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> |
6aa8b732 | 45 | |
e495606d | 46 | #include <asm/processor.h> |
e495606d AK |
47 | #include <asm/io.h> |
48 | #include <asm/uaccess.h> | |
3e021bf5 | 49 | #include <asm/pgtable.h> |
6aa8b732 | 50 | |
5f94c174 LV |
51 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
52 | #include "coalesced_mmio.h" | |
53 | #endif | |
54 | ||
8a98f664 XZ |
55 | #ifdef KVM_CAP_DEVICE_ASSIGNMENT |
56 | #include <linux/pci.h> | |
57 | #include <linux/interrupt.h> | |
58 | #include "irq.h" | |
59 | #endif | |
60 | ||
6aa8b732 AK |
61 | MODULE_AUTHOR("Qumranet"); |
62 | MODULE_LICENSE("GPL"); | |
63 | ||
e9b11c17 ZX |
64 | DEFINE_SPINLOCK(kvm_lock); |
65 | LIST_HEAD(vm_list); | |
133de902 | 66 | |
7f59f492 | 67 | static cpumask_var_t cpus_hardware_enabled; |
1b6c0168 | 68 | |
c16f862d RR |
69 | struct kmem_cache *kvm_vcpu_cache; |
70 | EXPORT_SYMBOL_GPL(kvm_vcpu_cache); | |
1165f5fe | 71 | |
15ad7146 AK |
72 | static __read_mostly struct preempt_ops kvm_preempt_ops; |
73 | ||
76f7c879 | 74 | struct dentry *kvm_debugfs_dir; |
6aa8b732 | 75 | |
bccf2150 AK |
76 | static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl, |
77 | unsigned long arg); | |
78 | ||
e8ba5d31 | 79 | static bool kvm_rebooting; |
4ecac3fd | 80 | |
8a98f664 XZ |
81 | #ifdef KVM_CAP_DEVICE_ASSIGNMENT |
82 | static struct kvm_assigned_dev_kernel *kvm_find_assigned_dev(struct list_head *head, | |
83 | int assigned_dev_id) | |
84 | { | |
85 | struct list_head *ptr; | |
86 | struct kvm_assigned_dev_kernel *match; | |
87 | ||
88 | list_for_each(ptr, head) { | |
89 | match = list_entry(ptr, struct kvm_assigned_dev_kernel, list); | |
90 | if (match->assigned_dev_id == assigned_dev_id) | |
91 | return match; | |
92 | } | |
93 | return NULL; | |
94 | } | |
95 | ||
2350bd1f SY |
96 | static int find_index_from_host_irq(struct kvm_assigned_dev_kernel |
97 | *assigned_dev, int irq) | |
98 | { | |
99 | int i, index; | |
100 | struct msix_entry *host_msix_entries; | |
101 | ||
102 | host_msix_entries = assigned_dev->host_msix_entries; | |
103 | ||
104 | index = -1; | |
105 | for (i = 0; i < assigned_dev->entries_nr; i++) | |
106 | if (irq == host_msix_entries[i].vector) { | |
107 | index = i; | |
108 | break; | |
109 | } | |
110 | if (index < 0) { | |
111 | printk(KERN_WARNING "Fail to find correlated MSI-X entry!\n"); | |
112 | return 0; | |
113 | } | |
114 | ||
115 | return index; | |
116 | } | |
117 | ||
8a98f664 XZ |
118 | static void kvm_assigned_dev_interrupt_work_handler(struct work_struct *work) |
119 | { | |
120 | struct kvm_assigned_dev_kernel *assigned_dev; | |
2350bd1f SY |
121 | struct kvm *kvm; |
122 | int irq, i; | |
8a98f664 XZ |
123 | |
124 | assigned_dev = container_of(work, struct kvm_assigned_dev_kernel, | |
125 | interrupt_work); | |
2350bd1f | 126 | kvm = assigned_dev->kvm; |
8a98f664 XZ |
127 | |
128 | /* This is taken to safely inject irq inside the guest. When | |
129 | * the interrupt injection (or the ioapic code) uses a | |
130 | * finer-grained lock, update this | |
131 | */ | |
2350bd1f | 132 | mutex_lock(&kvm->lock); |
e56d532f | 133 | if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_HOST_MSIX) { |
2350bd1f SY |
134 | struct kvm_guest_msix_entry *guest_entries = |
135 | assigned_dev->guest_msix_entries; | |
136 | for (i = 0; i < assigned_dev->entries_nr; i++) { | |
137 | if (!(guest_entries[i].flags & | |
138 | KVM_ASSIGNED_MSIX_PENDING)) | |
139 | continue; | |
140 | guest_entries[i].flags &= ~KVM_ASSIGNED_MSIX_PENDING; | |
141 | kvm_set_irq(assigned_dev->kvm, | |
142 | assigned_dev->irq_source_id, | |
143 | guest_entries[i].vector, 1); | |
144 | irq = assigned_dev->host_msix_entries[i].vector; | |
145 | if (irq != 0) | |
146 | enable_irq(irq); | |
147 | assigned_dev->host_irq_disabled = false; | |
148 | } | |
149 | } else { | |
150 | kvm_set_irq(assigned_dev->kvm, assigned_dev->irq_source_id, | |
151 | assigned_dev->guest_irq, 1); | |
152 | if (assigned_dev->irq_requested_type & | |
e56d532f | 153 | KVM_DEV_IRQ_GUEST_MSI) { |
2350bd1f SY |
154 | enable_irq(assigned_dev->host_irq); |
155 | assigned_dev->host_irq_disabled = false; | |
156 | } | |
6b9cc7fd | 157 | } |
2350bd1f | 158 | |
8a98f664 | 159 | mutex_unlock(&assigned_dev->kvm->lock); |
8a98f664 XZ |
160 | } |
161 | ||
8a98f664 XZ |
162 | static irqreturn_t kvm_assigned_dev_intr(int irq, void *dev_id) |
163 | { | |
164 | struct kvm_assigned_dev_kernel *assigned_dev = | |
165 | (struct kvm_assigned_dev_kernel *) dev_id; | |
166 | ||
e56d532f | 167 | if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_HOST_MSIX) { |
2350bd1f SY |
168 | int index = find_index_from_host_irq(assigned_dev, irq); |
169 | if (index < 0) | |
170 | return IRQ_HANDLED; | |
171 | assigned_dev->guest_msix_entries[index].flags |= | |
172 | KVM_ASSIGNED_MSIX_PENDING; | |
173 | } | |
174 | ||
8a98f664 | 175 | schedule_work(&assigned_dev->interrupt_work); |
defaf158 | 176 | |
8a98f664 | 177 | disable_irq_nosync(irq); |
defaf158 MM |
178 | assigned_dev->host_irq_disabled = true; |
179 | ||
8a98f664 XZ |
180 | return IRQ_HANDLED; |
181 | } | |
182 | ||
183 | /* Ack the irq line for an assigned device */ | |
184 | static void kvm_assigned_dev_ack_irq(struct kvm_irq_ack_notifier *kian) | |
185 | { | |
186 | struct kvm_assigned_dev_kernel *dev; | |
187 | ||
188 | if (kian->gsi == -1) | |
189 | return; | |
190 | ||
191 | dev = container_of(kian, struct kvm_assigned_dev_kernel, | |
192 | ack_notifier); | |
defaf158 | 193 | |
5550af4d | 194 | kvm_set_irq(dev->kvm, dev->irq_source_id, dev->guest_irq, 0); |
defaf158 MM |
195 | |
196 | /* The guest irq may be shared so this ack may be | |
197 | * from another device. | |
198 | */ | |
199 | if (dev->host_irq_disabled) { | |
200 | enable_irq(dev->host_irq); | |
201 | dev->host_irq_disabled = false; | |
202 | } | |
8a98f664 XZ |
203 | } |
204 | ||
e56d532f SY |
205 | static void deassign_guest_irq(struct kvm *kvm, |
206 | struct kvm_assigned_dev_kernel *assigned_dev) | |
8a98f664 | 207 | { |
e19e30ef | 208 | kvm_unregister_irq_ack_notifier(&assigned_dev->ack_notifier); |
e56d532f | 209 | assigned_dev->ack_notifier.gsi = -1; |
f29b2673 MM |
210 | |
211 | if (assigned_dev->irq_source_id != -1) | |
212 | kvm_free_irq_source_id(kvm, assigned_dev->irq_source_id); | |
213 | assigned_dev->irq_source_id = -1; | |
e56d532f SY |
214 | assigned_dev->irq_requested_type &= ~(KVM_DEV_IRQ_GUEST_MASK); |
215 | } | |
8a98f664 | 216 | |
e56d532f SY |
217 | /* The function implicit hold kvm->lock mutex due to cancel_work_sync() */ |
218 | static void deassign_host_irq(struct kvm *kvm, | |
219 | struct kvm_assigned_dev_kernel *assigned_dev) | |
220 | { | |
ba4cef31 SY |
221 | /* |
222 | * In kvm_free_device_irq, cancel_work_sync return true if: | |
223 | * 1. work is scheduled, and then cancelled. | |
224 | * 2. work callback is executed. | |
225 | * | |
226 | * The first one ensured that the irq is disabled and no more events | |
227 | * would happen. But for the second one, the irq may be enabled (e.g. | |
228 | * for MSI). So we disable irq here to prevent further events. | |
229 | * | |
230 | * Notice this maybe result in nested disable if the interrupt type is | |
231 | * INTx, but it's OK for we are going to free it. | |
232 | * | |
233 | * If this function is a part of VM destroy, please ensure that till | |
234 | * now, the kvm state is still legal for probably we also have to wait | |
235 | * interrupt_work done. | |
236 | */ | |
e56d532f | 237 | if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_HOST_MSIX) { |
d510d6cc SY |
238 | int i; |
239 | for (i = 0; i < assigned_dev->entries_nr; i++) | |
240 | disable_irq_nosync(assigned_dev-> | |
241 | host_msix_entries[i].vector); | |
242 | ||
243 | cancel_work_sync(&assigned_dev->interrupt_work); | |
244 | ||
245 | for (i = 0; i < assigned_dev->entries_nr; i++) | |
246 | free_irq(assigned_dev->host_msix_entries[i].vector, | |
247 | (void *)assigned_dev); | |
248 | ||
249 | assigned_dev->entries_nr = 0; | |
250 | kfree(assigned_dev->host_msix_entries); | |
251 | kfree(assigned_dev->guest_msix_entries); | |
252 | pci_disable_msix(assigned_dev->dev); | |
253 | } else { | |
254 | /* Deal with MSI and INTx */ | |
255 | disable_irq_nosync(assigned_dev->host_irq); | |
256 | cancel_work_sync(&assigned_dev->interrupt_work); | |
8a98f664 | 257 | |
d510d6cc | 258 | free_irq(assigned_dev->host_irq, (void *)assigned_dev); |
4a643be8 | 259 | |
e56d532f | 260 | if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_HOST_MSI) |
d510d6cc SY |
261 | pci_disable_msi(assigned_dev->dev); |
262 | } | |
4a643be8 | 263 | |
e56d532f SY |
264 | assigned_dev->irq_requested_type &= ~(KVM_DEV_IRQ_HOST_MASK); |
265 | } | |
266 | ||
267 | static int kvm_deassign_irq(struct kvm *kvm, | |
268 | struct kvm_assigned_dev_kernel *assigned_dev, | |
269 | unsigned long irq_requested_type) | |
270 | { | |
271 | unsigned long guest_irq_type, host_irq_type; | |
272 | ||
273 | if (!irqchip_in_kernel(kvm)) | |
274 | return -EINVAL; | |
275 | /* no irq assignment to deassign */ | |
276 | if (!assigned_dev->irq_requested_type) | |
277 | return -ENXIO; | |
278 | ||
279 | host_irq_type = irq_requested_type & KVM_DEV_IRQ_HOST_MASK; | |
280 | guest_irq_type = irq_requested_type & KVM_DEV_IRQ_GUEST_MASK; | |
281 | ||
282 | if (host_irq_type) | |
283 | deassign_host_irq(kvm, assigned_dev); | |
284 | if (guest_irq_type) | |
285 | deassign_guest_irq(kvm, assigned_dev); | |
286 | ||
287 | return 0; | |
4a643be8 MM |
288 | } |
289 | ||
e56d532f SY |
290 | static void kvm_free_assigned_irq(struct kvm *kvm, |
291 | struct kvm_assigned_dev_kernel *assigned_dev) | |
292 | { | |
293 | kvm_deassign_irq(kvm, assigned_dev, assigned_dev->irq_requested_type); | |
294 | } | |
4a643be8 MM |
295 | |
296 | static void kvm_free_assigned_device(struct kvm *kvm, | |
297 | struct kvm_assigned_dev_kernel | |
298 | *assigned_dev) | |
299 | { | |
300 | kvm_free_assigned_irq(kvm, assigned_dev); | |
301 | ||
6eb55818 SY |
302 | pci_reset_function(assigned_dev->dev); |
303 | ||
8a98f664 XZ |
304 | pci_release_regions(assigned_dev->dev); |
305 | pci_disable_device(assigned_dev->dev); | |
306 | pci_dev_put(assigned_dev->dev); | |
307 | ||
308 | list_del(&assigned_dev->list); | |
309 | kfree(assigned_dev); | |
310 | } | |
311 | ||
312 | void kvm_free_all_assigned_devices(struct kvm *kvm) | |
313 | { | |
314 | struct list_head *ptr, *ptr2; | |
315 | struct kvm_assigned_dev_kernel *assigned_dev; | |
316 | ||
317 | list_for_each_safe(ptr, ptr2, &kvm->arch.assigned_dev_head) { | |
318 | assigned_dev = list_entry(ptr, | |
319 | struct kvm_assigned_dev_kernel, | |
320 | list); | |
321 | ||
322 | kvm_free_assigned_device(kvm, assigned_dev); | |
323 | } | |
324 | } | |
325 | ||
e56d532f SY |
326 | static int assigned_device_enable_host_intx(struct kvm *kvm, |
327 | struct kvm_assigned_dev_kernel *dev) | |
00e3ed39 | 328 | { |
e56d532f SY |
329 | dev->host_irq = dev->dev->irq; |
330 | /* Even though this is PCI, we don't want to use shared | |
331 | * interrupts. Sharing host devices with guest-assigned devices | |
332 | * on the same interrupt line is not a happy situation: there | |
333 | * are going to be long delays in accepting, acking, etc. | |
334 | */ | |
335 | if (request_irq(dev->host_irq, kvm_assigned_dev_intr, | |
336 | 0, "kvm_assigned_intx_device", (void *)dev)) | |
337 | return -EIO; | |
338 | return 0; | |
339 | } | |
6b9cc7fd | 340 | |
e56d532f SY |
341 | #ifdef __KVM_HAVE_MSI |
342 | static int assigned_device_enable_host_msi(struct kvm *kvm, | |
343 | struct kvm_assigned_dev_kernel *dev) | |
344 | { | |
345 | int r; | |
00e3ed39 | 346 | |
e56d532f SY |
347 | if (!dev->dev->msi_enabled) { |
348 | r = pci_enable_msi(dev->dev); | |
349 | if (r) | |
350 | return r; | |
351 | } | |
00e3ed39 | 352 | |
e56d532f SY |
353 | dev->host_irq = dev->dev->irq; |
354 | if (request_irq(dev->host_irq, kvm_assigned_dev_intr, 0, | |
355 | "kvm_assigned_msi_device", (void *)dev)) { | |
356 | pci_disable_msi(dev->dev); | |
357 | return -EIO; | |
00e3ed39 SY |
358 | } |
359 | ||
00e3ed39 SY |
360 | return 0; |
361 | } | |
e56d532f | 362 | #endif |
00e3ed39 | 363 | |
e56d532f SY |
364 | #ifdef __KVM_HAVE_MSIX |
365 | static int assigned_device_enable_host_msix(struct kvm *kvm, | |
366 | struct kvm_assigned_dev_kernel *dev) | |
6b9cc7fd | 367 | { |
e56d532f | 368 | int i, r = -EINVAL; |
6b9cc7fd | 369 | |
e56d532f SY |
370 | /* host_msix_entries and guest_msix_entries should have been |
371 | * initialized */ | |
372 | if (dev->entries_nr == 0) | |
373 | return r; | |
6b9cc7fd | 374 | |
e56d532f SY |
375 | r = pci_enable_msix(dev->dev, dev->host_msix_entries, dev->entries_nr); |
376 | if (r) | |
377 | return r; | |
6b9cc7fd | 378 | |
e56d532f SY |
379 | for (i = 0; i < dev->entries_nr; i++) { |
380 | r = request_irq(dev->host_msix_entries[i].vector, | |
381 | kvm_assigned_dev_intr, 0, | |
382 | "kvm_assigned_msix_device", | |
383 | (void *)dev); | |
384 | /* FIXME: free requested_irq's on failure */ | |
385 | if (r) | |
386 | return r; | |
387 | } | |
5319c662 | 388 | |
e56d532f SY |
389 | return 0; |
390 | } | |
6b9cc7fd | 391 | |
e56d532f | 392 | #endif |
6b9cc7fd | 393 | |
e56d532f SY |
394 | static int assigned_device_enable_guest_intx(struct kvm *kvm, |
395 | struct kvm_assigned_dev_kernel *dev, | |
396 | struct kvm_assigned_irq *irq) | |
397 | { | |
398 | dev->guest_irq = irq->guest_irq; | |
399 | dev->ack_notifier.gsi = irq->guest_irq; | |
400 | return 0; | |
401 | } | |
5319c662 | 402 | |
e56d532f SY |
403 | #ifdef __KVM_HAVE_MSI |
404 | static int assigned_device_enable_guest_msi(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 = -1; | |
6b9cc7fd SY |
410 | return 0; |
411 | } | |
412 | #endif | |
e56d532f SY |
413 | #ifdef __KVM_HAVE_MSIX |
414 | static int assigned_device_enable_guest_msix(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; | |
420 | return 0; | |
421 | } | |
422 | #endif | |
423 | ||
424 | static int assign_host_irq(struct kvm *kvm, | |
425 | struct kvm_assigned_dev_kernel *dev, | |
426 | __u32 host_irq_type) | |
427 | { | |
428 | int r = -EEXIST; | |
429 | ||
430 | if (dev->irq_requested_type & KVM_DEV_IRQ_HOST_MASK) | |
431 | return r; | |
6b9cc7fd | 432 | |
e56d532f SY |
433 | switch (host_irq_type) { |
434 | case KVM_DEV_IRQ_HOST_INTX: | |
435 | r = assigned_device_enable_host_intx(kvm, dev); | |
436 | break; | |
437 | #ifdef __KVM_HAVE_MSI | |
438 | case KVM_DEV_IRQ_HOST_MSI: | |
439 | r = assigned_device_enable_host_msi(kvm, dev); | |
440 | break; | |
441 | #endif | |
d510d6cc | 442 | #ifdef __KVM_HAVE_MSIX |
e56d532f SY |
443 | case KVM_DEV_IRQ_HOST_MSIX: |
444 | r = assigned_device_enable_host_msix(kvm, dev); | |
445 | break; | |
446 | #endif | |
447 | default: | |
448 | r = -EINVAL; | |
449 | } | |
d510d6cc | 450 | |
e56d532f SY |
451 | if (!r) |
452 | dev->irq_requested_type |= host_irq_type; | |
d510d6cc | 453 | |
e56d532f SY |
454 | return r; |
455 | } | |
d510d6cc | 456 | |
e56d532f SY |
457 | static int assign_guest_irq(struct kvm *kvm, |
458 | struct kvm_assigned_dev_kernel *dev, | |
459 | struct kvm_assigned_irq *irq, | |
460 | unsigned long guest_irq_type) | |
461 | { | |
462 | int id; | |
463 | int r = -EEXIST; | |
d510d6cc | 464 | |
e56d532f SY |
465 | if (dev->irq_requested_type & KVM_DEV_IRQ_GUEST_MASK) |
466 | return r; | |
467 | ||
468 | id = kvm_request_irq_source_id(kvm); | |
469 | if (id < 0) | |
470 | return id; | |
471 | ||
472 | dev->irq_source_id = id; | |
473 | ||
474 | switch (guest_irq_type) { | |
475 | case KVM_DEV_IRQ_GUEST_INTX: | |
476 | r = assigned_device_enable_guest_intx(kvm, dev, irq); | |
477 | break; | |
478 | #ifdef __KVM_HAVE_MSI | |
479 | case KVM_DEV_IRQ_GUEST_MSI: | |
480 | r = assigned_device_enable_guest_msi(kvm, dev, irq); | |
481 | break; | |
482 | #endif | |
483 | #ifdef __KVM_HAVE_MSIX | |
484 | case KVM_DEV_IRQ_GUEST_MSIX: | |
485 | r = assigned_device_enable_guest_msix(kvm, dev, irq); | |
486 | break; | |
487 | #endif | |
488 | default: | |
489 | r = -EINVAL; | |
d510d6cc SY |
490 | } |
491 | ||
e56d532f SY |
492 | if (!r) { |
493 | dev->irq_requested_type |= guest_irq_type; | |
494 | kvm_register_irq_ack_notifier(kvm, &dev->ack_notifier); | |
495 | } else | |
496 | kvm_free_irq_source_id(kvm, dev->irq_source_id); | |
d510d6cc | 497 | |
e56d532f | 498 | return r; |
d510d6cc | 499 | } |
d510d6cc | 500 | |
e56d532f | 501 | /* TODO Deal with KVM_DEV_IRQ_ASSIGNED_MASK_MSIX */ |
8a98f664 | 502 | static int kvm_vm_ioctl_assign_irq(struct kvm *kvm, |
e56d532f | 503 | struct kvm_assigned_irq *assigned_irq) |
8a98f664 | 504 | { |
e56d532f | 505 | int r = -EINVAL; |
8a98f664 | 506 | struct kvm_assigned_dev_kernel *match; |
e56d532f | 507 | unsigned long host_irq_type, guest_irq_type; |
8a98f664 | 508 | |
e56d532f SY |
509 | if (!capable(CAP_SYS_RAWIO)) |
510 | return -EPERM; | |
8a98f664 | 511 | |
e56d532f SY |
512 | if (!irqchip_in_kernel(kvm)) |
513 | return r; | |
514 | ||
515 | mutex_lock(&kvm->lock); | |
516 | r = -ENODEV; | |
8a98f664 XZ |
517 | match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, |
518 | assigned_irq->assigned_dev_id); | |
e56d532f SY |
519 | if (!match) |
520 | goto out; | |
8a98f664 | 521 | |
e56d532f SY |
522 | host_irq_type = (assigned_irq->flags & KVM_DEV_IRQ_HOST_MASK); |
523 | guest_irq_type = (assigned_irq->flags & KVM_DEV_IRQ_GUEST_MASK); | |
17071fe7 | 524 | |
e56d532f SY |
525 | r = -EINVAL; |
526 | /* can only assign one type at a time */ | |
527 | if (hweight_long(host_irq_type) > 1) | |
528 | goto out; | |
529 | if (hweight_long(guest_irq_type) > 1) | |
530 | goto out; | |
531 | if (host_irq_type == 0 && guest_irq_type == 0) | |
532 | goto out; | |
17071fe7 | 533 | |
e56d532f SY |
534 | r = 0; |
535 | if (host_irq_type) | |
536 | r = assign_host_irq(kvm, match, host_irq_type); | |
537 | if (r) | |
538 | goto out; | |
8a98f664 | 539 | |
e56d532f SY |
540 | if (guest_irq_type) |
541 | r = assign_guest_irq(kvm, match, assigned_irq, guest_irq_type); | |
542 | out: | |
8a98f664 XZ |
543 | mutex_unlock(&kvm->lock); |
544 | return r; | |
e56d532f SY |
545 | } |
546 | ||
547 | static int kvm_vm_ioctl_deassign_dev_irq(struct kvm *kvm, | |
548 | struct kvm_assigned_irq | |
549 | *assigned_irq) | |
550 | { | |
551 | int r = -ENODEV; | |
552 | struct kvm_assigned_dev_kernel *match; | |
553 | ||
554 | mutex_lock(&kvm->lock); | |
555 | ||
556 | match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, | |
557 | assigned_irq->assigned_dev_id); | |
558 | if (!match) | |
559 | goto out; | |
560 | ||
561 | r = kvm_deassign_irq(kvm, match, assigned_irq->flags); | |
562 | out: | |
8a98f664 | 563 | mutex_unlock(&kvm->lock); |
8a98f664 XZ |
564 | return r; |
565 | } | |
566 | ||
567 | static int kvm_vm_ioctl_assign_device(struct kvm *kvm, | |
568 | struct kvm_assigned_pci_dev *assigned_dev) | |
569 | { | |
570 | int r = 0; | |
571 | struct kvm_assigned_dev_kernel *match; | |
572 | struct pci_dev *dev; | |
573 | ||
682edb4c | 574 | down_read(&kvm->slots_lock); |
8a98f664 XZ |
575 | mutex_lock(&kvm->lock); |
576 | ||
577 | match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, | |
578 | assigned_dev->assigned_dev_id); | |
579 | if (match) { | |
580 | /* device already assigned */ | |
e56d532f | 581 | r = -EEXIST; |
8a98f664 XZ |
582 | goto out; |
583 | } | |
584 | ||
585 | match = kzalloc(sizeof(struct kvm_assigned_dev_kernel), GFP_KERNEL); | |
586 | if (match == NULL) { | |
587 | printk(KERN_INFO "%s: Couldn't allocate memory\n", | |
588 | __func__); | |
589 | r = -ENOMEM; | |
590 | goto out; | |
591 | } | |
592 | dev = pci_get_bus_and_slot(assigned_dev->busnr, | |
593 | assigned_dev->devfn); | |
594 | if (!dev) { | |
595 | printk(KERN_INFO "%s: host device not found\n", __func__); | |
596 | r = -EINVAL; | |
597 | goto out_free; | |
598 | } | |
599 | if (pci_enable_device(dev)) { | |
600 | printk(KERN_INFO "%s: Could not enable PCI device\n", __func__); | |
601 | r = -EBUSY; | |
602 | goto out_put; | |
603 | } | |
604 | r = pci_request_regions(dev, "kvm_assigned_device"); | |
605 | if (r) { | |
606 | printk(KERN_INFO "%s: Could not get access to device regions\n", | |
607 | __func__); | |
608 | goto out_disable; | |
609 | } | |
6eb55818 SY |
610 | |
611 | pci_reset_function(dev); | |
612 | ||
8a98f664 XZ |
613 | match->assigned_dev_id = assigned_dev->assigned_dev_id; |
614 | match->host_busnr = assigned_dev->busnr; | |
615 | match->host_devfn = assigned_dev->devfn; | |
b653574a | 616 | match->flags = assigned_dev->flags; |
8a98f664 | 617 | match->dev = dev; |
f29b2673 | 618 | match->irq_source_id = -1; |
8a98f664 | 619 | match->kvm = kvm; |
e56d532f SY |
620 | match->ack_notifier.irq_acked = kvm_assigned_dev_ack_irq; |
621 | INIT_WORK(&match->interrupt_work, | |
622 | kvm_assigned_dev_interrupt_work_handler); | |
8a98f664 XZ |
623 | |
624 | list_add(&match->list, &kvm->arch.assigned_dev_head); | |
625 | ||
626 | if (assigned_dev->flags & KVM_DEV_ASSIGN_ENABLE_IOMMU) { | |
19de40a8 | 627 | if (!kvm->arch.iommu_domain) { |
260782bc WH |
628 | r = kvm_iommu_map_guest(kvm); |
629 | if (r) | |
630 | goto out_list_del; | |
631 | } | |
632 | r = kvm_assign_device(kvm, match); | |
8a98f664 XZ |
633 | if (r) |
634 | goto out_list_del; | |
635 | } | |
636 | ||
637 | out: | |
638 | mutex_unlock(&kvm->lock); | |
682edb4c | 639 | up_read(&kvm->slots_lock); |
8a98f664 XZ |
640 | return r; |
641 | out_list_del: | |
642 | list_del(&match->list); | |
643 | pci_release_regions(dev); | |
644 | out_disable: | |
645 | pci_disable_device(dev); | |
646 | out_put: | |
647 | pci_dev_put(dev); | |
648 | out_free: | |
649 | kfree(match); | |
650 | mutex_unlock(&kvm->lock); | |
682edb4c | 651 | up_read(&kvm->slots_lock); |
8a98f664 XZ |
652 | return r; |
653 | } | |
654 | #endif | |
655 | ||
0a920356 WH |
656 | #ifdef KVM_CAP_DEVICE_DEASSIGNMENT |
657 | static int kvm_vm_ioctl_deassign_device(struct kvm *kvm, | |
658 | struct kvm_assigned_pci_dev *assigned_dev) | |
659 | { | |
660 | int r = 0; | |
661 | struct kvm_assigned_dev_kernel *match; | |
662 | ||
663 | mutex_lock(&kvm->lock); | |
664 | ||
665 | match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, | |
666 | assigned_dev->assigned_dev_id); | |
667 | if (!match) { | |
668 | printk(KERN_INFO "%s: device hasn't been assigned before, " | |
669 | "so cannot be deassigned\n", __func__); | |
670 | r = -EINVAL; | |
671 | goto out; | |
672 | } | |
673 | ||
4a906e49 | 674 | if (match->flags & KVM_DEV_ASSIGN_ENABLE_IOMMU) |
0a920356 WH |
675 | kvm_deassign_device(kvm, match); |
676 | ||
677 | kvm_free_assigned_device(kvm, match); | |
678 | ||
679 | out: | |
680 | mutex_unlock(&kvm->lock); | |
681 | return r; | |
682 | } | |
683 | #endif | |
684 | ||
5aacf0ca JM |
685 | static inline int valid_vcpu(int n) |
686 | { | |
687 | return likely(n >= 0 && n < KVM_MAX_VCPUS); | |
688 | } | |
689 | ||
c77fb9dc | 690 | inline int kvm_is_mmio_pfn(pfn_t pfn) |
cbff90a7 | 691 | { |
fc5659c8 JR |
692 | if (pfn_valid(pfn)) { |
693 | struct page *page = compound_head(pfn_to_page(pfn)); | |
694 | return PageReserved(page); | |
695 | } | |
cbff90a7 BAY |
696 | |
697 | return true; | |
698 | } | |
699 | ||
bccf2150 AK |
700 | /* |
701 | * Switches to specified vcpu, until a matching vcpu_put() | |
702 | */ | |
313a3dc7 | 703 | void vcpu_load(struct kvm_vcpu *vcpu) |
6aa8b732 | 704 | { |
15ad7146 AK |
705 | int cpu; |
706 | ||
bccf2150 | 707 | mutex_lock(&vcpu->mutex); |
15ad7146 AK |
708 | cpu = get_cpu(); |
709 | preempt_notifier_register(&vcpu->preempt_notifier); | |
313a3dc7 | 710 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 | 711 | put_cpu(); |
6aa8b732 AK |
712 | } |
713 | ||
313a3dc7 | 714 | void vcpu_put(struct kvm_vcpu *vcpu) |
6aa8b732 | 715 | { |
15ad7146 | 716 | preempt_disable(); |
313a3dc7 | 717 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
718 | preempt_notifier_unregister(&vcpu->preempt_notifier); |
719 | preempt_enable(); | |
6aa8b732 AK |
720 | mutex_unlock(&vcpu->mutex); |
721 | } | |
722 | ||
d9e368d6 AK |
723 | static void ack_flush(void *_completed) |
724 | { | |
d9e368d6 AK |
725 | } |
726 | ||
49846896 | 727 | static bool make_all_cpus_request(struct kvm *kvm, unsigned int req) |
d9e368d6 | 728 | { |
597a5f55 | 729 | int i, cpu, me; |
6ef7a1bc RR |
730 | cpumask_var_t cpus; |
731 | bool called = true; | |
d9e368d6 | 732 | struct kvm_vcpu *vcpu; |
d9e368d6 | 733 | |
6ef7a1bc RR |
734 | if (alloc_cpumask_var(&cpus, GFP_ATOMIC)) |
735 | cpumask_clear(cpus); | |
736 | ||
597a5f55 | 737 | me = get_cpu(); |
fb3f0f51 RR |
738 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { |
739 | vcpu = kvm->vcpus[i]; | |
740 | if (!vcpu) | |
741 | continue; | |
49846896 | 742 | if (test_and_set_bit(req, &vcpu->requests)) |
d9e368d6 AK |
743 | continue; |
744 | cpu = vcpu->cpu; | |
6ef7a1bc RR |
745 | if (cpus != NULL && cpu != -1 && cpu != me) |
746 | cpumask_set_cpu(cpu, cpus); | |
49846896 | 747 | } |
6ef7a1bc RR |
748 | if (unlikely(cpus == NULL)) |
749 | smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1); | |
750 | else if (!cpumask_empty(cpus)) | |
751 | smp_call_function_many(cpus, ack_flush, NULL, 1); | |
752 | else | |
753 | called = false; | |
597a5f55 | 754 | put_cpu(); |
6ef7a1bc | 755 | free_cpumask_var(cpus); |
49846896 | 756 | return called; |
d9e368d6 AK |
757 | } |
758 | ||
49846896 | 759 | void kvm_flush_remote_tlbs(struct kvm *kvm) |
2e53d63a | 760 | { |
49846896 RR |
761 | if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH)) |
762 | ++kvm->stat.remote_tlb_flush; | |
2e53d63a MT |
763 | } |
764 | ||
49846896 RR |
765 | void kvm_reload_remote_mmus(struct kvm *kvm) |
766 | { | |
767 | make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD); | |
768 | } | |
2e53d63a | 769 | |
fb3f0f51 RR |
770 | int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id) |
771 | { | |
772 | struct page *page; | |
773 | int r; | |
774 | ||
775 | mutex_init(&vcpu->mutex); | |
776 | vcpu->cpu = -1; | |
fb3f0f51 RR |
777 | vcpu->kvm = kvm; |
778 | vcpu->vcpu_id = id; | |
b6958ce4 | 779 | init_waitqueue_head(&vcpu->wq); |
fb3f0f51 RR |
780 | |
781 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
782 | if (!page) { | |
783 | r = -ENOMEM; | |
784 | goto fail; | |
785 | } | |
786 | vcpu->run = page_address(page); | |
787 | ||
e9b11c17 | 788 | r = kvm_arch_vcpu_init(vcpu); |
fb3f0f51 | 789 | if (r < 0) |
e9b11c17 | 790 | goto fail_free_run; |
fb3f0f51 RR |
791 | return 0; |
792 | ||
fb3f0f51 RR |
793 | fail_free_run: |
794 | free_page((unsigned long)vcpu->run); | |
795 | fail: | |
76fafa5e | 796 | return r; |
fb3f0f51 RR |
797 | } |
798 | EXPORT_SYMBOL_GPL(kvm_vcpu_init); | |
799 | ||
800 | void kvm_vcpu_uninit(struct kvm_vcpu *vcpu) | |
801 | { | |
e9b11c17 | 802 | kvm_arch_vcpu_uninit(vcpu); |
fb3f0f51 RR |
803 | free_page((unsigned long)vcpu->run); |
804 | } | |
805 | EXPORT_SYMBOL_GPL(kvm_vcpu_uninit); | |
806 | ||
e930bffe AA |
807 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
808 | static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn) | |
809 | { | |
810 | return container_of(mn, struct kvm, mmu_notifier); | |
811 | } | |
812 | ||
813 | static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn, | |
814 | struct mm_struct *mm, | |
815 | unsigned long address) | |
816 | { | |
817 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
818 | int need_tlb_flush; | |
819 | ||
820 | /* | |
821 | * When ->invalidate_page runs, the linux pte has been zapped | |
822 | * already but the page is still allocated until | |
823 | * ->invalidate_page returns. So if we increase the sequence | |
824 | * here the kvm page fault will notice if the spte can't be | |
825 | * established because the page is going to be freed. If | |
826 | * instead the kvm page fault establishes the spte before | |
827 | * ->invalidate_page runs, kvm_unmap_hva will release it | |
828 | * before returning. | |
829 | * | |
830 | * The sequence increase only need to be seen at spin_unlock | |
831 | * time, and not at spin_lock time. | |
832 | * | |
833 | * Increasing the sequence after the spin_unlock would be | |
834 | * unsafe because the kvm page fault could then establish the | |
835 | * pte after kvm_unmap_hva returned, without noticing the page | |
836 | * is going to be freed. | |
837 | */ | |
838 | spin_lock(&kvm->mmu_lock); | |
839 | kvm->mmu_notifier_seq++; | |
840 | need_tlb_flush = kvm_unmap_hva(kvm, address); | |
841 | spin_unlock(&kvm->mmu_lock); | |
842 | ||
843 | /* we've to flush the tlb before the pages can be freed */ | |
844 | if (need_tlb_flush) | |
845 | kvm_flush_remote_tlbs(kvm); | |
846 | ||
847 | } | |
848 | ||
849 | static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn, | |
850 | struct mm_struct *mm, | |
851 | unsigned long start, | |
852 | unsigned long end) | |
853 | { | |
854 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
855 | int need_tlb_flush = 0; | |
856 | ||
857 | spin_lock(&kvm->mmu_lock); | |
858 | /* | |
859 | * The count increase must become visible at unlock time as no | |
860 | * spte can be established without taking the mmu_lock and | |
861 | * count is also read inside the mmu_lock critical section. | |
862 | */ | |
863 | kvm->mmu_notifier_count++; | |
864 | for (; start < end; start += PAGE_SIZE) | |
865 | need_tlb_flush |= kvm_unmap_hva(kvm, start); | |
866 | spin_unlock(&kvm->mmu_lock); | |
867 | ||
868 | /* we've to flush the tlb before the pages can be freed */ | |
869 | if (need_tlb_flush) | |
870 | kvm_flush_remote_tlbs(kvm); | |
871 | } | |
872 | ||
873 | static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn, | |
874 | struct mm_struct *mm, | |
875 | unsigned long start, | |
876 | unsigned long end) | |
877 | { | |
878 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
879 | ||
880 | spin_lock(&kvm->mmu_lock); | |
881 | /* | |
882 | * This sequence increase will notify the kvm page fault that | |
883 | * the page that is going to be mapped in the spte could have | |
884 | * been freed. | |
885 | */ | |
886 | kvm->mmu_notifier_seq++; | |
887 | /* | |
888 | * The above sequence increase must be visible before the | |
889 | * below count decrease but both values are read by the kvm | |
890 | * page fault under mmu_lock spinlock so we don't need to add | |
891 | * a smb_wmb() here in between the two. | |
892 | */ | |
893 | kvm->mmu_notifier_count--; | |
894 | spin_unlock(&kvm->mmu_lock); | |
895 | ||
896 | BUG_ON(kvm->mmu_notifier_count < 0); | |
897 | } | |
898 | ||
899 | static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn, | |
900 | struct mm_struct *mm, | |
901 | unsigned long address) | |
902 | { | |
903 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
904 | int young; | |
905 | ||
906 | spin_lock(&kvm->mmu_lock); | |
907 | young = kvm_age_hva(kvm, address); | |
908 | spin_unlock(&kvm->mmu_lock); | |
909 | ||
910 | if (young) | |
911 | kvm_flush_remote_tlbs(kvm); | |
912 | ||
913 | return young; | |
914 | } | |
915 | ||
85db06e5 MT |
916 | static void kvm_mmu_notifier_release(struct mmu_notifier *mn, |
917 | struct mm_struct *mm) | |
918 | { | |
919 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
920 | kvm_arch_flush_shadow(kvm); | |
921 | } | |
922 | ||
e930bffe AA |
923 | static const struct mmu_notifier_ops kvm_mmu_notifier_ops = { |
924 | .invalidate_page = kvm_mmu_notifier_invalidate_page, | |
925 | .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start, | |
926 | .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end, | |
927 | .clear_flush_young = kvm_mmu_notifier_clear_flush_young, | |
85db06e5 | 928 | .release = kvm_mmu_notifier_release, |
e930bffe AA |
929 | }; |
930 | #endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */ | |
931 | ||
f17abe9a | 932 | static struct kvm *kvm_create_vm(void) |
6aa8b732 | 933 | { |
d19a9cd2 | 934 | struct kvm *kvm = kvm_arch_create_vm(); |
5f94c174 LV |
935 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
936 | struct page *page; | |
937 | #endif | |
6aa8b732 | 938 | |
d19a9cd2 ZX |
939 | if (IS_ERR(kvm)) |
940 | goto out; | |
75858a84 | 941 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
399ec807 | 942 | INIT_LIST_HEAD(&kvm->irq_routing); |
75858a84 AK |
943 | INIT_HLIST_HEAD(&kvm->mask_notifier_list); |
944 | #endif | |
6aa8b732 | 945 | |
5f94c174 LV |
946 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
947 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
948 | if (!page) { | |
949 | kfree(kvm); | |
950 | return ERR_PTR(-ENOMEM); | |
951 | } | |
952 | kvm->coalesced_mmio_ring = | |
953 | (struct kvm_coalesced_mmio_ring *)page_address(page); | |
954 | #endif | |
955 | ||
e930bffe AA |
956 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
957 | { | |
958 | int err; | |
959 | kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops; | |
960 | err = mmu_notifier_register(&kvm->mmu_notifier, current->mm); | |
961 | if (err) { | |
962 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
963 | put_page(page); | |
964 | #endif | |
965 | kfree(kvm); | |
966 | return ERR_PTR(err); | |
967 | } | |
968 | } | |
969 | #endif | |
970 | ||
6d4e4c4f AK |
971 | kvm->mm = current->mm; |
972 | atomic_inc(&kvm->mm->mm_count); | |
aaee2c94 | 973 | spin_lock_init(&kvm->mmu_lock); |
74906345 | 974 | kvm_io_bus_init(&kvm->pio_bus); |
11ec2804 | 975 | mutex_init(&kvm->lock); |
2eeb2e94 | 976 | kvm_io_bus_init(&kvm->mmio_bus); |
72dc67a6 | 977 | init_rwsem(&kvm->slots_lock); |
d39f13b0 | 978 | atomic_set(&kvm->users_count, 1); |
5e58cfe4 RR |
979 | spin_lock(&kvm_lock); |
980 | list_add(&kvm->vm_list, &vm_list); | |
981 | spin_unlock(&kvm_lock); | |
5f94c174 LV |
982 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
983 | kvm_coalesced_mmio_init(kvm); | |
984 | #endif | |
d19a9cd2 | 985 | out: |
f17abe9a AK |
986 | return kvm; |
987 | } | |
988 | ||
6aa8b732 AK |
989 | /* |
990 | * Free any memory in @free but not in @dont. | |
991 | */ | |
992 | static void kvm_free_physmem_slot(struct kvm_memory_slot *free, | |
993 | struct kvm_memory_slot *dont) | |
994 | { | |
290fc38d IE |
995 | if (!dont || free->rmap != dont->rmap) |
996 | vfree(free->rmap); | |
6aa8b732 AK |
997 | |
998 | if (!dont || free->dirty_bitmap != dont->dirty_bitmap) | |
999 | vfree(free->dirty_bitmap); | |
1000 | ||
05da4558 MT |
1001 | if (!dont || free->lpage_info != dont->lpage_info) |
1002 | vfree(free->lpage_info); | |
1003 | ||
6aa8b732 | 1004 | free->npages = 0; |
8b6d44c7 | 1005 | free->dirty_bitmap = NULL; |
8d4e1288 | 1006 | free->rmap = NULL; |
05da4558 | 1007 | free->lpage_info = NULL; |
6aa8b732 AK |
1008 | } |
1009 | ||
d19a9cd2 | 1010 | void kvm_free_physmem(struct kvm *kvm) |
6aa8b732 AK |
1011 | { |
1012 | int i; | |
1013 | ||
1014 | for (i = 0; i < kvm->nmemslots; ++i) | |
8b6d44c7 | 1015 | kvm_free_physmem_slot(&kvm->memslots[i], NULL); |
6aa8b732 AK |
1016 | } |
1017 | ||
f17abe9a AK |
1018 | static void kvm_destroy_vm(struct kvm *kvm) |
1019 | { | |
6d4e4c4f AK |
1020 | struct mm_struct *mm = kvm->mm; |
1021 | ||
ad8ba2cd | 1022 | kvm_arch_sync_events(kvm); |
133de902 AK |
1023 | spin_lock(&kvm_lock); |
1024 | list_del(&kvm->vm_list); | |
1025 | spin_unlock(&kvm_lock); | |
399ec807 | 1026 | kvm_free_irq_routing(kvm); |
74906345 | 1027 | kvm_io_bus_destroy(&kvm->pio_bus); |
2eeb2e94 | 1028 | kvm_io_bus_destroy(&kvm->mmio_bus); |
5f94c174 LV |
1029 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
1030 | if (kvm->coalesced_mmio_ring != NULL) | |
1031 | free_page((unsigned long)kvm->coalesced_mmio_ring); | |
e930bffe AA |
1032 | #endif |
1033 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) | |
1034 | mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm); | |
5f94c174 | 1035 | #endif |
d19a9cd2 | 1036 | kvm_arch_destroy_vm(kvm); |
6d4e4c4f | 1037 | mmdrop(mm); |
f17abe9a AK |
1038 | } |
1039 | ||
d39f13b0 IE |
1040 | void kvm_get_kvm(struct kvm *kvm) |
1041 | { | |
1042 | atomic_inc(&kvm->users_count); | |
1043 | } | |
1044 | EXPORT_SYMBOL_GPL(kvm_get_kvm); | |
1045 | ||
1046 | void kvm_put_kvm(struct kvm *kvm) | |
1047 | { | |
1048 | if (atomic_dec_and_test(&kvm->users_count)) | |
1049 | kvm_destroy_vm(kvm); | |
1050 | } | |
1051 | EXPORT_SYMBOL_GPL(kvm_put_kvm); | |
1052 | ||
1053 | ||
f17abe9a AK |
1054 | static int kvm_vm_release(struct inode *inode, struct file *filp) |
1055 | { | |
1056 | struct kvm *kvm = filp->private_data; | |
1057 | ||
d39f13b0 | 1058 | kvm_put_kvm(kvm); |
6aa8b732 AK |
1059 | return 0; |
1060 | } | |
1061 | ||
6aa8b732 AK |
1062 | /* |
1063 | * Allocate some memory and give it an address in the guest physical address | |
1064 | * space. | |
1065 | * | |
1066 | * Discontiguous memory is allowed, mostly for framebuffers. | |
f78e0e2e | 1067 | * |
10589a46 | 1068 | * Must be called holding mmap_sem for write. |
6aa8b732 | 1069 | */ |
f78e0e2e SY |
1070 | int __kvm_set_memory_region(struct kvm *kvm, |
1071 | struct kvm_userspace_memory_region *mem, | |
1072 | int user_alloc) | |
6aa8b732 AK |
1073 | { |
1074 | int r; | |
1075 | gfn_t base_gfn; | |
1076 | unsigned long npages; | |
99894a79 | 1077 | int largepages; |
6aa8b732 AK |
1078 | unsigned long i; |
1079 | struct kvm_memory_slot *memslot; | |
1080 | struct kvm_memory_slot old, new; | |
6aa8b732 AK |
1081 | |
1082 | r = -EINVAL; | |
1083 | /* General sanity checks */ | |
1084 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
1085 | goto out; | |
1086 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
1087 | goto out; | |
e7cacd40 | 1088 | if (user_alloc && (mem->userspace_addr & (PAGE_SIZE - 1))) |
78749809 | 1089 | goto out; |
e0d62c7f | 1090 | if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS) |
6aa8b732 AK |
1091 | goto out; |
1092 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
1093 | goto out; | |
1094 | ||
1095 | memslot = &kvm->memslots[mem->slot]; | |
1096 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; | |
1097 | npages = mem->memory_size >> PAGE_SHIFT; | |
1098 | ||
1099 | if (!npages) | |
1100 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
1101 | ||
6aa8b732 AK |
1102 | new = old = *memslot; |
1103 | ||
1104 | new.base_gfn = base_gfn; | |
1105 | new.npages = npages; | |
1106 | new.flags = mem->flags; | |
1107 | ||
1108 | /* Disallow changing a memory slot's size. */ | |
1109 | r = -EINVAL; | |
1110 | if (npages && old.npages && npages != old.npages) | |
f78e0e2e | 1111 | goto out_free; |
6aa8b732 AK |
1112 | |
1113 | /* Check for overlaps */ | |
1114 | r = -EEXIST; | |
1115 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
1116 | struct kvm_memory_slot *s = &kvm->memslots[i]; | |
1117 | ||
4cd481f6 | 1118 | if (s == memslot || !s->npages) |
6aa8b732 AK |
1119 | continue; |
1120 | if (!((base_gfn + npages <= s->base_gfn) || | |
1121 | (base_gfn >= s->base_gfn + s->npages))) | |
f78e0e2e | 1122 | goto out_free; |
6aa8b732 | 1123 | } |
6aa8b732 | 1124 | |
6aa8b732 AK |
1125 | /* Free page dirty bitmap if unneeded */ |
1126 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 1127 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
1128 | |
1129 | r = -ENOMEM; | |
1130 | ||
1131 | /* Allocate if a slot is being created */ | |
eff0114a | 1132 | #ifndef CONFIG_S390 |
8d4e1288 | 1133 | if (npages && !new.rmap) { |
d77c26fc | 1134 | new.rmap = vmalloc(npages * sizeof(struct page *)); |
290fc38d IE |
1135 | |
1136 | if (!new.rmap) | |
f78e0e2e | 1137 | goto out_free; |
290fc38d | 1138 | |
290fc38d | 1139 | memset(new.rmap, 0, npages * sizeof(*new.rmap)); |
8d4e1288 | 1140 | |
80b14b5b | 1141 | new.user_alloc = user_alloc; |
604b38ac AA |
1142 | /* |
1143 | * hva_to_rmmap() serialzies with the mmu_lock and to be | |
1144 | * safe it has to ignore memslots with !user_alloc && | |
1145 | * !userspace_addr. | |
1146 | */ | |
1147 | if (user_alloc) | |
1148 | new.userspace_addr = mem->userspace_addr; | |
1149 | else | |
1150 | new.userspace_addr = 0; | |
6aa8b732 | 1151 | } |
05da4558 | 1152 | if (npages && !new.lpage_info) { |
99894a79 AK |
1153 | largepages = 1 + (base_gfn + npages - 1) / KVM_PAGES_PER_HPAGE; |
1154 | largepages -= base_gfn / KVM_PAGES_PER_HPAGE; | |
05da4558 MT |
1155 | |
1156 | new.lpage_info = vmalloc(largepages * sizeof(*new.lpage_info)); | |
1157 | ||
1158 | if (!new.lpage_info) | |
1159 | goto out_free; | |
1160 | ||
1161 | memset(new.lpage_info, 0, largepages * sizeof(*new.lpage_info)); | |
1162 | ||
1163 | if (base_gfn % KVM_PAGES_PER_HPAGE) | |
1164 | new.lpage_info[0].write_count = 1; | |
1165 | if ((base_gfn+npages) % KVM_PAGES_PER_HPAGE) | |
1166 | new.lpage_info[largepages-1].write_count = 1; | |
1167 | } | |
6aa8b732 AK |
1168 | |
1169 | /* Allocate page dirty bitmap if needed */ | |
1170 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
1171 | unsigned dirty_bytes = ALIGN(npages, BITS_PER_LONG) / 8; | |
1172 | ||
1173 | new.dirty_bitmap = vmalloc(dirty_bytes); | |
1174 | if (!new.dirty_bitmap) | |
f78e0e2e | 1175 | goto out_free; |
6aa8b732 AK |
1176 | memset(new.dirty_bitmap, 0, dirty_bytes); |
1177 | } | |
eff0114a | 1178 | #endif /* not defined CONFIG_S390 */ |
6aa8b732 | 1179 | |
34d4cb8f MT |
1180 | if (!npages) |
1181 | kvm_arch_flush_shadow(kvm); | |
1182 | ||
604b38ac AA |
1183 | spin_lock(&kvm->mmu_lock); |
1184 | if (mem->slot >= kvm->nmemslots) | |
1185 | kvm->nmemslots = mem->slot + 1; | |
1186 | ||
3ad82a7e | 1187 | *memslot = new; |
604b38ac | 1188 | spin_unlock(&kvm->mmu_lock); |
3ad82a7e | 1189 | |
0de10343 ZX |
1190 | r = kvm_arch_set_memory_region(kvm, mem, old, user_alloc); |
1191 | if (r) { | |
604b38ac | 1192 | spin_lock(&kvm->mmu_lock); |
0de10343 | 1193 | *memslot = old; |
604b38ac | 1194 | spin_unlock(&kvm->mmu_lock); |
0de10343 | 1195 | goto out_free; |
82ce2c96 IE |
1196 | } |
1197 | ||
6f897248 GC |
1198 | kvm_free_physmem_slot(&old, npages ? &new : NULL); |
1199 | /* Slot deletion case: we have to update the current slot */ | |
1200 | if (!npages) | |
1201 | *memslot = old; | |
8a98f664 | 1202 | #ifdef CONFIG_DMAR |
62c476c7 BAY |
1203 | /* map the pages in iommu page table */ |
1204 | r = kvm_iommu_map_pages(kvm, base_gfn, npages); | |
1205 | if (r) | |
1206 | goto out; | |
8a98f664 | 1207 | #endif |
6aa8b732 AK |
1208 | return 0; |
1209 | ||
f78e0e2e | 1210 | out_free: |
6aa8b732 AK |
1211 | kvm_free_physmem_slot(&new, &old); |
1212 | out: | |
1213 | return r; | |
210c7c4d IE |
1214 | |
1215 | } | |
f78e0e2e SY |
1216 | EXPORT_SYMBOL_GPL(__kvm_set_memory_region); |
1217 | ||
1218 | int kvm_set_memory_region(struct kvm *kvm, | |
1219 | struct kvm_userspace_memory_region *mem, | |
1220 | int user_alloc) | |
1221 | { | |
1222 | int r; | |
1223 | ||
72dc67a6 | 1224 | down_write(&kvm->slots_lock); |
f78e0e2e | 1225 | r = __kvm_set_memory_region(kvm, mem, user_alloc); |
72dc67a6 | 1226 | up_write(&kvm->slots_lock); |
f78e0e2e SY |
1227 | return r; |
1228 | } | |
210c7c4d IE |
1229 | EXPORT_SYMBOL_GPL(kvm_set_memory_region); |
1230 | ||
1fe779f8 CO |
1231 | int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
1232 | struct | |
1233 | kvm_userspace_memory_region *mem, | |
1234 | int user_alloc) | |
210c7c4d | 1235 | { |
e0d62c7f IE |
1236 | if (mem->slot >= KVM_MEMORY_SLOTS) |
1237 | return -EINVAL; | |
210c7c4d | 1238 | return kvm_set_memory_region(kvm, mem, user_alloc); |
6aa8b732 AK |
1239 | } |
1240 | ||
5bb064dc ZX |
1241 | int kvm_get_dirty_log(struct kvm *kvm, |
1242 | struct kvm_dirty_log *log, int *is_dirty) | |
6aa8b732 AK |
1243 | { |
1244 | struct kvm_memory_slot *memslot; | |
1245 | int r, i; | |
1246 | int n; | |
1247 | unsigned long any = 0; | |
1248 | ||
6aa8b732 AK |
1249 | r = -EINVAL; |
1250 | if (log->slot >= KVM_MEMORY_SLOTS) | |
1251 | goto out; | |
1252 | ||
1253 | memslot = &kvm->memslots[log->slot]; | |
1254 | r = -ENOENT; | |
1255 | if (!memslot->dirty_bitmap) | |
1256 | goto out; | |
1257 | ||
cd1a4a98 | 1258 | n = ALIGN(memslot->npages, BITS_PER_LONG) / 8; |
6aa8b732 | 1259 | |
cd1a4a98 | 1260 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
1261 | any = memslot->dirty_bitmap[i]; |
1262 | ||
1263 | r = -EFAULT; | |
1264 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
1265 | goto out; | |
1266 | ||
5bb064dc ZX |
1267 | if (any) |
1268 | *is_dirty = 1; | |
6aa8b732 AK |
1269 | |
1270 | r = 0; | |
6aa8b732 | 1271 | out: |
6aa8b732 AK |
1272 | return r; |
1273 | } | |
1274 | ||
cea7bb21 IE |
1275 | int is_error_page(struct page *page) |
1276 | { | |
1277 | return page == bad_page; | |
1278 | } | |
1279 | EXPORT_SYMBOL_GPL(is_error_page); | |
1280 | ||
35149e21 AL |
1281 | int is_error_pfn(pfn_t pfn) |
1282 | { | |
1283 | return pfn == bad_pfn; | |
1284 | } | |
1285 | EXPORT_SYMBOL_GPL(is_error_pfn); | |
1286 | ||
f9d46eb0 IE |
1287 | static inline unsigned long bad_hva(void) |
1288 | { | |
1289 | return PAGE_OFFSET; | |
1290 | } | |
1291 | ||
1292 | int kvm_is_error_hva(unsigned long addr) | |
1293 | { | |
1294 | return addr == bad_hva(); | |
1295 | } | |
1296 | EXPORT_SYMBOL_GPL(kvm_is_error_hva); | |
1297 | ||
2843099f | 1298 | struct kvm_memory_slot *gfn_to_memslot_unaliased(struct kvm *kvm, gfn_t gfn) |
6aa8b732 AK |
1299 | { |
1300 | int i; | |
1301 | ||
1302 | for (i = 0; i < kvm->nmemslots; ++i) { | |
1303 | struct kvm_memory_slot *memslot = &kvm->memslots[i]; | |
1304 | ||
1305 | if (gfn >= memslot->base_gfn | |
1306 | && gfn < memslot->base_gfn + memslot->npages) | |
1307 | return memslot; | |
1308 | } | |
8b6d44c7 | 1309 | return NULL; |
6aa8b732 | 1310 | } |
2843099f | 1311 | EXPORT_SYMBOL_GPL(gfn_to_memslot_unaliased); |
e8207547 AK |
1312 | |
1313 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
1314 | { | |
1315 | gfn = unalias_gfn(kvm, gfn); | |
2843099f | 1316 | return gfn_to_memslot_unaliased(kvm, gfn); |
e8207547 | 1317 | } |
6aa8b732 | 1318 | |
e0d62c7f IE |
1319 | int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
1320 | { | |
1321 | int i; | |
1322 | ||
1323 | gfn = unalias_gfn(kvm, gfn); | |
1324 | for (i = 0; i < KVM_MEMORY_SLOTS; ++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 1; | |
1330 | } | |
1331 | return 0; | |
1332 | } | |
1333 | EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); | |
1334 | ||
05da4558 | 1335 | unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn) |
539cb660 IE |
1336 | { |
1337 | struct kvm_memory_slot *slot; | |
1338 | ||
1339 | gfn = unalias_gfn(kvm, gfn); | |
2843099f | 1340 | slot = gfn_to_memslot_unaliased(kvm, gfn); |
539cb660 IE |
1341 | if (!slot) |
1342 | return bad_hva(); | |
1343 | return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE); | |
1344 | } | |
0d150298 | 1345 | EXPORT_SYMBOL_GPL(gfn_to_hva); |
539cb660 | 1346 | |
35149e21 | 1347 | pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn) |
954bbbc2 | 1348 | { |
8d4e1288 | 1349 | struct page *page[1]; |
539cb660 | 1350 | unsigned long addr; |
8d4e1288 | 1351 | int npages; |
2e2e3738 | 1352 | pfn_t pfn; |
954bbbc2 | 1353 | |
60395224 AK |
1354 | might_sleep(); |
1355 | ||
539cb660 IE |
1356 | addr = gfn_to_hva(kvm, gfn); |
1357 | if (kvm_is_error_hva(addr)) { | |
8a7ae055 | 1358 | get_page(bad_page); |
35149e21 | 1359 | return page_to_pfn(bad_page); |
8a7ae055 | 1360 | } |
8d4e1288 | 1361 | |
4c2155ce | 1362 | npages = get_user_pages_fast(addr, 1, 1, page); |
539cb660 | 1363 | |
2e2e3738 AL |
1364 | if (unlikely(npages != 1)) { |
1365 | struct vm_area_struct *vma; | |
1366 | ||
4c2155ce | 1367 | down_read(¤t->mm->mmap_sem); |
2e2e3738 | 1368 | vma = find_vma(current->mm, addr); |
4c2155ce | 1369 | |
2e2e3738 AL |
1370 | if (vma == NULL || addr < vma->vm_start || |
1371 | !(vma->vm_flags & VM_PFNMAP)) { | |
4c2155ce | 1372 | up_read(¤t->mm->mmap_sem); |
2e2e3738 AL |
1373 | get_page(bad_page); |
1374 | return page_to_pfn(bad_page); | |
1375 | } | |
1376 | ||
1377 | pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; | |
4c2155ce | 1378 | up_read(¤t->mm->mmap_sem); |
c77fb9dc | 1379 | BUG_ON(!kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
1380 | } else |
1381 | pfn = page_to_pfn(page[0]); | |
8d4e1288 | 1382 | |
2e2e3738 | 1383 | return pfn; |
35149e21 AL |
1384 | } |
1385 | ||
1386 | EXPORT_SYMBOL_GPL(gfn_to_pfn); | |
1387 | ||
1388 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) | |
1389 | { | |
2e2e3738 AL |
1390 | pfn_t pfn; |
1391 | ||
1392 | pfn = gfn_to_pfn(kvm, gfn); | |
c77fb9dc | 1393 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 AL |
1394 | return pfn_to_page(pfn); |
1395 | ||
c77fb9dc | 1396 | WARN_ON(kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
1397 | |
1398 | get_page(bad_page); | |
1399 | return bad_page; | |
954bbbc2 | 1400 | } |
aab61cc0 | 1401 | |
954bbbc2 AK |
1402 | EXPORT_SYMBOL_GPL(gfn_to_page); |
1403 | ||
b4231d61 IE |
1404 | void kvm_release_page_clean(struct page *page) |
1405 | { | |
35149e21 | 1406 | kvm_release_pfn_clean(page_to_pfn(page)); |
b4231d61 IE |
1407 | } |
1408 | EXPORT_SYMBOL_GPL(kvm_release_page_clean); | |
1409 | ||
35149e21 AL |
1410 | void kvm_release_pfn_clean(pfn_t pfn) |
1411 | { | |
c77fb9dc | 1412 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1413 | put_page(pfn_to_page(pfn)); |
35149e21 AL |
1414 | } |
1415 | EXPORT_SYMBOL_GPL(kvm_release_pfn_clean); | |
1416 | ||
b4231d61 | 1417 | void kvm_release_page_dirty(struct page *page) |
8a7ae055 | 1418 | { |
35149e21 AL |
1419 | kvm_release_pfn_dirty(page_to_pfn(page)); |
1420 | } | |
1421 | EXPORT_SYMBOL_GPL(kvm_release_page_dirty); | |
1422 | ||
1423 | void kvm_release_pfn_dirty(pfn_t pfn) | |
1424 | { | |
1425 | kvm_set_pfn_dirty(pfn); | |
1426 | kvm_release_pfn_clean(pfn); | |
1427 | } | |
1428 | EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty); | |
1429 | ||
1430 | void kvm_set_page_dirty(struct page *page) | |
1431 | { | |
1432 | kvm_set_pfn_dirty(page_to_pfn(page)); | |
1433 | } | |
1434 | EXPORT_SYMBOL_GPL(kvm_set_page_dirty); | |
1435 | ||
1436 | void kvm_set_pfn_dirty(pfn_t pfn) | |
1437 | { | |
c77fb9dc | 1438 | if (!kvm_is_mmio_pfn(pfn)) { |
2e2e3738 AL |
1439 | struct page *page = pfn_to_page(pfn); |
1440 | if (!PageReserved(page)) | |
1441 | SetPageDirty(page); | |
1442 | } | |
8a7ae055 | 1443 | } |
35149e21 AL |
1444 | EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty); |
1445 | ||
1446 | void kvm_set_pfn_accessed(pfn_t pfn) | |
1447 | { | |
c77fb9dc | 1448 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1449 | mark_page_accessed(pfn_to_page(pfn)); |
35149e21 AL |
1450 | } |
1451 | EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed); | |
1452 | ||
1453 | void kvm_get_pfn(pfn_t pfn) | |
1454 | { | |
c77fb9dc | 1455 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1456 | get_page(pfn_to_page(pfn)); |
35149e21 AL |
1457 | } |
1458 | EXPORT_SYMBOL_GPL(kvm_get_pfn); | |
8a7ae055 | 1459 | |
195aefde IE |
1460 | static int next_segment(unsigned long len, int offset) |
1461 | { | |
1462 | if (len > PAGE_SIZE - offset) | |
1463 | return PAGE_SIZE - offset; | |
1464 | else | |
1465 | return len; | |
1466 | } | |
1467 | ||
1468 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, | |
1469 | int len) | |
1470 | { | |
e0506bcb IE |
1471 | int r; |
1472 | unsigned long addr; | |
195aefde | 1473 | |
e0506bcb IE |
1474 | addr = gfn_to_hva(kvm, gfn); |
1475 | if (kvm_is_error_hva(addr)) | |
1476 | return -EFAULT; | |
1477 | r = copy_from_user(data, (void __user *)addr + offset, len); | |
1478 | if (r) | |
195aefde | 1479 | return -EFAULT; |
195aefde IE |
1480 | return 0; |
1481 | } | |
1482 | EXPORT_SYMBOL_GPL(kvm_read_guest_page); | |
1483 | ||
1484 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len) | |
1485 | { | |
1486 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1487 | int seg; | |
1488 | int offset = offset_in_page(gpa); | |
1489 | int ret; | |
1490 | ||
1491 | while ((seg = next_segment(len, offset)) != 0) { | |
1492 | ret = kvm_read_guest_page(kvm, gfn, data, offset, seg); | |
1493 | if (ret < 0) | |
1494 | return ret; | |
1495 | offset = 0; | |
1496 | len -= seg; | |
1497 | data += seg; | |
1498 | ++gfn; | |
1499 | } | |
1500 | return 0; | |
1501 | } | |
1502 | EXPORT_SYMBOL_GPL(kvm_read_guest); | |
1503 | ||
7ec54588 MT |
1504 | int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data, |
1505 | unsigned long len) | |
1506 | { | |
1507 | int r; | |
1508 | unsigned long addr; | |
1509 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1510 | int offset = offset_in_page(gpa); | |
1511 | ||
1512 | addr = gfn_to_hva(kvm, gfn); | |
1513 | if (kvm_is_error_hva(addr)) | |
1514 | return -EFAULT; | |
0aac03f0 | 1515 | pagefault_disable(); |
7ec54588 | 1516 | r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len); |
0aac03f0 | 1517 | pagefault_enable(); |
7ec54588 MT |
1518 | if (r) |
1519 | return -EFAULT; | |
1520 | return 0; | |
1521 | } | |
1522 | EXPORT_SYMBOL(kvm_read_guest_atomic); | |
1523 | ||
195aefde IE |
1524 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, |
1525 | int offset, int len) | |
1526 | { | |
e0506bcb IE |
1527 | int r; |
1528 | unsigned long addr; | |
195aefde | 1529 | |
e0506bcb IE |
1530 | addr = gfn_to_hva(kvm, gfn); |
1531 | if (kvm_is_error_hva(addr)) | |
1532 | return -EFAULT; | |
1533 | r = copy_to_user((void __user *)addr + offset, data, len); | |
1534 | if (r) | |
195aefde | 1535 | return -EFAULT; |
195aefde IE |
1536 | mark_page_dirty(kvm, gfn); |
1537 | return 0; | |
1538 | } | |
1539 | EXPORT_SYMBOL_GPL(kvm_write_guest_page); | |
1540 | ||
1541 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
1542 | unsigned long len) | |
1543 | { | |
1544 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1545 | int seg; | |
1546 | int offset = offset_in_page(gpa); | |
1547 | int ret; | |
1548 | ||
1549 | while ((seg = next_segment(len, offset)) != 0) { | |
1550 | ret = kvm_write_guest_page(kvm, gfn, data, offset, seg); | |
1551 | if (ret < 0) | |
1552 | return ret; | |
1553 | offset = 0; | |
1554 | len -= seg; | |
1555 | data += seg; | |
1556 | ++gfn; | |
1557 | } | |
1558 | return 0; | |
1559 | } | |
1560 | ||
1561 | int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len) | |
1562 | { | |
3e021bf5 | 1563 | return kvm_write_guest_page(kvm, gfn, empty_zero_page, offset, len); |
195aefde IE |
1564 | } |
1565 | EXPORT_SYMBOL_GPL(kvm_clear_guest_page); | |
1566 | ||
1567 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len) | |
1568 | { | |
1569 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1570 | int seg; | |
1571 | int offset = offset_in_page(gpa); | |
1572 | int ret; | |
1573 | ||
1574 | while ((seg = next_segment(len, offset)) != 0) { | |
1575 | ret = kvm_clear_guest_page(kvm, gfn, offset, seg); | |
1576 | if (ret < 0) | |
1577 | return ret; | |
1578 | offset = 0; | |
1579 | len -= seg; | |
1580 | ++gfn; | |
1581 | } | |
1582 | return 0; | |
1583 | } | |
1584 | EXPORT_SYMBOL_GPL(kvm_clear_guest); | |
1585 | ||
6aa8b732 AK |
1586 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
1587 | { | |
31389947 | 1588 | struct kvm_memory_slot *memslot; |
6aa8b732 | 1589 | |
3b6fff19 | 1590 | gfn = unalias_gfn(kvm, gfn); |
2843099f | 1591 | memslot = gfn_to_memslot_unaliased(kvm, gfn); |
7e9d619d RR |
1592 | if (memslot && memslot->dirty_bitmap) { |
1593 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 1594 | |
7e9d619d RR |
1595 | /* avoid RMW */ |
1596 | if (!test_bit(rel_gfn, memslot->dirty_bitmap)) | |
1597 | set_bit(rel_gfn, memslot->dirty_bitmap); | |
6aa8b732 AK |
1598 | } |
1599 | } | |
1600 | ||
b6958ce4 ED |
1601 | /* |
1602 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
1603 | */ | |
8776e519 | 1604 | void kvm_vcpu_block(struct kvm_vcpu *vcpu) |
d3bef15f | 1605 | { |
e5c239cf MT |
1606 | DEFINE_WAIT(wait); |
1607 | ||
1608 | for (;;) { | |
1609 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1610 | ||
d7690175 MT |
1611 | if (kvm_cpu_has_interrupt(vcpu) || |
1612 | kvm_cpu_has_pending_timer(vcpu) || | |
1613 | kvm_arch_vcpu_runnable(vcpu)) { | |
1614 | set_bit(KVM_REQ_UNHALT, &vcpu->requests); | |
e5c239cf | 1615 | break; |
d7690175 | 1616 | } |
e5c239cf MT |
1617 | if (signal_pending(current)) |
1618 | break; | |
1619 | ||
b6958ce4 ED |
1620 | vcpu_put(vcpu); |
1621 | schedule(); | |
1622 | vcpu_load(vcpu); | |
1623 | } | |
d3bef15f | 1624 | |
e5c239cf | 1625 | finish_wait(&vcpu->wq, &wait); |
b6958ce4 ED |
1626 | } |
1627 | ||
6aa8b732 AK |
1628 | void kvm_resched(struct kvm_vcpu *vcpu) |
1629 | { | |
3fca0365 YD |
1630 | if (!need_resched()) |
1631 | return; | |
6aa8b732 | 1632 | cond_resched(); |
6aa8b732 AK |
1633 | } |
1634 | EXPORT_SYMBOL_GPL(kvm_resched); | |
1635 | ||
e4a533a4 | 1636 | static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
9a2bb7f4 AK |
1637 | { |
1638 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
9a2bb7f4 AK |
1639 | struct page *page; |
1640 | ||
e4a533a4 | 1641 | if (vmf->pgoff == 0) |
039576c0 | 1642 | page = virt_to_page(vcpu->run); |
09566765 | 1643 | #ifdef CONFIG_X86 |
e4a533a4 | 1644 | else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET) |
ad312c7c | 1645 | page = virt_to_page(vcpu->arch.pio_data); |
5f94c174 LV |
1646 | #endif |
1647 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
1648 | else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET) | |
1649 | page = virt_to_page(vcpu->kvm->coalesced_mmio_ring); | |
09566765 | 1650 | #endif |
039576c0 | 1651 | else |
e4a533a4 | 1652 | return VM_FAULT_SIGBUS; |
9a2bb7f4 | 1653 | get_page(page); |
e4a533a4 | 1654 | vmf->page = page; |
1655 | return 0; | |
9a2bb7f4 AK |
1656 | } |
1657 | ||
1658 | static struct vm_operations_struct kvm_vcpu_vm_ops = { | |
e4a533a4 | 1659 | .fault = kvm_vcpu_fault, |
9a2bb7f4 AK |
1660 | }; |
1661 | ||
1662 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
1663 | { | |
1664 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
1665 | return 0; | |
1666 | } | |
1667 | ||
bccf2150 AK |
1668 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
1669 | { | |
1670 | struct kvm_vcpu *vcpu = filp->private_data; | |
1671 | ||
66c0b394 | 1672 | kvm_put_kvm(vcpu->kvm); |
bccf2150 AK |
1673 | return 0; |
1674 | } | |
1675 | ||
3d3aab1b | 1676 | static struct file_operations kvm_vcpu_fops = { |
bccf2150 AK |
1677 | .release = kvm_vcpu_release, |
1678 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
1679 | .compat_ioctl = kvm_vcpu_ioctl, | |
9a2bb7f4 | 1680 | .mmap = kvm_vcpu_mmap, |
bccf2150 AK |
1681 | }; |
1682 | ||
1683 | /* | |
1684 | * Allocates an inode for the vcpu. | |
1685 | */ | |
1686 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
1687 | { | |
7d9dbca3 | 1688 | int fd = anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, 0); |
2030a42c | 1689 | if (fd < 0) |
66c0b394 | 1690 | kvm_put_kvm(vcpu->kvm); |
bccf2150 | 1691 | return fd; |
bccf2150 AK |
1692 | } |
1693 | ||
c5ea7660 AK |
1694 | /* |
1695 | * Creates some virtual cpus. Good luck creating more than one. | |
1696 | */ | |
1697 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, int n) | |
1698 | { | |
1699 | int r; | |
1700 | struct kvm_vcpu *vcpu; | |
1701 | ||
c5ea7660 | 1702 | if (!valid_vcpu(n)) |
fb3f0f51 | 1703 | return -EINVAL; |
c5ea7660 | 1704 | |
e9b11c17 | 1705 | vcpu = kvm_arch_vcpu_create(kvm, n); |
fb3f0f51 RR |
1706 | if (IS_ERR(vcpu)) |
1707 | return PTR_ERR(vcpu); | |
c5ea7660 | 1708 | |
15ad7146 AK |
1709 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
1710 | ||
26e5215f AK |
1711 | r = kvm_arch_vcpu_setup(vcpu); |
1712 | if (r) | |
7d8fece6 | 1713 | return r; |
26e5215f | 1714 | |
11ec2804 | 1715 | mutex_lock(&kvm->lock); |
fb3f0f51 RR |
1716 | if (kvm->vcpus[n]) { |
1717 | r = -EEXIST; | |
e9b11c17 | 1718 | goto vcpu_destroy; |
fb3f0f51 RR |
1719 | } |
1720 | kvm->vcpus[n] = vcpu; | |
11ec2804 | 1721 | mutex_unlock(&kvm->lock); |
c5ea7660 | 1722 | |
fb3f0f51 | 1723 | /* Now it's all set up, let userspace reach it */ |
66c0b394 | 1724 | kvm_get_kvm(kvm); |
bccf2150 AK |
1725 | r = create_vcpu_fd(vcpu); |
1726 | if (r < 0) | |
fb3f0f51 RR |
1727 | goto unlink; |
1728 | return r; | |
39c3b86e | 1729 | |
fb3f0f51 | 1730 | unlink: |
11ec2804 | 1731 | mutex_lock(&kvm->lock); |
fb3f0f51 | 1732 | kvm->vcpus[n] = NULL; |
e9b11c17 | 1733 | vcpu_destroy: |
7d8fece6 | 1734 | mutex_unlock(&kvm->lock); |
d40ccc62 | 1735 | kvm_arch_vcpu_destroy(vcpu); |
c5ea7660 AK |
1736 | return r; |
1737 | } | |
1738 | ||
1961d276 AK |
1739 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
1740 | { | |
1741 | if (sigset) { | |
1742 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
1743 | vcpu->sigset_active = 1; | |
1744 | vcpu->sigset = *sigset; | |
1745 | } else | |
1746 | vcpu->sigset_active = 0; | |
1747 | return 0; | |
1748 | } | |
1749 | ||
c1e01514 SY |
1750 | #ifdef __KVM_HAVE_MSIX |
1751 | static int kvm_vm_ioctl_set_msix_nr(struct kvm *kvm, | |
1752 | struct kvm_assigned_msix_nr *entry_nr) | |
1753 | { | |
1754 | int r = 0; | |
1755 | struct kvm_assigned_dev_kernel *adev; | |
1756 | ||
1757 | mutex_lock(&kvm->lock); | |
1758 | ||
1759 | adev = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, | |
1760 | entry_nr->assigned_dev_id); | |
1761 | if (!adev) { | |
1762 | r = -EINVAL; | |
1763 | goto msix_nr_out; | |
1764 | } | |
1765 | ||
1766 | if (adev->entries_nr == 0) { | |
1767 | adev->entries_nr = entry_nr->entry_nr; | |
1768 | if (adev->entries_nr == 0 || | |
1769 | adev->entries_nr >= KVM_MAX_MSIX_PER_DEV) { | |
1770 | r = -EINVAL; | |
1771 | goto msix_nr_out; | |
1772 | } | |
1773 | ||
1774 | adev->host_msix_entries = kzalloc(sizeof(struct msix_entry) * | |
1775 | entry_nr->entry_nr, | |
1776 | GFP_KERNEL); | |
1777 | if (!adev->host_msix_entries) { | |
1778 | r = -ENOMEM; | |
1779 | goto msix_nr_out; | |
1780 | } | |
1781 | adev->guest_msix_entries = kzalloc( | |
1782 | sizeof(struct kvm_guest_msix_entry) * | |
1783 | entry_nr->entry_nr, GFP_KERNEL); | |
1784 | if (!adev->guest_msix_entries) { | |
1785 | kfree(adev->host_msix_entries); | |
1786 | r = -ENOMEM; | |
1787 | goto msix_nr_out; | |
1788 | } | |
1789 | } else /* Not allowed set MSI-X number twice */ | |
1790 | r = -EINVAL; | |
1791 | msix_nr_out: | |
1792 | mutex_unlock(&kvm->lock); | |
1793 | return r; | |
1794 | } | |
1795 | ||
1796 | static int kvm_vm_ioctl_set_msix_entry(struct kvm *kvm, | |
1797 | struct kvm_assigned_msix_entry *entry) | |
1798 | { | |
1799 | int r = 0, i; | |
1800 | struct kvm_assigned_dev_kernel *adev; | |
1801 | ||
1802 | mutex_lock(&kvm->lock); | |
1803 | ||
1804 | adev = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head, | |
1805 | entry->assigned_dev_id); | |
1806 | ||
1807 | if (!adev) { | |
1808 | r = -EINVAL; | |
1809 | goto msix_entry_out; | |
1810 | } | |
1811 | ||
1812 | for (i = 0; i < adev->entries_nr; i++) | |
1813 | if (adev->guest_msix_entries[i].vector == 0 || | |
1814 | adev->guest_msix_entries[i].entry == entry->entry) { | |
1815 | adev->guest_msix_entries[i].entry = entry->entry; | |
1816 | adev->guest_msix_entries[i].vector = entry->gsi; | |
1817 | adev->host_msix_entries[i].entry = entry->entry; | |
1818 | break; | |
1819 | } | |
1820 | if (i == adev->entries_nr) { | |
1821 | r = -ENOSPC; | |
1822 | goto msix_entry_out; | |
1823 | } | |
1824 | ||
1825 | msix_entry_out: | |
1826 | mutex_unlock(&kvm->lock); | |
1827 | ||
1828 | return r; | |
1829 | } | |
1830 | #endif | |
1831 | ||
bccf2150 AK |
1832 | static long kvm_vcpu_ioctl(struct file *filp, |
1833 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 1834 | { |
bccf2150 | 1835 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 1836 | void __user *argp = (void __user *)arg; |
313a3dc7 | 1837 | int r; |
fa3795a7 DH |
1838 | struct kvm_fpu *fpu = NULL; |
1839 | struct kvm_sregs *kvm_sregs = NULL; | |
6aa8b732 | 1840 | |
6d4e4c4f AK |
1841 | if (vcpu->kvm->mm != current->mm) |
1842 | return -EIO; | |
6aa8b732 | 1843 | switch (ioctl) { |
9a2bb7f4 | 1844 | case KVM_RUN: |
f0fe5108 AK |
1845 | r = -EINVAL; |
1846 | if (arg) | |
1847 | goto out; | |
b6c7a5dc | 1848 | r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run); |
6aa8b732 | 1849 | break; |
6aa8b732 | 1850 | case KVM_GET_REGS: { |
3e4bb3ac | 1851 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1852 | |
3e4bb3ac XZ |
1853 | r = -ENOMEM; |
1854 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1855 | if (!kvm_regs) | |
6aa8b732 | 1856 | goto out; |
3e4bb3ac XZ |
1857 | r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs); |
1858 | if (r) | |
1859 | goto out_free1; | |
6aa8b732 | 1860 | r = -EFAULT; |
3e4bb3ac XZ |
1861 | if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs))) |
1862 | goto out_free1; | |
6aa8b732 | 1863 | r = 0; |
3e4bb3ac XZ |
1864 | out_free1: |
1865 | kfree(kvm_regs); | |
6aa8b732 AK |
1866 | break; |
1867 | } | |
1868 | case KVM_SET_REGS: { | |
3e4bb3ac | 1869 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1870 | |
3e4bb3ac XZ |
1871 | r = -ENOMEM; |
1872 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1873 | if (!kvm_regs) | |
6aa8b732 | 1874 | goto out; |
3e4bb3ac XZ |
1875 | r = -EFAULT; |
1876 | if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs))) | |
1877 | goto out_free2; | |
1878 | r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs); | |
6aa8b732 | 1879 | if (r) |
3e4bb3ac | 1880 | goto out_free2; |
6aa8b732 | 1881 | r = 0; |
3e4bb3ac XZ |
1882 | out_free2: |
1883 | kfree(kvm_regs); | |
6aa8b732 AK |
1884 | break; |
1885 | } | |
1886 | case KVM_GET_SREGS: { | |
fa3795a7 DH |
1887 | kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1888 | r = -ENOMEM; | |
1889 | if (!kvm_sregs) | |
1890 | goto out; | |
1891 | r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs); | |
6aa8b732 AK |
1892 | if (r) |
1893 | goto out; | |
1894 | r = -EFAULT; | |
fa3795a7 | 1895 | if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs))) |
6aa8b732 AK |
1896 | goto out; |
1897 | r = 0; | |
1898 | break; | |
1899 | } | |
1900 | case KVM_SET_SREGS: { | |
fa3795a7 DH |
1901 | kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1902 | r = -ENOMEM; | |
1903 | if (!kvm_sregs) | |
1904 | goto out; | |
6aa8b732 | 1905 | r = -EFAULT; |
fa3795a7 | 1906 | if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs))) |
6aa8b732 | 1907 | goto out; |
fa3795a7 | 1908 | r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs); |
6aa8b732 AK |
1909 | if (r) |
1910 | goto out; | |
1911 | r = 0; | |
1912 | break; | |
1913 | } | |
62d9f0db MT |
1914 | case KVM_GET_MP_STATE: { |
1915 | struct kvm_mp_state mp_state; | |
1916 | ||
1917 | r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state); | |
1918 | if (r) | |
1919 | goto out; | |
1920 | r = -EFAULT; | |
1921 | if (copy_to_user(argp, &mp_state, sizeof mp_state)) | |
1922 | goto out; | |
1923 | r = 0; | |
1924 | break; | |
1925 | } | |
1926 | case KVM_SET_MP_STATE: { | |
1927 | struct kvm_mp_state mp_state; | |
1928 | ||
1929 | r = -EFAULT; | |
1930 | if (copy_from_user(&mp_state, argp, sizeof mp_state)) | |
1931 | goto out; | |
1932 | r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state); | |
1933 | if (r) | |
1934 | goto out; | |
1935 | r = 0; | |
1936 | break; | |
1937 | } | |
6aa8b732 AK |
1938 | case KVM_TRANSLATE: { |
1939 | struct kvm_translation tr; | |
1940 | ||
1941 | r = -EFAULT; | |
2f366987 | 1942 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 1943 | goto out; |
8b006791 | 1944 | r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
1945 | if (r) |
1946 | goto out; | |
1947 | r = -EFAULT; | |
2f366987 | 1948 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
1949 | goto out; |
1950 | r = 0; | |
1951 | break; | |
1952 | } | |
d0bfb940 JK |
1953 | case KVM_SET_GUEST_DEBUG: { |
1954 | struct kvm_guest_debug dbg; | |
6aa8b732 AK |
1955 | |
1956 | r = -EFAULT; | |
2f366987 | 1957 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 1958 | goto out; |
d0bfb940 | 1959 | r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg); |
6aa8b732 AK |
1960 | if (r) |
1961 | goto out; | |
1962 | r = 0; | |
1963 | break; | |
1964 | } | |
1961d276 AK |
1965 | case KVM_SET_SIGNAL_MASK: { |
1966 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
1967 | struct kvm_signal_mask kvm_sigmask; | |
1968 | sigset_t sigset, *p; | |
1969 | ||
1970 | p = NULL; | |
1971 | if (argp) { | |
1972 | r = -EFAULT; | |
1973 | if (copy_from_user(&kvm_sigmask, argp, | |
1974 | sizeof kvm_sigmask)) | |
1975 | goto out; | |
1976 | r = -EINVAL; | |
1977 | if (kvm_sigmask.len != sizeof sigset) | |
1978 | goto out; | |
1979 | r = -EFAULT; | |
1980 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
1981 | sizeof sigset)) | |
1982 | goto out; | |
1983 | p = &sigset; | |
1984 | } | |
1985 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset); | |
1986 | break; | |
1987 | } | |
b8836737 | 1988 | case KVM_GET_FPU: { |
fa3795a7 DH |
1989 | fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1990 | r = -ENOMEM; | |
1991 | if (!fpu) | |
1992 | goto out; | |
1993 | r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu); | |
b8836737 AK |
1994 | if (r) |
1995 | goto out; | |
1996 | r = -EFAULT; | |
fa3795a7 | 1997 | if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu))) |
b8836737 AK |
1998 | goto out; |
1999 | r = 0; | |
2000 | break; | |
2001 | } | |
2002 | case KVM_SET_FPU: { | |
fa3795a7 DH |
2003 | fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
2004 | r = -ENOMEM; | |
2005 | if (!fpu) | |
2006 | goto out; | |
b8836737 | 2007 | r = -EFAULT; |
fa3795a7 | 2008 | if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu))) |
b8836737 | 2009 | goto out; |
fa3795a7 | 2010 | r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu); |
b8836737 AK |
2011 | if (r) |
2012 | goto out; | |
2013 | r = 0; | |
2014 | break; | |
2015 | } | |
bccf2150 | 2016 | default: |
313a3dc7 | 2017 | r = kvm_arch_vcpu_ioctl(filp, ioctl, arg); |
bccf2150 AK |
2018 | } |
2019 | out: | |
fa3795a7 DH |
2020 | kfree(fpu); |
2021 | kfree(kvm_sregs); | |
bccf2150 AK |
2022 | return r; |
2023 | } | |
2024 | ||
2025 | static long kvm_vm_ioctl(struct file *filp, | |
2026 | unsigned int ioctl, unsigned long arg) | |
2027 | { | |
2028 | struct kvm *kvm = filp->private_data; | |
2029 | void __user *argp = (void __user *)arg; | |
1fe779f8 | 2030 | int r; |
bccf2150 | 2031 | |
6d4e4c4f AK |
2032 | if (kvm->mm != current->mm) |
2033 | return -EIO; | |
bccf2150 AK |
2034 | switch (ioctl) { |
2035 | case KVM_CREATE_VCPU: | |
2036 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
2037 | if (r < 0) | |
2038 | goto out; | |
2039 | break; | |
6fc138d2 IE |
2040 | case KVM_SET_USER_MEMORY_REGION: { |
2041 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
2042 | ||
2043 | r = -EFAULT; | |
2044 | if (copy_from_user(&kvm_userspace_mem, argp, | |
2045 | sizeof kvm_userspace_mem)) | |
2046 | goto out; | |
2047 | ||
2048 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1); | |
6aa8b732 AK |
2049 | if (r) |
2050 | goto out; | |
2051 | break; | |
2052 | } | |
2053 | case KVM_GET_DIRTY_LOG: { | |
2054 | struct kvm_dirty_log log; | |
2055 | ||
2056 | r = -EFAULT; | |
2f366987 | 2057 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 2058 | goto out; |
2c6f5df9 | 2059 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
2060 | if (r) |
2061 | goto out; | |
2062 | break; | |
2063 | } | |
5f94c174 LV |
2064 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
2065 | case KVM_REGISTER_COALESCED_MMIO: { | |
2066 | struct kvm_coalesced_mmio_zone zone; | |
2067 | r = -EFAULT; | |
2068 | if (copy_from_user(&zone, argp, sizeof zone)) | |
2069 | goto out; | |
2070 | r = -ENXIO; | |
2071 | r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone); | |
2072 | if (r) | |
2073 | goto out; | |
2074 | r = 0; | |
2075 | break; | |
2076 | } | |
2077 | case KVM_UNREGISTER_COALESCED_MMIO: { | |
2078 | struct kvm_coalesced_mmio_zone zone; | |
2079 | r = -EFAULT; | |
2080 | if (copy_from_user(&zone, argp, sizeof zone)) | |
2081 | goto out; | |
2082 | r = -ENXIO; | |
2083 | r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone); | |
2084 | if (r) | |
2085 | goto out; | |
2086 | r = 0; | |
2087 | break; | |
2088 | } | |
8a98f664 XZ |
2089 | #endif |
2090 | #ifdef KVM_CAP_DEVICE_ASSIGNMENT | |
2091 | case KVM_ASSIGN_PCI_DEVICE: { | |
2092 | struct kvm_assigned_pci_dev assigned_dev; | |
2093 | ||
2094 | r = -EFAULT; | |
2095 | if (copy_from_user(&assigned_dev, argp, sizeof assigned_dev)) | |
2096 | goto out; | |
2097 | r = kvm_vm_ioctl_assign_device(kvm, &assigned_dev); | |
2098 | if (r) | |
2099 | goto out; | |
2100 | break; | |
2101 | } | |
2102 | case KVM_ASSIGN_IRQ: { | |
e56d532f SY |
2103 | r = -EOPNOTSUPP; |
2104 | break; | |
2105 | } | |
2106 | #ifdef KVM_CAP_ASSIGN_DEV_IRQ | |
2107 | case KVM_ASSIGN_DEV_IRQ: { | |
8a98f664 XZ |
2108 | struct kvm_assigned_irq assigned_irq; |
2109 | ||
2110 | r = -EFAULT; | |
2111 | if (copy_from_user(&assigned_irq, argp, sizeof assigned_irq)) | |
2112 | goto out; | |
2113 | r = kvm_vm_ioctl_assign_irq(kvm, &assigned_irq); | |
2114 | if (r) | |
2115 | goto out; | |
2116 | break; | |
2117 | } | |
e56d532f SY |
2118 | case KVM_DEASSIGN_DEV_IRQ: { |
2119 | struct kvm_assigned_irq assigned_irq; | |
2120 | ||
2121 | r = -EFAULT; | |
2122 | if (copy_from_user(&assigned_irq, argp, sizeof assigned_irq)) | |
2123 | goto out; | |
2124 | r = kvm_vm_ioctl_deassign_dev_irq(kvm, &assigned_irq); | |
2125 | if (r) | |
2126 | goto out; | |
2127 | break; | |
2128 | } | |
2129 | #endif | |
0a920356 WH |
2130 | #endif |
2131 | #ifdef KVM_CAP_DEVICE_DEASSIGNMENT | |
2132 | case KVM_DEASSIGN_PCI_DEVICE: { | |
2133 | struct kvm_assigned_pci_dev assigned_dev; | |
2134 | ||
2135 | r = -EFAULT; | |
2136 | if (copy_from_user(&assigned_dev, argp, sizeof assigned_dev)) | |
2137 | goto out; | |
2138 | r = kvm_vm_ioctl_deassign_device(kvm, &assigned_dev); | |
2139 | if (r) | |
2140 | goto out; | |
2141 | break; | |
2142 | } | |
399ec807 AK |
2143 | #endif |
2144 | #ifdef KVM_CAP_IRQ_ROUTING | |
2145 | case KVM_SET_GSI_ROUTING: { | |
2146 | struct kvm_irq_routing routing; | |
2147 | struct kvm_irq_routing __user *urouting; | |
2148 | struct kvm_irq_routing_entry *entries; | |
2149 | ||
2150 | r = -EFAULT; | |
2151 | if (copy_from_user(&routing, argp, sizeof(routing))) | |
2152 | goto out; | |
2153 | r = -EINVAL; | |
2154 | if (routing.nr >= KVM_MAX_IRQ_ROUTES) | |
2155 | goto out; | |
2156 | if (routing.flags) | |
2157 | goto out; | |
2158 | r = -ENOMEM; | |
2159 | entries = vmalloc(routing.nr * sizeof(*entries)); | |
2160 | if (!entries) | |
2161 | goto out; | |
2162 | r = -EFAULT; | |
2163 | urouting = argp; | |
2164 | if (copy_from_user(entries, urouting->entries, | |
2165 | routing.nr * sizeof(*entries))) | |
2166 | goto out_free_irq_routing; | |
2167 | r = kvm_set_irq_routing(kvm, entries, routing.nr, | |
2168 | routing.flags); | |
2169 | out_free_irq_routing: | |
2170 | vfree(entries); | |
2171 | break; | |
2172 | } | |
c1e01514 SY |
2173 | #ifdef __KVM_HAVE_MSIX |
2174 | case KVM_ASSIGN_SET_MSIX_NR: { | |
2175 | struct kvm_assigned_msix_nr entry_nr; | |
2176 | r = -EFAULT; | |
2177 | if (copy_from_user(&entry_nr, argp, sizeof entry_nr)) | |
2178 | goto out; | |
2179 | r = kvm_vm_ioctl_set_msix_nr(kvm, &entry_nr); | |
2180 | if (r) | |
2181 | goto out; | |
2182 | break; | |
2183 | } | |
2184 | case KVM_ASSIGN_SET_MSIX_ENTRY: { | |
2185 | struct kvm_assigned_msix_entry entry; | |
2186 | r = -EFAULT; | |
2187 | if (copy_from_user(&entry, argp, sizeof entry)) | |
2188 | goto out; | |
2189 | r = kvm_vm_ioctl_set_msix_entry(kvm, &entry); | |
2190 | if (r) | |
2191 | goto out; | |
2192 | break; | |
2193 | } | |
5f94c174 | 2194 | #endif |
c1e01514 | 2195 | #endif /* KVM_CAP_IRQ_ROUTING */ |
f17abe9a | 2196 | default: |
1fe779f8 | 2197 | r = kvm_arch_vm_ioctl(filp, ioctl, arg); |
f17abe9a AK |
2198 | } |
2199 | out: | |
2200 | return r; | |
2201 | } | |
2202 | ||
e4a533a4 | 2203 | static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
f17abe9a | 2204 | { |
777b3f49 MT |
2205 | struct page *page[1]; |
2206 | unsigned long addr; | |
2207 | int npages; | |
2208 | gfn_t gfn = vmf->pgoff; | |
f17abe9a | 2209 | struct kvm *kvm = vma->vm_file->private_data; |
f17abe9a | 2210 | |
777b3f49 MT |
2211 | addr = gfn_to_hva(kvm, gfn); |
2212 | if (kvm_is_error_hva(addr)) | |
e4a533a4 | 2213 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
2214 | |
2215 | npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page, | |
2216 | NULL); | |
2217 | if (unlikely(npages != 1)) | |
e4a533a4 | 2218 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
2219 | |
2220 | vmf->page = page[0]; | |
e4a533a4 | 2221 | return 0; |
f17abe9a AK |
2222 | } |
2223 | ||
2224 | static struct vm_operations_struct kvm_vm_vm_ops = { | |
e4a533a4 | 2225 | .fault = kvm_vm_fault, |
f17abe9a AK |
2226 | }; |
2227 | ||
2228 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
2229 | { | |
2230 | vma->vm_ops = &kvm_vm_vm_ops; | |
2231 | return 0; | |
2232 | } | |
2233 | ||
3d3aab1b | 2234 | static struct file_operations kvm_vm_fops = { |
f17abe9a AK |
2235 | .release = kvm_vm_release, |
2236 | .unlocked_ioctl = kvm_vm_ioctl, | |
2237 | .compat_ioctl = kvm_vm_ioctl, | |
2238 | .mmap = kvm_vm_mmap, | |
2239 | }; | |
2240 | ||
2241 | static int kvm_dev_ioctl_create_vm(void) | |
2242 | { | |
2030a42c | 2243 | int fd; |
f17abe9a AK |
2244 | struct kvm *kvm; |
2245 | ||
f17abe9a | 2246 | kvm = kvm_create_vm(); |
d6d28168 AK |
2247 | if (IS_ERR(kvm)) |
2248 | return PTR_ERR(kvm); | |
7d9dbca3 | 2249 | fd = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, 0); |
2030a42c | 2250 | if (fd < 0) |
66c0b394 | 2251 | kvm_put_kvm(kvm); |
f17abe9a | 2252 | |
f17abe9a | 2253 | return fd; |
f17abe9a AK |
2254 | } |
2255 | ||
1a811b61 AK |
2256 | static long kvm_dev_ioctl_check_extension_generic(long arg) |
2257 | { | |
2258 | switch (arg) { | |
ca9edaee | 2259 | case KVM_CAP_USER_MEMORY: |
1a811b61 | 2260 | case KVM_CAP_DESTROY_MEMORY_REGION_WORKS: |
4cd481f6 | 2261 | case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS: |
1a811b61 | 2262 | return 1; |
399ec807 AK |
2263 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
2264 | case KVM_CAP_IRQ_ROUTING: | |
36463146 | 2265 | return KVM_MAX_IRQ_ROUTES; |
399ec807 | 2266 | #endif |
1a811b61 AK |
2267 | default: |
2268 | break; | |
2269 | } | |
2270 | return kvm_dev_ioctl_check_extension(arg); | |
2271 | } | |
2272 | ||
f17abe9a AK |
2273 | static long kvm_dev_ioctl(struct file *filp, |
2274 | unsigned int ioctl, unsigned long arg) | |
2275 | { | |
07c45a36 | 2276 | long r = -EINVAL; |
f17abe9a AK |
2277 | |
2278 | switch (ioctl) { | |
2279 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
2280 | r = -EINVAL; |
2281 | if (arg) | |
2282 | goto out; | |
f17abe9a AK |
2283 | r = KVM_API_VERSION; |
2284 | break; | |
2285 | case KVM_CREATE_VM: | |
f0fe5108 AK |
2286 | r = -EINVAL; |
2287 | if (arg) | |
2288 | goto out; | |
f17abe9a AK |
2289 | r = kvm_dev_ioctl_create_vm(); |
2290 | break; | |
018d00d2 | 2291 | case KVM_CHECK_EXTENSION: |
1a811b61 | 2292 | r = kvm_dev_ioctl_check_extension_generic(arg); |
5d308f45 | 2293 | break; |
07c45a36 AK |
2294 | case KVM_GET_VCPU_MMAP_SIZE: |
2295 | r = -EINVAL; | |
2296 | if (arg) | |
2297 | goto out; | |
adb1ff46 AK |
2298 | r = PAGE_SIZE; /* struct kvm_run */ |
2299 | #ifdef CONFIG_X86 | |
2300 | r += PAGE_SIZE; /* pio data page */ | |
5f94c174 LV |
2301 | #endif |
2302 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
2303 | r += PAGE_SIZE; /* coalesced mmio ring page */ | |
adb1ff46 | 2304 | #endif |
07c45a36 | 2305 | break; |
d4c9ff2d FEL |
2306 | case KVM_TRACE_ENABLE: |
2307 | case KVM_TRACE_PAUSE: | |
2308 | case KVM_TRACE_DISABLE: | |
2309 | r = kvm_trace_ioctl(ioctl, arg); | |
2310 | break; | |
6aa8b732 | 2311 | default: |
043405e1 | 2312 | return kvm_arch_dev_ioctl(filp, ioctl, arg); |
6aa8b732 AK |
2313 | } |
2314 | out: | |
2315 | return r; | |
2316 | } | |
2317 | ||
6aa8b732 | 2318 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
2319 | .unlocked_ioctl = kvm_dev_ioctl, |
2320 | .compat_ioctl = kvm_dev_ioctl, | |
6aa8b732 AK |
2321 | }; |
2322 | ||
2323 | static struct miscdevice kvm_dev = { | |
bbe4432e | 2324 | KVM_MINOR, |
6aa8b732 AK |
2325 | "kvm", |
2326 | &kvm_chardev_ops, | |
2327 | }; | |
2328 | ||
1b6c0168 AK |
2329 | static void hardware_enable(void *junk) |
2330 | { | |
2331 | int cpu = raw_smp_processor_id(); | |
2332 | ||
7f59f492 | 2333 | if (cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 2334 | return; |
7f59f492 | 2335 | cpumask_set_cpu(cpu, cpus_hardware_enabled); |
e9b11c17 | 2336 | kvm_arch_hardware_enable(NULL); |
1b6c0168 AK |
2337 | } |
2338 | ||
2339 | static void hardware_disable(void *junk) | |
2340 | { | |
2341 | int cpu = raw_smp_processor_id(); | |
2342 | ||
7f59f492 | 2343 | if (!cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 2344 | return; |
7f59f492 | 2345 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); |
e9b11c17 | 2346 | kvm_arch_hardware_disable(NULL); |
1b6c0168 AK |
2347 | } |
2348 | ||
774c47f1 AK |
2349 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
2350 | void *v) | |
2351 | { | |
2352 | int cpu = (long)v; | |
2353 | ||
1a6f4d7f | 2354 | val &= ~CPU_TASKS_FROZEN; |
774c47f1 | 2355 | switch (val) { |
cec9ad27 | 2356 | case CPU_DYING: |
6ec8a856 AK |
2357 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
2358 | cpu); | |
2359 | hardware_disable(NULL); | |
2360 | break; | |
774c47f1 | 2361 | case CPU_UP_CANCELED: |
43934a38 JK |
2362 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
2363 | cpu); | |
8691e5a8 | 2364 | smp_call_function_single(cpu, hardware_disable, NULL, 1); |
774c47f1 | 2365 | break; |
43934a38 JK |
2366 | case CPU_ONLINE: |
2367 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", | |
2368 | cpu); | |
8691e5a8 | 2369 | smp_call_function_single(cpu, hardware_enable, NULL, 1); |
774c47f1 AK |
2370 | break; |
2371 | } | |
2372 | return NOTIFY_OK; | |
2373 | } | |
2374 | ||
4ecac3fd AK |
2375 | |
2376 | asmlinkage void kvm_handle_fault_on_reboot(void) | |
2377 | { | |
2378 | if (kvm_rebooting) | |
2379 | /* spin while reset goes on */ | |
2380 | while (true) | |
2381 | ; | |
2382 | /* Fault while not rebooting. We want the trace. */ | |
2383 | BUG(); | |
2384 | } | |
2385 | EXPORT_SYMBOL_GPL(kvm_handle_fault_on_reboot); | |
2386 | ||
9a2b85c6 | 2387 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
d77c26fc | 2388 | void *v) |
9a2b85c6 RR |
2389 | { |
2390 | if (val == SYS_RESTART) { | |
2391 | /* | |
2392 | * Some (well, at least mine) BIOSes hang on reboot if | |
2393 | * in vmx root mode. | |
2394 | */ | |
2395 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
4ecac3fd | 2396 | kvm_rebooting = true; |
15c8b6c1 | 2397 | on_each_cpu(hardware_disable, NULL, 1); |
9a2b85c6 RR |
2398 | } |
2399 | return NOTIFY_OK; | |
2400 | } | |
2401 | ||
2402 | static struct notifier_block kvm_reboot_notifier = { | |
2403 | .notifier_call = kvm_reboot, | |
2404 | .priority = 0, | |
2405 | }; | |
2406 | ||
2eeb2e94 GH |
2407 | void kvm_io_bus_init(struct kvm_io_bus *bus) |
2408 | { | |
2409 | memset(bus, 0, sizeof(*bus)); | |
2410 | } | |
2411 | ||
2412 | void kvm_io_bus_destroy(struct kvm_io_bus *bus) | |
2413 | { | |
2414 | int i; | |
2415 | ||
2416 | for (i = 0; i < bus->dev_count; i++) { | |
2417 | struct kvm_io_device *pos = bus->devs[i]; | |
2418 | ||
2419 | kvm_iodevice_destructor(pos); | |
2420 | } | |
2421 | } | |
2422 | ||
92760499 LV |
2423 | struct kvm_io_device *kvm_io_bus_find_dev(struct kvm_io_bus *bus, |
2424 | gpa_t addr, int len, int is_write) | |
2eeb2e94 GH |
2425 | { |
2426 | int i; | |
2427 | ||
2428 | for (i = 0; i < bus->dev_count; i++) { | |
2429 | struct kvm_io_device *pos = bus->devs[i]; | |
2430 | ||
92760499 | 2431 | if (pos->in_range(pos, addr, len, is_write)) |
2eeb2e94 GH |
2432 | return pos; |
2433 | } | |
2434 | ||
2435 | return NULL; | |
2436 | } | |
2437 | ||
2438 | void kvm_io_bus_register_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev) | |
2439 | { | |
2440 | BUG_ON(bus->dev_count > (NR_IOBUS_DEVS-1)); | |
2441 | ||
2442 | bus->devs[bus->dev_count++] = dev; | |
2443 | } | |
2444 | ||
774c47f1 AK |
2445 | static struct notifier_block kvm_cpu_notifier = { |
2446 | .notifier_call = kvm_cpu_hotplug, | |
2447 | .priority = 20, /* must be > scheduler priority */ | |
2448 | }; | |
2449 | ||
8b88b099 | 2450 | static int vm_stat_get(void *_offset, u64 *val) |
ba1389b7 AK |
2451 | { |
2452 | unsigned offset = (long)_offset; | |
ba1389b7 AK |
2453 | struct kvm *kvm; |
2454 | ||
8b88b099 | 2455 | *val = 0; |
ba1389b7 AK |
2456 | spin_lock(&kvm_lock); |
2457 | list_for_each_entry(kvm, &vm_list, vm_list) | |
8b88b099 | 2458 | *val += *(u32 *)((void *)kvm + offset); |
ba1389b7 | 2459 | spin_unlock(&kvm_lock); |
8b88b099 | 2460 | return 0; |
ba1389b7 AK |
2461 | } |
2462 | ||
2463 | DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n"); | |
2464 | ||
8b88b099 | 2465 | static int vcpu_stat_get(void *_offset, u64 *val) |
1165f5fe AK |
2466 | { |
2467 | unsigned offset = (long)_offset; | |
1165f5fe AK |
2468 | struct kvm *kvm; |
2469 | struct kvm_vcpu *vcpu; | |
2470 | int i; | |
2471 | ||
8b88b099 | 2472 | *val = 0; |
1165f5fe AK |
2473 | spin_lock(&kvm_lock); |
2474 | list_for_each_entry(kvm, &vm_list, vm_list) | |
2475 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { | |
fb3f0f51 RR |
2476 | vcpu = kvm->vcpus[i]; |
2477 | if (vcpu) | |
8b88b099 | 2478 | *val += *(u32 *)((void *)vcpu + offset); |
1165f5fe AK |
2479 | } |
2480 | spin_unlock(&kvm_lock); | |
8b88b099 | 2481 | return 0; |
1165f5fe AK |
2482 | } |
2483 | ||
ba1389b7 AK |
2484 | DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n"); |
2485 | ||
2486 | static struct file_operations *stat_fops[] = { | |
2487 | [KVM_STAT_VCPU] = &vcpu_stat_fops, | |
2488 | [KVM_STAT_VM] = &vm_stat_fops, | |
2489 | }; | |
1165f5fe | 2490 | |
a16b043c | 2491 | static void kvm_init_debug(void) |
6aa8b732 AK |
2492 | { |
2493 | struct kvm_stats_debugfs_item *p; | |
2494 | ||
76f7c879 | 2495 | kvm_debugfs_dir = debugfs_create_dir("kvm", NULL); |
6aa8b732 | 2496 | for (p = debugfs_entries; p->name; ++p) |
76f7c879 | 2497 | p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir, |
1165f5fe | 2498 | (void *)(long)p->offset, |
ba1389b7 | 2499 | stat_fops[p->kind]); |
6aa8b732 AK |
2500 | } |
2501 | ||
2502 | static void kvm_exit_debug(void) | |
2503 | { | |
2504 | struct kvm_stats_debugfs_item *p; | |
2505 | ||
2506 | for (p = debugfs_entries; p->name; ++p) | |
2507 | debugfs_remove(p->dentry); | |
76f7c879 | 2508 | debugfs_remove(kvm_debugfs_dir); |
6aa8b732 AK |
2509 | } |
2510 | ||
59ae6c6b AK |
2511 | static int kvm_suspend(struct sys_device *dev, pm_message_t state) |
2512 | { | |
4267c41a | 2513 | hardware_disable(NULL); |
59ae6c6b AK |
2514 | return 0; |
2515 | } | |
2516 | ||
2517 | static int kvm_resume(struct sys_device *dev) | |
2518 | { | |
4267c41a | 2519 | hardware_enable(NULL); |
59ae6c6b AK |
2520 | return 0; |
2521 | } | |
2522 | ||
2523 | static struct sysdev_class kvm_sysdev_class = { | |
af5ca3f4 | 2524 | .name = "kvm", |
59ae6c6b AK |
2525 | .suspend = kvm_suspend, |
2526 | .resume = kvm_resume, | |
2527 | }; | |
2528 | ||
2529 | static struct sys_device kvm_sysdev = { | |
2530 | .id = 0, | |
2531 | .cls = &kvm_sysdev_class, | |
2532 | }; | |
2533 | ||
cea7bb21 | 2534 | struct page *bad_page; |
35149e21 | 2535 | pfn_t bad_pfn; |
6aa8b732 | 2536 | |
15ad7146 AK |
2537 | static inline |
2538 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
2539 | { | |
2540 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
2541 | } | |
2542 | ||
2543 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
2544 | { | |
2545 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2546 | ||
e9b11c17 | 2547 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 AK |
2548 | } |
2549 | ||
2550 | static void kvm_sched_out(struct preempt_notifier *pn, | |
2551 | struct task_struct *next) | |
2552 | { | |
2553 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2554 | ||
e9b11c17 | 2555 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
2556 | } |
2557 | ||
f8c16bba | 2558 | int kvm_init(void *opaque, unsigned int vcpu_size, |
c16f862d | 2559 | struct module *module) |
6aa8b732 AK |
2560 | { |
2561 | int r; | |
002c7f7c | 2562 | int cpu; |
6aa8b732 | 2563 | |
cb498ea2 ZX |
2564 | kvm_init_debug(); |
2565 | ||
f8c16bba ZX |
2566 | r = kvm_arch_init(opaque); |
2567 | if (r) | |
d2308784 | 2568 | goto out_fail; |
cb498ea2 ZX |
2569 | |
2570 | bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
2571 | ||
2572 | if (bad_page == NULL) { | |
2573 | r = -ENOMEM; | |
2574 | goto out; | |
2575 | } | |
2576 | ||
35149e21 AL |
2577 | bad_pfn = page_to_pfn(bad_page); |
2578 | ||
8437a617 | 2579 | if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) { |
7f59f492 RR |
2580 | r = -ENOMEM; |
2581 | goto out_free_0; | |
2582 | } | |
a4c0364b | 2583 | cpumask_clear(cpus_hardware_enabled); |
7f59f492 | 2584 | |
e9b11c17 | 2585 | r = kvm_arch_hardware_setup(); |
6aa8b732 | 2586 | if (r < 0) |
7f59f492 | 2587 | goto out_free_0a; |
6aa8b732 | 2588 | |
002c7f7c YS |
2589 | for_each_online_cpu(cpu) { |
2590 | smp_call_function_single(cpu, | |
e9b11c17 | 2591 | kvm_arch_check_processor_compat, |
8691e5a8 | 2592 | &r, 1); |
002c7f7c | 2593 | if (r < 0) |
d2308784 | 2594 | goto out_free_1; |
002c7f7c YS |
2595 | } |
2596 | ||
15c8b6c1 | 2597 | on_each_cpu(hardware_enable, NULL, 1); |
774c47f1 AK |
2598 | r = register_cpu_notifier(&kvm_cpu_notifier); |
2599 | if (r) | |
d2308784 | 2600 | goto out_free_2; |
6aa8b732 AK |
2601 | register_reboot_notifier(&kvm_reboot_notifier); |
2602 | ||
59ae6c6b AK |
2603 | r = sysdev_class_register(&kvm_sysdev_class); |
2604 | if (r) | |
d2308784 | 2605 | goto out_free_3; |
59ae6c6b AK |
2606 | |
2607 | r = sysdev_register(&kvm_sysdev); | |
2608 | if (r) | |
d2308784 | 2609 | goto out_free_4; |
59ae6c6b | 2610 | |
c16f862d RR |
2611 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
2612 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, | |
56919c5c JP |
2613 | __alignof__(struct kvm_vcpu), |
2614 | 0, NULL); | |
c16f862d RR |
2615 | if (!kvm_vcpu_cache) { |
2616 | r = -ENOMEM; | |
d2308784 | 2617 | goto out_free_5; |
c16f862d RR |
2618 | } |
2619 | ||
6aa8b732 | 2620 | kvm_chardev_ops.owner = module; |
3d3aab1b CB |
2621 | kvm_vm_fops.owner = module; |
2622 | kvm_vcpu_fops.owner = module; | |
6aa8b732 AK |
2623 | |
2624 | r = misc_register(&kvm_dev); | |
2625 | if (r) { | |
d77c26fc | 2626 | printk(KERN_ERR "kvm: misc device register failed\n"); |
6aa8b732 AK |
2627 | goto out_free; |
2628 | } | |
2629 | ||
15ad7146 AK |
2630 | kvm_preempt_ops.sched_in = kvm_sched_in; |
2631 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
2632 | ||
c7addb90 | 2633 | return 0; |
6aa8b732 AK |
2634 | |
2635 | out_free: | |
c16f862d | 2636 | kmem_cache_destroy(kvm_vcpu_cache); |
d2308784 | 2637 | out_free_5: |
59ae6c6b | 2638 | sysdev_unregister(&kvm_sysdev); |
d2308784 | 2639 | out_free_4: |
59ae6c6b | 2640 | sysdev_class_unregister(&kvm_sysdev_class); |
d2308784 | 2641 | out_free_3: |
6aa8b732 | 2642 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 | 2643 | unregister_cpu_notifier(&kvm_cpu_notifier); |
d2308784 | 2644 | out_free_2: |
15c8b6c1 | 2645 | on_each_cpu(hardware_disable, NULL, 1); |
d2308784 | 2646 | out_free_1: |
e9b11c17 | 2647 | kvm_arch_hardware_unsetup(); |
7f59f492 RR |
2648 | out_free_0a: |
2649 | free_cpumask_var(cpus_hardware_enabled); | |
d2308784 ZX |
2650 | out_free_0: |
2651 | __free_page(bad_page); | |
ca45aaae | 2652 | out: |
f8c16bba | 2653 | kvm_arch_exit(); |
cb498ea2 | 2654 | kvm_exit_debug(); |
d2308784 | 2655 | out_fail: |
6aa8b732 AK |
2656 | return r; |
2657 | } | |
cb498ea2 | 2658 | EXPORT_SYMBOL_GPL(kvm_init); |
6aa8b732 | 2659 | |
cb498ea2 | 2660 | void kvm_exit(void) |
6aa8b732 | 2661 | { |
d4c9ff2d | 2662 | kvm_trace_cleanup(); |
6aa8b732 | 2663 | misc_deregister(&kvm_dev); |
c16f862d | 2664 | kmem_cache_destroy(kvm_vcpu_cache); |
59ae6c6b AK |
2665 | sysdev_unregister(&kvm_sysdev); |
2666 | sysdev_class_unregister(&kvm_sysdev_class); | |
6aa8b732 | 2667 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 2668 | unregister_cpu_notifier(&kvm_cpu_notifier); |
15c8b6c1 | 2669 | on_each_cpu(hardware_disable, NULL, 1); |
e9b11c17 | 2670 | kvm_arch_hardware_unsetup(); |
f8c16bba | 2671 | kvm_arch_exit(); |
6aa8b732 | 2672 | kvm_exit_debug(); |
7f59f492 | 2673 | free_cpumask_var(cpus_hardware_enabled); |
cea7bb21 | 2674 | __free_page(bad_page); |
6aa8b732 | 2675 | } |
cb498ea2 | 2676 | EXPORT_SYMBOL_GPL(kvm_exit); |