]>
Commit | Line | Data |
---|---|---|
6aa8b732 AK |
1 | /* |
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 | ||
e2174021 | 18 | #include "iodev.h" |
6aa8b732 | 19 | |
edf88417 | 20 | #include <linux/kvm_host.h> |
6aa8b732 AK |
21 | #include <linux/kvm.h> |
22 | #include <linux/module.h> | |
23 | #include <linux/errno.h> | |
6aa8b732 AK |
24 | #include <linux/percpu.h> |
25 | #include <linux/gfp.h> | |
6aa8b732 AK |
26 | #include <linux/mm.h> |
27 | #include <linux/miscdevice.h> | |
28 | #include <linux/vmalloc.h> | |
6aa8b732 | 29 | #include <linux/reboot.h> |
6aa8b732 AK |
30 | #include <linux/debugfs.h> |
31 | #include <linux/highmem.h> | |
32 | #include <linux/file.h> | |
59ae6c6b | 33 | #include <linux/sysdev.h> |
774c47f1 | 34 | #include <linux/cpu.h> |
e8edc6e0 | 35 | #include <linux/sched.h> |
d9e368d6 AK |
36 | #include <linux/cpumask.h> |
37 | #include <linux/smp.h> | |
d6d28168 | 38 | #include <linux/anon_inodes.h> |
04d2cc77 | 39 | #include <linux/profile.h> |
7aa81cc0 | 40 | #include <linux/kvm_para.h> |
6fc138d2 | 41 | #include <linux/pagemap.h> |
8d4e1288 | 42 | #include <linux/mman.h> |
35149e21 | 43 | #include <linux/swap.h> |
6aa8b732 | 44 | |
e495606d | 45 | #include <asm/processor.h> |
e495606d AK |
46 | #include <asm/io.h> |
47 | #include <asm/uaccess.h> | |
3e021bf5 | 48 | #include <asm/pgtable.h> |
6aa8b732 | 49 | |
f64769eb SY |
50 | #ifdef CONFIG_X86 |
51 | #include <asm/msidef.h> | |
52 | #endif | |
53 | ||
5f94c174 LV |
54 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
55 | #include "coalesced_mmio.h" | |
56 | #endif | |
57 | ||
8a98f664 XZ |
58 | #ifdef KVM_CAP_DEVICE_ASSIGNMENT |
59 | #include <linux/pci.h> | |
60 | #include <linux/interrupt.h> | |
61 | #include "irq.h" | |
62 | #endif | |
63 | ||
6aa8b732 AK |
64 | MODULE_AUTHOR("Qumranet"); |
65 | MODULE_LICENSE("GPL"); | |
66 | ||
5319c662 SY |
67 | static int msi2intx = 1; |
68 | module_param(msi2intx, bool, 0); | |
69 | ||
e9b11c17 ZX |
70 | DEFINE_SPINLOCK(kvm_lock); |
71 | LIST_HEAD(vm_list); | |
133de902 | 72 | |
7f59f492 | 73 | static cpumask_var_t cpus_hardware_enabled; |
1b6c0168 | 74 | |
c16f862d RR |
75 | struct kmem_cache *kvm_vcpu_cache; |
76 | EXPORT_SYMBOL_GPL(kvm_vcpu_cache); | |
1165f5fe | 77 | |
15ad7146 AK |
78 | static __read_mostly struct preempt_ops kvm_preempt_ops; |
79 | ||
76f7c879 | 80 | struct dentry *kvm_debugfs_dir; |
6aa8b732 | 81 | |
bccf2150 AK |
82 | static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl, |
83 | unsigned long arg); | |
84 | ||
e8ba5d31 | 85 | static bool kvm_rebooting; |
4ecac3fd | 86 | |
8a98f664 | 87 | #ifdef KVM_CAP_DEVICE_ASSIGNMENT |
f64769eb SY |
88 | |
89 | #ifdef CONFIG_X86 | |
90 | static void assigned_device_msi_dispatch(struct kvm_assigned_dev_kernel *dev) | |
91 | { | |
92 | int vcpu_id; | |
93 | struct kvm_vcpu *vcpu; | |
94 | struct kvm_ioapic *ioapic = ioapic_irqchip(dev->kvm); | |
95 | int dest_id = (dev->guest_msi.address_lo & MSI_ADDR_DEST_ID_MASK) | |
96 | >> MSI_ADDR_DEST_ID_SHIFT; | |
97 | int vector = (dev->guest_msi.data & MSI_DATA_VECTOR_MASK) | |
98 | >> MSI_DATA_VECTOR_SHIFT; | |
99 | int dest_mode = test_bit(MSI_ADDR_DEST_MODE_SHIFT, | |
100 | (unsigned long *)&dev->guest_msi.address_lo); | |
101 | int trig_mode = test_bit(MSI_DATA_TRIGGER_SHIFT, | |
102 | (unsigned long *)&dev->guest_msi.data); | |
103 | int delivery_mode = test_bit(MSI_DATA_DELIVERY_MODE_SHIFT, | |
104 | (unsigned long *)&dev->guest_msi.data); | |
105 | u32 deliver_bitmask; | |
106 | ||
107 | BUG_ON(!ioapic); | |
108 | ||
109 | deliver_bitmask = kvm_ioapic_get_delivery_bitmask(ioapic, | |
110 | dest_id, dest_mode); | |
111 | /* IOAPIC delivery mode value is the same as MSI here */ | |
112 | switch (delivery_mode) { | |
113 | case IOAPIC_LOWEST_PRIORITY: | |
114 | vcpu = kvm_get_lowest_prio_vcpu(ioapic->kvm, vector, | |
115 | deliver_bitmask); | |
116 | if (vcpu != NULL) | |
117 | kvm_apic_set_irq(vcpu, vector, trig_mode); | |
118 | else | |
119 | printk(KERN_INFO "kvm: null lowest priority vcpu!\n"); | |
120 | break; | |
121 | case IOAPIC_FIXED: | |
122 | for (vcpu_id = 0; deliver_bitmask != 0; vcpu_id++) { | |
123 | if (!(deliver_bitmask & (1 << vcpu_id))) | |
124 | continue; | |
125 | deliver_bitmask &= ~(1 << vcpu_id); | |
126 | vcpu = ioapic->kvm->vcpus[vcpu_id]; | |
127 | if (vcpu) | |
128 | kvm_apic_set_irq(vcpu, vector, trig_mode); | |
129 | } | |
130 | break; | |
131 | default: | |
132 | printk(KERN_INFO "kvm: unsupported MSI delivery mode\n"); | |
133 | } | |
134 | } | |
135 | #else | |
136 | static void assigned_device_msi_dispatch(struct kvm_assigned_dev_kernel *dev) {} | |
137 | #endif | |
138 | ||
8a98f664 XZ |
139 | static struct kvm_assigned_dev_kernel *kvm_find_assigned_dev(struct list_head *head, |
140 | int assigned_dev_id) | |
141 | { | |
142 | struct list_head *ptr; | |
143 | struct kvm_assigned_dev_kernel *match; | |
144 | ||
145 | list_for_each(ptr, head) { | |
146 | match = list_entry(ptr, struct kvm_assigned_dev_kernel, list); | |
147 | if (match->assigned_dev_id == assigned_dev_id) | |
148 | return match; | |
149 | } | |
150 | return NULL; | |
151 | } | |
152 | ||
153 | static void kvm_assigned_dev_interrupt_work_handler(struct work_struct *work) | |
154 | { | |
155 | struct kvm_assigned_dev_kernel *assigned_dev; | |
156 | ||
157 | assigned_dev = container_of(work, struct kvm_assigned_dev_kernel, | |
158 | interrupt_work); | |
159 | ||
160 | /* This is taken to safely inject irq inside the guest. When | |
161 | * the interrupt injection (or the ioapic code) uses a | |
162 | * finer-grained lock, update this | |
163 | */ | |
164 | mutex_lock(&assigned_dev->kvm->lock); | |
6b9cc7fd SY |
165 | if (assigned_dev->irq_requested_type & KVM_ASSIGNED_DEV_GUEST_INTX) |
166 | kvm_set_irq(assigned_dev->kvm, | |
167 | assigned_dev->irq_source_id, | |
168 | assigned_dev->guest_irq, 1); | |
169 | else if (assigned_dev->irq_requested_type & | |
170 | KVM_ASSIGNED_DEV_GUEST_MSI) { | |
171 | assigned_device_msi_dispatch(assigned_dev); | |
172 | enable_irq(assigned_dev->host_irq); | |
defaf158 | 173 | assigned_dev->host_irq_disabled = false; |
6b9cc7fd | 174 | } |
8a98f664 XZ |
175 | mutex_unlock(&assigned_dev->kvm->lock); |
176 | kvm_put_kvm(assigned_dev->kvm); | |
177 | } | |
178 | ||
8a98f664 XZ |
179 | static irqreturn_t kvm_assigned_dev_intr(int irq, void *dev_id) |
180 | { | |
181 | struct kvm_assigned_dev_kernel *assigned_dev = | |
182 | (struct kvm_assigned_dev_kernel *) dev_id; | |
183 | ||
184 | kvm_get_kvm(assigned_dev->kvm); | |
defaf158 | 185 | |
8a98f664 | 186 | schedule_work(&assigned_dev->interrupt_work); |
defaf158 | 187 | |
8a98f664 | 188 | disable_irq_nosync(irq); |
defaf158 MM |
189 | assigned_dev->host_irq_disabled = true; |
190 | ||
8a98f664 XZ |
191 | return IRQ_HANDLED; |
192 | } | |
193 | ||
194 | /* Ack the irq line for an assigned device */ | |
195 | static void kvm_assigned_dev_ack_irq(struct kvm_irq_ack_notifier *kian) | |
196 | { | |
197 | struct kvm_assigned_dev_kernel *dev; | |
198 | ||
199 | if (kian->gsi == -1) | |
200 | return; | |
201 | ||
202 | dev = container_of(kian, struct kvm_assigned_dev_kernel, | |
203 | ack_notifier); | |
defaf158 | 204 | |
5550af4d | 205 | kvm_set_irq(dev->kvm, dev->irq_source_id, dev->guest_irq, 0); |
defaf158 MM |
206 | |
207 | /* The guest irq may be shared so this ack may be | |
208 | * from another device. | |
209 | */ | |
210 | if (dev->host_irq_disabled) { | |
211 | enable_irq(dev->host_irq); | |
212 | dev->host_irq_disabled = false; | |
213 | } | |
8a98f664 XZ |
214 | } |
215 | ||
4a643be8 MM |
216 | static void kvm_free_assigned_irq(struct kvm *kvm, |
217 | struct kvm_assigned_dev_kernel *assigned_dev) | |
8a98f664 | 218 | { |
4a643be8 MM |
219 | if (!irqchip_in_kernel(kvm)) |
220 | return; | |
8a98f664 | 221 | |
e19e30ef | 222 | kvm_unregister_irq_ack_notifier(&assigned_dev->ack_notifier); |
f29b2673 MM |
223 | |
224 | if (assigned_dev->irq_source_id != -1) | |
225 | kvm_free_irq_source_id(kvm, assigned_dev->irq_source_id); | |
226 | assigned_dev->irq_source_id = -1; | |
8a98f664 | 227 | |
4a643be8 MM |
228 | if (!assigned_dev->irq_requested_type) |
229 | return; | |
230 | ||
8a98f664 XZ |
231 | if (cancel_work_sync(&assigned_dev->interrupt_work)) |
232 | /* We had pending work. That means we will have to take | |
233 | * care of kvm_put_kvm. | |
234 | */ | |
235 | kvm_put_kvm(kvm); | |
236 | ||
4a643be8 MM |
237 | free_irq(assigned_dev->host_irq, (void *)assigned_dev); |
238 | ||
239 | if (assigned_dev->irq_requested_type & KVM_ASSIGNED_DEV_HOST_MSI) | |
240 | pci_disable_msi(assigned_dev->dev); | |
241 | ||
242 | assigned_dev->irq_requested_type = 0; | |
243 | } | |
244 | ||
245 | ||
246 | static void kvm_free_assigned_device(struct kvm *kvm, | |
247 | struct kvm_assigned_dev_kernel | |
248 | *assigned_dev) | |
249 | { | |
250 | kvm_free_assigned_irq(kvm, assigned_dev); | |
251 | ||
6eb55818 SY |
252 | pci_reset_function(assigned_dev->dev); |
253 | ||
8a98f664 XZ |
254 | pci_release_regions(assigned_dev->dev); |
255 | pci_disable_device(assigned_dev->dev); | |
256 | pci_dev_put(assigned_dev->dev); | |
257 | ||
258 | list_del(&assigned_dev->list); | |
259 | kfree(assigned_dev); | |
260 | } | |
261 | ||
262 | void kvm_free_all_assigned_devices(struct kvm *kvm) | |
263 | { | |
264 | struct list_head *ptr, *ptr2; | |
265 | struct kvm_assigned_dev_kernel *assigned_dev; | |
266 | ||
267 | list_for_each_safe(ptr, ptr2, &kvm->arch.assigned_dev_head) { | |
268 | assigned_dev = list_entry(ptr, | |
269 | struct kvm_assigned_dev_kernel, | |
270 | list); | |
271 | ||
272 | kvm_free_assigned_device(kvm, assigned_dev); | |
273 | } | |
274 | } | |
275 | ||
00e3ed39 SY |
276 | static int assigned_device_update_intx(struct kvm *kvm, |
277 | struct kvm_assigned_dev_kernel *adev, | |
278 | struct kvm_assigned_irq *airq) | |
279 | { | |
fbac7818 SY |
280 | adev->guest_irq = airq->guest_irq; |
281 | adev->ack_notifier.gsi = airq->guest_irq; | |
282 | ||
283 | if (adev->irq_requested_type & KVM_ASSIGNED_DEV_HOST_INTX) | |
00e3ed39 | 284 | return 0; |
00e3ed39 SY |
285 | |
286 | if (irqchip_in_kernel(kvm)) { | |
5319c662 SY |
287 | if (!msi2intx && |
288 | adev->irq_requested_type & KVM_ASSIGNED_DEV_HOST_MSI) { | |
6b9cc7fd SY |
289 | free_irq(adev->host_irq, (void *)kvm); |
290 | pci_disable_msi(adev->dev); | |
291 | } | |
292 | ||
00e3ed39 SY |
293 | if (!capable(CAP_SYS_RAWIO)) |
294 | return -EPERM; | |
295 | ||
296 | if (airq->host_irq) | |
297 | adev->host_irq = airq->host_irq; | |
298 | else | |
299 | adev->host_irq = adev->dev->irq; | |
00e3ed39 SY |
300 | |
301 | /* Even though this is PCI, we don't want to use shared | |
302 | * interrupts. Sharing host devices with guest-assigned devices | |
303 | * on the same interrupt line is not a happy situation: there | |
304 | * are going to be long delays in accepting, acking, etc. | |
305 | */ | |
306 | if (request_irq(adev->host_irq, kvm_assigned_dev_intr, | |
307 | 0, "kvm_assigned_intx_device", (void *)adev)) | |
308 | return -EIO; | |
309 | } | |
310 | ||
4f906c19 SY |
311 | adev->irq_requested_type = KVM_ASSIGNED_DEV_GUEST_INTX | |
312 | KVM_ASSIGNED_DEV_HOST_INTX; | |
00e3ed39 SY |
313 | return 0; |
314 | } | |
315 | ||
6b9cc7fd SY |
316 | #ifdef CONFIG_X86 |
317 | static int assigned_device_update_msi(struct kvm *kvm, | |
318 | struct kvm_assigned_dev_kernel *adev, | |
319 | struct kvm_assigned_irq *airq) | |
320 | { | |
321 | int r; | |
322 | ||
5319c662 SY |
323 | if (airq->flags & KVM_DEV_IRQ_ASSIGN_ENABLE_MSI) { |
324 | /* x86 don't care upper address of guest msi message addr */ | |
325 | adev->irq_requested_type |= KVM_ASSIGNED_DEV_GUEST_MSI; | |
326 | adev->irq_requested_type &= ~KVM_ASSIGNED_DEV_GUEST_INTX; | |
327 | adev->guest_msi.address_lo = airq->guest_msi.addr_lo; | |
328 | adev->guest_msi.data = airq->guest_msi.data; | |
329 | adev->ack_notifier.gsi = -1; | |
330 | } else if (msi2intx) { | |
331 | adev->irq_requested_type |= KVM_ASSIGNED_DEV_GUEST_INTX; | |
332 | adev->irq_requested_type &= ~KVM_ASSIGNED_DEV_GUEST_MSI; | |
333 | adev->guest_irq = airq->guest_irq; | |
334 | adev->ack_notifier.gsi = airq->guest_irq; | |
335 | } | |
6b9cc7fd SY |
336 | |
337 | if (adev->irq_requested_type & KVM_ASSIGNED_DEV_HOST_MSI) | |
338 | return 0; | |
339 | ||
340 | if (irqchip_in_kernel(kvm)) { | |
5319c662 SY |
341 | if (!msi2intx) { |
342 | if (adev->irq_requested_type & | |
343 | KVM_ASSIGNED_DEV_HOST_INTX) | |
344 | free_irq(adev->host_irq, (void *)adev); | |
345 | ||
346 | r = pci_enable_msi(adev->dev); | |
347 | if (r) | |
348 | return r; | |
349 | } | |
6b9cc7fd SY |
350 | |
351 | adev->host_irq = adev->dev->irq; | |
352 | if (request_irq(adev->host_irq, kvm_assigned_dev_intr, 0, | |
353 | "kvm_assigned_msi_device", (void *)adev)) | |
354 | return -EIO; | |
355 | } | |
356 | ||
5319c662 SY |
357 | if (!msi2intx) |
358 | adev->irq_requested_type = KVM_ASSIGNED_DEV_GUEST_MSI; | |
359 | ||
360 | adev->irq_requested_type |= KVM_ASSIGNED_DEV_HOST_MSI; | |
6b9cc7fd SY |
361 | return 0; |
362 | } | |
363 | #endif | |
364 | ||
8a98f664 XZ |
365 | static int kvm_vm_ioctl_assign_irq(struct kvm *kvm, |
366 | struct kvm_assigned_irq | |
367 | *assigned_irq) | |
368 | { | |
369 | int r = 0; | |
370 | struct kvm_assigned_dev_kernel *match; | |
371 | ||
372 | mutex_lock(&kvm->lock); | |
373 | ||
374 | match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, | |
375 | assigned_irq->assigned_dev_id); | |
376 | if (!match) { | |
377 | mutex_unlock(&kvm->lock); | |
378 | return -EINVAL; | |
379 | } | |
380 | ||
4f906c19 | 381 | if (!match->irq_requested_type) { |
342ffb93 SY |
382 | INIT_WORK(&match->interrupt_work, |
383 | kvm_assigned_dev_interrupt_work_handler); | |
384 | if (irqchip_in_kernel(kvm)) { | |
385 | /* Register ack nofitier */ | |
386 | match->ack_notifier.gsi = -1; | |
387 | match->ack_notifier.irq_acked = | |
388 | kvm_assigned_dev_ack_irq; | |
389 | kvm_register_irq_ack_notifier(kvm, | |
390 | &match->ack_notifier); | |
391 | ||
392 | /* Request IRQ source ID */ | |
393 | r = kvm_request_irq_source_id(kvm); | |
394 | if (r < 0) | |
395 | goto out_release; | |
396 | else | |
397 | match->irq_source_id = r; | |
5319c662 SY |
398 | |
399 | #ifdef CONFIG_X86 | |
400 | /* Determine host device irq type, we can know the | |
401 | * result from dev->msi_enabled */ | |
402 | if (msi2intx) | |
403 | pci_enable_msi(match->dev); | |
404 | #endif | |
342ffb93 | 405 | } |
8a98f664 XZ |
406 | } |
407 | ||
5319c662 SY |
408 | if ((!msi2intx && |
409 | (assigned_irq->flags & KVM_DEV_IRQ_ASSIGN_ENABLE_MSI)) || | |
410 | (msi2intx && match->dev->msi_enabled)) { | |
6b9cc7fd SY |
411 | #ifdef CONFIG_X86 |
412 | r = assigned_device_update_msi(kvm, match, assigned_irq); | |
413 | if (r) { | |
414 | printk(KERN_WARNING "kvm: failed to enable " | |
415 | "MSI device!\n"); | |
416 | goto out_release; | |
417 | } | |
418 | #else | |
419 | r = -ENOTTY; | |
420 | #endif | |
421 | } else if (assigned_irq->host_irq == 0 && match->dev->irq == 0) { | |
422 | /* Host device IRQ 0 means don't support INTx */ | |
5319c662 SY |
423 | if (!msi2intx) { |
424 | printk(KERN_WARNING | |
425 | "kvm: wait device to enable MSI!\n"); | |
426 | r = 0; | |
427 | } else { | |
428 | printk(KERN_WARNING | |
429 | "kvm: failed to enable MSI device!\n"); | |
430 | r = -ENOTTY; | |
431 | goto out_release; | |
432 | } | |
6b9cc7fd SY |
433 | } else { |
434 | /* Non-sharing INTx mode */ | |
435 | r = assigned_device_update_intx(kvm, match, assigned_irq); | |
436 | if (r) { | |
437 | printk(KERN_WARNING "kvm: failed to enable " | |
438 | "INTx device!\n"); | |
439 | goto out_release; | |
440 | } | |
441 | } | |
8a98f664 | 442 | |
8a98f664 XZ |
443 | mutex_unlock(&kvm->lock); |
444 | return r; | |
445 | out_release: | |
446 | mutex_unlock(&kvm->lock); | |
447 | kvm_free_assigned_device(kvm, match); | |
448 | return r; | |
449 | } | |
450 | ||
451 | static int kvm_vm_ioctl_assign_device(struct kvm *kvm, | |
452 | struct kvm_assigned_pci_dev *assigned_dev) | |
453 | { | |
454 | int r = 0; | |
455 | struct kvm_assigned_dev_kernel *match; | |
456 | struct pci_dev *dev; | |
457 | ||
458 | mutex_lock(&kvm->lock); | |
459 | ||
460 | match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, | |
461 | assigned_dev->assigned_dev_id); | |
462 | if (match) { | |
463 | /* device already assigned */ | |
464 | r = -EINVAL; | |
465 | goto out; | |
466 | } | |
467 | ||
468 | match = kzalloc(sizeof(struct kvm_assigned_dev_kernel), GFP_KERNEL); | |
469 | if (match == NULL) { | |
470 | printk(KERN_INFO "%s: Couldn't allocate memory\n", | |
471 | __func__); | |
472 | r = -ENOMEM; | |
473 | goto out; | |
474 | } | |
475 | dev = pci_get_bus_and_slot(assigned_dev->busnr, | |
476 | assigned_dev->devfn); | |
477 | if (!dev) { | |
478 | printk(KERN_INFO "%s: host device not found\n", __func__); | |
479 | r = -EINVAL; | |
480 | goto out_free; | |
481 | } | |
482 | if (pci_enable_device(dev)) { | |
483 | printk(KERN_INFO "%s: Could not enable PCI device\n", __func__); | |
484 | r = -EBUSY; | |
485 | goto out_put; | |
486 | } | |
487 | r = pci_request_regions(dev, "kvm_assigned_device"); | |
488 | if (r) { | |
489 | printk(KERN_INFO "%s: Could not get access to device regions\n", | |
490 | __func__); | |
491 | goto out_disable; | |
492 | } | |
6eb55818 SY |
493 | |
494 | pci_reset_function(dev); | |
495 | ||
8a98f664 XZ |
496 | match->assigned_dev_id = assigned_dev->assigned_dev_id; |
497 | match->host_busnr = assigned_dev->busnr; | |
498 | match->host_devfn = assigned_dev->devfn; | |
b653574a | 499 | match->flags = assigned_dev->flags; |
8a98f664 | 500 | match->dev = dev; |
f29b2673 | 501 | match->irq_source_id = -1; |
8a98f664 XZ |
502 | match->kvm = kvm; |
503 | ||
504 | list_add(&match->list, &kvm->arch.assigned_dev_head); | |
505 | ||
506 | if (assigned_dev->flags & KVM_DEV_ASSIGN_ENABLE_IOMMU) { | |
19de40a8 | 507 | if (!kvm->arch.iommu_domain) { |
260782bc WH |
508 | r = kvm_iommu_map_guest(kvm); |
509 | if (r) | |
510 | goto out_list_del; | |
511 | } | |
512 | r = kvm_assign_device(kvm, match); | |
8a98f664 XZ |
513 | if (r) |
514 | goto out_list_del; | |
515 | } | |
516 | ||
517 | out: | |
518 | mutex_unlock(&kvm->lock); | |
519 | return r; | |
520 | out_list_del: | |
521 | list_del(&match->list); | |
522 | pci_release_regions(dev); | |
523 | out_disable: | |
524 | pci_disable_device(dev); | |
525 | out_put: | |
526 | pci_dev_put(dev); | |
527 | out_free: | |
528 | kfree(match); | |
529 | mutex_unlock(&kvm->lock); | |
530 | return r; | |
531 | } | |
532 | #endif | |
533 | ||
0a920356 WH |
534 | #ifdef KVM_CAP_DEVICE_DEASSIGNMENT |
535 | static int kvm_vm_ioctl_deassign_device(struct kvm *kvm, | |
536 | struct kvm_assigned_pci_dev *assigned_dev) | |
537 | { | |
538 | int r = 0; | |
539 | struct kvm_assigned_dev_kernel *match; | |
540 | ||
541 | mutex_lock(&kvm->lock); | |
542 | ||
543 | match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, | |
544 | assigned_dev->assigned_dev_id); | |
545 | if (!match) { | |
546 | printk(KERN_INFO "%s: device hasn't been assigned before, " | |
547 | "so cannot be deassigned\n", __func__); | |
548 | r = -EINVAL; | |
549 | goto out; | |
550 | } | |
551 | ||
552 | if (assigned_dev->flags & KVM_DEV_ASSIGN_ENABLE_IOMMU) | |
553 | kvm_deassign_device(kvm, match); | |
554 | ||
555 | kvm_free_assigned_device(kvm, match); | |
556 | ||
557 | out: | |
558 | mutex_unlock(&kvm->lock); | |
559 | return r; | |
560 | } | |
561 | #endif | |
562 | ||
5aacf0ca JM |
563 | static inline int valid_vcpu(int n) |
564 | { | |
565 | return likely(n >= 0 && n < KVM_MAX_VCPUS); | |
566 | } | |
567 | ||
c77fb9dc | 568 | inline int kvm_is_mmio_pfn(pfn_t pfn) |
cbff90a7 BAY |
569 | { |
570 | if (pfn_valid(pfn)) | |
571 | return PageReserved(pfn_to_page(pfn)); | |
572 | ||
573 | return true; | |
574 | } | |
575 | ||
bccf2150 AK |
576 | /* |
577 | * Switches to specified vcpu, until a matching vcpu_put() | |
578 | */ | |
313a3dc7 | 579 | void vcpu_load(struct kvm_vcpu *vcpu) |
6aa8b732 | 580 | { |
15ad7146 AK |
581 | int cpu; |
582 | ||
bccf2150 | 583 | mutex_lock(&vcpu->mutex); |
15ad7146 AK |
584 | cpu = get_cpu(); |
585 | preempt_notifier_register(&vcpu->preempt_notifier); | |
313a3dc7 | 586 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 | 587 | put_cpu(); |
6aa8b732 AK |
588 | } |
589 | ||
313a3dc7 | 590 | void vcpu_put(struct kvm_vcpu *vcpu) |
6aa8b732 | 591 | { |
15ad7146 | 592 | preempt_disable(); |
313a3dc7 | 593 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
594 | preempt_notifier_unregister(&vcpu->preempt_notifier); |
595 | preempt_enable(); | |
6aa8b732 AK |
596 | mutex_unlock(&vcpu->mutex); |
597 | } | |
598 | ||
d9e368d6 AK |
599 | static void ack_flush(void *_completed) |
600 | { | |
d9e368d6 AK |
601 | } |
602 | ||
49846896 | 603 | static bool make_all_cpus_request(struct kvm *kvm, unsigned int req) |
d9e368d6 | 604 | { |
597a5f55 | 605 | int i, cpu, me; |
6ef7a1bc RR |
606 | cpumask_var_t cpus; |
607 | bool called = true; | |
d9e368d6 | 608 | struct kvm_vcpu *vcpu; |
d9e368d6 | 609 | |
6ef7a1bc RR |
610 | if (alloc_cpumask_var(&cpus, GFP_ATOMIC)) |
611 | cpumask_clear(cpus); | |
612 | ||
597a5f55 | 613 | me = get_cpu(); |
fb3f0f51 RR |
614 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { |
615 | vcpu = kvm->vcpus[i]; | |
616 | if (!vcpu) | |
617 | continue; | |
49846896 | 618 | if (test_and_set_bit(req, &vcpu->requests)) |
d9e368d6 AK |
619 | continue; |
620 | cpu = vcpu->cpu; | |
6ef7a1bc RR |
621 | if (cpus != NULL && cpu != -1 && cpu != me) |
622 | cpumask_set_cpu(cpu, cpus); | |
49846896 | 623 | } |
6ef7a1bc RR |
624 | if (unlikely(cpus == NULL)) |
625 | smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1); | |
626 | else if (!cpumask_empty(cpus)) | |
627 | smp_call_function_many(cpus, ack_flush, NULL, 1); | |
628 | else | |
629 | called = false; | |
597a5f55 | 630 | put_cpu(); |
6ef7a1bc | 631 | free_cpumask_var(cpus); |
49846896 | 632 | return called; |
d9e368d6 AK |
633 | } |
634 | ||
49846896 | 635 | void kvm_flush_remote_tlbs(struct kvm *kvm) |
2e53d63a | 636 | { |
49846896 RR |
637 | if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH)) |
638 | ++kvm->stat.remote_tlb_flush; | |
2e53d63a MT |
639 | } |
640 | ||
49846896 RR |
641 | void kvm_reload_remote_mmus(struct kvm *kvm) |
642 | { | |
643 | make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD); | |
644 | } | |
2e53d63a | 645 | |
fb3f0f51 RR |
646 | int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id) |
647 | { | |
648 | struct page *page; | |
649 | int r; | |
650 | ||
651 | mutex_init(&vcpu->mutex); | |
652 | vcpu->cpu = -1; | |
fb3f0f51 RR |
653 | vcpu->kvm = kvm; |
654 | vcpu->vcpu_id = id; | |
b6958ce4 | 655 | init_waitqueue_head(&vcpu->wq); |
fb3f0f51 RR |
656 | |
657 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
658 | if (!page) { | |
659 | r = -ENOMEM; | |
660 | goto fail; | |
661 | } | |
662 | vcpu->run = page_address(page); | |
663 | ||
e9b11c17 | 664 | r = kvm_arch_vcpu_init(vcpu); |
fb3f0f51 | 665 | if (r < 0) |
e9b11c17 | 666 | goto fail_free_run; |
fb3f0f51 RR |
667 | return 0; |
668 | ||
fb3f0f51 RR |
669 | fail_free_run: |
670 | free_page((unsigned long)vcpu->run); | |
671 | fail: | |
76fafa5e | 672 | return r; |
fb3f0f51 RR |
673 | } |
674 | EXPORT_SYMBOL_GPL(kvm_vcpu_init); | |
675 | ||
676 | void kvm_vcpu_uninit(struct kvm_vcpu *vcpu) | |
677 | { | |
e9b11c17 | 678 | kvm_arch_vcpu_uninit(vcpu); |
fb3f0f51 RR |
679 | free_page((unsigned long)vcpu->run); |
680 | } | |
681 | EXPORT_SYMBOL_GPL(kvm_vcpu_uninit); | |
682 | ||
e930bffe AA |
683 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
684 | static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn) | |
685 | { | |
686 | return container_of(mn, struct kvm, mmu_notifier); | |
687 | } | |
688 | ||
689 | static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn, | |
690 | struct mm_struct *mm, | |
691 | unsigned long address) | |
692 | { | |
693 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
694 | int need_tlb_flush; | |
695 | ||
696 | /* | |
697 | * When ->invalidate_page runs, the linux pte has been zapped | |
698 | * already but the page is still allocated until | |
699 | * ->invalidate_page returns. So if we increase the sequence | |
700 | * here the kvm page fault will notice if the spte can't be | |
701 | * established because the page is going to be freed. If | |
702 | * instead the kvm page fault establishes the spte before | |
703 | * ->invalidate_page runs, kvm_unmap_hva will release it | |
704 | * before returning. | |
705 | * | |
706 | * The sequence increase only need to be seen at spin_unlock | |
707 | * time, and not at spin_lock time. | |
708 | * | |
709 | * Increasing the sequence after the spin_unlock would be | |
710 | * unsafe because the kvm page fault could then establish the | |
711 | * pte after kvm_unmap_hva returned, without noticing the page | |
712 | * is going to be freed. | |
713 | */ | |
714 | spin_lock(&kvm->mmu_lock); | |
715 | kvm->mmu_notifier_seq++; | |
716 | need_tlb_flush = kvm_unmap_hva(kvm, address); | |
717 | spin_unlock(&kvm->mmu_lock); | |
718 | ||
719 | /* we've to flush the tlb before the pages can be freed */ | |
720 | if (need_tlb_flush) | |
721 | kvm_flush_remote_tlbs(kvm); | |
722 | ||
723 | } | |
724 | ||
725 | static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn, | |
726 | struct mm_struct *mm, | |
727 | unsigned long start, | |
728 | unsigned long end) | |
729 | { | |
730 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
731 | int need_tlb_flush = 0; | |
732 | ||
733 | spin_lock(&kvm->mmu_lock); | |
734 | /* | |
735 | * The count increase must become visible at unlock time as no | |
736 | * spte can be established without taking the mmu_lock and | |
737 | * count is also read inside the mmu_lock critical section. | |
738 | */ | |
739 | kvm->mmu_notifier_count++; | |
740 | for (; start < end; start += PAGE_SIZE) | |
741 | need_tlb_flush |= kvm_unmap_hva(kvm, start); | |
742 | spin_unlock(&kvm->mmu_lock); | |
743 | ||
744 | /* we've to flush the tlb before the pages can be freed */ | |
745 | if (need_tlb_flush) | |
746 | kvm_flush_remote_tlbs(kvm); | |
747 | } | |
748 | ||
749 | static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn, | |
750 | struct mm_struct *mm, | |
751 | unsigned long start, | |
752 | unsigned long end) | |
753 | { | |
754 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
755 | ||
756 | spin_lock(&kvm->mmu_lock); | |
757 | /* | |
758 | * This sequence increase will notify the kvm page fault that | |
759 | * the page that is going to be mapped in the spte could have | |
760 | * been freed. | |
761 | */ | |
762 | kvm->mmu_notifier_seq++; | |
763 | /* | |
764 | * The above sequence increase must be visible before the | |
765 | * below count decrease but both values are read by the kvm | |
766 | * page fault under mmu_lock spinlock so we don't need to add | |
767 | * a smb_wmb() here in between the two. | |
768 | */ | |
769 | kvm->mmu_notifier_count--; | |
770 | spin_unlock(&kvm->mmu_lock); | |
771 | ||
772 | BUG_ON(kvm->mmu_notifier_count < 0); | |
773 | } | |
774 | ||
775 | static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn, | |
776 | struct mm_struct *mm, | |
777 | unsigned long address) | |
778 | { | |
779 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
780 | int young; | |
781 | ||
782 | spin_lock(&kvm->mmu_lock); | |
783 | young = kvm_age_hva(kvm, address); | |
784 | spin_unlock(&kvm->mmu_lock); | |
785 | ||
786 | if (young) | |
787 | kvm_flush_remote_tlbs(kvm); | |
788 | ||
789 | return young; | |
790 | } | |
791 | ||
85db06e5 MT |
792 | static void kvm_mmu_notifier_release(struct mmu_notifier *mn, |
793 | struct mm_struct *mm) | |
794 | { | |
795 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
796 | kvm_arch_flush_shadow(kvm); | |
797 | } | |
798 | ||
e930bffe AA |
799 | static const struct mmu_notifier_ops kvm_mmu_notifier_ops = { |
800 | .invalidate_page = kvm_mmu_notifier_invalidate_page, | |
801 | .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start, | |
802 | .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end, | |
803 | .clear_flush_young = kvm_mmu_notifier_clear_flush_young, | |
85db06e5 | 804 | .release = kvm_mmu_notifier_release, |
e930bffe AA |
805 | }; |
806 | #endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */ | |
807 | ||
f17abe9a | 808 | static struct kvm *kvm_create_vm(void) |
6aa8b732 | 809 | { |
d19a9cd2 | 810 | struct kvm *kvm = kvm_arch_create_vm(); |
5f94c174 LV |
811 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
812 | struct page *page; | |
813 | #endif | |
6aa8b732 | 814 | |
d19a9cd2 ZX |
815 | if (IS_ERR(kvm)) |
816 | goto out; | |
6aa8b732 | 817 | |
5f94c174 LV |
818 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
819 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
820 | if (!page) { | |
821 | kfree(kvm); | |
822 | return ERR_PTR(-ENOMEM); | |
823 | } | |
824 | kvm->coalesced_mmio_ring = | |
825 | (struct kvm_coalesced_mmio_ring *)page_address(page); | |
826 | #endif | |
827 | ||
e930bffe AA |
828 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
829 | { | |
830 | int err; | |
831 | kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops; | |
832 | err = mmu_notifier_register(&kvm->mmu_notifier, current->mm); | |
833 | if (err) { | |
834 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
835 | put_page(page); | |
836 | #endif | |
837 | kfree(kvm); | |
838 | return ERR_PTR(err); | |
839 | } | |
840 | } | |
841 | #endif | |
842 | ||
6d4e4c4f AK |
843 | kvm->mm = current->mm; |
844 | atomic_inc(&kvm->mm->mm_count); | |
aaee2c94 | 845 | spin_lock_init(&kvm->mmu_lock); |
74906345 | 846 | kvm_io_bus_init(&kvm->pio_bus); |
11ec2804 | 847 | mutex_init(&kvm->lock); |
2eeb2e94 | 848 | kvm_io_bus_init(&kvm->mmio_bus); |
72dc67a6 | 849 | init_rwsem(&kvm->slots_lock); |
d39f13b0 | 850 | atomic_set(&kvm->users_count, 1); |
5e58cfe4 RR |
851 | spin_lock(&kvm_lock); |
852 | list_add(&kvm->vm_list, &vm_list); | |
853 | spin_unlock(&kvm_lock); | |
5f94c174 LV |
854 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
855 | kvm_coalesced_mmio_init(kvm); | |
856 | #endif | |
d19a9cd2 | 857 | out: |
f17abe9a AK |
858 | return kvm; |
859 | } | |
860 | ||
6aa8b732 AK |
861 | /* |
862 | * Free any memory in @free but not in @dont. | |
863 | */ | |
864 | static void kvm_free_physmem_slot(struct kvm_memory_slot *free, | |
865 | struct kvm_memory_slot *dont) | |
866 | { | |
290fc38d IE |
867 | if (!dont || free->rmap != dont->rmap) |
868 | vfree(free->rmap); | |
6aa8b732 AK |
869 | |
870 | if (!dont || free->dirty_bitmap != dont->dirty_bitmap) | |
871 | vfree(free->dirty_bitmap); | |
872 | ||
05da4558 MT |
873 | if (!dont || free->lpage_info != dont->lpage_info) |
874 | vfree(free->lpage_info); | |
875 | ||
6aa8b732 | 876 | free->npages = 0; |
8b6d44c7 | 877 | free->dirty_bitmap = NULL; |
8d4e1288 | 878 | free->rmap = NULL; |
05da4558 | 879 | free->lpage_info = NULL; |
6aa8b732 AK |
880 | } |
881 | ||
d19a9cd2 | 882 | void kvm_free_physmem(struct kvm *kvm) |
6aa8b732 AK |
883 | { |
884 | int i; | |
885 | ||
886 | for (i = 0; i < kvm->nmemslots; ++i) | |
8b6d44c7 | 887 | kvm_free_physmem_slot(&kvm->memslots[i], NULL); |
6aa8b732 AK |
888 | } |
889 | ||
f17abe9a AK |
890 | static void kvm_destroy_vm(struct kvm *kvm) |
891 | { | |
6d4e4c4f AK |
892 | struct mm_struct *mm = kvm->mm; |
893 | ||
133de902 AK |
894 | spin_lock(&kvm_lock); |
895 | list_del(&kvm->vm_list); | |
896 | spin_unlock(&kvm_lock); | |
74906345 | 897 | kvm_io_bus_destroy(&kvm->pio_bus); |
2eeb2e94 | 898 | kvm_io_bus_destroy(&kvm->mmio_bus); |
5f94c174 LV |
899 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
900 | if (kvm->coalesced_mmio_ring != NULL) | |
901 | free_page((unsigned long)kvm->coalesced_mmio_ring); | |
e930bffe AA |
902 | #endif |
903 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) | |
904 | mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm); | |
5f94c174 | 905 | #endif |
d19a9cd2 | 906 | kvm_arch_destroy_vm(kvm); |
6d4e4c4f | 907 | mmdrop(mm); |
f17abe9a AK |
908 | } |
909 | ||
d39f13b0 IE |
910 | void kvm_get_kvm(struct kvm *kvm) |
911 | { | |
912 | atomic_inc(&kvm->users_count); | |
913 | } | |
914 | EXPORT_SYMBOL_GPL(kvm_get_kvm); | |
915 | ||
916 | void kvm_put_kvm(struct kvm *kvm) | |
917 | { | |
918 | if (atomic_dec_and_test(&kvm->users_count)) | |
919 | kvm_destroy_vm(kvm); | |
920 | } | |
921 | EXPORT_SYMBOL_GPL(kvm_put_kvm); | |
922 | ||
923 | ||
f17abe9a AK |
924 | static int kvm_vm_release(struct inode *inode, struct file *filp) |
925 | { | |
926 | struct kvm *kvm = filp->private_data; | |
927 | ||
d39f13b0 | 928 | kvm_put_kvm(kvm); |
6aa8b732 AK |
929 | return 0; |
930 | } | |
931 | ||
6aa8b732 AK |
932 | /* |
933 | * Allocate some memory and give it an address in the guest physical address | |
934 | * space. | |
935 | * | |
936 | * Discontiguous memory is allowed, mostly for framebuffers. | |
f78e0e2e | 937 | * |
10589a46 | 938 | * Must be called holding mmap_sem for write. |
6aa8b732 | 939 | */ |
f78e0e2e SY |
940 | int __kvm_set_memory_region(struct kvm *kvm, |
941 | struct kvm_userspace_memory_region *mem, | |
942 | int user_alloc) | |
6aa8b732 AK |
943 | { |
944 | int r; | |
945 | gfn_t base_gfn; | |
946 | unsigned long npages; | |
947 | unsigned long i; | |
948 | struct kvm_memory_slot *memslot; | |
949 | struct kvm_memory_slot old, new; | |
6aa8b732 AK |
950 | |
951 | r = -EINVAL; | |
952 | /* General sanity checks */ | |
953 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
954 | goto out; | |
955 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
956 | goto out; | |
e7cacd40 | 957 | if (user_alloc && (mem->userspace_addr & (PAGE_SIZE - 1))) |
78749809 | 958 | goto out; |
e0d62c7f | 959 | if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS) |
6aa8b732 AK |
960 | goto out; |
961 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
962 | goto out; | |
963 | ||
964 | memslot = &kvm->memslots[mem->slot]; | |
965 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; | |
966 | npages = mem->memory_size >> PAGE_SHIFT; | |
967 | ||
968 | if (!npages) | |
969 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
970 | ||
6aa8b732 AK |
971 | new = old = *memslot; |
972 | ||
973 | new.base_gfn = base_gfn; | |
974 | new.npages = npages; | |
975 | new.flags = mem->flags; | |
976 | ||
977 | /* Disallow changing a memory slot's size. */ | |
978 | r = -EINVAL; | |
979 | if (npages && old.npages && npages != old.npages) | |
f78e0e2e | 980 | goto out_free; |
6aa8b732 AK |
981 | |
982 | /* Check for overlaps */ | |
983 | r = -EEXIST; | |
984 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
985 | struct kvm_memory_slot *s = &kvm->memslots[i]; | |
986 | ||
987 | if (s == memslot) | |
988 | continue; | |
989 | if (!((base_gfn + npages <= s->base_gfn) || | |
990 | (base_gfn >= s->base_gfn + s->npages))) | |
f78e0e2e | 991 | goto out_free; |
6aa8b732 | 992 | } |
6aa8b732 | 993 | |
6aa8b732 AK |
994 | /* Free page dirty bitmap if unneeded */ |
995 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 996 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
997 | |
998 | r = -ENOMEM; | |
999 | ||
1000 | /* Allocate if a slot is being created */ | |
eff0114a | 1001 | #ifndef CONFIG_S390 |
8d4e1288 | 1002 | if (npages && !new.rmap) { |
d77c26fc | 1003 | new.rmap = vmalloc(npages * sizeof(struct page *)); |
290fc38d IE |
1004 | |
1005 | if (!new.rmap) | |
f78e0e2e | 1006 | goto out_free; |
290fc38d | 1007 | |
290fc38d | 1008 | memset(new.rmap, 0, npages * sizeof(*new.rmap)); |
8d4e1288 | 1009 | |
80b14b5b | 1010 | new.user_alloc = user_alloc; |
604b38ac AA |
1011 | /* |
1012 | * hva_to_rmmap() serialzies with the mmu_lock and to be | |
1013 | * safe it has to ignore memslots with !user_alloc && | |
1014 | * !userspace_addr. | |
1015 | */ | |
1016 | if (user_alloc) | |
1017 | new.userspace_addr = mem->userspace_addr; | |
1018 | else | |
1019 | new.userspace_addr = 0; | |
6aa8b732 | 1020 | } |
05da4558 MT |
1021 | if (npages && !new.lpage_info) { |
1022 | int largepages = npages / KVM_PAGES_PER_HPAGE; | |
1023 | if (npages % KVM_PAGES_PER_HPAGE) | |
1024 | largepages++; | |
1025 | if (base_gfn % KVM_PAGES_PER_HPAGE) | |
1026 | largepages++; | |
1027 | ||
1028 | new.lpage_info = vmalloc(largepages * sizeof(*new.lpage_info)); | |
1029 | ||
1030 | if (!new.lpage_info) | |
1031 | goto out_free; | |
1032 | ||
1033 | memset(new.lpage_info, 0, largepages * sizeof(*new.lpage_info)); | |
1034 | ||
1035 | if (base_gfn % KVM_PAGES_PER_HPAGE) | |
1036 | new.lpage_info[0].write_count = 1; | |
1037 | if ((base_gfn+npages) % KVM_PAGES_PER_HPAGE) | |
1038 | new.lpage_info[largepages-1].write_count = 1; | |
1039 | } | |
6aa8b732 AK |
1040 | |
1041 | /* Allocate page dirty bitmap if needed */ | |
1042 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
1043 | unsigned dirty_bytes = ALIGN(npages, BITS_PER_LONG) / 8; | |
1044 | ||
1045 | new.dirty_bitmap = vmalloc(dirty_bytes); | |
1046 | if (!new.dirty_bitmap) | |
f78e0e2e | 1047 | goto out_free; |
6aa8b732 AK |
1048 | memset(new.dirty_bitmap, 0, dirty_bytes); |
1049 | } | |
eff0114a | 1050 | #endif /* not defined CONFIG_S390 */ |
6aa8b732 | 1051 | |
34d4cb8f MT |
1052 | if (!npages) |
1053 | kvm_arch_flush_shadow(kvm); | |
1054 | ||
604b38ac AA |
1055 | spin_lock(&kvm->mmu_lock); |
1056 | if (mem->slot >= kvm->nmemslots) | |
1057 | kvm->nmemslots = mem->slot + 1; | |
1058 | ||
3ad82a7e | 1059 | *memslot = new; |
604b38ac | 1060 | spin_unlock(&kvm->mmu_lock); |
3ad82a7e | 1061 | |
0de10343 ZX |
1062 | r = kvm_arch_set_memory_region(kvm, mem, old, user_alloc); |
1063 | if (r) { | |
604b38ac | 1064 | spin_lock(&kvm->mmu_lock); |
0de10343 | 1065 | *memslot = old; |
604b38ac | 1066 | spin_unlock(&kvm->mmu_lock); |
0de10343 | 1067 | goto out_free; |
82ce2c96 IE |
1068 | } |
1069 | ||
6f897248 GC |
1070 | kvm_free_physmem_slot(&old, npages ? &new : NULL); |
1071 | /* Slot deletion case: we have to update the current slot */ | |
1072 | if (!npages) | |
1073 | *memslot = old; | |
8a98f664 | 1074 | #ifdef CONFIG_DMAR |
62c476c7 BAY |
1075 | /* map the pages in iommu page table */ |
1076 | r = kvm_iommu_map_pages(kvm, base_gfn, npages); | |
1077 | if (r) | |
1078 | goto out; | |
8a98f664 | 1079 | #endif |
6aa8b732 AK |
1080 | return 0; |
1081 | ||
f78e0e2e | 1082 | out_free: |
6aa8b732 AK |
1083 | kvm_free_physmem_slot(&new, &old); |
1084 | out: | |
1085 | return r; | |
210c7c4d IE |
1086 | |
1087 | } | |
f78e0e2e SY |
1088 | EXPORT_SYMBOL_GPL(__kvm_set_memory_region); |
1089 | ||
1090 | int kvm_set_memory_region(struct kvm *kvm, | |
1091 | struct kvm_userspace_memory_region *mem, | |
1092 | int user_alloc) | |
1093 | { | |
1094 | int r; | |
1095 | ||
72dc67a6 | 1096 | down_write(&kvm->slots_lock); |
f78e0e2e | 1097 | r = __kvm_set_memory_region(kvm, mem, user_alloc); |
72dc67a6 | 1098 | up_write(&kvm->slots_lock); |
f78e0e2e SY |
1099 | return r; |
1100 | } | |
210c7c4d IE |
1101 | EXPORT_SYMBOL_GPL(kvm_set_memory_region); |
1102 | ||
1fe779f8 CO |
1103 | int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
1104 | struct | |
1105 | kvm_userspace_memory_region *mem, | |
1106 | int user_alloc) | |
210c7c4d | 1107 | { |
e0d62c7f IE |
1108 | if (mem->slot >= KVM_MEMORY_SLOTS) |
1109 | return -EINVAL; | |
210c7c4d | 1110 | return kvm_set_memory_region(kvm, mem, user_alloc); |
6aa8b732 AK |
1111 | } |
1112 | ||
5bb064dc ZX |
1113 | int kvm_get_dirty_log(struct kvm *kvm, |
1114 | struct kvm_dirty_log *log, int *is_dirty) | |
6aa8b732 AK |
1115 | { |
1116 | struct kvm_memory_slot *memslot; | |
1117 | int r, i; | |
1118 | int n; | |
1119 | unsigned long any = 0; | |
1120 | ||
6aa8b732 AK |
1121 | r = -EINVAL; |
1122 | if (log->slot >= KVM_MEMORY_SLOTS) | |
1123 | goto out; | |
1124 | ||
1125 | memslot = &kvm->memslots[log->slot]; | |
1126 | r = -ENOENT; | |
1127 | if (!memslot->dirty_bitmap) | |
1128 | goto out; | |
1129 | ||
cd1a4a98 | 1130 | n = ALIGN(memslot->npages, BITS_PER_LONG) / 8; |
6aa8b732 | 1131 | |
cd1a4a98 | 1132 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
1133 | any = memslot->dirty_bitmap[i]; |
1134 | ||
1135 | r = -EFAULT; | |
1136 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
1137 | goto out; | |
1138 | ||
5bb064dc ZX |
1139 | if (any) |
1140 | *is_dirty = 1; | |
6aa8b732 AK |
1141 | |
1142 | r = 0; | |
6aa8b732 | 1143 | out: |
6aa8b732 AK |
1144 | return r; |
1145 | } | |
1146 | ||
cea7bb21 IE |
1147 | int is_error_page(struct page *page) |
1148 | { | |
1149 | return page == bad_page; | |
1150 | } | |
1151 | EXPORT_SYMBOL_GPL(is_error_page); | |
1152 | ||
35149e21 AL |
1153 | int is_error_pfn(pfn_t pfn) |
1154 | { | |
1155 | return pfn == bad_pfn; | |
1156 | } | |
1157 | EXPORT_SYMBOL_GPL(is_error_pfn); | |
1158 | ||
f9d46eb0 IE |
1159 | static inline unsigned long bad_hva(void) |
1160 | { | |
1161 | return PAGE_OFFSET; | |
1162 | } | |
1163 | ||
1164 | int kvm_is_error_hva(unsigned long addr) | |
1165 | { | |
1166 | return addr == bad_hva(); | |
1167 | } | |
1168 | EXPORT_SYMBOL_GPL(kvm_is_error_hva); | |
1169 | ||
2843099f | 1170 | struct kvm_memory_slot *gfn_to_memslot_unaliased(struct kvm *kvm, gfn_t gfn) |
6aa8b732 AK |
1171 | { |
1172 | int i; | |
1173 | ||
1174 | for (i = 0; i < kvm->nmemslots; ++i) { | |
1175 | struct kvm_memory_slot *memslot = &kvm->memslots[i]; | |
1176 | ||
1177 | if (gfn >= memslot->base_gfn | |
1178 | && gfn < memslot->base_gfn + memslot->npages) | |
1179 | return memslot; | |
1180 | } | |
8b6d44c7 | 1181 | return NULL; |
6aa8b732 | 1182 | } |
2843099f | 1183 | EXPORT_SYMBOL_GPL(gfn_to_memslot_unaliased); |
e8207547 AK |
1184 | |
1185 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
1186 | { | |
1187 | gfn = unalias_gfn(kvm, gfn); | |
2843099f | 1188 | return gfn_to_memslot_unaliased(kvm, gfn); |
e8207547 | 1189 | } |
6aa8b732 | 1190 | |
e0d62c7f IE |
1191 | int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
1192 | { | |
1193 | int i; | |
1194 | ||
1195 | gfn = unalias_gfn(kvm, gfn); | |
1196 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
1197 | struct kvm_memory_slot *memslot = &kvm->memslots[i]; | |
1198 | ||
1199 | if (gfn >= memslot->base_gfn | |
1200 | && gfn < memslot->base_gfn + memslot->npages) | |
1201 | return 1; | |
1202 | } | |
1203 | return 0; | |
1204 | } | |
1205 | EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); | |
1206 | ||
05da4558 | 1207 | unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn) |
539cb660 IE |
1208 | { |
1209 | struct kvm_memory_slot *slot; | |
1210 | ||
1211 | gfn = unalias_gfn(kvm, gfn); | |
2843099f | 1212 | slot = gfn_to_memslot_unaliased(kvm, gfn); |
539cb660 IE |
1213 | if (!slot) |
1214 | return bad_hva(); | |
1215 | return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE); | |
1216 | } | |
0d150298 | 1217 | EXPORT_SYMBOL_GPL(gfn_to_hva); |
539cb660 | 1218 | |
35149e21 | 1219 | pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn) |
954bbbc2 | 1220 | { |
8d4e1288 | 1221 | struct page *page[1]; |
539cb660 | 1222 | unsigned long addr; |
8d4e1288 | 1223 | int npages; |
2e2e3738 | 1224 | pfn_t pfn; |
954bbbc2 | 1225 | |
60395224 AK |
1226 | might_sleep(); |
1227 | ||
539cb660 IE |
1228 | addr = gfn_to_hva(kvm, gfn); |
1229 | if (kvm_is_error_hva(addr)) { | |
8a7ae055 | 1230 | get_page(bad_page); |
35149e21 | 1231 | return page_to_pfn(bad_page); |
8a7ae055 | 1232 | } |
8d4e1288 | 1233 | |
4c2155ce | 1234 | npages = get_user_pages_fast(addr, 1, 1, page); |
539cb660 | 1235 | |
2e2e3738 AL |
1236 | if (unlikely(npages != 1)) { |
1237 | struct vm_area_struct *vma; | |
1238 | ||
4c2155ce | 1239 | down_read(¤t->mm->mmap_sem); |
2e2e3738 | 1240 | vma = find_vma(current->mm, addr); |
4c2155ce | 1241 | |
2e2e3738 AL |
1242 | if (vma == NULL || addr < vma->vm_start || |
1243 | !(vma->vm_flags & VM_PFNMAP)) { | |
4c2155ce | 1244 | up_read(¤t->mm->mmap_sem); |
2e2e3738 AL |
1245 | get_page(bad_page); |
1246 | return page_to_pfn(bad_page); | |
1247 | } | |
1248 | ||
1249 | pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; | |
4c2155ce | 1250 | up_read(¤t->mm->mmap_sem); |
c77fb9dc | 1251 | BUG_ON(!kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
1252 | } else |
1253 | pfn = page_to_pfn(page[0]); | |
8d4e1288 | 1254 | |
2e2e3738 | 1255 | return pfn; |
35149e21 AL |
1256 | } |
1257 | ||
1258 | EXPORT_SYMBOL_GPL(gfn_to_pfn); | |
1259 | ||
1260 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) | |
1261 | { | |
2e2e3738 AL |
1262 | pfn_t pfn; |
1263 | ||
1264 | pfn = gfn_to_pfn(kvm, gfn); | |
c77fb9dc | 1265 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 AL |
1266 | return pfn_to_page(pfn); |
1267 | ||
c77fb9dc | 1268 | WARN_ON(kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
1269 | |
1270 | get_page(bad_page); | |
1271 | return bad_page; | |
954bbbc2 | 1272 | } |
aab61cc0 | 1273 | |
954bbbc2 AK |
1274 | EXPORT_SYMBOL_GPL(gfn_to_page); |
1275 | ||
b4231d61 IE |
1276 | void kvm_release_page_clean(struct page *page) |
1277 | { | |
35149e21 | 1278 | kvm_release_pfn_clean(page_to_pfn(page)); |
b4231d61 IE |
1279 | } |
1280 | EXPORT_SYMBOL_GPL(kvm_release_page_clean); | |
1281 | ||
35149e21 AL |
1282 | void kvm_release_pfn_clean(pfn_t pfn) |
1283 | { | |
c77fb9dc | 1284 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1285 | put_page(pfn_to_page(pfn)); |
35149e21 AL |
1286 | } |
1287 | EXPORT_SYMBOL_GPL(kvm_release_pfn_clean); | |
1288 | ||
b4231d61 | 1289 | void kvm_release_page_dirty(struct page *page) |
8a7ae055 | 1290 | { |
35149e21 AL |
1291 | kvm_release_pfn_dirty(page_to_pfn(page)); |
1292 | } | |
1293 | EXPORT_SYMBOL_GPL(kvm_release_page_dirty); | |
1294 | ||
1295 | void kvm_release_pfn_dirty(pfn_t pfn) | |
1296 | { | |
1297 | kvm_set_pfn_dirty(pfn); | |
1298 | kvm_release_pfn_clean(pfn); | |
1299 | } | |
1300 | EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty); | |
1301 | ||
1302 | void kvm_set_page_dirty(struct page *page) | |
1303 | { | |
1304 | kvm_set_pfn_dirty(page_to_pfn(page)); | |
1305 | } | |
1306 | EXPORT_SYMBOL_GPL(kvm_set_page_dirty); | |
1307 | ||
1308 | void kvm_set_pfn_dirty(pfn_t pfn) | |
1309 | { | |
c77fb9dc | 1310 | if (!kvm_is_mmio_pfn(pfn)) { |
2e2e3738 AL |
1311 | struct page *page = pfn_to_page(pfn); |
1312 | if (!PageReserved(page)) | |
1313 | SetPageDirty(page); | |
1314 | } | |
8a7ae055 | 1315 | } |
35149e21 AL |
1316 | EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty); |
1317 | ||
1318 | void kvm_set_pfn_accessed(pfn_t pfn) | |
1319 | { | |
c77fb9dc | 1320 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1321 | mark_page_accessed(pfn_to_page(pfn)); |
35149e21 AL |
1322 | } |
1323 | EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed); | |
1324 | ||
1325 | void kvm_get_pfn(pfn_t pfn) | |
1326 | { | |
c77fb9dc | 1327 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1328 | get_page(pfn_to_page(pfn)); |
35149e21 AL |
1329 | } |
1330 | EXPORT_SYMBOL_GPL(kvm_get_pfn); | |
8a7ae055 | 1331 | |
195aefde IE |
1332 | static int next_segment(unsigned long len, int offset) |
1333 | { | |
1334 | if (len > PAGE_SIZE - offset) | |
1335 | return PAGE_SIZE - offset; | |
1336 | else | |
1337 | return len; | |
1338 | } | |
1339 | ||
1340 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, | |
1341 | int len) | |
1342 | { | |
e0506bcb IE |
1343 | int r; |
1344 | unsigned long addr; | |
195aefde | 1345 | |
e0506bcb IE |
1346 | addr = gfn_to_hva(kvm, gfn); |
1347 | if (kvm_is_error_hva(addr)) | |
1348 | return -EFAULT; | |
1349 | r = copy_from_user(data, (void __user *)addr + offset, len); | |
1350 | if (r) | |
195aefde | 1351 | return -EFAULT; |
195aefde IE |
1352 | return 0; |
1353 | } | |
1354 | EXPORT_SYMBOL_GPL(kvm_read_guest_page); | |
1355 | ||
1356 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len) | |
1357 | { | |
1358 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1359 | int seg; | |
1360 | int offset = offset_in_page(gpa); | |
1361 | int ret; | |
1362 | ||
1363 | while ((seg = next_segment(len, offset)) != 0) { | |
1364 | ret = kvm_read_guest_page(kvm, gfn, data, offset, seg); | |
1365 | if (ret < 0) | |
1366 | return ret; | |
1367 | offset = 0; | |
1368 | len -= seg; | |
1369 | data += seg; | |
1370 | ++gfn; | |
1371 | } | |
1372 | return 0; | |
1373 | } | |
1374 | EXPORT_SYMBOL_GPL(kvm_read_guest); | |
1375 | ||
7ec54588 MT |
1376 | int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data, |
1377 | unsigned long len) | |
1378 | { | |
1379 | int r; | |
1380 | unsigned long addr; | |
1381 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1382 | int offset = offset_in_page(gpa); | |
1383 | ||
1384 | addr = gfn_to_hva(kvm, gfn); | |
1385 | if (kvm_is_error_hva(addr)) | |
1386 | return -EFAULT; | |
0aac03f0 | 1387 | pagefault_disable(); |
7ec54588 | 1388 | r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len); |
0aac03f0 | 1389 | pagefault_enable(); |
7ec54588 MT |
1390 | if (r) |
1391 | return -EFAULT; | |
1392 | return 0; | |
1393 | } | |
1394 | EXPORT_SYMBOL(kvm_read_guest_atomic); | |
1395 | ||
195aefde IE |
1396 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, |
1397 | int offset, int len) | |
1398 | { | |
e0506bcb IE |
1399 | int r; |
1400 | unsigned long addr; | |
195aefde | 1401 | |
e0506bcb IE |
1402 | addr = gfn_to_hva(kvm, gfn); |
1403 | if (kvm_is_error_hva(addr)) | |
1404 | return -EFAULT; | |
1405 | r = copy_to_user((void __user *)addr + offset, data, len); | |
1406 | if (r) | |
195aefde | 1407 | return -EFAULT; |
195aefde IE |
1408 | mark_page_dirty(kvm, gfn); |
1409 | return 0; | |
1410 | } | |
1411 | EXPORT_SYMBOL_GPL(kvm_write_guest_page); | |
1412 | ||
1413 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
1414 | unsigned long len) | |
1415 | { | |
1416 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1417 | int seg; | |
1418 | int offset = offset_in_page(gpa); | |
1419 | int ret; | |
1420 | ||
1421 | while ((seg = next_segment(len, offset)) != 0) { | |
1422 | ret = kvm_write_guest_page(kvm, gfn, data, offset, seg); | |
1423 | if (ret < 0) | |
1424 | return ret; | |
1425 | offset = 0; | |
1426 | len -= seg; | |
1427 | data += seg; | |
1428 | ++gfn; | |
1429 | } | |
1430 | return 0; | |
1431 | } | |
1432 | ||
1433 | int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len) | |
1434 | { | |
3e021bf5 | 1435 | return kvm_write_guest_page(kvm, gfn, empty_zero_page, offset, len); |
195aefde IE |
1436 | } |
1437 | EXPORT_SYMBOL_GPL(kvm_clear_guest_page); | |
1438 | ||
1439 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len) | |
1440 | { | |
1441 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1442 | int seg; | |
1443 | int offset = offset_in_page(gpa); | |
1444 | int ret; | |
1445 | ||
1446 | while ((seg = next_segment(len, offset)) != 0) { | |
1447 | ret = kvm_clear_guest_page(kvm, gfn, offset, seg); | |
1448 | if (ret < 0) | |
1449 | return ret; | |
1450 | offset = 0; | |
1451 | len -= seg; | |
1452 | ++gfn; | |
1453 | } | |
1454 | return 0; | |
1455 | } | |
1456 | EXPORT_SYMBOL_GPL(kvm_clear_guest); | |
1457 | ||
6aa8b732 AK |
1458 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
1459 | { | |
31389947 | 1460 | struct kvm_memory_slot *memslot; |
6aa8b732 | 1461 | |
3b6fff19 | 1462 | gfn = unalias_gfn(kvm, gfn); |
2843099f | 1463 | memslot = gfn_to_memslot_unaliased(kvm, gfn); |
7e9d619d RR |
1464 | if (memslot && memslot->dirty_bitmap) { |
1465 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 1466 | |
7e9d619d RR |
1467 | /* avoid RMW */ |
1468 | if (!test_bit(rel_gfn, memslot->dirty_bitmap)) | |
1469 | set_bit(rel_gfn, memslot->dirty_bitmap); | |
6aa8b732 AK |
1470 | } |
1471 | } | |
1472 | ||
b6958ce4 ED |
1473 | /* |
1474 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
1475 | */ | |
8776e519 | 1476 | void kvm_vcpu_block(struct kvm_vcpu *vcpu) |
d3bef15f | 1477 | { |
e5c239cf MT |
1478 | DEFINE_WAIT(wait); |
1479 | ||
1480 | for (;;) { | |
1481 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1482 | ||
d7690175 MT |
1483 | if (kvm_cpu_has_interrupt(vcpu) || |
1484 | kvm_cpu_has_pending_timer(vcpu) || | |
1485 | kvm_arch_vcpu_runnable(vcpu)) { | |
1486 | set_bit(KVM_REQ_UNHALT, &vcpu->requests); | |
e5c239cf | 1487 | break; |
d7690175 | 1488 | } |
e5c239cf MT |
1489 | if (signal_pending(current)) |
1490 | break; | |
1491 | ||
b6958ce4 ED |
1492 | vcpu_put(vcpu); |
1493 | schedule(); | |
1494 | vcpu_load(vcpu); | |
1495 | } | |
d3bef15f | 1496 | |
e5c239cf | 1497 | finish_wait(&vcpu->wq, &wait); |
b6958ce4 ED |
1498 | } |
1499 | ||
6aa8b732 AK |
1500 | void kvm_resched(struct kvm_vcpu *vcpu) |
1501 | { | |
3fca0365 YD |
1502 | if (!need_resched()) |
1503 | return; | |
6aa8b732 | 1504 | cond_resched(); |
6aa8b732 AK |
1505 | } |
1506 | EXPORT_SYMBOL_GPL(kvm_resched); | |
1507 | ||
e4a533a4 | 1508 | static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
9a2bb7f4 AK |
1509 | { |
1510 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
9a2bb7f4 AK |
1511 | struct page *page; |
1512 | ||
e4a533a4 | 1513 | if (vmf->pgoff == 0) |
039576c0 | 1514 | page = virt_to_page(vcpu->run); |
09566765 | 1515 | #ifdef CONFIG_X86 |
e4a533a4 | 1516 | else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET) |
ad312c7c | 1517 | page = virt_to_page(vcpu->arch.pio_data); |
5f94c174 LV |
1518 | #endif |
1519 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
1520 | else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET) | |
1521 | page = virt_to_page(vcpu->kvm->coalesced_mmio_ring); | |
09566765 | 1522 | #endif |
039576c0 | 1523 | else |
e4a533a4 | 1524 | return VM_FAULT_SIGBUS; |
9a2bb7f4 | 1525 | get_page(page); |
e4a533a4 NP |
1526 | vmf->page = page; |
1527 | return 0; | |
9a2bb7f4 AK |
1528 | } |
1529 | ||
1530 | static struct vm_operations_struct kvm_vcpu_vm_ops = { | |
e4a533a4 | 1531 | .fault = kvm_vcpu_fault, |
9a2bb7f4 AK |
1532 | }; |
1533 | ||
1534 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
1535 | { | |
1536 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
1537 | return 0; | |
1538 | } | |
1539 | ||
bccf2150 AK |
1540 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
1541 | { | |
1542 | struct kvm_vcpu *vcpu = filp->private_data; | |
1543 | ||
66c0b394 | 1544 | kvm_put_kvm(vcpu->kvm); |
bccf2150 AK |
1545 | return 0; |
1546 | } | |
1547 | ||
3d3aab1b | 1548 | static struct file_operations kvm_vcpu_fops = { |
bccf2150 AK |
1549 | .release = kvm_vcpu_release, |
1550 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
1551 | .compat_ioctl = kvm_vcpu_ioctl, | |
9a2bb7f4 | 1552 | .mmap = kvm_vcpu_mmap, |
bccf2150 AK |
1553 | }; |
1554 | ||
1555 | /* | |
1556 | * Allocates an inode for the vcpu. | |
1557 | */ | |
1558 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
1559 | { | |
7d9dbca3 | 1560 | int fd = anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, 0); |
2030a42c | 1561 | if (fd < 0) |
66c0b394 | 1562 | kvm_put_kvm(vcpu->kvm); |
bccf2150 | 1563 | return fd; |
bccf2150 AK |
1564 | } |
1565 | ||
c5ea7660 AK |
1566 | /* |
1567 | * Creates some virtual cpus. Good luck creating more than one. | |
1568 | */ | |
1569 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, int n) | |
1570 | { | |
1571 | int r; | |
1572 | struct kvm_vcpu *vcpu; | |
1573 | ||
c5ea7660 | 1574 | if (!valid_vcpu(n)) |
fb3f0f51 | 1575 | return -EINVAL; |
c5ea7660 | 1576 | |
e9b11c17 | 1577 | vcpu = kvm_arch_vcpu_create(kvm, n); |
fb3f0f51 RR |
1578 | if (IS_ERR(vcpu)) |
1579 | return PTR_ERR(vcpu); | |
c5ea7660 | 1580 | |
15ad7146 AK |
1581 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
1582 | ||
26e5215f AK |
1583 | r = kvm_arch_vcpu_setup(vcpu); |
1584 | if (r) | |
7d8fece6 | 1585 | return r; |
26e5215f | 1586 | |
11ec2804 | 1587 | mutex_lock(&kvm->lock); |
fb3f0f51 RR |
1588 | if (kvm->vcpus[n]) { |
1589 | r = -EEXIST; | |
e9b11c17 | 1590 | goto vcpu_destroy; |
fb3f0f51 RR |
1591 | } |
1592 | kvm->vcpus[n] = vcpu; | |
11ec2804 | 1593 | mutex_unlock(&kvm->lock); |
c5ea7660 | 1594 | |
fb3f0f51 | 1595 | /* Now it's all set up, let userspace reach it */ |
66c0b394 | 1596 | kvm_get_kvm(kvm); |
bccf2150 AK |
1597 | r = create_vcpu_fd(vcpu); |
1598 | if (r < 0) | |
fb3f0f51 RR |
1599 | goto unlink; |
1600 | return r; | |
39c3b86e | 1601 | |
fb3f0f51 | 1602 | unlink: |
11ec2804 | 1603 | mutex_lock(&kvm->lock); |
fb3f0f51 | 1604 | kvm->vcpus[n] = NULL; |
e9b11c17 | 1605 | vcpu_destroy: |
7d8fece6 | 1606 | mutex_unlock(&kvm->lock); |
d40ccc62 | 1607 | kvm_arch_vcpu_destroy(vcpu); |
c5ea7660 AK |
1608 | return r; |
1609 | } | |
1610 | ||
1961d276 AK |
1611 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
1612 | { | |
1613 | if (sigset) { | |
1614 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
1615 | vcpu->sigset_active = 1; | |
1616 | vcpu->sigset = *sigset; | |
1617 | } else | |
1618 | vcpu->sigset_active = 0; | |
1619 | return 0; | |
1620 | } | |
1621 | ||
bccf2150 AK |
1622 | static long kvm_vcpu_ioctl(struct file *filp, |
1623 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 1624 | { |
bccf2150 | 1625 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 1626 | void __user *argp = (void __user *)arg; |
313a3dc7 | 1627 | int r; |
fa3795a7 DH |
1628 | struct kvm_fpu *fpu = NULL; |
1629 | struct kvm_sregs *kvm_sregs = NULL; | |
6aa8b732 | 1630 | |
6d4e4c4f AK |
1631 | if (vcpu->kvm->mm != current->mm) |
1632 | return -EIO; | |
6aa8b732 | 1633 | switch (ioctl) { |
9a2bb7f4 | 1634 | case KVM_RUN: |
f0fe5108 AK |
1635 | r = -EINVAL; |
1636 | if (arg) | |
1637 | goto out; | |
b6c7a5dc | 1638 | r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run); |
6aa8b732 | 1639 | break; |
6aa8b732 | 1640 | case KVM_GET_REGS: { |
3e4bb3ac | 1641 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1642 | |
3e4bb3ac XZ |
1643 | r = -ENOMEM; |
1644 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1645 | if (!kvm_regs) | |
6aa8b732 | 1646 | goto out; |
3e4bb3ac XZ |
1647 | r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs); |
1648 | if (r) | |
1649 | goto out_free1; | |
6aa8b732 | 1650 | r = -EFAULT; |
3e4bb3ac XZ |
1651 | if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs))) |
1652 | goto out_free1; | |
6aa8b732 | 1653 | r = 0; |
3e4bb3ac XZ |
1654 | out_free1: |
1655 | kfree(kvm_regs); | |
6aa8b732 AK |
1656 | break; |
1657 | } | |
1658 | case KVM_SET_REGS: { | |
3e4bb3ac | 1659 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1660 | |
3e4bb3ac XZ |
1661 | r = -ENOMEM; |
1662 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1663 | if (!kvm_regs) | |
6aa8b732 | 1664 | goto out; |
3e4bb3ac XZ |
1665 | r = -EFAULT; |
1666 | if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs))) | |
1667 | goto out_free2; | |
1668 | r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs); | |
6aa8b732 | 1669 | if (r) |
3e4bb3ac | 1670 | goto out_free2; |
6aa8b732 | 1671 | r = 0; |
3e4bb3ac XZ |
1672 | out_free2: |
1673 | kfree(kvm_regs); | |
6aa8b732 AK |
1674 | break; |
1675 | } | |
1676 | case KVM_GET_SREGS: { | |
fa3795a7 DH |
1677 | kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1678 | r = -ENOMEM; | |
1679 | if (!kvm_sregs) | |
1680 | goto out; | |
1681 | r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs); | |
6aa8b732 AK |
1682 | if (r) |
1683 | goto out; | |
1684 | r = -EFAULT; | |
fa3795a7 | 1685 | if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs))) |
6aa8b732 AK |
1686 | goto out; |
1687 | r = 0; | |
1688 | break; | |
1689 | } | |
1690 | case KVM_SET_SREGS: { | |
fa3795a7 DH |
1691 | kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1692 | r = -ENOMEM; | |
1693 | if (!kvm_sregs) | |
1694 | goto out; | |
6aa8b732 | 1695 | r = -EFAULT; |
fa3795a7 | 1696 | if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs))) |
6aa8b732 | 1697 | goto out; |
fa3795a7 | 1698 | r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs); |
6aa8b732 AK |
1699 | if (r) |
1700 | goto out; | |
1701 | r = 0; | |
1702 | break; | |
1703 | } | |
62d9f0db MT |
1704 | case KVM_GET_MP_STATE: { |
1705 | struct kvm_mp_state mp_state; | |
1706 | ||
1707 | r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state); | |
1708 | if (r) | |
1709 | goto out; | |
1710 | r = -EFAULT; | |
1711 | if (copy_to_user(argp, &mp_state, sizeof mp_state)) | |
1712 | goto out; | |
1713 | r = 0; | |
1714 | break; | |
1715 | } | |
1716 | case KVM_SET_MP_STATE: { | |
1717 | struct kvm_mp_state mp_state; | |
1718 | ||
1719 | r = -EFAULT; | |
1720 | if (copy_from_user(&mp_state, argp, sizeof mp_state)) | |
1721 | goto out; | |
1722 | r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state); | |
1723 | if (r) | |
1724 | goto out; | |
1725 | r = 0; | |
1726 | break; | |
1727 | } | |
6aa8b732 AK |
1728 | case KVM_TRANSLATE: { |
1729 | struct kvm_translation tr; | |
1730 | ||
1731 | r = -EFAULT; | |
2f366987 | 1732 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 1733 | goto out; |
8b006791 | 1734 | r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
1735 | if (r) |
1736 | goto out; | |
1737 | r = -EFAULT; | |
2f366987 | 1738 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
1739 | goto out; |
1740 | r = 0; | |
1741 | break; | |
1742 | } | |
6aa8b732 AK |
1743 | case KVM_DEBUG_GUEST: { |
1744 | struct kvm_debug_guest dbg; | |
1745 | ||
1746 | r = -EFAULT; | |
2f366987 | 1747 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 1748 | goto out; |
b6c7a5dc | 1749 | r = kvm_arch_vcpu_ioctl_debug_guest(vcpu, &dbg); |
6aa8b732 AK |
1750 | if (r) |
1751 | goto out; | |
1752 | r = 0; | |
1753 | break; | |
1754 | } | |
1961d276 AK |
1755 | case KVM_SET_SIGNAL_MASK: { |
1756 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
1757 | struct kvm_signal_mask kvm_sigmask; | |
1758 | sigset_t sigset, *p; | |
1759 | ||
1760 | p = NULL; | |
1761 | if (argp) { | |
1762 | r = -EFAULT; | |
1763 | if (copy_from_user(&kvm_sigmask, argp, | |
1764 | sizeof kvm_sigmask)) | |
1765 | goto out; | |
1766 | r = -EINVAL; | |
1767 | if (kvm_sigmask.len != sizeof sigset) | |
1768 | goto out; | |
1769 | r = -EFAULT; | |
1770 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
1771 | sizeof sigset)) | |
1772 | goto out; | |
1773 | p = &sigset; | |
1774 | } | |
1775 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset); | |
1776 | break; | |
1777 | } | |
b8836737 | 1778 | case KVM_GET_FPU: { |
fa3795a7 DH |
1779 | fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1780 | r = -ENOMEM; | |
1781 | if (!fpu) | |
1782 | goto out; | |
1783 | r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu); | |
b8836737 AK |
1784 | if (r) |
1785 | goto out; | |
1786 | r = -EFAULT; | |
fa3795a7 | 1787 | if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu))) |
b8836737 AK |
1788 | goto out; |
1789 | r = 0; | |
1790 | break; | |
1791 | } | |
1792 | case KVM_SET_FPU: { | |
fa3795a7 DH |
1793 | fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1794 | r = -ENOMEM; | |
1795 | if (!fpu) | |
1796 | goto out; | |
b8836737 | 1797 | r = -EFAULT; |
fa3795a7 | 1798 | if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu))) |
b8836737 | 1799 | goto out; |
fa3795a7 | 1800 | r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu); |
b8836737 AK |
1801 | if (r) |
1802 | goto out; | |
1803 | r = 0; | |
1804 | break; | |
1805 | } | |
bccf2150 | 1806 | default: |
313a3dc7 | 1807 | r = kvm_arch_vcpu_ioctl(filp, ioctl, arg); |
bccf2150 AK |
1808 | } |
1809 | out: | |
fa3795a7 DH |
1810 | kfree(fpu); |
1811 | kfree(kvm_sregs); | |
bccf2150 AK |
1812 | return r; |
1813 | } | |
1814 | ||
1815 | static long kvm_vm_ioctl(struct file *filp, | |
1816 | unsigned int ioctl, unsigned long arg) | |
1817 | { | |
1818 | struct kvm *kvm = filp->private_data; | |
1819 | void __user *argp = (void __user *)arg; | |
1fe779f8 | 1820 | int r; |
bccf2150 | 1821 | |
6d4e4c4f AK |
1822 | if (kvm->mm != current->mm) |
1823 | return -EIO; | |
bccf2150 AK |
1824 | switch (ioctl) { |
1825 | case KVM_CREATE_VCPU: | |
1826 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
1827 | if (r < 0) | |
1828 | goto out; | |
1829 | break; | |
6fc138d2 IE |
1830 | case KVM_SET_USER_MEMORY_REGION: { |
1831 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
1832 | ||
1833 | r = -EFAULT; | |
1834 | if (copy_from_user(&kvm_userspace_mem, argp, | |
1835 | sizeof kvm_userspace_mem)) | |
1836 | goto out; | |
1837 | ||
1838 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1); | |
6aa8b732 AK |
1839 | if (r) |
1840 | goto out; | |
1841 | break; | |
1842 | } | |
1843 | case KVM_GET_DIRTY_LOG: { | |
1844 | struct kvm_dirty_log log; | |
1845 | ||
1846 | r = -EFAULT; | |
2f366987 | 1847 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 1848 | goto out; |
2c6f5df9 | 1849 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
1850 | if (r) |
1851 | goto out; | |
1852 | break; | |
1853 | } | |
5f94c174 LV |
1854 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
1855 | case KVM_REGISTER_COALESCED_MMIO: { | |
1856 | struct kvm_coalesced_mmio_zone zone; | |
1857 | r = -EFAULT; | |
1858 | if (copy_from_user(&zone, argp, sizeof zone)) | |
1859 | goto out; | |
1860 | r = -ENXIO; | |
1861 | r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone); | |
1862 | if (r) | |
1863 | goto out; | |
1864 | r = 0; | |
1865 | break; | |
1866 | } | |
1867 | case KVM_UNREGISTER_COALESCED_MMIO: { | |
1868 | struct kvm_coalesced_mmio_zone zone; | |
1869 | r = -EFAULT; | |
1870 | if (copy_from_user(&zone, argp, sizeof zone)) | |
1871 | goto out; | |
1872 | r = -ENXIO; | |
1873 | r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone); | |
1874 | if (r) | |
1875 | goto out; | |
1876 | r = 0; | |
1877 | break; | |
1878 | } | |
8a98f664 XZ |
1879 | #endif |
1880 | #ifdef KVM_CAP_DEVICE_ASSIGNMENT | |
1881 | case KVM_ASSIGN_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_assign_device(kvm, &assigned_dev); | |
1888 | if (r) | |
1889 | goto out; | |
1890 | break; | |
1891 | } | |
1892 | case KVM_ASSIGN_IRQ: { | |
1893 | struct kvm_assigned_irq assigned_irq; | |
1894 | ||
1895 | r = -EFAULT; | |
1896 | if (copy_from_user(&assigned_irq, argp, sizeof assigned_irq)) | |
1897 | goto out; | |
1898 | r = kvm_vm_ioctl_assign_irq(kvm, &assigned_irq); | |
1899 | if (r) | |
1900 | goto out; | |
1901 | break; | |
1902 | } | |
0a920356 WH |
1903 | #endif |
1904 | #ifdef KVM_CAP_DEVICE_DEASSIGNMENT | |
1905 | case KVM_DEASSIGN_PCI_DEVICE: { | |
1906 | struct kvm_assigned_pci_dev assigned_dev; | |
1907 | ||
1908 | r = -EFAULT; | |
1909 | if (copy_from_user(&assigned_dev, argp, sizeof assigned_dev)) | |
1910 | goto out; | |
1911 | r = kvm_vm_ioctl_deassign_device(kvm, &assigned_dev); | |
1912 | if (r) | |
1913 | goto out; | |
1914 | break; | |
1915 | } | |
5f94c174 | 1916 | #endif |
f17abe9a | 1917 | default: |
1fe779f8 | 1918 | r = kvm_arch_vm_ioctl(filp, ioctl, arg); |
f17abe9a AK |
1919 | } |
1920 | out: | |
1921 | return r; | |
1922 | } | |
1923 | ||
e4a533a4 | 1924 | static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
f17abe9a | 1925 | { |
777b3f49 MT |
1926 | struct page *page[1]; |
1927 | unsigned long addr; | |
1928 | int npages; | |
1929 | gfn_t gfn = vmf->pgoff; | |
f17abe9a | 1930 | struct kvm *kvm = vma->vm_file->private_data; |
f17abe9a | 1931 | |
777b3f49 MT |
1932 | addr = gfn_to_hva(kvm, gfn); |
1933 | if (kvm_is_error_hva(addr)) | |
e4a533a4 | 1934 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
1935 | |
1936 | npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page, | |
1937 | NULL); | |
1938 | if (unlikely(npages != 1)) | |
e4a533a4 | 1939 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
1940 | |
1941 | vmf->page = page[0]; | |
e4a533a4 | 1942 | return 0; |
f17abe9a AK |
1943 | } |
1944 | ||
1945 | static struct vm_operations_struct kvm_vm_vm_ops = { | |
e4a533a4 | 1946 | .fault = kvm_vm_fault, |
f17abe9a AK |
1947 | }; |
1948 | ||
1949 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
1950 | { | |
1951 | vma->vm_ops = &kvm_vm_vm_ops; | |
1952 | return 0; | |
1953 | } | |
1954 | ||
3d3aab1b | 1955 | static struct file_operations kvm_vm_fops = { |
f17abe9a AK |
1956 | .release = kvm_vm_release, |
1957 | .unlocked_ioctl = kvm_vm_ioctl, | |
1958 | .compat_ioctl = kvm_vm_ioctl, | |
1959 | .mmap = kvm_vm_mmap, | |
1960 | }; | |
1961 | ||
1962 | static int kvm_dev_ioctl_create_vm(void) | |
1963 | { | |
2030a42c | 1964 | int fd; |
f17abe9a AK |
1965 | struct kvm *kvm; |
1966 | ||
f17abe9a | 1967 | kvm = kvm_create_vm(); |
d6d28168 AK |
1968 | if (IS_ERR(kvm)) |
1969 | return PTR_ERR(kvm); | |
7d9dbca3 | 1970 | fd = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, 0); |
2030a42c | 1971 | if (fd < 0) |
66c0b394 | 1972 | kvm_put_kvm(kvm); |
f17abe9a | 1973 | |
f17abe9a | 1974 | return fd; |
f17abe9a AK |
1975 | } |
1976 | ||
1a811b61 AK |
1977 | static long kvm_dev_ioctl_check_extension_generic(long arg) |
1978 | { | |
1979 | switch (arg) { | |
ca9edaee | 1980 | case KVM_CAP_USER_MEMORY: |
1a811b61 AK |
1981 | case KVM_CAP_DESTROY_MEMORY_REGION_WORKS: |
1982 | return 1; | |
1983 | default: | |
1984 | break; | |
1985 | } | |
1986 | return kvm_dev_ioctl_check_extension(arg); | |
1987 | } | |
1988 | ||
f17abe9a AK |
1989 | static long kvm_dev_ioctl(struct file *filp, |
1990 | unsigned int ioctl, unsigned long arg) | |
1991 | { | |
07c45a36 | 1992 | long r = -EINVAL; |
f17abe9a AK |
1993 | |
1994 | switch (ioctl) { | |
1995 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
1996 | r = -EINVAL; |
1997 | if (arg) | |
1998 | goto out; | |
f17abe9a AK |
1999 | r = KVM_API_VERSION; |
2000 | break; | |
2001 | case KVM_CREATE_VM: | |
f0fe5108 AK |
2002 | r = -EINVAL; |
2003 | if (arg) | |
2004 | goto out; | |
f17abe9a AK |
2005 | r = kvm_dev_ioctl_create_vm(); |
2006 | break; | |
018d00d2 | 2007 | case KVM_CHECK_EXTENSION: |
1a811b61 | 2008 | r = kvm_dev_ioctl_check_extension_generic(arg); |
5d308f45 | 2009 | break; |
07c45a36 AK |
2010 | case KVM_GET_VCPU_MMAP_SIZE: |
2011 | r = -EINVAL; | |
2012 | if (arg) | |
2013 | goto out; | |
adb1ff46 AK |
2014 | r = PAGE_SIZE; /* struct kvm_run */ |
2015 | #ifdef CONFIG_X86 | |
2016 | r += PAGE_SIZE; /* pio data page */ | |
5f94c174 LV |
2017 | #endif |
2018 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
2019 | r += PAGE_SIZE; /* coalesced mmio ring page */ | |
adb1ff46 | 2020 | #endif |
07c45a36 | 2021 | break; |
d4c9ff2d FEL |
2022 | case KVM_TRACE_ENABLE: |
2023 | case KVM_TRACE_PAUSE: | |
2024 | case KVM_TRACE_DISABLE: | |
2025 | r = kvm_trace_ioctl(ioctl, arg); | |
2026 | break; | |
6aa8b732 | 2027 | default: |
043405e1 | 2028 | return kvm_arch_dev_ioctl(filp, ioctl, arg); |
6aa8b732 AK |
2029 | } |
2030 | out: | |
2031 | return r; | |
2032 | } | |
2033 | ||
6aa8b732 | 2034 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
2035 | .unlocked_ioctl = kvm_dev_ioctl, |
2036 | .compat_ioctl = kvm_dev_ioctl, | |
6aa8b732 AK |
2037 | }; |
2038 | ||
2039 | static struct miscdevice kvm_dev = { | |
bbe4432e | 2040 | KVM_MINOR, |
6aa8b732 AK |
2041 | "kvm", |
2042 | &kvm_chardev_ops, | |
2043 | }; | |
2044 | ||
1b6c0168 AK |
2045 | static void hardware_enable(void *junk) |
2046 | { | |
2047 | int cpu = raw_smp_processor_id(); | |
2048 | ||
7f59f492 | 2049 | if (cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 2050 | return; |
7f59f492 | 2051 | cpumask_set_cpu(cpu, cpus_hardware_enabled); |
e9b11c17 | 2052 | kvm_arch_hardware_enable(NULL); |
1b6c0168 AK |
2053 | } |
2054 | ||
2055 | static void hardware_disable(void *junk) | |
2056 | { | |
2057 | int cpu = raw_smp_processor_id(); | |
2058 | ||
7f59f492 | 2059 | if (!cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 2060 | return; |
7f59f492 | 2061 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); |
e9b11c17 | 2062 | kvm_arch_hardware_disable(NULL); |
1b6c0168 AK |
2063 | } |
2064 | ||
774c47f1 AK |
2065 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
2066 | void *v) | |
2067 | { | |
2068 | int cpu = (long)v; | |
2069 | ||
1a6f4d7f | 2070 | val &= ~CPU_TASKS_FROZEN; |
774c47f1 | 2071 | switch (val) { |
cec9ad27 | 2072 | case CPU_DYING: |
6ec8a856 AK |
2073 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
2074 | cpu); | |
2075 | hardware_disable(NULL); | |
2076 | break; | |
774c47f1 | 2077 | case CPU_UP_CANCELED: |
43934a38 JK |
2078 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
2079 | cpu); | |
8691e5a8 | 2080 | smp_call_function_single(cpu, hardware_disable, NULL, 1); |
774c47f1 | 2081 | break; |
43934a38 JK |
2082 | case CPU_ONLINE: |
2083 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", | |
2084 | cpu); | |
8691e5a8 | 2085 | smp_call_function_single(cpu, hardware_enable, NULL, 1); |
774c47f1 AK |
2086 | break; |
2087 | } | |
2088 | return NOTIFY_OK; | |
2089 | } | |
2090 | ||
4ecac3fd AK |
2091 | |
2092 | asmlinkage void kvm_handle_fault_on_reboot(void) | |
2093 | { | |
2094 | if (kvm_rebooting) | |
2095 | /* spin while reset goes on */ | |
2096 | while (true) | |
2097 | ; | |
2098 | /* Fault while not rebooting. We want the trace. */ | |
2099 | BUG(); | |
2100 | } | |
2101 | EXPORT_SYMBOL_GPL(kvm_handle_fault_on_reboot); | |
2102 | ||
9a2b85c6 | 2103 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
d77c26fc | 2104 | void *v) |
9a2b85c6 RR |
2105 | { |
2106 | if (val == SYS_RESTART) { | |
2107 | /* | |
2108 | * Some (well, at least mine) BIOSes hang on reboot if | |
2109 | * in vmx root mode. | |
2110 | */ | |
2111 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
4ecac3fd | 2112 | kvm_rebooting = true; |
15c8b6c1 | 2113 | on_each_cpu(hardware_disable, NULL, 1); |
9a2b85c6 RR |
2114 | } |
2115 | return NOTIFY_OK; | |
2116 | } | |
2117 | ||
2118 | static struct notifier_block kvm_reboot_notifier = { | |
2119 | .notifier_call = kvm_reboot, | |
2120 | .priority = 0, | |
2121 | }; | |
2122 | ||
2eeb2e94 GH |
2123 | void kvm_io_bus_init(struct kvm_io_bus *bus) |
2124 | { | |
2125 | memset(bus, 0, sizeof(*bus)); | |
2126 | } | |
2127 | ||
2128 | void kvm_io_bus_destroy(struct kvm_io_bus *bus) | |
2129 | { | |
2130 | int i; | |
2131 | ||
2132 | for (i = 0; i < bus->dev_count; i++) { | |
2133 | struct kvm_io_device *pos = bus->devs[i]; | |
2134 | ||
2135 | kvm_iodevice_destructor(pos); | |
2136 | } | |
2137 | } | |
2138 | ||
92760499 LV |
2139 | struct kvm_io_device *kvm_io_bus_find_dev(struct kvm_io_bus *bus, |
2140 | gpa_t addr, int len, int is_write) | |
2eeb2e94 GH |
2141 | { |
2142 | int i; | |
2143 | ||
2144 | for (i = 0; i < bus->dev_count; i++) { | |
2145 | struct kvm_io_device *pos = bus->devs[i]; | |
2146 | ||
92760499 | 2147 | if (pos->in_range(pos, addr, len, is_write)) |
2eeb2e94 GH |
2148 | return pos; |
2149 | } | |
2150 | ||
2151 | return NULL; | |
2152 | } | |
2153 | ||
2154 | void kvm_io_bus_register_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev) | |
2155 | { | |
2156 | BUG_ON(bus->dev_count > (NR_IOBUS_DEVS-1)); | |
2157 | ||
2158 | bus->devs[bus->dev_count++] = dev; | |
2159 | } | |
2160 | ||
774c47f1 AK |
2161 | static struct notifier_block kvm_cpu_notifier = { |
2162 | .notifier_call = kvm_cpu_hotplug, | |
2163 | .priority = 20, /* must be > scheduler priority */ | |
2164 | }; | |
2165 | ||
8b88b099 | 2166 | static int vm_stat_get(void *_offset, u64 *val) |
ba1389b7 AK |
2167 | { |
2168 | unsigned offset = (long)_offset; | |
ba1389b7 AK |
2169 | struct kvm *kvm; |
2170 | ||
8b88b099 | 2171 | *val = 0; |
ba1389b7 AK |
2172 | spin_lock(&kvm_lock); |
2173 | list_for_each_entry(kvm, &vm_list, vm_list) | |
8b88b099 | 2174 | *val += *(u32 *)((void *)kvm + offset); |
ba1389b7 | 2175 | spin_unlock(&kvm_lock); |
8b88b099 | 2176 | return 0; |
ba1389b7 AK |
2177 | } |
2178 | ||
2179 | DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n"); | |
2180 | ||
8b88b099 | 2181 | static int vcpu_stat_get(void *_offset, u64 *val) |
1165f5fe AK |
2182 | { |
2183 | unsigned offset = (long)_offset; | |
1165f5fe AK |
2184 | struct kvm *kvm; |
2185 | struct kvm_vcpu *vcpu; | |
2186 | int i; | |
2187 | ||
8b88b099 | 2188 | *val = 0; |
1165f5fe AK |
2189 | spin_lock(&kvm_lock); |
2190 | list_for_each_entry(kvm, &vm_list, vm_list) | |
2191 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { | |
fb3f0f51 RR |
2192 | vcpu = kvm->vcpus[i]; |
2193 | if (vcpu) | |
8b88b099 | 2194 | *val += *(u32 *)((void *)vcpu + offset); |
1165f5fe AK |
2195 | } |
2196 | spin_unlock(&kvm_lock); | |
8b88b099 | 2197 | return 0; |
1165f5fe AK |
2198 | } |
2199 | ||
ba1389b7 AK |
2200 | DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n"); |
2201 | ||
2202 | static struct file_operations *stat_fops[] = { | |
2203 | [KVM_STAT_VCPU] = &vcpu_stat_fops, | |
2204 | [KVM_STAT_VM] = &vm_stat_fops, | |
2205 | }; | |
1165f5fe | 2206 | |
a16b043c | 2207 | static void kvm_init_debug(void) |
6aa8b732 AK |
2208 | { |
2209 | struct kvm_stats_debugfs_item *p; | |
2210 | ||
76f7c879 | 2211 | kvm_debugfs_dir = debugfs_create_dir("kvm", NULL); |
6aa8b732 | 2212 | for (p = debugfs_entries; p->name; ++p) |
76f7c879 | 2213 | p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir, |
1165f5fe | 2214 | (void *)(long)p->offset, |
ba1389b7 | 2215 | stat_fops[p->kind]); |
6aa8b732 AK |
2216 | } |
2217 | ||
2218 | static void kvm_exit_debug(void) | |
2219 | { | |
2220 | struct kvm_stats_debugfs_item *p; | |
2221 | ||
2222 | for (p = debugfs_entries; p->name; ++p) | |
2223 | debugfs_remove(p->dentry); | |
76f7c879 | 2224 | debugfs_remove(kvm_debugfs_dir); |
6aa8b732 AK |
2225 | } |
2226 | ||
59ae6c6b AK |
2227 | static int kvm_suspend(struct sys_device *dev, pm_message_t state) |
2228 | { | |
4267c41a | 2229 | hardware_disable(NULL); |
59ae6c6b AK |
2230 | return 0; |
2231 | } | |
2232 | ||
2233 | static int kvm_resume(struct sys_device *dev) | |
2234 | { | |
4267c41a | 2235 | hardware_enable(NULL); |
59ae6c6b AK |
2236 | return 0; |
2237 | } | |
2238 | ||
2239 | static struct sysdev_class kvm_sysdev_class = { | |
af5ca3f4 | 2240 | .name = "kvm", |
59ae6c6b AK |
2241 | .suspend = kvm_suspend, |
2242 | .resume = kvm_resume, | |
2243 | }; | |
2244 | ||
2245 | static struct sys_device kvm_sysdev = { | |
2246 | .id = 0, | |
2247 | .cls = &kvm_sysdev_class, | |
2248 | }; | |
2249 | ||
cea7bb21 | 2250 | struct page *bad_page; |
35149e21 | 2251 | pfn_t bad_pfn; |
6aa8b732 | 2252 | |
15ad7146 AK |
2253 | static inline |
2254 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
2255 | { | |
2256 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
2257 | } | |
2258 | ||
2259 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
2260 | { | |
2261 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2262 | ||
e9b11c17 | 2263 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 AK |
2264 | } |
2265 | ||
2266 | static void kvm_sched_out(struct preempt_notifier *pn, | |
2267 | struct task_struct *next) | |
2268 | { | |
2269 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2270 | ||
e9b11c17 | 2271 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
2272 | } |
2273 | ||
f8c16bba | 2274 | int kvm_init(void *opaque, unsigned int vcpu_size, |
c16f862d | 2275 | struct module *module) |
6aa8b732 AK |
2276 | { |
2277 | int r; | |
002c7f7c | 2278 | int cpu; |
6aa8b732 | 2279 | |
cb498ea2 ZX |
2280 | kvm_init_debug(); |
2281 | ||
f8c16bba ZX |
2282 | r = kvm_arch_init(opaque); |
2283 | if (r) | |
d2308784 | 2284 | goto out_fail; |
cb498ea2 ZX |
2285 | |
2286 | bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
2287 | ||
2288 | if (bad_page == NULL) { | |
2289 | r = -ENOMEM; | |
2290 | goto out; | |
2291 | } | |
2292 | ||
35149e21 AL |
2293 | bad_pfn = page_to_pfn(bad_page); |
2294 | ||
7f59f492 RR |
2295 | if (!alloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) { |
2296 | r = -ENOMEM; | |
2297 | goto out_free_0; | |
2298 | } | |
2299 | ||
e9b11c17 | 2300 | r = kvm_arch_hardware_setup(); |
6aa8b732 | 2301 | if (r < 0) |
7f59f492 | 2302 | goto out_free_0a; |
6aa8b732 | 2303 | |
002c7f7c YS |
2304 | for_each_online_cpu(cpu) { |
2305 | smp_call_function_single(cpu, | |
e9b11c17 | 2306 | kvm_arch_check_processor_compat, |
8691e5a8 | 2307 | &r, 1); |
002c7f7c | 2308 | if (r < 0) |
d2308784 | 2309 | goto out_free_1; |
002c7f7c YS |
2310 | } |
2311 | ||
15c8b6c1 | 2312 | on_each_cpu(hardware_enable, NULL, 1); |
774c47f1 AK |
2313 | r = register_cpu_notifier(&kvm_cpu_notifier); |
2314 | if (r) | |
d2308784 | 2315 | goto out_free_2; |
6aa8b732 AK |
2316 | register_reboot_notifier(&kvm_reboot_notifier); |
2317 | ||
59ae6c6b AK |
2318 | r = sysdev_class_register(&kvm_sysdev_class); |
2319 | if (r) | |
d2308784 | 2320 | goto out_free_3; |
59ae6c6b AK |
2321 | |
2322 | r = sysdev_register(&kvm_sysdev); | |
2323 | if (r) | |
d2308784 | 2324 | goto out_free_4; |
59ae6c6b | 2325 | |
c16f862d RR |
2326 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
2327 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, | |
56919c5c JP |
2328 | __alignof__(struct kvm_vcpu), |
2329 | 0, NULL); | |
c16f862d RR |
2330 | if (!kvm_vcpu_cache) { |
2331 | r = -ENOMEM; | |
d2308784 | 2332 | goto out_free_5; |
c16f862d RR |
2333 | } |
2334 | ||
6aa8b732 | 2335 | kvm_chardev_ops.owner = module; |
3d3aab1b CB |
2336 | kvm_vm_fops.owner = module; |
2337 | kvm_vcpu_fops.owner = module; | |
6aa8b732 AK |
2338 | |
2339 | r = misc_register(&kvm_dev); | |
2340 | if (r) { | |
d77c26fc | 2341 | printk(KERN_ERR "kvm: misc device register failed\n"); |
6aa8b732 AK |
2342 | goto out_free; |
2343 | } | |
2344 | ||
15ad7146 AK |
2345 | kvm_preempt_ops.sched_in = kvm_sched_in; |
2346 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
5319c662 SY |
2347 | #ifndef CONFIG_X86 |
2348 | msi2intx = 0; | |
2349 | #endif | |
15ad7146 | 2350 | |
c7addb90 | 2351 | return 0; |
6aa8b732 AK |
2352 | |
2353 | out_free: | |
c16f862d | 2354 | kmem_cache_destroy(kvm_vcpu_cache); |
d2308784 | 2355 | out_free_5: |
59ae6c6b | 2356 | sysdev_unregister(&kvm_sysdev); |
d2308784 | 2357 | out_free_4: |
59ae6c6b | 2358 | sysdev_class_unregister(&kvm_sysdev_class); |
d2308784 | 2359 | out_free_3: |
6aa8b732 | 2360 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 | 2361 | unregister_cpu_notifier(&kvm_cpu_notifier); |
d2308784 | 2362 | out_free_2: |
15c8b6c1 | 2363 | on_each_cpu(hardware_disable, NULL, 1); |
d2308784 | 2364 | out_free_1: |
e9b11c17 | 2365 | kvm_arch_hardware_unsetup(); |
7f59f492 RR |
2366 | out_free_0a: |
2367 | free_cpumask_var(cpus_hardware_enabled); | |
d2308784 ZX |
2368 | out_free_0: |
2369 | __free_page(bad_page); | |
ca45aaae | 2370 | out: |
f8c16bba | 2371 | kvm_arch_exit(); |
cb498ea2 | 2372 | kvm_exit_debug(); |
d2308784 | 2373 | out_fail: |
6aa8b732 AK |
2374 | return r; |
2375 | } | |
cb498ea2 | 2376 | EXPORT_SYMBOL_GPL(kvm_init); |
6aa8b732 | 2377 | |
cb498ea2 | 2378 | void kvm_exit(void) |
6aa8b732 | 2379 | { |
d4c9ff2d | 2380 | kvm_trace_cleanup(); |
6aa8b732 | 2381 | misc_deregister(&kvm_dev); |
c16f862d | 2382 | kmem_cache_destroy(kvm_vcpu_cache); |
59ae6c6b AK |
2383 | sysdev_unregister(&kvm_sysdev); |
2384 | sysdev_class_unregister(&kvm_sysdev_class); | |
6aa8b732 | 2385 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 2386 | unregister_cpu_notifier(&kvm_cpu_notifier); |
15c8b6c1 | 2387 | on_each_cpu(hardware_disable, NULL, 1); |
e9b11c17 | 2388 | kvm_arch_hardware_unsetup(); |
f8c16bba | 2389 | kvm_arch_exit(); |
6aa8b732 | 2390 | kvm_exit_debug(); |
7f59f492 | 2391 | free_cpumask_var(cpus_hardware_enabled); |
cea7bb21 | 2392 | __free_page(bad_page); |
6aa8b732 | 2393 | } |
cb498ea2 | 2394 | EXPORT_SYMBOL_GPL(kvm_exit); |