]>
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> |
e56d532f | 44 | #include <linux/bitops.h> |
547de29e | 45 | #include <linux/spinlock.h> |
6aa8b732 | 46 | |
e495606d | 47 | #include <asm/processor.h> |
e495606d AK |
48 | #include <asm/io.h> |
49 | #include <asm/uaccess.h> | |
3e021bf5 | 50 | #include <asm/pgtable.h> |
6aa8b732 | 51 | |
5f94c174 LV |
52 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
53 | #include "coalesced_mmio.h" | |
54 | #endif | |
55 | ||
8a98f664 XZ |
56 | #ifdef KVM_CAP_DEVICE_ASSIGNMENT |
57 | #include <linux/pci.h> | |
58 | #include <linux/interrupt.h> | |
59 | #include "irq.h" | |
60 | #endif | |
61 | ||
6aa8b732 AK |
62 | MODULE_AUTHOR("Qumranet"); |
63 | MODULE_LICENSE("GPL"); | |
64 | ||
e9b11c17 ZX |
65 | DEFINE_SPINLOCK(kvm_lock); |
66 | LIST_HEAD(vm_list); | |
133de902 | 67 | |
7f59f492 | 68 | static cpumask_var_t cpus_hardware_enabled; |
1b6c0168 | 69 | |
c16f862d RR |
70 | struct kmem_cache *kvm_vcpu_cache; |
71 | EXPORT_SYMBOL_GPL(kvm_vcpu_cache); | |
1165f5fe | 72 | |
15ad7146 AK |
73 | static __read_mostly struct preempt_ops kvm_preempt_ops; |
74 | ||
76f7c879 | 75 | struct dentry *kvm_debugfs_dir; |
6aa8b732 | 76 | |
bccf2150 AK |
77 | static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl, |
78 | unsigned long arg); | |
79 | ||
e8ba5d31 | 80 | static bool kvm_rebooting; |
4ecac3fd | 81 | |
8a98f664 XZ |
82 | #ifdef KVM_CAP_DEVICE_ASSIGNMENT |
83 | static struct kvm_assigned_dev_kernel *kvm_find_assigned_dev(struct list_head *head, | |
84 | int assigned_dev_id) | |
85 | { | |
86 | struct list_head *ptr; | |
87 | struct kvm_assigned_dev_kernel *match; | |
88 | ||
89 | list_for_each(ptr, head) { | |
90 | match = list_entry(ptr, struct kvm_assigned_dev_kernel, list); | |
91 | if (match->assigned_dev_id == assigned_dev_id) | |
92 | return match; | |
93 | } | |
94 | return NULL; | |
95 | } | |
96 | ||
2350bd1f SY |
97 | static int find_index_from_host_irq(struct kvm_assigned_dev_kernel |
98 | *assigned_dev, int irq) | |
99 | { | |
100 | int i, index; | |
101 | struct msix_entry *host_msix_entries; | |
102 | ||
103 | host_msix_entries = assigned_dev->host_msix_entries; | |
104 | ||
105 | index = -1; | |
106 | for (i = 0; i < assigned_dev->entries_nr; i++) | |
107 | if (irq == host_msix_entries[i].vector) { | |
108 | index = i; | |
109 | break; | |
110 | } | |
111 | if (index < 0) { | |
112 | printk(KERN_WARNING "Fail to find correlated MSI-X entry!\n"); | |
113 | return 0; | |
114 | } | |
115 | ||
116 | return index; | |
117 | } | |
118 | ||
8a98f664 XZ |
119 | static void kvm_assigned_dev_interrupt_work_handler(struct work_struct *work) |
120 | { | |
121 | struct kvm_assigned_dev_kernel *assigned_dev; | |
2350bd1f SY |
122 | struct kvm *kvm; |
123 | int irq, i; | |
8a98f664 XZ |
124 | |
125 | assigned_dev = container_of(work, struct kvm_assigned_dev_kernel, | |
126 | interrupt_work); | |
2350bd1f | 127 | kvm = assigned_dev->kvm; |
8a98f664 XZ |
128 | |
129 | /* This is taken to safely inject irq inside the guest. When | |
130 | * the interrupt injection (or the ioapic code) uses a | |
131 | * finer-grained lock, update this | |
132 | */ | |
2350bd1f | 133 | mutex_lock(&kvm->lock); |
547de29e | 134 | spin_lock_irq(&assigned_dev->assigned_dev_lock); |
e56d532f | 135 | if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_HOST_MSIX) { |
2350bd1f SY |
136 | struct kvm_guest_msix_entry *guest_entries = |
137 | assigned_dev->guest_msix_entries; | |
138 | for (i = 0; i < assigned_dev->entries_nr; i++) { | |
139 | if (!(guest_entries[i].flags & | |
140 | KVM_ASSIGNED_MSIX_PENDING)) | |
141 | continue; | |
142 | guest_entries[i].flags &= ~KVM_ASSIGNED_MSIX_PENDING; | |
143 | kvm_set_irq(assigned_dev->kvm, | |
144 | assigned_dev->irq_source_id, | |
145 | guest_entries[i].vector, 1); | |
146 | irq = assigned_dev->host_msix_entries[i].vector; | |
147 | if (irq != 0) | |
148 | enable_irq(irq); | |
149 | assigned_dev->host_irq_disabled = false; | |
150 | } | |
151 | } else { | |
152 | kvm_set_irq(assigned_dev->kvm, assigned_dev->irq_source_id, | |
153 | assigned_dev->guest_irq, 1); | |
154 | if (assigned_dev->irq_requested_type & | |
e56d532f | 155 | KVM_DEV_IRQ_GUEST_MSI) { |
2350bd1f SY |
156 | enable_irq(assigned_dev->host_irq); |
157 | assigned_dev->host_irq_disabled = false; | |
158 | } | |
6b9cc7fd | 159 | } |
2350bd1f | 160 | |
547de29e | 161 | spin_unlock_irq(&assigned_dev->assigned_dev_lock); |
8a98f664 | 162 | mutex_unlock(&assigned_dev->kvm->lock); |
8a98f664 XZ |
163 | } |
164 | ||
8a98f664 XZ |
165 | static irqreturn_t kvm_assigned_dev_intr(int irq, void *dev_id) |
166 | { | |
547de29e | 167 | unsigned long flags; |
8a98f664 XZ |
168 | struct kvm_assigned_dev_kernel *assigned_dev = |
169 | (struct kvm_assigned_dev_kernel *) dev_id; | |
170 | ||
547de29e | 171 | spin_lock_irqsave(&assigned_dev->assigned_dev_lock, flags); |
e56d532f | 172 | if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_HOST_MSIX) { |
2350bd1f SY |
173 | int index = find_index_from_host_irq(assigned_dev, irq); |
174 | if (index < 0) | |
547de29e | 175 | goto out; |
2350bd1f SY |
176 | assigned_dev->guest_msix_entries[index].flags |= |
177 | KVM_ASSIGNED_MSIX_PENDING; | |
178 | } | |
179 | ||
8a98f664 | 180 | schedule_work(&assigned_dev->interrupt_work); |
defaf158 | 181 | |
8a98f664 | 182 | disable_irq_nosync(irq); |
defaf158 MM |
183 | assigned_dev->host_irq_disabled = true; |
184 | ||
547de29e MT |
185 | out: |
186 | spin_unlock_irqrestore(&assigned_dev->assigned_dev_lock, flags); | |
8a98f664 XZ |
187 | return IRQ_HANDLED; |
188 | } | |
189 | ||
190 | /* Ack the irq line for an assigned device */ | |
191 | static void kvm_assigned_dev_ack_irq(struct kvm_irq_ack_notifier *kian) | |
192 | { | |
193 | struct kvm_assigned_dev_kernel *dev; | |
547de29e | 194 | unsigned long flags; |
8a98f664 XZ |
195 | |
196 | if (kian->gsi == -1) | |
197 | return; | |
198 | ||
199 | dev = container_of(kian, struct kvm_assigned_dev_kernel, | |
200 | ack_notifier); | |
defaf158 | 201 | |
5550af4d | 202 | kvm_set_irq(dev->kvm, dev->irq_source_id, dev->guest_irq, 0); |
defaf158 MM |
203 | |
204 | /* The guest irq may be shared so this ack may be | |
205 | * from another device. | |
206 | */ | |
547de29e | 207 | spin_lock_irqsave(&dev->assigned_dev_lock, flags); |
defaf158 MM |
208 | if (dev->host_irq_disabled) { |
209 | enable_irq(dev->host_irq); | |
210 | dev->host_irq_disabled = false; | |
211 | } | |
547de29e | 212 | spin_unlock_irqrestore(&dev->assigned_dev_lock, flags); |
8a98f664 XZ |
213 | } |
214 | ||
e56d532f SY |
215 | static void deassign_guest_irq(struct kvm *kvm, |
216 | struct kvm_assigned_dev_kernel *assigned_dev) | |
8a98f664 | 217 | { |
e19e30ef | 218 | kvm_unregister_irq_ack_notifier(&assigned_dev->ack_notifier); |
e56d532f | 219 | assigned_dev->ack_notifier.gsi = -1; |
f29b2673 MM |
220 | |
221 | if (assigned_dev->irq_source_id != -1) | |
222 | kvm_free_irq_source_id(kvm, assigned_dev->irq_source_id); | |
223 | assigned_dev->irq_source_id = -1; | |
e56d532f SY |
224 | assigned_dev->irq_requested_type &= ~(KVM_DEV_IRQ_GUEST_MASK); |
225 | } | |
8a98f664 | 226 | |
e56d532f SY |
227 | /* The function implicit hold kvm->lock mutex due to cancel_work_sync() */ |
228 | static void deassign_host_irq(struct kvm *kvm, | |
229 | struct kvm_assigned_dev_kernel *assigned_dev) | |
230 | { | |
ba4cef31 SY |
231 | /* |
232 | * In kvm_free_device_irq, cancel_work_sync return true if: | |
233 | * 1. work is scheduled, and then cancelled. | |
234 | * 2. work callback is executed. | |
235 | * | |
236 | * The first one ensured that the irq is disabled and no more events | |
237 | * would happen. But for the second one, the irq may be enabled (e.g. | |
238 | * for MSI). So we disable irq here to prevent further events. | |
239 | * | |
240 | * Notice this maybe result in nested disable if the interrupt type is | |
241 | * INTx, but it's OK for we are going to free it. | |
242 | * | |
243 | * If this function is a part of VM destroy, please ensure that till | |
244 | * now, the kvm state is still legal for probably we also have to wait | |
245 | * interrupt_work done. | |
246 | */ | |
e56d532f | 247 | if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_HOST_MSIX) { |
d510d6cc SY |
248 | int i; |
249 | for (i = 0; i < assigned_dev->entries_nr; i++) | |
250 | disable_irq_nosync(assigned_dev-> | |
251 | host_msix_entries[i].vector); | |
252 | ||
253 | cancel_work_sync(&assigned_dev->interrupt_work); | |
254 | ||
255 | for (i = 0; i < assigned_dev->entries_nr; i++) | |
256 | free_irq(assigned_dev->host_msix_entries[i].vector, | |
257 | (void *)assigned_dev); | |
258 | ||
259 | assigned_dev->entries_nr = 0; | |
260 | kfree(assigned_dev->host_msix_entries); | |
261 | kfree(assigned_dev->guest_msix_entries); | |
262 | pci_disable_msix(assigned_dev->dev); | |
263 | } else { | |
264 | /* Deal with MSI and INTx */ | |
265 | disable_irq_nosync(assigned_dev->host_irq); | |
266 | cancel_work_sync(&assigned_dev->interrupt_work); | |
8a98f664 | 267 | |
d510d6cc | 268 | free_irq(assigned_dev->host_irq, (void *)assigned_dev); |
4a643be8 | 269 | |
e56d532f | 270 | if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_HOST_MSI) |
d510d6cc SY |
271 | pci_disable_msi(assigned_dev->dev); |
272 | } | |
4a643be8 | 273 | |
e56d532f SY |
274 | assigned_dev->irq_requested_type &= ~(KVM_DEV_IRQ_HOST_MASK); |
275 | } | |
276 | ||
277 | static int kvm_deassign_irq(struct kvm *kvm, | |
278 | struct kvm_assigned_dev_kernel *assigned_dev, | |
279 | unsigned long irq_requested_type) | |
280 | { | |
281 | unsigned long guest_irq_type, host_irq_type; | |
282 | ||
283 | if (!irqchip_in_kernel(kvm)) | |
284 | return -EINVAL; | |
285 | /* no irq assignment to deassign */ | |
286 | if (!assigned_dev->irq_requested_type) | |
287 | return -ENXIO; | |
288 | ||
289 | host_irq_type = irq_requested_type & KVM_DEV_IRQ_HOST_MASK; | |
290 | guest_irq_type = irq_requested_type & KVM_DEV_IRQ_GUEST_MASK; | |
291 | ||
292 | if (host_irq_type) | |
293 | deassign_host_irq(kvm, assigned_dev); | |
294 | if (guest_irq_type) | |
295 | deassign_guest_irq(kvm, assigned_dev); | |
296 | ||
297 | return 0; | |
4a643be8 MM |
298 | } |
299 | ||
e56d532f SY |
300 | static void kvm_free_assigned_irq(struct kvm *kvm, |
301 | struct kvm_assigned_dev_kernel *assigned_dev) | |
302 | { | |
303 | kvm_deassign_irq(kvm, assigned_dev, assigned_dev->irq_requested_type); | |
304 | } | |
4a643be8 MM |
305 | |
306 | static void kvm_free_assigned_device(struct kvm *kvm, | |
307 | struct kvm_assigned_dev_kernel | |
308 | *assigned_dev) | |
309 | { | |
310 | kvm_free_assigned_irq(kvm, assigned_dev); | |
311 | ||
6eb55818 SY |
312 | pci_reset_function(assigned_dev->dev); |
313 | ||
8a98f664 XZ |
314 | pci_release_regions(assigned_dev->dev); |
315 | pci_disable_device(assigned_dev->dev); | |
316 | pci_dev_put(assigned_dev->dev); | |
317 | ||
318 | list_del(&assigned_dev->list); | |
319 | kfree(assigned_dev); | |
320 | } | |
321 | ||
322 | void kvm_free_all_assigned_devices(struct kvm *kvm) | |
323 | { | |
324 | struct list_head *ptr, *ptr2; | |
325 | struct kvm_assigned_dev_kernel *assigned_dev; | |
326 | ||
327 | list_for_each_safe(ptr, ptr2, &kvm->arch.assigned_dev_head) { | |
328 | assigned_dev = list_entry(ptr, | |
329 | struct kvm_assigned_dev_kernel, | |
330 | list); | |
331 | ||
332 | kvm_free_assigned_device(kvm, assigned_dev); | |
333 | } | |
334 | } | |
335 | ||
e56d532f SY |
336 | static int assigned_device_enable_host_intx(struct kvm *kvm, |
337 | struct kvm_assigned_dev_kernel *dev) | |
00e3ed39 | 338 | { |
e56d532f SY |
339 | dev->host_irq = dev->dev->irq; |
340 | /* Even though this is PCI, we don't want to use shared | |
341 | * interrupts. Sharing host devices with guest-assigned devices | |
342 | * on the same interrupt line is not a happy situation: there | |
343 | * are going to be long delays in accepting, acking, etc. | |
344 | */ | |
345 | if (request_irq(dev->host_irq, kvm_assigned_dev_intr, | |
346 | 0, "kvm_assigned_intx_device", (void *)dev)) | |
347 | return -EIO; | |
348 | return 0; | |
349 | } | |
6b9cc7fd | 350 | |
e56d532f SY |
351 | #ifdef __KVM_HAVE_MSI |
352 | static int assigned_device_enable_host_msi(struct kvm *kvm, | |
353 | struct kvm_assigned_dev_kernel *dev) | |
354 | { | |
355 | int r; | |
00e3ed39 | 356 | |
e56d532f SY |
357 | if (!dev->dev->msi_enabled) { |
358 | r = pci_enable_msi(dev->dev); | |
359 | if (r) | |
360 | return r; | |
361 | } | |
00e3ed39 | 362 | |
e56d532f SY |
363 | dev->host_irq = dev->dev->irq; |
364 | if (request_irq(dev->host_irq, kvm_assigned_dev_intr, 0, | |
365 | "kvm_assigned_msi_device", (void *)dev)) { | |
366 | pci_disable_msi(dev->dev); | |
367 | return -EIO; | |
00e3ed39 SY |
368 | } |
369 | ||
00e3ed39 SY |
370 | return 0; |
371 | } | |
e56d532f | 372 | #endif |
00e3ed39 | 373 | |
e56d532f SY |
374 | #ifdef __KVM_HAVE_MSIX |
375 | static int assigned_device_enable_host_msix(struct kvm *kvm, | |
376 | struct kvm_assigned_dev_kernel *dev) | |
6b9cc7fd | 377 | { |
e56d532f | 378 | int i, r = -EINVAL; |
6b9cc7fd | 379 | |
e56d532f SY |
380 | /* host_msix_entries and guest_msix_entries should have been |
381 | * initialized */ | |
382 | if (dev->entries_nr == 0) | |
383 | return r; | |
6b9cc7fd | 384 | |
e56d532f SY |
385 | r = pci_enable_msix(dev->dev, dev->host_msix_entries, dev->entries_nr); |
386 | if (r) | |
387 | return r; | |
6b9cc7fd | 388 | |
e56d532f SY |
389 | for (i = 0; i < dev->entries_nr; i++) { |
390 | r = request_irq(dev->host_msix_entries[i].vector, | |
391 | kvm_assigned_dev_intr, 0, | |
392 | "kvm_assigned_msix_device", | |
393 | (void *)dev); | |
394 | /* FIXME: free requested_irq's on failure */ | |
395 | if (r) | |
396 | return r; | |
397 | } | |
5319c662 | 398 | |
e56d532f SY |
399 | return 0; |
400 | } | |
6b9cc7fd | 401 | |
e56d532f | 402 | #endif |
6b9cc7fd | 403 | |
e56d532f SY |
404 | static int assigned_device_enable_guest_intx(struct kvm *kvm, |
405 | struct kvm_assigned_dev_kernel *dev, | |
406 | struct kvm_assigned_irq *irq) | |
407 | { | |
408 | dev->guest_irq = irq->guest_irq; | |
409 | dev->ack_notifier.gsi = irq->guest_irq; | |
410 | return 0; | |
411 | } | |
5319c662 | 412 | |
e56d532f SY |
413 | #ifdef __KVM_HAVE_MSI |
414 | static int assigned_device_enable_guest_msi(struct kvm *kvm, | |
415 | struct kvm_assigned_dev_kernel *dev, | |
416 | struct kvm_assigned_irq *irq) | |
417 | { | |
418 | dev->guest_irq = irq->guest_irq; | |
419 | dev->ack_notifier.gsi = -1; | |
6b9cc7fd SY |
420 | return 0; |
421 | } | |
422 | #endif | |
e56d532f SY |
423 | #ifdef __KVM_HAVE_MSIX |
424 | static int assigned_device_enable_guest_msix(struct kvm *kvm, | |
425 | struct kvm_assigned_dev_kernel *dev, | |
426 | struct kvm_assigned_irq *irq) | |
427 | { | |
428 | dev->guest_irq = irq->guest_irq; | |
429 | dev->ack_notifier.gsi = -1; | |
430 | return 0; | |
431 | } | |
432 | #endif | |
433 | ||
434 | static int assign_host_irq(struct kvm *kvm, | |
435 | struct kvm_assigned_dev_kernel *dev, | |
436 | __u32 host_irq_type) | |
437 | { | |
438 | int r = -EEXIST; | |
439 | ||
440 | if (dev->irq_requested_type & KVM_DEV_IRQ_HOST_MASK) | |
441 | return r; | |
6b9cc7fd | 442 | |
e56d532f SY |
443 | switch (host_irq_type) { |
444 | case KVM_DEV_IRQ_HOST_INTX: | |
445 | r = assigned_device_enable_host_intx(kvm, dev); | |
446 | break; | |
447 | #ifdef __KVM_HAVE_MSI | |
448 | case KVM_DEV_IRQ_HOST_MSI: | |
449 | r = assigned_device_enable_host_msi(kvm, dev); | |
450 | break; | |
451 | #endif | |
d510d6cc | 452 | #ifdef __KVM_HAVE_MSIX |
e56d532f SY |
453 | case KVM_DEV_IRQ_HOST_MSIX: |
454 | r = assigned_device_enable_host_msix(kvm, dev); | |
455 | break; | |
456 | #endif | |
457 | default: | |
458 | r = -EINVAL; | |
459 | } | |
d510d6cc | 460 | |
e56d532f SY |
461 | if (!r) |
462 | dev->irq_requested_type |= host_irq_type; | |
d510d6cc | 463 | |
e56d532f SY |
464 | return r; |
465 | } | |
d510d6cc | 466 | |
e56d532f SY |
467 | static int assign_guest_irq(struct kvm *kvm, |
468 | struct kvm_assigned_dev_kernel *dev, | |
469 | struct kvm_assigned_irq *irq, | |
470 | unsigned long guest_irq_type) | |
471 | { | |
472 | int id; | |
473 | int r = -EEXIST; | |
d510d6cc | 474 | |
e56d532f SY |
475 | if (dev->irq_requested_type & KVM_DEV_IRQ_GUEST_MASK) |
476 | return r; | |
477 | ||
478 | id = kvm_request_irq_source_id(kvm); | |
479 | if (id < 0) | |
480 | return id; | |
481 | ||
482 | dev->irq_source_id = id; | |
483 | ||
484 | switch (guest_irq_type) { | |
485 | case KVM_DEV_IRQ_GUEST_INTX: | |
486 | r = assigned_device_enable_guest_intx(kvm, dev, irq); | |
487 | break; | |
488 | #ifdef __KVM_HAVE_MSI | |
489 | case KVM_DEV_IRQ_GUEST_MSI: | |
490 | r = assigned_device_enable_guest_msi(kvm, dev, irq); | |
491 | break; | |
492 | #endif | |
493 | #ifdef __KVM_HAVE_MSIX | |
494 | case KVM_DEV_IRQ_GUEST_MSIX: | |
495 | r = assigned_device_enable_guest_msix(kvm, dev, irq); | |
496 | break; | |
497 | #endif | |
498 | default: | |
499 | r = -EINVAL; | |
d510d6cc SY |
500 | } |
501 | ||
e56d532f SY |
502 | if (!r) { |
503 | dev->irq_requested_type |= guest_irq_type; | |
504 | kvm_register_irq_ack_notifier(kvm, &dev->ack_notifier); | |
505 | } else | |
506 | kvm_free_irq_source_id(kvm, dev->irq_source_id); | |
d510d6cc | 507 | |
e56d532f | 508 | return r; |
d510d6cc | 509 | } |
d510d6cc | 510 | |
e56d532f | 511 | /* TODO Deal with KVM_DEV_IRQ_ASSIGNED_MASK_MSIX */ |
8a98f664 | 512 | static int kvm_vm_ioctl_assign_irq(struct kvm *kvm, |
e56d532f | 513 | struct kvm_assigned_irq *assigned_irq) |
8a98f664 | 514 | { |
e56d532f | 515 | int r = -EINVAL; |
8a98f664 | 516 | struct kvm_assigned_dev_kernel *match; |
e56d532f | 517 | unsigned long host_irq_type, guest_irq_type; |
8a98f664 | 518 | |
e56d532f SY |
519 | if (!capable(CAP_SYS_RAWIO)) |
520 | return -EPERM; | |
8a98f664 | 521 | |
e56d532f SY |
522 | if (!irqchip_in_kernel(kvm)) |
523 | return r; | |
524 | ||
525 | mutex_lock(&kvm->lock); | |
526 | r = -ENODEV; | |
8a98f664 XZ |
527 | match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, |
528 | assigned_irq->assigned_dev_id); | |
e56d532f SY |
529 | if (!match) |
530 | goto out; | |
8a98f664 | 531 | |
e56d532f SY |
532 | host_irq_type = (assigned_irq->flags & KVM_DEV_IRQ_HOST_MASK); |
533 | guest_irq_type = (assigned_irq->flags & KVM_DEV_IRQ_GUEST_MASK); | |
17071fe7 | 534 | |
e56d532f SY |
535 | r = -EINVAL; |
536 | /* can only assign one type at a time */ | |
537 | if (hweight_long(host_irq_type) > 1) | |
538 | goto out; | |
539 | if (hweight_long(guest_irq_type) > 1) | |
540 | goto out; | |
541 | if (host_irq_type == 0 && guest_irq_type == 0) | |
542 | goto out; | |
17071fe7 | 543 | |
e56d532f SY |
544 | r = 0; |
545 | if (host_irq_type) | |
546 | r = assign_host_irq(kvm, match, host_irq_type); | |
547 | if (r) | |
548 | goto out; | |
8a98f664 | 549 | |
e56d532f SY |
550 | if (guest_irq_type) |
551 | r = assign_guest_irq(kvm, match, assigned_irq, guest_irq_type); | |
552 | out: | |
8a98f664 XZ |
553 | mutex_unlock(&kvm->lock); |
554 | return r; | |
e56d532f SY |
555 | } |
556 | ||
557 | static int kvm_vm_ioctl_deassign_dev_irq(struct kvm *kvm, | |
558 | struct kvm_assigned_irq | |
559 | *assigned_irq) | |
560 | { | |
561 | int r = -ENODEV; | |
562 | struct kvm_assigned_dev_kernel *match; | |
563 | ||
564 | mutex_lock(&kvm->lock); | |
565 | ||
566 | match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, | |
567 | assigned_irq->assigned_dev_id); | |
568 | if (!match) | |
569 | goto out; | |
570 | ||
571 | r = kvm_deassign_irq(kvm, match, assigned_irq->flags); | |
572 | out: | |
8a98f664 | 573 | mutex_unlock(&kvm->lock); |
8a98f664 XZ |
574 | return r; |
575 | } | |
576 | ||
577 | static int kvm_vm_ioctl_assign_device(struct kvm *kvm, | |
578 | struct kvm_assigned_pci_dev *assigned_dev) | |
579 | { | |
580 | int r = 0; | |
581 | struct kvm_assigned_dev_kernel *match; | |
582 | struct pci_dev *dev; | |
583 | ||
682edb4c | 584 | down_read(&kvm->slots_lock); |
8a98f664 XZ |
585 | mutex_lock(&kvm->lock); |
586 | ||
587 | match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, | |
588 | assigned_dev->assigned_dev_id); | |
589 | if (match) { | |
590 | /* device already assigned */ | |
e56d532f | 591 | r = -EEXIST; |
8a98f664 XZ |
592 | goto out; |
593 | } | |
594 | ||
595 | match = kzalloc(sizeof(struct kvm_assigned_dev_kernel), GFP_KERNEL); | |
596 | if (match == NULL) { | |
597 | printk(KERN_INFO "%s: Couldn't allocate memory\n", | |
598 | __func__); | |
599 | r = -ENOMEM; | |
600 | goto out; | |
601 | } | |
602 | dev = pci_get_bus_and_slot(assigned_dev->busnr, | |
603 | assigned_dev->devfn); | |
604 | if (!dev) { | |
605 | printk(KERN_INFO "%s: host device not found\n", __func__); | |
606 | r = -EINVAL; | |
607 | goto out_free; | |
608 | } | |
609 | if (pci_enable_device(dev)) { | |
610 | printk(KERN_INFO "%s: Could not enable PCI device\n", __func__); | |
611 | r = -EBUSY; | |
612 | goto out_put; | |
613 | } | |
614 | r = pci_request_regions(dev, "kvm_assigned_device"); | |
615 | if (r) { | |
616 | printk(KERN_INFO "%s: Could not get access to device regions\n", | |
617 | __func__); | |
618 | goto out_disable; | |
619 | } | |
6eb55818 SY |
620 | |
621 | pci_reset_function(dev); | |
622 | ||
8a98f664 XZ |
623 | match->assigned_dev_id = assigned_dev->assigned_dev_id; |
624 | match->host_busnr = assigned_dev->busnr; | |
625 | match->host_devfn = assigned_dev->devfn; | |
b653574a | 626 | match->flags = assigned_dev->flags; |
8a98f664 | 627 | match->dev = dev; |
547de29e | 628 | spin_lock_init(&match->assigned_dev_lock); |
f29b2673 | 629 | match->irq_source_id = -1; |
8a98f664 | 630 | match->kvm = kvm; |
e56d532f SY |
631 | match->ack_notifier.irq_acked = kvm_assigned_dev_ack_irq; |
632 | INIT_WORK(&match->interrupt_work, | |
633 | kvm_assigned_dev_interrupt_work_handler); | |
8a98f664 XZ |
634 | |
635 | list_add(&match->list, &kvm->arch.assigned_dev_head); | |
636 | ||
637 | if (assigned_dev->flags & KVM_DEV_ASSIGN_ENABLE_IOMMU) { | |
19de40a8 | 638 | if (!kvm->arch.iommu_domain) { |
260782bc WH |
639 | r = kvm_iommu_map_guest(kvm); |
640 | if (r) | |
641 | goto out_list_del; | |
642 | } | |
643 | r = kvm_assign_device(kvm, match); | |
8a98f664 XZ |
644 | if (r) |
645 | goto out_list_del; | |
646 | } | |
647 | ||
648 | out: | |
649 | mutex_unlock(&kvm->lock); | |
682edb4c | 650 | up_read(&kvm->slots_lock); |
8a98f664 XZ |
651 | return r; |
652 | out_list_del: | |
653 | list_del(&match->list); | |
654 | pci_release_regions(dev); | |
655 | out_disable: | |
656 | pci_disable_device(dev); | |
657 | out_put: | |
658 | pci_dev_put(dev); | |
659 | out_free: | |
660 | kfree(match); | |
661 | mutex_unlock(&kvm->lock); | |
682edb4c | 662 | up_read(&kvm->slots_lock); |
8a98f664 XZ |
663 | return r; |
664 | } | |
665 | #endif | |
666 | ||
0a920356 WH |
667 | #ifdef KVM_CAP_DEVICE_DEASSIGNMENT |
668 | static int kvm_vm_ioctl_deassign_device(struct kvm *kvm, | |
669 | struct kvm_assigned_pci_dev *assigned_dev) | |
670 | { | |
671 | int r = 0; | |
672 | struct kvm_assigned_dev_kernel *match; | |
673 | ||
674 | mutex_lock(&kvm->lock); | |
675 | ||
676 | match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, | |
677 | assigned_dev->assigned_dev_id); | |
678 | if (!match) { | |
679 | printk(KERN_INFO "%s: device hasn't been assigned before, " | |
680 | "so cannot be deassigned\n", __func__); | |
681 | r = -EINVAL; | |
682 | goto out; | |
683 | } | |
684 | ||
4a906e49 | 685 | if (match->flags & KVM_DEV_ASSIGN_ENABLE_IOMMU) |
0a920356 WH |
686 | kvm_deassign_device(kvm, match); |
687 | ||
688 | kvm_free_assigned_device(kvm, match); | |
689 | ||
690 | out: | |
691 | mutex_unlock(&kvm->lock); | |
692 | return r; | |
693 | } | |
694 | #endif | |
695 | ||
5aacf0ca JM |
696 | static inline int valid_vcpu(int n) |
697 | { | |
698 | return likely(n >= 0 && n < KVM_MAX_VCPUS); | |
699 | } | |
700 | ||
c77fb9dc | 701 | inline int kvm_is_mmio_pfn(pfn_t pfn) |
cbff90a7 | 702 | { |
fc5659c8 JR |
703 | if (pfn_valid(pfn)) { |
704 | struct page *page = compound_head(pfn_to_page(pfn)); | |
705 | return PageReserved(page); | |
706 | } | |
cbff90a7 BAY |
707 | |
708 | return true; | |
709 | } | |
710 | ||
bccf2150 AK |
711 | /* |
712 | * Switches to specified vcpu, until a matching vcpu_put() | |
713 | */ | |
313a3dc7 | 714 | void vcpu_load(struct kvm_vcpu *vcpu) |
6aa8b732 | 715 | { |
15ad7146 AK |
716 | int cpu; |
717 | ||
bccf2150 | 718 | mutex_lock(&vcpu->mutex); |
15ad7146 AK |
719 | cpu = get_cpu(); |
720 | preempt_notifier_register(&vcpu->preempt_notifier); | |
313a3dc7 | 721 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 | 722 | put_cpu(); |
6aa8b732 AK |
723 | } |
724 | ||
313a3dc7 | 725 | void vcpu_put(struct kvm_vcpu *vcpu) |
6aa8b732 | 726 | { |
15ad7146 | 727 | preempt_disable(); |
313a3dc7 | 728 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
729 | preempt_notifier_unregister(&vcpu->preempt_notifier); |
730 | preempt_enable(); | |
6aa8b732 AK |
731 | mutex_unlock(&vcpu->mutex); |
732 | } | |
733 | ||
d9e368d6 AK |
734 | static void ack_flush(void *_completed) |
735 | { | |
d9e368d6 AK |
736 | } |
737 | ||
49846896 | 738 | static bool make_all_cpus_request(struct kvm *kvm, unsigned int req) |
d9e368d6 | 739 | { |
597a5f55 | 740 | int i, cpu, me; |
6ef7a1bc RR |
741 | cpumask_var_t cpus; |
742 | bool called = true; | |
d9e368d6 | 743 | struct kvm_vcpu *vcpu; |
d9e368d6 | 744 | |
6ef7a1bc RR |
745 | if (alloc_cpumask_var(&cpus, GFP_ATOMIC)) |
746 | cpumask_clear(cpus); | |
747 | ||
597a5f55 | 748 | me = get_cpu(); |
fb3f0f51 RR |
749 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { |
750 | vcpu = kvm->vcpus[i]; | |
751 | if (!vcpu) | |
752 | continue; | |
49846896 | 753 | if (test_and_set_bit(req, &vcpu->requests)) |
d9e368d6 AK |
754 | continue; |
755 | cpu = vcpu->cpu; | |
6ef7a1bc RR |
756 | if (cpus != NULL && cpu != -1 && cpu != me) |
757 | cpumask_set_cpu(cpu, cpus); | |
49846896 | 758 | } |
6ef7a1bc RR |
759 | if (unlikely(cpus == NULL)) |
760 | smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1); | |
761 | else if (!cpumask_empty(cpus)) | |
762 | smp_call_function_many(cpus, ack_flush, NULL, 1); | |
763 | else | |
764 | called = false; | |
597a5f55 | 765 | put_cpu(); |
6ef7a1bc | 766 | free_cpumask_var(cpus); |
49846896 | 767 | return called; |
d9e368d6 AK |
768 | } |
769 | ||
49846896 | 770 | void kvm_flush_remote_tlbs(struct kvm *kvm) |
2e53d63a | 771 | { |
49846896 RR |
772 | if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH)) |
773 | ++kvm->stat.remote_tlb_flush; | |
2e53d63a MT |
774 | } |
775 | ||
49846896 RR |
776 | void kvm_reload_remote_mmus(struct kvm *kvm) |
777 | { | |
778 | make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD); | |
779 | } | |
2e53d63a | 780 | |
fb3f0f51 RR |
781 | int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id) |
782 | { | |
783 | struct page *page; | |
784 | int r; | |
785 | ||
786 | mutex_init(&vcpu->mutex); | |
787 | vcpu->cpu = -1; | |
fb3f0f51 RR |
788 | vcpu->kvm = kvm; |
789 | vcpu->vcpu_id = id; | |
b6958ce4 | 790 | init_waitqueue_head(&vcpu->wq); |
fb3f0f51 RR |
791 | |
792 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
793 | if (!page) { | |
794 | r = -ENOMEM; | |
795 | goto fail; | |
796 | } | |
797 | vcpu->run = page_address(page); | |
798 | ||
e9b11c17 | 799 | r = kvm_arch_vcpu_init(vcpu); |
fb3f0f51 | 800 | if (r < 0) |
e9b11c17 | 801 | goto fail_free_run; |
fb3f0f51 RR |
802 | return 0; |
803 | ||
fb3f0f51 RR |
804 | fail_free_run: |
805 | free_page((unsigned long)vcpu->run); | |
806 | fail: | |
76fafa5e | 807 | return r; |
fb3f0f51 RR |
808 | } |
809 | EXPORT_SYMBOL_GPL(kvm_vcpu_init); | |
810 | ||
811 | void kvm_vcpu_uninit(struct kvm_vcpu *vcpu) | |
812 | { | |
e9b11c17 | 813 | kvm_arch_vcpu_uninit(vcpu); |
fb3f0f51 RR |
814 | free_page((unsigned long)vcpu->run); |
815 | } | |
816 | EXPORT_SYMBOL_GPL(kvm_vcpu_uninit); | |
817 | ||
e930bffe AA |
818 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
819 | static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn) | |
820 | { | |
821 | return container_of(mn, struct kvm, mmu_notifier); | |
822 | } | |
823 | ||
824 | static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn, | |
825 | struct mm_struct *mm, | |
826 | unsigned long address) | |
827 | { | |
828 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
829 | int need_tlb_flush; | |
830 | ||
831 | /* | |
832 | * When ->invalidate_page runs, the linux pte has been zapped | |
833 | * already but the page is still allocated until | |
834 | * ->invalidate_page returns. So if we increase the sequence | |
835 | * here the kvm page fault will notice if the spte can't be | |
836 | * established because the page is going to be freed. If | |
837 | * instead the kvm page fault establishes the spte before | |
838 | * ->invalidate_page runs, kvm_unmap_hva will release it | |
839 | * before returning. | |
840 | * | |
841 | * The sequence increase only need to be seen at spin_unlock | |
842 | * time, and not at spin_lock time. | |
843 | * | |
844 | * Increasing the sequence after the spin_unlock would be | |
845 | * unsafe because the kvm page fault could then establish the | |
846 | * pte after kvm_unmap_hva returned, without noticing the page | |
847 | * is going to be freed. | |
848 | */ | |
849 | spin_lock(&kvm->mmu_lock); | |
850 | kvm->mmu_notifier_seq++; | |
851 | need_tlb_flush = kvm_unmap_hva(kvm, address); | |
852 | spin_unlock(&kvm->mmu_lock); | |
853 | ||
854 | /* we've to flush the tlb before the pages can be freed */ | |
855 | if (need_tlb_flush) | |
856 | kvm_flush_remote_tlbs(kvm); | |
857 | ||
858 | } | |
859 | ||
860 | static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn, | |
861 | struct mm_struct *mm, | |
862 | unsigned long start, | |
863 | unsigned long end) | |
864 | { | |
865 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
866 | int need_tlb_flush = 0; | |
867 | ||
868 | spin_lock(&kvm->mmu_lock); | |
869 | /* | |
870 | * The count increase must become visible at unlock time as no | |
871 | * spte can be established without taking the mmu_lock and | |
872 | * count is also read inside the mmu_lock critical section. | |
873 | */ | |
874 | kvm->mmu_notifier_count++; | |
875 | for (; start < end; start += PAGE_SIZE) | |
876 | need_tlb_flush |= kvm_unmap_hva(kvm, start); | |
877 | spin_unlock(&kvm->mmu_lock); | |
878 | ||
879 | /* we've to flush the tlb before the pages can be freed */ | |
880 | if (need_tlb_flush) | |
881 | kvm_flush_remote_tlbs(kvm); | |
882 | } | |
883 | ||
884 | static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn, | |
885 | struct mm_struct *mm, | |
886 | unsigned long start, | |
887 | unsigned long end) | |
888 | { | |
889 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
890 | ||
891 | spin_lock(&kvm->mmu_lock); | |
892 | /* | |
893 | * This sequence increase will notify the kvm page fault that | |
894 | * the page that is going to be mapped in the spte could have | |
895 | * been freed. | |
896 | */ | |
897 | kvm->mmu_notifier_seq++; | |
898 | /* | |
899 | * The above sequence increase must be visible before the | |
900 | * below count decrease but both values are read by the kvm | |
901 | * page fault under mmu_lock spinlock so we don't need to add | |
902 | * a smb_wmb() here in between the two. | |
903 | */ | |
904 | kvm->mmu_notifier_count--; | |
905 | spin_unlock(&kvm->mmu_lock); | |
906 | ||
907 | BUG_ON(kvm->mmu_notifier_count < 0); | |
908 | } | |
909 | ||
910 | static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn, | |
911 | struct mm_struct *mm, | |
912 | unsigned long address) | |
913 | { | |
914 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
915 | int young; | |
916 | ||
917 | spin_lock(&kvm->mmu_lock); | |
918 | young = kvm_age_hva(kvm, address); | |
919 | spin_unlock(&kvm->mmu_lock); | |
920 | ||
921 | if (young) | |
922 | kvm_flush_remote_tlbs(kvm); | |
923 | ||
924 | return young; | |
925 | } | |
926 | ||
85db06e5 MT |
927 | static void kvm_mmu_notifier_release(struct mmu_notifier *mn, |
928 | struct mm_struct *mm) | |
929 | { | |
930 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
931 | kvm_arch_flush_shadow(kvm); | |
932 | } | |
933 | ||
e930bffe AA |
934 | static const struct mmu_notifier_ops kvm_mmu_notifier_ops = { |
935 | .invalidate_page = kvm_mmu_notifier_invalidate_page, | |
936 | .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start, | |
937 | .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end, | |
938 | .clear_flush_young = kvm_mmu_notifier_clear_flush_young, | |
85db06e5 | 939 | .release = kvm_mmu_notifier_release, |
e930bffe AA |
940 | }; |
941 | #endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */ | |
942 | ||
f17abe9a | 943 | static struct kvm *kvm_create_vm(void) |
6aa8b732 | 944 | { |
d19a9cd2 | 945 | struct kvm *kvm = kvm_arch_create_vm(); |
5f94c174 LV |
946 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
947 | struct page *page; | |
948 | #endif | |
6aa8b732 | 949 | |
d19a9cd2 ZX |
950 | if (IS_ERR(kvm)) |
951 | goto out; | |
75858a84 | 952 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
399ec807 | 953 | INIT_LIST_HEAD(&kvm->irq_routing); |
75858a84 AK |
954 | INIT_HLIST_HEAD(&kvm->mask_notifier_list); |
955 | #endif | |
6aa8b732 | 956 | |
5f94c174 LV |
957 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
958 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
959 | if (!page) { | |
960 | kfree(kvm); | |
961 | return ERR_PTR(-ENOMEM); | |
962 | } | |
963 | kvm->coalesced_mmio_ring = | |
964 | (struct kvm_coalesced_mmio_ring *)page_address(page); | |
965 | #endif | |
966 | ||
e930bffe AA |
967 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
968 | { | |
969 | int err; | |
970 | kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops; | |
971 | err = mmu_notifier_register(&kvm->mmu_notifier, current->mm); | |
972 | if (err) { | |
973 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
974 | put_page(page); | |
975 | #endif | |
976 | kfree(kvm); | |
977 | return ERR_PTR(err); | |
978 | } | |
979 | } | |
980 | #endif | |
981 | ||
6d4e4c4f AK |
982 | kvm->mm = current->mm; |
983 | atomic_inc(&kvm->mm->mm_count); | |
aaee2c94 | 984 | spin_lock_init(&kvm->mmu_lock); |
74906345 | 985 | kvm_io_bus_init(&kvm->pio_bus); |
11ec2804 | 986 | mutex_init(&kvm->lock); |
2eeb2e94 | 987 | kvm_io_bus_init(&kvm->mmio_bus); |
72dc67a6 | 988 | init_rwsem(&kvm->slots_lock); |
d39f13b0 | 989 | atomic_set(&kvm->users_count, 1); |
5e58cfe4 RR |
990 | spin_lock(&kvm_lock); |
991 | list_add(&kvm->vm_list, &vm_list); | |
992 | spin_unlock(&kvm_lock); | |
5f94c174 LV |
993 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
994 | kvm_coalesced_mmio_init(kvm); | |
995 | #endif | |
d19a9cd2 | 996 | out: |
f17abe9a AK |
997 | return kvm; |
998 | } | |
999 | ||
6aa8b732 AK |
1000 | /* |
1001 | * Free any memory in @free but not in @dont. | |
1002 | */ | |
1003 | static void kvm_free_physmem_slot(struct kvm_memory_slot *free, | |
1004 | struct kvm_memory_slot *dont) | |
1005 | { | |
290fc38d IE |
1006 | if (!dont || free->rmap != dont->rmap) |
1007 | vfree(free->rmap); | |
6aa8b732 AK |
1008 | |
1009 | if (!dont || free->dirty_bitmap != dont->dirty_bitmap) | |
1010 | vfree(free->dirty_bitmap); | |
1011 | ||
05da4558 MT |
1012 | if (!dont || free->lpage_info != dont->lpage_info) |
1013 | vfree(free->lpage_info); | |
1014 | ||
6aa8b732 | 1015 | free->npages = 0; |
8b6d44c7 | 1016 | free->dirty_bitmap = NULL; |
8d4e1288 | 1017 | free->rmap = NULL; |
05da4558 | 1018 | free->lpage_info = NULL; |
6aa8b732 AK |
1019 | } |
1020 | ||
d19a9cd2 | 1021 | void kvm_free_physmem(struct kvm *kvm) |
6aa8b732 AK |
1022 | { |
1023 | int i; | |
1024 | ||
1025 | for (i = 0; i < kvm->nmemslots; ++i) | |
8b6d44c7 | 1026 | kvm_free_physmem_slot(&kvm->memslots[i], NULL); |
6aa8b732 AK |
1027 | } |
1028 | ||
f17abe9a AK |
1029 | static void kvm_destroy_vm(struct kvm *kvm) |
1030 | { | |
6d4e4c4f AK |
1031 | struct mm_struct *mm = kvm->mm; |
1032 | ||
ad8ba2cd | 1033 | kvm_arch_sync_events(kvm); |
133de902 AK |
1034 | spin_lock(&kvm_lock); |
1035 | list_del(&kvm->vm_list); | |
1036 | spin_unlock(&kvm_lock); | |
399ec807 | 1037 | kvm_free_irq_routing(kvm); |
74906345 | 1038 | kvm_io_bus_destroy(&kvm->pio_bus); |
2eeb2e94 | 1039 | kvm_io_bus_destroy(&kvm->mmio_bus); |
5f94c174 LV |
1040 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
1041 | if (kvm->coalesced_mmio_ring != NULL) | |
1042 | free_page((unsigned long)kvm->coalesced_mmio_ring); | |
e930bffe AA |
1043 | #endif |
1044 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) | |
1045 | mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm); | |
f00be0ca GN |
1046 | #else |
1047 | kvm_arch_flush_shadow(kvm); | |
5f94c174 | 1048 | #endif |
d19a9cd2 | 1049 | kvm_arch_destroy_vm(kvm); |
6d4e4c4f | 1050 | mmdrop(mm); |
f17abe9a AK |
1051 | } |
1052 | ||
d39f13b0 IE |
1053 | void kvm_get_kvm(struct kvm *kvm) |
1054 | { | |
1055 | atomic_inc(&kvm->users_count); | |
1056 | } | |
1057 | EXPORT_SYMBOL_GPL(kvm_get_kvm); | |
1058 | ||
1059 | void kvm_put_kvm(struct kvm *kvm) | |
1060 | { | |
1061 | if (atomic_dec_and_test(&kvm->users_count)) | |
1062 | kvm_destroy_vm(kvm); | |
1063 | } | |
1064 | EXPORT_SYMBOL_GPL(kvm_put_kvm); | |
1065 | ||
1066 | ||
f17abe9a AK |
1067 | static int kvm_vm_release(struct inode *inode, struct file *filp) |
1068 | { | |
1069 | struct kvm *kvm = filp->private_data; | |
1070 | ||
d39f13b0 | 1071 | kvm_put_kvm(kvm); |
6aa8b732 AK |
1072 | return 0; |
1073 | } | |
1074 | ||
6aa8b732 AK |
1075 | /* |
1076 | * Allocate some memory and give it an address in the guest physical address | |
1077 | * space. | |
1078 | * | |
1079 | * Discontiguous memory is allowed, mostly for framebuffers. | |
f78e0e2e | 1080 | * |
10589a46 | 1081 | * Must be called holding mmap_sem for write. |
6aa8b732 | 1082 | */ |
f78e0e2e SY |
1083 | int __kvm_set_memory_region(struct kvm *kvm, |
1084 | struct kvm_userspace_memory_region *mem, | |
1085 | int user_alloc) | |
6aa8b732 AK |
1086 | { |
1087 | int r; | |
1088 | gfn_t base_gfn; | |
ac04527f | 1089 | unsigned long npages, ugfn; |
09f8ca74 | 1090 | unsigned long largepages, i; |
6aa8b732 AK |
1091 | struct kvm_memory_slot *memslot; |
1092 | struct kvm_memory_slot old, new; | |
6aa8b732 AK |
1093 | |
1094 | r = -EINVAL; | |
1095 | /* General sanity checks */ | |
1096 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
1097 | goto out; | |
1098 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
1099 | goto out; | |
e7cacd40 | 1100 | if (user_alloc && (mem->userspace_addr & (PAGE_SIZE - 1))) |
78749809 | 1101 | goto out; |
e0d62c7f | 1102 | if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS) |
6aa8b732 AK |
1103 | goto out; |
1104 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
1105 | goto out; | |
1106 | ||
1107 | memslot = &kvm->memslots[mem->slot]; | |
1108 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; | |
1109 | npages = mem->memory_size >> PAGE_SHIFT; | |
1110 | ||
1111 | if (!npages) | |
1112 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
1113 | ||
6aa8b732 AK |
1114 | new = old = *memslot; |
1115 | ||
1116 | new.base_gfn = base_gfn; | |
1117 | new.npages = npages; | |
1118 | new.flags = mem->flags; | |
1119 | ||
1120 | /* Disallow changing a memory slot's size. */ | |
1121 | r = -EINVAL; | |
1122 | if (npages && old.npages && npages != old.npages) | |
f78e0e2e | 1123 | goto out_free; |
6aa8b732 AK |
1124 | |
1125 | /* Check for overlaps */ | |
1126 | r = -EEXIST; | |
1127 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
1128 | struct kvm_memory_slot *s = &kvm->memslots[i]; | |
1129 | ||
4cd481f6 | 1130 | if (s == memslot || !s->npages) |
6aa8b732 AK |
1131 | continue; |
1132 | if (!((base_gfn + npages <= s->base_gfn) || | |
1133 | (base_gfn >= s->base_gfn + s->npages))) | |
f78e0e2e | 1134 | goto out_free; |
6aa8b732 | 1135 | } |
6aa8b732 | 1136 | |
6aa8b732 AK |
1137 | /* Free page dirty bitmap if unneeded */ |
1138 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 1139 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
1140 | |
1141 | r = -ENOMEM; | |
1142 | ||
1143 | /* Allocate if a slot is being created */ | |
eff0114a | 1144 | #ifndef CONFIG_S390 |
8d4e1288 | 1145 | if (npages && !new.rmap) { |
d77c26fc | 1146 | new.rmap = vmalloc(npages * sizeof(struct page *)); |
290fc38d IE |
1147 | |
1148 | if (!new.rmap) | |
f78e0e2e | 1149 | goto out_free; |
290fc38d | 1150 | |
290fc38d | 1151 | memset(new.rmap, 0, npages * sizeof(*new.rmap)); |
8d4e1288 | 1152 | |
80b14b5b | 1153 | new.user_alloc = user_alloc; |
604b38ac AA |
1154 | /* |
1155 | * hva_to_rmmap() serialzies with the mmu_lock and to be | |
1156 | * safe it has to ignore memslots with !user_alloc && | |
1157 | * !userspace_addr. | |
1158 | */ | |
1159 | if (user_alloc) | |
1160 | new.userspace_addr = mem->userspace_addr; | |
1161 | else | |
1162 | new.userspace_addr = 0; | |
6aa8b732 | 1163 | } |
05da4558 | 1164 | if (npages && !new.lpage_info) { |
99894a79 AK |
1165 | largepages = 1 + (base_gfn + npages - 1) / KVM_PAGES_PER_HPAGE; |
1166 | largepages -= base_gfn / KVM_PAGES_PER_HPAGE; | |
05da4558 MT |
1167 | |
1168 | new.lpage_info = vmalloc(largepages * sizeof(*new.lpage_info)); | |
1169 | ||
1170 | if (!new.lpage_info) | |
1171 | goto out_free; | |
1172 | ||
1173 | memset(new.lpage_info, 0, largepages * sizeof(*new.lpage_info)); | |
1174 | ||
1175 | if (base_gfn % KVM_PAGES_PER_HPAGE) | |
1176 | new.lpage_info[0].write_count = 1; | |
1177 | if ((base_gfn+npages) % KVM_PAGES_PER_HPAGE) | |
1178 | new.lpage_info[largepages-1].write_count = 1; | |
ac04527f AK |
1179 | ugfn = new.userspace_addr >> PAGE_SHIFT; |
1180 | /* | |
1181 | * If the gfn and userspace address are not aligned wrt each | |
1182 | * other, disable large page support for this slot | |
1183 | */ | |
1184 | if ((base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE - 1)) | |
1185 | for (i = 0; i < largepages; ++i) | |
1186 | new.lpage_info[i].write_count = 1; | |
05da4558 | 1187 | } |
6aa8b732 AK |
1188 | |
1189 | /* Allocate page dirty bitmap if needed */ | |
1190 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
1191 | unsigned dirty_bytes = ALIGN(npages, BITS_PER_LONG) / 8; | |
1192 | ||
1193 | new.dirty_bitmap = vmalloc(dirty_bytes); | |
1194 | if (!new.dirty_bitmap) | |
f78e0e2e | 1195 | goto out_free; |
6aa8b732 AK |
1196 | memset(new.dirty_bitmap, 0, dirty_bytes); |
1197 | } | |
eff0114a | 1198 | #endif /* not defined CONFIG_S390 */ |
6aa8b732 | 1199 | |
34d4cb8f MT |
1200 | if (!npages) |
1201 | kvm_arch_flush_shadow(kvm); | |
1202 | ||
604b38ac AA |
1203 | spin_lock(&kvm->mmu_lock); |
1204 | if (mem->slot >= kvm->nmemslots) | |
1205 | kvm->nmemslots = mem->slot + 1; | |
1206 | ||
3ad82a7e | 1207 | *memslot = new; |
604b38ac | 1208 | spin_unlock(&kvm->mmu_lock); |
3ad82a7e | 1209 | |
0de10343 ZX |
1210 | r = kvm_arch_set_memory_region(kvm, mem, old, user_alloc); |
1211 | if (r) { | |
604b38ac | 1212 | spin_lock(&kvm->mmu_lock); |
0de10343 | 1213 | *memslot = old; |
604b38ac | 1214 | spin_unlock(&kvm->mmu_lock); |
0de10343 | 1215 | goto out_free; |
82ce2c96 IE |
1216 | } |
1217 | ||
6f897248 GC |
1218 | kvm_free_physmem_slot(&old, npages ? &new : NULL); |
1219 | /* Slot deletion case: we have to update the current slot */ | |
b43b1901 | 1220 | spin_lock(&kvm->mmu_lock); |
6f897248 GC |
1221 | if (!npages) |
1222 | *memslot = old; | |
b43b1901 | 1223 | spin_unlock(&kvm->mmu_lock); |
8a98f664 | 1224 | #ifdef CONFIG_DMAR |
62c476c7 BAY |
1225 | /* map the pages in iommu page table */ |
1226 | r = kvm_iommu_map_pages(kvm, base_gfn, npages); | |
1227 | if (r) | |
1228 | goto out; | |
8a98f664 | 1229 | #endif |
6aa8b732 AK |
1230 | return 0; |
1231 | ||
f78e0e2e | 1232 | out_free: |
6aa8b732 AK |
1233 | kvm_free_physmem_slot(&new, &old); |
1234 | out: | |
1235 | return r; | |
210c7c4d IE |
1236 | |
1237 | } | |
f78e0e2e SY |
1238 | EXPORT_SYMBOL_GPL(__kvm_set_memory_region); |
1239 | ||
1240 | int kvm_set_memory_region(struct kvm *kvm, | |
1241 | struct kvm_userspace_memory_region *mem, | |
1242 | int user_alloc) | |
1243 | { | |
1244 | int r; | |
1245 | ||
72dc67a6 | 1246 | down_write(&kvm->slots_lock); |
f78e0e2e | 1247 | r = __kvm_set_memory_region(kvm, mem, user_alloc); |
72dc67a6 | 1248 | up_write(&kvm->slots_lock); |
f78e0e2e SY |
1249 | return r; |
1250 | } | |
210c7c4d IE |
1251 | EXPORT_SYMBOL_GPL(kvm_set_memory_region); |
1252 | ||
1fe779f8 CO |
1253 | int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
1254 | struct | |
1255 | kvm_userspace_memory_region *mem, | |
1256 | int user_alloc) | |
210c7c4d | 1257 | { |
e0d62c7f IE |
1258 | if (mem->slot >= KVM_MEMORY_SLOTS) |
1259 | return -EINVAL; | |
210c7c4d | 1260 | return kvm_set_memory_region(kvm, mem, user_alloc); |
6aa8b732 AK |
1261 | } |
1262 | ||
5bb064dc ZX |
1263 | int kvm_get_dirty_log(struct kvm *kvm, |
1264 | struct kvm_dirty_log *log, int *is_dirty) | |
6aa8b732 AK |
1265 | { |
1266 | struct kvm_memory_slot *memslot; | |
1267 | int r, i; | |
1268 | int n; | |
1269 | unsigned long any = 0; | |
1270 | ||
6aa8b732 AK |
1271 | r = -EINVAL; |
1272 | if (log->slot >= KVM_MEMORY_SLOTS) | |
1273 | goto out; | |
1274 | ||
1275 | memslot = &kvm->memslots[log->slot]; | |
1276 | r = -ENOENT; | |
1277 | if (!memslot->dirty_bitmap) | |
1278 | goto out; | |
1279 | ||
cd1a4a98 | 1280 | n = ALIGN(memslot->npages, BITS_PER_LONG) / 8; |
6aa8b732 | 1281 | |
cd1a4a98 | 1282 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
1283 | any = memslot->dirty_bitmap[i]; |
1284 | ||
1285 | r = -EFAULT; | |
1286 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
1287 | goto out; | |
1288 | ||
5bb064dc ZX |
1289 | if (any) |
1290 | *is_dirty = 1; | |
6aa8b732 AK |
1291 | |
1292 | r = 0; | |
6aa8b732 | 1293 | out: |
6aa8b732 AK |
1294 | return r; |
1295 | } | |
1296 | ||
cea7bb21 IE |
1297 | int is_error_page(struct page *page) |
1298 | { | |
1299 | return page == bad_page; | |
1300 | } | |
1301 | EXPORT_SYMBOL_GPL(is_error_page); | |
1302 | ||
35149e21 AL |
1303 | int is_error_pfn(pfn_t pfn) |
1304 | { | |
1305 | return pfn == bad_pfn; | |
1306 | } | |
1307 | EXPORT_SYMBOL_GPL(is_error_pfn); | |
1308 | ||
f9d46eb0 IE |
1309 | static inline unsigned long bad_hva(void) |
1310 | { | |
1311 | return PAGE_OFFSET; | |
1312 | } | |
1313 | ||
1314 | int kvm_is_error_hva(unsigned long addr) | |
1315 | { | |
1316 | return addr == bad_hva(); | |
1317 | } | |
1318 | EXPORT_SYMBOL_GPL(kvm_is_error_hva); | |
1319 | ||
2843099f | 1320 | struct kvm_memory_slot *gfn_to_memslot_unaliased(struct kvm *kvm, gfn_t gfn) |
6aa8b732 AK |
1321 | { |
1322 | int i; | |
1323 | ||
1324 | for (i = 0; i < kvm->nmemslots; ++i) { | |
1325 | struct kvm_memory_slot *memslot = &kvm->memslots[i]; | |
1326 | ||
1327 | if (gfn >= memslot->base_gfn | |
1328 | && gfn < memslot->base_gfn + memslot->npages) | |
1329 | return memslot; | |
1330 | } | |
8b6d44c7 | 1331 | return NULL; |
6aa8b732 | 1332 | } |
2843099f | 1333 | EXPORT_SYMBOL_GPL(gfn_to_memslot_unaliased); |
e8207547 AK |
1334 | |
1335 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
1336 | { | |
1337 | gfn = unalias_gfn(kvm, gfn); | |
2843099f | 1338 | return gfn_to_memslot_unaliased(kvm, gfn); |
e8207547 | 1339 | } |
6aa8b732 | 1340 | |
e0d62c7f IE |
1341 | int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
1342 | { | |
1343 | int i; | |
1344 | ||
1345 | gfn = unalias_gfn(kvm, gfn); | |
1346 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
1347 | struct kvm_memory_slot *memslot = &kvm->memslots[i]; | |
1348 | ||
1349 | if (gfn >= memslot->base_gfn | |
1350 | && gfn < memslot->base_gfn + memslot->npages) | |
1351 | return 1; | |
1352 | } | |
1353 | return 0; | |
1354 | } | |
1355 | EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); | |
1356 | ||
05da4558 | 1357 | unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn) |
539cb660 IE |
1358 | { |
1359 | struct kvm_memory_slot *slot; | |
1360 | ||
1361 | gfn = unalias_gfn(kvm, gfn); | |
2843099f | 1362 | slot = gfn_to_memslot_unaliased(kvm, gfn); |
539cb660 IE |
1363 | if (!slot) |
1364 | return bad_hva(); | |
1365 | return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE); | |
1366 | } | |
0d150298 | 1367 | EXPORT_SYMBOL_GPL(gfn_to_hva); |
539cb660 | 1368 | |
35149e21 | 1369 | pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn) |
954bbbc2 | 1370 | { |
8d4e1288 | 1371 | struct page *page[1]; |
539cb660 | 1372 | unsigned long addr; |
8d4e1288 | 1373 | int npages; |
2e2e3738 | 1374 | pfn_t pfn; |
954bbbc2 | 1375 | |
60395224 AK |
1376 | might_sleep(); |
1377 | ||
539cb660 IE |
1378 | addr = gfn_to_hva(kvm, gfn); |
1379 | if (kvm_is_error_hva(addr)) { | |
8a7ae055 | 1380 | get_page(bad_page); |
35149e21 | 1381 | return page_to_pfn(bad_page); |
8a7ae055 | 1382 | } |
8d4e1288 | 1383 | |
4c2155ce | 1384 | npages = get_user_pages_fast(addr, 1, 1, page); |
539cb660 | 1385 | |
2e2e3738 AL |
1386 | if (unlikely(npages != 1)) { |
1387 | struct vm_area_struct *vma; | |
1388 | ||
4c2155ce | 1389 | down_read(¤t->mm->mmap_sem); |
2e2e3738 | 1390 | vma = find_vma(current->mm, addr); |
4c2155ce | 1391 | |
2e2e3738 AL |
1392 | if (vma == NULL || addr < vma->vm_start || |
1393 | !(vma->vm_flags & VM_PFNMAP)) { | |
4c2155ce | 1394 | up_read(¤t->mm->mmap_sem); |
2e2e3738 AL |
1395 | get_page(bad_page); |
1396 | return page_to_pfn(bad_page); | |
1397 | } | |
1398 | ||
1399 | pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; | |
4c2155ce | 1400 | up_read(¤t->mm->mmap_sem); |
c77fb9dc | 1401 | BUG_ON(!kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
1402 | } else |
1403 | pfn = page_to_pfn(page[0]); | |
8d4e1288 | 1404 | |
2e2e3738 | 1405 | return pfn; |
35149e21 AL |
1406 | } |
1407 | ||
1408 | EXPORT_SYMBOL_GPL(gfn_to_pfn); | |
1409 | ||
1410 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) | |
1411 | { | |
2e2e3738 AL |
1412 | pfn_t pfn; |
1413 | ||
1414 | pfn = gfn_to_pfn(kvm, gfn); | |
c77fb9dc | 1415 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 AL |
1416 | return pfn_to_page(pfn); |
1417 | ||
c77fb9dc | 1418 | WARN_ON(kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
1419 | |
1420 | get_page(bad_page); | |
1421 | return bad_page; | |
954bbbc2 | 1422 | } |
aab61cc0 | 1423 | |
954bbbc2 AK |
1424 | EXPORT_SYMBOL_GPL(gfn_to_page); |
1425 | ||
b4231d61 IE |
1426 | void kvm_release_page_clean(struct page *page) |
1427 | { | |
35149e21 | 1428 | kvm_release_pfn_clean(page_to_pfn(page)); |
b4231d61 IE |
1429 | } |
1430 | EXPORT_SYMBOL_GPL(kvm_release_page_clean); | |
1431 | ||
35149e21 AL |
1432 | void kvm_release_pfn_clean(pfn_t pfn) |
1433 | { | |
c77fb9dc | 1434 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1435 | put_page(pfn_to_page(pfn)); |
35149e21 AL |
1436 | } |
1437 | EXPORT_SYMBOL_GPL(kvm_release_pfn_clean); | |
1438 | ||
b4231d61 | 1439 | void kvm_release_page_dirty(struct page *page) |
8a7ae055 | 1440 | { |
35149e21 AL |
1441 | kvm_release_pfn_dirty(page_to_pfn(page)); |
1442 | } | |
1443 | EXPORT_SYMBOL_GPL(kvm_release_page_dirty); | |
1444 | ||
1445 | void kvm_release_pfn_dirty(pfn_t pfn) | |
1446 | { | |
1447 | kvm_set_pfn_dirty(pfn); | |
1448 | kvm_release_pfn_clean(pfn); | |
1449 | } | |
1450 | EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty); | |
1451 | ||
1452 | void kvm_set_page_dirty(struct page *page) | |
1453 | { | |
1454 | kvm_set_pfn_dirty(page_to_pfn(page)); | |
1455 | } | |
1456 | EXPORT_SYMBOL_GPL(kvm_set_page_dirty); | |
1457 | ||
1458 | void kvm_set_pfn_dirty(pfn_t pfn) | |
1459 | { | |
c77fb9dc | 1460 | if (!kvm_is_mmio_pfn(pfn)) { |
2e2e3738 AL |
1461 | struct page *page = pfn_to_page(pfn); |
1462 | if (!PageReserved(page)) | |
1463 | SetPageDirty(page); | |
1464 | } | |
8a7ae055 | 1465 | } |
35149e21 AL |
1466 | EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty); |
1467 | ||
1468 | void kvm_set_pfn_accessed(pfn_t pfn) | |
1469 | { | |
c77fb9dc | 1470 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1471 | mark_page_accessed(pfn_to_page(pfn)); |
35149e21 AL |
1472 | } |
1473 | EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed); | |
1474 | ||
1475 | void kvm_get_pfn(pfn_t pfn) | |
1476 | { | |
c77fb9dc | 1477 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1478 | get_page(pfn_to_page(pfn)); |
35149e21 AL |
1479 | } |
1480 | EXPORT_SYMBOL_GPL(kvm_get_pfn); | |
8a7ae055 | 1481 | |
195aefde IE |
1482 | static int next_segment(unsigned long len, int offset) |
1483 | { | |
1484 | if (len > PAGE_SIZE - offset) | |
1485 | return PAGE_SIZE - offset; | |
1486 | else | |
1487 | return len; | |
1488 | } | |
1489 | ||
1490 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, | |
1491 | int len) | |
1492 | { | |
e0506bcb IE |
1493 | int r; |
1494 | unsigned long addr; | |
195aefde | 1495 | |
e0506bcb IE |
1496 | addr = gfn_to_hva(kvm, gfn); |
1497 | if (kvm_is_error_hva(addr)) | |
1498 | return -EFAULT; | |
1499 | r = copy_from_user(data, (void __user *)addr + offset, len); | |
1500 | if (r) | |
195aefde | 1501 | return -EFAULT; |
195aefde IE |
1502 | return 0; |
1503 | } | |
1504 | EXPORT_SYMBOL_GPL(kvm_read_guest_page); | |
1505 | ||
1506 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len) | |
1507 | { | |
1508 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1509 | int seg; | |
1510 | int offset = offset_in_page(gpa); | |
1511 | int ret; | |
1512 | ||
1513 | while ((seg = next_segment(len, offset)) != 0) { | |
1514 | ret = kvm_read_guest_page(kvm, gfn, data, offset, seg); | |
1515 | if (ret < 0) | |
1516 | return ret; | |
1517 | offset = 0; | |
1518 | len -= seg; | |
1519 | data += seg; | |
1520 | ++gfn; | |
1521 | } | |
1522 | return 0; | |
1523 | } | |
1524 | EXPORT_SYMBOL_GPL(kvm_read_guest); | |
1525 | ||
7ec54588 MT |
1526 | int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data, |
1527 | unsigned long len) | |
1528 | { | |
1529 | int r; | |
1530 | unsigned long addr; | |
1531 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1532 | int offset = offset_in_page(gpa); | |
1533 | ||
1534 | addr = gfn_to_hva(kvm, gfn); | |
1535 | if (kvm_is_error_hva(addr)) | |
1536 | return -EFAULT; | |
0aac03f0 | 1537 | pagefault_disable(); |
7ec54588 | 1538 | r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len); |
0aac03f0 | 1539 | pagefault_enable(); |
7ec54588 MT |
1540 | if (r) |
1541 | return -EFAULT; | |
1542 | return 0; | |
1543 | } | |
1544 | EXPORT_SYMBOL(kvm_read_guest_atomic); | |
1545 | ||
195aefde IE |
1546 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, |
1547 | int offset, int len) | |
1548 | { | |
e0506bcb IE |
1549 | int r; |
1550 | unsigned long addr; | |
195aefde | 1551 | |
e0506bcb IE |
1552 | addr = gfn_to_hva(kvm, gfn); |
1553 | if (kvm_is_error_hva(addr)) | |
1554 | return -EFAULT; | |
1555 | r = copy_to_user((void __user *)addr + offset, data, len); | |
1556 | if (r) | |
195aefde | 1557 | return -EFAULT; |
195aefde IE |
1558 | mark_page_dirty(kvm, gfn); |
1559 | return 0; | |
1560 | } | |
1561 | EXPORT_SYMBOL_GPL(kvm_write_guest_page); | |
1562 | ||
1563 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
1564 | unsigned long len) | |
1565 | { | |
1566 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1567 | int seg; | |
1568 | int offset = offset_in_page(gpa); | |
1569 | int ret; | |
1570 | ||
1571 | while ((seg = next_segment(len, offset)) != 0) { | |
1572 | ret = kvm_write_guest_page(kvm, gfn, data, offset, seg); | |
1573 | if (ret < 0) | |
1574 | return ret; | |
1575 | offset = 0; | |
1576 | len -= seg; | |
1577 | data += seg; | |
1578 | ++gfn; | |
1579 | } | |
1580 | return 0; | |
1581 | } | |
1582 | ||
1583 | int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len) | |
1584 | { | |
3e021bf5 | 1585 | return kvm_write_guest_page(kvm, gfn, empty_zero_page, offset, len); |
195aefde IE |
1586 | } |
1587 | EXPORT_SYMBOL_GPL(kvm_clear_guest_page); | |
1588 | ||
1589 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len) | |
1590 | { | |
1591 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1592 | int seg; | |
1593 | int offset = offset_in_page(gpa); | |
1594 | int ret; | |
1595 | ||
1596 | while ((seg = next_segment(len, offset)) != 0) { | |
1597 | ret = kvm_clear_guest_page(kvm, gfn, offset, seg); | |
1598 | if (ret < 0) | |
1599 | return ret; | |
1600 | offset = 0; | |
1601 | len -= seg; | |
1602 | ++gfn; | |
1603 | } | |
1604 | return 0; | |
1605 | } | |
1606 | EXPORT_SYMBOL_GPL(kvm_clear_guest); | |
1607 | ||
6aa8b732 AK |
1608 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
1609 | { | |
31389947 | 1610 | struct kvm_memory_slot *memslot; |
6aa8b732 | 1611 | |
3b6fff19 | 1612 | gfn = unalias_gfn(kvm, gfn); |
2843099f | 1613 | memslot = gfn_to_memslot_unaliased(kvm, gfn); |
7e9d619d RR |
1614 | if (memslot && memslot->dirty_bitmap) { |
1615 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 1616 | |
7e9d619d RR |
1617 | /* avoid RMW */ |
1618 | if (!test_bit(rel_gfn, memslot->dirty_bitmap)) | |
1619 | set_bit(rel_gfn, memslot->dirty_bitmap); | |
6aa8b732 AK |
1620 | } |
1621 | } | |
1622 | ||
b6958ce4 ED |
1623 | /* |
1624 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
1625 | */ | |
8776e519 | 1626 | void kvm_vcpu_block(struct kvm_vcpu *vcpu) |
d3bef15f | 1627 | { |
e5c239cf MT |
1628 | DEFINE_WAIT(wait); |
1629 | ||
1630 | for (;;) { | |
1631 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1632 | ||
78646121 GN |
1633 | if ((kvm_arch_interrupt_allowed(vcpu) && |
1634 | kvm_cpu_has_interrupt(vcpu)) || | |
09cec754 | 1635 | kvm_arch_vcpu_runnable(vcpu)) { |
d7690175 | 1636 | set_bit(KVM_REQ_UNHALT, &vcpu->requests); |
e5c239cf | 1637 | break; |
d7690175 | 1638 | } |
09cec754 GN |
1639 | if (kvm_cpu_has_pending_timer(vcpu)) |
1640 | break; | |
e5c239cf MT |
1641 | if (signal_pending(current)) |
1642 | break; | |
1643 | ||
b6958ce4 ED |
1644 | vcpu_put(vcpu); |
1645 | schedule(); | |
1646 | vcpu_load(vcpu); | |
1647 | } | |
d3bef15f | 1648 | |
e5c239cf | 1649 | finish_wait(&vcpu->wq, &wait); |
b6958ce4 ED |
1650 | } |
1651 | ||
6aa8b732 AK |
1652 | void kvm_resched(struct kvm_vcpu *vcpu) |
1653 | { | |
3fca0365 YD |
1654 | if (!need_resched()) |
1655 | return; | |
6aa8b732 | 1656 | cond_resched(); |
6aa8b732 AK |
1657 | } |
1658 | EXPORT_SYMBOL_GPL(kvm_resched); | |
1659 | ||
e4a533a4 | 1660 | static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
9a2bb7f4 AK |
1661 | { |
1662 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
9a2bb7f4 AK |
1663 | struct page *page; |
1664 | ||
e4a533a4 | 1665 | if (vmf->pgoff == 0) |
039576c0 | 1666 | page = virt_to_page(vcpu->run); |
09566765 | 1667 | #ifdef CONFIG_X86 |
e4a533a4 | 1668 | else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET) |
ad312c7c | 1669 | page = virt_to_page(vcpu->arch.pio_data); |
5f94c174 LV |
1670 | #endif |
1671 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
1672 | else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET) | |
1673 | page = virt_to_page(vcpu->kvm->coalesced_mmio_ring); | |
09566765 | 1674 | #endif |
039576c0 | 1675 | else |
e4a533a4 | 1676 | return VM_FAULT_SIGBUS; |
9a2bb7f4 | 1677 | get_page(page); |
e4a533a4 | 1678 | vmf->page = page; |
1679 | return 0; | |
9a2bb7f4 AK |
1680 | } |
1681 | ||
1682 | static struct vm_operations_struct kvm_vcpu_vm_ops = { | |
e4a533a4 | 1683 | .fault = kvm_vcpu_fault, |
9a2bb7f4 AK |
1684 | }; |
1685 | ||
1686 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
1687 | { | |
1688 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
1689 | return 0; | |
1690 | } | |
1691 | ||
bccf2150 AK |
1692 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
1693 | { | |
1694 | struct kvm_vcpu *vcpu = filp->private_data; | |
1695 | ||
66c0b394 | 1696 | kvm_put_kvm(vcpu->kvm); |
bccf2150 AK |
1697 | return 0; |
1698 | } | |
1699 | ||
3d3aab1b | 1700 | static struct file_operations kvm_vcpu_fops = { |
bccf2150 AK |
1701 | .release = kvm_vcpu_release, |
1702 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
1703 | .compat_ioctl = kvm_vcpu_ioctl, | |
9a2bb7f4 | 1704 | .mmap = kvm_vcpu_mmap, |
bccf2150 AK |
1705 | }; |
1706 | ||
1707 | /* | |
1708 | * Allocates an inode for the vcpu. | |
1709 | */ | |
1710 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
1711 | { | |
7d9dbca3 | 1712 | int fd = anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, 0); |
2030a42c | 1713 | if (fd < 0) |
66c0b394 | 1714 | kvm_put_kvm(vcpu->kvm); |
bccf2150 | 1715 | return fd; |
bccf2150 AK |
1716 | } |
1717 | ||
c5ea7660 AK |
1718 | /* |
1719 | * Creates some virtual cpus. Good luck creating more than one. | |
1720 | */ | |
1721 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, int n) | |
1722 | { | |
1723 | int r; | |
1724 | struct kvm_vcpu *vcpu; | |
1725 | ||
c5ea7660 | 1726 | if (!valid_vcpu(n)) |
fb3f0f51 | 1727 | return -EINVAL; |
c5ea7660 | 1728 | |
e9b11c17 | 1729 | vcpu = kvm_arch_vcpu_create(kvm, n); |
fb3f0f51 RR |
1730 | if (IS_ERR(vcpu)) |
1731 | return PTR_ERR(vcpu); | |
c5ea7660 | 1732 | |
15ad7146 AK |
1733 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
1734 | ||
26e5215f AK |
1735 | r = kvm_arch_vcpu_setup(vcpu); |
1736 | if (r) | |
7d8fece6 | 1737 | return r; |
26e5215f | 1738 | |
11ec2804 | 1739 | mutex_lock(&kvm->lock); |
fb3f0f51 RR |
1740 | if (kvm->vcpus[n]) { |
1741 | r = -EEXIST; | |
e9b11c17 | 1742 | goto vcpu_destroy; |
fb3f0f51 RR |
1743 | } |
1744 | kvm->vcpus[n] = vcpu; | |
11ec2804 | 1745 | mutex_unlock(&kvm->lock); |
c5ea7660 | 1746 | |
fb3f0f51 | 1747 | /* Now it's all set up, let userspace reach it */ |
66c0b394 | 1748 | kvm_get_kvm(kvm); |
bccf2150 AK |
1749 | r = create_vcpu_fd(vcpu); |
1750 | if (r < 0) | |
fb3f0f51 RR |
1751 | goto unlink; |
1752 | return r; | |
39c3b86e | 1753 | |
fb3f0f51 | 1754 | unlink: |
11ec2804 | 1755 | mutex_lock(&kvm->lock); |
fb3f0f51 | 1756 | kvm->vcpus[n] = NULL; |
e9b11c17 | 1757 | vcpu_destroy: |
7d8fece6 | 1758 | mutex_unlock(&kvm->lock); |
d40ccc62 | 1759 | kvm_arch_vcpu_destroy(vcpu); |
c5ea7660 AK |
1760 | return r; |
1761 | } | |
1762 | ||
1961d276 AK |
1763 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
1764 | { | |
1765 | if (sigset) { | |
1766 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
1767 | vcpu->sigset_active = 1; | |
1768 | vcpu->sigset = *sigset; | |
1769 | } else | |
1770 | vcpu->sigset_active = 0; | |
1771 | return 0; | |
1772 | } | |
1773 | ||
c1e01514 SY |
1774 | #ifdef __KVM_HAVE_MSIX |
1775 | static int kvm_vm_ioctl_set_msix_nr(struct kvm *kvm, | |
1776 | struct kvm_assigned_msix_nr *entry_nr) | |
1777 | { | |
1778 | int r = 0; | |
1779 | struct kvm_assigned_dev_kernel *adev; | |
1780 | ||
1781 | mutex_lock(&kvm->lock); | |
1782 | ||
1783 | adev = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, | |
1784 | entry_nr->assigned_dev_id); | |
1785 | if (!adev) { | |
1786 | r = -EINVAL; | |
1787 | goto msix_nr_out; | |
1788 | } | |
1789 | ||
1790 | if (adev->entries_nr == 0) { | |
1791 | adev->entries_nr = entry_nr->entry_nr; | |
1792 | if (adev->entries_nr == 0 || | |
1793 | adev->entries_nr >= KVM_MAX_MSIX_PER_DEV) { | |
1794 | r = -EINVAL; | |
1795 | goto msix_nr_out; | |
1796 | } | |
1797 | ||
1798 | adev->host_msix_entries = kzalloc(sizeof(struct msix_entry) * | |
1799 | entry_nr->entry_nr, | |
1800 | GFP_KERNEL); | |
1801 | if (!adev->host_msix_entries) { | |
1802 | r = -ENOMEM; | |
1803 | goto msix_nr_out; | |
1804 | } | |
1805 | adev->guest_msix_entries = kzalloc( | |
1806 | sizeof(struct kvm_guest_msix_entry) * | |
1807 | entry_nr->entry_nr, GFP_KERNEL); | |
1808 | if (!adev->guest_msix_entries) { | |
1809 | kfree(adev->host_msix_entries); | |
1810 | r = -ENOMEM; | |
1811 | goto msix_nr_out; | |
1812 | } | |
1813 | } else /* Not allowed set MSI-X number twice */ | |
1814 | r = -EINVAL; | |
1815 | msix_nr_out: | |
1816 | mutex_unlock(&kvm->lock); | |
1817 | return r; | |
1818 | } | |
1819 | ||
1820 | static int kvm_vm_ioctl_set_msix_entry(struct kvm *kvm, | |
1821 | struct kvm_assigned_msix_entry *entry) | |
1822 | { | |
1823 | int r = 0, i; | |
1824 | struct kvm_assigned_dev_kernel *adev; | |
1825 | ||
1826 | mutex_lock(&kvm->lock); | |
1827 | ||
1828 | adev = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, | |
1829 | entry->assigned_dev_id); | |
1830 | ||
1831 | if (!adev) { | |
1832 | r = -EINVAL; | |
1833 | goto msix_entry_out; | |
1834 | } | |
1835 | ||
1836 | for (i = 0; i < adev->entries_nr; i++) | |
1837 | if (adev->guest_msix_entries[i].vector == 0 || | |
1838 | adev->guest_msix_entries[i].entry == entry->entry) { | |
1839 | adev->guest_msix_entries[i].entry = entry->entry; | |
1840 | adev->guest_msix_entries[i].vector = entry->gsi; | |
1841 | adev->host_msix_entries[i].entry = entry->entry; | |
1842 | break; | |
1843 | } | |
1844 | if (i == adev->entries_nr) { | |
1845 | r = -ENOSPC; | |
1846 | goto msix_entry_out; | |
1847 | } | |
1848 | ||
1849 | msix_entry_out: | |
1850 | mutex_unlock(&kvm->lock); | |
1851 | ||
1852 | return r; | |
1853 | } | |
1854 | #endif | |
1855 | ||
bccf2150 AK |
1856 | static long kvm_vcpu_ioctl(struct file *filp, |
1857 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 1858 | { |
bccf2150 | 1859 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 1860 | void __user *argp = (void __user *)arg; |
313a3dc7 | 1861 | int r; |
fa3795a7 DH |
1862 | struct kvm_fpu *fpu = NULL; |
1863 | struct kvm_sregs *kvm_sregs = NULL; | |
6aa8b732 | 1864 | |
6d4e4c4f AK |
1865 | if (vcpu->kvm->mm != current->mm) |
1866 | return -EIO; | |
6aa8b732 | 1867 | switch (ioctl) { |
9a2bb7f4 | 1868 | case KVM_RUN: |
f0fe5108 AK |
1869 | r = -EINVAL; |
1870 | if (arg) | |
1871 | goto out; | |
b6c7a5dc | 1872 | r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run); |
6aa8b732 | 1873 | break; |
6aa8b732 | 1874 | case KVM_GET_REGS: { |
3e4bb3ac | 1875 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1876 | |
3e4bb3ac XZ |
1877 | r = -ENOMEM; |
1878 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1879 | if (!kvm_regs) | |
6aa8b732 | 1880 | goto out; |
3e4bb3ac XZ |
1881 | r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs); |
1882 | if (r) | |
1883 | goto out_free1; | |
6aa8b732 | 1884 | r = -EFAULT; |
3e4bb3ac XZ |
1885 | if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs))) |
1886 | goto out_free1; | |
6aa8b732 | 1887 | r = 0; |
3e4bb3ac XZ |
1888 | out_free1: |
1889 | kfree(kvm_regs); | |
6aa8b732 AK |
1890 | break; |
1891 | } | |
1892 | case KVM_SET_REGS: { | |
3e4bb3ac | 1893 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1894 | |
3e4bb3ac XZ |
1895 | r = -ENOMEM; |
1896 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1897 | if (!kvm_regs) | |
6aa8b732 | 1898 | goto out; |
3e4bb3ac XZ |
1899 | r = -EFAULT; |
1900 | if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs))) | |
1901 | goto out_free2; | |
1902 | r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs); | |
6aa8b732 | 1903 | if (r) |
3e4bb3ac | 1904 | goto out_free2; |
6aa8b732 | 1905 | r = 0; |
3e4bb3ac XZ |
1906 | out_free2: |
1907 | kfree(kvm_regs); | |
6aa8b732 AK |
1908 | break; |
1909 | } | |
1910 | case KVM_GET_SREGS: { | |
fa3795a7 DH |
1911 | kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1912 | r = -ENOMEM; | |
1913 | if (!kvm_sregs) | |
1914 | goto out; | |
1915 | r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs); | |
6aa8b732 AK |
1916 | if (r) |
1917 | goto out; | |
1918 | r = -EFAULT; | |
fa3795a7 | 1919 | if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs))) |
6aa8b732 AK |
1920 | goto out; |
1921 | r = 0; | |
1922 | break; | |
1923 | } | |
1924 | case KVM_SET_SREGS: { | |
fa3795a7 DH |
1925 | kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1926 | r = -ENOMEM; | |
1927 | if (!kvm_sregs) | |
1928 | goto out; | |
6aa8b732 | 1929 | r = -EFAULT; |
fa3795a7 | 1930 | if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs))) |
6aa8b732 | 1931 | goto out; |
fa3795a7 | 1932 | r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs); |
6aa8b732 AK |
1933 | if (r) |
1934 | goto out; | |
1935 | r = 0; | |
1936 | break; | |
1937 | } | |
62d9f0db MT |
1938 | case KVM_GET_MP_STATE: { |
1939 | struct kvm_mp_state mp_state; | |
1940 | ||
1941 | r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state); | |
1942 | if (r) | |
1943 | goto out; | |
1944 | r = -EFAULT; | |
1945 | if (copy_to_user(argp, &mp_state, sizeof mp_state)) | |
1946 | goto out; | |
1947 | r = 0; | |
1948 | break; | |
1949 | } | |
1950 | case KVM_SET_MP_STATE: { | |
1951 | struct kvm_mp_state mp_state; | |
1952 | ||
1953 | r = -EFAULT; | |
1954 | if (copy_from_user(&mp_state, argp, sizeof mp_state)) | |
1955 | goto out; | |
1956 | r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state); | |
1957 | if (r) | |
1958 | goto out; | |
1959 | r = 0; | |
1960 | break; | |
1961 | } | |
6aa8b732 AK |
1962 | case KVM_TRANSLATE: { |
1963 | struct kvm_translation tr; | |
1964 | ||
1965 | r = -EFAULT; | |
2f366987 | 1966 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 1967 | goto out; |
8b006791 | 1968 | r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
1969 | if (r) |
1970 | goto out; | |
1971 | r = -EFAULT; | |
2f366987 | 1972 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
1973 | goto out; |
1974 | r = 0; | |
1975 | break; | |
1976 | } | |
d0bfb940 JK |
1977 | case KVM_SET_GUEST_DEBUG: { |
1978 | struct kvm_guest_debug dbg; | |
6aa8b732 AK |
1979 | |
1980 | r = -EFAULT; | |
2f366987 | 1981 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 1982 | goto out; |
d0bfb940 | 1983 | r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg); |
6aa8b732 AK |
1984 | if (r) |
1985 | goto out; | |
1986 | r = 0; | |
1987 | break; | |
1988 | } | |
1961d276 AK |
1989 | case KVM_SET_SIGNAL_MASK: { |
1990 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
1991 | struct kvm_signal_mask kvm_sigmask; | |
1992 | sigset_t sigset, *p; | |
1993 | ||
1994 | p = NULL; | |
1995 | if (argp) { | |
1996 | r = -EFAULT; | |
1997 | if (copy_from_user(&kvm_sigmask, argp, | |
1998 | sizeof kvm_sigmask)) | |
1999 | goto out; | |
2000 | r = -EINVAL; | |
2001 | if (kvm_sigmask.len != sizeof sigset) | |
2002 | goto out; | |
2003 | r = -EFAULT; | |
2004 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
2005 | sizeof sigset)) | |
2006 | goto out; | |
2007 | p = &sigset; | |
2008 | } | |
2009 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset); | |
2010 | break; | |
2011 | } | |
b8836737 | 2012 | case KVM_GET_FPU: { |
fa3795a7 DH |
2013 | fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
2014 | r = -ENOMEM; | |
2015 | if (!fpu) | |
2016 | goto out; | |
2017 | r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu); | |
b8836737 AK |
2018 | if (r) |
2019 | goto out; | |
2020 | r = -EFAULT; | |
fa3795a7 | 2021 | if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu))) |
b8836737 AK |
2022 | goto out; |
2023 | r = 0; | |
2024 | break; | |
2025 | } | |
2026 | case KVM_SET_FPU: { | |
fa3795a7 DH |
2027 | fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
2028 | r = -ENOMEM; | |
2029 | if (!fpu) | |
2030 | goto out; | |
b8836737 | 2031 | r = -EFAULT; |
fa3795a7 | 2032 | if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu))) |
b8836737 | 2033 | goto out; |
fa3795a7 | 2034 | r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu); |
b8836737 AK |
2035 | if (r) |
2036 | goto out; | |
2037 | r = 0; | |
2038 | break; | |
2039 | } | |
bccf2150 | 2040 | default: |
313a3dc7 | 2041 | r = kvm_arch_vcpu_ioctl(filp, ioctl, arg); |
bccf2150 AK |
2042 | } |
2043 | out: | |
fa3795a7 DH |
2044 | kfree(fpu); |
2045 | kfree(kvm_sregs); | |
bccf2150 AK |
2046 | return r; |
2047 | } | |
2048 | ||
2049 | static long kvm_vm_ioctl(struct file *filp, | |
2050 | unsigned int ioctl, unsigned long arg) | |
2051 | { | |
2052 | struct kvm *kvm = filp->private_data; | |
2053 | void __user *argp = (void __user *)arg; | |
1fe779f8 | 2054 | int r; |
bccf2150 | 2055 | |
6d4e4c4f AK |
2056 | if (kvm->mm != current->mm) |
2057 | return -EIO; | |
bccf2150 AK |
2058 | switch (ioctl) { |
2059 | case KVM_CREATE_VCPU: | |
2060 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
2061 | if (r < 0) | |
2062 | goto out; | |
2063 | break; | |
6fc138d2 IE |
2064 | case KVM_SET_USER_MEMORY_REGION: { |
2065 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
2066 | ||
2067 | r = -EFAULT; | |
2068 | if (copy_from_user(&kvm_userspace_mem, argp, | |
2069 | sizeof kvm_userspace_mem)) | |
2070 | goto out; | |
2071 | ||
2072 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1); | |
6aa8b732 AK |
2073 | if (r) |
2074 | goto out; | |
2075 | break; | |
2076 | } | |
2077 | case KVM_GET_DIRTY_LOG: { | |
2078 | struct kvm_dirty_log log; | |
2079 | ||
2080 | r = -EFAULT; | |
2f366987 | 2081 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 2082 | goto out; |
2c6f5df9 | 2083 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
2084 | if (r) |
2085 | goto out; | |
2086 | break; | |
2087 | } | |
5f94c174 LV |
2088 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
2089 | case KVM_REGISTER_COALESCED_MMIO: { | |
2090 | struct kvm_coalesced_mmio_zone zone; | |
2091 | r = -EFAULT; | |
2092 | if (copy_from_user(&zone, argp, sizeof zone)) | |
2093 | goto out; | |
2094 | r = -ENXIO; | |
2095 | r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone); | |
2096 | if (r) | |
2097 | goto out; | |
2098 | r = 0; | |
2099 | break; | |
2100 | } | |
2101 | case KVM_UNREGISTER_COALESCED_MMIO: { | |
2102 | struct kvm_coalesced_mmio_zone zone; | |
2103 | r = -EFAULT; | |
2104 | if (copy_from_user(&zone, argp, sizeof zone)) | |
2105 | goto out; | |
2106 | r = -ENXIO; | |
2107 | r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone); | |
2108 | if (r) | |
2109 | goto out; | |
2110 | r = 0; | |
2111 | break; | |
2112 | } | |
8a98f664 XZ |
2113 | #endif |
2114 | #ifdef KVM_CAP_DEVICE_ASSIGNMENT | |
2115 | case KVM_ASSIGN_PCI_DEVICE: { | |
2116 | struct kvm_assigned_pci_dev assigned_dev; | |
2117 | ||
2118 | r = -EFAULT; | |
2119 | if (copy_from_user(&assigned_dev, argp, sizeof assigned_dev)) | |
2120 | goto out; | |
2121 | r = kvm_vm_ioctl_assign_device(kvm, &assigned_dev); | |
2122 | if (r) | |
2123 | goto out; | |
2124 | break; | |
2125 | } | |
2126 | case KVM_ASSIGN_IRQ: { | |
e56d532f SY |
2127 | r = -EOPNOTSUPP; |
2128 | break; | |
2129 | } | |
2130 | #ifdef KVM_CAP_ASSIGN_DEV_IRQ | |
2131 | case KVM_ASSIGN_DEV_IRQ: { | |
8a98f664 XZ |
2132 | struct kvm_assigned_irq assigned_irq; |
2133 | ||
2134 | r = -EFAULT; | |
2135 | if (copy_from_user(&assigned_irq, argp, sizeof assigned_irq)) | |
2136 | goto out; | |
2137 | r = kvm_vm_ioctl_assign_irq(kvm, &assigned_irq); | |
2138 | if (r) | |
2139 | goto out; | |
2140 | break; | |
2141 | } | |
e56d532f SY |
2142 | case KVM_DEASSIGN_DEV_IRQ: { |
2143 | struct kvm_assigned_irq assigned_irq; | |
2144 | ||
2145 | r = -EFAULT; | |
2146 | if (copy_from_user(&assigned_irq, argp, sizeof assigned_irq)) | |
2147 | goto out; | |
2148 | r = kvm_vm_ioctl_deassign_dev_irq(kvm, &assigned_irq); | |
2149 | if (r) | |
2150 | goto out; | |
2151 | break; | |
2152 | } | |
2153 | #endif | |
0a920356 WH |
2154 | #endif |
2155 | #ifdef KVM_CAP_DEVICE_DEASSIGNMENT | |
2156 | case KVM_DEASSIGN_PCI_DEVICE: { | |
2157 | struct kvm_assigned_pci_dev assigned_dev; | |
2158 | ||
2159 | r = -EFAULT; | |
2160 | if (copy_from_user(&assigned_dev, argp, sizeof assigned_dev)) | |
2161 | goto out; | |
2162 | r = kvm_vm_ioctl_deassign_device(kvm, &assigned_dev); | |
2163 | if (r) | |
2164 | goto out; | |
2165 | break; | |
2166 | } | |
399ec807 AK |
2167 | #endif |
2168 | #ifdef KVM_CAP_IRQ_ROUTING | |
2169 | case KVM_SET_GSI_ROUTING: { | |
2170 | struct kvm_irq_routing routing; | |
2171 | struct kvm_irq_routing __user *urouting; | |
2172 | struct kvm_irq_routing_entry *entries; | |
2173 | ||
2174 | r = -EFAULT; | |
2175 | if (copy_from_user(&routing, argp, sizeof(routing))) | |
2176 | goto out; | |
2177 | r = -EINVAL; | |
2178 | if (routing.nr >= KVM_MAX_IRQ_ROUTES) | |
2179 | goto out; | |
2180 | if (routing.flags) | |
2181 | goto out; | |
2182 | r = -ENOMEM; | |
2183 | entries = vmalloc(routing.nr * sizeof(*entries)); | |
2184 | if (!entries) | |
2185 | goto out; | |
2186 | r = -EFAULT; | |
2187 | urouting = argp; | |
2188 | if (copy_from_user(entries, urouting->entries, | |
2189 | routing.nr * sizeof(*entries))) | |
2190 | goto out_free_irq_routing; | |
2191 | r = kvm_set_irq_routing(kvm, entries, routing.nr, | |
2192 | routing.flags); | |
2193 | out_free_irq_routing: | |
2194 | vfree(entries); | |
2195 | break; | |
2196 | } | |
c1e01514 SY |
2197 | #ifdef __KVM_HAVE_MSIX |
2198 | case KVM_ASSIGN_SET_MSIX_NR: { | |
2199 | struct kvm_assigned_msix_nr entry_nr; | |
2200 | r = -EFAULT; | |
2201 | if (copy_from_user(&entry_nr, argp, sizeof entry_nr)) | |
2202 | goto out; | |
2203 | r = kvm_vm_ioctl_set_msix_nr(kvm, &entry_nr); | |
2204 | if (r) | |
2205 | goto out; | |
2206 | break; | |
2207 | } | |
2208 | case KVM_ASSIGN_SET_MSIX_ENTRY: { | |
2209 | struct kvm_assigned_msix_entry entry; | |
2210 | r = -EFAULT; | |
2211 | if (copy_from_user(&entry, argp, sizeof entry)) | |
2212 | goto out; | |
2213 | r = kvm_vm_ioctl_set_msix_entry(kvm, &entry); | |
2214 | if (r) | |
2215 | goto out; | |
2216 | break; | |
2217 | } | |
5f94c174 | 2218 | #endif |
c1e01514 | 2219 | #endif /* KVM_CAP_IRQ_ROUTING */ |
f17abe9a | 2220 | default: |
1fe779f8 | 2221 | r = kvm_arch_vm_ioctl(filp, ioctl, arg); |
f17abe9a AK |
2222 | } |
2223 | out: | |
2224 | return r; | |
2225 | } | |
2226 | ||
e4a533a4 | 2227 | static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
f17abe9a | 2228 | { |
777b3f49 MT |
2229 | struct page *page[1]; |
2230 | unsigned long addr; | |
2231 | int npages; | |
2232 | gfn_t gfn = vmf->pgoff; | |
f17abe9a | 2233 | struct kvm *kvm = vma->vm_file->private_data; |
f17abe9a | 2234 | |
777b3f49 MT |
2235 | addr = gfn_to_hva(kvm, gfn); |
2236 | if (kvm_is_error_hva(addr)) | |
e4a533a4 | 2237 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
2238 | |
2239 | npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page, | |
2240 | NULL); | |
2241 | if (unlikely(npages != 1)) | |
e4a533a4 | 2242 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
2243 | |
2244 | vmf->page = page[0]; | |
e4a533a4 | 2245 | return 0; |
f17abe9a AK |
2246 | } |
2247 | ||
2248 | static struct vm_operations_struct kvm_vm_vm_ops = { | |
e4a533a4 | 2249 | .fault = kvm_vm_fault, |
f17abe9a AK |
2250 | }; |
2251 | ||
2252 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
2253 | { | |
2254 | vma->vm_ops = &kvm_vm_vm_ops; | |
2255 | return 0; | |
2256 | } | |
2257 | ||
3d3aab1b | 2258 | static struct file_operations kvm_vm_fops = { |
f17abe9a AK |
2259 | .release = kvm_vm_release, |
2260 | .unlocked_ioctl = kvm_vm_ioctl, | |
2261 | .compat_ioctl = kvm_vm_ioctl, | |
2262 | .mmap = kvm_vm_mmap, | |
2263 | }; | |
2264 | ||
2265 | static int kvm_dev_ioctl_create_vm(void) | |
2266 | { | |
2030a42c | 2267 | int fd; |
f17abe9a AK |
2268 | struct kvm *kvm; |
2269 | ||
f17abe9a | 2270 | kvm = kvm_create_vm(); |
d6d28168 AK |
2271 | if (IS_ERR(kvm)) |
2272 | return PTR_ERR(kvm); | |
7d9dbca3 | 2273 | fd = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, 0); |
2030a42c | 2274 | if (fd < 0) |
66c0b394 | 2275 | kvm_put_kvm(kvm); |
f17abe9a | 2276 | |
f17abe9a | 2277 | return fd; |
f17abe9a AK |
2278 | } |
2279 | ||
1a811b61 AK |
2280 | static long kvm_dev_ioctl_check_extension_generic(long arg) |
2281 | { | |
2282 | switch (arg) { | |
ca9edaee | 2283 | case KVM_CAP_USER_MEMORY: |
1a811b61 | 2284 | case KVM_CAP_DESTROY_MEMORY_REGION_WORKS: |
4cd481f6 | 2285 | case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS: |
1a811b61 | 2286 | return 1; |
399ec807 AK |
2287 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
2288 | case KVM_CAP_IRQ_ROUTING: | |
36463146 | 2289 | return KVM_MAX_IRQ_ROUTES; |
399ec807 | 2290 | #endif |
1a811b61 AK |
2291 | default: |
2292 | break; | |
2293 | } | |
2294 | return kvm_dev_ioctl_check_extension(arg); | |
2295 | } | |
2296 | ||
f17abe9a AK |
2297 | static long kvm_dev_ioctl(struct file *filp, |
2298 | unsigned int ioctl, unsigned long arg) | |
2299 | { | |
07c45a36 | 2300 | long r = -EINVAL; |
f17abe9a AK |
2301 | |
2302 | switch (ioctl) { | |
2303 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
2304 | r = -EINVAL; |
2305 | if (arg) | |
2306 | goto out; | |
f17abe9a AK |
2307 | r = KVM_API_VERSION; |
2308 | break; | |
2309 | case KVM_CREATE_VM: | |
f0fe5108 AK |
2310 | r = -EINVAL; |
2311 | if (arg) | |
2312 | goto out; | |
f17abe9a AK |
2313 | r = kvm_dev_ioctl_create_vm(); |
2314 | break; | |
018d00d2 | 2315 | case KVM_CHECK_EXTENSION: |
1a811b61 | 2316 | r = kvm_dev_ioctl_check_extension_generic(arg); |
5d308f45 | 2317 | break; |
07c45a36 AK |
2318 | case KVM_GET_VCPU_MMAP_SIZE: |
2319 | r = -EINVAL; | |
2320 | if (arg) | |
2321 | goto out; | |
adb1ff46 AK |
2322 | r = PAGE_SIZE; /* struct kvm_run */ |
2323 | #ifdef CONFIG_X86 | |
2324 | r += PAGE_SIZE; /* pio data page */ | |
5f94c174 LV |
2325 | #endif |
2326 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
2327 | r += PAGE_SIZE; /* coalesced mmio ring page */ | |
adb1ff46 | 2328 | #endif |
07c45a36 | 2329 | break; |
d4c9ff2d FEL |
2330 | case KVM_TRACE_ENABLE: |
2331 | case KVM_TRACE_PAUSE: | |
2332 | case KVM_TRACE_DISABLE: | |
2333 | r = kvm_trace_ioctl(ioctl, arg); | |
2334 | break; | |
6aa8b732 | 2335 | default: |
043405e1 | 2336 | return kvm_arch_dev_ioctl(filp, ioctl, arg); |
6aa8b732 AK |
2337 | } |
2338 | out: | |
2339 | return r; | |
2340 | } | |
2341 | ||
6aa8b732 | 2342 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
2343 | .unlocked_ioctl = kvm_dev_ioctl, |
2344 | .compat_ioctl = kvm_dev_ioctl, | |
6aa8b732 AK |
2345 | }; |
2346 | ||
2347 | static struct miscdevice kvm_dev = { | |
bbe4432e | 2348 | KVM_MINOR, |
6aa8b732 AK |
2349 | "kvm", |
2350 | &kvm_chardev_ops, | |
2351 | }; | |
2352 | ||
1b6c0168 AK |
2353 | static void hardware_enable(void *junk) |
2354 | { | |
2355 | int cpu = raw_smp_processor_id(); | |
2356 | ||
7f59f492 | 2357 | if (cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 2358 | return; |
7f59f492 | 2359 | cpumask_set_cpu(cpu, cpus_hardware_enabled); |
e9b11c17 | 2360 | kvm_arch_hardware_enable(NULL); |
1b6c0168 AK |
2361 | } |
2362 | ||
2363 | static void hardware_disable(void *junk) | |
2364 | { | |
2365 | int cpu = raw_smp_processor_id(); | |
2366 | ||
7f59f492 | 2367 | if (!cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 2368 | return; |
7f59f492 | 2369 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); |
e9b11c17 | 2370 | kvm_arch_hardware_disable(NULL); |
1b6c0168 AK |
2371 | } |
2372 | ||
774c47f1 AK |
2373 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
2374 | void *v) | |
2375 | { | |
2376 | int cpu = (long)v; | |
2377 | ||
1a6f4d7f | 2378 | val &= ~CPU_TASKS_FROZEN; |
774c47f1 | 2379 | switch (val) { |
cec9ad27 | 2380 | case CPU_DYING: |
6ec8a856 AK |
2381 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
2382 | cpu); | |
2383 | hardware_disable(NULL); | |
2384 | break; | |
774c47f1 | 2385 | case CPU_UP_CANCELED: |
43934a38 JK |
2386 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
2387 | cpu); | |
8691e5a8 | 2388 | smp_call_function_single(cpu, hardware_disable, NULL, 1); |
774c47f1 | 2389 | break; |
43934a38 JK |
2390 | case CPU_ONLINE: |
2391 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", | |
2392 | cpu); | |
8691e5a8 | 2393 | smp_call_function_single(cpu, hardware_enable, NULL, 1); |
774c47f1 AK |
2394 | break; |
2395 | } | |
2396 | return NOTIFY_OK; | |
2397 | } | |
2398 | ||
4ecac3fd AK |
2399 | |
2400 | asmlinkage void kvm_handle_fault_on_reboot(void) | |
2401 | { | |
2402 | if (kvm_rebooting) | |
2403 | /* spin while reset goes on */ | |
2404 | while (true) | |
2405 | ; | |
2406 | /* Fault while not rebooting. We want the trace. */ | |
2407 | BUG(); | |
2408 | } | |
2409 | EXPORT_SYMBOL_GPL(kvm_handle_fault_on_reboot); | |
2410 | ||
9a2b85c6 | 2411 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
d77c26fc | 2412 | void *v) |
9a2b85c6 | 2413 | { |
8e1c1815 SY |
2414 | /* |
2415 | * Some (well, at least mine) BIOSes hang on reboot if | |
2416 | * in vmx root mode. | |
2417 | * | |
2418 | * And Intel TXT required VMX off for all cpu when system shutdown. | |
2419 | */ | |
2420 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
2421 | kvm_rebooting = true; | |
2422 | on_each_cpu(hardware_disable, NULL, 1); | |
9a2b85c6 RR |
2423 | return NOTIFY_OK; |
2424 | } | |
2425 | ||
2426 | static struct notifier_block kvm_reboot_notifier = { | |
2427 | .notifier_call = kvm_reboot, | |
2428 | .priority = 0, | |
2429 | }; | |
2430 | ||
2eeb2e94 GH |
2431 | void kvm_io_bus_init(struct kvm_io_bus *bus) |
2432 | { | |
2433 | memset(bus, 0, sizeof(*bus)); | |
2434 | } | |
2435 | ||
2436 | void kvm_io_bus_destroy(struct kvm_io_bus *bus) | |
2437 | { | |
2438 | int i; | |
2439 | ||
2440 | for (i = 0; i < bus->dev_count; i++) { | |
2441 | struct kvm_io_device *pos = bus->devs[i]; | |
2442 | ||
2443 | kvm_iodevice_destructor(pos); | |
2444 | } | |
2445 | } | |
2446 | ||
92760499 LV |
2447 | struct kvm_io_device *kvm_io_bus_find_dev(struct kvm_io_bus *bus, |
2448 | gpa_t addr, int len, int is_write) | |
2eeb2e94 GH |
2449 | { |
2450 | int i; | |
2451 | ||
2452 | for (i = 0; i < bus->dev_count; i++) { | |
2453 | struct kvm_io_device *pos = bus->devs[i]; | |
2454 | ||
92760499 | 2455 | if (pos->in_range(pos, addr, len, is_write)) |
2eeb2e94 GH |
2456 | return pos; |
2457 | } | |
2458 | ||
2459 | return NULL; | |
2460 | } | |
2461 | ||
2462 | void kvm_io_bus_register_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev) | |
2463 | { | |
2464 | BUG_ON(bus->dev_count > (NR_IOBUS_DEVS-1)); | |
2465 | ||
2466 | bus->devs[bus->dev_count++] = dev; | |
2467 | } | |
2468 | ||
774c47f1 AK |
2469 | static struct notifier_block kvm_cpu_notifier = { |
2470 | .notifier_call = kvm_cpu_hotplug, | |
2471 | .priority = 20, /* must be > scheduler priority */ | |
2472 | }; | |
2473 | ||
8b88b099 | 2474 | static int vm_stat_get(void *_offset, u64 *val) |
ba1389b7 AK |
2475 | { |
2476 | unsigned offset = (long)_offset; | |
ba1389b7 AK |
2477 | struct kvm *kvm; |
2478 | ||
8b88b099 | 2479 | *val = 0; |
ba1389b7 AK |
2480 | spin_lock(&kvm_lock); |
2481 | list_for_each_entry(kvm, &vm_list, vm_list) | |
8b88b099 | 2482 | *val += *(u32 *)((void *)kvm + offset); |
ba1389b7 | 2483 | spin_unlock(&kvm_lock); |
8b88b099 | 2484 | return 0; |
ba1389b7 AK |
2485 | } |
2486 | ||
2487 | DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n"); | |
2488 | ||
8b88b099 | 2489 | static int vcpu_stat_get(void *_offset, u64 *val) |
1165f5fe AK |
2490 | { |
2491 | unsigned offset = (long)_offset; | |
1165f5fe AK |
2492 | struct kvm *kvm; |
2493 | struct kvm_vcpu *vcpu; | |
2494 | int i; | |
2495 | ||
8b88b099 | 2496 | *val = 0; |
1165f5fe AK |
2497 | spin_lock(&kvm_lock); |
2498 | list_for_each_entry(kvm, &vm_list, vm_list) | |
2499 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { | |
fb3f0f51 RR |
2500 | vcpu = kvm->vcpus[i]; |
2501 | if (vcpu) | |
8b88b099 | 2502 | *val += *(u32 *)((void *)vcpu + offset); |
1165f5fe AK |
2503 | } |
2504 | spin_unlock(&kvm_lock); | |
8b88b099 | 2505 | return 0; |
1165f5fe AK |
2506 | } |
2507 | ||
ba1389b7 AK |
2508 | DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n"); |
2509 | ||
2510 | static struct file_operations *stat_fops[] = { | |
2511 | [KVM_STAT_VCPU] = &vcpu_stat_fops, | |
2512 | [KVM_STAT_VM] = &vm_stat_fops, | |
2513 | }; | |
1165f5fe | 2514 | |
a16b043c | 2515 | static void kvm_init_debug(void) |
6aa8b732 AK |
2516 | { |
2517 | struct kvm_stats_debugfs_item *p; | |
2518 | ||
76f7c879 | 2519 | kvm_debugfs_dir = debugfs_create_dir("kvm", NULL); |
6aa8b732 | 2520 | for (p = debugfs_entries; p->name; ++p) |
76f7c879 | 2521 | p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir, |
1165f5fe | 2522 | (void *)(long)p->offset, |
ba1389b7 | 2523 | stat_fops[p->kind]); |
6aa8b732 AK |
2524 | } |
2525 | ||
2526 | static void kvm_exit_debug(void) | |
2527 | { | |
2528 | struct kvm_stats_debugfs_item *p; | |
2529 | ||
2530 | for (p = debugfs_entries; p->name; ++p) | |
2531 | debugfs_remove(p->dentry); | |
76f7c879 | 2532 | debugfs_remove(kvm_debugfs_dir); |
6aa8b732 AK |
2533 | } |
2534 | ||
59ae6c6b AK |
2535 | static int kvm_suspend(struct sys_device *dev, pm_message_t state) |
2536 | { | |
4267c41a | 2537 | hardware_disable(NULL); |
59ae6c6b AK |
2538 | return 0; |
2539 | } | |
2540 | ||
2541 | static int kvm_resume(struct sys_device *dev) | |
2542 | { | |
4267c41a | 2543 | hardware_enable(NULL); |
59ae6c6b AK |
2544 | return 0; |
2545 | } | |
2546 | ||
2547 | static struct sysdev_class kvm_sysdev_class = { | |
af5ca3f4 | 2548 | .name = "kvm", |
59ae6c6b AK |
2549 | .suspend = kvm_suspend, |
2550 | .resume = kvm_resume, | |
2551 | }; | |
2552 | ||
2553 | static struct sys_device kvm_sysdev = { | |
2554 | .id = 0, | |
2555 | .cls = &kvm_sysdev_class, | |
2556 | }; | |
2557 | ||
cea7bb21 | 2558 | struct page *bad_page; |
35149e21 | 2559 | pfn_t bad_pfn; |
6aa8b732 | 2560 | |
15ad7146 AK |
2561 | static inline |
2562 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
2563 | { | |
2564 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
2565 | } | |
2566 | ||
2567 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
2568 | { | |
2569 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2570 | ||
e9b11c17 | 2571 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 AK |
2572 | } |
2573 | ||
2574 | static void kvm_sched_out(struct preempt_notifier *pn, | |
2575 | struct task_struct *next) | |
2576 | { | |
2577 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2578 | ||
e9b11c17 | 2579 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
2580 | } |
2581 | ||
f8c16bba | 2582 | int kvm_init(void *opaque, unsigned int vcpu_size, |
c16f862d | 2583 | struct module *module) |
6aa8b732 AK |
2584 | { |
2585 | int r; | |
002c7f7c | 2586 | int cpu; |
6aa8b732 | 2587 | |
cb498ea2 ZX |
2588 | kvm_init_debug(); |
2589 | ||
f8c16bba ZX |
2590 | r = kvm_arch_init(opaque); |
2591 | if (r) | |
d2308784 | 2592 | goto out_fail; |
cb498ea2 ZX |
2593 | |
2594 | bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
2595 | ||
2596 | if (bad_page == NULL) { | |
2597 | r = -ENOMEM; | |
2598 | goto out; | |
2599 | } | |
2600 | ||
35149e21 AL |
2601 | bad_pfn = page_to_pfn(bad_page); |
2602 | ||
8437a617 | 2603 | if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) { |
7f59f492 RR |
2604 | r = -ENOMEM; |
2605 | goto out_free_0; | |
2606 | } | |
2607 | ||
e9b11c17 | 2608 | r = kvm_arch_hardware_setup(); |
6aa8b732 | 2609 | if (r < 0) |
7f59f492 | 2610 | goto out_free_0a; |
6aa8b732 | 2611 | |
002c7f7c YS |
2612 | for_each_online_cpu(cpu) { |
2613 | smp_call_function_single(cpu, | |
e9b11c17 | 2614 | kvm_arch_check_processor_compat, |
8691e5a8 | 2615 | &r, 1); |
002c7f7c | 2616 | if (r < 0) |
d2308784 | 2617 | goto out_free_1; |
002c7f7c YS |
2618 | } |
2619 | ||
15c8b6c1 | 2620 | on_each_cpu(hardware_enable, NULL, 1); |
774c47f1 AK |
2621 | r = register_cpu_notifier(&kvm_cpu_notifier); |
2622 | if (r) | |
d2308784 | 2623 | goto out_free_2; |
6aa8b732 AK |
2624 | register_reboot_notifier(&kvm_reboot_notifier); |
2625 | ||
59ae6c6b AK |
2626 | r = sysdev_class_register(&kvm_sysdev_class); |
2627 | if (r) | |
d2308784 | 2628 | goto out_free_3; |
59ae6c6b AK |
2629 | |
2630 | r = sysdev_register(&kvm_sysdev); | |
2631 | if (r) | |
d2308784 | 2632 | goto out_free_4; |
59ae6c6b | 2633 | |
c16f862d RR |
2634 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
2635 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, | |
56919c5c JP |
2636 | __alignof__(struct kvm_vcpu), |
2637 | 0, NULL); | |
c16f862d RR |
2638 | if (!kvm_vcpu_cache) { |
2639 | r = -ENOMEM; | |
d2308784 | 2640 | goto out_free_5; |
c16f862d RR |
2641 | } |
2642 | ||
6aa8b732 | 2643 | kvm_chardev_ops.owner = module; |
3d3aab1b CB |
2644 | kvm_vm_fops.owner = module; |
2645 | kvm_vcpu_fops.owner = module; | |
6aa8b732 AK |
2646 | |
2647 | r = misc_register(&kvm_dev); | |
2648 | if (r) { | |
d77c26fc | 2649 | printk(KERN_ERR "kvm: misc device register failed\n"); |
6aa8b732 AK |
2650 | goto out_free; |
2651 | } | |
2652 | ||
15ad7146 AK |
2653 | kvm_preempt_ops.sched_in = kvm_sched_in; |
2654 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
2655 | ||
c7addb90 | 2656 | return 0; |
6aa8b732 AK |
2657 | |
2658 | out_free: | |
c16f862d | 2659 | kmem_cache_destroy(kvm_vcpu_cache); |
d2308784 | 2660 | out_free_5: |
59ae6c6b | 2661 | sysdev_unregister(&kvm_sysdev); |
d2308784 | 2662 | out_free_4: |
59ae6c6b | 2663 | sysdev_class_unregister(&kvm_sysdev_class); |
d2308784 | 2664 | out_free_3: |
6aa8b732 | 2665 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 | 2666 | unregister_cpu_notifier(&kvm_cpu_notifier); |
d2308784 | 2667 | out_free_2: |
15c8b6c1 | 2668 | on_each_cpu(hardware_disable, NULL, 1); |
d2308784 | 2669 | out_free_1: |
e9b11c17 | 2670 | kvm_arch_hardware_unsetup(); |
7f59f492 RR |
2671 | out_free_0a: |
2672 | free_cpumask_var(cpus_hardware_enabled); | |
d2308784 ZX |
2673 | out_free_0: |
2674 | __free_page(bad_page); | |
ca45aaae | 2675 | out: |
f8c16bba | 2676 | kvm_arch_exit(); |
cb498ea2 | 2677 | kvm_exit_debug(); |
d2308784 | 2678 | out_fail: |
6aa8b732 AK |
2679 | return r; |
2680 | } | |
cb498ea2 | 2681 | EXPORT_SYMBOL_GPL(kvm_init); |
6aa8b732 | 2682 | |
cb498ea2 | 2683 | void kvm_exit(void) |
6aa8b732 | 2684 | { |
d4c9ff2d | 2685 | kvm_trace_cleanup(); |
6aa8b732 | 2686 | misc_deregister(&kvm_dev); |
c16f862d | 2687 | kmem_cache_destroy(kvm_vcpu_cache); |
59ae6c6b AK |
2688 | sysdev_unregister(&kvm_sysdev); |
2689 | sysdev_class_unregister(&kvm_sysdev_class); | |
6aa8b732 | 2690 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 2691 | unregister_cpu_notifier(&kvm_cpu_notifier); |
15c8b6c1 | 2692 | on_each_cpu(hardware_disable, NULL, 1); |
e9b11c17 | 2693 | kvm_arch_hardware_unsetup(); |
f8c16bba | 2694 | kvm_arch_exit(); |
6aa8b732 | 2695 | kvm_exit_debug(); |
7f59f492 | 2696 | free_cpumask_var(cpus_hardware_enabled); |
cea7bb21 | 2697 | __free_page(bad_page); |
6aa8b732 | 2698 | } |
cb498ea2 | 2699 | EXPORT_SYMBOL_GPL(kvm_exit); |