm_memory_slot *dont)
840{
Izik Eidus290fc382007-09-27 14:11:22 +0200841 if (!dont || free->rmap != dont->rmap)
842 vfree(free->rmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800843
844 if (!dont || free->dirty_bitmap != dont->dirty_bitmap)
845 vfree(free->dirty_bitmap);
846
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300847 if (!dont || free->lpage_info != dont->lpage_info)
848 vfree(free->lpage_info);
849
Avi Kivity6aa8b732006-12-10 02:21:36 -0800850 free->npages = 0;
Al Viro8b6d44c2007-02-09 16:38:40 +0000851 free->dirty_bitmap = NULL;
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500852 free->rmap = NULL;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300853 free->lpage_info = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800854}
855
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800856void kvm_free_physmem(struct kvm *kvm)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800857{
858 int i;
859
860 for (i = 0; i < kvm->nmemslots; ++i)
Al Viro8b6d44c2007-02-09 16:38:40 +0000861 kvm_free_physmem_slot(&kvm->memslots[i], NULL);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800862}
863
Avi Kivityf17abe92007-02-21 19:28:04 +0200864static void kvm_destroy_vm(struct kvm *kvm)
865{
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200866 struct mm_struct *mm = kvm->mm;
867
Sheng Yangad8ba2c2009-01-06 10:03:02 +0800868 kvm_arch_sync_events(kvm);
Avi Kivity133de902007-02-12 00:54:44 -0800869 spin_lock(&kvm_lock);
870 list_del(&kvm->vm_list);
871 spin_unlock(&kvm_lock);
Avi Kivity399ec802008-11-19 13:58:46 +0200872 kvm_free_irq_routing(kvm);
Eddie Dong74906342007-06-19 18:05:03 +0300873 kvm_io_bus_destroy(&kvm->pio_bus);
Gregory Haskins2eeb2e92007-05-31 14:08:53 -0400874 kvm_io_bus_destroy(&kvm->mmio_bus);
Laurent Vivier5f94c172008-05-30 16:05:54 +0200875#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
876 if (kvm->coalesced_mmio_ring != NULL)
877 free_page((unsigned long)kvm->coalesced_mmio_ring);
878#endif
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200879#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
880 mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
881#endif
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800882 kvm_arch_destroy_vm(kvm);
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200883 mmdrop(mm);
Avi Kivityf17abe92007-02-21 19:28:04 +0200884}
885
Izik Eidusd39f13b2008-03-30 16:01:25 +0300886void kvm_get_kvm(struct kvm *kvm)
887{
888 atomic_inc(&kvm->users_count);
889}
890EXPORT_SYMBOL_GPL(kvm_get_kvm);
891
892void kvm_put_kvm(struct kvm *kvm)
893{
894 if (atomic_dec_and_test(&kvm->users_count))
895 kvm_destroy_vm(kvm);
896}
897EXPORT_SYMBOL_GPL(kvm_put_kvm);
898
899
Avi Kivityf17abe92007-02-21 19:28:04 +0200900static int kvm_vm_release(struct inode *inode, struct file *filp)
901{
902 struct kvm *kvm = filp->private_data;
903
Izik Eidusd39f13b2008-03-30 16:01:25 +0300904 kvm_put_kvm(kvm);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800905 return 0;
906}
907
Avi Kivity6aa8b732006-12-10 02:21:36 -0800908/*
Avi Kivity6aa8b732006-12-10 02:21:36 -0800909 * Allocate some memory and give it an address in the guest physical address
910 * space.
911 *
912 * Discontiguous memory is allowed, mostly for framebuffers.
Sheng Yangf78e0e22007-10-29 09:40:42 +0800913 *
Marcelo Tosatti10589a42007-12-20 19:18:22 -0500914 * Must be called holding mmap_sem for write.
Avi Kivity6aa8b732006-12-10 02:21:36 -0800915 */
Sheng Yangf78e0e22007-10-29 09:40:42 +0800916int __kvm_set_memory_region(struct kvm *kvm,
917 struct kvm_userspace_memory_region *mem,
918 int user_alloc)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800919{
920 int r;
921 gfn_t base_gfn;
922 unsigned long npages;
923 unsigned long i;
924 struct kvm_memory_slot *memslot;
925 struct kvm_memory_slot old, new;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800926
927 r = -EINVAL;
928 /* General sanity checks */
929 if (mem->memory_size & (PAGE_SIZE - 1))
930 goto out;
931 if (mem->guest_phys_addr & (PAGE_SIZE - 1))
932 goto out;
Sheng Yange7cacd42008-11-11 15:30:40 +0800933 if (user_alloc && (mem->userspace_addr & (PAGE_SIZE - 1)))
Hollis Blanchard78749802008-11-07 13:32:12 -0600934 goto out;
Izik Eiduse0d62c72007-10-24 23:57:46 +0200935 if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800936 goto out;
937 if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
938 goto out;
939
940 memslot = &kvm->memslots[mem->slot];
941 base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
942 npages = mem->memory_size >> PAGE_SHIFT;
943
944 if (!npages)
945 mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES;
946
Avi Kivity6aa8b732006-12-10 02:21:36 -0800947 new = old = *memslot;
948
949 new.base_gfn = base_gfn;
950 new.npages = npages;
951 new.flags = mem->flags;
952
953 /* Disallow changing a memory slot's size. */
954 r = -EINVAL;
955 if (npages && old.npages && npages != old.npages)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800956 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800957
958 /* Check for overlaps */
959 r = -EEXIST;
960 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
961 struct kvm_memory_slot *s = &kvm->memslots[i];
962
963 if (s == memslot)
964 continue;
965 if (!((base_gfn + npages <= s->base_gfn) ||
966 (base_gfn >= s->base_gfn + s->npages)))
Sheng Yangf78e0e22007-10-29 09:40:42 +0800967 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800968 }
Avi Kivity6aa8b732006-12-10 02:21:36 -0800969
Avi Kivity6aa8b732006-12-10 02:21:36 -0800970 /* Free page dirty bitmap if unneeded */
971 if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
Al Viro8b6d44c2007-02-09 16:38:40 +0000972 new.dirty_bitmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800973
974 r = -ENOMEM;
975
976 /* Allocate if a slot is being created */
Carsten Otteeff01142008-06-27 15:05:31 +0200977#ifndef CONFIG_S390
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500978 if (npages && !new.rmap) {
Mike Dayd77c26f2007-10-08 09:02:08 -0400979 new.rmap = vmalloc(npages * sizeof(struct page *));
Izik Eidus290fc382007-09-27 14:11:22 +0200980
981 if (!new.rmap)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800982 goto out_free;
Izik Eidus290fc382007-09-27 14:11:22 +0200983
Izik Eidus290fc382007-09-27 14:11:22 +0200984 memset(new.rmap, 0, npages * sizeof(*new.rmap));
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500985
Izik Eidus80b14b52007-10-25 11:54:04 +0200986 new.user_alloc = user_alloc;
Andrea Arcangeli604b38a2008-07-25 16:32:03 +0200987 /*
988 * hva_to_rmmap() serialzies with the mmu_lock and to be
989 * safe it has to ignore memslots with !user_alloc &&
990 * !userspace_addr.
991 */
992 if (user_alloc)
993 new.userspace_addr = mem->userspace_addr;
994 else
995 new.userspace_addr = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800996 }
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300997 if (npages && !new.lpage_info) {
998 int largepages = npages / KVM_PAGES_PER_HPAGE;
999 if (npages % KVM_PAGES_PER_HPAGE)
1000 largepages++;
1001 if (base_gfn % KVM_PAGES_PER_HPAGE)
1002 largepages++;
1003
1004 new.lpage_info = vmalloc(largepages * sizeof(*new.lpage_info));
1005
1006 if (!new.lpage_info)
1007 goto out_free;
1008
1009 memset(new.lpage_info, 0, largepages * sizeof(*new.lpage_info));
1010
1011 if (base_gfn % KVM_PAGES_PER_HPAGE)
1012 new.lpage_info[0].write_count = 1;
1013 if ((base_gfn+npages) % KVM_PAGES_PER_HPAGE)
1014 new.lpage_info[largepages-1].write_count = 1;
1015 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08001016
1017 /* Allocate page dirty bitmap if needed */
1018 if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
1019 unsigned dirty_bytes = ALIGN(npages, BITS_PER_LONG) / 8;
1020
1021 new.dirty_bitmap = vmalloc(dirty_bytes);
1022 if (!new.dirty_bitmap)
Sheng Yangf78e0e22007-10-29 09:40:42 +08001023 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001024 memset(new.dirty_bitmap, 0, dirty_bytes);
1025 }
Carsten Otteeff01142008-06-27 15:05:31 +02001026#endif /* not defined CONFIG_S390 */
Avi Kivity6aa8b732006-12-10 02:21:36 -08001027
Marcelo Tosatti34d4cb82008-07-10 20:49:31 -03001028 if (!npages)
1029 kvm_arch_flush_shadow(kvm);
1030
Andrea Arcangeli604b38a2008-07-25 16:32:03 +02001031 spin_lock(&kvm->mmu_lock);
1032 if (mem->slot >= kvm->nmemslots)
1033 kvm->nmemslots = mem->slot + 1;
1034
Avi Kivity6aa8b732006-12-10 02:21:36 -08001035 *memslot = new;
Andrea Arcangeli604b38a2008-07-25 16:32:03 +02001036 spin_unlock(&kvm->mmu_lock);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001037
Zhang Xiantao0de10342007-11-20 16:25:04 +08001038 r = kvm_arch_set_memory_region(kvm, mem, old, user_alloc);
1039 if (r) {
Andrea Arcangeli604b38a2008-07-25 16:32:03 +02001040 spin_lock(&kvm->mmu_lock);
Zhang Xiantao0de10342007-11-20 16:25:04 +08001041 *memslot = old;
Andrea Arcangeli604b38a2008-07-25 16:32:03 +02001042 spin_unlock(&kvm->mmu_lock);
Zhang Xiantao0de10342007-11-20 16:25:04 +08001043 goto out_free;
Zhang Xiantao3ad82a72007-11-20 13:11:38 +08001044 }
1045
Glauber Costa6f897242008-12-03 13:40:51 -02001046 kvm_free_physmem_slot(&old, npages ? &new : NULL);
1047 /* Slot deletion case: we have to update the current slot */
1048 if (!npages)
1049 *memslot = old;
Xiantao Zhang8a98f662008-10-06 13:47:38 +08001050#ifdef CONFIG_DMAR
Ben-Ami Yassour62c476c2008-09-14 03:48:28 +03001051 /* map the pages in iommu page table */
1052 r = kvm_iommu_map_pages(kvm, base_gfn, npages);
1053 if (r)
1054 goto out;
Xiantao Zhang8a98f662008-10-06 13:47:38 +08001055#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -08001056 return 0;
1057
Sheng Yangf78e0e22007-10-29 09:40:42 +08001058out_free:
Avi Kivity6aa8b732006-12-10 02:21:36 -08001059 kvm_free_physmem_slot(&new, &old);
1060out:
1061 return r;
Izik Eidus210c7c42007-10-24 23:52:57 +02001062
1063}
Sheng Yangf78e0e22007-10-29 09:40:42 +08001064EXPORT_SYMBOL_GPL(__kvm_set_memory_region);
1065
1066int kvm_set_memory_region(struct kvm *kvm,
1067 struct kvm_userspace_memory_region *mem,
1068 int user_alloc)
1069{
1070 int r;
1071
Izik Eidus72dc67a2008-02-10 18:04:15 +02001072 down_write(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +08001073 r = __kvm_set_memory_region(kvm, mem, user_alloc);
Izik Eidus72dc67a2008-02-10 18:04:15 +02001074 up_write(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +08001075 return r;
1076}
Izik Eidus210c7c42007-10-24 23:52:57 +02001077EXPORT_SYMBOL_GPL(kvm_set_memory_region);
1078
Carsten Otte1fe779f2007-10-29 16:08:35 +01001079int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
1080 struct
1081 kvm_userspace_memory_region *mem,
1082 int user_alloc)
Izik Eidus210c7c42007-10-24 23:52:57 +02001083{
Izik Eiduse0d62c72007-10-24 23:57:46 +02001084 if (mem->slot >= KVM_MEMORY_SLOTS)
1085 return -EINVAL;
Izik Eidus210c7c42007-10-24 23:52:57 +02001086 return kvm_set_memory_region(kvm, mem, user_alloc);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001087}
1088
Zhang Xiantao5bb064d2007-11-18 20:29:43 +08001089int kvm_get_dirty_log(struct kvm *kvm,
1090 struct kvm_dirty_log *log, int *is_dirty)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001091{
1092 struct kvm_memory_slot *memslot;
1093 int r, i;
1094 int n;
1095 unsigned long any = 0;
1096
Avi Kivity6aa8b732006-12-10 02:21:36 -08001097 r = -EINVAL;
1098 if (log->slot >= KVM_MEMORY_SLOTS)
1099 goto out;
1100
1101 memslot = &kvm->memslots[log->slot];
1102 r = -ENOENT;
1103 if (!memslot->dirty_bitmap)
1104 goto out;
1105
Uri Lublincd1a4a92007-02-22 16:43:09 +02001106 n = ALIGN(memslot->npages, BITS_PER_LONG) / 8;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001107
Uri Lublincd1a4a92007-02-22 16:43:09 +02001108 for (i = 0; !any && i < n/sizeof(long); ++i)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001109 any = memslot->dirty_bitmap[i];
1110
1111 r = -EFAULT;
1112 if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
1113 goto out;
1114
Zhang Xiantao5bb064d2007-11-18 20:29:43 +08001115 if (any)
1116 *is_dirty = 1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001117
1118 r = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001119out:
Avi Kivity6aa8b732006-12-10 02:21:36 -08001120 return r;
1121}
1122
Izik Eiduscea7bb22007-10-17 19:17:48 +02001123int is_error_page(struct page *page)
1124{
1125 return page == bad_page;
1126}
1127EXPORT_SYMBOL_GPL(is_error_page);
1128
Anthony Liguori35149e22008-04-02 14:46:56 -05001129int is_error_pfn(pfn_t pfn)
1130{
1131 return pfn == bad_pfn;
1132}
1133EXPORT_SYMBOL_GPL(is_error_pfn);
1134
Izik Eidusf9d46eb2007-11-11 22:02:22 +02001135static inline unsigned long bad_hva(void)
1136{
1137 return PAGE_OFFSET;
1138}
1139
1140int kvm_is_error_hva(unsigned long addr)
1141{
1142 return addr == bad_hva();
1143}
1144EXPORT_SYMBOL_GPL(kvm_is_error_hva);
1145
Izik Eidus28430992008-10-03 17:40:32 +03001146struct kvm_memory_slot *gfn_to_memslot_unaliased(struct kvm *kvm, gfn_t gfn)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001147{
1148 int i;
1149
1150 for (i = 0; i < kvm->nmemslots; ++i) {
1151 struct kvm_memory_slot *memslot = &kvm->memslots[i];
1152
1153 if (gfn >= memslot->base_gfn
1154 && gfn < memslot->base_gfn + memslot->npages)
1155 return memslot;
1156 }
Al Viro8b6d44c2007-02-09 16:38:40 +00001157 return NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001158}
Izik Eidus28430992008-10-03 17:40:32 +03001159EXPORT_SYMBOL_GPL(gfn_to_memslot_unaliased);
Avi Kivitye8207542007-03-30 16:54:30 +03001160
1161struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn)
1162{
1163 gfn = unalias_gfn(kvm, gfn);
Izik Eidus28430992008-10-03 17:40:32 +03001164 return gfn_to_memslot_unaliased(kvm, gfn);
Avi Kivitye8207542007-03-30 16:54:30 +03001165}
Avi Kivity6aa8b732006-12-10 02:21:36 -08001166
Izik Eiduse0d62c72007-10-24 23:57:46 +02001167int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
1168{
1169 int i;
1170
1171 gfn = unalias_gfn(kvm, gfn);
1172 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
1173 struct kvm_memory_slot *memslot = &kvm->memslots[i];
1174
1175 if (gfn >= memslot->base_gfn
1176 && gfn < memslot->base_gfn + memslot->npages)
1177 return 1;
1178 }
1179 return 0;
1180}
1181EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);
1182
Marcelo Tosatti05da4552008-02-23 11:44:30 -03001183unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
Izik Eidus539cb662007-11-11 22:05:04 +02001184{
1185 struct kvm_memory_slot *slot;
1186
1187 gfn = unalias_gfn(kvm, gfn);
Izik Eidus28430992008-10-03 17:40:32 +03001188 slot = gfn_to_memslot_unaliased(kvm, gfn);
Izik Eidus539cb662007-11-11 22:05:04 +02001189 if (!slot)
1190 return bad_hva();
1191 return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE);
1192}
Sheng Yang0d150292008-04-25 21:44:50 +08001193EXPORT_SYMBOL_GPL(gfn_to_hva);
Izik Eidus539cb662007-11-11 22:05:04 +02001194
Anthony Liguori35149e22008-04-02 14:46:56 -05001195pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
Avi Kivity954bbbc22007-03-30 14:02:32 +03001196{
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001197 struct page *page[1];
Izik Eidus539cb662007-11-11 22:05:04 +02001198 unsigned long addr;
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001199 int npages;
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001200 pfn_t pfn;
Avi Kivity954bbbc22007-03-30 14:02:32 +03001201
Avi Kivity60395222007-10-21 11:03:36 +02001202 might_sleep();
1203
Izik Eidus539cb662007-11-11 22:05:04 +02001204 addr = gfn_to_hva(kvm, gfn);
1205 if (kvm_is_error_hva(addr)) {
Izik Eidus8a7ae052007-10-18 11:09:33 +02001206 get_page(bad_page);
Anthony Liguori35149e22008-04-02 14:46:56 -05001207 return page_to_pfn(bad_page);
Izik Eidus8a7ae052007-10-18 11:09:33 +02001208 }
Izik Eidus8a7ae052007-10-18 11:09:33 +02001209
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001210 npages = get_user_pages_fast(addr, 1, 1, page);
Izik Eidus539cb662007-11-11 22:05:04 +02001211
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001212 if (unlikely(npages != 1)) {
1213 struct vm_area_struct *vma;
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001214
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001215 down_read(&current->mm->mmap_sem);
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001216 vma = find_vma(current->mm, addr);
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001217
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001218 if (vma == NULL || addr < vma->vm_start ||
1219 !(vma->vm_flags & VM_PFNMAP)) {
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001220 up_read(&current->mm->mmap_sem);
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001221 get_page(bad_page);
1222 return page_to_pfn(bad_page);
1223 }
1224
1225 pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001226 up_read(&current->mm->mmap_sem);
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001227 BUG_ON(!kvm_is_mmio_pfn(pfn));
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001228 } else
1229 pfn = page_to_pfn(page[0]);
1230
1231 return pfn;
Anthony Liguori35149e22008-04-02 14:46:56 -05001232}
1233
1234EXPORT_SYMBOL_GPL(gfn_to_pfn);
1235
1236struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
1237{
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001238 pfn_t pfn;
1239
1240 pfn = gfn_to_pfn(kvm, gfn);
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001241 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001242 return pfn_to_page(pfn);
1243
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001244 WARN_ON(kvm_is_mmio_pfn(pfn));
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001245
1246 get_page(bad_page);
1247 return bad_page;
Avi Kivity954bbbc22007-03-30 14:02:32 +03001248}
Anthony Liguoriaab61cc2007-10-29 15:15:20 -05001249
Avi Kivity954bbbc22007-03-30 14:02:32 +03001250EXPORT_SYMBOL_GPL(gfn_to_page);
1251
Izik Eidusb4231d62007-11-20 11:49:33 +02001252void kvm_release_page_clean(struct page *page)
1253{
Anthony Liguori35149e22008-04-02 14:46:56 -05001254 kvm_release_pfn_clean(page_to_pfn(page));
Izik Eidusb4231d62007-11-20 11:49:33 +02001255}
1256EXPORT_SYMBOL_GPL(kvm_release_page_clean);
1257
Anthony Liguori35149e22008-04-02 14:46:56 -05001258void kvm_release_pfn_clean(pfn_t pfn)
1259{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001260 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001261 put_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001262}
1263EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);
1264
Izik Eidusb4231d62007-11-20 11:49:33 +02001265void kvm_release_page_dirty(struct page *page)
Izik Eidus8a7ae052007-10-18 11:09:33 +02001266{
Anthony Liguori35149e22008-04-02 14:46:56 -05001267 kvm_release_pfn_dirty(page_to_pfn(page));
Izik Eidus8a7ae052007-10-18 11:09:33 +02001268}
Izik Eidusb4231d62007-11-20 11:49:33 +02001269EXPORT_SYMBOL_GPL(kvm_release_page_dirty);
Izik Eidus8a7ae052007-10-18 11:09:33 +02001270
Anthony Liguori35149e22008-04-02 14:46:56 -05001271void kvm_release_pfn_dirty(pfn_t pfn)
1272{
1273 kvm_set_pfn_dirty(pfn);
1274 kvm_release_pfn_clean(pfn);
1275}
1276EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
1277
1278void kvm_set_page_dirty(struct page *page)
1279{
1280 kvm_set_pfn_dirty(page_to_pfn(page));
1281}
1282EXPORT_SYMBOL_GPL(kvm_set_page_dirty);
1283
1284void kvm_set_pfn_dirty(pfn_t pfn)
1285{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001286 if (!kvm_is_mmio_pfn(pfn)) {
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001287 struct page *page = pfn_to_page(pfn);
1288 if (!PageReserved(page))
1289 SetPageDirty(page);
1290 }
Anthony Liguori35149e22008-04-02 14:46:56 -05001291}
1292EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);
1293
1294void kvm_set_pfn_accessed(pfn_t pfn)
1295{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001296 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001297 mark_page_accessed(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001298}
1299EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);
1300
1301void kvm_get_pfn(pfn_t pfn)
1302{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001303 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001304 get_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001305}
1306EXPORT_SYMBOL_GPL(kvm_get_pfn);
1307
Izik Eidus195aefd2007-10-01 22:14:18 +02001308static int next_segment(unsigned long len, int offset)
1309{
1310 if (len > PAGE_SIZE - offset)
1311 return PAGE_SIZE - offset;
1312 else
1313 return len;
1314}
1315
1316int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
1317 int len)
1318{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001319 int r;
1320 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001321
Izik Eiduse0506bc2007-11-11 22:10:22 +02001322 addr = gfn_to_hva(kvm, gfn);
1323 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001324 return -EFAULT;
Izik Eiduse0506bc2007-11-11 22:10:22 +02001325 r = copy_from_user(data, (void __user *)addr + offset, len);
1326 if (r)
1327 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001328 return 0;
1329}
1330EXPORT_SYMBOL_GPL(kvm_read_guest_page);
1331
1332int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len)
1333{
1334 gfn_t gfn = gpa >> PAGE_SHIFT;
1335 int seg;
1336 int offset = offset_in_page(gpa);
1337 int ret;
1338
1339 while ((seg = next_segment(len, offset)) != 0) {
1340 ret = kvm_read_guest_page(kvm, gfn, data, offset, seg);
1341 if (ret < 0)
1342 return ret;
1343 offset = 0;
1344 len -= seg;
1345 data += seg;
1346 ++gfn;
1347 }
1348 return 0;
1349}
1350EXPORT_SYMBOL_GPL(kvm_read_guest);
1351
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001352int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
1353 unsigned long len)
1354{
1355 int r;
1356 unsigned long addr;
1357 gfn_t gfn = gpa >> PAGE_SHIFT;
1358 int offset = offset_in_page(gpa);
1359
1360 addr = gfn_to_hva(kvm, gfn);
1361 if (kvm_is_error_hva(addr))
1362 return -EFAULT;
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001363 pagefault_disable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001364 r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001365 pagefault_enable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001366 if (r)
1367 return -EFAULT;
1368 return 0;
1369}
1370EXPORT_SYMBOL(kvm_read_guest_atomic);
1371
Izik Eidus195aefd2007-10-01 22:14:18 +02001372int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
1373 int offset, int len)
1374{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001375 int r;
1376 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001377
Izik Eiduse0506bc2007-11-11 22:10:22 +02001378 addr = gfn_to_hva(kvm, gfn);
1379 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001380 return -EFAULT;
Izik Eiduse0506bc2007-11-11 22:10:22 +02001381 r = copy_to_user((void __user *)addr + offset, data, len);
1382 if (r)
1383 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001384 mark_page_dirty(kvm, gfn);
1385 return 0;
1386}
1387EXPORT_SYMBOL_GPL(kvm_write_guest_page);
1388
1389int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
1390 unsigned long len)
1391{
1392 gfn_t gfn = gpa >> PAGE_SHIFT;
1393 int seg;
1394 int offset = offset_in_page(gpa);
1395 int ret;
1396
1397 while ((seg = next_segment(len, offset)) != 0) {
1398 ret = kvm_write_guest_page(kvm, gfn, data, offset, seg);
1399 if (ret < 0)
1400 return ret;
1401 offset = 0;
1402 len -= seg;
1403 data += seg;
1404 ++gfn;
1405 }
1406 return 0;
1407}
1408
1409int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
1410{
Izik Eidus3e021bf2007-11-19 11:16:57 +02001411 return kvm_write_guest_page(kvm, gfn, empty_zero_page, offset, len);
Izik Eidus195aefd2007-10-01 22:14:18 +02001412}
1413EXPORT_SYMBOL_GPL(kvm_clear_guest_page);
1414
1415int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len)
1416{
1417 gfn_t gfn = gpa >> PAGE_SHIFT;
1418 int seg;
1419 int offset = offset_in_page(gpa);
1420 int ret;
1421
1422 while ((seg = next_segment(len, offset)) != 0) {
1423 ret = kvm_clear_guest_page(kvm, gfn, offset, seg);
1424 if (ret < 0)
1425 return ret;
1426 offset = 0;
1427 len -= seg;
1428 ++gfn;
1429 }
1430 return 0;
1431}
1432EXPORT_SYMBOL_GPL(kvm_clear_guest);
1433
Avi Kivity6aa8b732006-12-10 02:21:36 -08001434void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
1435{
Nguyen Anh Quynh31389942007-06-05 10:35:19 +03001436 struct kvm_memory_slot *memslot;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001437
Uri Lublin3b6fff12007-10-30 10:42:09 +02001438 gfn = unalias_gfn(kvm, gfn);
Izik Eidus28430992008-10-03 17:40:32 +03001439 memslot = gfn_to_memslot_unaliased(kvm, gfn);
Rusty Russell7e9d6192007-07-31 20:41:14 +10001440 if (memslot && memslot->dirty_bitmap) {
1441 unsigned long rel_gfn = gfn - memslot->base_gfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001442
Rusty Russell7e9d6192007-07-31 20:41:14 +10001443 /* avoid RMW */
1444 if (!test_bit(rel_gfn, memslot->dirty_bitmap))
1445 set_bit(rel_gfn, memslot->dirty_bitmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001446 }
1447}
1448
Eddie Dongb6958ce2007-07-18 12:15:21 +03001449/*
1450 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
1451 */
Hollis Blanchard8776e512007-10-31 17:24:24 -05001452void kvm_vcpu_block(struct kvm_vcpu *vcpu)
Eddie Dongb6958ce2007-07-18 12:15:21 +03001453{
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001454 DEFINE_WAIT(wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001455
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001456 for (;;) {
1457 prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001458
Marcelo Tosattid7690172008-09-08 15:23:48 -03001459 if (kvm_cpu_has_interrupt(vcpu) ||
1460 kvm_cpu_has_pending_timer(vcpu) ||
1461 kvm_arch_vcpu_runnable(vcpu)) {
1462 set_bit(KVM_REQ_UNHALT, &vcpu->requests);
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001463 break;
Marcelo Tosattid7690172008-09-08 15:23:48 -03001464 }
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001465 if (signal_pending(current))
1466 break;
1467
Eddie Dongb6958ce2007-07-18 12:15:21 +03001468 vcpu_put(vcpu);
1469 schedule();
1470 vcpu_load(vcpu);
1471 }
1472
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001473 finish_wait(&vcpu->wq, &wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001474}
1475
Avi Kivity6aa8b732006-12-10 02:21:36 -08001476void kvm_resched(struct kvm_vcpu *vcpu)
1477{
Yaozu Dong3fca0362007-04-25 16:49:19 +03001478 if (!need_resched())
1479 return;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001480 cond_resched();
Avi Kivity6aa8b732006-12-10 02:21:36 -08001481}
1482EXPORT_SYMBOL_GPL(kvm_resched);
1483
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001484static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001485{
1486 struct kvm_vcpu *vcpu = vma->vm_file->private_data;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001487 struct page *page;
1488
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001489 if (vmf->pgoff == 0)
Avi Kivity039576c2007-03-20 12:46:50 +02001490 page = virt_to_page(vcpu->run);
Avi Kivity09566762008-01-23 18:14:23 +02001491#ifdef CONFIG_X86
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001492 else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
Zhang Xiantaoad312c72007-12-13 23:50:52 +08001493 page = virt_to_page(vcpu->arch.pio_data);
Avi Kivity09566762008-01-23 18:14:23 +02001494#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02001495#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1496 else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
1497 page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
1498#endif
Avi Kivity039576c2007-03-20 12:46:50 +02001499 else
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001500 return VM_FAULT_SIGBUS;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001501 get_page(page);
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001502 vmf->page = page;
1503 return 0;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001504}
1505
1506static struct vm_operations_struct kvm_vcpu_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001507 .fault = kvm_vcpu_fault,
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001508};
1509
1510static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma)
1511{
1512 vma->vm_ops = &kvm_vcpu_vm_ops;
1513 return 0;
1514}
1515
Avi Kivitybccf2152007-02-21 18:04:26 +02001516static int kvm_vcpu_release(struct inode *inode, struct file *filp)
1517{
1518 struct kvm_vcpu *vcpu = filp->private_data;
1519
Al Viro66c0b392008-04-19 20:33:56 +01001520 kvm_put_kvm(vcpu->kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001521 return 0;
1522}
1523
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01001524static struct file_operations kvm_vcpu_fops = {
Avi Kivitybccf2152007-02-21 18:04:26 +02001525 .release = kvm_vcpu_release,
1526 .unlocked_ioctl = kvm_vcpu_ioctl,
1527 .compat_ioctl = kvm_vcpu_ioctl,
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001528 .mmap = kvm_vcpu_mmap,
Avi Kivitybccf2152007-02-21 18:04:26 +02001529};
1530
1531/*
1532 * Allocates an inode for the vcpu.
1533 */
1534static int create_vcpu_fd(struct kvm_vcpu *vcpu)
1535{
Ulrich Drepper7d9dbca2008-07-23 21:29:22 -07001536 int fd = anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, 0);
Al Viro2030a422008-02-23 06:46:49 -05001537 if (fd < 0)
Al Viro66c0b392008-04-19 20:33:56 +01001538 kvm_put_kvm(vcpu->kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001539 return fd;
Avi Kivitybccf2152007-02-21 18:04:26 +02001540}
1541
Avi Kivityc5ea7662007-02-20 18:41:05 +02001542/*
1543 * Creates some virtual cpus. Good luck creating more than one.
1544 */
1545static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, int n)
1546{
1547 int r;
1548 struct kvm_vcpu *vcpu;
1549
Avi Kivityc5ea7662007-02-20 18:41:05 +02001550 if (!valid_vcpu(n))
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001551 return -EINVAL;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001552
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001553 vcpu = kvm_arch_vcpu_create(kvm, n);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001554 if (IS_ERR(vcpu))
1555 return PTR_ERR(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001556
Avi Kivity15ad7142007-07-11 18:17:21 +03001557 preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);
1558
Avi Kivity26e52152007-11-20 15:30:24 +02001559 r = kvm_arch_vcpu_setup(vcpu);
1560 if (r)
Glauber Costa7d8fece2008-09-17 23:16:59 -03001561 return r;
Avi Kivity26e52152007-11-20 15:30:24 +02001562
Shaohua Li11ec2802007-07-23 14:51:37 +08001563 mutex_lock(&kvm->lock);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001564 if (kvm->vcpus[n]) {
1565 r = -EEXIST;
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001566 goto vcpu_destroy;
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001567 }
1568 kvm->vcpus[n] = vcpu;
Shaohua Li11ec2802007-07-23 14:51:37 +08001569 mutex_unlock(&kvm->lock);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001570
1571 /* Now it's all set up, let userspace reach it */
Al Viro66c0b392008-04-19 20:33:56 +01001572 kvm_get_kvm(kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001573 r = create_vcpu_fd(vcpu);
1574 if (r < 0)
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001575 goto unlink;
Avi Kivitybccf2152007-02-21 18:04:26 +02001576 return r;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001577
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001578unlink:
Shaohua Li11ec2802007-07-23 14:51:37 +08001579 mutex_lock(&kvm->lock);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001580 kvm->vcpus[n] = NULL;
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001581vcpu_destroy:
Glauber Costa7d8fece2008-09-17 23:16:59 -03001582 mutex_unlock(&kvm->lock);
Hollis Blanchardd40ccc62007-11-19 14:04:43 -06001583 kvm_arch_vcpu_destroy(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001584 return r;
1585}
1586
Avi Kivity1961d272007-03-05 19:46:05 +02001587static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset)
1588{
1589 if (sigset) {
1590 sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP));
1591 vcpu->sigset_active = 1;
1592 vcpu->sigset = *sigset;
1593 } else
1594 vcpu->sigset_active = 0;
1595 return 0;
1596}
1597
Avi Kivitybccf2152007-02-21 18:04:26 +02001598static long kvm_vcpu_ioctl(struct file *filp,
1599 unsigned int ioctl, unsigned long arg)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001600{
Avi Kivitybccf2152007-02-21 18:04:26 +02001601 struct kvm_vcpu *vcpu = filp->private_data;
Al Viro2f366982007-02-09 16:38:35 +00001602 void __user *argp = (void __user *)arg;
Carsten Otte313a3dc2007-10-11 19:16:52 +02001603 int r;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001604 struct kvm_fpu *fpu = NULL;
1605 struct kvm_sregs *kvm_sregs = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001606
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001607 if (vcpu->kvm->mm != current->mm)
1608 return -EIO;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001609 switch (ioctl) {
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001610 case KVM_RUN:
Avi Kivityf0fe5102007-03-07 13:11:17 +02001611 r = -EINVAL;
1612 if (arg)
1613 goto out;
Hollis Blanchardb6c7a5d2007-11-01 14:16:10 -05001614 r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001615 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001616 case KVM_GET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001617 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001618
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001619 r = -ENOMEM;
1620 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
1621 if (!kvm_regs)
1622 goto out;
1623 r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001624 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001625 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001626 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001627 if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
1628 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001629 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001630out_free1:
1631 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001632 break;
1633 }
1634 case KVM_SET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001635 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001636
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001637 r = -ENOMEM;
1638 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
1639 if (!kvm_regs)
1640 goto out;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001641 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001642 if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs)))
1643 goto out_free2;
1644 r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001645 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001646 goto out_free2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001647 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001648out_free2:
1649 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001650 break;
1651 }
1652 case KVM_GET_SREGS: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001653 kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL);
1654 r = -ENOMEM;
1655 if (!kvm_sregs)
1656 goto out;
1657 r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001658 if (r)
1659 goto out;
1660 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001661 if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001662 goto out;
1663 r = 0;
1664 break;
1665 }
1666 case KVM_SET_SREGS: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001667 kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL);
1668 r = -ENOMEM;
1669 if (!kvm_sregs)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001670 goto out;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001671 r = -EFAULT;
1672 if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs)))
1673 goto out;
1674 r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001675 if (r)
1676 goto out;
1677 r = 0;
1678 break;
1679 }
Marcelo Tosatti62d9f0d2008-04-11 13:24:45 -03001680 case KVM_GET_MP_STATE: {
1681 struct kvm_mp_state mp_state;
1682
1683 r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state);
1684 if (r)
1685 goto out;
1686 r = -EFAULT;
1687 if (copy_to_user(argp, &mp_state, sizeof mp_state))
1688 goto out;
1689 r = 0;
1690 break;
1691 }
1692 case KVM_SET_MP_STATE: {
1693 struct kvm_mp_state mp_state;
1694
1695 r = -EFAULT;
1696 if (copy_from_user(&mp_state, argp, sizeof mp_state))
1697 goto out;
1698 r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
1699 if (r)
1700 goto out;
1701 r = 0;
1702 break;
1703 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08001704 case KVM_TRANSLATE: {
1705 struct kvm_translation tr;
1706
1707 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001708 if (copy_from_user(&tr, argp, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001709 goto out;
Zhang Xiantao8b006792007-11-16 13:05:55 +08001710 r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001711 if (r)
1712 goto out;
1713 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001714 if (copy_to_user(argp, &tr, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001715 goto out;
1716 r = 0;
1717 break;
1718 }
Jan Kiszkad0bfb942008-12-15 13:52:10 +01001719 case KVM_SET_GUEST_DEBUG: {
1720 struct kvm_guest_debug dbg;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001721
1722 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001723 if (copy_from_user(&dbg, argp, sizeof dbg))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001724 goto out;
Jan Kiszkad0bfb942008-12-15 13:52:10 +01001725 r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001726 if (r)
1727 goto out;
1728 r = 0;
1729 break;
1730 }
Avi Kivity1961d272007-03-05 19:46:05 +02001731 case KVM_SET_SIGNAL_MASK: {
1732 struct kvm_signal_mask __user *sigmask_arg = argp;
1733 struct kvm_signal_mask kvm_sigmask;
1734 sigset_t sigset, *p;
1735
1736 p = NULL;
1737 if (argp) {
1738 r = -EFAULT;
1739 if (copy_from_user(&kvm_sigmask, argp,
1740 sizeof kvm_sigmask))
1741 goto out;
1742 r = -EINVAL;
1743 if (kvm_sigmask.len != sizeof sigset)
1744 goto out;
1745 r = -EFAULT;
1746 if (copy_from_user(&sigset, sigmask_arg->sigset,
1747 sizeof sigset))
1748 goto out;
1749 p = &sigset;
1750 }
1751 r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
1752 break;
1753 }
Avi Kivityb8836732007-04-01 16:34:31 +03001754 case KVM_GET_FPU: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001755 fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL);
1756 r = -ENOMEM;
1757 if (!fpu)
1758 goto out;
1759 r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001760 if (r)
1761 goto out;
1762 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001763 if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
Avi Kivityb8836732007-04-01 16:34:31 +03001764 goto out;
1765 r = 0;
1766 break;
1767 }
1768 case KVM_SET_FPU: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001769 fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL);
1770 r = -ENOMEM;
1771 if (!fpu)
Avi Kivityb8836732007-04-01 16:34:31 +03001772 goto out;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001773 r = -EFAULT;
1774 if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu)))
1775 goto out;
1776 r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001777 if (r)
1778 goto out;
1779 r = 0;
1780 break;
1781 }
Avi Kivitybccf2152007-02-21 18:04:26 +02001782 default:
Carsten Otte313a3dc2007-10-11 19:16:52 +02001783 r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
Avi Kivitybccf2152007-02-21 18:04:26 +02001784 }
1785out:
Dave Hansenfa3795a2008-08-11 10:01:46 -07001786 kfree(fpu);
1787 kfree(kvm_sregs);
Avi Kivitybccf2152007-02-21 18:04:26 +02001788 return r;
1789}
1790
1791static long kvm_vm_ioctl(struct file *filp,
1792 unsigned int ioctl, unsigned long arg)
1793{
1794 struct kvm *kvm = filp->private_data;
1795 void __user *argp = (void __user *)arg;
Carsten Otte1fe779f2007-10-29 16:08:35 +01001796 int r;
Avi Kivitybccf2152007-02-21 18:04:26 +02001797
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001798 if (kvm->mm != current->mm)
1799 return -EIO;
Avi Kivitybccf2152007-02-21 18:04:26 +02001800 switch (ioctl) {
1801 case KVM_CREATE_VCPU:
1802 r = kvm_vm_ioctl_create_vcpu(kvm, arg);
1803 if (r < 0)
1804 goto out;
1805 break;
Izik Eidus6fc138d2007-10-09 19:20:39 +02001806 case KVM_SET_USER_MEMORY_REGION: {
1807 struct kvm_userspace_memory_region kvm_userspace_mem;
1808
1809 r = -EFAULT;
1810 if (copy_from_user(&kvm_userspace_mem, argp,
1811 sizeof kvm_userspace_mem))
1812 goto out;
1813
1814 r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001815 if (r)
1816 goto out;
1817 break;
1818 }
1819 case KVM_GET_DIRTY_LOG: {
1820 struct kvm_dirty_log log;
1821
1822 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001823 if (copy_from_user(&log, argp, sizeof log))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001824 goto out;
Avi Kivity2c6f5df2007-02-20 18:27:58 +02001825 r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001826 if (r)
1827 goto out;
1828 break;
1829 }
Laurent Vivier5f94c172008-05-30 16:05:54 +02001830#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1831 case KVM_REGISTER_COALESCED_MMIO: {
1832 struct kvm_coalesced_mmio_zone zone;
1833 r = -EFAULT;
1834 if (copy_from_user(&zone, argp, sizeof zone))
1835 goto out;
1836 r = -ENXIO;
1837 r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
1838 if (r)
1839 goto out;
1840 r = 0;
1841 break;
1842 }
1843 case KVM_UNREGISTER_COALESCED_MMIO: {
1844 struct kvm_coalesced_mmio_zone zone;
1845 r = -EFAULT;
1846 if (copy_from_user(&zone, argp, sizeof zone))
1847 goto out;
1848 r = -ENXIO;
1849 r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
1850 if (r)
1851 goto out;
1852 r = 0;
1853 break;
1854 }
1855#endif
Xiantao Zhang8a98f662008-10-06 13:47:38 +08001856#ifdef KVM_CAP_DEVICE_ASSIGNMENT
1857 case KVM_ASSIGN_PCI_DEVICE: {
1858 struct kvm_assigned_pci_dev assigned_dev;
1859
1860 r = -EFAULT;
1861 if (copy_from_user(&assigned_dev, argp, sizeof assigned_dev))
1862 goto out;
1863 r = kvm_vm_ioctl_assign_device(kvm, &assigned_dev);
1864 if (r)
1865 goto out;
1866 break;
1867 }
1868 case KVM_ASSIGN_IRQ: {
1869 struct kvm_assigned_irq assigned_irq;
1870
1871 r = -EFAULT;
1872 if (copy_from_user(&assigned_irq, argp, sizeof assigned_irq))
1873 goto out;
1874 r = kvm_vm_ioctl_assign_irq(kvm, &assigned_irq);
1875 if (r)
1876 goto out;
1877 break;
1878 }
1879#endif
Weidong Han0a920352008-12-02 21:24:23 +08001880#ifdef KVM_CAP_DEVICE_DEASSIGNMENT
1881 case KVM_DEASSIGN_PCI_DEVICE: {
1882 struct kvm_assigned_pci_dev assigned_dev;
1883
1884 r = -EFAULT;
1885 if (copy_from_user(&assigned_dev, argp, sizeof assigned_dev))
1886 goto out;
1887 r = kvm_vm_ioctl_deassign_device(kvm, &assigned_dev);
1888 if (r)
1889 goto out;
1890 break;
1891 }
1892#endif
Avi Kivity399ec802008-11-19 13:58:46 +02001893#ifdef KVM_CAP_IRQ_ROUTING
1894 case KVM_SET_GSI_ROUTING: {
1895 struct kvm_irq_routing routing;
1896 struct kvm_irq_routing __user *urouting;
1897 struct kvm_irq_routing_entry *entries;
1898
1899 r = -EFAULT;
1900 if (copy_from_user(&routing, argp, sizeof(routing)))
1901 goto out;
1902 r = -EINVAL;
1903 if (routing.nr >= KVM_MAX_IRQ_ROUTES)
1904 goto out;
1905 if (routing.flags)
1906 goto out;
1907 r = -ENOMEM;
1908 entries = vmalloc(routing.nr * sizeof(*entries));
1909 if (!entries)
1910 goto out;
1911 r = -EFAULT;
1912 urouting = argp;
1913 if (copy_from_user(entries, urouting->entries,
1914 routing.nr * sizeof(*entries)))
1915 goto out_free_irq_routing;
1916 r = kvm_set_irq_routing(kvm, entries, routing.nr,
1917 routing.flags);
1918 out_free_irq_routing:
1919 vfree(entries);
1920 break;
1921 }
1922#endif
Avi Kivityf17abe92007-02-21 19:28:04 +02001923 default:
Carsten Otte1fe779f2007-10-29 16:08:35 +01001924 r = kvm_arch_vm_ioctl(filp, ioctl, arg);
Avi Kivityf17abe92007-02-21 19:28:04 +02001925 }
1926out:
1927 return r;
1928}
1929
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001930static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivityf17abe92007-02-21 19:28:04 +02001931{
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03001932 struct page *page[1];
1933 unsigned long addr;
1934 int npages;
1935 gfn_t gfn = vmf->pgoff;
Avi Kivityf17abe92007-02-21 19:28:04 +02001936 struct kvm *kvm = vma->vm_file->private_data;
Avi Kivityf17abe92007-02-21 19:28:04 +02001937
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03001938 addr = gfn_to_hva(kvm, gfn);
1939 if (kvm_is_error_hva(addr))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001940 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03001941
1942 npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page,
1943 NULL);
1944 if (unlikely(npages != 1))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001945 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03001946
1947 vmf->page = page[0];
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001948 return 0;
Avi Kivityf17abe92007-02-21 19:28:04 +02001949}
1950
1951static struct vm_operations_struct kvm_vm_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001952 .fault = kvm_vm_fault,
Avi Kivityf17abe92007-02-21 19:28:04 +02001953};
1954
1955static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma)
1956{
1957 vma->vm_ops = &kvm_vm_vm_ops;
1958 return 0;
1959}
1960
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01001961static struct file_operations kvm_vm_fops = {
Avi Kivityf17abe92007-02-21 19:28:04 +02001962 .release = kvm_vm_release,
1963 .unlocked_ioctl = kvm_vm_ioctl,
1964 .compat_ioctl = kvm_vm_ioctl,
1965 .mmap = kvm_vm_mmap,
1966};
1967
1968static int kvm_dev_ioctl_create_vm(void)
1969{
Al Viro2030a422008-02-23 06:46:49 -05001970 int fd;
Avi Kivityf17abe92007-02-21 19:28:04 +02001971 struct kvm *kvm;
1972
Avi Kivityf17abe92007-02-21 19:28:04 +02001973 kvm = kvm_create_vm();
Avi Kivityd6d28162007-06-28 08:38:16 -04001974 if (IS_ERR(kvm))
1975 return PTR_ERR(kvm);
Ulrich Drepper7d9dbca2008-07-23 21:29:22 -07001976 fd = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, 0);
Al Viro2030a422008-02-23 06:46:49 -05001977 if (fd < 0)
Al Viro66c0b392008-04-19 20:33:56 +01001978 kvm_put_kvm(kvm);
Avi Kivityf17abe92007-02-21 19:28:04 +02001979
Avi Kivityf17abe92007-02-21 19:28:04 +02001980 return fd;
Avi Kivityf17abe92007-02-21 19:28:04 +02001981}
1982
Avi Kivity1a811b62008-12-08 18:25:27 +02001983static long kvm_dev_ioctl_check_extension_generic(long arg)
1984{
1985 switch (arg) {
Avi Kivityca9edae2008-12-08 18:29:29 +02001986 case KVM_CAP_USER_MEMORY:
Avi Kivity1a811b62008-12-08 18:25:27 +02001987 case KVM_CAP_DESTROY_MEMORY_REGION_WORKS:
1988 return 1;
Avi Kivity399ec802008-11-19 13:58:46 +02001989#ifdef CONFIG_HAVE_KVM_IRQCHIP
1990 case KVM_CAP_IRQ_ROUTING:
Sheng Yang36463142009-03-16 16:33:43 +08001991 return KVM_MAX_IRQ_ROUTES;
Avi Kivity399ec802008-11-19 13:58:46 +02001992#endif
Avi Kivity1a811b62008-12-08 18:25:27 +02001993 default:
1994 break;
1995 }
1996 return kvm_dev_ioctl_check_extension(arg);
1997}
1998
Avi Kivityf17abe92007-02-21 19:28:04 +02001999static long kvm_dev_ioctl(struct file *filp,
2000 unsigned int ioctl, unsigned long arg)
2001{
Avi Kivity07c45a32007-03-07 13:05:38 +02002002 long r = -EINVAL;
Avi Kivityf17abe92007-02-21 19:28:04 +02002003
2004 switch (ioctl) {
2005 case KVM_GET_API_VERSION:
Avi Kivityf0fe5102007-03-07 13:11:17 +02002006 r = -EINVAL;
2007 if (arg)
2008 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02002009 r = KVM_API_VERSION;
2010 break;
2011 case KVM_CREATE_VM:
Avi Kivityf0fe5102007-03-07 13:11:17 +02002012 r = -EINVAL;
2013 if (arg)
2014 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02002015 r = kvm_dev_ioctl_create_vm();
2016 break;
Zhang Xiantao018d00d2007-11-15 23:07:47 +08002017 case KVM_CHECK_EXTENSION:
Avi Kivity1a811b62008-12-08 18:25:27 +02002018 r = kvm_dev_ioctl_check_extension_generic(arg);
Avi Kivity5d308f42007-03-01 17:56:20 +02002019 break;
Avi Kivity07c45a32007-03-07 13:05:38 +02002020 case KVM_GET_VCPU_MMAP_SIZE:
2021 r = -EINVAL;
2022 if (arg)
2023 goto out;
Avi Kivityadb1ff42008-01-24 15:13:08 +02002024 r = PAGE_SIZE; /* struct kvm_run */
2025#ifdef CONFIG_X86
2026 r += PAGE_SIZE; /* pio data page */
2027#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02002028#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
2029 r += PAGE_SIZE; /* coalesced mmio ring page */
2030#endif
Avi Kivity07c45a32007-03-07 13:05:38 +02002031 break;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04002032 case KVM_TRACE_ENABLE:
2033 case KVM_TRACE_PAUSE:
2034 case KVM_TRACE_DISABLE:
2035 r = kvm_trace_ioctl(ioctl, arg);
2036 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002037 default:
Carsten Otte043405e2007-10-10 17:16:19 +02002038 return kvm_arch_dev_ioctl(filp, ioctl, arg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002039 }
2040out:
2041 return r;
2042}
2043
Avi Kivity6aa8b732006-12-10 02:21:36 -08002044static struct file_operations kvm_chardev_ops = {
Avi Kivity6aa8b732006-12-10 02:21:36 -08002045 .unlocked_ioctl = kvm_dev_ioctl,
2046 .compat_ioctl = kvm_dev_ioctl,
Avi Kivity6aa8b732006-12-10 02:21:36 -08002047};
2048
2049static struct miscdevice kvm_dev = {
Avi Kivitybbe44322007-03-04 13:27:36 +02002050 KVM_MINOR,
Avi Kivity6aa8b732006-12-10 02:21:36 -08002051 "kvm",
2052 &kvm_chardev_ops,
2053};
2054
Avi Kivity1b6c0162007-05-24 13:03:52 +03002055static void hardware_enable(void *junk)
2056{
2057 int cpu = raw_smp_processor_id();
2058
Rusty Russell7f59f492008-12-07 21:25:45 +10302059 if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03002060 return;
Rusty Russell7f59f492008-12-07 21:25:45 +10302061 cpumask_set_cpu(cpu, cpus_hardware_enabled);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002062 kvm_arch_hardware_enable(NULL);
Avi Kivity1b6c0162007-05-24 13:03:52 +03002063}
2064
2065static void hardware_disable(void *junk)
2066{
2067 int cpu = raw_smp_processor_id();
2068
Rusty Russell7f59f492008-12-07 21:25:45 +10302069 if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03002070 return;
Rusty Russell7f59f492008-12-07 21:25:45 +10302071 cpumask_clear_cpu(cpu, cpus_hardware_enabled);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002072 kvm_arch_hardware_disable(NULL);
Avi Kivity1b6c0162007-05-24 13:03:52 +03002073}
2074
Avi Kivity774c47f2007-02-12 00:54:47 -08002075static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val,
2076 void *v)
2077{
2078 int cpu = (long)v;
2079
Avi Kivity1a6f4d72007-11-11 18:37:32 +02002080 val &= ~CPU_TASKS_FROZEN;
Avi Kivity774c47f2007-02-12 00:54:47 -08002081 switch (val) {
Avi Kivitycec9ad22007-05-24 13:11:41 +03002082 case CPU_DYING:
Avi Kivity6ec8a852007-08-19 15:57:26 +03002083 printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n",
2084 cpu);
2085 hardware_disable(NULL);
2086 break;
Avi Kivity774c47f2007-02-12 00:54:47 -08002087 case CPU_UP_CANCELED:
Jeremy Katz43934a32007-02-19 14:37:46 +02002088 printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n",
2089 cpu);
Jens Axboe8691e5a2008-06-06 11:18:06 +02002090 smp_call_function_single(cpu, hardware_disable, NULL, 1);
Avi Kivity774c47f2007-02-12 00:54:47 -08002091 break;
Jeremy Katz43934a32007-02-19 14:37:46 +02002092 case CPU_ONLINE:
2093 printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n",
2094 cpu);
Jens Axboe8691e5a2008-06-06 11:18:06 +02002095 smp_call_function_single(cpu, hardware_enable, NULL, 1);
Avi Kivity774c47f2007-02-12 00:54:47 -08002096 break;
2097 }
2098 return NOTIFY_OK;
2099}
2100
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002101
2102asmlinkage void kvm_handle_fault_on_reboot(void)
2103{
2104 if (kvm_rebooting)
2105 /* spin while reset goes on */
2106 while (true)
2107 ;
2108 /* Fault while not rebooting. We want the trace. */
2109 BUG();
2110}
2111EXPORT_SYMBOL_GPL(kvm_handle_fault_on_reboot);
2112
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002113static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
Mike Dayd77c26f2007-10-08 09:02:08 -04002114 void *v)
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002115{
2116 if (val == SYS_RESTART) {
2117 /*
2118 * Some (well, at least mine) BIOSes hang on reboot if
2119 * in vmx root mode.
2120 */
2121 printk(KERN_INFO "kvm: exiting hardware virtualization\n");
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002122 kvm_rebooting = true;
Jens Axboe15c8b6c2008-05-09 09:39:44 +02002123 on_each_cpu(hardware_disable, NULL, 1);
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002124 }
2125 return NOTIFY_OK;
2126}
2127
2128static struct notifier_block kvm_reboot_notifier = {
2129 .notifier_call = kvm_reboot,
2130 .priority = 0,
2131};
2132
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002133void kvm_io_bus_init(struct kvm_io_bus *bus)
2134{
2135 memset(bus, 0, sizeof(*bus));
2136}
2137
2138void kvm_io_bus_destroy(struct kvm_io_bus *bus)
2139{
2140 int i;
2141
2142 for (i = 0; i < bus->dev_count; i++) {
2143 struct kvm_io_device *pos = bus->devs[i];
2144
2145 kvm_iodevice_destructor(pos);
2146 }
2147}
2148
Laurent Vivier92760492008-05-30 16:05:53 +02002149struct kvm_io_device *kvm_io_bus_find_dev(struct kvm_io_bus *bus,
2150 gpa_t addr, int len, int is_write)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002151{
2152 int i;
2153
2154 for (i = 0; i < bus->dev_count; i++) {
2155 struct kvm_io_device *pos = bus->devs[i];
2156
Laurent Vivier92760492008-05-30 16:05:53 +02002157 if (pos->in_range(pos, addr, len, is_write))
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002158 return pos;
2159 }
2160
2161 return NULL;
2162}
2163
2164void kvm_io_bus_register_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev)
2165{
2166 BUG_ON(bus->dev_count > (NR_IOBUS_DEVS-1));
2167
2168 bus->devs[bus->dev_count++] = dev;
2169}
2170
Avi Kivity774c47f2007-02-12 00:54:47 -08002171static struct notifier_block kvm_cpu_notifier = {
2172 .notifier_call = kvm_cpu_hotplug,
2173 .priority = 20, /* must be > scheduler priority */
2174};
2175
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002176static int vm_stat_get(void *_offset, u64 *val)
Avi Kivityba1389b2007-11-18 16:24:12 +02002177{
2178 unsigned offset = (long)_offset;
Avi Kivityba1389b2007-11-18 16:24:12 +02002179 struct kvm *kvm;
2180
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002181 *val = 0;
Avi Kivityba1389b2007-11-18 16:24:12 +02002182 spin_lock(&kvm_lock);
2183 list_for_each_entry(kvm, &vm_list, vm_list)
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002184 *val += *(u32 *)((void *)kvm + offset);
Avi Kivityba1389b2007-11-18 16:24:12 +02002185 spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002186 return 0;
Avi Kivityba1389b2007-11-18 16:24:12 +02002187}
2188
2189DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n");
2190
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002191static int vcpu_stat_get(void *_offset, u64 *val)
Avi Kivity1165f5f2007-04-19 17:27:43 +03002192{
2193 unsigned offset = (long)_offset;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002194 struct kvm *kvm;
2195 struct kvm_vcpu *vcpu;
2196 int i;
2197
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002198 *val = 0;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002199 spin_lock(&kvm_lock);
2200 list_for_each_entry(kvm, &vm_list, vm_list)
2201 for (i = 0; i < KVM_MAX_VCPUS; ++i) {
Rusty Russellfb3f0f52007-07-27 17:16:56 +10002202 vcpu = kvm->vcpus[i];
2203 if (vcpu)
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002204 *val += *(u32 *)((void *)vcpu + offset);
Avi Kivity1165f5f2007-04-19 17:27:43 +03002205 }
2206 spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002207 return 0;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002208}
2209
Avi Kivityba1389b2007-11-18 16:24:12 +02002210DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n");
2211
2212static struct file_operations *stat_fops[] = {
2213 [KVM_STAT_VCPU] = &vcpu_stat_fops,
2214 [KVM_STAT_VM] = &vm_stat_fops,
2215};
Avi Kivity1165f5f2007-04-19 17:27:43 +03002216
Zhang Xiantaoa16b0432007-11-16 14:38:21 +08002217static void kvm_init_debug(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002218{
2219 struct kvm_stats_debugfs_item *p;
2220
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002221 kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002222 for (p = debugfs_entries; p->name; ++p)
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002223 p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir,
Avi Kivity1165f5f2007-04-19 17:27:43 +03002224 (void *)(long)p->offset,
Avi Kivityba1389b2007-11-18 16:24:12 +02002225 stat_fops[p->kind]);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002226}
2227
2228static void kvm_exit_debug(void)
2229{
2230 struct kvm_stats_debugfs_item *p;
2231
2232 for (p = debugfs_entries; p->name; ++p)
2233 debugfs_remove(p->dentry);
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002234 debugfs_remove(kvm_debugfs_dir);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002235}
2236
Avi Kivity59ae6c62007-02-12 00:54:48 -08002237static int kvm_suspend(struct sys_device *dev, pm_message_t state)
2238{
Avi Kivity4267c412007-05-24 13:09:41 +03002239 hardware_disable(NULL);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002240 return 0;
2241}
2242
2243static int kvm_resume(struct sys_device *dev)
2244{
Avi Kivity4267c412007-05-24 13:09:41 +03002245 hardware_enable(NULL);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002246 return 0;
2247}
2248
2249static struct sysdev_class kvm_sysdev_class = {
Kay Sieversaf5ca3f2007-12-20 02:09:39 +01002250 .name = "kvm",
Avi Kivity59ae6c62007-02-12 00:54:48 -08002251 .suspend = kvm_suspend,
2252 .resume = kvm_resume,
2253};
2254
2255static struct sys_device kvm_sysdev = {
2256 .id = 0,
2257 .cls = &kvm_sysdev_class,
2258};
2259
Izik Eiduscea7bb22007-10-17 19:17:48 +02002260struct page *bad_page;
Anthony Liguori35149e22008-04-02 14:46:56 -05002261pfn_t bad_pfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002262
Avi Kivity15ad7142007-07-11 18:17:21 +03002263static inline
2264struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn)
2265{
2266 return container_of(pn, struct kvm_vcpu, preempt_notifier);
2267}
2268
2269static void kvm_sched_in(struct preempt_notifier *pn, int cpu)
2270{
2271 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2272
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002273 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002274}
2275
2276static void kvm_sched_out(struct preempt_notifier *pn,
2277 struct task_struct *next)
2278{
2279 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2280
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002281 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002282}
2283
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002284int kvm_init(void *opaque, unsigned int vcpu_size,
Rusty Russellc16f8622007-07-30 21:12:19 +10002285 struct module *module)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002286{
2287 int r;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002288 int cpu;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002289
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002290 kvm_init_debug();
2291
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002292 r = kvm_arch_init(opaque);
2293 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002294 goto out_fail;
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002295
2296 bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2297
2298 if (bad_page == NULL) {
2299 r = -ENOMEM;
2300 goto out;
2301 }
2302
Anthony Liguori35149e22008-04-02 14:46:56 -05002303 bad_pfn = page_to_pfn(bad_page);
2304
Rusty Russell7f59f492008-12-07 21:25:45 +10302305 if (!alloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
2306 r = -ENOMEM;
2307 goto out_free_0;
2308 }
2309
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002310 r = kvm_arch_hardware_setup();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002311 if (r < 0)
Rusty Russell7f59f492008-12-07 21:25:45 +10302312 goto out_free_0a;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002313
Yang, Sheng002c7f72007-07-31 14:23:01 +03002314 for_each_online_cpu(cpu) {
2315 smp_call_function_single(cpu,
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002316 kvm_arch_check_processor_compat,
Jens Axboe8691e5a2008-06-06 11:18:06 +02002317 &r, 1);
Yang, Sheng002c7f72007-07-31 14:23:01 +03002318 if (r < 0)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002319 goto out_free_1;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002320 }
2321
Jens Axboe15c8b6c2008-05-09 09:39:44 +02002322 on_each_cpu(hardware_enable, NULL, 1);
Avi Kivity774c47f2007-02-12 00:54:47 -08002323 r = register_cpu_notifier(&kvm_cpu_notifier);
2324 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002325 goto out_free_2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002326 register_reboot_notifier(&kvm_reboot_notifier);
2327
Avi Kivity59ae6c62007-02-12 00:54:48 -08002328 r = sysdev_class_register(&kvm_sysdev_class);
2329 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002330 goto out_free_3;
Avi Kivity59ae6c62007-02-12 00:54:48 -08002331
2332 r = sysdev_register(&kvm_sysdev);
2333 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002334 goto out_free_4;
Avi Kivity59ae6c62007-02-12 00:54:48 -08002335
Rusty Russellc16f8622007-07-30 21:12:19 +10002336 /* A kmem cache lets us meet the alignment requirements of fx_save. */
2337 kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size,
Joe Perches56919c52007-11-12 20:06:51 -08002338 __alignof__(struct kvm_vcpu),
2339 0, NULL);
Rusty Russellc16f8622007-07-30 21:12:19 +10002340 if (!kvm_vcpu_cache) {
2341 r = -ENOMEM;
Zhang Xiantaod23087842007-11-29 15:35:39 +08002342 goto out_free_5;
Rusty Russellc16f8622007-07-30 21:12:19 +10002343 }
2344
Avi Kivity6aa8b732006-12-10 02:21:36 -08002345 kvm_chardev_ops.owner = module;
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01002346 kvm_vm_fops.owner = module;
2347 kvm_vcpu_fops.owner = module;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002348
2349 r = misc_register(&kvm_dev);
2350 if (r) {
Mike Dayd77c26f2007-10-08 09:02:08 -04002351 printk(KERN_ERR "kvm: misc device register failed\n");
Avi Kivity6aa8b732006-12-10 02:21:36 -08002352 goto out_free;
2353 }
2354
Avi Kivity15ad7142007-07-11 18:17:21 +03002355 kvm_preempt_ops.sched_in = kvm_sched_in;
2356 kvm_preempt_ops.sched_out = kvm_sched_out;
Sheng Yang5319c662008-11-24 14:32:57 +08002357#ifndef CONFIG_X86
2358 msi2intx = 0;
2359#endif
Avi Kivity15ad7142007-07-11 18:17:21 +03002360
Avi Kivityc7addb92007-09-16 18:58:32 +02002361 return 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002362
2363out_free:
Rusty Russellc16f8622007-07-30 21:12:19 +10002364 kmem_cache_destroy(kvm_vcpu_cache);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002365out_free_5:
Avi Kivity59ae6c62007-02-12 00:54:48 -08002366 sysdev_unregister(&kvm_sysdev);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002367out_free_4:
Avi Kivity59ae6c62007-02-12 00:54:48 -08002368 sysdev_class_unregister(&kvm_sysdev_class);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002369out_free_3:
Avi Kivity6aa8b732006-12-10 02:21:36 -08002370 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity774c47f2007-02-12 00:54:47 -08002371 unregister_cpu_notifier(&kvm_cpu_notifier);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002372out_free_2:
Jens Axboe15c8b6c2008-05-09 09:39:44 +02002373 on_each_cpu(hardware_disable, NULL, 1);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002374out_free_1:
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002375 kvm_arch_hardware_unsetup();
Rusty Russell7f59f492008-12-07 21:25:45 +10302376out_free_0a:
2377 free_cpumask_var(cpus_hardware_enabled);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002378out_free_0:
2379 __free_page(bad_page);
Avi Kivityca45aaa2007-03-01 19:21:03 +02002380out:
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002381 kvm_arch_exit();
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002382 kvm_exit_debug();
Zhang Xiantaod23087842007-11-29 15:35:39 +08002383out_fail:
Avi Kivity6aa8b732006-12-10 02:21:36 -08002384 return r;
2385}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002386EXPORT_SYMBOL_GPL(kvm_init);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002387
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002388void kvm_exit(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002389{
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04002390 kvm_trace_cleanup();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002391 misc_deregister(&kvm_dev);
Rusty Russellc16f8622007-07-30 21:12:19 +10002392 kmem_cache_destroy(kvm_vcpu_cache);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002393 sysdev_unregister(&kvm_sysdev);
2394 sysdev_class_unregister(&kvm_sysdev_class);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002395 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002396 unregister_cpu_notifier(&kvm_cpu_notifier);
Jens Axboe15c8b6c2008-05-09 09:39:44 +02002397 on_each_cpu(hardware_disable, NULL, 1);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002398 kvm_arch_hardware_unsetup();
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002399 kvm_arch_exit();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002400 kvm_exit_debug();
Rusty Russell7f59f492008-12-07 21:25:45 +10302401 free_cpumask_var(cpus_hardware_enabled);
Izik Eiduscea7bb22007-10-17 19:17:48 +02002402 __free_page(bad_page);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002403}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002404EXPORT_SYMBOL_GPL(kvm_exit);