Commit | Line | Data |
---|---|---|
043405e1 CO |
1 | /* |
2 | * Kernel-based Virtual Machine driver for Linux | |
3 | * | |
4 | * derived from drivers/kvm/kvm_main.c | |
5 | * | |
6 | * Copyright (C) 2006 Qumranet, Inc. | |
4d5c5d0f BAY |
7 | * Copyright (C) 2008 Qumranet, Inc. |
8 | * Copyright IBM Corporation, 2008 | |
9611c187 | 9 | * Copyright 2010 Red Hat, Inc. and/or its affiliates. |
043405e1 CO |
10 | * |
11 | * Authors: | |
12 | * Avi Kivity <avi@qumranet.com> | |
13 | * Yaniv Kamay <yaniv@qumranet.com> | |
4d5c5d0f BAY |
14 | * Amit Shah <amit.shah@qumranet.com> |
15 | * Ben-Ami Yassour <benami@il.ibm.com> | |
043405e1 CO |
16 | * |
17 | * This work is licensed under the terms of the GNU GPL, version 2. See | |
18 | * the COPYING file in the top-level directory. | |
19 | * | |
20 | */ | |
21 | ||
edf88417 | 22 | #include <linux/kvm_host.h> |
313a3dc7 | 23 | #include "irq.h" |
1d737c8a | 24 | #include "mmu.h" |
7837699f | 25 | #include "i8254.h" |
37817f29 | 26 | #include "tss.h" |
5fdbf976 | 27 | #include "kvm_cache_regs.h" |
26eef70c | 28 | #include "x86.h" |
313a3dc7 | 29 | |
18068523 | 30 | #include <linux/clocksource.h> |
4d5c5d0f | 31 | #include <linux/interrupt.h> |
313a3dc7 CO |
32 | #include <linux/kvm.h> |
33 | #include <linux/fs.h> | |
34 | #include <linux/vmalloc.h> | |
5fb76f9b | 35 | #include <linux/module.h> |
0de10343 | 36 | #include <linux/mman.h> |
2bacc55c | 37 | #include <linux/highmem.h> |
19de40a8 | 38 | #include <linux/iommu.h> |
62c476c7 | 39 | #include <linux/intel-iommu.h> |
c8076604 | 40 | #include <linux/cpufreq.h> |
18863bdd | 41 | #include <linux/user-return-notifier.h> |
a983fb23 | 42 | #include <linux/srcu.h> |
5a0e3ad6 | 43 | #include <linux/slab.h> |
ff9d07a0 | 44 | #include <linux/perf_event.h> |
7bee342a | 45 | #include <linux/uaccess.h> |
af585b92 | 46 | #include <linux/hash.h> |
aec51dc4 | 47 | #include <trace/events/kvm.h> |
2ed152af | 48 | |
229456fc MT |
49 | #define CREATE_TRACE_POINTS |
50 | #include "trace.h" | |
043405e1 | 51 | |
24f1e32c | 52 | #include <asm/debugreg.h> |
d825ed0a | 53 | #include <asm/msr.h> |
a5f61300 | 54 | #include <asm/desc.h> |
0bed3b56 | 55 | #include <asm/mtrr.h> |
890ca9ae | 56 | #include <asm/mce.h> |
7cf30855 | 57 | #include <asm/i387.h> |
98918833 | 58 | #include <asm/xcr.h> |
1d5f066e | 59 | #include <asm/pvclock.h> |
217fc9cf | 60 | #include <asm/div64.h> |
043405e1 | 61 | |
313a3dc7 | 62 | #define MAX_IO_MSRS 256 |
a03490ed CO |
63 | #define CR0_RESERVED_BITS \ |
64 | (~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \ | |
65 | | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \ | |
66 | | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG)) | |
67 | #define CR4_RESERVED_BITS \ | |
68 | (~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\ | |
69 | | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE \ | |
70 | | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR \ | |
2acf923e | 71 | | X86_CR4_OSXSAVE \ |
a03490ed CO |
72 | | X86_CR4_OSXMMEXCPT | X86_CR4_VMXE)) |
73 | ||
74 | #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR) | |
890ca9ae HY |
75 | |
76 | #define KVM_MAX_MCE_BANKS 32 | |
5854dbca | 77 | #define KVM_MCE_CAP_SUPPORTED (MCG_CTL_P | MCG_SER_P) |
890ca9ae | 78 | |
50a37eb4 JR |
79 | /* EFER defaults: |
80 | * - enable syscall per default because its emulated by KVM | |
81 | * - enable LME and LMA per default on 64 bit KVM | |
82 | */ | |
83 | #ifdef CONFIG_X86_64 | |
84 | static u64 __read_mostly efer_reserved_bits = 0xfffffffffffffafeULL; | |
85 | #else | |
86 | static u64 __read_mostly efer_reserved_bits = 0xfffffffffffffffeULL; | |
87 | #endif | |
313a3dc7 | 88 | |
ba1389b7 AK |
89 | #define VM_STAT(x) offsetof(struct kvm, stat.x), KVM_STAT_VM |
90 | #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU | |
417bc304 | 91 | |
cb142eb7 | 92 | static void update_cr8_intercept(struct kvm_vcpu *vcpu); |
674eea0f AK |
93 | static int kvm_dev_ioctl_get_supported_cpuid(struct kvm_cpuid2 *cpuid, |
94 | struct kvm_cpuid_entry2 __user *entries); | |
95 | ||
97896d04 | 96 | struct kvm_x86_ops *kvm_x86_ops; |
5fdbf976 | 97 | EXPORT_SYMBOL_GPL(kvm_x86_ops); |
97896d04 | 98 | |
ed85c068 AP |
99 | int ignore_msrs = 0; |
100 | module_param_named(ignore_msrs, ignore_msrs, bool, S_IRUGO | S_IWUSR); | |
101 | ||
18863bdd AK |
102 | #define KVM_NR_SHARED_MSRS 16 |
103 | ||
104 | struct kvm_shared_msrs_global { | |
105 | int nr; | |
2bf78fa7 | 106 | u32 msrs[KVM_NR_SHARED_MSRS]; |
18863bdd AK |
107 | }; |
108 | ||
109 | struct kvm_shared_msrs { | |
110 | struct user_return_notifier urn; | |
111 | bool registered; | |
2bf78fa7 SY |
112 | struct kvm_shared_msr_values { |
113 | u64 host; | |
114 | u64 curr; | |
115 | } values[KVM_NR_SHARED_MSRS]; | |
18863bdd AK |
116 | }; |
117 | ||
118 | static struct kvm_shared_msrs_global __read_mostly shared_msrs_global; | |
119 | static DEFINE_PER_CPU(struct kvm_shared_msrs, shared_msrs); | |
120 | ||
417bc304 | 121 | struct kvm_stats_debugfs_item debugfs_entries[] = { |
ba1389b7 AK |
122 | { "pf_fixed", VCPU_STAT(pf_fixed) }, |
123 | { "pf_guest", VCPU_STAT(pf_guest) }, | |
124 | { "tlb_flush", VCPU_STAT(tlb_flush) }, | |
125 | { "invlpg", VCPU_STAT(invlpg) }, | |
126 | { "exits", VCPU_STAT(exits) }, | |
127 | { "io_exits", VCPU_STAT(io_exits) }, | |
128 | { "mmio_exits", VCPU_STAT(mmio_exits) }, | |
129 | { "signal_exits", VCPU_STAT(signal_exits) }, | |
130 | { "irq_window", VCPU_STAT(irq_window_exits) }, | |
f08864b4 | 131 | { "nmi_window", VCPU_STAT(nmi_window_exits) }, |
ba1389b7 AK |
132 | { "halt_exits", VCPU_STAT(halt_exits) }, |
133 | { "halt_wakeup", VCPU_STAT(halt_wakeup) }, | |
f11c3a8d | 134 | { "hypercalls", VCPU_STAT(hypercalls) }, |
ba1389b7 AK |
135 | { "request_irq", VCPU_STAT(request_irq_exits) }, |
136 | { "irq_exits", VCPU_STAT(irq_exits) }, | |
137 | { "host_state_reload", VCPU_STAT(host_state_reload) }, | |
138 | { "efer_reload", VCPU_STAT(efer_reload) }, | |
139 | { "fpu_reload", VCPU_STAT(fpu_reload) }, | |
140 | { "insn_emulation", VCPU_STAT(insn_emulation) }, | |
141 | { "insn_emulation_fail", VCPU_STAT(insn_emulation_fail) }, | |
fa89a817 | 142 | { "irq_injections", VCPU_STAT(irq_injections) }, |
c4abb7c9 | 143 | { "nmi_injections", VCPU_STAT(nmi_injections) }, |
4cee5764 AK |
144 | { "mmu_shadow_zapped", VM_STAT(mmu_shadow_zapped) }, |
145 | { "mmu_pte_write", VM_STAT(mmu_pte_write) }, | |
146 | { "mmu_pte_updated", VM_STAT(mmu_pte_updated) }, | |
147 | { "mmu_pde_zapped", VM_STAT(mmu_pde_zapped) }, | |
148 | { "mmu_flooded", VM_STAT(mmu_flooded) }, | |
149 | { "mmu_recycled", VM_STAT(mmu_recycled) }, | |
dfc5aa00 | 150 | { "mmu_cache_miss", VM_STAT(mmu_cache_miss) }, |
4731d4c7 | 151 | { "mmu_unsync", VM_STAT(mmu_unsync) }, |
0f74a24c | 152 | { "remote_tlb_flush", VM_STAT(remote_tlb_flush) }, |
05da4558 | 153 | { "largepages", VM_STAT(lpages) }, |
417bc304 HB |
154 | { NULL } |
155 | }; | |
156 | ||
2acf923e DC |
157 | u64 __read_mostly host_xcr0; |
158 | ||
af585b92 GN |
159 | static inline void kvm_async_pf_hash_reset(struct kvm_vcpu *vcpu) |
160 | { | |
161 | int i; | |
162 | for (i = 0; i < roundup_pow_of_two(ASYNC_PF_PER_VCPU); i++) | |
163 | vcpu->arch.apf.gfns[i] = ~0; | |
164 | } | |
165 | ||
18863bdd AK |
166 | static void kvm_on_user_return(struct user_return_notifier *urn) |
167 | { | |
168 | unsigned slot; | |
18863bdd AK |
169 | struct kvm_shared_msrs *locals |
170 | = container_of(urn, struct kvm_shared_msrs, urn); | |
2bf78fa7 | 171 | struct kvm_shared_msr_values *values; |
18863bdd AK |
172 | |
173 | for (slot = 0; slot < shared_msrs_global.nr; ++slot) { | |
2bf78fa7 SY |
174 | values = &locals->values[slot]; |
175 | if (values->host != values->curr) { | |
176 | wrmsrl(shared_msrs_global.msrs[slot], values->host); | |
177 | values->curr = values->host; | |
18863bdd AK |
178 | } |
179 | } | |
180 | locals->registered = false; | |
181 | user_return_notifier_unregister(urn); | |
182 | } | |
183 | ||
2bf78fa7 | 184 | static void shared_msr_update(unsigned slot, u32 msr) |
18863bdd | 185 | { |
2bf78fa7 | 186 | struct kvm_shared_msrs *smsr; |
18863bdd AK |
187 | u64 value; |
188 | ||
2bf78fa7 SY |
189 | smsr = &__get_cpu_var(shared_msrs); |
190 | /* only read, and nobody should modify it at this time, | |
191 | * so don't need lock */ | |
192 | if (slot >= shared_msrs_global.nr) { | |
193 | printk(KERN_ERR "kvm: invalid MSR slot!"); | |
194 | return; | |
195 | } | |
196 | rdmsrl_safe(msr, &value); | |
197 | smsr->values[slot].host = value; | |
198 | smsr->values[slot].curr = value; | |
199 | } | |
200 | ||
201 | void kvm_define_shared_msr(unsigned slot, u32 msr) | |
202 | { | |
18863bdd AK |
203 | if (slot >= shared_msrs_global.nr) |
204 | shared_msrs_global.nr = slot + 1; | |
2bf78fa7 SY |
205 | shared_msrs_global.msrs[slot] = msr; |
206 | /* we need ensured the shared_msr_global have been updated */ | |
207 | smp_wmb(); | |
18863bdd AK |
208 | } |
209 | EXPORT_SYMBOL_GPL(kvm_define_shared_msr); | |
210 | ||
211 | static void kvm_shared_msr_cpu_online(void) | |
212 | { | |
213 | unsigned i; | |
18863bdd AK |
214 | |
215 | for (i = 0; i < shared_msrs_global.nr; ++i) | |
2bf78fa7 | 216 | shared_msr_update(i, shared_msrs_global.msrs[i]); |
18863bdd AK |
217 | } |
218 | ||
d5696725 | 219 | void kvm_set_shared_msr(unsigned slot, u64 value, u64 mask) |
18863bdd AK |
220 | { |
221 | struct kvm_shared_msrs *smsr = &__get_cpu_var(shared_msrs); | |
222 | ||
2bf78fa7 | 223 | if (((value ^ smsr->values[slot].curr) & mask) == 0) |
18863bdd | 224 | return; |
2bf78fa7 SY |
225 | smsr->values[slot].curr = value; |
226 | wrmsrl(shared_msrs_global.msrs[slot], value); | |
18863bdd AK |
227 | if (!smsr->registered) { |
228 | smsr->urn.on_user_return = kvm_on_user_return; | |
229 | user_return_notifier_register(&smsr->urn); | |
230 | smsr->registered = true; | |
231 | } | |
232 | } | |
233 | EXPORT_SYMBOL_GPL(kvm_set_shared_msr); | |
234 | ||
3548bab5 AK |
235 | static void drop_user_return_notifiers(void *ignore) |
236 | { | |
237 | struct kvm_shared_msrs *smsr = &__get_cpu_var(shared_msrs); | |
238 | ||
239 | if (smsr->registered) | |
240 | kvm_on_user_return(&smsr->urn); | |
241 | } | |
242 | ||
6866b83e CO |
243 | u64 kvm_get_apic_base(struct kvm_vcpu *vcpu) |
244 | { | |
245 | if (irqchip_in_kernel(vcpu->kvm)) | |
ad312c7c | 246 | return vcpu->arch.apic_base; |
6866b83e | 247 | else |
ad312c7c | 248 | return vcpu->arch.apic_base; |
6866b83e CO |
249 | } |
250 | EXPORT_SYMBOL_GPL(kvm_get_apic_base); | |
251 | ||
252 | void kvm_set_apic_base(struct kvm_vcpu *vcpu, u64 data) | |
253 | { | |
254 | /* TODO: reserve bits check */ | |
255 | if (irqchip_in_kernel(vcpu->kvm)) | |
256 | kvm_lapic_set_base(vcpu, data); | |
257 | else | |
ad312c7c | 258 | vcpu->arch.apic_base = data; |
6866b83e CO |
259 | } |
260 | EXPORT_SYMBOL_GPL(kvm_set_apic_base); | |
261 | ||
3fd28fce ED |
262 | #define EXCPT_BENIGN 0 |
263 | #define EXCPT_CONTRIBUTORY 1 | |
264 | #define EXCPT_PF 2 | |
265 | ||
266 | static int exception_class(int vector) | |
267 | { | |
268 | switch (vector) { | |
269 | case PF_VECTOR: | |
270 | return EXCPT_PF; | |
271 | case DE_VECTOR: | |
272 | case TS_VECTOR: | |
273 | case NP_VECTOR: | |
274 | case SS_VECTOR: | |
275 | case GP_VECTOR: | |
276 | return EXCPT_CONTRIBUTORY; | |
277 | default: | |
278 | break; | |
279 | } | |
280 | return EXCPT_BENIGN; | |
281 | } | |
282 | ||
283 | static void kvm_multiple_exception(struct kvm_vcpu *vcpu, | |
ce7ddec4 JR |
284 | unsigned nr, bool has_error, u32 error_code, |
285 | bool reinject) | |
3fd28fce ED |
286 | { |
287 | u32 prev_nr; | |
288 | int class1, class2; | |
289 | ||
3842d135 AK |
290 | kvm_make_request(KVM_REQ_EVENT, vcpu); |
291 | ||
3fd28fce ED |
292 | if (!vcpu->arch.exception.pending) { |
293 | queue: | |
294 | vcpu->arch.exception.pending = true; | |
295 | vcpu->arch.exception.has_error_code = has_error; | |
296 | vcpu->arch.exception.nr = nr; | |
297 | vcpu->arch.exception.error_code = error_code; | |
3f0fd292 | 298 | vcpu->arch.exception.reinject = reinject; |
3fd28fce ED |
299 | return; |
300 | } | |
301 | ||
302 | /* to check exception */ | |
303 | prev_nr = vcpu->arch.exception.nr; | |
304 | if (prev_nr == DF_VECTOR) { | |
305 | /* triple fault -> shutdown */ | |
a8eeb04a | 306 | kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu); |
3fd28fce ED |
307 | return; |
308 | } | |
309 | class1 = exception_class(prev_nr); | |
310 | class2 = exception_class(nr); | |
311 | if ((class1 == EXCPT_CONTRIBUTORY && class2 == EXCPT_CONTRIBUTORY) | |
312 | || (class1 == EXCPT_PF && class2 != EXCPT_BENIGN)) { | |
313 | /* generate double fault per SDM Table 5-5 */ | |
314 | vcpu->arch.exception.pending = true; | |
315 | vcpu->arch.exception.has_error_code = true; | |
316 | vcpu->arch.exception.nr = DF_VECTOR; | |
317 | vcpu->arch.exception.error_code = 0; | |
318 | } else | |
319 | /* replace previous exception with a new one in a hope | |
320 | that instruction re-execution will regenerate lost | |
321 | exception */ | |
322 | goto queue; | |
323 | } | |
324 | ||
298101da AK |
325 | void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr) |
326 | { | |
ce7ddec4 | 327 | kvm_multiple_exception(vcpu, nr, false, 0, false); |
298101da AK |
328 | } |
329 | EXPORT_SYMBOL_GPL(kvm_queue_exception); | |
330 | ||
ce7ddec4 JR |
331 | void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr) |
332 | { | |
333 | kvm_multiple_exception(vcpu, nr, false, 0, true); | |
334 | } | |
335 | EXPORT_SYMBOL_GPL(kvm_requeue_exception); | |
336 | ||
6389ee94 | 337 | void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault) |
c3c91fee AK |
338 | { |
339 | ++vcpu->stat.pf_guest; | |
6389ee94 AK |
340 | vcpu->arch.cr2 = fault->address; |
341 | kvm_queue_exception_e(vcpu, PF_VECTOR, fault->error_code); | |
c3c91fee AK |
342 | } |
343 | ||
6389ee94 | 344 | void kvm_propagate_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault) |
d4f8cf66 | 345 | { |
6389ee94 AK |
346 | if (mmu_is_nested(vcpu) && !fault->nested_page_fault) |
347 | vcpu->arch.nested_mmu.inject_page_fault(vcpu, fault); | |
d4f8cf66 | 348 | else |
6389ee94 | 349 | vcpu->arch.mmu.inject_page_fault(vcpu, fault); |
d4f8cf66 JR |
350 | } |
351 | ||
3419ffc8 SY |
352 | void kvm_inject_nmi(struct kvm_vcpu *vcpu) |
353 | { | |
3842d135 | 354 | kvm_make_request(KVM_REQ_EVENT, vcpu); |
3419ffc8 SY |
355 | vcpu->arch.nmi_pending = 1; |
356 | } | |
357 | EXPORT_SYMBOL_GPL(kvm_inject_nmi); | |
358 | ||
298101da AK |
359 | void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code) |
360 | { | |
ce7ddec4 | 361 | kvm_multiple_exception(vcpu, nr, true, error_code, false); |
298101da AK |
362 | } |
363 | EXPORT_SYMBOL_GPL(kvm_queue_exception_e); | |
364 | ||
ce7ddec4 JR |
365 | void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code) |
366 | { | |
367 | kvm_multiple_exception(vcpu, nr, true, error_code, true); | |
368 | } | |
369 | EXPORT_SYMBOL_GPL(kvm_requeue_exception_e); | |
370 | ||
0a79b009 AK |
371 | /* |
372 | * Checks if cpl <= required_cpl; if true, return true. Otherwise queue | |
373 | * a #GP and return false. | |
374 | */ | |
375 | bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl) | |
298101da | 376 | { |
0a79b009 AK |
377 | if (kvm_x86_ops->get_cpl(vcpu) <= required_cpl) |
378 | return true; | |
379 | kvm_queue_exception_e(vcpu, GP_VECTOR, 0); | |
380 | return false; | |
298101da | 381 | } |
0a79b009 | 382 | EXPORT_SYMBOL_GPL(kvm_require_cpl); |
298101da | 383 | |
ec92fe44 JR |
384 | /* |
385 | * This function will be used to read from the physical memory of the currently | |
386 | * running guest. The difference to kvm_read_guest_page is that this function | |
387 | * can read from guest physical or from the guest's guest physical memory. | |
388 | */ | |
389 | int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, | |
390 | gfn_t ngfn, void *data, int offset, int len, | |
391 | u32 access) | |
392 | { | |
393 | gfn_t real_gfn; | |
394 | gpa_t ngpa; | |
395 | ||
396 | ngpa = gfn_to_gpa(ngfn); | |
397 | real_gfn = mmu->translate_gpa(vcpu, ngpa, access); | |
398 | if (real_gfn == UNMAPPED_GVA) | |
399 | return -EFAULT; | |
400 | ||
401 | real_gfn = gpa_to_gfn(real_gfn); | |
402 | ||
403 | return kvm_read_guest_page(vcpu->kvm, real_gfn, data, offset, len); | |
404 | } | |
405 | EXPORT_SYMBOL_GPL(kvm_read_guest_page_mmu); | |
406 | ||
3d06b8bf JR |
407 | int kvm_read_nested_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, |
408 | void *data, int offset, int len, u32 access) | |
409 | { | |
410 | return kvm_read_guest_page_mmu(vcpu, vcpu->arch.walk_mmu, gfn, | |
411 | data, offset, len, access); | |
412 | } | |
413 | ||
a03490ed CO |
414 | /* |
415 | * Load the pae pdptrs. Return true is they are all valid. | |
416 | */ | |
ff03a073 | 417 | int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3) |
a03490ed CO |
418 | { |
419 | gfn_t pdpt_gfn = cr3 >> PAGE_SHIFT; | |
420 | unsigned offset = ((cr3 & (PAGE_SIZE-1)) >> 5) << 2; | |
421 | int i; | |
422 | int ret; | |
ff03a073 | 423 | u64 pdpte[ARRAY_SIZE(mmu->pdptrs)]; |
a03490ed | 424 | |
ff03a073 JR |
425 | ret = kvm_read_guest_page_mmu(vcpu, mmu, pdpt_gfn, pdpte, |
426 | offset * sizeof(u64), sizeof(pdpte), | |
427 | PFERR_USER_MASK|PFERR_WRITE_MASK); | |
a03490ed CO |
428 | if (ret < 0) { |
429 | ret = 0; | |
430 | goto out; | |
431 | } | |
432 | for (i = 0; i < ARRAY_SIZE(pdpte); ++i) { | |
43a3795a | 433 | if (is_present_gpte(pdpte[i]) && |
20c466b5 | 434 | (pdpte[i] & vcpu->arch.mmu.rsvd_bits_mask[0][2])) { |
a03490ed CO |
435 | ret = 0; |
436 | goto out; | |
437 | } | |
438 | } | |
439 | ret = 1; | |
440 | ||
ff03a073 | 441 | memcpy(mmu->pdptrs, pdpte, sizeof(mmu->pdptrs)); |
6de4f3ad AK |
442 | __set_bit(VCPU_EXREG_PDPTR, |
443 | (unsigned long *)&vcpu->arch.regs_avail); | |
444 | __set_bit(VCPU_EXREG_PDPTR, | |
445 | (unsigned long *)&vcpu->arch.regs_dirty); | |
a03490ed | 446 | out: |
a03490ed CO |
447 | |
448 | return ret; | |
449 | } | |
cc4b6871 | 450 | EXPORT_SYMBOL_GPL(load_pdptrs); |
a03490ed | 451 | |
d835dfec AK |
452 | static bool pdptrs_changed(struct kvm_vcpu *vcpu) |
453 | { | |
ff03a073 | 454 | u64 pdpte[ARRAY_SIZE(vcpu->arch.walk_mmu->pdptrs)]; |
d835dfec | 455 | bool changed = true; |
3d06b8bf JR |
456 | int offset; |
457 | gfn_t gfn; | |
d835dfec AK |
458 | int r; |
459 | ||
460 | if (is_long_mode(vcpu) || !is_pae(vcpu)) | |
461 | return false; | |
462 | ||
6de4f3ad AK |
463 | if (!test_bit(VCPU_EXREG_PDPTR, |
464 | (unsigned long *)&vcpu->arch.regs_avail)) | |
465 | return true; | |
466 | ||
3d06b8bf JR |
467 | gfn = (vcpu->arch.cr3 & ~31u) >> PAGE_SHIFT; |
468 | offset = (vcpu->arch.cr3 & ~31u) & (PAGE_SIZE - 1); | |
469 | r = kvm_read_nested_guest_page(vcpu, gfn, pdpte, offset, sizeof(pdpte), | |
470 | PFERR_USER_MASK | PFERR_WRITE_MASK); | |
d835dfec AK |
471 | if (r < 0) |
472 | goto out; | |
ff03a073 | 473 | changed = memcmp(pdpte, vcpu->arch.walk_mmu->pdptrs, sizeof(pdpte)) != 0; |
d835dfec | 474 | out: |
d835dfec AK |
475 | |
476 | return changed; | |
477 | } | |
478 | ||
49a9b07e | 479 | int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0) |
a03490ed | 480 | { |
aad82703 SY |
481 | unsigned long old_cr0 = kvm_read_cr0(vcpu); |
482 | unsigned long update_bits = X86_CR0_PG | X86_CR0_WP | | |
483 | X86_CR0_CD | X86_CR0_NW; | |
484 | ||
f9a48e6a AK |
485 | cr0 |= X86_CR0_ET; |
486 | ||
ab344828 | 487 | #ifdef CONFIG_X86_64 |
0f12244f GN |
488 | if (cr0 & 0xffffffff00000000UL) |
489 | return 1; | |
ab344828 GN |
490 | #endif |
491 | ||
492 | cr0 &= ~CR0_RESERVED_BITS; | |
a03490ed | 493 | |
0f12244f GN |
494 | if ((cr0 & X86_CR0_NW) && !(cr0 & X86_CR0_CD)) |
495 | return 1; | |
a03490ed | 496 | |
0f12244f GN |
497 | if ((cr0 & X86_CR0_PG) && !(cr0 & X86_CR0_PE)) |
498 | return 1; | |
a03490ed CO |
499 | |
500 | if (!is_paging(vcpu) && (cr0 & X86_CR0_PG)) { | |
501 | #ifdef CONFIG_X86_64 | |
f6801dff | 502 | if ((vcpu->arch.efer & EFER_LME)) { |
a03490ed CO |
503 | int cs_db, cs_l; |
504 | ||
0f12244f GN |
505 | if (!is_pae(vcpu)) |
506 | return 1; | |
a03490ed | 507 | kvm_x86_ops->get_cs_db_l_bits(vcpu, &cs_db, &cs_l); |
0f12244f GN |
508 | if (cs_l) |
509 | return 1; | |
a03490ed CO |
510 | } else |
511 | #endif | |
ff03a073 JR |
512 | if (is_pae(vcpu) && !load_pdptrs(vcpu, vcpu->arch.walk_mmu, |
513 | vcpu->arch.cr3)) | |
0f12244f | 514 | return 1; |
a03490ed CO |
515 | } |
516 | ||
517 | kvm_x86_ops->set_cr0(vcpu, cr0); | |
a03490ed | 518 | |
e5f3f027 XG |
519 | if ((cr0 ^ old_cr0) & X86_CR0_PG) |
520 | kvm_clear_async_pf_completion_queue(vcpu); | |
521 | ||
aad82703 SY |
522 | if ((cr0 ^ old_cr0) & update_bits) |
523 | kvm_mmu_reset_context(vcpu); | |
0f12244f GN |
524 | return 0; |
525 | } | |
2d3ad1f4 | 526 | EXPORT_SYMBOL_GPL(kvm_set_cr0); |
a03490ed | 527 | |
2d3ad1f4 | 528 | void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw) |
a03490ed | 529 | { |
49a9b07e | 530 | (void)kvm_set_cr0(vcpu, kvm_read_cr0_bits(vcpu, ~0x0eul) | (msw & 0x0f)); |
a03490ed | 531 | } |
2d3ad1f4 | 532 | EXPORT_SYMBOL_GPL(kvm_lmsw); |
a03490ed | 533 | |
2acf923e DC |
534 | int __kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr) |
535 | { | |
536 | u64 xcr0; | |
537 | ||
538 | /* Only support XCR_XFEATURE_ENABLED_MASK(xcr0) now */ | |
539 | if (index != XCR_XFEATURE_ENABLED_MASK) | |
540 | return 1; | |
541 | xcr0 = xcr; | |
542 | if (kvm_x86_ops->get_cpl(vcpu) != 0) | |
543 | return 1; | |
544 | if (!(xcr0 & XSTATE_FP)) | |
545 | return 1; | |
546 | if ((xcr0 & XSTATE_YMM) && !(xcr0 & XSTATE_SSE)) | |
547 | return 1; | |
548 | if (xcr0 & ~host_xcr0) | |
549 | return 1; | |
550 | vcpu->arch.xcr0 = xcr0; | |
551 | vcpu->guest_xcr0_loaded = 0; | |
552 | return 0; | |
553 | } | |
554 | ||
555 | int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr) | |
556 | { | |
557 | if (__kvm_set_xcr(vcpu, index, xcr)) { | |
558 | kvm_inject_gp(vcpu, 0); | |
559 | return 1; | |
560 | } | |
561 | return 0; | |
562 | } | |
563 | EXPORT_SYMBOL_GPL(kvm_set_xcr); | |
564 | ||
565 | static bool guest_cpuid_has_xsave(struct kvm_vcpu *vcpu) | |
566 | { | |
567 | struct kvm_cpuid_entry2 *best; | |
568 | ||
569 | best = kvm_find_cpuid_entry(vcpu, 1, 0); | |
570 | return best && (best->ecx & bit(X86_FEATURE_XSAVE)); | |
571 | } | |
572 | ||
573 | static void update_cpuid(struct kvm_vcpu *vcpu) | |
574 | { | |
575 | struct kvm_cpuid_entry2 *best; | |
576 | ||
577 | best = kvm_find_cpuid_entry(vcpu, 1, 0); | |
578 | if (!best) | |
579 | return; | |
580 | ||
581 | /* Update OSXSAVE bit */ | |
582 | if (cpu_has_xsave && best->function == 0x1) { | |
583 | best->ecx &= ~(bit(X86_FEATURE_OSXSAVE)); | |
584 | if (kvm_read_cr4_bits(vcpu, X86_CR4_OSXSAVE)) | |
585 | best->ecx |= bit(X86_FEATURE_OSXSAVE); | |
586 | } | |
587 | } | |
588 | ||
a83b29c6 | 589 | int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4) |
a03490ed | 590 | { |
fc78f519 | 591 | unsigned long old_cr4 = kvm_read_cr4(vcpu); |
a2edf57f AK |
592 | unsigned long pdptr_bits = X86_CR4_PGE | X86_CR4_PSE | X86_CR4_PAE; |
593 | ||
0f12244f GN |
594 | if (cr4 & CR4_RESERVED_BITS) |
595 | return 1; | |
a03490ed | 596 | |
2acf923e DC |
597 | if (!guest_cpuid_has_xsave(vcpu) && (cr4 & X86_CR4_OSXSAVE)) |
598 | return 1; | |
599 | ||
a03490ed | 600 | if (is_long_mode(vcpu)) { |
0f12244f GN |
601 | if (!(cr4 & X86_CR4_PAE)) |
602 | return 1; | |
a2edf57f AK |
603 | } else if (is_paging(vcpu) && (cr4 & X86_CR4_PAE) |
604 | && ((cr4 ^ old_cr4) & pdptr_bits) | |
ff03a073 | 605 | && !load_pdptrs(vcpu, vcpu->arch.walk_mmu, vcpu->arch.cr3)) |
0f12244f GN |
606 | return 1; |
607 | ||
608 | if (cr4 & X86_CR4_VMXE) | |
609 | return 1; | |
a03490ed | 610 | |
a03490ed | 611 | kvm_x86_ops->set_cr4(vcpu, cr4); |
62ad0755 | 612 | |
aad82703 SY |
613 | if ((cr4 ^ old_cr4) & pdptr_bits) |
614 | kvm_mmu_reset_context(vcpu); | |
0f12244f | 615 | |
2acf923e DC |
616 | if ((cr4 ^ old_cr4) & X86_CR4_OSXSAVE) |
617 | update_cpuid(vcpu); | |
618 | ||
0f12244f GN |
619 | return 0; |
620 | } | |
2d3ad1f4 | 621 | EXPORT_SYMBOL_GPL(kvm_set_cr4); |
a03490ed | 622 | |
2390218b | 623 | int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3) |
a03490ed | 624 | { |
ad312c7c | 625 | if (cr3 == vcpu->arch.cr3 && !pdptrs_changed(vcpu)) { |
0ba73cda | 626 | kvm_mmu_sync_roots(vcpu); |
d835dfec | 627 | kvm_mmu_flush_tlb(vcpu); |
0f12244f | 628 | return 0; |
d835dfec AK |
629 | } |
630 | ||
a03490ed | 631 | if (is_long_mode(vcpu)) { |
0f12244f GN |
632 | if (cr3 & CR3_L_MODE_RESERVED_BITS) |
633 | return 1; | |
a03490ed CO |
634 | } else { |
635 | if (is_pae(vcpu)) { | |
0f12244f GN |
636 | if (cr3 & CR3_PAE_RESERVED_BITS) |
637 | return 1; | |
ff03a073 JR |
638 | if (is_paging(vcpu) && |
639 | !load_pdptrs(vcpu, vcpu->arch.walk_mmu, cr3)) | |
0f12244f | 640 | return 1; |
a03490ed CO |
641 | } |
642 | /* | |
643 | * We don't check reserved bits in nonpae mode, because | |
644 | * this isn't enforced, and VMware depends on this. | |
645 | */ | |
646 | } | |
647 | ||
a03490ed CO |
648 | /* |
649 | * Does the new cr3 value map to physical memory? (Note, we | |
650 | * catch an invalid cr3 even in real-mode, because it would | |
651 | * cause trouble later on when we turn on paging anyway.) | |
652 | * | |
653 | * A real CPU would silently accept an invalid cr3 and would | |
654 | * attempt to use it - with largely undefined (and often hard | |
655 | * to debug) behavior on the guest side. | |
656 | */ | |
657 | if (unlikely(!gfn_to_memslot(vcpu->kvm, cr3 >> PAGE_SHIFT))) | |
0f12244f GN |
658 | return 1; |
659 | vcpu->arch.cr3 = cr3; | |
660 | vcpu->arch.mmu.new_cr3(vcpu); | |
661 | return 0; | |
662 | } | |
2d3ad1f4 | 663 | EXPORT_SYMBOL_GPL(kvm_set_cr3); |
a03490ed | 664 | |
0f12244f | 665 | int __kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8) |
a03490ed | 666 | { |
0f12244f GN |
667 | if (cr8 & CR8_RESERVED_BITS) |
668 | return 1; | |
a03490ed CO |
669 | if (irqchip_in_kernel(vcpu->kvm)) |
670 | kvm_lapic_set_tpr(vcpu, cr8); | |
671 | else | |
ad312c7c | 672 | vcpu->arch.cr8 = cr8; |
0f12244f GN |
673 | return 0; |
674 | } | |
675 | ||
676 | void kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8) | |
677 | { | |
678 | if (__kvm_set_cr8(vcpu, cr8)) | |
679 | kvm_inject_gp(vcpu, 0); | |
a03490ed | 680 | } |
2d3ad1f4 | 681 | EXPORT_SYMBOL_GPL(kvm_set_cr8); |
a03490ed | 682 | |
2d3ad1f4 | 683 | unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu) |
a03490ed CO |
684 | { |
685 | if (irqchip_in_kernel(vcpu->kvm)) | |
686 | return kvm_lapic_get_cr8(vcpu); | |
687 | else | |
ad312c7c | 688 | return vcpu->arch.cr8; |
a03490ed | 689 | } |
2d3ad1f4 | 690 | EXPORT_SYMBOL_GPL(kvm_get_cr8); |
a03490ed | 691 | |
338dbc97 | 692 | static int __kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val) |
020df079 GN |
693 | { |
694 | switch (dr) { | |
695 | case 0 ... 3: | |
696 | vcpu->arch.db[dr] = val; | |
697 | if (!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP)) | |
698 | vcpu->arch.eff_db[dr] = val; | |
699 | break; | |
700 | case 4: | |
338dbc97 GN |
701 | if (kvm_read_cr4_bits(vcpu, X86_CR4_DE)) |
702 | return 1; /* #UD */ | |
020df079 GN |
703 | /* fall through */ |
704 | case 6: | |
338dbc97 GN |
705 | if (val & 0xffffffff00000000ULL) |
706 | return -1; /* #GP */ | |
020df079 GN |
707 | vcpu->arch.dr6 = (val & DR6_VOLATILE) | DR6_FIXED_1; |
708 | break; | |
709 | case 5: | |
338dbc97 GN |
710 | if (kvm_read_cr4_bits(vcpu, X86_CR4_DE)) |
711 | return 1; /* #UD */ | |
020df079 GN |
712 | /* fall through */ |
713 | default: /* 7 */ | |
338dbc97 GN |
714 | if (val & 0xffffffff00000000ULL) |
715 | return -1; /* #GP */ | |
020df079 GN |
716 | vcpu->arch.dr7 = (val & DR7_VOLATILE) | DR7_FIXED_1; |
717 | if (!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP)) { | |
718 | kvm_x86_ops->set_dr7(vcpu, vcpu->arch.dr7); | |
719 | vcpu->arch.switch_db_regs = (val & DR7_BP_EN_MASK); | |
720 | } | |
721 | break; | |
722 | } | |
723 | ||
724 | return 0; | |
725 | } | |
338dbc97 GN |
726 | |
727 | int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val) | |
728 | { | |
729 | int res; | |
730 | ||
731 | res = __kvm_set_dr(vcpu, dr, val); | |
732 | if (res > 0) | |
733 | kvm_queue_exception(vcpu, UD_VECTOR); | |
734 | else if (res < 0) | |
735 | kvm_inject_gp(vcpu, 0); | |
736 | ||
737 | return res; | |
738 | } | |
020df079 GN |
739 | EXPORT_SYMBOL_GPL(kvm_set_dr); |
740 | ||
338dbc97 | 741 | static int _kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val) |
020df079 GN |
742 | { |
743 | switch (dr) { | |
744 | case 0 ... 3: | |
745 | *val = vcpu->arch.db[dr]; | |
746 | break; | |
747 | case 4: | |
338dbc97 | 748 | if (kvm_read_cr4_bits(vcpu, X86_CR4_DE)) |
020df079 | 749 | return 1; |
020df079 GN |
750 | /* fall through */ |
751 | case 6: | |
752 | *val = vcpu->arch.dr6; | |
753 | break; | |
754 | case 5: | |
338dbc97 | 755 | if (kvm_read_cr4_bits(vcpu, X86_CR4_DE)) |
020df079 | 756 | return 1; |
020df079 GN |
757 | /* fall through */ |
758 | default: /* 7 */ | |
759 | *val = vcpu->arch.dr7; | |
760 | break; | |
761 | } | |
762 | ||
763 | return 0; | |
764 | } | |
338dbc97 GN |
765 | |
766 | int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val) | |
767 | { | |
768 | if (_kvm_get_dr(vcpu, dr, val)) { | |
769 | kvm_queue_exception(vcpu, UD_VECTOR); | |
770 | return 1; | |
771 | } | |
772 | return 0; | |
773 | } | |
020df079 GN |
774 | EXPORT_SYMBOL_GPL(kvm_get_dr); |
775 | ||
043405e1 CO |
776 | /* |
777 | * List of msr numbers which we expose to userspace through KVM_GET_MSRS | |
778 | * and KVM_SET_MSRS, and KVM_GET_MSR_INDEX_LIST. | |
779 | * | |
780 | * This list is modified at module load time to reflect the | |
e3267cbb GC |
781 | * capabilities of the host cpu. This capabilities test skips MSRs that are |
782 | * kvm-specific. Those are put in the beginning of the list. | |
043405e1 | 783 | */ |
e3267cbb | 784 | |
344d9588 | 785 | #define KVM_SAVE_MSRS_BEGIN 8 |
043405e1 | 786 | static u32 msrs_to_save[] = { |
e3267cbb | 787 | MSR_KVM_SYSTEM_TIME, MSR_KVM_WALL_CLOCK, |
11c6bffa | 788 | MSR_KVM_SYSTEM_TIME_NEW, MSR_KVM_WALL_CLOCK_NEW, |
55cd8e5a | 789 | HV_X64_MSR_GUEST_OS_ID, HV_X64_MSR_HYPERCALL, |
344d9588 | 790 | HV_X64_MSR_APIC_ASSIST_PAGE, MSR_KVM_ASYNC_PF_EN, |
043405e1 | 791 | MSR_IA32_SYSENTER_CS, MSR_IA32_SYSENTER_ESP, MSR_IA32_SYSENTER_EIP, |
8c06585d | 792 | MSR_STAR, |
043405e1 CO |
793 | #ifdef CONFIG_X86_64 |
794 | MSR_CSTAR, MSR_KERNEL_GS_BASE, MSR_SYSCALL_MASK, MSR_LSTAR, | |
795 | #endif | |
e90aa41e | 796 | MSR_IA32_TSC, MSR_IA32_CR_PAT, MSR_VM_HSAVE_PA |
043405e1 CO |
797 | }; |
798 | ||
799 | static unsigned num_msrs_to_save; | |
800 | ||
801 | static u32 emulated_msrs[] = { | |
802 | MSR_IA32_MISC_ENABLE, | |
908e75f3 AK |
803 | MSR_IA32_MCG_STATUS, |
804 | MSR_IA32_MCG_CTL, | |
043405e1 CO |
805 | }; |
806 | ||
b69e8cae | 807 | static int set_efer(struct kvm_vcpu *vcpu, u64 efer) |
15c4a640 | 808 | { |
aad82703 SY |
809 | u64 old_efer = vcpu->arch.efer; |
810 | ||
b69e8cae RJ |
811 | if (efer & efer_reserved_bits) |
812 | return 1; | |
15c4a640 CO |
813 | |
814 | if (is_paging(vcpu) | |
b69e8cae RJ |
815 | && (vcpu->arch.efer & EFER_LME) != (efer & EFER_LME)) |
816 | return 1; | |
15c4a640 | 817 | |
1b2fd70c AG |
818 | if (efer & EFER_FFXSR) { |
819 | struct kvm_cpuid_entry2 *feat; | |
820 | ||
821 | feat = kvm_find_cpuid_entry(vcpu, 0x80000001, 0); | |
b69e8cae RJ |
822 | if (!feat || !(feat->edx & bit(X86_FEATURE_FXSR_OPT))) |
823 | return 1; | |
1b2fd70c AG |
824 | } |
825 | ||
d8017474 AG |
826 | if (efer & EFER_SVME) { |
827 | struct kvm_cpuid_entry2 *feat; | |
828 | ||
829 | feat = kvm_find_cpuid_entry(vcpu, 0x80000001, 0); | |
b69e8cae RJ |
830 | if (!feat || !(feat->ecx & bit(X86_FEATURE_SVM))) |
831 | return 1; | |
d8017474 AG |
832 | } |
833 | ||
15c4a640 | 834 | efer &= ~EFER_LMA; |
f6801dff | 835 | efer |= vcpu->arch.efer & EFER_LMA; |
15c4a640 | 836 | |
a3d204e2 SY |
837 | kvm_x86_ops->set_efer(vcpu, efer); |
838 | ||
9645bb56 | 839 | vcpu->arch.mmu.base_role.nxe = (efer & EFER_NX) && !tdp_enabled; |
b69e8cae | 840 | |
aad82703 SY |
841 | /* Update reserved bits */ |
842 | if ((efer ^ old_efer) & EFER_NX) | |
843 | kvm_mmu_reset_context(vcpu); | |
844 | ||
b69e8cae | 845 | return 0; |
15c4a640 CO |
846 | } |
847 | ||
f2b4b7dd JR |
848 | void kvm_enable_efer_bits(u64 mask) |
849 | { | |
850 | efer_reserved_bits &= ~mask; | |
851 | } | |
852 | EXPORT_SYMBOL_GPL(kvm_enable_efer_bits); | |
853 | ||
854 | ||
15c4a640 CO |
855 | /* |
856 | * Writes msr value into into the appropriate "register". | |
857 | * Returns 0 on success, non-0 otherwise. | |
858 | * Assumes vcpu_load() was already called. | |
859 | */ | |
860 | int kvm_set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data) | |
861 | { | |
862 | return kvm_x86_ops->set_msr(vcpu, msr_index, data); | |
863 | } | |
864 | ||
313a3dc7 CO |
865 | /* |
866 | * Adapt set_msr() to msr_io()'s calling convention | |
867 | */ | |
868 | static int do_set_msr(struct kvm_vcpu *vcpu, unsigned index, u64 *data) | |
869 | { | |
870 | return kvm_set_msr(vcpu, index, *data); | |
871 | } | |
872 | ||
18068523 GOC |
873 | static void kvm_write_wall_clock(struct kvm *kvm, gpa_t wall_clock) |
874 | { | |
9ed3c444 AK |
875 | int version; |
876 | int r; | |
50d0a0f9 | 877 | struct pvclock_wall_clock wc; |
923de3cf | 878 | struct timespec boot; |
18068523 GOC |
879 | |
880 | if (!wall_clock) | |
881 | return; | |
882 | ||
9ed3c444 AK |
883 | r = kvm_read_guest(kvm, wall_clock, &version, sizeof(version)); |
884 | if (r) | |
885 | return; | |
886 | ||
887 | if (version & 1) | |
888 | ++version; /* first time write, random junk */ | |
889 | ||
890 | ++version; | |
18068523 | 891 | |
18068523 GOC |
892 | kvm_write_guest(kvm, wall_clock, &version, sizeof(version)); |
893 | ||
50d0a0f9 GH |
894 | /* |
895 | * The guest calculates current wall clock time by adding | |
34c238a1 | 896 | * system time (updated by kvm_guest_time_update below) to the |
50d0a0f9 GH |
897 | * wall clock specified here. guest system time equals host |
898 | * system time for us, thus we must fill in host boot time here. | |
899 | */ | |
923de3cf | 900 | getboottime(&boot); |
50d0a0f9 GH |
901 | |
902 | wc.sec = boot.tv_sec; | |
903 | wc.nsec = boot.tv_nsec; | |
904 | wc.version = version; | |
18068523 GOC |
905 | |
906 | kvm_write_guest(kvm, wall_clock, &wc, sizeof(wc)); | |
907 | ||
908 | version++; | |
909 | kvm_write_guest(kvm, wall_clock, &version, sizeof(version)); | |
18068523 GOC |
910 | } |
911 | ||
50d0a0f9 GH |
912 | static uint32_t div_frac(uint32_t dividend, uint32_t divisor) |
913 | { | |
914 | uint32_t quotient, remainder; | |
915 | ||
916 | /* Don't try to replace with do_div(), this one calculates | |
917 | * "(dividend << 32) / divisor" */ | |
918 | __asm__ ( "divl %4" | |
919 | : "=a" (quotient), "=d" (remainder) | |
920 | : "0" (0), "1" (dividend), "r" (divisor) ); | |
921 | return quotient; | |
922 | } | |
923 | ||
5f4e3f88 ZA |
924 | static void kvm_get_time_scale(uint32_t scaled_khz, uint32_t base_khz, |
925 | s8 *pshift, u32 *pmultiplier) | |
50d0a0f9 | 926 | { |
5f4e3f88 | 927 | uint64_t scaled64; |
50d0a0f9 GH |
928 | int32_t shift = 0; |
929 | uint64_t tps64; | |
930 | uint32_t tps32; | |
931 | ||
5f4e3f88 ZA |
932 | tps64 = base_khz * 1000LL; |
933 | scaled64 = scaled_khz * 1000LL; | |
50933623 | 934 | while (tps64 > scaled64*2 || tps64 & 0xffffffff00000000ULL) { |
50d0a0f9 GH |
935 | tps64 >>= 1; |
936 | shift--; | |
937 | } | |
938 | ||
939 | tps32 = (uint32_t)tps64; | |
50933623 JK |
940 | while (tps32 <= scaled64 || scaled64 & 0xffffffff00000000ULL) { |
941 | if (scaled64 & 0xffffffff00000000ULL || tps32 & 0x80000000) | |
5f4e3f88 ZA |
942 | scaled64 >>= 1; |
943 | else | |
944 | tps32 <<= 1; | |
50d0a0f9 GH |
945 | shift++; |
946 | } | |
947 | ||
5f4e3f88 ZA |
948 | *pshift = shift; |
949 | *pmultiplier = div_frac(scaled64, tps32); | |
50d0a0f9 | 950 | |
5f4e3f88 ZA |
951 | pr_debug("%s: base_khz %u => %u, shift %d, mul %u\n", |
952 | __func__, base_khz, scaled_khz, shift, *pmultiplier); | |
50d0a0f9 GH |
953 | } |
954 | ||
759379dd ZA |
955 | static inline u64 get_kernel_ns(void) |
956 | { | |
957 | struct timespec ts; | |
958 | ||
959 | WARN_ON(preemptible()); | |
960 | ktime_get_ts(&ts); | |
961 | monotonic_to_bootbased(&ts); | |
962 | return timespec_to_ns(&ts); | |
50d0a0f9 GH |
963 | } |
964 | ||
c8076604 | 965 | static DEFINE_PER_CPU(unsigned long, cpu_tsc_khz); |
c285545f | 966 | unsigned long max_tsc_khz; |
c8076604 | 967 | |
8cfdc000 ZA |
968 | static inline int kvm_tsc_changes_freq(void) |
969 | { | |
970 | int cpu = get_cpu(); | |
971 | int ret = !boot_cpu_has(X86_FEATURE_CONSTANT_TSC) && | |
972 | cpufreq_quick_get(cpu) != 0; | |
973 | put_cpu(); | |
974 | return ret; | |
975 | } | |
976 | ||
759379dd ZA |
977 | static inline u64 nsec_to_cycles(u64 nsec) |
978 | { | |
217fc9cf AK |
979 | u64 ret; |
980 | ||
759379dd ZA |
981 | WARN_ON(preemptible()); |
982 | if (kvm_tsc_changes_freq()) | |
983 | printk_once(KERN_WARNING | |
984 | "kvm: unreliable cycle conversion on adjustable rate TSC\n"); | |
217fc9cf AK |
985 | ret = nsec * __get_cpu_var(cpu_tsc_khz); |
986 | do_div(ret, USEC_PER_SEC); | |
987 | return ret; | |
759379dd ZA |
988 | } |
989 | ||
c285545f ZA |
990 | static void kvm_arch_set_tsc_khz(struct kvm *kvm, u32 this_tsc_khz) |
991 | { | |
992 | /* Compute a scale to convert nanoseconds in TSC cycles */ | |
993 | kvm_get_time_scale(this_tsc_khz, NSEC_PER_SEC / 1000, | |
994 | &kvm->arch.virtual_tsc_shift, | |
995 | &kvm->arch.virtual_tsc_mult); | |
996 | kvm->arch.virtual_tsc_khz = this_tsc_khz; | |
997 | } | |
998 | ||
999 | static u64 compute_guest_tsc(struct kvm_vcpu *vcpu, s64 kernel_ns) | |
1000 | { | |
1001 | u64 tsc = pvclock_scale_delta(kernel_ns-vcpu->arch.last_tsc_nsec, | |
1002 | vcpu->kvm->arch.virtual_tsc_mult, | |
1003 | vcpu->kvm->arch.virtual_tsc_shift); | |
1004 | tsc += vcpu->arch.last_tsc_write; | |
1005 | return tsc; | |
1006 | } | |
1007 | ||
99e3e30a ZA |
1008 | void kvm_write_tsc(struct kvm_vcpu *vcpu, u64 data) |
1009 | { | |
1010 | struct kvm *kvm = vcpu->kvm; | |
f38e098f | 1011 | u64 offset, ns, elapsed; |
99e3e30a | 1012 | unsigned long flags; |
46543ba4 | 1013 | s64 sdiff; |
99e3e30a ZA |
1014 | |
1015 | spin_lock_irqsave(&kvm->arch.tsc_write_lock, flags); | |
1016 | offset = data - native_read_tsc(); | |
759379dd | 1017 | ns = get_kernel_ns(); |
f38e098f | 1018 | elapsed = ns - kvm->arch.last_tsc_nsec; |
46543ba4 ZA |
1019 | sdiff = data - kvm->arch.last_tsc_write; |
1020 | if (sdiff < 0) | |
1021 | sdiff = -sdiff; | |
f38e098f ZA |
1022 | |
1023 | /* | |
46543ba4 | 1024 | * Special case: close write to TSC within 5 seconds of |
f38e098f | 1025 | * another CPU is interpreted as an attempt to synchronize |
46543ba4 ZA |
1026 | * The 5 seconds is to accomodate host load / swapping as |
1027 | * well as any reset of TSC during the boot process. | |
f38e098f ZA |
1028 | * |
1029 | * In that case, for a reliable TSC, we can match TSC offsets, | |
46543ba4 | 1030 | * or make a best guest using elapsed value. |
f38e098f | 1031 | */ |
46543ba4 ZA |
1032 | if (sdiff < nsec_to_cycles(5ULL * NSEC_PER_SEC) && |
1033 | elapsed < 5ULL * NSEC_PER_SEC) { | |
f38e098f ZA |
1034 | if (!check_tsc_unstable()) { |
1035 | offset = kvm->arch.last_tsc_offset; | |
1036 | pr_debug("kvm: matched tsc offset for %llu\n", data); | |
1037 | } else { | |
759379dd ZA |
1038 | u64 delta = nsec_to_cycles(elapsed); |
1039 | offset += delta; | |
1040 | pr_debug("kvm: adjusted tsc offset by %llu\n", delta); | |
f38e098f ZA |
1041 | } |
1042 | ns = kvm->arch.last_tsc_nsec; | |
1043 | } | |
1044 | kvm->arch.last_tsc_nsec = ns; | |
1045 | kvm->arch.last_tsc_write = data; | |
1046 | kvm->arch.last_tsc_offset = offset; | |
99e3e30a ZA |
1047 | kvm_x86_ops->write_tsc_offset(vcpu, offset); |
1048 | spin_unlock_irqrestore(&kvm->arch.tsc_write_lock, flags); | |
1049 | ||
1050 | /* Reset of TSC must disable overshoot protection below */ | |
1051 | vcpu->arch.hv_clock.tsc_timestamp = 0; | |
c285545f ZA |
1052 | vcpu->arch.last_tsc_write = data; |
1053 | vcpu->arch.last_tsc_nsec = ns; | |
99e3e30a ZA |
1054 | } |
1055 | EXPORT_SYMBOL_GPL(kvm_write_tsc); | |
1056 | ||
34c238a1 | 1057 | static int kvm_guest_time_update(struct kvm_vcpu *v) |
18068523 | 1058 | { |
18068523 GOC |
1059 | unsigned long flags; |
1060 | struct kvm_vcpu_arch *vcpu = &v->arch; | |
1061 | void *shared_kaddr; | |
463656c0 | 1062 | unsigned long this_tsc_khz; |
1d5f066e ZA |
1063 | s64 kernel_ns, max_kernel_ns; |
1064 | u64 tsc_timestamp; | |
18068523 | 1065 | |
18068523 GOC |
1066 | /* Keep irq disabled to prevent changes to the clock */ |
1067 | local_irq_save(flags); | |
1d5f066e | 1068 | kvm_get_msr(v, MSR_IA32_TSC, &tsc_timestamp); |
759379dd | 1069 | kernel_ns = get_kernel_ns(); |
8cfdc000 | 1070 | this_tsc_khz = __get_cpu_var(cpu_tsc_khz); |
18068523 | 1071 | |
8cfdc000 | 1072 | if (unlikely(this_tsc_khz == 0)) { |
c285545f | 1073 | local_irq_restore(flags); |
34c238a1 | 1074 | kvm_make_request(KVM_REQ_CLOCK_UPDATE, v); |
8cfdc000 ZA |
1075 | return 1; |
1076 | } | |
18068523 | 1077 | |
c285545f ZA |
1078 | /* |
1079 | * We may have to catch up the TSC to match elapsed wall clock | |
1080 | * time for two reasons, even if kvmclock is used. | |
1081 | * 1) CPU could have been running below the maximum TSC rate | |
1082 | * 2) Broken TSC compensation resets the base at each VCPU | |
1083 | * entry to avoid unknown leaps of TSC even when running | |
1084 | * again on the same CPU. This may cause apparent elapsed | |
1085 | * time to disappear, and the guest to stand still or run | |
1086 | * very slowly. | |
1087 | */ | |
1088 | if (vcpu->tsc_catchup) { | |
1089 | u64 tsc = compute_guest_tsc(v, kernel_ns); | |
1090 | if (tsc > tsc_timestamp) { | |
1091 | kvm_x86_ops->adjust_tsc_offset(v, tsc - tsc_timestamp); | |
1092 | tsc_timestamp = tsc; | |
1093 | } | |
50d0a0f9 GH |
1094 | } |
1095 | ||
18068523 GOC |
1096 | local_irq_restore(flags); |
1097 | ||
c285545f ZA |
1098 | if (!vcpu->time_page) |
1099 | return 0; | |
18068523 | 1100 | |
1d5f066e ZA |
1101 | /* |
1102 | * Time as measured by the TSC may go backwards when resetting the base | |
1103 | * tsc_timestamp. The reason for this is that the TSC resolution is | |
1104 | * higher than the resolution of the other clock scales. Thus, many | |
1105 | * possible measurments of the TSC correspond to one measurement of any | |
1106 | * other clock, and so a spread of values is possible. This is not a | |
1107 | * problem for the computation of the nanosecond clock; with TSC rates | |
1108 | * around 1GHZ, there can only be a few cycles which correspond to one | |
1109 | * nanosecond value, and any path through this code will inevitably | |
1110 | * take longer than that. However, with the kernel_ns value itself, | |
1111 | * the precision may be much lower, down to HZ granularity. If the | |
1112 | * first sampling of TSC against kernel_ns ends in the low part of the | |
1113 | * range, and the second in the high end of the range, we can get: | |
1114 | * | |
1115 | * (TSC - offset_low) * S + kns_old > (TSC - offset_high) * S + kns_new | |
1116 | * | |
1117 | * As the sampling errors potentially range in the thousands of cycles, | |
1118 | * it is possible such a time value has already been observed by the | |
1119 | * guest. To protect against this, we must compute the system time as | |
1120 | * observed by the guest and ensure the new system time is greater. | |
1121 | */ | |
1122 | max_kernel_ns = 0; | |
1123 | if (vcpu->hv_clock.tsc_timestamp && vcpu->last_guest_tsc) { | |
1124 | max_kernel_ns = vcpu->last_guest_tsc - | |
1125 | vcpu->hv_clock.tsc_timestamp; | |
1126 | max_kernel_ns = pvclock_scale_delta(max_kernel_ns, | |
1127 | vcpu->hv_clock.tsc_to_system_mul, | |
1128 | vcpu->hv_clock.tsc_shift); | |
1129 | max_kernel_ns += vcpu->last_kernel_ns; | |
1130 | } | |
afbcf7ab | 1131 | |
e48672fa | 1132 | if (unlikely(vcpu->hw_tsc_khz != this_tsc_khz)) { |
5f4e3f88 ZA |
1133 | kvm_get_time_scale(NSEC_PER_SEC / 1000, this_tsc_khz, |
1134 | &vcpu->hv_clock.tsc_shift, | |
1135 | &vcpu->hv_clock.tsc_to_system_mul); | |
e48672fa | 1136 | vcpu->hw_tsc_khz = this_tsc_khz; |
8cfdc000 ZA |
1137 | } |
1138 | ||
1d5f066e ZA |
1139 | if (max_kernel_ns > kernel_ns) |
1140 | kernel_ns = max_kernel_ns; | |
1141 | ||
8cfdc000 | 1142 | /* With all the info we got, fill in the values */ |
1d5f066e | 1143 | vcpu->hv_clock.tsc_timestamp = tsc_timestamp; |
759379dd | 1144 | vcpu->hv_clock.system_time = kernel_ns + v->kvm->arch.kvmclock_offset; |
1d5f066e | 1145 | vcpu->last_kernel_ns = kernel_ns; |
28e4639a | 1146 | vcpu->last_guest_tsc = tsc_timestamp; |
371bcf64 GC |
1147 | vcpu->hv_clock.flags = 0; |
1148 | ||
18068523 GOC |
1149 | /* |
1150 | * The interface expects us to write an even number signaling that the | |
1151 | * update is finished. Since the guest won't see the intermediate | |
50d0a0f9 | 1152 | * state, we just increase by 2 at the end. |
18068523 | 1153 | */ |
50d0a0f9 | 1154 | vcpu->hv_clock.version += 2; |
18068523 GOC |
1155 | |
1156 | shared_kaddr = kmap_atomic(vcpu->time_page, KM_USER0); | |
1157 | ||
1158 | memcpy(shared_kaddr + vcpu->time_offset, &vcpu->hv_clock, | |
50d0a0f9 | 1159 | sizeof(vcpu->hv_clock)); |
18068523 GOC |
1160 | |
1161 | kunmap_atomic(shared_kaddr, KM_USER0); | |
1162 | ||
1163 | mark_page_dirty(v->kvm, vcpu->time >> PAGE_SHIFT); | |
8cfdc000 | 1164 | return 0; |
c8076604 GH |
1165 | } |
1166 | ||
9ba075a6 AK |
1167 | static bool msr_mtrr_valid(unsigned msr) |
1168 | { | |
1169 | switch (msr) { | |
1170 | case 0x200 ... 0x200 + 2 * KVM_NR_VAR_MTRR - 1: | |
1171 | case MSR_MTRRfix64K_00000: | |
1172 | case MSR_MTRRfix16K_80000: | |
1173 | case MSR_MTRRfix16K_A0000: | |
1174 | case MSR_MTRRfix4K_C0000: | |
1175 | case MSR_MTRRfix4K_C8000: | |
1176 | case MSR_MTRRfix4K_D0000: | |
1177 | case MSR_MTRRfix4K_D8000: | |
1178 | case MSR_MTRRfix4K_E0000: | |
1179 | case MSR_MTRRfix4K_E8000: | |
1180 | case MSR_MTRRfix4K_F0000: | |
1181 | case MSR_MTRRfix4K_F8000: | |
1182 | case MSR_MTRRdefType: | |
1183 | case MSR_IA32_CR_PAT: | |
1184 | return true; | |
1185 | case 0x2f8: | |
1186 | return true; | |
1187 | } | |
1188 | return false; | |
1189 | } | |
1190 | ||
d6289b93 MT |
1191 | static bool valid_pat_type(unsigned t) |
1192 | { | |
1193 | return t < 8 && (1 << t) & 0xf3; /* 0, 1, 4, 5, 6, 7 */ | |
1194 | } | |
1195 | ||
1196 | static bool valid_mtrr_type(unsigned t) | |
1197 | { | |
1198 | return t < 8 && (1 << t) & 0x73; /* 0, 1, 4, 5, 6 */ | |
1199 | } | |
1200 | ||
1201 | static bool mtrr_valid(struct kvm_vcpu *vcpu, u32 msr, u64 data) | |
1202 | { | |
1203 | int i; | |
1204 | ||
1205 | if (!msr_mtrr_valid(msr)) | |
1206 | return false; | |
1207 | ||
1208 | if (msr == MSR_IA32_CR_PAT) { | |
1209 | for (i = 0; i < 8; i++) | |
1210 | if (!valid_pat_type((data >> (i * 8)) & 0xff)) | |
1211 | return false; | |
1212 | return true; | |
1213 | } else if (msr == MSR_MTRRdefType) { | |
1214 | if (data & ~0xcff) | |
1215 | return false; | |
1216 | return valid_mtrr_type(data & 0xff); | |
1217 | } else if (msr >= MSR_MTRRfix64K_00000 && msr <= MSR_MTRRfix4K_F8000) { | |
1218 | for (i = 0; i < 8 ; i++) | |
1219 | if (!valid_mtrr_type((data >> (i * 8)) & 0xff)) | |
1220 | return false; | |
1221 | return true; | |
1222 | } | |
1223 | ||
1224 | /* variable MTRRs */ | |
1225 | return valid_mtrr_type(data & 0xff); | |
1226 | } | |
1227 | ||
9ba075a6 AK |
1228 | static int set_msr_mtrr(struct kvm_vcpu *vcpu, u32 msr, u64 data) |
1229 | { | |
0bed3b56 SY |
1230 | u64 *p = (u64 *)&vcpu->arch.mtrr_state.fixed_ranges; |
1231 | ||
d6289b93 | 1232 | if (!mtrr_valid(vcpu, msr, data)) |
9ba075a6 AK |
1233 | return 1; |
1234 | ||
0bed3b56 SY |
1235 | if (msr == MSR_MTRRdefType) { |
1236 | vcpu->arch.mtrr_state.def_type = data; | |
1237 | vcpu->arch.mtrr_state.enabled = (data & 0xc00) >> 10; | |
1238 | } else if (msr == MSR_MTRRfix64K_00000) | |
1239 | p[0] = data; | |
1240 | else if (msr == MSR_MTRRfix16K_80000 || msr == MSR_MTRRfix16K_A0000) | |
1241 | p[1 + msr - MSR_MTRRfix16K_80000] = data; | |
1242 | else if (msr >= MSR_MTRRfix4K_C0000 && msr <= MSR_MTRRfix4K_F8000) | |
1243 | p[3 + msr - MSR_MTRRfix4K_C0000] = data; | |
1244 | else if (msr == MSR_IA32_CR_PAT) | |
1245 | vcpu->arch.pat = data; | |
1246 | else { /* Variable MTRRs */ | |
1247 | int idx, is_mtrr_mask; | |
1248 | u64 *pt; | |
1249 | ||
1250 | idx = (msr - 0x200) / 2; | |
1251 | is_mtrr_mask = msr - 0x200 - 2 * idx; | |
1252 | if (!is_mtrr_mask) | |
1253 | pt = | |
1254 | (u64 *)&vcpu->arch.mtrr_state.var_ranges[idx].base_lo; | |
1255 | else | |
1256 | pt = | |
1257 | (u64 *)&vcpu->arch.mtrr_state.var_ranges[idx].mask_lo; | |
1258 | *pt = data; | |
1259 | } | |
1260 | ||
1261 | kvm_mmu_reset_context(vcpu); | |
9ba075a6 AK |
1262 | return 0; |
1263 | } | |
15c4a640 | 1264 | |
890ca9ae | 1265 | static int set_msr_mce(struct kvm_vcpu *vcpu, u32 msr, u64 data) |
15c4a640 | 1266 | { |
890ca9ae HY |
1267 | u64 mcg_cap = vcpu->arch.mcg_cap; |
1268 | unsigned bank_num = mcg_cap & 0xff; | |
1269 | ||
15c4a640 | 1270 | switch (msr) { |
15c4a640 | 1271 | case MSR_IA32_MCG_STATUS: |
890ca9ae | 1272 | vcpu->arch.mcg_status = data; |
15c4a640 | 1273 | break; |
c7ac679c | 1274 | case MSR_IA32_MCG_CTL: |
890ca9ae HY |
1275 | if (!(mcg_cap & MCG_CTL_P)) |
1276 | return 1; | |
1277 | if (data != 0 && data != ~(u64)0) | |
1278 | return -1; | |
1279 | vcpu->arch.mcg_ctl = data; | |
1280 | break; | |
1281 | default: | |
1282 | if (msr >= MSR_IA32_MC0_CTL && | |
1283 | msr < MSR_IA32_MC0_CTL + 4 * bank_num) { | |
1284 | u32 offset = msr - MSR_IA32_MC0_CTL; | |
114be429 AP |
1285 | /* only 0 or all 1s can be written to IA32_MCi_CTL |
1286 | * some Linux kernels though clear bit 10 in bank 4 to | |
1287 | * workaround a BIOS/GART TBL issue on AMD K8s, ignore | |
1288 | * this to avoid an uncatched #GP in the guest | |
1289 | */ | |
890ca9ae | 1290 | if ((offset & 0x3) == 0 && |
114be429 | 1291 | data != 0 && (data | (1 << 10)) != ~(u64)0) |
890ca9ae HY |
1292 | return -1; |
1293 | vcpu->arch.mce_banks[offset] = data; | |
1294 | break; | |
1295 | } | |
1296 | return 1; | |
1297 | } | |
1298 | return 0; | |
1299 | } | |
1300 | ||
ffde22ac ES |
1301 | static int xen_hvm_config(struct kvm_vcpu *vcpu, u64 data) |
1302 | { | |
1303 | struct kvm *kvm = vcpu->kvm; | |
1304 | int lm = is_long_mode(vcpu); | |
1305 | u8 *blob_addr = lm ? (u8 *)(long)kvm->arch.xen_hvm_config.blob_addr_64 | |
1306 | : (u8 *)(long)kvm->arch.xen_hvm_config.blob_addr_32; | |
1307 | u8 blob_size = lm ? kvm->arch.xen_hvm_config.blob_size_64 | |
1308 | : kvm->arch.xen_hvm_config.blob_size_32; | |
1309 | u32 page_num = data & ~PAGE_MASK; | |
1310 | u64 page_addr = data & PAGE_MASK; | |
1311 | u8 *page; | |
1312 | int r; | |
1313 | ||
1314 | r = -E2BIG; | |
1315 | if (page_num >= blob_size) | |
1316 | goto out; | |
1317 | r = -ENOMEM; | |
1318 | page = kzalloc(PAGE_SIZE, GFP_KERNEL); | |
1319 | if (!page) | |
1320 | goto out; | |
1321 | r = -EFAULT; | |
1322 | if (copy_from_user(page, blob_addr + (page_num * PAGE_SIZE), PAGE_SIZE)) | |
1323 | goto out_free; | |
1324 | if (kvm_write_guest(kvm, page_addr, page, PAGE_SIZE)) | |
1325 | goto out_free; | |
1326 | r = 0; | |
1327 | out_free: | |
1328 | kfree(page); | |
1329 | out: | |
1330 | return r; | |
1331 | } | |
1332 | ||
55cd8e5a GN |
1333 | static bool kvm_hv_hypercall_enabled(struct kvm *kvm) |
1334 | { | |
1335 | return kvm->arch.hv_hypercall & HV_X64_MSR_HYPERCALL_ENABLE; | |
1336 | } | |
1337 | ||
1338 | static bool kvm_hv_msr_partition_wide(u32 msr) | |
1339 | { | |
1340 | bool r = false; | |
1341 | switch (msr) { | |
1342 | case HV_X64_MSR_GUEST_OS_ID: | |
1343 | case HV_X64_MSR_HYPERCALL: | |
1344 | r = true; | |
1345 | break; | |
1346 | } | |
1347 | ||
1348 | return r; | |
1349 | } | |
1350 | ||
1351 | static int set_msr_hyperv_pw(struct kvm_vcpu *vcpu, u32 msr, u64 data) | |
1352 | { | |
1353 | struct kvm *kvm = vcpu->kvm; | |
1354 | ||
1355 | switch (msr) { | |
1356 | case HV_X64_MSR_GUEST_OS_ID: | |
1357 | kvm->arch.hv_guest_os_id = data; | |
1358 | /* setting guest os id to zero disables hypercall page */ | |
1359 | if (!kvm->arch.hv_guest_os_id) | |
1360 | kvm->arch.hv_hypercall &= ~HV_X64_MSR_HYPERCALL_ENABLE; | |
1361 | break; | |
1362 | case HV_X64_MSR_HYPERCALL: { | |
1363 | u64 gfn; | |
1364 | unsigned long addr; | |
1365 | u8 instructions[4]; | |
1366 | ||
1367 | /* if guest os id is not set hypercall should remain disabled */ | |
1368 | if (!kvm->arch.hv_guest_os_id) | |
1369 | break; | |
1370 | if (!(data & HV_X64_MSR_HYPERCALL_ENABLE)) { | |
1371 | kvm->arch.hv_hypercall = data; | |
1372 | break; | |
1373 | } | |
1374 | gfn = data >> HV_X64_MSR_HYPERCALL_PAGE_ADDRESS_SHIFT; | |
1375 | addr = gfn_to_hva(kvm, gfn); | |
1376 | if (kvm_is_error_hva(addr)) | |
1377 | return 1; | |
1378 | kvm_x86_ops->patch_hypercall(vcpu, instructions); | |
1379 | ((unsigned char *)instructions)[3] = 0xc3; /* ret */ | |
1380 | if (copy_to_user((void __user *)addr, instructions, 4)) | |
1381 | return 1; | |
1382 | kvm->arch.hv_hypercall = data; | |
1383 | break; | |
1384 | } | |
1385 | default: | |
1386 | pr_unimpl(vcpu, "HYPER-V unimplemented wrmsr: 0x%x " | |
1387 | "data 0x%llx\n", msr, data); | |
1388 | return 1; | |
1389 | } | |
1390 | return 0; | |
1391 | } | |
1392 | ||
1393 | static int set_msr_hyperv(struct kvm_vcpu *vcpu, u32 msr, u64 data) | |
1394 | { | |
10388a07 GN |
1395 | switch (msr) { |
1396 | case HV_X64_MSR_APIC_ASSIST_PAGE: { | |
1397 | unsigned long addr; | |
55cd8e5a | 1398 | |
10388a07 GN |
1399 | if (!(data & HV_X64_MSR_APIC_ASSIST_PAGE_ENABLE)) { |
1400 | vcpu->arch.hv_vapic = data; | |
1401 | break; | |
1402 | } | |
1403 | addr = gfn_to_hva(vcpu->kvm, data >> | |
1404 | HV_X64_MSR_APIC_ASSIST_PAGE_ADDRESS_SHIFT); | |
1405 | if (kvm_is_error_hva(addr)) | |
1406 | return 1; | |
1407 | if (clear_user((void __user *)addr, PAGE_SIZE)) | |
1408 | return 1; | |
1409 | vcpu->arch.hv_vapic = data; | |
1410 | break; | |
1411 | } | |
1412 | case HV_X64_MSR_EOI: | |
1413 | return kvm_hv_vapic_msr_write(vcpu, APIC_EOI, data); | |
1414 | case HV_X64_MSR_ICR: | |
1415 | return kvm_hv_vapic_msr_write(vcpu, APIC_ICR, data); | |
1416 | case HV_X64_MSR_TPR: | |
1417 | return kvm_hv_vapic_msr_write(vcpu, APIC_TASKPRI, data); | |
1418 | default: | |
1419 | pr_unimpl(vcpu, "HYPER-V unimplemented wrmsr: 0x%x " | |
1420 | "data 0x%llx\n", msr, data); | |
1421 | return 1; | |
1422 | } | |
1423 | ||
1424 | return 0; | |
55cd8e5a GN |
1425 | } |
1426 | ||
344d9588 GN |
1427 | static int kvm_pv_enable_async_pf(struct kvm_vcpu *vcpu, u64 data) |
1428 | { | |
1429 | gpa_t gpa = data & ~0x3f; | |
1430 | ||
6adba527 GN |
1431 | /* Bits 2:5 are resrved, Should be zero */ |
1432 | if (data & 0x3c) | |
344d9588 GN |
1433 | return 1; |
1434 | ||
1435 | vcpu->arch.apf.msr_val = data; | |
1436 | ||
1437 | if (!(data & KVM_ASYNC_PF_ENABLED)) { | |
1438 | kvm_clear_async_pf_completion_queue(vcpu); | |
1439 | kvm_async_pf_hash_reset(vcpu); | |
1440 | return 0; | |
1441 | } | |
1442 | ||
1443 | if (kvm_gfn_to_hva_cache_init(vcpu->kvm, &vcpu->arch.apf.data, gpa)) | |
1444 | return 1; | |
1445 | ||
6adba527 | 1446 | vcpu->arch.apf.send_user_only = !(data & KVM_ASYNC_PF_SEND_ALWAYS); |
344d9588 GN |
1447 | kvm_async_pf_wakeup_all(vcpu); |
1448 | return 0; | |
1449 | } | |
1450 | ||
15c4a640 CO |
1451 | int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data) |
1452 | { | |
1453 | switch (msr) { | |
15c4a640 | 1454 | case MSR_EFER: |
b69e8cae | 1455 | return set_efer(vcpu, data); |
8f1589d9 AP |
1456 | case MSR_K7_HWCR: |
1457 | data &= ~(u64)0x40; /* ignore flush filter disable */ | |
82494028 | 1458 | data &= ~(u64)0x100; /* ignore ignne emulation enable */ |
8f1589d9 AP |
1459 | if (data != 0) { |
1460 | pr_unimpl(vcpu, "unimplemented HWCR wrmsr: 0x%llx\n", | |
1461 | data); | |
1462 | return 1; | |
1463 | } | |
15c4a640 | 1464 | break; |
f7c6d140 AP |
1465 | case MSR_FAM10H_MMIO_CONF_BASE: |
1466 | if (data != 0) { | |
1467 | pr_unimpl(vcpu, "unimplemented MMIO_CONF_BASE wrmsr: " | |
1468 | "0x%llx\n", data); | |
1469 | return 1; | |
1470 | } | |
15c4a640 | 1471 | break; |
c323c0e5 | 1472 | case MSR_AMD64_NB_CFG: |
c7ac679c | 1473 | break; |
b5e2fec0 AG |
1474 | case MSR_IA32_DEBUGCTLMSR: |
1475 | if (!data) { | |
1476 | /* We support the non-activated case already */ | |
1477 | break; | |
1478 | } else if (data & ~(DEBUGCTLMSR_LBR | DEBUGCTLMSR_BTF)) { | |
1479 | /* Values other than LBR and BTF are vendor-specific, | |
1480 | thus reserved and should throw a #GP */ | |
1481 | return 1; | |
1482 | } | |
1483 | pr_unimpl(vcpu, "%s: MSR_IA32_DEBUGCTLMSR 0x%llx, nop\n", | |
1484 | __func__, data); | |
1485 | break; | |
15c4a640 CO |
1486 | case MSR_IA32_UCODE_REV: |
1487 | case MSR_IA32_UCODE_WRITE: | |
61a6bd67 | 1488 | case MSR_VM_HSAVE_PA: |
6098ca93 | 1489 | case MSR_AMD64_PATCH_LOADER: |
15c4a640 | 1490 | break; |
9ba075a6 AK |
1491 | case 0x200 ... 0x2ff: |
1492 | return set_msr_mtrr(vcpu, msr, data); | |
15c4a640 CO |
1493 | case MSR_IA32_APICBASE: |
1494 | kvm_set_apic_base(vcpu, data); | |
1495 | break; | |
0105d1a5 GN |
1496 | case APIC_BASE_MSR ... APIC_BASE_MSR + 0x3ff: |
1497 | return kvm_x2apic_msr_write(vcpu, msr, data); | |
15c4a640 | 1498 | case MSR_IA32_MISC_ENABLE: |
ad312c7c | 1499 | vcpu->arch.ia32_misc_enable_msr = data; |
15c4a640 | 1500 | break; |
11c6bffa | 1501 | case MSR_KVM_WALL_CLOCK_NEW: |
18068523 GOC |
1502 | case MSR_KVM_WALL_CLOCK: |
1503 | vcpu->kvm->arch.wall_clock = data; | |
1504 | kvm_write_wall_clock(vcpu->kvm, data); | |
1505 | break; | |
11c6bffa | 1506 | case MSR_KVM_SYSTEM_TIME_NEW: |
18068523 GOC |
1507 | case MSR_KVM_SYSTEM_TIME: { |
1508 | if (vcpu->arch.time_page) { | |
1509 | kvm_release_page_dirty(vcpu->arch.time_page); | |
1510 | vcpu->arch.time_page = NULL; | |
1511 | } | |
1512 | ||
1513 | vcpu->arch.time = data; | |
c285545f | 1514 | kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu); |
18068523 GOC |
1515 | |
1516 | /* we verify if the enable bit is set... */ | |
1517 | if (!(data & 1)) | |
1518 | break; | |
1519 | ||
1520 | /* ...but clean it before doing the actual write */ | |
1521 | vcpu->arch.time_offset = data & ~(PAGE_MASK | 1); | |
1522 | ||
18068523 GOC |
1523 | vcpu->arch.time_page = |
1524 | gfn_to_page(vcpu->kvm, data >> PAGE_SHIFT); | |
18068523 GOC |
1525 | |
1526 | if (is_error_page(vcpu->arch.time_page)) { | |
1527 | kvm_release_page_clean(vcpu->arch.time_page); | |
1528 | vcpu->arch.time_page = NULL; | |
1529 | } | |
18068523 GOC |
1530 | break; |
1531 | } | |
344d9588 GN |
1532 | case MSR_KVM_ASYNC_PF_EN: |
1533 | if (kvm_pv_enable_async_pf(vcpu, data)) | |
1534 | return 1; | |
1535 | break; | |
890ca9ae HY |
1536 | case MSR_IA32_MCG_CTL: |
1537 | case MSR_IA32_MCG_STATUS: | |
1538 | case MSR_IA32_MC0_CTL ... MSR_IA32_MC0_CTL + 4 * KVM_MAX_MCE_BANKS - 1: | |
1539 | return set_msr_mce(vcpu, msr, data); | |
71db6023 AP |
1540 | |
1541 | /* Performance counters are not protected by a CPUID bit, | |
1542 | * so we should check all of them in the generic path for the sake of | |
1543 | * cross vendor migration. | |
1544 | * Writing a zero into the event select MSRs disables them, | |
1545 | * which we perfectly emulate ;-). Any other value should be at least | |
1546 | * reported, some guests depend on them. | |
1547 | */ | |
1548 | case MSR_P6_EVNTSEL0: | |
1549 | case MSR_P6_EVNTSEL1: | |
1550 | case MSR_K7_EVNTSEL0: | |
1551 | case MSR_K7_EVNTSEL1: | |
1552 | case MSR_K7_EVNTSEL2: | |
1553 | case MSR_K7_EVNTSEL3: | |
1554 | if (data != 0) | |
1555 | pr_unimpl(vcpu, "unimplemented perfctr wrmsr: " | |
1556 | "0x%x data 0x%llx\n", msr, data); | |
1557 | break; | |
1558 | /* at least RHEL 4 unconditionally writes to the perfctr registers, | |
1559 | * so we ignore writes to make it happy. | |
1560 | */ | |
1561 | case MSR_P6_PERFCTR0: | |
1562 | case MSR_P6_PERFCTR1: | |
1563 | case MSR_K7_PERFCTR0: | |
1564 | case MSR_K7_PERFCTR1: | |
1565 | case MSR_K7_PERFCTR2: | |
1566 | case MSR_K7_PERFCTR3: | |
1567 | pr_unimpl(vcpu, "unimplemented perfctr wrmsr: " | |
1568 | "0x%x data 0x%llx\n", msr, data); | |
1569 | break; | |
84e0cefa JS |
1570 | case MSR_K7_CLK_CTL: |
1571 | /* | |
1572 | * Ignore all writes to this no longer documented MSR. | |
1573 | * Writes are only relevant for old K7 processors, | |
1574 | * all pre-dating SVM, but a recommended workaround from | |
1575 | * AMD for these chips. It is possible to speicify the | |
1576 | * affected processor models on the command line, hence | |
1577 | * the need to ignore the workaround. | |
1578 | */ | |
1579 | break; | |
55cd8e5a GN |
1580 | case HV_X64_MSR_GUEST_OS_ID ... HV_X64_MSR_SINT15: |
1581 | if (kvm_hv_msr_partition_wide(msr)) { | |
1582 | int r; | |
1583 | mutex_lock(&vcpu->kvm->lock); | |
1584 | r = set_msr_hyperv_pw(vcpu, msr, data); | |
1585 | mutex_unlock(&vcpu->kvm->lock); | |
1586 | return r; | |
1587 | } else | |
1588 | return set_msr_hyperv(vcpu, msr, data); | |
1589 | break; | |
15c4a640 | 1590 | default: |
ffde22ac ES |
1591 | if (msr && (msr == vcpu->kvm->arch.xen_hvm_config.msr)) |
1592 | return xen_hvm_config(vcpu, data); | |
ed85c068 AP |
1593 | if (!ignore_msrs) { |
1594 | pr_unimpl(vcpu, "unhandled wrmsr: 0x%x data %llx\n", | |
1595 | msr, data); | |
1596 | return 1; | |
1597 | } else { | |
1598 | pr_unimpl(vcpu, "ignored wrmsr: 0x%x data %llx\n", | |
1599 | msr, data); | |
1600 | break; | |
1601 | } | |
15c4a640 CO |
1602 | } |
1603 | return 0; | |
1604 | } | |
1605 | EXPORT_SYMBOL_GPL(kvm_set_msr_common); | |
1606 | ||
1607 | ||
1608 | /* | |
1609 | * Reads an msr value (of 'msr_index') into 'pdata'. | |
1610 | * Returns 0 on success, non-0 otherwise. | |
1611 | * Assumes vcpu_load() was already called. | |
1612 | */ | |
1613 | int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata) | |
1614 | { | |
1615 | return kvm_x86_ops->get_msr(vcpu, msr_index, pdata); | |
1616 | } | |
1617 | ||
9ba075a6 AK |
1618 | static int get_msr_mtrr(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata) |
1619 | { | |
0bed3b56 SY |
1620 | u64 *p = (u64 *)&vcpu->arch.mtrr_state.fixed_ranges; |
1621 | ||
9ba075a6 AK |
1622 | if (!msr_mtrr_valid(msr)) |
1623 | return 1; | |
1624 | ||
0bed3b56 SY |
1625 | if (msr == MSR_MTRRdefType) |
1626 | *pdata = vcpu->arch.mtrr_state.def_type + | |
1627 | (vcpu->arch.mtrr_state.enabled << 10); | |
1628 | else if (msr == MSR_MTRRfix64K_00000) | |
1629 | *pdata = p[0]; | |
1630 | else if (msr == MSR_MTRRfix16K_80000 || msr == MSR_MTRRfix16K_A0000) | |
1631 | *pdata = p[1 + msr - MSR_MTRRfix16K_80000]; | |
1632 | else if (msr >= MSR_MTRRfix4K_C0000 && msr <= MSR_MTRRfix4K_F8000) | |
1633 | *pdata = p[3 + msr - MSR_MTRRfix4K_C0000]; | |
1634 | else if (msr == MSR_IA32_CR_PAT) | |
1635 | *pdata = vcpu->arch.pat; | |
1636 | else { /* Variable MTRRs */ | |
1637 | int idx, is_mtrr_mask; | |
1638 | u64 *pt; | |
1639 | ||
1640 | idx = (msr - 0x200) / 2; | |
1641 | is_mtrr_mask = msr - 0x200 - 2 * idx; | |
1642 | if (!is_mtrr_mask) | |
1643 | pt = | |
1644 | (u64 *)&vcpu->arch.mtrr_state.var_ranges[idx].base_lo; | |
1645 | else | |
1646 | pt = | |
1647 | (u64 *)&vcpu->arch.mtrr_state.var_ranges[idx].mask_lo; | |
1648 | *pdata = *pt; | |
1649 | } | |
1650 | ||
9ba075a6 AK |
1651 | return 0; |
1652 | } | |
1653 | ||
890ca9ae | 1654 | static int get_msr_mce(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata) |
15c4a640 CO |
1655 | { |
1656 | u64 data; | |
890ca9ae HY |
1657 | u64 mcg_cap = vcpu->arch.mcg_cap; |
1658 | unsigned bank_num = mcg_cap & 0xff; | |
15c4a640 CO |
1659 | |
1660 | switch (msr) { | |
15c4a640 CO |
1661 | case MSR_IA32_P5_MC_ADDR: |
1662 | case MSR_IA32_P5_MC_TYPE: | |
890ca9ae HY |
1663 | data = 0; |
1664 | break; | |
15c4a640 | 1665 | case MSR_IA32_MCG_CAP: |
890ca9ae HY |
1666 | data = vcpu->arch.mcg_cap; |
1667 | break; | |
c7ac679c | 1668 | case MSR_IA32_MCG_CTL: |
890ca9ae HY |
1669 | if (!(mcg_cap & MCG_CTL_P)) |
1670 | return 1; | |
1671 | data = vcpu->arch.mcg_ctl; | |
1672 | break; | |
1673 | case MSR_IA32_MCG_STATUS: | |
1674 | data = vcpu->arch.mcg_status; | |
1675 | break; | |
1676 | default: | |
1677 | if (msr >= MSR_IA32_MC0_CTL && | |
1678 | msr < MSR_IA32_MC0_CTL + 4 * bank_num) { | |
1679 | u32 offset = msr - MSR_IA32_MC0_CTL; | |
1680 | data = vcpu->arch.mce_banks[offset]; | |
1681 | break; | |
1682 | } | |
1683 | return 1; | |
1684 | } | |
1685 | *pdata = data; | |
1686 | return 0; | |
1687 | } | |
1688 | ||
55cd8e5a GN |
1689 | static int get_msr_hyperv_pw(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata) |
1690 | { | |
1691 | u64 data = 0; | |
1692 | struct kvm *kvm = vcpu->kvm; | |
1693 | ||
1694 | switch (msr) { | |
1695 | case HV_X64_MSR_GUEST_OS_ID: | |
1696 | data = kvm->arch.hv_guest_os_id; | |
1697 | break; | |
1698 | case HV_X64_MSR_HYPERCALL: | |
1699 | data = kvm->arch.hv_hypercall; | |
1700 | break; | |
1701 | default: | |
1702 | pr_unimpl(vcpu, "Hyper-V unhandled rdmsr: 0x%x\n", msr); | |
1703 | return 1; | |
1704 | } | |
1705 | ||
1706 | *pdata = data; | |
1707 | return 0; | |
1708 | } | |
1709 | ||
1710 | static int get_msr_hyperv(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata) | |
1711 | { | |
1712 | u64 data = 0; | |
1713 | ||
1714 | switch (msr) { | |
1715 | case HV_X64_MSR_VP_INDEX: { | |
1716 | int r; | |
1717 | struct kvm_vcpu *v; | |
1718 | kvm_for_each_vcpu(r, v, vcpu->kvm) | |
1719 | if (v == vcpu) | |
1720 | data = r; | |
1721 | break; | |
1722 | } | |
10388a07 GN |
1723 | case HV_X64_MSR_EOI: |
1724 | return kvm_hv_vapic_msr_read(vcpu, APIC_EOI, pdata); | |
1725 | case HV_X64_MSR_ICR: | |
1726 | return kvm_hv_vapic_msr_read(vcpu, APIC_ICR, pdata); | |
1727 | case HV_X64_MSR_TPR: | |
1728 | return kvm_hv_vapic_msr_read(vcpu, APIC_TASKPRI, pdata); | |
55cd8e5a GN |
1729 | default: |
1730 | pr_unimpl(vcpu, "Hyper-V unhandled rdmsr: 0x%x\n", msr); | |
1731 | return 1; | |
1732 | } | |
1733 | *pdata = data; | |
1734 | return 0; | |
1735 | } | |
1736 | ||
890ca9ae HY |
1737 | int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata) |
1738 | { | |
1739 | u64 data; | |
1740 | ||
1741 | switch (msr) { | |
890ca9ae | 1742 | case MSR_IA32_PLATFORM_ID: |
15c4a640 | 1743 | case MSR_IA32_UCODE_REV: |
15c4a640 | 1744 | case MSR_IA32_EBL_CR_POWERON: |
b5e2fec0 AG |
1745 | case MSR_IA32_DEBUGCTLMSR: |
1746 | case MSR_IA32_LASTBRANCHFROMIP: | |
1747 | case MSR_IA32_LASTBRANCHTOIP: | |
1748 | case MSR_IA32_LASTINTFROMIP: | |
1749 | case MSR_IA32_LASTINTTOIP: | |
60af2ecd JSR |
1750 | case MSR_K8_SYSCFG: |
1751 | case MSR_K7_HWCR: | |
61a6bd67 | 1752 | case MSR_VM_HSAVE_PA: |
1f3ee616 AS |
1753 | case MSR_P6_PERFCTR0: |
1754 | case MSR_P6_PERFCTR1: | |
7fe29e0f AS |
1755 | case MSR_P6_EVNTSEL0: |
1756 | case MSR_P6_EVNTSEL1: | |
9e699624 | 1757 | case MSR_K7_EVNTSEL0: |
1f3ee616 | 1758 | case MSR_K7_PERFCTR0: |
1fdbd48c | 1759 | case MSR_K8_INT_PENDING_MSG: |
c323c0e5 | 1760 | case MSR_AMD64_NB_CFG: |
f7c6d140 | 1761 | case MSR_FAM10H_MMIO_CONF_BASE: |
15c4a640 CO |
1762 | data = 0; |
1763 | break; | |
9ba075a6 AK |
1764 | case MSR_MTRRcap: |
1765 | data = 0x500 | KVM_NR_VAR_MTRR; | |
1766 | break; | |
1767 | case 0x200 ... 0x2ff: | |
1768 | return get_msr_mtrr(vcpu, msr, pdata); | |
15c4a640 CO |
1769 | case 0xcd: /* fsb frequency */ |
1770 | data = 3; | |
1771 | break; | |
7b914098 JS |
1772 | /* |
1773 | * MSR_EBC_FREQUENCY_ID | |
1774 | * Conservative value valid for even the basic CPU models. | |
1775 | * Models 0,1: 000 in bits 23:21 indicating a bus speed of | |
1776 | * 100MHz, model 2 000 in bits 18:16 indicating 100MHz, | |
1777 | * and 266MHz for model 3, or 4. Set Core Clock | |
1778 | * Frequency to System Bus Frequency Ratio to 1 (bits | |
1779 | * 31:24) even though these are only valid for CPU | |
1780 | * models > 2, however guests may end up dividing or | |
1781 | * multiplying by zero otherwise. | |
1782 | */ | |
1783 | case MSR_EBC_FREQUENCY_ID: | |
1784 | data = 1 << 24; | |
1785 | break; | |
15c4a640 CO |
1786 | case MSR_IA32_APICBASE: |
1787 | data = kvm_get_apic_base(vcpu); | |
1788 | break; | |
0105d1a5 GN |
1789 | case APIC_BASE_MSR ... APIC_BASE_MSR + 0x3ff: |
1790 | return kvm_x2apic_msr_read(vcpu, msr, pdata); | |
1791 | break; | |
15c4a640 | 1792 | case MSR_IA32_MISC_ENABLE: |
ad312c7c | 1793 | data = vcpu->arch.ia32_misc_enable_msr; |
15c4a640 | 1794 | break; |
847f0ad8 AG |
1795 | case MSR_IA32_PERF_STATUS: |
1796 | /* TSC increment by tick */ | |
1797 | data = 1000ULL; | |
1798 | /* CPU multiplier */ | |
1799 | data |= (((uint64_t)4ULL) << 40); | |
1800 | break; | |
15c4a640 | 1801 | case MSR_EFER: |
f6801dff | 1802 | data = vcpu->arch.efer; |
15c4a640 | 1803 | break; |
18068523 | 1804 | case MSR_KVM_WALL_CLOCK: |
11c6bffa | 1805 | case MSR_KVM_WALL_CLOCK_NEW: |
18068523 GOC |
1806 | data = vcpu->kvm->arch.wall_clock; |
1807 | break; | |
1808 | case MSR_KVM_SYSTEM_TIME: | |
11c6bffa | 1809 | case MSR_KVM_SYSTEM_TIME_NEW: |
18068523 GOC |
1810 | data = vcpu->arch.time; |
1811 | break; | |
344d9588 GN |
1812 | case MSR_KVM_ASYNC_PF_EN: |
1813 | data = vcpu->arch.apf.msr_val; | |
1814 | break; | |
890ca9ae HY |
1815 | case MSR_IA32_P5_MC_ADDR: |
1816 | case MSR_IA32_P5_MC_TYPE: | |
1817 | case MSR_IA32_MCG_CAP: | |
1818 | case MSR_IA32_MCG_CTL: | |
1819 | case MSR_IA32_MCG_STATUS: | |
1820 | case MSR_IA32_MC0_CTL ... MSR_IA32_MC0_CTL + 4 * KVM_MAX_MCE_BANKS - 1: | |
1821 | return get_msr_mce(vcpu, msr, pdata); | |
84e0cefa JS |
1822 | case MSR_K7_CLK_CTL: |
1823 | /* | |
1824 | * Provide expected ramp-up count for K7. All other | |
1825 | * are set to zero, indicating minimum divisors for | |
1826 | * every field. | |
1827 | * | |
1828 | * This prevents guest kernels on AMD host with CPU | |
1829 | * type 6, model 8 and higher from exploding due to | |
1830 | * the rdmsr failing. | |
1831 | */ | |
1832 | data = 0x20000000; | |
1833 | break; | |
55cd8e5a GN |
1834 | case HV_X64_MSR_GUEST_OS_ID ... HV_X64_MSR_SINT15: |
1835 | if (kvm_hv_msr_partition_wide(msr)) { | |
1836 | int r; | |
1837 | mutex_lock(&vcpu->kvm->lock); | |
1838 | r = get_msr_hyperv_pw(vcpu, msr, pdata); | |
1839 | mutex_unlock(&vcpu->kvm->lock); | |
1840 | return r; | |
1841 | } else | |
1842 | return get_msr_hyperv(vcpu, msr, pdata); | |
1843 | break; | |
15c4a640 | 1844 | default: |
ed85c068 AP |
1845 | if (!ignore_msrs) { |
1846 | pr_unimpl(vcpu, "unhandled rdmsr: 0x%x\n", msr); | |
1847 | return 1; | |
1848 | } else { | |
1849 | pr_unimpl(vcpu, "ignored rdmsr: 0x%x\n", msr); | |
1850 | data = 0; | |
1851 | } | |
1852 | break; | |
15c4a640 CO |
1853 | } |
1854 | *pdata = data; | |
1855 | return 0; | |
1856 | } | |
1857 | EXPORT_SYMBOL_GPL(kvm_get_msr_common); | |
1858 | ||
313a3dc7 CO |
1859 | /* |
1860 | * Read or write a bunch of msrs. All parameters are kernel addresses. | |
1861 | * | |
1862 | * @return number of msrs set successfully. | |
1863 | */ | |
1864 | static int __msr_io(struct kvm_vcpu *vcpu, struct kvm_msrs *msrs, | |
1865 | struct kvm_msr_entry *entries, | |
1866 | int (*do_msr)(struct kvm_vcpu *vcpu, | |
1867 | unsigned index, u64 *data)) | |
1868 | { | |
f656ce01 | 1869 | int i, idx; |
313a3dc7 | 1870 | |
f656ce01 | 1871 | idx = srcu_read_lock(&vcpu->kvm->srcu); |
313a3dc7 CO |
1872 | for (i = 0; i < msrs->nmsrs; ++i) |
1873 | if (do_msr(vcpu, entries[i].index, &entries[i].data)) | |
1874 | break; | |
f656ce01 | 1875 | srcu_read_unlock(&vcpu->kvm->srcu, idx); |
313a3dc7 | 1876 | |
313a3dc7 CO |
1877 | return i; |
1878 | } | |
1879 | ||
1880 | /* | |
1881 | * Read or write a bunch of msrs. Parameters are user addresses. | |
1882 | * | |
1883 | * @return number of msrs set successfully. | |
1884 | */ | |
1885 | static int msr_io(struct kvm_vcpu *vcpu, struct kvm_msrs __user *user_msrs, | |
1886 | int (*do_msr)(struct kvm_vcpu *vcpu, | |
1887 | unsigned index, u64 *data), | |
1888 | int writeback) | |
1889 | { | |
1890 | struct kvm_msrs msrs; | |
1891 | struct kvm_msr_entry *entries; | |
1892 | int r, n; | |
1893 | unsigned size; | |
1894 | ||
1895 | r = -EFAULT; | |
1896 | if (copy_from_user(&msrs, user_msrs, sizeof msrs)) | |
1897 | goto out; | |
1898 | ||
1899 | r = -E2BIG; | |
1900 | if (msrs.nmsrs >= MAX_IO_MSRS) | |
1901 | goto out; | |
1902 | ||
1903 | r = -ENOMEM; | |
1904 | size = sizeof(struct kvm_msr_entry) * msrs.nmsrs; | |
7a73c028 | 1905 | entries = kmalloc(size, GFP_KERNEL); |
313a3dc7 CO |
1906 | if (!entries) |
1907 | goto out; | |
1908 | ||
1909 | r = -EFAULT; | |
1910 | if (copy_from_user(entries, user_msrs->entries, size)) | |
1911 | goto out_free; | |
1912 | ||
1913 | r = n = __msr_io(vcpu, &msrs, entries, do_msr); | |
1914 | if (r < 0) | |
1915 | goto out_free; | |
1916 | ||
1917 | r = -EFAULT; | |
1918 | if (writeback && copy_to_user(user_msrs->entries, entries, size)) | |
1919 | goto out_free; | |
1920 | ||
1921 | r = n; | |
1922 | ||
1923 | out_free: | |
7a73c028 | 1924 | kfree(entries); |
313a3dc7 CO |
1925 | out: |
1926 | return r; | |
1927 | } | |
1928 | ||
018d00d2 ZX |
1929 | int kvm_dev_ioctl_check_extension(long ext) |
1930 | { | |
1931 | int r; | |
1932 | ||
1933 | switch (ext) { | |
1934 | case KVM_CAP_IRQCHIP: | |
1935 | case KVM_CAP_HLT: | |
1936 | case KVM_CAP_MMU_SHADOW_CACHE_CONTROL: | |
018d00d2 | 1937 | case KVM_CAP_SET_TSS_ADDR: |
07716717 | 1938 | case KVM_CAP_EXT_CPUID: |
c8076604 | 1939 | case KVM_CAP_CLOCKSOURCE: |
7837699f | 1940 | case KVM_CAP_PIT: |
a28e4f5a | 1941 | case KVM_CAP_NOP_IO_DELAY: |
62d9f0db | 1942 | case KVM_CAP_MP_STATE: |
ed848624 | 1943 | case KVM_CAP_SYNC_MMU: |
52d939a0 | 1944 | case KVM_CAP_REINJECT_CONTROL: |
4925663a | 1945 | case KVM_CAP_IRQ_INJECT_STATUS: |
e56d532f | 1946 | case KVM_CAP_ASSIGN_DEV_IRQ: |
721eecbf | 1947 | case KVM_CAP_IRQFD: |
d34e6b17 | 1948 | case KVM_CAP_IOEVENTFD: |
c5ff41ce | 1949 | case KVM_CAP_PIT2: |
e9f42757 | 1950 | case KVM_CAP_PIT_STATE2: |
b927a3ce | 1951 | case KVM_CAP_SET_IDENTITY_MAP_ADDR: |
ffde22ac | 1952 | case KVM_CAP_XEN_HVM: |
afbcf7ab | 1953 | case KVM_CAP_ADJUST_CLOCK: |
3cfc3092 | 1954 | case KVM_CAP_VCPU_EVENTS: |
55cd8e5a | 1955 | case KVM_CAP_HYPERV: |
10388a07 | 1956 | case KVM_CAP_HYPERV_VAPIC: |
c25bc163 | 1957 | case KVM_CAP_HYPERV_SPIN: |
ab9f4ecb | 1958 | case KVM_CAP_PCI_SEGMENT: |
a1efbe77 | 1959 | case KVM_CAP_DEBUGREGS: |
d2be1651 | 1960 | case KVM_CAP_X86_ROBUST_SINGLESTEP: |
2d5b5a66 | 1961 | case KVM_CAP_XSAVE: |
344d9588 | 1962 | case KVM_CAP_ASYNC_PF: |
018d00d2 ZX |
1963 | r = 1; |
1964 | break; | |
542472b5 LV |
1965 | case KVM_CAP_COALESCED_MMIO: |
1966 | r = KVM_COALESCED_MMIO_PAGE_OFFSET; | |
1967 | break; | |
774ead3a AK |
1968 | case KVM_CAP_VAPIC: |
1969 | r = !kvm_x86_ops->cpu_has_accelerated_tpr(); | |
1970 | break; | |
f725230a AK |
1971 | case KVM_CAP_NR_VCPUS: |
1972 | r = KVM_MAX_VCPUS; | |
1973 | break; | |
a988b910 AK |
1974 | case KVM_CAP_NR_MEMSLOTS: |
1975 | r = KVM_MEMORY_SLOTS; | |
1976 | break; | |
a68a6a72 MT |
1977 | case KVM_CAP_PV_MMU: /* obsolete */ |
1978 | r = 0; | |
2f333bcb | 1979 | break; |
62c476c7 | 1980 | case KVM_CAP_IOMMU: |
19de40a8 | 1981 | r = iommu_found(); |
62c476c7 | 1982 | break; |
890ca9ae HY |
1983 | case KVM_CAP_MCE: |
1984 | r = KVM_MAX_MCE_BANKS; | |
1985 | break; | |
2d5b5a66 SY |
1986 | case KVM_CAP_XCRS: |
1987 | r = cpu_has_xsave; | |
1988 | break; | |
018d00d2 ZX |
1989 | default: |
1990 | r = 0; | |
1991 | break; | |
1992 | } | |
1993 | return r; | |
1994 | ||
1995 | } | |
1996 | ||
043405e1 CO |
1997 | long kvm_arch_dev_ioctl(struct file *filp, |
1998 | unsigned int ioctl, unsigned long arg) | |
1999 | { | |
2000 | void __user *argp = (void __user *)arg; | |
2001 | long r; | |
2002 | ||
2003 | switch (ioctl) { | |
2004 | case KVM_GET_MSR_INDEX_LIST: { | |
2005 | struct kvm_msr_list __user *user_msr_list = argp; | |
2006 | struct kvm_msr_list msr_list; | |
2007 | unsigned n; | |
2008 | ||
2009 | r = -EFAULT; | |
2010 | if (copy_from_user(&msr_list, user_msr_list, sizeof msr_list)) | |
2011 | goto out; | |
2012 | n = msr_list.nmsrs; | |
2013 | msr_list.nmsrs = num_msrs_to_save + ARRAY_SIZE(emulated_msrs); | |
2014 | if (copy_to_user(user_msr_list, &msr_list, sizeof msr_list)) | |
2015 | goto out; | |
2016 | r = -E2BIG; | |
e125e7b6 | 2017 | if (n < msr_list.nmsrs) |
043405e1 CO |
2018 | goto out; |
2019 | r = -EFAULT; | |
2020 | if (copy_to_user(user_msr_list->indices, &msrs_to_save, | |
2021 | num_msrs_to_save * sizeof(u32))) | |
2022 | goto out; | |
e125e7b6 | 2023 | if (copy_to_user(user_msr_list->indices + num_msrs_to_save, |
043405e1 CO |
2024 | &emulated_msrs, |
2025 | ARRAY_SIZE(emulated_msrs) * sizeof(u32))) | |
2026 | goto out; | |
2027 | r = 0; | |
2028 | break; | |
2029 | } | |
674eea0f AK |
2030 | case KVM_GET_SUPPORTED_CPUID: { |
2031 | struct kvm_cpuid2 __user *cpuid_arg = argp; | |
2032 | struct kvm_cpuid2 cpuid; | |
2033 | ||
2034 | r = -EFAULT; | |
2035 | if (copy_from_user(&cpuid, cpuid_arg, sizeof cpuid)) | |
2036 | goto out; | |
2037 | r = kvm_dev_ioctl_get_supported_cpuid(&cpuid, | |
19355475 | 2038 | cpuid_arg->entries); |
674eea0f AK |
2039 | if (r) |
2040 | goto out; | |
2041 | ||
2042 | r = -EFAULT; | |
2043 | if (copy_to_user(cpuid_arg, &cpuid, sizeof cpuid)) | |
2044 | goto out; | |
2045 | r = 0; | |
2046 | break; | |
2047 | } | |
890ca9ae HY |
2048 | case KVM_X86_GET_MCE_CAP_SUPPORTED: { |
2049 | u64 mce_cap; | |
2050 | ||
2051 | mce_cap = KVM_MCE_CAP_SUPPORTED; | |
2052 | r = -EFAULT; | |
2053 | if (copy_to_user(argp, &mce_cap, sizeof mce_cap)) | |
2054 | goto out; | |
2055 | r = 0; | |
2056 | break; | |
2057 | } | |
043405e1 CO |
2058 | default: |
2059 | r = -EINVAL; | |
2060 | } | |
2061 | out: | |
2062 | return r; | |
2063 | } | |
2064 | ||
f5f48ee1 SY |
2065 | static void wbinvd_ipi(void *garbage) |
2066 | { | |
2067 | wbinvd(); | |
2068 | } | |
2069 | ||
2070 | static bool need_emulate_wbinvd(struct kvm_vcpu *vcpu) | |
2071 | { | |
2072 | return vcpu->kvm->arch.iommu_domain && | |
2073 | !(vcpu->kvm->arch.iommu_flags & KVM_IOMMU_CACHE_COHERENCY); | |
2074 | } | |
2075 | ||
313a3dc7 CO |
2076 | void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu) |
2077 | { | |
f5f48ee1 SY |
2078 | /* Address WBINVD may be executed by guest */ |
2079 | if (need_emulate_wbinvd(vcpu)) { | |
2080 | if (kvm_x86_ops->has_wbinvd_exit()) | |
2081 | cpumask_set_cpu(cpu, vcpu->arch.wbinvd_dirty_mask); | |
2082 | else if (vcpu->cpu != -1 && vcpu->cpu != cpu) | |
2083 | smp_call_function_single(vcpu->cpu, | |
2084 | wbinvd_ipi, NULL, 1); | |
2085 | } | |
2086 | ||
313a3dc7 | 2087 | kvm_x86_ops->vcpu_load(vcpu, cpu); |
48434c20 | 2088 | if (unlikely(vcpu->cpu != cpu) || check_tsc_unstable()) { |
e48672fa ZA |
2089 | /* Make sure TSC doesn't go backwards */ |
2090 | s64 tsc_delta = !vcpu->arch.last_host_tsc ? 0 : | |
2091 | native_read_tsc() - vcpu->arch.last_host_tsc; | |
2092 | if (tsc_delta < 0) | |
2093 | mark_tsc_unstable("KVM discovered backwards TSC"); | |
c285545f | 2094 | if (check_tsc_unstable()) { |
e48672fa | 2095 | kvm_x86_ops->adjust_tsc_offset(vcpu, -tsc_delta); |
c285545f ZA |
2096 | vcpu->arch.tsc_catchup = 1; |
2097 | kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu); | |
2098 | } | |
2099 | if (vcpu->cpu != cpu) | |
2100 | kvm_migrate_timers(vcpu); | |
e48672fa | 2101 | vcpu->cpu = cpu; |
6b7d7e76 | 2102 | } |
313a3dc7 CO |
2103 | } |
2104 | ||
2105 | void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu) | |
2106 | { | |
02daab21 | 2107 | kvm_x86_ops->vcpu_put(vcpu); |
1c11e713 | 2108 | kvm_put_guest_fpu(vcpu); |
e48672fa | 2109 | vcpu->arch.last_host_tsc = native_read_tsc(); |
313a3dc7 CO |
2110 | } |
2111 | ||
07716717 | 2112 | static int is_efer_nx(void) |
313a3dc7 | 2113 | { |
e286e86e | 2114 | unsigned long long efer = 0; |
313a3dc7 | 2115 | |
e286e86e | 2116 | rdmsrl_safe(MSR_EFER, &efer); |
07716717 DK |
2117 | return efer & EFER_NX; |
2118 | } | |
2119 | ||
2120 | static void cpuid_fix_nx_cap(struct kvm_vcpu *vcpu) | |
2121 | { | |
2122 | int i; | |
2123 | struct kvm_cpuid_entry2 *e, *entry; | |
2124 | ||
313a3dc7 | 2125 | entry = NULL; |
ad312c7c ZX |
2126 | for (i = 0; i < vcpu->arch.cpuid_nent; ++i) { |
2127 | e = &vcpu->arch.cpuid_entries[i]; | |
313a3dc7 CO |
2128 | if (e->function == 0x80000001) { |
2129 | entry = e; | |
2130 | break; | |
2131 | } | |
2132 | } | |
07716717 | 2133 | if (entry && (entry->edx & (1 << 20)) && !is_efer_nx()) { |
313a3dc7 CO |
2134 | entry->edx &= ~(1 << 20); |
2135 | printk(KERN_INFO "kvm: guest NX capability removed\n"); | |
2136 | } | |
2137 | } | |
2138 | ||
07716717 | 2139 | /* when an old userspace process fills a new kernel module */ |
313a3dc7 CO |
2140 | static int kvm_vcpu_ioctl_set_cpuid(struct kvm_vcpu *vcpu, |
2141 | struct kvm_cpuid *cpuid, | |
2142 | struct kvm_cpuid_entry __user *entries) | |
07716717 DK |
2143 | { |
2144 | int r, i; | |
2145 | struct kvm_cpuid_entry *cpuid_entries; | |
2146 | ||
2147 | r = -E2BIG; | |
2148 | if (cpuid->nent > KVM_MAX_CPUID_ENTRIES) | |
2149 | goto out; | |
2150 | r = -ENOMEM; | |
2151 | cpuid_entries = vmalloc(sizeof(struct kvm_cpuid_entry) * cpuid->nent); | |
2152 | if (!cpuid_entries) | |
2153 | goto out; | |
2154 | r = -EFAULT; | |
2155 | if (copy_from_user(cpuid_entries, entries, | |
2156 | cpuid->nent * sizeof(struct kvm_cpuid_entry))) | |
2157 | goto out_free; | |
2158 | for (i = 0; i < cpuid->nent; i++) { | |
ad312c7c ZX |
2159 | vcpu->arch.cpuid_entries[i].function = cpuid_entries[i].function; |
2160 | vcpu->arch.cpuid_entries[i].eax = cpuid_entries[i].eax; | |
2161 | vcpu->arch.cpuid_entries[i].ebx = cpuid_entries[i].ebx; | |
2162 | vcpu->arch.cpuid_entries[i].ecx = cpuid_entries[i].ecx; | |
2163 | vcpu->arch.cpuid_entries[i].edx = cpuid_entries[i].edx; | |
2164 | vcpu->arch.cpuid_entries[i].index = 0; | |
2165 | vcpu->arch.cpuid_entries[i].flags = 0; | |
2166 | vcpu->arch.cpuid_entries[i].padding[0] = 0; | |
2167 | vcpu->arch.cpuid_entries[i].padding[1] = 0; | |
2168 | vcpu->arch.cpuid_entries[i].padding[2] = 0; | |
2169 | } | |
2170 | vcpu->arch.cpuid_nent = cpuid->nent; | |
07716717 DK |
2171 | cpuid_fix_nx_cap(vcpu); |
2172 | r = 0; | |
fc61b800 | 2173 | kvm_apic_set_version(vcpu); |
0e851880 | 2174 | kvm_x86_ops->cpuid_update(vcpu); |
2acf923e | 2175 | update_cpuid(vcpu); |
07716717 DK |
2176 | |
2177 | out_free: | |
2178 | vfree(cpuid_entries); | |
2179 | out: | |
2180 | return r; | |
2181 | } | |
2182 | ||
2183 | static int kvm_vcpu_ioctl_set_cpuid2(struct kvm_vcpu *vcpu, | |
19355475 AS |
2184 | struct kvm_cpuid2 *cpuid, |
2185 | struct kvm_cpuid_entry2 __user *entries) | |
313a3dc7 CO |
2186 | { |
2187 | int r; | |
2188 | ||
2189 | r = -E2BIG; | |
2190 | if (cpuid->nent > KVM_MAX_CPUID_ENTRIES) | |
2191 | goto out; | |
2192 | r = -EFAULT; | |
ad312c7c | 2193 | if (copy_from_user(&vcpu->arch.cpuid_entries, entries, |
07716717 | 2194 | cpuid->nent * sizeof(struct kvm_cpuid_entry2))) |
313a3dc7 | 2195 | goto out; |
ad312c7c | 2196 | vcpu->arch.cpuid_nent = cpuid->nent; |
fc61b800 | 2197 | kvm_apic_set_version(vcpu); |
0e851880 | 2198 | kvm_x86_ops->cpuid_update(vcpu); |
2acf923e | 2199 | update_cpuid(vcpu); |
313a3dc7 CO |
2200 | return 0; |
2201 | ||
2202 | out: | |
2203 | return r; | |
2204 | } | |
2205 | ||
07716717 | 2206 | static int kvm_vcpu_ioctl_get_cpuid2(struct kvm_vcpu *vcpu, |
19355475 AS |
2207 | struct kvm_cpuid2 *cpuid, |
2208 | struct kvm_cpuid_entry2 __user *entries) | |
07716717 DK |
2209 | { |
2210 | int r; | |
2211 | ||
2212 | r = -E2BIG; | |
ad312c7c | 2213 | if (cpuid->nent < vcpu->arch.cpuid_nent) |
07716717 DK |
2214 | goto out; |
2215 | r = -EFAULT; | |
ad312c7c | 2216 | if (copy_to_user(entries, &vcpu->arch.cpuid_entries, |
19355475 | 2217 | vcpu->arch.cpuid_nent * sizeof(struct kvm_cpuid_entry2))) |
07716717 DK |
2218 | goto out; |
2219 | return 0; | |
2220 | ||
2221 | out: | |
ad312c7c | 2222 | cpuid->nent = vcpu->arch.cpuid_nent; |
07716717 DK |
2223 | return r; |
2224 | } | |
2225 | ||
945ee35e AK |
2226 | static void cpuid_mask(u32 *word, int wordnum) |
2227 | { | |
2228 | *word &= boot_cpu_data.x86_capability[wordnum]; | |
2229 | } | |
2230 | ||
07716717 | 2231 | static void do_cpuid_1_ent(struct kvm_cpuid_entry2 *entry, u32 function, |
19355475 | 2232 | u32 index) |
07716717 DK |
2233 | { |
2234 | entry->function = function; | |
2235 | entry->index = index; | |
2236 | cpuid_count(entry->function, entry->index, | |
19355475 | 2237 | &entry->eax, &entry->ebx, &entry->ecx, &entry->edx); |
07716717 DK |
2238 | entry->flags = 0; |
2239 | } | |
2240 | ||
7faa4ee1 AK |
2241 | #define F(x) bit(X86_FEATURE_##x) |
2242 | ||
07716717 DK |
2243 | static void do_cpuid_ent(struct kvm_cpuid_entry2 *entry, u32 function, |
2244 | u32 index, int *nent, int maxnent) | |
2245 | { | |
7faa4ee1 | 2246 | unsigned f_nx = is_efer_nx() ? F(NX) : 0; |
07716717 | 2247 | #ifdef CONFIG_X86_64 |
17cc3935 SY |
2248 | unsigned f_gbpages = (kvm_x86_ops->get_lpage_level() == PT_PDPE_LEVEL) |
2249 | ? F(GBPAGES) : 0; | |
7faa4ee1 AK |
2250 | unsigned f_lm = F(LM); |
2251 | #else | |
17cc3935 | 2252 | unsigned f_gbpages = 0; |
7faa4ee1 | 2253 | unsigned f_lm = 0; |
07716717 | 2254 | #endif |
4e47c7a6 | 2255 | unsigned f_rdtscp = kvm_x86_ops->rdtscp_supported() ? F(RDTSCP) : 0; |
7faa4ee1 AK |
2256 | |
2257 | /* cpuid 1.edx */ | |
2258 | const u32 kvm_supported_word0_x86_features = | |
2259 | F(FPU) | F(VME) | F(DE) | F(PSE) | | |
2260 | F(TSC) | F(MSR) | F(PAE) | F(MCE) | | |
2261 | F(CX8) | F(APIC) | 0 /* Reserved */ | F(SEP) | | |
2262 | F(MTRR) | F(PGE) | F(MCA) | F(CMOV) | | |
2263 | F(PAT) | F(PSE36) | 0 /* PSN */ | F(CLFLSH) | | |
2264 | 0 /* Reserved, DS, ACPI */ | F(MMX) | | |
2265 | F(FXSR) | F(XMM) | F(XMM2) | F(SELFSNOOP) | | |
2266 | 0 /* HTT, TM, Reserved, PBE */; | |
2267 | /* cpuid 0x80000001.edx */ | |
2268 | const u32 kvm_supported_word1_x86_features = | |
2269 | F(FPU) | F(VME) | F(DE) | F(PSE) | | |
2270 | F(TSC) | F(MSR) | F(PAE) | F(MCE) | | |
2271 | F(CX8) | F(APIC) | 0 /* Reserved */ | F(SYSCALL) | | |
2272 | F(MTRR) | F(PGE) | F(MCA) | F(CMOV) | | |
2273 | F(PAT) | F(PSE36) | 0 /* Reserved */ | | |
2274 | f_nx | 0 /* Reserved */ | F(MMXEXT) | F(MMX) | | |
4e47c7a6 | 2275 | F(FXSR) | F(FXSR_OPT) | f_gbpages | f_rdtscp | |
7faa4ee1 AK |
2276 | 0 /* Reserved */ | f_lm | F(3DNOWEXT) | F(3DNOW); |
2277 | /* cpuid 1.ecx */ | |
2278 | const u32 kvm_supported_word4_x86_features = | |
6c3f6041 | 2279 | F(XMM3) | F(PCLMULQDQ) | 0 /* DTES64, MONITOR */ | |
d149c731 AK |
2280 | 0 /* DS-CPL, VMX, SMX, EST */ | |
2281 | 0 /* TM2 */ | F(SSSE3) | 0 /* CNXT-ID */ | 0 /* Reserved */ | | |
2282 | 0 /* Reserved */ | F(CX16) | 0 /* xTPR Update, PDCM */ | | |
2283 | 0 /* Reserved, DCA */ | F(XMM4_1) | | |
0105d1a5 | 2284 | F(XMM4_2) | F(X2APIC) | F(MOVBE) | F(POPCNT) | |
6d886fd0 AP |
2285 | 0 /* Reserved*/ | F(AES) | F(XSAVE) | 0 /* OSXSAVE */ | F(AVX) | |
2286 | F(F16C); | |
7faa4ee1 | 2287 | /* cpuid 0x80000001.ecx */ |
07716717 | 2288 | const u32 kvm_supported_word6_x86_features = |
4c62a2dc | 2289 | F(LAHF_LM) | F(CMP_LEGACY) | 0 /*SVM*/ | 0 /* ExtApicSpace */ | |
7faa4ee1 | 2290 | F(CR8_LEGACY) | F(ABM) | F(SSE4A) | F(MISALIGNSSE) | |
7ef8aa72 | 2291 | F(3DNOWPREFETCH) | 0 /* OSVW */ | 0 /* IBS */ | F(XOP) | |
6d886fd0 | 2292 | 0 /* SKINIT, WDT, LWP */ | F(FMA4) | F(TBM); |
07716717 | 2293 | |
19355475 | 2294 | /* all calls to cpuid_count() should be made on the same cpu */ |
07716717 DK |
2295 | get_cpu(); |
2296 | do_cpuid_1_ent(entry, function, index); | |
2297 | ++*nent; | |
2298 | ||
2299 | switch (function) { | |
2300 | case 0: | |
2acf923e | 2301 | entry->eax = min(entry->eax, (u32)0xd); |
07716717 DK |
2302 | break; |
2303 | case 1: | |
2304 | entry->edx &= kvm_supported_word0_x86_features; | |
945ee35e | 2305 | cpuid_mask(&entry->edx, 0); |
7faa4ee1 | 2306 | entry->ecx &= kvm_supported_word4_x86_features; |
945ee35e | 2307 | cpuid_mask(&entry->ecx, 4); |
0d1de2d9 GN |
2308 | /* we support x2apic emulation even if host does not support |
2309 | * it since we emulate x2apic in software */ | |
2310 | entry->ecx |= F(X2APIC); | |
07716717 DK |
2311 | break; |
2312 | /* function 2 entries are STATEFUL. That is, repeated cpuid commands | |
2313 | * may return different values. This forces us to get_cpu() before | |
2314 | * issuing the first command, and also to emulate this annoying behavior | |
2315 | * in kvm_emulate_cpuid() using KVM_CPUID_FLAG_STATE_READ_NEXT */ | |
2316 | case 2: { | |
2317 | int t, times = entry->eax & 0xff; | |
2318 | ||
2319 | entry->flags |= KVM_CPUID_FLAG_STATEFUL_FUNC; | |
0fdf8e59 | 2320 | entry->flags |= KVM_CPUID_FLAG_STATE_READ_NEXT; |
07716717 DK |
2321 | for (t = 1; t < times && *nent < maxnent; ++t) { |
2322 | do_cpuid_1_ent(&entry[t], function, 0); | |
2323 | entry[t].flags |= KVM_CPUID_FLAG_STATEFUL_FUNC; | |
2324 | ++*nent; | |
2325 | } | |
2326 | break; | |
2327 | } | |
2328 | /* function 4 and 0xb have additional index. */ | |
2329 | case 4: { | |
14af3f3c | 2330 | int i, cache_type; |
07716717 DK |
2331 | |
2332 | entry->flags |= KVM_CPUID_FLAG_SIGNIFCANT_INDEX; | |
2333 | /* read more entries until cache_type is zero */ | |
14af3f3c HH |
2334 | for (i = 1; *nent < maxnent; ++i) { |
2335 | cache_type = entry[i - 1].eax & 0x1f; | |
07716717 DK |
2336 | if (!cache_type) |
2337 | break; | |
14af3f3c HH |
2338 | do_cpuid_1_ent(&entry[i], function, i); |
2339 | entry[i].flags |= | |
07716717 DK |
2340 | KVM_CPUID_FLAG_SIGNIFCANT_INDEX; |
2341 | ++*nent; | |
2342 | } | |
2343 | break; | |
2344 | } | |
2345 | case 0xb: { | |
14af3f3c | 2346 | int i, level_type; |
07716717 DK |
2347 | |
2348 | entry->flags |= KVM_CPUID_FLAG_SIGNIFCANT_INDEX; | |
2349 | /* read more entries until level_type is zero */ | |
14af3f3c | 2350 | for (i = 1; *nent < maxnent; ++i) { |
0853d2c1 | 2351 | level_type = entry[i - 1].ecx & 0xff00; |
07716717 DK |
2352 | if (!level_type) |
2353 | break; | |
14af3f3c HH |
2354 | do_cpuid_1_ent(&entry[i], function, i); |
2355 | entry[i].flags |= | |
07716717 DK |
2356 | KVM_CPUID_FLAG_SIGNIFCANT_INDEX; |
2357 | ++*nent; | |
2358 | } | |
2359 | break; | |
2360 | } | |
2acf923e DC |
2361 | case 0xd: { |
2362 | int i; | |
2363 | ||
2364 | entry->flags |= KVM_CPUID_FLAG_SIGNIFCANT_INDEX; | |
2365 | for (i = 1; *nent < maxnent; ++i) { | |
2366 | if (entry[i - 1].eax == 0 && i != 2) | |
2367 | break; | |
2368 | do_cpuid_1_ent(&entry[i], function, i); | |
2369 | entry[i].flags |= | |
2370 | KVM_CPUID_FLAG_SIGNIFCANT_INDEX; | |
2371 | ++*nent; | |
2372 | } | |
2373 | break; | |
2374 | } | |
84478c82 GC |
2375 | case KVM_CPUID_SIGNATURE: { |
2376 | char signature[12] = "KVMKVMKVM\0\0"; | |
2377 | u32 *sigptr = (u32 *)signature; | |
2378 | entry->eax = 0; | |
2379 | entry->ebx = sigptr[0]; | |
2380 | entry->ecx = sigptr[1]; | |
2381 | entry->edx = sigptr[2]; | |
2382 | break; | |
2383 | } | |
2384 | case KVM_CPUID_FEATURES: | |
2385 | entry->eax = (1 << KVM_FEATURE_CLOCKSOURCE) | | |
2386 | (1 << KVM_FEATURE_NOP_IO_DELAY) | | |
371bcf64 GC |
2387 | (1 << KVM_FEATURE_CLOCKSOURCE2) | |
2388 | (1 << KVM_FEATURE_CLOCKSOURCE_STABLE_BIT); | |
84478c82 GC |
2389 | entry->ebx = 0; |
2390 | entry->ecx = 0; | |
2391 | entry->edx = 0; | |
2392 | break; | |
07716717 DK |
2393 | case 0x80000000: |
2394 | entry->eax = min(entry->eax, 0x8000001a); | |
2395 | break; | |
2396 | case 0x80000001: | |
2397 | entry->edx &= kvm_supported_word1_x86_features; | |
945ee35e | 2398 | cpuid_mask(&entry->edx, 1); |
07716717 | 2399 | entry->ecx &= kvm_supported_word6_x86_features; |
945ee35e | 2400 | cpuid_mask(&entry->ecx, 6); |
07716717 DK |
2401 | break; |
2402 | } | |
d4330ef2 JR |
2403 | |
2404 | kvm_x86_ops->set_supported_cpuid(function, entry); | |
2405 | ||
07716717 DK |
2406 | put_cpu(); |
2407 | } | |
2408 | ||
7faa4ee1 AK |
2409 | #undef F |
2410 | ||
674eea0f | 2411 | static int kvm_dev_ioctl_get_supported_cpuid(struct kvm_cpuid2 *cpuid, |
19355475 | 2412 | struct kvm_cpuid_entry2 __user *entries) |
07716717 DK |
2413 | { |
2414 | struct kvm_cpuid_entry2 *cpuid_entries; | |
2415 | int limit, nent = 0, r = -E2BIG; | |
2416 | u32 func; | |
2417 | ||
2418 | if (cpuid->nent < 1) | |
2419 | goto out; | |
6a544355 AK |
2420 | if (cpuid->nent > KVM_MAX_CPUID_ENTRIES) |
2421 | cpuid->nent = KVM_MAX_CPUID_ENTRIES; | |
07716717 DK |
2422 | r = -ENOMEM; |
2423 | cpuid_entries = vmalloc(sizeof(struct kvm_cpuid_entry2) * cpuid->nent); | |
2424 | if (!cpuid_entries) | |
2425 | goto out; | |
2426 | ||
2427 | do_cpuid_ent(&cpuid_entries[0], 0, 0, &nent, cpuid->nent); | |
2428 | limit = cpuid_entries[0].eax; | |
2429 | for (func = 1; func <= limit && nent < cpuid->nent; ++func) | |
2430 | do_cpuid_ent(&cpuid_entries[nent], func, 0, | |
19355475 | 2431 | &nent, cpuid->nent); |
07716717 DK |
2432 | r = -E2BIG; |
2433 | if (nent >= cpuid->nent) | |
2434 | goto out_free; | |
2435 | ||
2436 | do_cpuid_ent(&cpuid_entries[nent], 0x80000000, 0, &nent, cpuid->nent); | |
2437 | limit = cpuid_entries[nent - 1].eax; | |
2438 | for (func = 0x80000001; func <= limit && nent < cpuid->nent; ++func) | |
2439 | do_cpuid_ent(&cpuid_entries[nent], func, 0, | |
19355475 | 2440 | &nent, cpuid->nent); |
84478c82 GC |
2441 | |
2442 | ||
2443 | ||
2444 | r = -E2BIG; | |
2445 | if (nent >= cpuid->nent) | |
2446 | goto out_free; | |
2447 | ||
2448 | do_cpuid_ent(&cpuid_entries[nent], KVM_CPUID_SIGNATURE, 0, &nent, | |
2449 | cpuid->nent); | |
2450 | ||
2451 | r = -E2BIG; | |
2452 | if (nent >= cpuid->nent) | |
2453 | goto out_free; | |
2454 | ||
2455 | do_cpuid_ent(&cpuid_entries[nent], KVM_CPUID_FEATURES, 0, &nent, | |
2456 | cpuid->nent); | |
2457 | ||
cb007648 MM |
2458 | r = -E2BIG; |
2459 | if (nent >= cpuid->nent) | |
2460 | goto out_free; | |
2461 | ||
07716717 DK |
2462 | r = -EFAULT; |
2463 | if (copy_to_user(entries, cpuid_entries, | |
19355475 | 2464 | nent * sizeof(struct kvm_cpuid_entry2))) |
07716717 DK |
2465 | goto out_free; |
2466 | cpuid->nent = nent; | |
2467 | r = 0; | |
2468 | ||
2469 | out_free: | |
2470 | vfree(cpuid_entries); | |
2471 | out: | |
2472 | return r; | |
2473 | } | |
2474 | ||
313a3dc7 CO |
2475 | static int kvm_vcpu_ioctl_get_lapic(struct kvm_vcpu *vcpu, |
2476 | struct kvm_lapic_state *s) | |
2477 | { | |
ad312c7c | 2478 | memcpy(s->regs, vcpu->arch.apic->regs, sizeof *s); |
313a3dc7 CO |
2479 | |
2480 | return 0; | |
2481 | } | |
2482 | ||
2483 | static int kvm_vcpu_ioctl_set_lapic(struct kvm_vcpu *vcpu, | |
2484 | struct kvm_lapic_state *s) | |
2485 | { | |
ad312c7c | 2486 | memcpy(vcpu->arch.apic->regs, s->regs, sizeof *s); |
313a3dc7 | 2487 | kvm_apic_post_state_restore(vcpu); |
cb142eb7 | 2488 | update_cr8_intercept(vcpu); |
313a3dc7 CO |
2489 | |
2490 | return 0; | |
2491 | } | |
2492 | ||
f77bc6a4 ZX |
2493 | static int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, |
2494 | struct kvm_interrupt *irq) | |
2495 | { | |
2496 | if (irq->irq < 0 || irq->irq >= 256) | |
2497 | return -EINVAL; | |
2498 | if (irqchip_in_kernel(vcpu->kvm)) | |
2499 | return -ENXIO; | |
f77bc6a4 | 2500 | |
66fd3f7f | 2501 | kvm_queue_interrupt(vcpu, irq->irq, false); |
3842d135 | 2502 | kvm_make_request(KVM_REQ_EVENT, vcpu); |
f77bc6a4 | 2503 | |
f77bc6a4 ZX |
2504 | return 0; |
2505 | } | |
2506 | ||
c4abb7c9 JK |
2507 | static int kvm_vcpu_ioctl_nmi(struct kvm_vcpu *vcpu) |
2508 | { | |
c4abb7c9 | 2509 | kvm_inject_nmi(vcpu); |
c4abb7c9 JK |
2510 | |
2511 | return 0; | |
2512 | } | |
2513 | ||
b209749f AK |
2514 | static int vcpu_ioctl_tpr_access_reporting(struct kvm_vcpu *vcpu, |
2515 | struct kvm_tpr_access_ctl *tac) | |
2516 | { | |
2517 | if (tac->flags) | |
2518 | return -EINVAL; | |
2519 | vcpu->arch.tpr_access_reporting = !!tac->enabled; | |
2520 | return 0; | |
2521 | } | |
2522 | ||
890ca9ae HY |
2523 | static int kvm_vcpu_ioctl_x86_setup_mce(struct kvm_vcpu *vcpu, |
2524 | u64 mcg_cap) | |
2525 | { | |
2526 | int r; | |
2527 | unsigned bank_num = mcg_cap & 0xff, bank; | |
2528 | ||
2529 | r = -EINVAL; | |
a9e38c3e | 2530 | if (!bank_num || bank_num >= KVM_MAX_MCE_BANKS) |
890ca9ae HY |
2531 | goto out; |
2532 | if (mcg_cap & ~(KVM_MCE_CAP_SUPPORTED | 0xff | 0xff0000)) | |
2533 | goto out; | |
2534 | r = 0; | |
2535 | vcpu->arch.mcg_cap = mcg_cap; | |
2536 | /* Init IA32_MCG_CTL to all 1s */ | |
2537 | if (mcg_cap & MCG_CTL_P) | |
2538 | vcpu->arch.mcg_ctl = ~(u64)0; | |
2539 | /* Init IA32_MCi_CTL to all 1s */ | |
2540 | for (bank = 0; bank < bank_num; bank++) | |
2541 | vcpu->arch.mce_banks[bank*4] = ~(u64)0; | |
2542 | out: | |
2543 | return r; | |
2544 | } | |
2545 | ||
2546 | static int kvm_vcpu_ioctl_x86_set_mce(struct kvm_vcpu *vcpu, | |
2547 | struct kvm_x86_mce *mce) | |
2548 | { | |
2549 | u64 mcg_cap = vcpu->arch.mcg_cap; | |
2550 | unsigned bank_num = mcg_cap & 0xff; | |
2551 | u64 *banks = vcpu->arch.mce_banks; | |
2552 | ||
2553 | if (mce->bank >= bank_num || !(mce->status & MCI_STATUS_VAL)) | |
2554 | return -EINVAL; | |
2555 | /* | |
2556 | * if IA32_MCG_CTL is not all 1s, the uncorrected error | |
2557 | * reporting is disabled | |
2558 | */ | |
2559 | if ((mce->status & MCI_STATUS_UC) && (mcg_cap & MCG_CTL_P) && | |
2560 | vcpu->arch.mcg_ctl != ~(u64)0) | |
2561 | return 0; | |
2562 | banks += 4 * mce->bank; | |
2563 | /* | |
2564 | * if IA32_MCi_CTL is not all 1s, the uncorrected error | |
2565 | * reporting is disabled for the bank | |
2566 | */ | |
2567 | if ((mce->status & MCI_STATUS_UC) && banks[0] != ~(u64)0) | |
2568 | return 0; | |
2569 | if (mce->status & MCI_STATUS_UC) { | |
2570 | if ((vcpu->arch.mcg_status & MCG_STATUS_MCIP) || | |
fc78f519 | 2571 | !kvm_read_cr4_bits(vcpu, X86_CR4_MCE)) { |
890ca9ae HY |
2572 | printk(KERN_DEBUG "kvm: set_mce: " |
2573 | "injects mce exception while " | |
2574 | "previous one is in progress!\n"); | |
a8eeb04a | 2575 | kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu); |
890ca9ae HY |
2576 | return 0; |
2577 | } | |
2578 | if (banks[1] & MCI_STATUS_VAL) | |
2579 | mce->status |= MCI_STATUS_OVER; | |
2580 | banks[2] = mce->addr; | |
2581 | banks[3] = mce->misc; | |
2582 | vcpu->arch.mcg_status = mce->mcg_status; | |
2583 | banks[1] = mce->status; | |
2584 | kvm_queue_exception(vcpu, MC_VECTOR); | |
2585 | } else if (!(banks[1] & MCI_STATUS_VAL) | |
2586 | || !(banks[1] & MCI_STATUS_UC)) { | |
2587 | if (banks[1] & MCI_STATUS_VAL) | |
2588 | mce->status |= MCI_STATUS_OVER; | |
2589 | banks[2] = mce->addr; | |
2590 | banks[3] = mce->misc; | |
2591 | banks[1] = mce->status; | |
2592 | } else | |
2593 | banks[1] |= MCI_STATUS_OVER; | |
2594 | return 0; | |
2595 | } | |
2596 | ||
3cfc3092 JK |
2597 | static void kvm_vcpu_ioctl_x86_get_vcpu_events(struct kvm_vcpu *vcpu, |
2598 | struct kvm_vcpu_events *events) | |
2599 | { | |
03b82a30 JK |
2600 | events->exception.injected = |
2601 | vcpu->arch.exception.pending && | |
2602 | !kvm_exception_is_soft(vcpu->arch.exception.nr); | |
3cfc3092 JK |
2603 | events->exception.nr = vcpu->arch.exception.nr; |
2604 | events->exception.has_error_code = vcpu->arch.exception.has_error_code; | |
97e69aa6 | 2605 | events->exception.pad = 0; |
3cfc3092 JK |
2606 | events->exception.error_code = vcpu->arch.exception.error_code; |
2607 | ||
03b82a30 JK |
2608 | events->interrupt.injected = |
2609 | vcpu->arch.interrupt.pending && !vcpu->arch.interrupt.soft; | |
3cfc3092 | 2610 | events->interrupt.nr = vcpu->arch.interrupt.nr; |
03b82a30 | 2611 | events->interrupt.soft = 0; |
48005f64 JK |
2612 | events->interrupt.shadow = |
2613 | kvm_x86_ops->get_interrupt_shadow(vcpu, | |
2614 | KVM_X86_SHADOW_INT_MOV_SS | KVM_X86_SHADOW_INT_STI); | |
3cfc3092 JK |
2615 | |
2616 | events->nmi.injected = vcpu->arch.nmi_injected; | |
2617 | events->nmi.pending = vcpu->arch.nmi_pending; | |
2618 | events->nmi.masked = kvm_x86_ops->get_nmi_mask(vcpu); | |
97e69aa6 | 2619 | events->nmi.pad = 0; |
3cfc3092 JK |
2620 | |
2621 | events->sipi_vector = vcpu->arch.sipi_vector; | |
2622 | ||
dab4b911 | 2623 | events->flags = (KVM_VCPUEVENT_VALID_NMI_PENDING |
48005f64 JK |
2624 | | KVM_VCPUEVENT_VALID_SIPI_VECTOR |
2625 | | KVM_VCPUEVENT_VALID_SHADOW); | |
97e69aa6 | 2626 | memset(&events->reserved, 0, sizeof(events->reserved)); |
3cfc3092 JK |
2627 | } |
2628 | ||
2629 | static int kvm_vcpu_ioctl_x86_set_vcpu_events(struct kvm_vcpu *vcpu, | |
2630 | struct kvm_vcpu_events *events) | |
2631 | { | |
dab4b911 | 2632 | if (events->flags & ~(KVM_VCPUEVENT_VALID_NMI_PENDING |
48005f64 JK |
2633 | | KVM_VCPUEVENT_VALID_SIPI_VECTOR |
2634 | | KVM_VCPUEVENT_VALID_SHADOW)) | |
3cfc3092 JK |
2635 | return -EINVAL; |
2636 | ||
3cfc3092 JK |
2637 | vcpu->arch.exception.pending = events->exception.injected; |
2638 | vcpu->arch.exception.nr = events->exception.nr; | |
2639 | vcpu->arch.exception.has_error_code = events->exception.has_error_code; | |
2640 | vcpu->arch.exception.error_code = events->exception.error_code; | |
2641 | ||
2642 | vcpu->arch.interrupt.pending = events->interrupt.injected; | |
2643 | vcpu->arch.interrupt.nr = events->interrupt.nr; | |
2644 | vcpu->arch.interrupt.soft = events->interrupt.soft; | |
2645 | if (vcpu->arch.interrupt.pending && irqchip_in_kernel(vcpu->kvm)) | |
2646 | kvm_pic_clear_isr_ack(vcpu->kvm); | |
48005f64 JK |
2647 | if (events->flags & KVM_VCPUEVENT_VALID_SHADOW) |
2648 | kvm_x86_ops->set_interrupt_shadow(vcpu, | |
2649 | events->interrupt.shadow); | |
3cfc3092 JK |
2650 | |
2651 | vcpu->arch.nmi_injected = events->nmi.injected; | |
dab4b911 JK |
2652 | if (events->flags & KVM_VCPUEVENT_VALID_NMI_PENDING) |
2653 | vcpu->arch.nmi_pending = events->nmi.pending; | |
3cfc3092 JK |
2654 | kvm_x86_ops->set_nmi_mask(vcpu, events->nmi.masked); |
2655 | ||
dab4b911 JK |
2656 | if (events->flags & KVM_VCPUEVENT_VALID_SIPI_VECTOR) |
2657 | vcpu->arch.sipi_vector = events->sipi_vector; | |
3cfc3092 | 2658 | |
3842d135 AK |
2659 | kvm_make_request(KVM_REQ_EVENT, vcpu); |
2660 | ||
3cfc3092 JK |
2661 | return 0; |
2662 | } | |
2663 | ||
a1efbe77 JK |
2664 | static void kvm_vcpu_ioctl_x86_get_debugregs(struct kvm_vcpu *vcpu, |
2665 | struct kvm_debugregs *dbgregs) | |
2666 | { | |
a1efbe77 JK |
2667 | memcpy(dbgregs->db, vcpu->arch.db, sizeof(vcpu->arch.db)); |
2668 | dbgregs->dr6 = vcpu->arch.dr6; | |
2669 | dbgregs->dr7 = vcpu->arch.dr7; | |
2670 | dbgregs->flags = 0; | |
97e69aa6 | 2671 | memset(&dbgregs->reserved, 0, sizeof(dbgregs->reserved)); |
a1efbe77 JK |
2672 | } |
2673 | ||
2674 | static int kvm_vcpu_ioctl_x86_set_debugregs(struct kvm_vcpu *vcpu, | |
2675 | struct kvm_debugregs *dbgregs) | |
2676 | { | |
2677 | if (dbgregs->flags) | |
2678 | return -EINVAL; | |
2679 | ||
a1efbe77 JK |
2680 | memcpy(vcpu->arch.db, dbgregs->db, sizeof(vcpu->arch.db)); |
2681 | vcpu->arch.dr6 = dbgregs->dr6; | |
2682 | vcpu->arch.dr7 = dbgregs->dr7; | |
2683 | ||
a1efbe77 JK |
2684 | return 0; |
2685 | } | |
2686 | ||
2d5b5a66 SY |
2687 | static void kvm_vcpu_ioctl_x86_get_xsave(struct kvm_vcpu *vcpu, |
2688 | struct kvm_xsave *guest_xsave) | |
2689 | { | |
2690 | if (cpu_has_xsave) | |
2691 | memcpy(guest_xsave->region, | |
2692 | &vcpu->arch.guest_fpu.state->xsave, | |
f45755b8 | 2693 | xstate_size); |
2d5b5a66 SY |
2694 | else { |
2695 | memcpy(guest_xsave->region, | |
2696 | &vcpu->arch.guest_fpu.state->fxsave, | |
2697 | sizeof(struct i387_fxsave_struct)); | |
2698 | *(u64 *)&guest_xsave->region[XSAVE_HDR_OFFSET / sizeof(u32)] = | |
2699 | XSTATE_FPSSE; | |
2700 | } | |
2701 | } | |
2702 | ||
2703 | static int kvm_vcpu_ioctl_x86_set_xsave(struct kvm_vcpu *vcpu, | |
2704 | struct kvm_xsave *guest_xsave) | |
2705 | { | |
2706 | u64 xstate_bv = | |
2707 | *(u64 *)&guest_xsave->region[XSAVE_HDR_OFFSET / sizeof(u32)]; | |
2708 | ||
2709 | if (cpu_has_xsave) | |
2710 | memcpy(&vcpu->arch.guest_fpu.state->xsave, | |
f45755b8 | 2711 | guest_xsave->region, xstate_size); |
2d5b5a66 SY |
2712 | else { |
2713 | if (xstate_bv & ~XSTATE_FPSSE) | |
2714 | return -EINVAL; | |
2715 | memcpy(&vcpu->arch.guest_fpu.state->fxsave, | |
2716 | guest_xsave->region, sizeof(struct i387_fxsave_struct)); | |
2717 | } | |
2718 | return 0; | |
2719 | } | |
2720 | ||
2721 | static void kvm_vcpu_ioctl_x86_get_xcrs(struct kvm_vcpu *vcpu, | |
2722 | struct kvm_xcrs *guest_xcrs) | |
2723 | { | |
2724 | if (!cpu_has_xsave) { | |
2725 | guest_xcrs->nr_xcrs = 0; | |
2726 | return; | |
2727 | } | |
2728 | ||
2729 | guest_xcrs->nr_xcrs = 1; | |
2730 | guest_xcrs->flags = 0; | |
2731 | guest_xcrs->xcrs[0].xcr = XCR_XFEATURE_ENABLED_MASK; | |
2732 | guest_xcrs->xcrs[0].value = vcpu->arch.xcr0; | |
2733 | } | |
2734 | ||
2735 | static int kvm_vcpu_ioctl_x86_set_xcrs(struct kvm_vcpu *vcpu, | |
2736 | struct kvm_xcrs *guest_xcrs) | |
2737 | { | |
2738 | int i, r = 0; | |
2739 | ||
2740 | if (!cpu_has_xsave) | |
2741 | return -EINVAL; | |
2742 | ||
2743 | if (guest_xcrs->nr_xcrs > KVM_MAX_XCRS || guest_xcrs->flags) | |
2744 | return -EINVAL; | |
2745 | ||
2746 | for (i = 0; i < guest_xcrs->nr_xcrs; i++) | |
2747 | /* Only support XCR0 currently */ | |
2748 | if (guest_xcrs->xcrs[0].xcr == XCR_XFEATURE_ENABLED_MASK) { | |
2749 | r = __kvm_set_xcr(vcpu, XCR_XFEATURE_ENABLED_MASK, | |
2750 | guest_xcrs->xcrs[0].value); | |
2751 | break; | |
2752 | } | |
2753 | if (r) | |
2754 | r = -EINVAL; | |
2755 | return r; | |
2756 | } | |
2757 | ||
313a3dc7 CO |
2758 | long kvm_arch_vcpu_ioctl(struct file *filp, |
2759 | unsigned int ioctl, unsigned long arg) | |
2760 | { | |
2761 | struct kvm_vcpu *vcpu = filp->private_data; | |
2762 | void __user *argp = (void __user *)arg; | |
2763 | int r; | |
d1ac91d8 AK |
2764 | union { |
2765 | struct kvm_lapic_state *lapic; | |
2766 | struct kvm_xsave *xsave; | |
2767 | struct kvm_xcrs *xcrs; | |
2768 | void *buffer; | |
2769 | } u; | |
2770 | ||
2771 | u.buffer = NULL; | |
313a3dc7 CO |
2772 | switch (ioctl) { |
2773 | case KVM_GET_LAPIC: { | |
2204ae3c MT |
2774 | r = -EINVAL; |
2775 | if (!vcpu->arch.apic) | |
2776 | goto out; | |
d1ac91d8 | 2777 | u.lapic = kzalloc(sizeof(struct kvm_lapic_state), GFP_KERNEL); |
313a3dc7 | 2778 | |
b772ff36 | 2779 | r = -ENOMEM; |
d1ac91d8 | 2780 | if (!u.lapic) |
b772ff36 | 2781 | goto out; |
d1ac91d8 | 2782 | r = kvm_vcpu_ioctl_get_lapic(vcpu, u.lapic); |
313a3dc7 CO |
2783 | if (r) |
2784 | goto out; | |
2785 | r = -EFAULT; | |
d1ac91d8 | 2786 | if (copy_to_user(argp, u.lapic, sizeof(struct kvm_lapic_state))) |
313a3dc7 CO |
2787 | goto out; |
2788 | r = 0; | |
2789 | break; | |
2790 | } | |
2791 | case KVM_SET_LAPIC: { | |
2204ae3c MT |
2792 | r = -EINVAL; |
2793 | if (!vcpu->arch.apic) | |
2794 | goto out; | |
d1ac91d8 | 2795 | u.lapic = kmalloc(sizeof(struct kvm_lapic_state), GFP_KERNEL); |
b772ff36 | 2796 | r = -ENOMEM; |
d1ac91d8 | 2797 | if (!u.lapic) |
b772ff36 | 2798 | goto out; |
313a3dc7 | 2799 | r = -EFAULT; |
d1ac91d8 | 2800 | if (copy_from_user(u.lapic, argp, sizeof(struct kvm_lapic_state))) |
313a3dc7 | 2801 | goto out; |
d1ac91d8 | 2802 | r = kvm_vcpu_ioctl_set_lapic(vcpu, u.lapic); |
313a3dc7 CO |
2803 | if (r) |
2804 | goto out; | |
2805 | r = 0; | |
2806 | break; | |
2807 | } | |
f77bc6a4 ZX |
2808 | case KVM_INTERRUPT: { |
2809 | struct kvm_interrupt irq; | |
2810 | ||
2811 | r = -EFAULT; | |
2812 | if (copy_from_user(&irq, argp, sizeof irq)) | |
2813 | goto out; | |
2814 | r = kvm_vcpu_ioctl_interrupt(vcpu, &irq); | |
2815 | if (r) | |
2816 | goto out; | |
2817 | r = 0; | |
2818 | break; | |
2819 | } | |
c4abb7c9 JK |
2820 | case KVM_NMI: { |
2821 | r = kvm_vcpu_ioctl_nmi(vcpu); | |
2822 | if (r) | |
2823 | goto out; | |
2824 | r = 0; | |
2825 | break; | |
2826 | } | |
313a3dc7 CO |
2827 | case KVM_SET_CPUID: { |
2828 | struct kvm_cpuid __user *cpuid_arg = argp; | |
2829 | struct kvm_cpuid cpuid; | |
2830 | ||
2831 | r = -EFAULT; | |
2832 | if (copy_from_user(&cpuid, cpuid_arg, sizeof cpuid)) | |
2833 | goto out; | |
2834 | r = kvm_vcpu_ioctl_set_cpuid(vcpu, &cpuid, cpuid_arg->entries); | |
2835 | if (r) | |
2836 | goto out; | |
2837 | break; | |
2838 | } | |
07716717 DK |
2839 | case KVM_SET_CPUID2: { |
2840 | struct kvm_cpuid2 __user *cpuid_arg = argp; | |
2841 | struct kvm_cpuid2 cpuid; | |
2842 | ||
2843 | r = -EFAULT; | |
2844 | if (copy_from_user(&cpuid, cpuid_arg, sizeof cpuid)) | |
2845 | goto out; | |
2846 | r = kvm_vcpu_ioctl_set_cpuid2(vcpu, &cpuid, | |
19355475 | 2847 | cpuid_arg->entries); |
07716717 DK |
2848 | if (r) |
2849 | goto out; | |
2850 | break; | |
2851 | } | |
2852 | case KVM_GET_CPUID2: { | |
2853 | struct kvm_cpuid2 __user *cpuid_arg = argp; | |
2854 | struct kvm_cpuid2 cpuid; | |
2855 | ||
2856 | r = -EFAULT; | |
2857 | if (copy_from_user(&cpuid, cpuid_arg, sizeof cpuid)) | |
2858 | goto out; | |
2859 | r = kvm_vcpu_ioctl_get_cpuid2(vcpu, &cpuid, | |
19355475 | 2860 | cpuid_arg->entries); |
07716717 DK |
2861 | if (r) |
2862 | goto out; | |
2863 | r = -EFAULT; | |
2864 | if (copy_to_user(cpuid_arg, &cpuid, sizeof cpuid)) | |
2865 | goto out; | |
2866 | r = 0; | |
2867 | break; | |
2868 | } | |
313a3dc7 CO |
2869 | case KVM_GET_MSRS: |
2870 | r = msr_io(vcpu, argp, kvm_get_msr, 1); | |
2871 | break; | |
2872 | case KVM_SET_MSRS: | |
2873 | r = msr_io(vcpu, argp, do_set_msr, 0); | |
2874 | break; | |
b209749f AK |
2875 | case KVM_TPR_ACCESS_REPORTING: { |
2876 | struct kvm_tpr_access_ctl tac; | |
2877 | ||
2878 | r = -EFAULT; | |
2879 | if (copy_from_user(&tac, argp, sizeof tac)) | |
2880 | goto out; | |
2881 | r = vcpu_ioctl_tpr_access_reporting(vcpu, &tac); | |
2882 | if (r) | |
2883 | goto out; | |
2884 | r = -EFAULT; | |
2885 | if (copy_to_user(argp, &tac, sizeof tac)) | |
2886 | goto out; | |
2887 | r = 0; | |
2888 | break; | |
2889 | }; | |
b93463aa AK |
2890 | case KVM_SET_VAPIC_ADDR: { |
2891 | struct kvm_vapic_addr va; | |
2892 | ||
2893 | r = -EINVAL; | |
2894 | if (!irqchip_in_kernel(vcpu->kvm)) | |
2895 | goto out; | |
2896 | r = -EFAULT; | |
2897 | if (copy_from_user(&va, argp, sizeof va)) | |
2898 | goto out; | |
2899 | r = 0; | |
2900 | kvm_lapic_set_vapic_addr(vcpu, va.vapic_addr); | |
2901 | break; | |
2902 | } | |
890ca9ae HY |
2903 | case KVM_X86_SETUP_MCE: { |
2904 | u64 mcg_cap; | |
2905 | ||
2906 | r = -EFAULT; | |
2907 | if (copy_from_user(&mcg_cap, argp, sizeof mcg_cap)) | |
2908 | goto out; | |
2909 | r = kvm_vcpu_ioctl_x86_setup_mce(vcpu, mcg_cap); | |
2910 | break; | |
2911 | } | |
2912 | case KVM_X86_SET_MCE: { | |
2913 | struct kvm_x86_mce mce; | |
2914 | ||
2915 | r = -EFAULT; | |
2916 | if (copy_from_user(&mce, argp, sizeof mce)) | |
2917 | goto out; | |
2918 | r = kvm_vcpu_ioctl_x86_set_mce(vcpu, &mce); | |
2919 | break; | |
2920 | } | |
3cfc3092 JK |
2921 | case KVM_GET_VCPU_EVENTS: { |
2922 | struct kvm_vcpu_events events; | |
2923 | ||
2924 | kvm_vcpu_ioctl_x86_get_vcpu_events(vcpu, &events); | |
2925 | ||
2926 | r = -EFAULT; | |
2927 | if (copy_to_user(argp, &events, sizeof(struct kvm_vcpu_events))) | |
2928 | break; | |
2929 | r = 0; | |
2930 | break; | |
2931 | } | |
2932 | case KVM_SET_VCPU_EVENTS: { | |
2933 | struct kvm_vcpu_events events; | |
2934 | ||
2935 | r = -EFAULT; | |
2936 | if (copy_from_user(&events, argp, sizeof(struct kvm_vcpu_events))) | |
2937 | break; | |
2938 | ||
2939 | r = kvm_vcpu_ioctl_x86_set_vcpu_events(vcpu, &events); | |
2940 | break; | |
2941 | } | |
a1efbe77 JK |
2942 | case KVM_GET_DEBUGREGS: { |
2943 | struct kvm_debugregs dbgregs; | |
2944 | ||
2945 | kvm_vcpu_ioctl_x86_get_debugregs(vcpu, &dbgregs); | |
2946 | ||
2947 | r = -EFAULT; | |
2948 | if (copy_to_user(argp, &dbgregs, | |
2949 | sizeof(struct kvm_debugregs))) | |
2950 | break; | |
2951 | r = 0; | |
2952 | break; | |
2953 | } | |
2954 | case KVM_SET_DEBUGREGS: { | |
2955 | struct kvm_debugregs dbgregs; | |
2956 | ||
2957 | r = -EFAULT; | |
2958 | if (copy_from_user(&dbgregs, argp, | |
2959 | sizeof(struct kvm_debugregs))) | |
2960 | break; | |
2961 | ||
2962 | r = kvm_vcpu_ioctl_x86_set_debugregs(vcpu, &dbgregs); | |
2963 | break; | |
2964 | } | |
2d5b5a66 | 2965 | case KVM_GET_XSAVE: { |
d1ac91d8 | 2966 | u.xsave = kzalloc(sizeof(struct kvm_xsave), GFP_KERNEL); |
2d5b5a66 | 2967 | r = -ENOMEM; |
d1ac91d8 | 2968 | if (!u.xsave) |
2d5b5a66 SY |
2969 | break; |
2970 | ||
d1ac91d8 | 2971 | kvm_vcpu_ioctl_x86_get_xsave(vcpu, u.xsave); |
2d5b5a66 SY |
2972 | |
2973 | r = -EFAULT; | |
d1ac91d8 | 2974 | if (copy_to_user(argp, u.xsave, sizeof(struct kvm_xsave))) |
2d5b5a66 SY |
2975 | break; |
2976 | r = 0; | |
2977 | break; | |
2978 | } | |
2979 | case KVM_SET_XSAVE: { | |
d1ac91d8 | 2980 | u.xsave = kzalloc(sizeof(struct kvm_xsave), GFP_KERNEL); |
2d5b5a66 | 2981 | r = -ENOMEM; |
d1ac91d8 | 2982 | if (!u.xsave) |
2d5b5a66 SY |
2983 | break; |
2984 | ||
2985 | r = -EFAULT; | |
d1ac91d8 | 2986 | if (copy_from_user(u.xsave, argp, sizeof(struct kvm_xsave))) |
2d5b5a66 SY |
2987 | break; |
2988 | ||
d1ac91d8 | 2989 | r = kvm_vcpu_ioctl_x86_set_xsave(vcpu, u.xsave); |
2d5b5a66 SY |
2990 | break; |
2991 | } | |
2992 | case KVM_GET_XCRS: { | |
d1ac91d8 | 2993 | u.xcrs = kzalloc(sizeof(struct kvm_xcrs), GFP_KERNEL); |
2d5b5a66 | 2994 | r = -ENOMEM; |
d1ac91d8 | 2995 | if (!u.xcrs) |
2d5b5a66 SY |
2996 | break; |
2997 | ||
d1ac91d8 | 2998 | kvm_vcpu_ioctl_x86_get_xcrs(vcpu, u.xcrs); |
2d5b5a66 SY |
2999 | |
3000 | r = -EFAULT; | |
d1ac91d8 | 3001 | if (copy_to_user(argp, u.xcrs, |
2d5b5a66 SY |
3002 | sizeof(struct kvm_xcrs))) |
3003 | break; | |
3004 | r = 0; | |
3005 | break; | |
3006 | } | |
3007 | case KVM_SET_XCRS: { | |
d1ac91d8 | 3008 | u.xcrs = kzalloc(sizeof(struct kvm_xcrs), GFP_KERNEL); |
2d5b5a66 | 3009 | r = -ENOMEM; |
d1ac91d8 | 3010 | if (!u.xcrs) |
2d5b5a66 SY |
3011 | break; |
3012 | ||
3013 | r = -EFAULT; | |
d1ac91d8 | 3014 | if (copy_from_user(u.xcrs, argp, |
2d5b5a66 SY |
3015 | sizeof(struct kvm_xcrs))) |
3016 | break; | |
3017 | ||
d1ac91d8 | 3018 | r = kvm_vcpu_ioctl_x86_set_xcrs(vcpu, u.xcrs); |
2d5b5a66 SY |
3019 | break; |
3020 | } | |
313a3dc7 CO |
3021 | default: |
3022 | r = -EINVAL; | |
3023 | } | |
3024 | out: | |
d1ac91d8 | 3025 | kfree(u.buffer); |
313a3dc7 CO |
3026 | return r; |
3027 | } | |
3028 | ||
1fe779f8 CO |
3029 | static int kvm_vm_ioctl_set_tss_addr(struct kvm *kvm, unsigned long addr) |
3030 | { | |
3031 | int ret; | |
3032 | ||
3033 | if (addr > (unsigned int)(-3 * PAGE_SIZE)) | |
3034 | return -1; | |
3035 | ret = kvm_x86_ops->set_tss_addr(kvm, addr); | |
3036 | return ret; | |
3037 | } | |
3038 | ||
b927a3ce SY |
3039 | static int kvm_vm_ioctl_set_identity_map_addr(struct kvm *kvm, |
3040 | u64 ident_addr) | |
3041 | { | |
3042 | kvm->arch.ept_identity_map_addr = ident_addr; | |
3043 | return 0; | |
3044 | } | |
3045 | ||
1fe779f8 CO |
3046 | static int kvm_vm_ioctl_set_nr_mmu_pages(struct kvm *kvm, |
3047 | u32 kvm_nr_mmu_pages) | |
3048 | { | |
3049 | if (kvm_nr_mmu_pages < KVM_MIN_ALLOC_MMU_PAGES) | |
3050 | return -EINVAL; | |
3051 | ||
79fac95e | 3052 | mutex_lock(&kvm->slots_lock); |
7c8a83b7 | 3053 | spin_lock(&kvm->mmu_lock); |
1fe779f8 CO |
3054 | |
3055 | kvm_mmu_change_mmu_pages(kvm, kvm_nr_mmu_pages); | |
f05e70ac | 3056 | kvm->arch.n_requested_mmu_pages = kvm_nr_mmu_pages; |
1fe779f8 | 3057 | |
7c8a83b7 | 3058 | spin_unlock(&kvm->mmu_lock); |
79fac95e | 3059 | mutex_unlock(&kvm->slots_lock); |
1fe779f8 CO |
3060 | return 0; |
3061 | } | |
3062 | ||
3063 | static int kvm_vm_ioctl_get_nr_mmu_pages(struct kvm *kvm) | |
3064 | { | |
39de71ec | 3065 | return kvm->arch.n_max_mmu_pages; |
1fe779f8 CO |
3066 | } |
3067 | ||
1fe779f8 CO |
3068 | static int kvm_vm_ioctl_get_irqchip(struct kvm *kvm, struct kvm_irqchip *chip) |
3069 | { | |
3070 | int r; | |
3071 | ||
3072 | r = 0; | |
3073 | switch (chip->chip_id) { | |
3074 | case KVM_IRQCHIP_PIC_MASTER: | |
3075 | memcpy(&chip->chip.pic, | |
3076 | &pic_irqchip(kvm)->pics[0], | |
3077 | sizeof(struct kvm_pic_state)); | |
3078 | break; | |
3079 | case KVM_IRQCHIP_PIC_SLAVE: | |
3080 | memcpy(&chip->chip.pic, | |
3081 | &pic_irqchip(kvm)->pics[1], | |
3082 | sizeof(struct kvm_pic_state)); | |
3083 | break; | |
3084 | case KVM_IRQCHIP_IOAPIC: | |
eba0226b | 3085 | r = kvm_get_ioapic(kvm, &chip->chip.ioapic); |
1fe779f8 CO |
3086 | break; |
3087 | default: | |
3088 | r = -EINVAL; | |
3089 | break; | |
3090 | } | |
3091 | return r; | |
3092 | } | |
3093 | ||
3094 | static int kvm_vm_ioctl_set_irqchip(struct kvm *kvm, struct kvm_irqchip *chip) | |
3095 | { | |
3096 | int r; | |
3097 | ||
3098 | r = 0; | |
3099 | switch (chip->chip_id) { | |
3100 | case KVM_IRQCHIP_PIC_MASTER: | |
f4f51050 | 3101 | spin_lock(&pic_irqchip(kvm)->lock); |
1fe779f8 CO |
3102 | memcpy(&pic_irqchip(kvm)->pics[0], |
3103 | &chip->chip.pic, | |
3104 | sizeof(struct kvm_pic_state)); | |
f4f51050 | 3105 | spin_unlock(&pic_irqchip(kvm)->lock); |
1fe779f8 CO |
3106 | break; |
3107 | case KVM_IRQCHIP_PIC_SLAVE: | |
f4f51050 | 3108 | spin_lock(&pic_irqchip(kvm)->lock); |
1fe779f8 CO |
3109 | memcpy(&pic_irqchip(kvm)->pics[1], |
3110 | &chip->chip.pic, | |
3111 | sizeof(struct kvm_pic_state)); | |
f4f51050 | 3112 | spin_unlock(&pic_irqchip(kvm)->lock); |
1fe779f8 CO |
3113 | break; |
3114 | case KVM_IRQCHIP_IOAPIC: | |
eba0226b | 3115 | r = kvm_set_ioapic(kvm, &chip->chip.ioapic); |
1fe779f8 CO |
3116 | break; |
3117 | default: | |
3118 | r = -EINVAL; | |
3119 | break; | |
3120 | } | |
3121 | kvm_pic_update_irq(pic_irqchip(kvm)); | |
3122 | return r; | |
3123 | } | |
3124 | ||
e0f63cb9 SY |
3125 | static int kvm_vm_ioctl_get_pit(struct kvm *kvm, struct kvm_pit_state *ps) |
3126 | { | |
3127 | int r = 0; | |
3128 | ||
894a9c55 | 3129 | mutex_lock(&kvm->arch.vpit->pit_state.lock); |
e0f63cb9 | 3130 | memcpy(ps, &kvm->arch.vpit->pit_state, sizeof(struct kvm_pit_state)); |
894a9c55 | 3131 | mutex_unlock(&kvm->arch.vpit->pit_state.lock); |
e0f63cb9 SY |
3132 | return r; |
3133 | } | |
3134 | ||
3135 | static int kvm_vm_ioctl_set_pit(struct kvm *kvm, struct kvm_pit_state *ps) | |
3136 | { | |
3137 | int r = 0; | |
3138 | ||
894a9c55 | 3139 | mutex_lock(&kvm->arch.vpit->pit_state.lock); |
e0f63cb9 | 3140 | memcpy(&kvm->arch.vpit->pit_state, ps, sizeof(struct kvm_pit_state)); |
e9f42757 BK |
3141 | kvm_pit_load_count(kvm, 0, ps->channels[0].count, 0); |
3142 | mutex_unlock(&kvm->arch.vpit->pit_state.lock); | |
3143 | return r; | |
3144 | } | |
3145 | ||
3146 | static int kvm_vm_ioctl_get_pit2(struct kvm *kvm, struct kvm_pit_state2 *ps) | |
3147 | { | |
3148 | int r = 0; | |
3149 | ||
3150 | mutex_lock(&kvm->arch.vpit->pit_state.lock); | |
3151 | memcpy(ps->channels, &kvm->arch.vpit->pit_state.channels, | |
3152 | sizeof(ps->channels)); | |
3153 | ps->flags = kvm->arch.vpit->pit_state.flags; | |
3154 | mutex_unlock(&kvm->arch.vpit->pit_state.lock); | |
97e69aa6 | 3155 | memset(&ps->reserved, 0, sizeof(ps->reserved)); |
e9f42757 BK |
3156 | return r; |
3157 | } | |
3158 | ||
3159 | static int kvm_vm_ioctl_set_pit2(struct kvm *kvm, struct kvm_pit_state2 *ps) | |
3160 | { | |
3161 | int r = 0, start = 0; | |
3162 | u32 prev_legacy, cur_legacy; | |
3163 | mutex_lock(&kvm->arch.vpit->pit_state.lock); | |
3164 | prev_legacy = kvm->arch.vpit->pit_state.flags & KVM_PIT_FLAGS_HPET_LEGACY; | |
3165 | cur_legacy = ps->flags & KVM_PIT_FLAGS_HPET_LEGACY; | |
3166 | if (!prev_legacy && cur_legacy) | |
3167 | start = 1; | |
3168 | memcpy(&kvm->arch.vpit->pit_state.channels, &ps->channels, | |
3169 | sizeof(kvm->arch.vpit->pit_state.channels)); | |
3170 | kvm->arch.vpit->pit_state.flags = ps->flags; | |
3171 | kvm_pit_load_count(kvm, 0, kvm->arch.vpit->pit_state.channels[0].count, start); | |
894a9c55 | 3172 | mutex_unlock(&kvm->arch.vpit->pit_state.lock); |
e0f63cb9 SY |
3173 | return r; |
3174 | } | |
3175 | ||
52d939a0 MT |
3176 | static int kvm_vm_ioctl_reinject(struct kvm *kvm, |
3177 | struct kvm_reinject_control *control) | |
3178 | { | |
3179 | if (!kvm->arch.vpit) | |
3180 | return -ENXIO; | |
894a9c55 | 3181 | mutex_lock(&kvm->arch.vpit->pit_state.lock); |
52d939a0 | 3182 | kvm->arch.vpit->pit_state.pit_timer.reinject = control->pit_reinject; |
894a9c55 | 3183 | mutex_unlock(&kvm->arch.vpit->pit_state.lock); |
52d939a0 MT |
3184 | return 0; |
3185 | } | |
3186 | ||
5bb064dc ZX |
3187 | /* |
3188 | * Get (and clear) the dirty memory log for a memory slot. | |
3189 | */ | |
3190 | int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, | |
3191 | struct kvm_dirty_log *log) | |
3192 | { | |
87bf6e7d | 3193 | int r, i; |
5bb064dc | 3194 | struct kvm_memory_slot *memslot; |
87bf6e7d | 3195 | unsigned long n; |
b050b015 | 3196 | unsigned long is_dirty = 0; |
5bb064dc | 3197 | |
79fac95e | 3198 | mutex_lock(&kvm->slots_lock); |
5bb064dc | 3199 | |
b050b015 MT |
3200 | r = -EINVAL; |
3201 | if (log->slot >= KVM_MEMORY_SLOTS) | |
3202 | goto out; | |
3203 | ||
3204 | memslot = &kvm->memslots->memslots[log->slot]; | |
3205 | r = -ENOENT; | |
3206 | if (!memslot->dirty_bitmap) | |
3207 | goto out; | |
3208 | ||
87bf6e7d | 3209 | n = kvm_dirty_bitmap_bytes(memslot); |
b050b015 | 3210 | |
b050b015 MT |
3211 | for (i = 0; !is_dirty && i < n/sizeof(long); i++) |
3212 | is_dirty = memslot->dirty_bitmap[i]; | |
5bb064dc ZX |
3213 | |
3214 | /* If nothing is dirty, don't bother messing with page tables. */ | |
3215 | if (is_dirty) { | |
b050b015 | 3216 | struct kvm_memslots *slots, *old_slots; |
914ebccd | 3217 | unsigned long *dirty_bitmap; |
b050b015 | 3218 | |
515a0127 TY |
3219 | dirty_bitmap = memslot->dirty_bitmap_head; |
3220 | if (memslot->dirty_bitmap == dirty_bitmap) | |
3221 | dirty_bitmap += n / sizeof(long); | |
914ebccd | 3222 | memset(dirty_bitmap, 0, n); |
b050b015 | 3223 | |
914ebccd TY |
3224 | r = -ENOMEM; |
3225 | slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
515a0127 | 3226 | if (!slots) |
914ebccd | 3227 | goto out; |
b050b015 MT |
3228 | memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots)); |
3229 | slots->memslots[log->slot].dirty_bitmap = dirty_bitmap; | |
49c7754c | 3230 | slots->generation++; |
b050b015 MT |
3231 | |
3232 | old_slots = kvm->memslots; | |
3233 | rcu_assign_pointer(kvm->memslots, slots); | |
3234 | synchronize_srcu_expedited(&kvm->srcu); | |
3235 | dirty_bitmap = old_slots->memslots[log->slot].dirty_bitmap; | |
3236 | kfree(old_slots); | |
914ebccd | 3237 | |
edde99ce MT |
3238 | spin_lock(&kvm->mmu_lock); |
3239 | kvm_mmu_slot_remove_write_access(kvm, log->slot); | |
3240 | spin_unlock(&kvm->mmu_lock); | |
3241 | ||
914ebccd | 3242 | r = -EFAULT; |
515a0127 | 3243 | if (copy_to_user(log->dirty_bitmap, dirty_bitmap, n)) |
914ebccd | 3244 | goto out; |
914ebccd TY |
3245 | } else { |
3246 | r = -EFAULT; | |
3247 | if (clear_user(log->dirty_bitmap, n)) | |
3248 | goto out; | |
5bb064dc | 3249 | } |
b050b015 | 3250 | |
5bb064dc ZX |
3251 | r = 0; |
3252 | out: | |
79fac95e | 3253 | mutex_unlock(&kvm->slots_lock); |
5bb064dc ZX |
3254 | return r; |
3255 | } | |
3256 | ||
1fe779f8 CO |
3257 | long kvm_arch_vm_ioctl(struct file *filp, |
3258 | unsigned int ioctl, unsigned long arg) | |
3259 | { | |
3260 | struct kvm *kvm = filp->private_data; | |
3261 | void __user *argp = (void __user *)arg; | |
367e1319 | 3262 | int r = -ENOTTY; |
f0d66275 DH |
3263 | /* |
3264 | * This union makes it completely explicit to gcc-3.x | |
3265 | * that these two variables' stack usage should be | |
3266 | * combined, not added together. | |
3267 | */ | |
3268 | union { | |
3269 | struct kvm_pit_state ps; | |
e9f42757 | 3270 | struct kvm_pit_state2 ps2; |
c5ff41ce | 3271 | struct kvm_pit_config pit_config; |
f0d66275 | 3272 | } u; |
1fe779f8 CO |
3273 | |
3274 | switch (ioctl) { | |
3275 | case KVM_SET_TSS_ADDR: | |
3276 | r = kvm_vm_ioctl_set_tss_addr(kvm, arg); | |
3277 | if (r < 0) | |
3278 | goto out; | |
3279 | break; | |
b927a3ce SY |
3280 | case KVM_SET_IDENTITY_MAP_ADDR: { |
3281 | u64 ident_addr; | |
3282 | ||
3283 | r = -EFAULT; | |
3284 | if (copy_from_user(&ident_addr, argp, sizeof ident_addr)) | |
3285 | goto out; | |
3286 | r = kvm_vm_ioctl_set_identity_map_addr(kvm, ident_addr); | |
3287 | if (r < 0) | |
3288 | goto out; | |
3289 | break; | |
3290 | } | |
1fe779f8 CO |
3291 | case KVM_SET_NR_MMU_PAGES: |
3292 | r = kvm_vm_ioctl_set_nr_mmu_pages(kvm, arg); | |
3293 | if (r) | |
3294 | goto out; | |
3295 | break; | |
3296 | case KVM_GET_NR_MMU_PAGES: | |
3297 | r = kvm_vm_ioctl_get_nr_mmu_pages(kvm); | |
3298 | break; | |
3ddea128 MT |
3299 | case KVM_CREATE_IRQCHIP: { |
3300 | struct kvm_pic *vpic; | |
3301 | ||
3302 | mutex_lock(&kvm->lock); | |
3303 | r = -EEXIST; | |
3304 | if (kvm->arch.vpic) | |
3305 | goto create_irqchip_unlock; | |
1fe779f8 | 3306 | r = -ENOMEM; |
3ddea128 MT |
3307 | vpic = kvm_create_pic(kvm); |
3308 | if (vpic) { | |
1fe779f8 CO |
3309 | r = kvm_ioapic_init(kvm); |
3310 | if (r) { | |
72bb2fcd WY |
3311 | kvm_io_bus_unregister_dev(kvm, KVM_PIO_BUS, |
3312 | &vpic->dev); | |
3ddea128 MT |
3313 | kfree(vpic); |
3314 | goto create_irqchip_unlock; | |
1fe779f8 CO |
3315 | } |
3316 | } else | |
3ddea128 MT |
3317 | goto create_irqchip_unlock; |
3318 | smp_wmb(); | |
3319 | kvm->arch.vpic = vpic; | |
3320 | smp_wmb(); | |
399ec807 AK |
3321 | r = kvm_setup_default_irq_routing(kvm); |
3322 | if (r) { | |
3ddea128 | 3323 | mutex_lock(&kvm->irq_lock); |
72bb2fcd WY |
3324 | kvm_ioapic_destroy(kvm); |
3325 | kvm_destroy_pic(kvm); | |
3ddea128 | 3326 | mutex_unlock(&kvm->irq_lock); |
399ec807 | 3327 | } |
3ddea128 MT |
3328 | create_irqchip_unlock: |
3329 | mutex_unlock(&kvm->lock); | |
1fe779f8 | 3330 | break; |
3ddea128 | 3331 | } |
7837699f | 3332 | case KVM_CREATE_PIT: |
c5ff41ce JK |
3333 | u.pit_config.flags = KVM_PIT_SPEAKER_DUMMY; |
3334 | goto create_pit; | |
3335 | case KVM_CREATE_PIT2: | |
3336 | r = -EFAULT; | |
3337 | if (copy_from_user(&u.pit_config, argp, | |
3338 | sizeof(struct kvm_pit_config))) | |
3339 | goto out; | |
3340 | create_pit: | |
79fac95e | 3341 | mutex_lock(&kvm->slots_lock); |
269e05e4 AK |
3342 | r = -EEXIST; |
3343 | if (kvm->arch.vpit) | |
3344 | goto create_pit_unlock; | |
7837699f | 3345 | r = -ENOMEM; |
c5ff41ce | 3346 | kvm->arch.vpit = kvm_create_pit(kvm, u.pit_config.flags); |
7837699f SY |
3347 | if (kvm->arch.vpit) |
3348 | r = 0; | |
269e05e4 | 3349 | create_pit_unlock: |
79fac95e | 3350 | mutex_unlock(&kvm->slots_lock); |
7837699f | 3351 | break; |
4925663a | 3352 | case KVM_IRQ_LINE_STATUS: |
1fe779f8 CO |
3353 | case KVM_IRQ_LINE: { |
3354 | struct kvm_irq_level irq_event; | |
3355 | ||
3356 | r = -EFAULT; | |
3357 | if (copy_from_user(&irq_event, argp, sizeof irq_event)) | |
3358 | goto out; | |
160d2f6c | 3359 | r = -ENXIO; |
1fe779f8 | 3360 | if (irqchip_in_kernel(kvm)) { |
4925663a | 3361 | __s32 status; |
4925663a GN |
3362 | status = kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID, |
3363 | irq_event.irq, irq_event.level); | |
4925663a | 3364 | if (ioctl == KVM_IRQ_LINE_STATUS) { |
160d2f6c | 3365 | r = -EFAULT; |
4925663a GN |
3366 | irq_event.status = status; |
3367 | if (copy_to_user(argp, &irq_event, | |
3368 | sizeof irq_event)) | |
3369 | goto out; | |
3370 | } | |
1fe779f8 CO |
3371 | r = 0; |
3372 | } | |
3373 | break; | |
3374 | } | |
3375 | case KVM_GET_IRQCHIP: { | |
3376 | /* 0: PIC master, 1: PIC slave, 2: IOAPIC */ | |
f0d66275 | 3377 | struct kvm_irqchip *chip = kmalloc(sizeof(*chip), GFP_KERNEL); |
1fe779f8 | 3378 | |
f0d66275 DH |
3379 | r = -ENOMEM; |
3380 | if (!chip) | |
1fe779f8 | 3381 | goto out; |
f0d66275 DH |
3382 | r = -EFAULT; |
3383 | if (copy_from_user(chip, argp, sizeof *chip)) | |
3384 | goto get_irqchip_out; | |
1fe779f8 CO |
3385 | r = -ENXIO; |
3386 | if (!irqchip_in_kernel(kvm)) | |
f0d66275 DH |
3387 | goto get_irqchip_out; |
3388 | r = kvm_vm_ioctl_get_irqchip(kvm, chip); | |
1fe779f8 | 3389 | if (r) |
f0d66275 | 3390 | goto get_irqchip_out; |
1fe779f8 | 3391 | r = -EFAULT; |
f0d66275 DH |
3392 | if (copy_to_user(argp, chip, sizeof *chip)) |
3393 | goto get_irqchip_out; | |
1fe779f8 | 3394 | r = 0; |
f0d66275 DH |
3395 | get_irqchip_out: |
3396 | kfree(chip); | |
3397 | if (r) | |
3398 | goto out; | |
1fe779f8 CO |
3399 | break; |
3400 | } | |
3401 | case KVM_SET_IRQCHIP: { | |
3402 | /* 0: PIC master, 1: PIC slave, 2: IOAPIC */ | |
f0d66275 | 3403 | struct kvm_irqchip *chip = kmalloc(sizeof(*chip), GFP_KERNEL); |
1fe779f8 | 3404 | |
f0d66275 DH |
3405 | r = -ENOMEM; |
3406 | if (!chip) | |
1fe779f8 | 3407 | goto out; |
f0d66275 DH |
3408 | r = -EFAULT; |
3409 | if (copy_from_user(chip, argp, sizeof *chip)) | |
3410 | goto set_irqchip_out; | |
1fe779f8 CO |
3411 | r = -ENXIO; |
3412 | if (!irqchip_in_kernel(kvm)) | |
f0d66275 DH |
3413 | goto set_irqchip_out; |
3414 | r = kvm_vm_ioctl_set_irqchip(kvm, chip); | |
1fe779f8 | 3415 | if (r) |
f0d66275 | 3416 | goto set_irqchip_out; |
1fe779f8 | 3417 | r = 0; |
f0d66275 DH |
3418 | set_irqchip_out: |
3419 | kfree(chip); | |
3420 | if (r) | |
3421 | goto out; | |
1fe779f8 CO |
3422 | break; |
3423 | } | |
e0f63cb9 | 3424 | case KVM_GET_PIT: { |
e0f63cb9 | 3425 | r = -EFAULT; |
f0d66275 | 3426 | if (copy_from_user(&u.ps, argp, sizeof(struct kvm_pit_state))) |
e0f63cb9 SY |
3427 | goto out; |
3428 | r = -ENXIO; | |
3429 | if (!kvm->arch.vpit) | |
3430 | goto out; | |
f0d66275 | 3431 | r = kvm_vm_ioctl_get_pit(kvm, &u.ps); |
e0f63cb9 SY |
3432 | if (r) |
3433 | goto out; | |
3434 | r = -EFAULT; | |
f0d66275 | 3435 | if (copy_to_user(argp, &u.ps, sizeof(struct kvm_pit_state))) |
e0f63cb9 SY |
3436 | goto out; |
3437 | r = 0; | |
3438 | break; | |
3439 | } | |
3440 | case KVM_SET_PIT: { | |
e0f63cb9 | 3441 | r = -EFAULT; |
f0d66275 | 3442 | if (copy_from_user(&u.ps, argp, sizeof u.ps)) |
e0f63cb9 SY |
3443 | goto out; |
3444 | r = -ENXIO; | |
3445 | if (!kvm->arch.vpit) | |
3446 | goto out; | |
f0d66275 | 3447 | r = kvm_vm_ioctl_set_pit(kvm, &u.ps); |
e0f63cb9 SY |
3448 | if (r) |
3449 | goto out; | |
3450 | r = 0; | |
3451 | break; | |
3452 | } | |
e9f42757 BK |
3453 | case KVM_GET_PIT2: { |
3454 | r = -ENXIO; | |
3455 | if (!kvm->arch.vpit) | |
3456 | goto out; | |
3457 | r = kvm_vm_ioctl_get_pit2(kvm, &u.ps2); | |
3458 | if (r) | |
3459 | goto out; | |
3460 | r = -EFAULT; | |
3461 | if (copy_to_user(argp, &u.ps2, sizeof(u.ps2))) | |
3462 | goto out; | |
3463 | r = 0; | |
3464 | break; | |
3465 | } | |
3466 | case KVM_SET_PIT2: { | |
3467 | r = -EFAULT; | |
3468 | if (copy_from_user(&u.ps2, argp, sizeof(u.ps2))) | |
3469 | goto out; | |
3470 | r = -ENXIO; | |
3471 | if (!kvm->arch.vpit) | |
3472 | goto out; | |
3473 | r = kvm_vm_ioctl_set_pit2(kvm, &u.ps2); | |
3474 | if (r) | |
3475 | goto out; | |
3476 | r = 0; | |
3477 | break; | |
3478 | } | |
52d939a0 MT |
3479 | case KVM_REINJECT_CONTROL: { |
3480 | struct kvm_reinject_control control; | |
3481 | r = -EFAULT; | |
3482 | if (copy_from_user(&control, argp, sizeof(control))) | |
3483 | goto out; | |
3484 | r = kvm_vm_ioctl_reinject(kvm, &control); | |
3485 | if (r) | |
3486 | goto out; | |
3487 | r = 0; | |
3488 | break; | |
3489 | } | |
ffde22ac ES |
3490 | case KVM_XEN_HVM_CONFIG: { |
3491 | r = -EFAULT; | |
3492 | if (copy_from_user(&kvm->arch.xen_hvm_config, argp, | |
3493 | sizeof(struct kvm_xen_hvm_config))) | |
3494 | goto out; | |
3495 | r = -EINVAL; | |
3496 | if (kvm->arch.xen_hvm_config.flags) | |
3497 | goto out; | |
3498 | r = 0; | |
3499 | break; | |
3500 | } | |
afbcf7ab | 3501 | case KVM_SET_CLOCK: { |
afbcf7ab GC |
3502 | struct kvm_clock_data user_ns; |
3503 | u64 now_ns; | |
3504 | s64 delta; | |
3505 | ||
3506 | r = -EFAULT; | |
3507 | if (copy_from_user(&user_ns, argp, sizeof(user_ns))) | |
3508 | goto out; | |
3509 | ||
3510 | r = -EINVAL; | |
3511 | if (user_ns.flags) | |
3512 | goto out; | |
3513 | ||
3514 | r = 0; | |
395c6b0a | 3515 | local_irq_disable(); |
759379dd | 3516 | now_ns = get_kernel_ns(); |
afbcf7ab | 3517 | delta = user_ns.clock - now_ns; |
395c6b0a | 3518 | local_irq_enable(); |
afbcf7ab GC |
3519 | kvm->arch.kvmclock_offset = delta; |
3520 | break; | |
3521 | } | |
3522 | case KVM_GET_CLOCK: { | |
afbcf7ab GC |
3523 | struct kvm_clock_data user_ns; |
3524 | u64 now_ns; | |
3525 | ||
395c6b0a | 3526 | local_irq_disable(); |
759379dd | 3527 | now_ns = get_kernel_ns(); |
afbcf7ab | 3528 | user_ns.clock = kvm->arch.kvmclock_offset + now_ns; |
395c6b0a | 3529 | local_irq_enable(); |
afbcf7ab | 3530 | user_ns.flags = 0; |
97e69aa6 | 3531 | memset(&user_ns.pad, 0, sizeof(user_ns.pad)); |
afbcf7ab GC |
3532 | |
3533 | r = -EFAULT; | |
3534 | if (copy_to_user(argp, &user_ns, sizeof(user_ns))) | |
3535 | goto out; | |
3536 | r = 0; | |
3537 | break; | |
3538 | } | |
3539 | ||
1fe779f8 CO |
3540 | default: |
3541 | ; | |
3542 | } | |
3543 | out: | |
3544 | return r; | |
3545 | } | |
3546 | ||
a16b043c | 3547 | static void kvm_init_msr_list(void) |
043405e1 CO |
3548 | { |
3549 | u32 dummy[2]; | |
3550 | unsigned i, j; | |
3551 | ||
e3267cbb GC |
3552 | /* skip the first msrs in the list. KVM-specific */ |
3553 | for (i = j = KVM_SAVE_MSRS_BEGIN; i < ARRAY_SIZE(msrs_to_save); i++) { | |
043405e1 CO |
3554 | if (rdmsr_safe(msrs_to_save[i], &dummy[0], &dummy[1]) < 0) |
3555 | continue; | |
3556 | if (j < i) | |
3557 | msrs_to_save[j] = msrs_to_save[i]; | |
3558 | j++; | |
3559 | } | |
3560 | num_msrs_to_save = j; | |
3561 | } | |
3562 | ||
bda9020e MT |
3563 | static int vcpu_mmio_write(struct kvm_vcpu *vcpu, gpa_t addr, int len, |
3564 | const void *v) | |
bbd9b64e | 3565 | { |
bda9020e MT |
3566 | if (vcpu->arch.apic && |
3567 | !kvm_iodevice_write(&vcpu->arch.apic->dev, addr, len, v)) | |
3568 | return 0; | |
bbd9b64e | 3569 | |
e93f8a0f | 3570 | return kvm_io_bus_write(vcpu->kvm, KVM_MMIO_BUS, addr, len, v); |
bbd9b64e CO |
3571 | } |
3572 | ||
bda9020e | 3573 | static int vcpu_mmio_read(struct kvm_vcpu *vcpu, gpa_t addr, int len, void *v) |
bbd9b64e | 3574 | { |
bda9020e MT |
3575 | if (vcpu->arch.apic && |
3576 | !kvm_iodevice_read(&vcpu->arch.apic->dev, addr, len, v)) | |
3577 | return 0; | |
bbd9b64e | 3578 | |
e93f8a0f | 3579 | return kvm_io_bus_read(vcpu->kvm, KVM_MMIO_BUS, addr, len, v); |
bbd9b64e CO |
3580 | } |
3581 | ||
2dafc6c2 GN |
3582 | static void kvm_set_segment(struct kvm_vcpu *vcpu, |
3583 | struct kvm_segment *var, int seg) | |
3584 | { | |
3585 | kvm_x86_ops->set_segment(vcpu, var, seg); | |
3586 | } | |
3587 | ||
3588 | void kvm_get_segment(struct kvm_vcpu *vcpu, | |
3589 | struct kvm_segment *var, int seg) | |
3590 | { | |
3591 | kvm_x86_ops->get_segment(vcpu, var, seg); | |
3592 | } | |
3593 | ||
c30a358d JR |
3594 | static gpa_t translate_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access) |
3595 | { | |
3596 | return gpa; | |
3597 | } | |
3598 | ||
02f59dc9 JR |
3599 | static gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access) |
3600 | { | |
3601 | gpa_t t_gpa; | |
ab9ae313 | 3602 | struct x86_exception exception; |
02f59dc9 JR |
3603 | |
3604 | BUG_ON(!mmu_is_nested(vcpu)); | |
3605 | ||
3606 | /* NPT walks are always user-walks */ | |
3607 | access |= PFERR_USER_MASK; | |
ab9ae313 | 3608 | t_gpa = vcpu->arch.mmu.gva_to_gpa(vcpu, gpa, access, &exception); |
02f59dc9 JR |
3609 | |
3610 | return t_gpa; | |
3611 | } | |
3612 | ||
ab9ae313 AK |
3613 | gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva, |
3614 | struct x86_exception *exception) | |
1871c602 GN |
3615 | { |
3616 | u32 access = (kvm_x86_ops->get_cpl(vcpu) == 3) ? PFERR_USER_MASK : 0; | |
ab9ae313 | 3617 | return vcpu->arch.walk_mmu->gva_to_gpa(vcpu, gva, access, exception); |
1871c602 GN |
3618 | } |
3619 | ||
ab9ae313 AK |
3620 | gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva, |
3621 | struct x86_exception *exception) | |
1871c602 GN |
3622 | { |
3623 | u32 access = (kvm_x86_ops->get_cpl(vcpu) == 3) ? PFERR_USER_MASK : 0; | |
3624 | access |= PFERR_FETCH_MASK; | |
ab9ae313 | 3625 | return vcpu->arch.walk_mmu->gva_to_gpa(vcpu, gva, access, exception); |
1871c602 GN |
3626 | } |
3627 | ||
ab9ae313 AK |
3628 | gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva, |
3629 | struct x86_exception *exception) | |
1871c602 GN |
3630 | { |
3631 | u32 access = (kvm_x86_ops->get_cpl(vcpu) == 3) ? PFERR_USER_MASK : 0; | |
3632 | access |= PFERR_WRITE_MASK; | |
ab9ae313 | 3633 | return vcpu->arch.walk_mmu->gva_to_gpa(vcpu, gva, access, exception); |
1871c602 GN |
3634 | } |
3635 | ||
3636 | /* uses this to access any guest's mapped memory without checking CPL */ | |
ab9ae313 AK |
3637 | gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva, |
3638 | struct x86_exception *exception) | |
bcc55cba | 3639 | { |
ab9ae313 | 3640 | return vcpu->arch.walk_mmu->gva_to_gpa(vcpu, gva, 0, exception); |
bcc55cba AK |
3641 | } |
3642 | ||
1871c602 GN |
3643 | static int kvm_read_guest_virt_helper(gva_t addr, void *val, unsigned int bytes, |
3644 | struct kvm_vcpu *vcpu, u32 access, | |
bcc55cba | 3645 | struct x86_exception *exception) |
bbd9b64e CO |
3646 | { |
3647 | void *data = val; | |
10589a46 | 3648 | int r = X86EMUL_CONTINUE; |
bbd9b64e CO |
3649 | |
3650 | while (bytes) { | |
14dfe855 | 3651 | gpa_t gpa = vcpu->arch.walk_mmu->gva_to_gpa(vcpu, addr, access, |
ab9ae313 | 3652 | exception); |
bbd9b64e | 3653 | unsigned offset = addr & (PAGE_SIZE-1); |
77c2002e | 3654 | unsigned toread = min(bytes, (unsigned)PAGE_SIZE - offset); |
bbd9b64e CO |
3655 | int ret; |
3656 | ||
bcc55cba | 3657 | if (gpa == UNMAPPED_GVA) |
ab9ae313 | 3658 | return X86EMUL_PROPAGATE_FAULT; |
77c2002e | 3659 | ret = kvm_read_guest(vcpu->kvm, gpa, data, toread); |
10589a46 | 3660 | if (ret < 0) { |
c3cd7ffa | 3661 | r = X86EMUL_IO_NEEDED; |
10589a46 MT |
3662 | goto out; |
3663 | } | |
bbd9b64e | 3664 | |
77c2002e IE |
3665 | bytes -= toread; |
3666 | data += toread; | |
3667 | addr += toread; | |
bbd9b64e | 3668 | } |
10589a46 | 3669 | out: |
10589a46 | 3670 | return r; |
bbd9b64e | 3671 | } |
77c2002e | 3672 | |
1871c602 GN |
3673 | /* used for instruction fetching */ |
3674 | static int kvm_fetch_guest_virt(gva_t addr, void *val, unsigned int bytes, | |
bcc55cba AK |
3675 | struct kvm_vcpu *vcpu, |
3676 | struct x86_exception *exception) | |
1871c602 GN |
3677 | { |
3678 | u32 access = (kvm_x86_ops->get_cpl(vcpu) == 3) ? PFERR_USER_MASK : 0; | |
3679 | return kvm_read_guest_virt_helper(addr, val, bytes, vcpu, | |
bcc55cba AK |
3680 | access | PFERR_FETCH_MASK, |
3681 | exception); | |
1871c602 GN |
3682 | } |
3683 | ||
3684 | static int kvm_read_guest_virt(gva_t addr, void *val, unsigned int bytes, | |
bcc55cba AK |
3685 | struct kvm_vcpu *vcpu, |
3686 | struct x86_exception *exception) | |
1871c602 GN |
3687 | { |
3688 | u32 access = (kvm_x86_ops->get_cpl(vcpu) == 3) ? PFERR_USER_MASK : 0; | |
3689 | return kvm_read_guest_virt_helper(addr, val, bytes, vcpu, access, | |
bcc55cba | 3690 | exception); |
1871c602 GN |
3691 | } |
3692 | ||
3693 | static int kvm_read_guest_virt_system(gva_t addr, void *val, unsigned int bytes, | |
bcc55cba AK |
3694 | struct kvm_vcpu *vcpu, |
3695 | struct x86_exception *exception) | |
1871c602 | 3696 | { |
bcc55cba | 3697 | return kvm_read_guest_virt_helper(addr, val, bytes, vcpu, 0, exception); |
1871c602 GN |
3698 | } |
3699 | ||
7972995b | 3700 | static int kvm_write_guest_virt_system(gva_t addr, void *val, |
2dafc6c2 | 3701 | unsigned int bytes, |
7972995b | 3702 | struct kvm_vcpu *vcpu, |
bcc55cba | 3703 | struct x86_exception *exception) |
77c2002e IE |
3704 | { |
3705 | void *data = val; | |
3706 | int r = X86EMUL_CONTINUE; | |
3707 | ||
3708 | while (bytes) { | |
14dfe855 JR |
3709 | gpa_t gpa = vcpu->arch.walk_mmu->gva_to_gpa(vcpu, addr, |
3710 | PFERR_WRITE_MASK, | |
ab9ae313 | 3711 | exception); |
77c2002e IE |
3712 | unsigned offset = addr & (PAGE_SIZE-1); |
3713 | unsigned towrite = min(bytes, (unsigned)PAGE_SIZE - offset); | |
3714 | int ret; | |
3715 | ||
bcc55cba | 3716 | if (gpa == UNMAPPED_GVA) |
ab9ae313 | 3717 | return X86EMUL_PROPAGATE_FAULT; |
77c2002e IE |
3718 | ret = kvm_write_guest(vcpu->kvm, gpa, data, towrite); |
3719 | if (ret < 0) { | |
c3cd7ffa | 3720 | r = X86EMUL_IO_NEEDED; |
77c2002e IE |
3721 | goto out; |
3722 | } | |
3723 | ||
3724 | bytes -= towrite; | |
3725 | data += towrite; | |
3726 | addr += towrite; | |
3727 | } | |
3728 | out: | |
3729 | return r; | |
3730 | } | |
3731 | ||
bbd9b64e CO |
3732 | static int emulator_read_emulated(unsigned long addr, |
3733 | void *val, | |
3734 | unsigned int bytes, | |
bcc55cba | 3735 | struct x86_exception *exception, |
bbd9b64e CO |
3736 | struct kvm_vcpu *vcpu) |
3737 | { | |
bbd9b64e CO |
3738 | gpa_t gpa; |
3739 | ||
3740 | if (vcpu->mmio_read_completed) { | |
3741 | memcpy(val, vcpu->mmio_data, bytes); | |
aec51dc4 AK |
3742 | trace_kvm_mmio(KVM_TRACE_MMIO_READ, bytes, |
3743 | vcpu->mmio_phys_addr, *(u64 *)val); | |
bbd9b64e CO |
3744 | vcpu->mmio_read_completed = 0; |
3745 | return X86EMUL_CONTINUE; | |
3746 | } | |
3747 | ||
ab9ae313 | 3748 | gpa = kvm_mmu_gva_to_gpa_read(vcpu, addr, exception); |
1871c602 | 3749 | |
8fe681e9 | 3750 | if (gpa == UNMAPPED_GVA) |
ab9ae313 | 3751 | return X86EMUL_PROPAGATE_FAULT; |
bbd9b64e CO |
3752 | |
3753 | /* For APIC access vmexit */ | |
3754 | if ((gpa & PAGE_MASK) == APIC_DEFAULT_PHYS_BASE) | |
3755 | goto mmio; | |
3756 | ||
bcc55cba AK |
3757 | if (kvm_read_guest_virt(addr, val, bytes, vcpu, exception) |
3758 | == X86EMUL_CONTINUE) | |
bbd9b64e | 3759 | return X86EMUL_CONTINUE; |
bbd9b64e CO |
3760 | |
3761 | mmio: | |
3762 | /* | |
3763 | * Is this MMIO handled locally? | |
3764 | */ | |
aec51dc4 AK |
3765 | if (!vcpu_mmio_read(vcpu, gpa, bytes, val)) { |
3766 | trace_kvm_mmio(KVM_TRACE_MMIO_READ, bytes, gpa, *(u64 *)val); | |
bbd9b64e CO |
3767 | return X86EMUL_CONTINUE; |
3768 | } | |
aec51dc4 AK |
3769 | |
3770 | trace_kvm_mmio(KVM_TRACE_MMIO_READ_UNSATISFIED, bytes, gpa, 0); | |
bbd9b64e CO |
3771 | |
3772 | vcpu->mmio_needed = 1; | |
411c35b7 GN |
3773 | vcpu->run->exit_reason = KVM_EXIT_MMIO; |
3774 | vcpu->run->mmio.phys_addr = vcpu->mmio_phys_addr = gpa; | |
3775 | vcpu->run->mmio.len = vcpu->mmio_size = bytes; | |
3776 | vcpu->run->mmio.is_write = vcpu->mmio_is_write = 0; | |
bbd9b64e | 3777 | |
c3cd7ffa | 3778 | return X86EMUL_IO_NEEDED; |
bbd9b64e CO |
3779 | } |
3780 | ||
3200f405 | 3781 | int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa, |
bcc55cba | 3782 | const void *val, int bytes) |
bbd9b64e CO |
3783 | { |
3784 | int ret; | |
3785 | ||
3786 | ret = kvm_write_guest(vcpu->kvm, gpa, val, bytes); | |
9f811285 | 3787 | if (ret < 0) |
bbd9b64e | 3788 | return 0; |
ad218f85 | 3789 | kvm_mmu_pte_write(vcpu, gpa, val, bytes, 1); |
bbd9b64e CO |
3790 | return 1; |
3791 | } | |
3792 | ||
3793 | static int emulator_write_emulated_onepage(unsigned long addr, | |
3794 | const void *val, | |
3795 | unsigned int bytes, | |
bcc55cba | 3796 | struct x86_exception *exception, |
bbd9b64e CO |
3797 | struct kvm_vcpu *vcpu) |
3798 | { | |
10589a46 MT |
3799 | gpa_t gpa; |
3800 | ||
ab9ae313 | 3801 | gpa = kvm_mmu_gva_to_gpa_write(vcpu, addr, exception); |
bbd9b64e | 3802 | |
8fe681e9 | 3803 | if (gpa == UNMAPPED_GVA) |
ab9ae313 | 3804 | return X86EMUL_PROPAGATE_FAULT; |
bbd9b64e CO |
3805 | |
3806 | /* For APIC access vmexit */ | |
3807 | if ((gpa & PAGE_MASK) == APIC_DEFAULT_PHYS_BASE) | |
3808 | goto mmio; | |
3809 | ||
3810 | if (emulator_write_phys(vcpu, gpa, val, bytes)) | |
3811 | return X86EMUL_CONTINUE; | |
3812 | ||
3813 | mmio: | |
aec51dc4 | 3814 | trace_kvm_mmio(KVM_TRACE_MMIO_WRITE, bytes, gpa, *(u64 *)val); |
bbd9b64e CO |
3815 | /* |
3816 | * Is this MMIO handled locally? | |
3817 | */ | |
bda9020e | 3818 | if (!vcpu_mmio_write(vcpu, gpa, bytes, val)) |
bbd9b64e | 3819 | return X86EMUL_CONTINUE; |
bbd9b64e CO |
3820 | |
3821 | vcpu->mmio_needed = 1; | |
411c35b7 GN |
3822 | vcpu->run->exit_reason = KVM_EXIT_MMIO; |
3823 | vcpu->run->mmio.phys_addr = vcpu->mmio_phys_addr = gpa; | |
3824 | vcpu->run->mmio.len = vcpu->mmio_size = bytes; | |
3825 | vcpu->run->mmio.is_write = vcpu->mmio_is_write = 1; | |
3826 | memcpy(vcpu->run->mmio.data, val, bytes); | |
bbd9b64e CO |
3827 | |
3828 | return X86EMUL_CONTINUE; | |
3829 | } | |
3830 | ||
3831 | int emulator_write_emulated(unsigned long addr, | |
8f6abd06 GN |
3832 | const void *val, |
3833 | unsigned int bytes, | |
bcc55cba | 3834 | struct x86_exception *exception, |
8f6abd06 | 3835 | struct kvm_vcpu *vcpu) |
bbd9b64e CO |
3836 | { |
3837 | /* Crossing a page boundary? */ | |
3838 | if (((addr + bytes - 1) ^ addr) & PAGE_MASK) { | |
3839 | int rc, now; | |
3840 | ||
3841 | now = -addr & ~PAGE_MASK; | |
bcc55cba | 3842 | rc = emulator_write_emulated_onepage(addr, val, now, exception, |
8fe681e9 | 3843 | vcpu); |
bbd9b64e CO |
3844 | if (rc != X86EMUL_CONTINUE) |
3845 | return rc; | |
3846 | addr += now; | |
3847 | val += now; | |
3848 | bytes -= now; | |
3849 | } | |
bcc55cba | 3850 | return emulator_write_emulated_onepage(addr, val, bytes, exception, |
8fe681e9 | 3851 | vcpu); |
bbd9b64e | 3852 | } |
bbd9b64e | 3853 | |
daea3e73 AK |
3854 | #define CMPXCHG_TYPE(t, ptr, old, new) \ |
3855 | (cmpxchg((t *)(ptr), *(t *)(old), *(t *)(new)) == *(t *)(old)) | |
3856 | ||
3857 | #ifdef CONFIG_X86_64 | |
3858 | # define CMPXCHG64(ptr, old, new) CMPXCHG_TYPE(u64, ptr, old, new) | |
3859 | #else | |
3860 | # define CMPXCHG64(ptr, old, new) \ | |
9749a6c0 | 3861 | (cmpxchg64((u64 *)(ptr), *(u64 *)(old), *(u64 *)(new)) == *(u64 *)(old)) |
daea3e73 AK |
3862 | #endif |
3863 | ||
bbd9b64e CO |
3864 | static int emulator_cmpxchg_emulated(unsigned long addr, |
3865 | const void *old, | |
3866 | const void *new, | |
3867 | unsigned int bytes, | |
bcc55cba | 3868 | struct x86_exception *exception, |
bbd9b64e CO |
3869 | struct kvm_vcpu *vcpu) |
3870 | { | |
daea3e73 AK |
3871 | gpa_t gpa; |
3872 | struct page *page; | |
3873 | char *kaddr; | |
3874 | bool exchanged; | |
2bacc55c | 3875 | |
daea3e73 AK |
3876 | /* guests cmpxchg8b have to be emulated atomically */ |
3877 | if (bytes > 8 || (bytes & (bytes - 1))) | |
3878 | goto emul_write; | |
10589a46 | 3879 | |
daea3e73 | 3880 | gpa = kvm_mmu_gva_to_gpa_write(vcpu, addr, NULL); |
2bacc55c | 3881 | |
daea3e73 AK |
3882 | if (gpa == UNMAPPED_GVA || |
3883 | (gpa & PAGE_MASK) == APIC_DEFAULT_PHYS_BASE) | |
3884 | goto emul_write; | |
2bacc55c | 3885 | |
daea3e73 AK |
3886 | if (((gpa + bytes - 1) & PAGE_MASK) != (gpa & PAGE_MASK)) |
3887 | goto emul_write; | |
72dc67a6 | 3888 | |
daea3e73 | 3889 | page = gfn_to_page(vcpu->kvm, gpa >> PAGE_SHIFT); |
c19b8bd6 WY |
3890 | if (is_error_page(page)) { |
3891 | kvm_release_page_clean(page); | |
3892 | goto emul_write; | |
3893 | } | |
72dc67a6 | 3894 | |
daea3e73 AK |
3895 | kaddr = kmap_atomic(page, KM_USER0); |
3896 | kaddr += offset_in_page(gpa); | |
3897 | switch (bytes) { | |
3898 | case 1: | |
3899 | exchanged = CMPXCHG_TYPE(u8, kaddr, old, new); | |
3900 | break; | |
3901 | case 2: | |
3902 | exchanged = CMPXCHG_TYPE(u16, kaddr, old, new); | |
3903 | break; | |
3904 | case 4: | |
3905 | exchanged = CMPXCHG_TYPE(u32, kaddr, old, new); | |
3906 | break; | |
3907 | case 8: | |
3908 | exchanged = CMPXCHG64(kaddr, old, new); | |
3909 | break; | |
3910 | default: | |
3911 | BUG(); | |
2bacc55c | 3912 | } |
daea3e73 AK |
3913 | kunmap_atomic(kaddr, KM_USER0); |
3914 | kvm_release_page_dirty(page); | |
3915 | ||
3916 | if (!exchanged) | |
3917 | return X86EMUL_CMPXCHG_FAILED; | |
3918 | ||
8f6abd06 GN |
3919 | kvm_mmu_pte_write(vcpu, gpa, new, bytes, 1); |
3920 | ||
3921 | return X86EMUL_CONTINUE; | |
4a5f48f6 | 3922 | |
3200f405 | 3923 | emul_write: |
daea3e73 | 3924 | printk_once(KERN_WARNING "kvm: emulating exchange as write\n"); |
2bacc55c | 3925 | |
bcc55cba | 3926 | return emulator_write_emulated(addr, new, bytes, exception, vcpu); |
bbd9b64e CO |
3927 | } |
3928 | ||
cf8f70bf GN |
3929 | static int kernel_pio(struct kvm_vcpu *vcpu, void *pd) |
3930 | { | |
3931 | /* TODO: String I/O for in kernel device */ | |
3932 | int r; | |
3933 | ||
3934 | if (vcpu->arch.pio.in) | |
3935 | r = kvm_io_bus_read(vcpu->kvm, KVM_PIO_BUS, vcpu->arch.pio.port, | |
3936 | vcpu->arch.pio.size, pd); | |
3937 | else | |
3938 | r = kvm_io_bus_write(vcpu->kvm, KVM_PIO_BUS, | |
3939 | vcpu->arch.pio.port, vcpu->arch.pio.size, | |
3940 | pd); | |
3941 | return r; | |
3942 | } | |
3943 | ||
3944 | ||
3945 | static int emulator_pio_in_emulated(int size, unsigned short port, void *val, | |
3946 | unsigned int count, struct kvm_vcpu *vcpu) | |
3947 | { | |
7972995b | 3948 | if (vcpu->arch.pio.count) |
cf8f70bf GN |
3949 | goto data_avail; |
3950 | ||
c41a15dd | 3951 | trace_kvm_pio(0, port, size, 1); |
cf8f70bf GN |
3952 | |
3953 | vcpu->arch.pio.port = port; | |
3954 | vcpu->arch.pio.in = 1; | |
7972995b | 3955 | vcpu->arch.pio.count = count; |
cf8f70bf GN |
3956 | vcpu->arch.pio.size = size; |
3957 | ||
3958 | if (!kernel_pio(vcpu, vcpu->arch.pio_data)) { | |
3959 | data_avail: | |
3960 | memcpy(val, vcpu->arch.pio_data, size * count); | |
7972995b | 3961 | vcpu->arch.pio.count = 0; |
cf8f70bf GN |
3962 | return 1; |
3963 | } | |
3964 | ||
3965 | vcpu->run->exit_reason = KVM_EXIT_IO; | |
3966 | vcpu->run->io.direction = KVM_EXIT_IO_IN; | |
3967 | vcpu->run->io.size = size; | |
3968 | vcpu->run->io.data_offset = KVM_PIO_PAGE_OFFSET * PAGE_SIZE; | |
3969 | vcpu->run->io.count = count; | |
3970 | vcpu->run->io.port = port; | |
3971 | ||
3972 | return 0; | |
3973 | } | |
3974 | ||
3975 | static int emulator_pio_out_emulated(int size, unsigned short port, | |
3976 | const void *val, unsigned int count, | |
3977 | struct kvm_vcpu *vcpu) | |
3978 | { | |
c41a15dd | 3979 | trace_kvm_pio(1, port, size, 1); |
cf8f70bf GN |
3980 | |
3981 | vcpu->arch.pio.port = port; | |
3982 | vcpu->arch.pio.in = 0; | |
7972995b | 3983 | vcpu->arch.pio.count = count; |
cf8f70bf GN |
3984 | vcpu->arch.pio.size = size; |
3985 | ||
3986 | memcpy(vcpu->arch.pio_data, val, size * count); | |
3987 | ||
3988 | if (!kernel_pio(vcpu, vcpu->arch.pio_data)) { | |
7972995b | 3989 | vcpu->arch.pio.count = 0; |
cf8f70bf GN |
3990 | return 1; |
3991 | } | |
3992 | ||
3993 | vcpu->run->exit_reason = KVM_EXIT_IO; | |
3994 | vcpu->run->io.direction = KVM_EXIT_IO_OUT; | |
3995 | vcpu->run->io.size = size; | |
3996 | vcpu->run->io.data_offset = KVM_PIO_PAGE_OFFSET * PAGE_SIZE; | |
3997 | vcpu->run->io.count = count; | |
3998 | vcpu->run->io.port = port; | |
3999 | ||
4000 | return 0; | |
4001 | } | |
4002 | ||
bbd9b64e CO |
4003 | static unsigned long get_segment_base(struct kvm_vcpu *vcpu, int seg) |
4004 | { | |
4005 | return kvm_x86_ops->get_segment_base(vcpu, seg); | |
4006 | } | |
4007 | ||
4008 | int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address) | |
4009 | { | |
a7052897 | 4010 | kvm_mmu_invlpg(vcpu, address); |
bbd9b64e CO |
4011 | return X86EMUL_CONTINUE; |
4012 | } | |
4013 | ||
f5f48ee1 SY |
4014 | int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu) |
4015 | { | |
4016 | if (!need_emulate_wbinvd(vcpu)) | |
4017 | return X86EMUL_CONTINUE; | |
4018 | ||
4019 | if (kvm_x86_ops->has_wbinvd_exit()) { | |
2eec7343 JK |
4020 | int cpu = get_cpu(); |
4021 | ||
4022 | cpumask_set_cpu(cpu, vcpu->arch.wbinvd_dirty_mask); | |
f5f48ee1 SY |
4023 | smp_call_function_many(vcpu->arch.wbinvd_dirty_mask, |
4024 | wbinvd_ipi, NULL, 1); | |
2eec7343 | 4025 | put_cpu(); |
f5f48ee1 | 4026 | cpumask_clear(vcpu->arch.wbinvd_dirty_mask); |
2eec7343 JK |
4027 | } else |
4028 | wbinvd(); | |
f5f48ee1 SY |
4029 | return X86EMUL_CONTINUE; |
4030 | } | |
4031 | EXPORT_SYMBOL_GPL(kvm_emulate_wbinvd); | |
4032 | ||
bbd9b64e CO |
4033 | int emulate_clts(struct kvm_vcpu *vcpu) |
4034 | { | |
4d4ec087 | 4035 | kvm_x86_ops->set_cr0(vcpu, kvm_read_cr0_bits(vcpu, ~X86_CR0_TS)); |
6b52d186 | 4036 | kvm_x86_ops->fpu_activate(vcpu); |
bbd9b64e CO |
4037 | return X86EMUL_CONTINUE; |
4038 | } | |
4039 | ||
35aa5375 | 4040 | int emulator_get_dr(int dr, unsigned long *dest, struct kvm_vcpu *vcpu) |
bbd9b64e | 4041 | { |
338dbc97 | 4042 | return _kvm_get_dr(vcpu, dr, dest); |
bbd9b64e CO |
4043 | } |
4044 | ||
35aa5375 | 4045 | int emulator_set_dr(int dr, unsigned long value, struct kvm_vcpu *vcpu) |
bbd9b64e | 4046 | { |
338dbc97 GN |
4047 | |
4048 | return __kvm_set_dr(vcpu, dr, value); | |
bbd9b64e CO |
4049 | } |
4050 | ||
52a46617 | 4051 | static u64 mk_cr_64(u64 curr_cr, u32 new_val) |
5fdbf976 | 4052 | { |
52a46617 | 4053 | return (curr_cr & ~((1ULL << 32) - 1)) | new_val; |
5fdbf976 MT |
4054 | } |
4055 | ||
52a46617 | 4056 | static unsigned long emulator_get_cr(int cr, struct kvm_vcpu *vcpu) |
bbd9b64e | 4057 | { |
52a46617 GN |
4058 | unsigned long value; |
4059 | ||
4060 | switch (cr) { | |
4061 | case 0: | |
4062 | value = kvm_read_cr0(vcpu); | |
4063 | break; | |
4064 | case 2: | |
4065 | value = vcpu->arch.cr2; | |
4066 | break; | |
4067 | case 3: | |
4068 | value = vcpu->arch.cr3; | |
4069 | break; | |
4070 | case 4: | |
4071 | value = kvm_read_cr4(vcpu); | |
4072 | break; | |
4073 | case 8: | |
4074 | value = kvm_get_cr8(vcpu); | |
4075 | break; | |
4076 | default: | |
4077 | vcpu_printf(vcpu, "%s: unexpected cr %u\n", __func__, cr); | |
4078 | return 0; | |
4079 | } | |
4080 | ||
4081 | return value; | |
4082 | } | |
4083 | ||
0f12244f | 4084 | static int emulator_set_cr(int cr, unsigned long val, struct kvm_vcpu *vcpu) |
52a46617 | 4085 | { |
0f12244f GN |
4086 | int res = 0; |
4087 | ||
52a46617 GN |
4088 | switch (cr) { |
4089 | case 0: | |
49a9b07e | 4090 | res = kvm_set_cr0(vcpu, mk_cr_64(kvm_read_cr0(vcpu), val)); |
52a46617 GN |
4091 | break; |
4092 | case 2: | |
4093 | vcpu->arch.cr2 = val; | |
4094 | break; | |
4095 | case 3: | |
2390218b | 4096 | res = kvm_set_cr3(vcpu, val); |
52a46617 GN |
4097 | break; |
4098 | case 4: | |
a83b29c6 | 4099 | res = kvm_set_cr4(vcpu, mk_cr_64(kvm_read_cr4(vcpu), val)); |
52a46617 GN |
4100 | break; |
4101 | case 8: | |
0f12244f | 4102 | res = __kvm_set_cr8(vcpu, val & 0xfUL); |
52a46617 GN |
4103 | break; |
4104 | default: | |
4105 | vcpu_printf(vcpu, "%s: unexpected cr %u\n", __func__, cr); | |
0f12244f | 4106 | res = -1; |
52a46617 | 4107 | } |
0f12244f GN |
4108 | |
4109 | return res; | |
52a46617 GN |
4110 | } |
4111 | ||
9c537244 GN |
4112 | static int emulator_get_cpl(struct kvm_vcpu *vcpu) |
4113 | { | |
4114 | return kvm_x86_ops->get_cpl(vcpu); | |
4115 | } | |
4116 | ||
2dafc6c2 GN |
4117 | static void emulator_get_gdt(struct desc_ptr *dt, struct kvm_vcpu *vcpu) |
4118 | { | |
4119 | kvm_x86_ops->get_gdt(vcpu, dt); | |
4120 | } | |
4121 | ||
160ce1f1 MG |
4122 | static void emulator_get_idt(struct desc_ptr *dt, struct kvm_vcpu *vcpu) |
4123 | { | |
4124 | kvm_x86_ops->get_idt(vcpu, dt); | |
4125 | } | |
4126 | ||
5951c442 GN |
4127 | static unsigned long emulator_get_cached_segment_base(int seg, |
4128 | struct kvm_vcpu *vcpu) | |
4129 | { | |
4130 | return get_segment_base(vcpu, seg); | |
4131 | } | |
4132 | ||
2dafc6c2 GN |
4133 | static bool emulator_get_cached_descriptor(struct desc_struct *desc, int seg, |
4134 | struct kvm_vcpu *vcpu) | |
4135 | { | |
4136 | struct kvm_segment var; | |
4137 | ||
4138 | kvm_get_segment(vcpu, &var, seg); | |
4139 | ||
4140 | if (var.unusable) | |
4141 | return false; | |
4142 | ||
4143 | if (var.g) | |
4144 | var.limit >>= 12; | |
4145 | set_desc_limit(desc, var.limit); | |
4146 | set_desc_base(desc, (unsigned long)var.base); | |
4147 | desc->type = var.type; | |
4148 | desc->s = var.s; | |
4149 | desc->dpl = var.dpl; | |
4150 | desc->p = var.present; | |
4151 | desc->avl = var.avl; | |
4152 | desc->l = var.l; | |
4153 | desc->d = var.db; | |
4154 | desc->g = var.g; | |
4155 | ||
4156 | return true; | |
4157 | } | |
4158 | ||
4159 | static void emulator_set_cached_descriptor(struct desc_struct *desc, int seg, | |
4160 | struct kvm_vcpu *vcpu) | |
4161 | { | |
4162 | struct kvm_segment var; | |
4163 | ||
4164 | /* needed to preserve selector */ | |
4165 | kvm_get_segment(vcpu, &var, seg); | |
4166 | ||
4167 | var.base = get_desc_base(desc); | |
4168 | var.limit = get_desc_limit(desc); | |
4169 | if (desc->g) | |
4170 | var.limit = (var.limit << 12) | 0xfff; | |
4171 | var.type = desc->type; | |
4172 | var.present = desc->p; | |
4173 | var.dpl = desc->dpl; | |
4174 | var.db = desc->d; | |
4175 | var.s = desc->s; | |
4176 | var.l = desc->l; | |
4177 | var.g = desc->g; | |
4178 | var.avl = desc->avl; | |
4179 | var.present = desc->p; | |
4180 | var.unusable = !var.present; | |
4181 | var.padding = 0; | |
4182 | ||
4183 | kvm_set_segment(vcpu, &var, seg); | |
4184 | return; | |
4185 | } | |
4186 | ||
4187 | static u16 emulator_get_segment_selector(int seg, struct kvm_vcpu *vcpu) | |
4188 | { | |
4189 | struct kvm_segment kvm_seg; | |
4190 | ||
4191 | kvm_get_segment(vcpu, &kvm_seg, seg); | |
4192 | return kvm_seg.selector; | |
4193 | } | |
4194 | ||
4195 | static void emulator_set_segment_selector(u16 sel, int seg, | |
4196 | struct kvm_vcpu *vcpu) | |
4197 | { | |
4198 | struct kvm_segment kvm_seg; | |
4199 | ||
4200 | kvm_get_segment(vcpu, &kvm_seg, seg); | |
4201 | kvm_seg.selector = sel; | |
4202 | kvm_set_segment(vcpu, &kvm_seg, seg); | |
4203 | } | |
4204 | ||
14af3f3c | 4205 | static struct x86_emulate_ops emulate_ops = { |
1871c602 | 4206 | .read_std = kvm_read_guest_virt_system, |
2dafc6c2 | 4207 | .write_std = kvm_write_guest_virt_system, |
1871c602 | 4208 | .fetch = kvm_fetch_guest_virt, |
bbd9b64e CO |
4209 | .read_emulated = emulator_read_emulated, |
4210 | .write_emulated = emulator_write_emulated, | |
4211 | .cmpxchg_emulated = emulator_cmpxchg_emulated, | |
cf8f70bf GN |
4212 | .pio_in_emulated = emulator_pio_in_emulated, |
4213 | .pio_out_emulated = emulator_pio_out_emulated, | |
2dafc6c2 GN |
4214 | .get_cached_descriptor = emulator_get_cached_descriptor, |
4215 | .set_cached_descriptor = emulator_set_cached_descriptor, | |
4216 | .get_segment_selector = emulator_get_segment_selector, | |
4217 | .set_segment_selector = emulator_set_segment_selector, | |
5951c442 | 4218 | .get_cached_segment_base = emulator_get_cached_segment_base, |
2dafc6c2 | 4219 | .get_gdt = emulator_get_gdt, |
160ce1f1 | 4220 | .get_idt = emulator_get_idt, |
52a46617 GN |
4221 | .get_cr = emulator_get_cr, |
4222 | .set_cr = emulator_set_cr, | |
9c537244 | 4223 | .cpl = emulator_get_cpl, |
35aa5375 GN |
4224 | .get_dr = emulator_get_dr, |
4225 | .set_dr = emulator_set_dr, | |
3fb1b5db GN |
4226 | .set_msr = kvm_set_msr, |
4227 | .get_msr = kvm_get_msr, | |
bbd9b64e CO |
4228 | }; |
4229 | ||
5fdbf976 MT |
4230 | static void cache_all_regs(struct kvm_vcpu *vcpu) |
4231 | { | |
4232 | kvm_register_read(vcpu, VCPU_REGS_RAX); | |
4233 | kvm_register_read(vcpu, VCPU_REGS_RSP); | |
4234 | kvm_register_read(vcpu, VCPU_REGS_RIP); | |
4235 | vcpu->arch.regs_dirty = ~0; | |
4236 | } | |
4237 | ||
95cb2295 GN |
4238 | static void toggle_interruptibility(struct kvm_vcpu *vcpu, u32 mask) |
4239 | { | |
4240 | u32 int_shadow = kvm_x86_ops->get_interrupt_shadow(vcpu, mask); | |
4241 | /* | |
4242 | * an sti; sti; sequence only disable interrupts for the first | |
4243 | * instruction. So, if the last instruction, be it emulated or | |
4244 | * not, left the system with the INT_STI flag enabled, it | |
4245 | * means that the last instruction is an sti. We should not | |
4246 | * leave the flag on in this case. The same goes for mov ss | |
4247 | */ | |
4248 | if (!(int_shadow & mask)) | |
4249 | kvm_x86_ops->set_interrupt_shadow(vcpu, mask); | |
4250 | } | |
4251 | ||
54b8486f GN |
4252 | static void inject_emulated_exception(struct kvm_vcpu *vcpu) |
4253 | { | |
4254 | struct x86_emulate_ctxt *ctxt = &vcpu->arch.emulate_ctxt; | |
da9cb575 | 4255 | if (ctxt->exception.vector == PF_VECTOR) |
6389ee94 | 4256 | kvm_propagate_fault(vcpu, &ctxt->exception); |
da9cb575 AK |
4257 | else if (ctxt->exception.error_code_valid) |
4258 | kvm_queue_exception_e(vcpu, ctxt->exception.vector, | |
4259 | ctxt->exception.error_code); | |
54b8486f | 4260 | else |
da9cb575 | 4261 | kvm_queue_exception(vcpu, ctxt->exception.vector); |
54b8486f GN |
4262 | } |
4263 | ||
8ec4722d MG |
4264 | static void init_emulate_ctxt(struct kvm_vcpu *vcpu) |
4265 | { | |
4266 | struct decode_cache *c = &vcpu->arch.emulate_ctxt.decode; | |
4267 | int cs_db, cs_l; | |
4268 | ||
4269 | cache_all_regs(vcpu); | |
4270 | ||
4271 | kvm_x86_ops->get_cs_db_l_bits(vcpu, &cs_db, &cs_l); | |
4272 | ||
4273 | vcpu->arch.emulate_ctxt.vcpu = vcpu; | |
4274 | vcpu->arch.emulate_ctxt.eflags = kvm_x86_ops->get_rflags(vcpu); | |
4275 | vcpu->arch.emulate_ctxt.eip = kvm_rip_read(vcpu); | |
4276 | vcpu->arch.emulate_ctxt.mode = | |
4277 | (!is_protmode(vcpu)) ? X86EMUL_MODE_REAL : | |
4278 | (vcpu->arch.emulate_ctxt.eflags & X86_EFLAGS_VM) | |
4279 | ? X86EMUL_MODE_VM86 : cs_l | |
4280 | ? X86EMUL_MODE_PROT64 : cs_db | |
4281 | ? X86EMUL_MODE_PROT32 : X86EMUL_MODE_PROT16; | |
4282 | memset(c, 0, sizeof(struct decode_cache)); | |
4283 | memcpy(c->regs, vcpu->arch.regs, sizeof c->regs); | |
4284 | } | |
4285 | ||
63995653 MG |
4286 | int kvm_inject_realmode_interrupt(struct kvm_vcpu *vcpu, int irq) |
4287 | { | |
4288 | struct decode_cache *c = &vcpu->arch.emulate_ctxt.decode; | |
4289 | int ret; | |
4290 | ||
4291 | init_emulate_ctxt(vcpu); | |
4292 | ||
4293 | vcpu->arch.emulate_ctxt.decode.op_bytes = 2; | |
4294 | vcpu->arch.emulate_ctxt.decode.ad_bytes = 2; | |
4295 | vcpu->arch.emulate_ctxt.decode.eip = vcpu->arch.emulate_ctxt.eip; | |
4296 | ret = emulate_int_real(&vcpu->arch.emulate_ctxt, &emulate_ops, irq); | |
4297 | ||
4298 | if (ret != X86EMUL_CONTINUE) | |
4299 | return EMULATE_FAIL; | |
4300 | ||
4301 | vcpu->arch.emulate_ctxt.eip = c->eip; | |
4302 | memcpy(vcpu->arch.regs, c->regs, sizeof c->regs); | |
4303 | kvm_rip_write(vcpu, vcpu->arch.emulate_ctxt.eip); | |
4304 | kvm_x86_ops->set_rflags(vcpu, vcpu->arch.emulate_ctxt.eflags); | |
4305 | ||
4306 | if (irq == NMI_VECTOR) | |
4307 | vcpu->arch.nmi_pending = false; | |
4308 | else | |
4309 | vcpu->arch.interrupt.pending = false; | |
4310 | ||
4311 | return EMULATE_DONE; | |
4312 | } | |
4313 | EXPORT_SYMBOL_GPL(kvm_inject_realmode_interrupt); | |
4314 | ||
6d77dbfc GN |
4315 | static int handle_emulation_failure(struct kvm_vcpu *vcpu) |
4316 | { | |
6d77dbfc GN |
4317 | ++vcpu->stat.insn_emulation_fail; |
4318 | trace_kvm_emulate_insn_failed(vcpu); | |
4319 | vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; | |
4320 | vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_EMULATION; | |
4321 | vcpu->run->internal.ndata = 0; | |
4322 | kvm_queue_exception(vcpu, UD_VECTOR); | |
4323 | return EMULATE_FAIL; | |
4324 | } | |
4325 | ||
a6f177ef GN |
4326 | static bool reexecute_instruction(struct kvm_vcpu *vcpu, gva_t gva) |
4327 | { | |
4328 | gpa_t gpa; | |
4329 | ||
68be0803 GN |
4330 | if (tdp_enabled) |
4331 | return false; | |
4332 | ||
a6f177ef GN |
4333 | /* |
4334 | * if emulation was due to access to shadowed page table | |
4335 | * and it failed try to unshadow page and re-entetr the | |
4336 | * guest to let CPU execute the instruction. | |
4337 | */ | |
4338 | if (kvm_mmu_unprotect_page_virt(vcpu, gva)) | |
4339 | return true; | |
4340 | ||
4341 | gpa = kvm_mmu_gva_to_gpa_system(vcpu, gva, NULL); | |
4342 | ||
4343 | if (gpa == UNMAPPED_GVA) | |
4344 | return true; /* let cpu generate fault */ | |
4345 | ||
4346 | if (!kvm_is_error_hva(gfn_to_hva(vcpu->kvm, gpa >> PAGE_SHIFT))) | |
4347 | return true; | |
4348 | ||
4349 | return false; | |
4350 | } | |
4351 | ||
bbd9b64e | 4352 | int emulate_instruction(struct kvm_vcpu *vcpu, |
bbd9b64e CO |
4353 | unsigned long cr2, |
4354 | u16 error_code, | |
571008da | 4355 | int emulation_type) |
bbd9b64e | 4356 | { |
95cb2295 | 4357 | int r; |
4d2179e1 | 4358 | struct decode_cache *c = &vcpu->arch.emulate_ctxt.decode; |
bbd9b64e | 4359 | |
26eef70c | 4360 | kvm_clear_exception_queue(vcpu); |
ad312c7c | 4361 | vcpu->arch.mmio_fault_cr2 = cr2; |
5fdbf976 | 4362 | /* |
56e82318 | 4363 | * TODO: fix emulate.c to use guest_read/write_register |
5fdbf976 MT |
4364 | * instead of direct ->regs accesses, can save hundred cycles |
4365 | * on Intel for instructions that don't read/change RSP, for | |
4366 | * for example. | |
4367 | */ | |
4368 | cache_all_regs(vcpu); | |
bbd9b64e | 4369 | |
571008da | 4370 | if (!(emulation_type & EMULTYPE_NO_DECODE)) { |
8ec4722d | 4371 | init_emulate_ctxt(vcpu); |
95cb2295 | 4372 | vcpu->arch.emulate_ctxt.interruptibility = 0; |
da9cb575 | 4373 | vcpu->arch.emulate_ctxt.have_exception = false; |
4fc40f07 | 4374 | vcpu->arch.emulate_ctxt.perm_ok = false; |
bbd9b64e | 4375 | |
9aabc88f | 4376 | r = x86_decode_insn(&vcpu->arch.emulate_ctxt); |
d47f00a6 JR |
4377 | if (r == X86EMUL_PROPAGATE_FAULT) |
4378 | goto done; | |
bbd9b64e | 4379 | |
e46479f8 | 4380 | trace_kvm_emulate_insn_start(vcpu); |
571008da | 4381 | |
0cb5762e AP |
4382 | /* Only allow emulation of specific instructions on #UD |
4383 | * (namely VMMCALL, sysenter, sysexit, syscall)*/ | |
0cb5762e AP |
4384 | if (emulation_type & EMULTYPE_TRAP_UD) { |
4385 | if (!c->twobyte) | |
4386 | return EMULATE_FAIL; | |
4387 | switch (c->b) { | |
4388 | case 0x01: /* VMMCALL */ | |
4389 | if (c->modrm_mod != 3 || c->modrm_rm != 1) | |
4390 | return EMULATE_FAIL; | |
4391 | break; | |
4392 | case 0x34: /* sysenter */ | |
4393 | case 0x35: /* sysexit */ | |
4394 | if (c->modrm_mod != 0 || c->modrm_rm != 0) | |
4395 | return EMULATE_FAIL; | |
4396 | break; | |
4397 | case 0x05: /* syscall */ | |
4398 | if (c->modrm_mod != 0 || c->modrm_rm != 0) | |
4399 | return EMULATE_FAIL; | |
4400 | break; | |
4401 | default: | |
4402 | return EMULATE_FAIL; | |
4403 | } | |
4404 | ||
4405 | if (!(c->modrm_reg == 0 || c->modrm_reg == 3)) | |
4406 | return EMULATE_FAIL; | |
4407 | } | |
571008da | 4408 | |
f2b5756b | 4409 | ++vcpu->stat.insn_emulation; |
bbd9b64e | 4410 | if (r) { |
a6f177ef | 4411 | if (reexecute_instruction(vcpu, cr2)) |
bbd9b64e | 4412 | return EMULATE_DONE; |
6d77dbfc GN |
4413 | if (emulation_type & EMULTYPE_SKIP) |
4414 | return EMULATE_FAIL; | |
4415 | return handle_emulation_failure(vcpu); | |
bbd9b64e CO |
4416 | } |
4417 | } | |
4418 | ||
ba8afb6b GN |
4419 | if (emulation_type & EMULTYPE_SKIP) { |
4420 | kvm_rip_write(vcpu, vcpu->arch.emulate_ctxt.decode.eip); | |
4421 | return EMULATE_DONE; | |
4422 | } | |
4423 | ||
4d2179e1 GN |
4424 | /* this is needed for vmware backdor interface to work since it |
4425 | changes registers values during IO operation */ | |
4426 | memcpy(c->regs, vcpu->arch.regs, sizeof c->regs); | |
4427 | ||
5cd21917 | 4428 | restart: |
9aabc88f | 4429 | r = x86_emulate_insn(&vcpu->arch.emulate_ctxt); |
bbd9b64e | 4430 | |
d2ddd1c4 | 4431 | if (r == EMULATION_FAILED) { |
a6f177ef | 4432 | if (reexecute_instruction(vcpu, cr2)) |
c3cd7ffa GN |
4433 | return EMULATE_DONE; |
4434 | ||
6d77dbfc | 4435 | return handle_emulation_failure(vcpu); |
bbd9b64e CO |
4436 | } |
4437 | ||
d47f00a6 | 4438 | done: |
da9cb575 | 4439 | if (vcpu->arch.emulate_ctxt.have_exception) { |
54b8486f | 4440 | inject_emulated_exception(vcpu); |
d2ddd1c4 GN |
4441 | r = EMULATE_DONE; |
4442 | } else if (vcpu->arch.pio.count) { | |
3457e419 GN |
4443 | if (!vcpu->arch.pio.in) |
4444 | vcpu->arch.pio.count = 0; | |
e85d28f8 GN |
4445 | r = EMULATE_DO_MMIO; |
4446 | } else if (vcpu->mmio_needed) { | |
3457e419 GN |
4447 | if (vcpu->mmio_is_write) |
4448 | vcpu->mmio_needed = 0; | |
e85d28f8 | 4449 | r = EMULATE_DO_MMIO; |
d2ddd1c4 | 4450 | } else if (r == EMULATION_RESTART) |
5cd21917 | 4451 | goto restart; |
d2ddd1c4 GN |
4452 | else |
4453 | r = EMULATE_DONE; | |
f850e2e6 | 4454 | |
e85d28f8 GN |
4455 | toggle_interruptibility(vcpu, vcpu->arch.emulate_ctxt.interruptibility); |
4456 | kvm_x86_ops->set_rflags(vcpu, vcpu->arch.emulate_ctxt.eflags); | |
3842d135 | 4457 | kvm_make_request(KVM_REQ_EVENT, vcpu); |
e85d28f8 GN |
4458 | memcpy(vcpu->arch.regs, c->regs, sizeof c->regs); |
4459 | kvm_rip_write(vcpu, vcpu->arch.emulate_ctxt.eip); | |
4460 | ||
4461 | return r; | |
de7d789a | 4462 | } |
bbd9b64e | 4463 | EXPORT_SYMBOL_GPL(emulate_instruction); |
de7d789a | 4464 | |
cf8f70bf | 4465 | int kvm_fast_pio_out(struct kvm_vcpu *vcpu, int size, unsigned short port) |
de7d789a | 4466 | { |
cf8f70bf GN |
4467 | unsigned long val = kvm_register_read(vcpu, VCPU_REGS_RAX); |
4468 | int ret = emulator_pio_out_emulated(size, port, &val, 1, vcpu); | |
4469 | /* do not return to emulator after return from userspace */ | |
7972995b | 4470 | vcpu->arch.pio.count = 0; |
de7d789a CO |
4471 | return ret; |
4472 | } | |
cf8f70bf | 4473 | EXPORT_SYMBOL_GPL(kvm_fast_pio_out); |
de7d789a | 4474 | |
8cfdc000 ZA |
4475 | static void tsc_bad(void *info) |
4476 | { | |
4477 | __get_cpu_var(cpu_tsc_khz) = 0; | |
4478 | } | |
4479 | ||
4480 | static void tsc_khz_changed(void *data) | |
c8076604 | 4481 | { |
8cfdc000 ZA |
4482 | struct cpufreq_freqs *freq = data; |
4483 | unsigned long khz = 0; | |
4484 | ||
4485 | if (data) | |
4486 | khz = freq->new; | |
4487 | else if (!boot_cpu_has(X86_FEATURE_CONSTANT_TSC)) | |
4488 | khz = cpufreq_quick_get(raw_smp_processor_id()); | |
4489 | if (!khz) | |
4490 | khz = tsc_khz; | |
4491 | __get_cpu_var(cpu_tsc_khz) = khz; | |
c8076604 GH |
4492 | } |
4493 | ||
c8076604 GH |
4494 | static int kvmclock_cpufreq_notifier(struct notifier_block *nb, unsigned long val, |
4495 | void *data) | |
4496 | { | |
4497 | struct cpufreq_freqs *freq = data; | |
4498 | struct kvm *kvm; | |
4499 | struct kvm_vcpu *vcpu; | |
4500 | int i, send_ipi = 0; | |
4501 | ||
8cfdc000 ZA |
4502 | /* |
4503 | * We allow guests to temporarily run on slowing clocks, | |
4504 | * provided we notify them after, or to run on accelerating | |
4505 | * clocks, provided we notify them before. Thus time never | |
4506 | * goes backwards. | |
4507 | * | |
4508 | * However, we have a problem. We can't atomically update | |
4509 | * the frequency of a given CPU from this function; it is | |
4510 | * merely a notifier, which can be called from any CPU. | |
4511 | * Changing the TSC frequency at arbitrary points in time | |
4512 | * requires a recomputation of local variables related to | |
4513 | * the TSC for each VCPU. We must flag these local variables | |
4514 | * to be updated and be sure the update takes place with the | |
4515 | * new frequency before any guests proceed. | |
4516 | * | |
4517 | * Unfortunately, the combination of hotplug CPU and frequency | |
4518 | * change creates an intractable locking scenario; the order | |
4519 | * of when these callouts happen is undefined with respect to | |
4520 | * CPU hotplug, and they can race with each other. As such, | |
4521 | * merely setting per_cpu(cpu_tsc_khz) = X during a hotadd is | |
4522 | * undefined; you can actually have a CPU frequency change take | |
4523 | * place in between the computation of X and the setting of the | |
4524 | * variable. To protect against this problem, all updates of | |
4525 | * the per_cpu tsc_khz variable are done in an interrupt | |
4526 | * protected IPI, and all callers wishing to update the value | |
4527 | * must wait for a synchronous IPI to complete (which is trivial | |
4528 | * if the caller is on the CPU already). This establishes the | |
4529 | * necessary total order on variable updates. | |
4530 | * | |
4531 | * Note that because a guest time update may take place | |
4532 | * anytime after the setting of the VCPU's request bit, the | |
4533 | * correct TSC value must be set before the request. However, | |
4534 | * to ensure the update actually makes it to any guest which | |
4535 | * starts running in hardware virtualization between the set | |
4536 | * and the acquisition of the spinlock, we must also ping the | |
4537 | * CPU after setting the request bit. | |
4538 | * | |
4539 | */ | |
4540 | ||
c8076604 GH |
4541 | if (val == CPUFREQ_PRECHANGE && freq->old > freq->new) |
4542 | return 0; | |
4543 | if (val == CPUFREQ_POSTCHANGE && freq->old < freq->new) | |
4544 | return 0; | |
8cfdc000 ZA |
4545 | |
4546 | smp_call_function_single(freq->cpu, tsc_khz_changed, freq, 1); | |
c8076604 GH |
4547 | |
4548 | spin_lock(&kvm_lock); | |
4549 | list_for_each_entry(kvm, &vm_list, vm_list) { | |
988a2cae | 4550 | kvm_for_each_vcpu(i, vcpu, kvm) { |
c8076604 GH |
4551 | if (vcpu->cpu != freq->cpu) |
4552 | continue; | |
c285545f | 4553 | kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu); |
c8076604 | 4554 | if (vcpu->cpu != smp_processor_id()) |
8cfdc000 | 4555 | send_ipi = 1; |
c8076604 GH |
4556 | } |
4557 | } | |
4558 | spin_unlock(&kvm_lock); | |
4559 | ||
4560 | if (freq->old < freq->new && send_ipi) { | |
4561 | /* | |
4562 | * We upscale the frequency. Must make the guest | |
4563 | * doesn't see old kvmclock values while running with | |
4564 | * the new frequency, otherwise we risk the guest sees | |
4565 | * time go backwards. | |
4566 | * | |
4567 | * In case we update the frequency for another cpu | |
4568 | * (which might be in guest context) send an interrupt | |
4569 | * to kick the cpu out of guest context. Next time | |
4570 | * guest context is entered kvmclock will be updated, | |
4571 | * so the guest will not see stale values. | |
4572 | */ | |
8cfdc000 | 4573 | smp_call_function_single(freq->cpu, tsc_khz_changed, freq, 1); |
c8076604 GH |
4574 | } |
4575 | return 0; | |
4576 | } | |
4577 | ||
4578 | static struct notifier_block kvmclock_cpufreq_notifier_block = { | |
8cfdc000 ZA |
4579 | .notifier_call = kvmclock_cpufreq_notifier |
4580 | }; | |
4581 | ||
4582 | static int kvmclock_cpu_notifier(struct notifier_block *nfb, | |
4583 | unsigned long action, void *hcpu) | |
4584 | { | |
4585 | unsigned int cpu = (unsigned long)hcpu; | |
4586 | ||
4587 | switch (action) { | |
4588 | case CPU_ONLINE: | |
4589 | case CPU_DOWN_FAILED: | |
4590 | smp_call_function_single(cpu, tsc_khz_changed, NULL, 1); | |
4591 | break; | |
4592 | case CPU_DOWN_PREPARE: | |
4593 | smp_call_function_single(cpu, tsc_bad, NULL, 1); | |
4594 | break; | |
4595 | } | |
4596 | return NOTIFY_OK; | |
4597 | } | |
4598 | ||
4599 | static struct notifier_block kvmclock_cpu_notifier_block = { | |
4600 | .notifier_call = kvmclock_cpu_notifier, | |
4601 | .priority = -INT_MAX | |
c8076604 GH |
4602 | }; |
4603 | ||
b820cc0c ZA |
4604 | static void kvm_timer_init(void) |
4605 | { | |
4606 | int cpu; | |
4607 | ||
c285545f | 4608 | max_tsc_khz = tsc_khz; |
8cfdc000 | 4609 | register_hotcpu_notifier(&kvmclock_cpu_notifier_block); |
b820cc0c | 4610 | if (!boot_cpu_has(X86_FEATURE_CONSTANT_TSC)) { |
c285545f ZA |
4611 | #ifdef CONFIG_CPU_FREQ |
4612 | struct cpufreq_policy policy; | |
4613 | memset(&policy, 0, sizeof(policy)); | |
3e26f230 AK |
4614 | cpu = get_cpu(); |
4615 | cpufreq_get_policy(&policy, cpu); | |
c285545f ZA |
4616 | if (policy.cpuinfo.max_freq) |
4617 | max_tsc_khz = policy.cpuinfo.max_freq; | |
3e26f230 | 4618 | put_cpu(); |
c285545f | 4619 | #endif |
b820cc0c ZA |
4620 | cpufreq_register_notifier(&kvmclock_cpufreq_notifier_block, |
4621 | CPUFREQ_TRANSITION_NOTIFIER); | |
4622 | } | |
c285545f | 4623 | pr_debug("kvm: max_tsc_khz = %ld\n", max_tsc_khz); |
8cfdc000 ZA |
4624 | for_each_online_cpu(cpu) |
4625 | smp_call_function_single(cpu, tsc_khz_changed, NULL, 1); | |
b820cc0c ZA |
4626 | } |
4627 | ||
ff9d07a0 ZY |
4628 | static DEFINE_PER_CPU(struct kvm_vcpu *, current_vcpu); |
4629 | ||
4630 | static int kvm_is_in_guest(void) | |
4631 | { | |
4632 | return percpu_read(current_vcpu) != NULL; | |
4633 | } | |
4634 | ||
4635 | static int kvm_is_user_mode(void) | |
4636 | { | |
4637 | int user_mode = 3; | |
dcf46b94 | 4638 | |
ff9d07a0 ZY |
4639 | if (percpu_read(current_vcpu)) |
4640 | user_mode = kvm_x86_ops->get_cpl(percpu_read(current_vcpu)); | |
dcf46b94 | 4641 | |
ff9d07a0 ZY |
4642 | return user_mode != 0; |
4643 | } | |
4644 | ||
4645 | static unsigned long kvm_get_guest_ip(void) | |
4646 | { | |
4647 | unsigned long ip = 0; | |
dcf46b94 | 4648 | |
ff9d07a0 ZY |
4649 | if (percpu_read(current_vcpu)) |
4650 | ip = kvm_rip_read(percpu_read(current_vcpu)); | |
dcf46b94 | 4651 | |
ff9d07a0 ZY |
4652 | return ip; |
4653 | } | |
4654 | ||
4655 | static struct perf_guest_info_callbacks kvm_guest_cbs = { | |
4656 | .is_in_guest = kvm_is_in_guest, | |
4657 | .is_user_mode = kvm_is_user_mode, | |
4658 | .get_guest_ip = kvm_get_guest_ip, | |
4659 | }; | |
4660 | ||
4661 | void kvm_before_handle_nmi(struct kvm_vcpu *vcpu) | |
4662 | { | |
4663 | percpu_write(current_vcpu, vcpu); | |
4664 | } | |
4665 | EXPORT_SYMBOL_GPL(kvm_before_handle_nmi); | |
4666 | ||
4667 | void kvm_after_handle_nmi(struct kvm_vcpu *vcpu) | |
4668 | { | |
4669 | percpu_write(current_vcpu, NULL); | |
4670 | } | |
4671 | EXPORT_SYMBOL_GPL(kvm_after_handle_nmi); | |
4672 | ||
f8c16bba | 4673 | int kvm_arch_init(void *opaque) |
043405e1 | 4674 | { |
b820cc0c | 4675 | int r; |
f8c16bba ZX |
4676 | struct kvm_x86_ops *ops = (struct kvm_x86_ops *)opaque; |
4677 | ||
f8c16bba ZX |
4678 | if (kvm_x86_ops) { |
4679 | printk(KERN_ERR "kvm: already loaded the other module\n"); | |
56c6d28a ZX |
4680 | r = -EEXIST; |
4681 | goto out; | |
f8c16bba ZX |
4682 | } |
4683 | ||
4684 | if (!ops->cpu_has_kvm_support()) { | |
4685 | printk(KERN_ERR "kvm: no hardware support\n"); | |
56c6d28a ZX |
4686 | r = -EOPNOTSUPP; |
4687 | goto out; | |
f8c16bba ZX |
4688 | } |
4689 | if (ops->disabled_by_bios()) { | |
4690 | printk(KERN_ERR "kvm: disabled by bios\n"); | |
56c6d28a ZX |
4691 | r = -EOPNOTSUPP; |
4692 | goto out; | |
f8c16bba ZX |
4693 | } |
4694 | ||
97db56ce AK |
4695 | r = kvm_mmu_module_init(); |
4696 | if (r) | |
4697 | goto out; | |
4698 | ||
4699 | kvm_init_msr_list(); | |
4700 | ||
f8c16bba | 4701 | kvm_x86_ops = ops; |
56c6d28a | 4702 | kvm_mmu_set_nonpresent_ptes(0ull, 0ull); |
7b52345e | 4703 | kvm_mmu_set_mask_ptes(PT_USER_MASK, PT_ACCESSED_MASK, |
4b12f0de | 4704 | PT_DIRTY_MASK, PT64_NX_MASK, 0); |
c8076604 | 4705 | |
b820cc0c | 4706 | kvm_timer_init(); |
c8076604 | 4707 | |
ff9d07a0 ZY |
4708 | perf_register_guest_info_callbacks(&kvm_guest_cbs); |
4709 | ||
2acf923e DC |
4710 | if (cpu_has_xsave) |
4711 | host_xcr0 = xgetbv(XCR_XFEATURE_ENABLED_MASK); | |
4712 | ||
f8c16bba | 4713 | return 0; |
56c6d28a ZX |
4714 | |
4715 | out: | |
56c6d28a | 4716 | return r; |
043405e1 | 4717 | } |
8776e519 | 4718 | |
f8c16bba ZX |
4719 | void kvm_arch_exit(void) |
4720 | { | |
ff9d07a0 ZY |
4721 | perf_unregister_guest_info_callbacks(&kvm_guest_cbs); |
4722 | ||
888d256e JK |
4723 | if (!boot_cpu_has(X86_FEATURE_CONSTANT_TSC)) |
4724 | cpufreq_unregister_notifier(&kvmclock_cpufreq_notifier_block, | |
4725 | CPUFREQ_TRANSITION_NOTIFIER); | |
8cfdc000 | 4726 | unregister_hotcpu_notifier(&kvmclock_cpu_notifier_block); |
f8c16bba | 4727 | kvm_x86_ops = NULL; |
56c6d28a ZX |
4728 | kvm_mmu_module_exit(); |
4729 | } | |
f8c16bba | 4730 | |
8776e519 HB |
4731 | int kvm_emulate_halt(struct kvm_vcpu *vcpu) |
4732 | { | |
4733 | ++vcpu->stat.halt_exits; | |
4734 | if (irqchip_in_kernel(vcpu->kvm)) { | |
a4535290 | 4735 | vcpu->arch.mp_state = KVM_MP_STATE_HALTED; |
8776e519 HB |
4736 | return 1; |
4737 | } else { | |
4738 | vcpu->run->exit_reason = KVM_EXIT_HLT; | |
4739 | return 0; | |
4740 | } | |
4741 | } | |
4742 | EXPORT_SYMBOL_GPL(kvm_emulate_halt); | |
4743 | ||
2f333bcb MT |
4744 | static inline gpa_t hc_gpa(struct kvm_vcpu *vcpu, unsigned long a0, |
4745 | unsigned long a1) | |
4746 | { | |
4747 | if (is_long_mode(vcpu)) | |
4748 | return a0; | |
4749 | else | |
4750 | return a0 | ((gpa_t)a1 << 32); | |
4751 | } | |
4752 | ||
55cd8e5a GN |
4753 | int kvm_hv_hypercall(struct kvm_vcpu *vcpu) |
4754 | { | |
4755 | u64 param, ingpa, outgpa, ret; | |
4756 | uint16_t code, rep_idx, rep_cnt, res = HV_STATUS_SUCCESS, rep_done = 0; | |
4757 | bool fast, longmode; | |
4758 | int cs_db, cs_l; | |
4759 | ||
4760 | /* | |
4761 | * hypercall generates UD from non zero cpl and real mode | |
4762 | * per HYPER-V spec | |
4763 | */ | |
3eeb3288 | 4764 | if (kvm_x86_ops->get_cpl(vcpu) != 0 || !is_protmode(vcpu)) { |
55cd8e5a GN |
4765 | kvm_queue_exception(vcpu, UD_VECTOR); |
4766 | return 0; | |
4767 | } | |
4768 | ||
4769 | kvm_x86_ops->get_cs_db_l_bits(vcpu, &cs_db, &cs_l); | |
4770 | longmode = is_long_mode(vcpu) && cs_l == 1; | |
4771 | ||
4772 | if (!longmode) { | |
ccd46936 GN |
4773 | param = ((u64)kvm_register_read(vcpu, VCPU_REGS_RDX) << 32) | |
4774 | (kvm_register_read(vcpu, VCPU_REGS_RAX) & 0xffffffff); | |
4775 | ingpa = ((u64)kvm_register_read(vcpu, VCPU_REGS_RBX) << 32) | | |
4776 | (kvm_register_read(vcpu, VCPU_REGS_RCX) & 0xffffffff); | |
4777 | outgpa = ((u64)kvm_register_read(vcpu, VCPU_REGS_RDI) << 32) | | |
4778 | (kvm_register_read(vcpu, VCPU_REGS_RSI) & 0xffffffff); | |
55cd8e5a GN |
4779 | } |
4780 | #ifdef CONFIG_X86_64 | |
4781 | else { | |
4782 | param = kvm_register_read(vcpu, VCPU_REGS_RCX); | |
4783 | ingpa = kvm_register_read(vcpu, VCPU_REGS_RDX); | |
4784 | outgpa = kvm_register_read(vcpu, VCPU_REGS_R8); | |
4785 | } | |
4786 | #endif | |
4787 | ||
4788 | code = param & 0xffff; | |
4789 | fast = (param >> 16) & 0x1; | |
4790 | rep_cnt = (param >> 32) & 0xfff; | |
4791 | rep_idx = (param >> 48) & 0xfff; | |
4792 | ||
4793 | trace_kvm_hv_hypercall(code, fast, rep_cnt, rep_idx, ingpa, outgpa); | |
4794 | ||
c25bc163 GN |
4795 | switch (code) { |
4796 | case HV_X64_HV_NOTIFY_LONG_SPIN_WAIT: | |
4797 | kvm_vcpu_on_spin(vcpu); | |
4798 | break; | |
4799 | default: | |
4800 | res = HV_STATUS_INVALID_HYPERCALL_CODE; | |
4801 | break; | |
4802 | } | |
55cd8e5a GN |
4803 | |
4804 | ret = res | (((u64)rep_done & 0xfff) << 32); | |
4805 | if (longmode) { | |
4806 | kvm_register_write(vcpu, VCPU_REGS_RAX, ret); | |
4807 | } else { | |
4808 | kvm_register_write(vcpu, VCPU_REGS_RDX, ret >> 32); | |
4809 | kvm_register_write(vcpu, VCPU_REGS_RAX, ret & 0xffffffff); | |
4810 | } | |
4811 | ||
4812 | return 1; | |
4813 | } | |
4814 | ||
8776e519 HB |
4815 | int kvm_emulate_hypercall(struct kvm_vcpu *vcpu) |
4816 | { | |
4817 | unsigned long nr, a0, a1, a2, a3, ret; | |
2f333bcb | 4818 | int r = 1; |
8776e519 | 4819 | |
55cd8e5a GN |
4820 | if (kvm_hv_hypercall_enabled(vcpu->kvm)) |
4821 | return kvm_hv_hypercall(vcpu); | |
4822 | ||
5fdbf976 MT |
4823 | nr = kvm_register_read(vcpu, VCPU_REGS_RAX); |
4824 | a0 = kvm_register_read(vcpu, VCPU_REGS_RBX); | |
4825 | a1 = kvm_register_read(vcpu, VCPU_REGS_RCX); | |
4826 | a2 = kvm_register_read(vcpu, VCPU_REGS_RDX); | |
4827 | a3 = kvm_register_read(vcpu, VCPU_REGS_RSI); | |
8776e519 | 4828 | |
229456fc | 4829 | trace_kvm_hypercall(nr, a0, a1, a2, a3); |
2714d1d3 | 4830 | |
8776e519 HB |
4831 | if (!is_long_mode(vcpu)) { |
4832 | nr &= 0xFFFFFFFF; | |
4833 | a0 &= 0xFFFFFFFF; | |
4834 | a1 &= 0xFFFFFFFF; | |
4835 | a2 &= 0xFFFFFFFF; | |
4836 | a3 &= 0xFFFFFFFF; | |
4837 | } | |
4838 | ||
07708c4a JK |
4839 | if (kvm_x86_ops->get_cpl(vcpu) != 0) { |
4840 | ret = -KVM_EPERM; | |
4841 | goto out; | |
4842 | } | |
4843 | ||
8776e519 | 4844 | switch (nr) { |
b93463aa AK |
4845 | case KVM_HC_VAPIC_POLL_IRQ: |
4846 | ret = 0; | |
4847 | break; | |
2f333bcb MT |
4848 | case KVM_HC_MMU_OP: |
4849 | r = kvm_pv_mmu_op(vcpu, a0, hc_gpa(vcpu, a1, a2), &ret); | |
4850 | break; | |
8776e519 HB |
4851 | default: |
4852 | ret = -KVM_ENOSYS; | |
4853 | break; | |
4854 | } | |
07708c4a | 4855 | out: |
5fdbf976 | 4856 | kvm_register_write(vcpu, VCPU_REGS_RAX, ret); |
f11c3a8d | 4857 | ++vcpu->stat.hypercalls; |
2f333bcb | 4858 | return r; |
8776e519 HB |
4859 | } |
4860 | EXPORT_SYMBOL_GPL(kvm_emulate_hypercall); | |
4861 | ||
4862 | int kvm_fix_hypercall(struct kvm_vcpu *vcpu) | |
4863 | { | |
4864 | char instruction[3]; | |
5fdbf976 | 4865 | unsigned long rip = kvm_rip_read(vcpu); |
8776e519 | 4866 | |
8776e519 HB |
4867 | /* |
4868 | * Blow out the MMU to ensure that no other VCPU has an active mapping | |
4869 | * to ensure that the updated hypercall appears atomically across all | |
4870 | * VCPUs. | |
4871 | */ | |
4872 | kvm_mmu_zap_all(vcpu->kvm); | |
4873 | ||
8776e519 | 4874 | kvm_x86_ops->patch_hypercall(vcpu, instruction); |
8776e519 | 4875 | |
8fe681e9 | 4876 | return emulator_write_emulated(rip, instruction, 3, NULL, vcpu); |
8776e519 HB |
4877 | } |
4878 | ||
8776e519 HB |
4879 | void realmode_lgdt(struct kvm_vcpu *vcpu, u16 limit, unsigned long base) |
4880 | { | |
89a27f4d | 4881 | struct desc_ptr dt = { limit, base }; |
8776e519 HB |
4882 | |
4883 | kvm_x86_ops->set_gdt(vcpu, &dt); | |
4884 | } | |
4885 | ||
4886 | void realmode_lidt(struct kvm_vcpu *vcpu, u16 limit, unsigned long base) | |
4887 | { | |
89a27f4d | 4888 | struct desc_ptr dt = { limit, base }; |
8776e519 HB |
4889 | |
4890 | kvm_x86_ops->set_idt(vcpu, &dt); | |
4891 | } | |
4892 | ||
07716717 DK |
4893 | static int move_to_next_stateful_cpuid_entry(struct kvm_vcpu *vcpu, int i) |
4894 | { | |
ad312c7c ZX |
4895 | struct kvm_cpuid_entry2 *e = &vcpu->arch.cpuid_entries[i]; |
4896 | int j, nent = vcpu->arch.cpuid_nent; | |
07716717 DK |
4897 | |
4898 | e->flags &= ~KVM_CPUID_FLAG_STATE_READ_NEXT; | |
4899 | /* when no next entry is found, the current entry[i] is reselected */ | |
0fdf8e59 | 4900 | for (j = i + 1; ; j = (j + 1) % nent) { |
ad312c7c | 4901 | struct kvm_cpuid_entry2 *ej = &vcpu->arch.cpuid_entries[j]; |
07716717 DK |
4902 | if (ej->function == e->function) { |
4903 | ej->flags |= KVM_CPUID_FLAG_STATE_READ_NEXT; | |
4904 | return j; | |
4905 | } | |
4906 | } | |
4907 | return 0; /* silence gcc, even though control never reaches here */ | |
4908 | } | |
4909 | ||
4910 | /* find an entry with matching function, matching index (if needed), and that | |
4911 | * should be read next (if it's stateful) */ | |
4912 | static int is_matching_cpuid_entry(struct kvm_cpuid_entry2 *e, | |
4913 | u32 function, u32 index) | |
4914 | { | |
4915 | if (e->function != function) | |
4916 | return 0; | |
4917 | if ((e->flags & KVM_CPUID_FLAG_SIGNIFCANT_INDEX) && e->index != index) | |
4918 | return 0; | |
4919 | if ((e->flags & KVM_CPUID_FLAG_STATEFUL_FUNC) && | |
19355475 | 4920 | !(e->flags & KVM_CPUID_FLAG_STATE_READ_NEXT)) |
07716717 DK |
4921 | return 0; |
4922 | return 1; | |
4923 | } | |
4924 | ||
d8017474 AG |
4925 | struct kvm_cpuid_entry2 *kvm_find_cpuid_entry(struct kvm_vcpu *vcpu, |
4926 | u32 function, u32 index) | |
8776e519 HB |
4927 | { |
4928 | int i; | |
d8017474 | 4929 | struct kvm_cpuid_entry2 *best = NULL; |
8776e519 | 4930 | |
ad312c7c | 4931 | for (i = 0; i < vcpu->arch.cpuid_nent; ++i) { |
d8017474 AG |
4932 | struct kvm_cpuid_entry2 *e; |
4933 | ||
ad312c7c | 4934 | e = &vcpu->arch.cpuid_entries[i]; |
07716717 DK |
4935 | if (is_matching_cpuid_entry(e, function, index)) { |
4936 | if (e->flags & KVM_CPUID_FLAG_STATEFUL_FUNC) | |
4937 | move_to_next_stateful_cpuid_entry(vcpu, i); | |
8776e519 HB |
4938 | best = e; |
4939 | break; | |
4940 | } | |
4941 | /* | |
4942 | * Both basic or both extended? | |
4943 | */ | |
4944 | if (((e->function ^ function) & 0x80000000) == 0) | |
4945 | if (!best || e->function > best->function) | |
4946 | best = e; | |
4947 | } | |
d8017474 AG |
4948 | return best; |
4949 | } | |
0e851880 | 4950 | EXPORT_SYMBOL_GPL(kvm_find_cpuid_entry); |
d8017474 | 4951 | |
82725b20 DE |
4952 | int cpuid_maxphyaddr(struct kvm_vcpu *vcpu) |
4953 | { | |
4954 | struct kvm_cpuid_entry2 *best; | |
4955 | ||
f7a71197 AK |
4956 | best = kvm_find_cpuid_entry(vcpu, 0x80000000, 0); |
4957 | if (!best || best->eax < 0x80000008) | |
4958 | goto not_found; | |
82725b20 DE |
4959 | best = kvm_find_cpuid_entry(vcpu, 0x80000008, 0); |
4960 | if (best) | |
4961 | return best->eax & 0xff; | |
f7a71197 | 4962 | not_found: |
82725b20 DE |
4963 | return 36; |
4964 | } | |
4965 | ||
d8017474 AG |
4966 | void kvm_emulate_cpuid(struct kvm_vcpu *vcpu) |
4967 | { | |
4968 | u32 function, index; | |
4969 | struct kvm_cpuid_entry2 *best; | |
4970 | ||
4971 | function = kvm_register_read(vcpu, VCPU_REGS_RAX); | |
4972 | index = kvm_register_read(vcpu, VCPU_REGS_RCX); | |
4973 | kvm_register_write(vcpu, VCPU_REGS_RAX, 0); | |
4974 | kvm_register_write(vcpu, VCPU_REGS_RBX, 0); | |
4975 | kvm_register_write(vcpu, VCPU_REGS_RCX, 0); | |
4976 | kvm_register_write(vcpu, VCPU_REGS_RDX, 0); | |
4977 | best = kvm_find_cpuid_entry(vcpu, function, index); | |
8776e519 | 4978 | if (best) { |
5fdbf976 MT |
4979 | kvm_register_write(vcpu, VCPU_REGS_RAX, best->eax); |
4980 | kvm_register_write(vcpu, VCPU_REGS_RBX, best->ebx); | |
4981 | kvm_register_write(vcpu, VCPU_REGS_RCX, best->ecx); | |
4982 | kvm_register_write(vcpu, VCPU_REGS_RDX, best->edx); | |
8776e519 | 4983 | } |
8776e519 | 4984 | kvm_x86_ops->skip_emulated_instruction(vcpu); |
229456fc MT |
4985 | trace_kvm_cpuid(function, |
4986 | kvm_register_read(vcpu, VCPU_REGS_RAX), | |
4987 | kvm_register_read(vcpu, VCPU_REGS_RBX), | |
4988 | kvm_register_read(vcpu, VCPU_REGS_RCX), | |
4989 | kvm_register_read(vcpu, VCPU_REGS_RDX)); | |
8776e519 HB |
4990 | } |
4991 | EXPORT_SYMBOL_GPL(kvm_emulate_cpuid); | |
d0752060 | 4992 | |
b6c7a5dc HB |
4993 | /* |
4994 | * Check if userspace requested an interrupt window, and that the | |
4995 | * interrupt window is open. | |
4996 | * | |
4997 | * No need to exit to userspace if we already have an interrupt queued. | |
4998 | */ | |
851ba692 | 4999 | static int dm_request_for_irq_injection(struct kvm_vcpu *vcpu) |
b6c7a5dc | 5000 | { |
8061823a | 5001 | return (!irqchip_in_kernel(vcpu->kvm) && !kvm_cpu_has_interrupt(vcpu) && |
851ba692 | 5002 | vcpu->run->request_interrupt_window && |
5df56646 | 5003 | kvm_arch_interrupt_allowed(vcpu)); |
b6c7a5dc HB |
5004 | } |
5005 | ||
851ba692 | 5006 | static void post_kvm_run_save(struct kvm_vcpu *vcpu) |
b6c7a5dc | 5007 | { |
851ba692 AK |
5008 | struct kvm_run *kvm_run = vcpu->run; |
5009 | ||
91586a3b | 5010 | kvm_run->if_flag = (kvm_get_rflags(vcpu) & X86_EFLAGS_IF) != 0; |
2d3ad1f4 | 5011 | kvm_run->cr8 = kvm_get_cr8(vcpu); |
b6c7a5dc | 5012 | kvm_run->apic_base = kvm_get_apic_base(vcpu); |
4531220b | 5013 | if (irqchip_in_kernel(vcpu->kvm)) |
b6c7a5dc | 5014 | kvm_run->ready_for_interrupt_injection = 1; |
4531220b | 5015 | else |
b6c7a5dc | 5016 | kvm_run->ready_for_interrupt_injection = |
fa9726b0 GN |
5017 | kvm_arch_interrupt_allowed(vcpu) && |
5018 | !kvm_cpu_has_interrupt(vcpu) && | |
5019 | !kvm_event_needs_reinjection(vcpu); | |
b6c7a5dc HB |
5020 | } |
5021 | ||
b93463aa AK |
5022 | static void vapic_enter(struct kvm_vcpu *vcpu) |
5023 | { | |
5024 | struct kvm_lapic *apic = vcpu->arch.apic; | |
5025 | struct page *page; | |
5026 | ||
5027 | if (!apic || !apic->vapic_addr) | |
5028 | return; | |
5029 | ||
5030 | page = gfn_to_page(vcpu->kvm, apic->vapic_addr >> PAGE_SHIFT); | |
72dc67a6 IE |
5031 | |
5032 | vcpu->arch.apic->vapic_page = page; | |
b93463aa AK |
5033 | } |
5034 | ||
5035 | static void vapic_exit(struct kvm_vcpu *vcpu) | |
5036 | { | |
5037 | struct kvm_lapic *apic = vcpu->arch.apic; | |
f656ce01 | 5038 | int idx; |
b93463aa AK |
5039 | |
5040 | if (!apic || !apic->vapic_addr) | |
5041 | return; | |
5042 | ||
f656ce01 | 5043 | idx = srcu_read_lock(&vcpu->kvm->srcu); |
b93463aa AK |
5044 | kvm_release_page_dirty(apic->vapic_page); |
5045 | mark_page_dirty(vcpu->kvm, apic->vapic_addr >> PAGE_SHIFT); | |
f656ce01 | 5046 | srcu_read_unlock(&vcpu->kvm->srcu, idx); |
b93463aa AK |
5047 | } |
5048 | ||
95ba8273 GN |
5049 | static void update_cr8_intercept(struct kvm_vcpu *vcpu) |
5050 | { | |
5051 | int max_irr, tpr; | |
5052 | ||
5053 | if (!kvm_x86_ops->update_cr8_intercept) | |
5054 | return; | |
5055 | ||
88c808fd AK |
5056 | if (!vcpu->arch.apic) |
5057 | return; | |
5058 | ||
8db3baa2 GN |
5059 | if (!vcpu->arch.apic->vapic_addr) |
5060 | max_irr = kvm_lapic_find_highest_irr(vcpu); | |
5061 | else | |
5062 | max_irr = -1; | |
95ba8273 GN |
5063 | |
5064 | if (max_irr != -1) | |
5065 | max_irr >>= 4; | |
5066 | ||
5067 | tpr = kvm_lapic_get_cr8(vcpu); | |
5068 | ||
5069 | kvm_x86_ops->update_cr8_intercept(vcpu, tpr, max_irr); | |
5070 | } | |
5071 | ||
851ba692 | 5072 | static void inject_pending_event(struct kvm_vcpu *vcpu) |
95ba8273 GN |
5073 | { |
5074 | /* try to reinject previous events if any */ | |
b59bb7bd | 5075 | if (vcpu->arch.exception.pending) { |
5c1c85d0 AK |
5076 | trace_kvm_inj_exception(vcpu->arch.exception.nr, |
5077 | vcpu->arch.exception.has_error_code, | |
5078 | vcpu->arch.exception.error_code); | |
b59bb7bd GN |
5079 | kvm_x86_ops->queue_exception(vcpu, vcpu->arch.exception.nr, |
5080 | vcpu->arch.exception.has_error_code, | |
ce7ddec4 JR |
5081 | vcpu->arch.exception.error_code, |
5082 | vcpu->arch.exception.reinject); | |
b59bb7bd GN |
5083 | return; |
5084 | } | |
5085 | ||
95ba8273 GN |
5086 | if (vcpu->arch.nmi_injected) { |
5087 | kvm_x86_ops->set_nmi(vcpu); | |
5088 | return; | |
5089 | } | |
5090 | ||
5091 | if (vcpu->arch.interrupt.pending) { | |
66fd3f7f | 5092 | kvm_x86_ops->set_irq(vcpu); |
95ba8273 GN |
5093 | return; |
5094 | } | |
5095 | ||
5096 | /* try to inject new event if pending */ | |
5097 | if (vcpu->arch.nmi_pending) { | |
5098 | if (kvm_x86_ops->nmi_allowed(vcpu)) { | |
5099 | vcpu->arch.nmi_pending = false; | |
5100 | vcpu->arch.nmi_injected = true; | |
5101 | kvm_x86_ops->set_nmi(vcpu); | |
5102 | } | |
5103 | } else if (kvm_cpu_has_interrupt(vcpu)) { | |
5104 | if (kvm_x86_ops->interrupt_allowed(vcpu)) { | |
66fd3f7f GN |
5105 | kvm_queue_interrupt(vcpu, kvm_cpu_get_interrupt(vcpu), |
5106 | false); | |
5107 | kvm_x86_ops->set_irq(vcpu); | |
95ba8273 GN |
5108 | } |
5109 | } | |
5110 | } | |
5111 | ||
2acf923e DC |
5112 | static void kvm_load_guest_xcr0(struct kvm_vcpu *vcpu) |
5113 | { | |
5114 | if (kvm_read_cr4_bits(vcpu, X86_CR4_OSXSAVE) && | |
5115 | !vcpu->guest_xcr0_loaded) { | |
5116 | /* kvm_set_xcr() also depends on this */ | |
5117 | xsetbv(XCR_XFEATURE_ENABLED_MASK, vcpu->arch.xcr0); | |
5118 | vcpu->guest_xcr0_loaded = 1; | |
5119 | } | |
5120 | } | |
5121 | ||
5122 | static void kvm_put_guest_xcr0(struct kvm_vcpu *vcpu) | |
5123 | { | |
5124 | if (vcpu->guest_xcr0_loaded) { | |
5125 | if (vcpu->arch.xcr0 != host_xcr0) | |
5126 | xsetbv(XCR_XFEATURE_ENABLED_MASK, host_xcr0); | |
5127 | vcpu->guest_xcr0_loaded = 0; | |
5128 | } | |
5129 | } | |
5130 | ||
851ba692 | 5131 | static int vcpu_enter_guest(struct kvm_vcpu *vcpu) |
b6c7a5dc HB |
5132 | { |
5133 | int r; | |
6a8b1d13 | 5134 | bool req_int_win = !irqchip_in_kernel(vcpu->kvm) && |
851ba692 | 5135 | vcpu->run->request_interrupt_window; |
b6c7a5dc | 5136 | |
3e007509 | 5137 | if (vcpu->requests) { |
a8eeb04a | 5138 | if (kvm_check_request(KVM_REQ_MMU_RELOAD, vcpu)) |
2e53d63a | 5139 | kvm_mmu_unload(vcpu); |
a8eeb04a | 5140 | if (kvm_check_request(KVM_REQ_MIGRATE_TIMER, vcpu)) |
2f599714 | 5141 | __kvm_migrate_timers(vcpu); |
34c238a1 ZA |
5142 | if (kvm_check_request(KVM_REQ_CLOCK_UPDATE, vcpu)) { |
5143 | r = kvm_guest_time_update(vcpu); | |
8cfdc000 ZA |
5144 | if (unlikely(r)) |
5145 | goto out; | |
5146 | } | |
a8eeb04a | 5147 | if (kvm_check_request(KVM_REQ_MMU_SYNC, vcpu)) |
4731d4c7 | 5148 | kvm_mmu_sync_roots(vcpu); |
a8eeb04a | 5149 | if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu)) |
d4acf7e7 | 5150 | kvm_x86_ops->tlb_flush(vcpu); |
a8eeb04a | 5151 | if (kvm_check_request(KVM_REQ_REPORT_TPR_ACCESS, vcpu)) { |
851ba692 | 5152 | vcpu->run->exit_reason = KVM_EXIT_TPR_ACCESS; |
b93463aa AK |
5153 | r = 0; |
5154 | goto out; | |
5155 | } | |
a8eeb04a | 5156 | if (kvm_check_request(KVM_REQ_TRIPLE_FAULT, vcpu)) { |
851ba692 | 5157 | vcpu->run->exit_reason = KVM_EXIT_SHUTDOWN; |
71c4dfaf JR |
5158 | r = 0; |
5159 | goto out; | |
5160 | } | |
a8eeb04a | 5161 | if (kvm_check_request(KVM_REQ_DEACTIVATE_FPU, vcpu)) { |
02daab21 AK |
5162 | vcpu->fpu_active = 0; |
5163 | kvm_x86_ops->fpu_deactivate(vcpu); | |
5164 | } | |
af585b92 GN |
5165 | if (kvm_check_request(KVM_REQ_APF_HALT, vcpu)) { |
5166 | /* Page is swapped out. Do synthetic halt */ | |
5167 | vcpu->arch.apf.halted = true; | |
5168 | r = 1; | |
5169 | goto out; | |
5170 | } | |
2f52d58c | 5171 | } |
b93463aa | 5172 | |
3e007509 AK |
5173 | r = kvm_mmu_reload(vcpu); |
5174 | if (unlikely(r)) | |
5175 | goto out; | |
5176 | ||
b463a6f7 AK |
5177 | if (kvm_check_request(KVM_REQ_EVENT, vcpu) || req_int_win) { |
5178 | inject_pending_event(vcpu); | |
5179 | ||
5180 | /* enable NMI/IRQ window open exits if needed */ | |
5181 | if (vcpu->arch.nmi_pending) | |
5182 | kvm_x86_ops->enable_nmi_window(vcpu); | |
5183 | else if (kvm_cpu_has_interrupt(vcpu) || req_int_win) | |
5184 | kvm_x86_ops->enable_irq_window(vcpu); | |
5185 | ||
5186 | if (kvm_lapic_enabled(vcpu)) { | |
5187 | update_cr8_intercept(vcpu); | |
5188 | kvm_lapic_sync_to_vapic(vcpu); | |
5189 | } | |
5190 | } | |
5191 | ||
b6c7a5dc HB |
5192 | preempt_disable(); |
5193 | ||
5194 | kvm_x86_ops->prepare_guest_switch(vcpu); | |
2608d7a1 AK |
5195 | if (vcpu->fpu_active) |
5196 | kvm_load_guest_fpu(vcpu); | |
2acf923e | 5197 | kvm_load_guest_xcr0(vcpu); |
b6c7a5dc | 5198 | |
d94e1dc9 AK |
5199 | atomic_set(&vcpu->guest_mode, 1); |
5200 | smp_wmb(); | |
b6c7a5dc | 5201 | |
d94e1dc9 | 5202 | local_irq_disable(); |
32f88400 | 5203 | |
d94e1dc9 AK |
5204 | if (!atomic_read(&vcpu->guest_mode) || vcpu->requests |
5205 | || need_resched() || signal_pending(current)) { | |
5206 | atomic_set(&vcpu->guest_mode, 0); | |
5207 | smp_wmb(); | |
6c142801 AK |
5208 | local_irq_enable(); |
5209 | preempt_enable(); | |
b463a6f7 | 5210 | kvm_x86_ops->cancel_injection(vcpu); |
6c142801 AK |
5211 | r = 1; |
5212 | goto out; | |
5213 | } | |
5214 | ||
f656ce01 | 5215 | srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx); |
3200f405 | 5216 | |
b6c7a5dc HB |
5217 | kvm_guest_enter(); |
5218 | ||
42dbaa5a | 5219 | if (unlikely(vcpu->arch.switch_db_regs)) { |
42dbaa5a JK |
5220 | set_debugreg(0, 7); |
5221 | set_debugreg(vcpu->arch.eff_db[0], 0); | |
5222 | set_debugreg(vcpu->arch.eff_db[1], 1); | |
5223 | set_debugreg(vcpu->arch.eff_db[2], 2); | |
5224 | set_debugreg(vcpu->arch.eff_db[3], 3); | |
5225 | } | |
b6c7a5dc | 5226 | |
229456fc | 5227 | trace_kvm_entry(vcpu->vcpu_id); |
851ba692 | 5228 | kvm_x86_ops->run(vcpu); |
b6c7a5dc | 5229 | |
24f1e32c FW |
5230 | /* |
5231 | * If the guest has used debug registers, at least dr7 | |
5232 | * will be disabled while returning to the host. | |
5233 | * If we don't have active breakpoints in the host, we don't | |
5234 | * care about the messed up debug address registers. But if | |
5235 | * we have some of them active, restore the old state. | |
5236 | */ | |
59d8eb53 | 5237 | if (hw_breakpoint_active()) |
24f1e32c | 5238 | hw_breakpoint_restore(); |
42dbaa5a | 5239 | |
1d5f066e ZA |
5240 | kvm_get_msr(vcpu, MSR_IA32_TSC, &vcpu->arch.last_guest_tsc); |
5241 | ||
d94e1dc9 AK |
5242 | atomic_set(&vcpu->guest_mode, 0); |
5243 | smp_wmb(); | |
b6c7a5dc HB |
5244 | local_irq_enable(); |
5245 | ||
5246 | ++vcpu->stat.exits; | |
5247 | ||
5248 | /* | |
5249 | * We must have an instruction between local_irq_enable() and | |
5250 | * kvm_guest_exit(), so the timer interrupt isn't delayed by | |
5251 | * the interrupt shadow. The stat.exits increment will do nicely. | |
5252 | * But we need to prevent reordering, hence this barrier(): | |
5253 | */ | |
5254 | barrier(); | |
5255 | ||
5256 | kvm_guest_exit(); | |
5257 | ||
5258 | preempt_enable(); | |
5259 | ||
f656ce01 | 5260 | vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu); |
3200f405 | 5261 | |
b6c7a5dc HB |
5262 | /* |
5263 | * Profile KVM exit RIPs: | |
5264 | */ | |
5265 | if (unlikely(prof_on == KVM_PROFILING)) { | |
5fdbf976 MT |
5266 | unsigned long rip = kvm_rip_read(vcpu); |
5267 | profile_hit(KVM_PROFILING, (void *)rip); | |
b6c7a5dc HB |
5268 | } |
5269 | ||
298101da | 5270 | |
b93463aa AK |
5271 | kvm_lapic_sync_from_vapic(vcpu); |
5272 | ||
851ba692 | 5273 | r = kvm_x86_ops->handle_exit(vcpu); |
d7690175 MT |
5274 | out: |
5275 | return r; | |
5276 | } | |
b6c7a5dc | 5277 | |
09cec754 | 5278 | |
851ba692 | 5279 | static int __vcpu_run(struct kvm_vcpu *vcpu) |
d7690175 MT |
5280 | { |
5281 | int r; | |
f656ce01 | 5282 | struct kvm *kvm = vcpu->kvm; |
d7690175 MT |
5283 | |
5284 | if (unlikely(vcpu->arch.mp_state == KVM_MP_STATE_SIPI_RECEIVED)) { | |
1b10bf31 JK |
5285 | pr_debug("vcpu %d received sipi with vector # %x\n", |
5286 | vcpu->vcpu_id, vcpu->arch.sipi_vector); | |
d7690175 | 5287 | kvm_lapic_reset(vcpu); |
5f179287 | 5288 | r = kvm_arch_vcpu_reset(vcpu); |
d7690175 MT |
5289 | if (r) |
5290 | return r; | |
5291 | vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE; | |
b6c7a5dc HB |
5292 | } |
5293 | ||
f656ce01 | 5294 | vcpu->srcu_idx = srcu_read_lock(&kvm->srcu); |
d7690175 MT |
5295 | vapic_enter(vcpu); |
5296 | ||
5297 | r = 1; | |
5298 | while (r > 0) { | |
af585b92 GN |
5299 | if (vcpu->arch.mp_state == KVM_MP_STATE_RUNNABLE && |
5300 | !vcpu->arch.apf.halted) | |
851ba692 | 5301 | r = vcpu_enter_guest(vcpu); |
d7690175 | 5302 | else { |
f656ce01 | 5303 | srcu_read_unlock(&kvm->srcu, vcpu->srcu_idx); |
d7690175 | 5304 | kvm_vcpu_block(vcpu); |
f656ce01 | 5305 | vcpu->srcu_idx = srcu_read_lock(&kvm->srcu); |
a8eeb04a | 5306 | if (kvm_check_request(KVM_REQ_UNHALT, vcpu)) |
09cec754 GN |
5307 | { |
5308 | switch(vcpu->arch.mp_state) { | |
5309 | case KVM_MP_STATE_HALTED: | |
d7690175 | 5310 | vcpu->arch.mp_state = |
09cec754 GN |
5311 | KVM_MP_STATE_RUNNABLE; |
5312 | case KVM_MP_STATE_RUNNABLE: | |
af585b92 | 5313 | vcpu->arch.apf.halted = false; |
09cec754 GN |
5314 | break; |
5315 | case KVM_MP_STATE_SIPI_RECEIVED: | |
5316 | default: | |
5317 | r = -EINTR; | |
5318 | break; | |
5319 | } | |
5320 | } | |
d7690175 MT |
5321 | } |
5322 | ||
09cec754 GN |
5323 | if (r <= 0) |
5324 | break; | |
5325 | ||
5326 | clear_bit(KVM_REQ_PENDING_TIMER, &vcpu->requests); | |
5327 | if (kvm_cpu_has_pending_timer(vcpu)) | |
5328 | kvm_inject_pending_timer_irqs(vcpu); | |
5329 | ||
851ba692 | 5330 | if (dm_request_for_irq_injection(vcpu)) { |
09cec754 | 5331 | r = -EINTR; |
851ba692 | 5332 | vcpu->run->exit_reason = KVM_EXIT_INTR; |
09cec754 GN |
5333 | ++vcpu->stat.request_irq_exits; |
5334 | } | |
af585b92 GN |
5335 | |
5336 | kvm_check_async_pf_completion(vcpu); | |
5337 | ||
09cec754 GN |
5338 | if (signal_pending(current)) { |
5339 | r = -EINTR; | |
851ba692 | 5340 | vcpu->run->exit_reason = KVM_EXIT_INTR; |
09cec754 GN |
5341 | ++vcpu->stat.signal_exits; |
5342 | } | |
5343 | if (need_resched()) { | |
f656ce01 | 5344 | srcu_read_unlock(&kvm->srcu, vcpu->srcu_idx); |
09cec754 | 5345 | kvm_resched(vcpu); |
f656ce01 | 5346 | vcpu->srcu_idx = srcu_read_lock(&kvm->srcu); |
d7690175 | 5347 | } |
b6c7a5dc HB |
5348 | } |
5349 | ||
f656ce01 | 5350 | srcu_read_unlock(&kvm->srcu, vcpu->srcu_idx); |
b6c7a5dc | 5351 | |
b93463aa AK |
5352 | vapic_exit(vcpu); |
5353 | ||
b6c7a5dc HB |
5354 | return r; |
5355 | } | |
5356 | ||
5357 | int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) | |
5358 | { | |
5359 | int r; | |
5360 | sigset_t sigsaved; | |
5361 | ||
ac9f6dc0 AK |
5362 | if (vcpu->sigset_active) |
5363 | sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved); | |
5364 | ||
a4535290 | 5365 | if (unlikely(vcpu->arch.mp_state == KVM_MP_STATE_UNINITIALIZED)) { |
b6c7a5dc | 5366 | kvm_vcpu_block(vcpu); |
d7690175 | 5367 | clear_bit(KVM_REQ_UNHALT, &vcpu->requests); |
ac9f6dc0 AK |
5368 | r = -EAGAIN; |
5369 | goto out; | |
b6c7a5dc HB |
5370 | } |
5371 | ||
b6c7a5dc HB |
5372 | /* re-sync apic's tpr */ |
5373 | if (!irqchip_in_kernel(vcpu->kvm)) | |
2d3ad1f4 | 5374 | kvm_set_cr8(vcpu, kvm_run->cr8); |
b6c7a5dc | 5375 | |
d2ddd1c4 | 5376 | if (vcpu->arch.pio.count || vcpu->mmio_needed) { |
92bf9748 GN |
5377 | if (vcpu->mmio_needed) { |
5378 | memcpy(vcpu->mmio_data, kvm_run->mmio.data, 8); | |
5379 | vcpu->mmio_read_completed = 1; | |
5380 | vcpu->mmio_needed = 0; | |
b6c7a5dc | 5381 | } |
f656ce01 | 5382 | vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu); |
5cd21917 | 5383 | r = emulate_instruction(vcpu, 0, 0, EMULTYPE_NO_DECODE); |
f656ce01 | 5384 | srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx); |
6d77dbfc | 5385 | if (r != EMULATE_DONE) { |
b6c7a5dc HB |
5386 | r = 0; |
5387 | goto out; | |
5388 | } | |
5389 | } | |
5fdbf976 MT |
5390 | if (kvm_run->exit_reason == KVM_EXIT_HYPERCALL) |
5391 | kvm_register_write(vcpu, VCPU_REGS_RAX, | |
5392 | kvm_run->hypercall.ret); | |
b6c7a5dc | 5393 | |
851ba692 | 5394 | r = __vcpu_run(vcpu); |
b6c7a5dc HB |
5395 | |
5396 | out: | |
f1d86e46 | 5397 | post_kvm_run_save(vcpu); |
b6c7a5dc HB |
5398 | if (vcpu->sigset_active) |
5399 | sigprocmask(SIG_SETMASK, &sigsaved, NULL); | |
5400 | ||
b6c7a5dc HB |
5401 | return r; |
5402 | } | |
5403 | ||
5404 | int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs) | |
5405 | { | |
5fdbf976 MT |
5406 | regs->rax = kvm_register_read(vcpu, VCPU_REGS_RAX); |
5407 | regs->rbx = kvm_register_read(vcpu, VCPU_REGS_RBX); | |
5408 | regs->rcx = kvm_register_read(vcpu, VCPU_REGS_RCX); | |
5409 | regs->rdx = kvm_register_read(vcpu, VCPU_REGS_RDX); | |
5410 | regs->rsi = kvm_register_read(vcpu, VCPU_REGS_RSI); | |
5411 | regs->rdi = kvm_register_read(vcpu, VCPU_REGS_RDI); | |
5412 | regs->rsp = kvm_register_read(vcpu, VCPU_REGS_RSP); | |
5413 | regs->rbp = kvm_register_read(vcpu, VCPU_REGS_RBP); | |
b6c7a5dc | 5414 | #ifdef CONFIG_X86_64 |
5fdbf976 MT |
5415 | regs->r8 = kvm_register_read(vcpu, VCPU_REGS_R8); |
5416 | regs->r9 = kvm_register_read(vcpu, VCPU_REGS_R9); | |
5417 | regs->r10 = kvm_register_read(vcpu, VCPU_REGS_R10); | |
5418 | regs->r11 = kvm_register_read(vcpu, VCPU_REGS_R11); | |
5419 | regs->r12 = kvm_register_read(vcpu, VCPU_REGS_R12); | |
5420 | regs->r13 = kvm_register_read(vcpu, VCPU_REGS_R13); | |
5421 | regs->r14 = kvm_register_read(vcpu, VCPU_REGS_R14); | |
5422 | regs->r15 = kvm_register_read(vcpu, VCPU_REGS_R15); | |
b6c7a5dc HB |
5423 | #endif |
5424 | ||
5fdbf976 | 5425 | regs->rip = kvm_rip_read(vcpu); |
91586a3b | 5426 | regs->rflags = kvm_get_rflags(vcpu); |
b6c7a5dc | 5427 | |
b6c7a5dc HB |
5428 | return 0; |
5429 | } | |
5430 | ||
5431 | int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs) | |
5432 | { | |
5fdbf976 MT |
5433 | kvm_register_write(vcpu, VCPU_REGS_RAX, regs->rax); |
5434 | kvm_register_write(vcpu, VCPU_REGS_RBX, regs->rbx); | |
5435 | kvm_register_write(vcpu, VCPU_REGS_RCX, regs->rcx); | |
5436 | kvm_register_write(vcpu, VCPU_REGS_RDX, regs->rdx); | |
5437 | kvm_register_write(vcpu, VCPU_REGS_RSI, regs->rsi); | |
5438 | kvm_register_write(vcpu, VCPU_REGS_RDI, regs->rdi); | |
5439 | kvm_register_write(vcpu, VCPU_REGS_RSP, regs->rsp); | |
5440 | kvm_register_write(vcpu, VCPU_REGS_RBP, regs->rbp); | |
b6c7a5dc | 5441 | #ifdef CONFIG_X86_64 |
5fdbf976 MT |
5442 | kvm_register_write(vcpu, VCPU_REGS_R8, regs->r8); |
5443 | kvm_register_write(vcpu, VCPU_REGS_R9, regs->r9); | |
5444 | kvm_register_write(vcpu, VCPU_REGS_R10, regs->r10); | |
5445 | kvm_register_write(vcpu, VCPU_REGS_R11, regs->r11); | |
5446 | kvm_register_write(vcpu, VCPU_REGS_R12, regs->r12); | |
5447 | kvm_register_write(vcpu, VCPU_REGS_R13, regs->r13); | |
5448 | kvm_register_write(vcpu, VCPU_REGS_R14, regs->r14); | |
5449 | kvm_register_write(vcpu, VCPU_REGS_R15, regs->r15); | |
b6c7a5dc HB |
5450 | #endif |
5451 | ||
5fdbf976 | 5452 | kvm_rip_write(vcpu, regs->rip); |
91586a3b | 5453 | kvm_set_rflags(vcpu, regs->rflags); |
b6c7a5dc | 5454 | |
b4f14abd JK |
5455 | vcpu->arch.exception.pending = false; |
5456 | ||
3842d135 AK |
5457 | kvm_make_request(KVM_REQ_EVENT, vcpu); |
5458 | ||
b6c7a5dc HB |
5459 | return 0; |
5460 | } | |
5461 | ||
b6c7a5dc HB |
5462 | void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l) |
5463 | { | |
5464 | struct kvm_segment cs; | |
5465 | ||
3e6e0aab | 5466 | kvm_get_segment(vcpu, &cs, VCPU_SREG_CS); |
b6c7a5dc HB |
5467 | *db = cs.db; |
5468 | *l = cs.l; | |
5469 | } | |
5470 | EXPORT_SYMBOL_GPL(kvm_get_cs_db_l_bits); | |
5471 | ||
5472 | int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu, | |
5473 | struct kvm_sregs *sregs) | |
5474 | { | |
89a27f4d | 5475 | struct desc_ptr dt; |
b6c7a5dc | 5476 | |
3e6e0aab GT |
5477 | kvm_get_segment(vcpu, &sregs->cs, VCPU_SREG_CS); |
5478 | kvm_get_segment(vcpu, &sregs->ds, VCPU_SREG_DS); | |
5479 | kvm_get_segment(vcpu, &sregs->es, VCPU_SREG_ES); | |
5480 | kvm_get_segment(vcpu, &sregs->fs, VCPU_SREG_FS); | |
5481 | kvm_get_segment(vcpu, &sregs->gs, VCPU_SREG_GS); | |
5482 | kvm_get_segment(vcpu, &sregs->ss, VCPU_SREG_SS); | |
b6c7a5dc | 5483 | |
3e6e0aab GT |
5484 | kvm_get_segment(vcpu, &sregs->tr, VCPU_SREG_TR); |
5485 | kvm_get_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR); | |
b6c7a5dc HB |
5486 | |
5487 | kvm_x86_ops->get_idt(vcpu, &dt); | |
89a27f4d GN |
5488 | sregs->idt.limit = dt.size; |
5489 | sregs->idt.base = dt.address; | |
b6c7a5dc | 5490 | kvm_x86_ops->get_gdt(vcpu, &dt); |
89a27f4d GN |
5491 | sregs->gdt.limit = dt.size; |
5492 | sregs->gdt.base = dt.address; | |
b6c7a5dc | 5493 | |
4d4ec087 | 5494 | sregs->cr0 = kvm_read_cr0(vcpu); |
ad312c7c ZX |
5495 | sregs->cr2 = vcpu->arch.cr2; |
5496 | sregs->cr3 = vcpu->arch.cr3; | |
fc78f519 | 5497 | sregs->cr4 = kvm_read_cr4(vcpu); |
2d3ad1f4 | 5498 | sregs->cr8 = kvm_get_cr8(vcpu); |
f6801dff | 5499 | sregs->efer = vcpu->arch.efer; |
b6c7a5dc HB |
5500 | sregs->apic_base = kvm_get_apic_base(vcpu); |
5501 | ||
923c61bb | 5502 | memset(sregs->interrupt_bitmap, 0, sizeof sregs->interrupt_bitmap); |
b6c7a5dc | 5503 | |
36752c9b | 5504 | if (vcpu->arch.interrupt.pending && !vcpu->arch.interrupt.soft) |
14d0bc1f GN |
5505 | set_bit(vcpu->arch.interrupt.nr, |
5506 | (unsigned long *)sregs->interrupt_bitmap); | |
16d7a191 | 5507 | |
b6c7a5dc HB |
5508 | return 0; |
5509 | } | |
5510 | ||
62d9f0db MT |
5511 | int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu, |
5512 | struct kvm_mp_state *mp_state) | |
5513 | { | |
62d9f0db | 5514 | mp_state->mp_state = vcpu->arch.mp_state; |
62d9f0db MT |
5515 | return 0; |
5516 | } | |
5517 | ||
5518 | int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu, | |
5519 | struct kvm_mp_state *mp_state) | |
5520 | { | |
62d9f0db | 5521 | vcpu->arch.mp_state = mp_state->mp_state; |
3842d135 | 5522 | kvm_make_request(KVM_REQ_EVENT, vcpu); |
62d9f0db MT |
5523 | return 0; |
5524 | } | |
5525 | ||
e269fb21 JK |
5526 | int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int reason, |
5527 | bool has_error_code, u32 error_code) | |
b6c7a5dc | 5528 | { |
4d2179e1 | 5529 | struct decode_cache *c = &vcpu->arch.emulate_ctxt.decode; |
8ec4722d | 5530 | int ret; |
e01c2426 | 5531 | |
8ec4722d | 5532 | init_emulate_ctxt(vcpu); |
c697518a | 5533 | |
9aabc88f | 5534 | ret = emulator_task_switch(&vcpu->arch.emulate_ctxt, |
e269fb21 JK |
5535 | tss_selector, reason, has_error_code, |
5536 | error_code); | |
c697518a | 5537 | |
c697518a | 5538 | if (ret) |
19d04437 | 5539 | return EMULATE_FAIL; |
37817f29 | 5540 | |
4d2179e1 | 5541 | memcpy(vcpu->arch.regs, c->regs, sizeof c->regs); |
95c55886 | 5542 | kvm_rip_write(vcpu, vcpu->arch.emulate_ctxt.eip); |
19d04437 | 5543 | kvm_x86_ops->set_rflags(vcpu, vcpu->arch.emulate_ctxt.eflags); |
3842d135 | 5544 | kvm_make_request(KVM_REQ_EVENT, vcpu); |
19d04437 | 5545 | return EMULATE_DONE; |
37817f29 IE |
5546 | } |
5547 | EXPORT_SYMBOL_GPL(kvm_task_switch); | |
5548 | ||
b6c7a5dc HB |
5549 | int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu, |
5550 | struct kvm_sregs *sregs) | |
5551 | { | |
5552 | int mmu_reset_needed = 0; | |
923c61bb | 5553 | int pending_vec, max_bits; |
89a27f4d | 5554 | struct desc_ptr dt; |
b6c7a5dc | 5555 | |
89a27f4d GN |
5556 | dt.size = sregs->idt.limit; |
5557 | dt.address = sregs->idt.base; | |
b6c7a5dc | 5558 | kvm_x86_ops->set_idt(vcpu, &dt); |
89a27f4d GN |
5559 | dt.size = sregs->gdt.limit; |
5560 | dt.address = sregs->gdt.base; | |
b6c7a5dc HB |
5561 | kvm_x86_ops->set_gdt(vcpu, &dt); |
5562 | ||
ad312c7c ZX |
5563 | vcpu->arch.cr2 = sregs->cr2; |
5564 | mmu_reset_needed |= vcpu->arch.cr3 != sregs->cr3; | |
dc7e795e | 5565 | vcpu->arch.cr3 = sregs->cr3; |
b6c7a5dc | 5566 | |
2d3ad1f4 | 5567 | kvm_set_cr8(vcpu, sregs->cr8); |
b6c7a5dc | 5568 | |
f6801dff | 5569 | mmu_reset_needed |= vcpu->arch.efer != sregs->efer; |
b6c7a5dc | 5570 | kvm_x86_ops->set_efer(vcpu, sregs->efer); |
b6c7a5dc HB |
5571 | kvm_set_apic_base(vcpu, sregs->apic_base); |
5572 | ||
4d4ec087 | 5573 | mmu_reset_needed |= kvm_read_cr0(vcpu) != sregs->cr0; |
b6c7a5dc | 5574 | kvm_x86_ops->set_cr0(vcpu, sregs->cr0); |
d7306163 | 5575 | vcpu->arch.cr0 = sregs->cr0; |
b6c7a5dc | 5576 | |
fc78f519 | 5577 | mmu_reset_needed |= kvm_read_cr4(vcpu) != sregs->cr4; |
b6c7a5dc | 5578 | kvm_x86_ops->set_cr4(vcpu, sregs->cr4); |
3ea3aa8c SY |
5579 | if (sregs->cr4 & X86_CR4_OSXSAVE) |
5580 | update_cpuid(vcpu); | |
7c93be44 | 5581 | if (!is_long_mode(vcpu) && is_pae(vcpu)) { |
ff03a073 | 5582 | load_pdptrs(vcpu, vcpu->arch.walk_mmu, vcpu->arch.cr3); |
7c93be44 MT |
5583 | mmu_reset_needed = 1; |
5584 | } | |
b6c7a5dc HB |
5585 | |
5586 | if (mmu_reset_needed) | |
5587 | kvm_mmu_reset_context(vcpu); | |
5588 | ||
923c61bb GN |
5589 | max_bits = (sizeof sregs->interrupt_bitmap) << 3; |
5590 | pending_vec = find_first_bit( | |
5591 | (const unsigned long *)sregs->interrupt_bitmap, max_bits); | |
5592 | if (pending_vec < max_bits) { | |
66fd3f7f | 5593 | kvm_queue_interrupt(vcpu, pending_vec, false); |
923c61bb GN |
5594 | pr_debug("Set back pending irq %d\n", pending_vec); |
5595 | if (irqchip_in_kernel(vcpu->kvm)) | |
5596 | kvm_pic_clear_isr_ack(vcpu->kvm); | |
b6c7a5dc HB |
5597 | } |
5598 | ||
3e6e0aab GT |
5599 | kvm_set_segment(vcpu, &sregs->cs, VCPU_SREG_CS); |
5600 | kvm_set_segment(vcpu, &sregs->ds, VCPU_SREG_DS); | |
5601 | kvm_set_segment(vcpu, &sregs->es, VCPU_SREG_ES); | |
5602 | kvm_set_segment(vcpu, &sregs->fs, VCPU_SREG_FS); | |
5603 | kvm_set_segment(vcpu, &sregs->gs, VCPU_SREG_GS); | |
5604 | kvm_set_segment(vcpu, &sregs->ss, VCPU_SREG_SS); | |
b6c7a5dc | 5605 | |
3e6e0aab GT |
5606 | kvm_set_segment(vcpu, &sregs->tr, VCPU_SREG_TR); |
5607 | kvm_set_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR); | |
b6c7a5dc | 5608 | |
5f0269f5 ME |
5609 | update_cr8_intercept(vcpu); |
5610 | ||
9c3e4aab | 5611 | /* Older userspace won't unhalt the vcpu on reset. */ |
c5af89b6 | 5612 | if (kvm_vcpu_is_bsp(vcpu) && kvm_rip_read(vcpu) == 0xfff0 && |
9c3e4aab | 5613 | sregs->cs.selector == 0xf000 && sregs->cs.base == 0xffff0000 && |
3eeb3288 | 5614 | !is_protmode(vcpu)) |
9c3e4aab MT |
5615 | vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE; |
5616 | ||
3842d135 AK |
5617 | kvm_make_request(KVM_REQ_EVENT, vcpu); |
5618 | ||
b6c7a5dc HB |
5619 | return 0; |
5620 | } | |
5621 | ||
d0bfb940 JK |
5622 | int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu, |
5623 | struct kvm_guest_debug *dbg) | |
b6c7a5dc | 5624 | { |
355be0b9 | 5625 | unsigned long rflags; |
ae675ef0 | 5626 | int i, r; |
b6c7a5dc | 5627 | |
4f926bf2 JK |
5628 | if (dbg->control & (KVM_GUESTDBG_INJECT_DB | KVM_GUESTDBG_INJECT_BP)) { |
5629 | r = -EBUSY; | |
5630 | if (vcpu->arch.exception.pending) | |
2122ff5e | 5631 | goto out; |
4f926bf2 JK |
5632 | if (dbg->control & KVM_GUESTDBG_INJECT_DB) |
5633 | kvm_queue_exception(vcpu, DB_VECTOR); | |
5634 | else | |
5635 | kvm_queue_exception(vcpu, BP_VECTOR); | |
5636 | } | |
5637 | ||
91586a3b JK |
5638 | /* |
5639 | * Read rflags as long as potentially injected trace flags are still | |
5640 | * filtered out. | |
5641 | */ | |
5642 | rflags = kvm_get_rflags(vcpu); | |
355be0b9 JK |
5643 | |
5644 | vcpu->guest_debug = dbg->control; | |
5645 | if (!(vcpu->guest_debug & KVM_GUESTDBG_ENABLE)) | |
5646 | vcpu->guest_debug = 0; | |
5647 | ||
5648 | if (vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP) { | |
ae675ef0 JK |
5649 | for (i = 0; i < KVM_NR_DB_REGS; ++i) |
5650 | vcpu->arch.eff_db[i] = dbg->arch.debugreg[i]; | |
5651 | vcpu->arch.switch_db_regs = | |
5652 | (dbg->arch.debugreg[7] & DR7_BP_EN_MASK); | |
5653 | } else { | |
5654 | for (i = 0; i < KVM_NR_DB_REGS; i++) | |
5655 | vcpu->arch.eff_db[i] = vcpu->arch.db[i]; | |
5656 | vcpu->arch.switch_db_regs = (vcpu->arch.dr7 & DR7_BP_EN_MASK); | |
5657 | } | |
5658 | ||
f92653ee JK |
5659 | if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP) |
5660 | vcpu->arch.singlestep_rip = kvm_rip_read(vcpu) + | |
5661 | get_segment_base(vcpu, VCPU_SREG_CS); | |
94fe45da | 5662 | |
91586a3b JK |
5663 | /* |
5664 | * Trigger an rflags update that will inject or remove the trace | |
5665 | * flags. | |
5666 | */ | |
5667 | kvm_set_rflags(vcpu, rflags); | |
b6c7a5dc | 5668 | |
355be0b9 | 5669 | kvm_x86_ops->set_guest_debug(vcpu, dbg); |
b6c7a5dc | 5670 | |
4f926bf2 | 5671 | r = 0; |
d0bfb940 | 5672 | |
2122ff5e | 5673 | out: |
b6c7a5dc HB |
5674 | |
5675 | return r; | |
5676 | } | |
5677 | ||
8b006791 ZX |
5678 | /* |
5679 | * Translate a guest virtual address to a guest physical address. | |
5680 | */ | |
5681 | int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu, | |
5682 | struct kvm_translation *tr) | |
5683 | { | |
5684 | unsigned long vaddr = tr->linear_address; | |
5685 | gpa_t gpa; | |
f656ce01 | 5686 | int idx; |
8b006791 | 5687 | |
f656ce01 | 5688 | idx = srcu_read_lock(&vcpu->kvm->srcu); |
1871c602 | 5689 | gpa = kvm_mmu_gva_to_gpa_system(vcpu, vaddr, NULL); |
f656ce01 | 5690 | srcu_read_unlock(&vcpu->kvm->srcu, idx); |
8b006791 ZX |
5691 | tr->physical_address = gpa; |
5692 | tr->valid = gpa != UNMAPPED_GVA; | |
5693 | tr->writeable = 1; | |
5694 | tr->usermode = 0; | |
8b006791 ZX |
5695 | |
5696 | return 0; | |
5697 | } | |
5698 | ||
d0752060 HB |
5699 | int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu) |
5700 | { | |
98918833 SY |
5701 | struct i387_fxsave_struct *fxsave = |
5702 | &vcpu->arch.guest_fpu.state->fxsave; | |
d0752060 | 5703 | |
d0752060 HB |
5704 | memcpy(fpu->fpr, fxsave->st_space, 128); |
5705 | fpu->fcw = fxsave->cwd; | |
5706 | fpu->fsw = fxsave->swd; | |
5707 | fpu->ftwx = fxsave->twd; | |
5708 | fpu->last_opcode = fxsave->fop; | |
5709 | fpu->last_ip = fxsave->rip; | |
5710 | fpu->last_dp = fxsave->rdp; | |
5711 | memcpy(fpu->xmm, fxsave->xmm_space, sizeof fxsave->xmm_space); | |
5712 | ||
d0752060 HB |
5713 | return 0; |
5714 | } | |
5715 | ||
5716 | int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu) | |
5717 | { | |
98918833 SY |
5718 | struct i387_fxsave_struct *fxsave = |
5719 | &vcpu->arch.guest_fpu.state->fxsave; | |
d0752060 | 5720 | |
d0752060 HB |
5721 | memcpy(fxsave->st_space, fpu->fpr, 128); |
5722 | fxsave->cwd = fpu->fcw; | |
5723 | fxsave->swd = fpu->fsw; | |
5724 | fxsave->twd = fpu->ftwx; | |
5725 | fxsave->fop = fpu->last_opcode; | |
5726 | fxsave->rip = fpu->last_ip; | |
5727 | fxsave->rdp = fpu->last_dp; | |
5728 | memcpy(fxsave->xmm_space, fpu->xmm, sizeof fxsave->xmm_space); | |
5729 | ||
d0752060 HB |
5730 | return 0; |
5731 | } | |
5732 | ||
10ab25cd | 5733 | int fx_init(struct kvm_vcpu *vcpu) |
d0752060 | 5734 | { |
10ab25cd JK |
5735 | int err; |
5736 | ||
5737 | err = fpu_alloc(&vcpu->arch.guest_fpu); | |
5738 | if (err) | |
5739 | return err; | |
5740 | ||
98918833 | 5741 | fpu_finit(&vcpu->arch.guest_fpu); |
d0752060 | 5742 | |
2acf923e DC |
5743 | /* |
5744 | * Ensure guest xcr0 is valid for loading | |
5745 | */ | |
5746 | vcpu->arch.xcr0 = XSTATE_FP; | |
5747 | ||
ad312c7c | 5748 | vcpu->arch.cr0 |= X86_CR0_ET; |
10ab25cd JK |
5749 | |
5750 | return 0; | |
d0752060 HB |
5751 | } |
5752 | EXPORT_SYMBOL_GPL(fx_init); | |
5753 | ||
98918833 SY |
5754 | static void fx_free(struct kvm_vcpu *vcpu) |
5755 | { | |
5756 | fpu_free(&vcpu->arch.guest_fpu); | |
5757 | } | |
5758 | ||
d0752060 HB |
5759 | void kvm_load_guest_fpu(struct kvm_vcpu *vcpu) |
5760 | { | |
2608d7a1 | 5761 | if (vcpu->guest_fpu_loaded) |
d0752060 HB |
5762 | return; |
5763 | ||
2acf923e DC |
5764 | /* |
5765 | * Restore all possible states in the guest, | |
5766 | * and assume host would use all available bits. | |
5767 | * Guest xcr0 would be loaded later. | |
5768 | */ | |
5769 | kvm_put_guest_xcr0(vcpu); | |
d0752060 | 5770 | vcpu->guest_fpu_loaded = 1; |
7cf30855 | 5771 | unlazy_fpu(current); |
98918833 | 5772 | fpu_restore_checking(&vcpu->arch.guest_fpu); |
0c04851c | 5773 | trace_kvm_fpu(1); |
d0752060 | 5774 | } |
d0752060 HB |
5775 | |
5776 | void kvm_put_guest_fpu(struct kvm_vcpu *vcpu) | |
5777 | { | |
2acf923e DC |
5778 | kvm_put_guest_xcr0(vcpu); |
5779 | ||
d0752060 HB |
5780 | if (!vcpu->guest_fpu_loaded) |
5781 | return; | |
5782 | ||
5783 | vcpu->guest_fpu_loaded = 0; | |
98918833 | 5784 | fpu_save_init(&vcpu->arch.guest_fpu); |
f096ed85 | 5785 | ++vcpu->stat.fpu_reload; |
a8eeb04a | 5786 | kvm_make_request(KVM_REQ_DEACTIVATE_FPU, vcpu); |
0c04851c | 5787 | trace_kvm_fpu(0); |
d0752060 | 5788 | } |
e9b11c17 ZX |
5789 | |
5790 | void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu) | |
5791 | { | |
7f1ea208 JR |
5792 | if (vcpu->arch.time_page) { |
5793 | kvm_release_page_dirty(vcpu->arch.time_page); | |
5794 | vcpu->arch.time_page = NULL; | |
5795 | } | |
5796 | ||
f5f48ee1 | 5797 | free_cpumask_var(vcpu->arch.wbinvd_dirty_mask); |
98918833 | 5798 | fx_free(vcpu); |
e9b11c17 ZX |
5799 | kvm_x86_ops->vcpu_free(vcpu); |
5800 | } | |
5801 | ||
5802 | struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, | |
5803 | unsigned int id) | |
5804 | { | |
6755bae8 ZA |
5805 | if (check_tsc_unstable() && atomic_read(&kvm->online_vcpus) != 0) |
5806 | printk_once(KERN_WARNING | |
5807 | "kvm: SMP vm created on host with unstable TSC; " | |
5808 | "guest TSC will not be reliable\n"); | |
26e5215f AK |
5809 | return kvm_x86_ops->vcpu_create(kvm, id); |
5810 | } | |
e9b11c17 | 5811 | |
26e5215f AK |
5812 | int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu) |
5813 | { | |
5814 | int r; | |
e9b11c17 | 5815 | |
0bed3b56 | 5816 | vcpu->arch.mtrr_state.have_fixed = 1; |
e9b11c17 ZX |
5817 | vcpu_load(vcpu); |
5818 | r = kvm_arch_vcpu_reset(vcpu); | |
5819 | if (r == 0) | |
5820 | r = kvm_mmu_setup(vcpu); | |
5821 | vcpu_put(vcpu); | |
5822 | if (r < 0) | |
5823 | goto free_vcpu; | |
5824 | ||
26e5215f | 5825 | return 0; |
e9b11c17 ZX |
5826 | free_vcpu: |
5827 | kvm_x86_ops->vcpu_free(vcpu); | |
26e5215f | 5828 | return r; |
e9b11c17 ZX |
5829 | } |
5830 | ||
d40ccc62 | 5831 | void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu) |
e9b11c17 | 5832 | { |
344d9588 GN |
5833 | vcpu->arch.apf.msr_val = 0; |
5834 | ||
e9b11c17 ZX |
5835 | vcpu_load(vcpu); |
5836 | kvm_mmu_unload(vcpu); | |
5837 | vcpu_put(vcpu); | |
5838 | ||
98918833 | 5839 | fx_free(vcpu); |
e9b11c17 ZX |
5840 | kvm_x86_ops->vcpu_free(vcpu); |
5841 | } | |
5842 | ||
5843 | int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu) | |
5844 | { | |
448fa4a9 JK |
5845 | vcpu->arch.nmi_pending = false; |
5846 | vcpu->arch.nmi_injected = false; | |
5847 | ||
42dbaa5a JK |
5848 | vcpu->arch.switch_db_regs = 0; |
5849 | memset(vcpu->arch.db, 0, sizeof(vcpu->arch.db)); | |
5850 | vcpu->arch.dr6 = DR6_FIXED_1; | |
5851 | vcpu->arch.dr7 = DR7_FIXED_1; | |
5852 | ||
3842d135 | 5853 | kvm_make_request(KVM_REQ_EVENT, vcpu); |
344d9588 | 5854 | vcpu->arch.apf.msr_val = 0; |
3842d135 | 5855 | |
af585b92 GN |
5856 | kvm_clear_async_pf_completion_queue(vcpu); |
5857 | kvm_async_pf_hash_reset(vcpu); | |
5858 | vcpu->arch.apf.halted = false; | |
5859 | ||
e9b11c17 ZX |
5860 | return kvm_x86_ops->vcpu_reset(vcpu); |
5861 | } | |
5862 | ||
10474ae8 | 5863 | int kvm_arch_hardware_enable(void *garbage) |
e9b11c17 | 5864 | { |
ca84d1a2 ZA |
5865 | struct kvm *kvm; |
5866 | struct kvm_vcpu *vcpu; | |
5867 | int i; | |
18863bdd AK |
5868 | |
5869 | kvm_shared_msr_cpu_online(); | |
ca84d1a2 ZA |
5870 | list_for_each_entry(kvm, &vm_list, vm_list) |
5871 | kvm_for_each_vcpu(i, vcpu, kvm) | |
5872 | if (vcpu->cpu == smp_processor_id()) | |
c285545f | 5873 | kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu); |
10474ae8 | 5874 | return kvm_x86_ops->hardware_enable(garbage); |
e9b11c17 ZX |
5875 | } |
5876 | ||
5877 | void kvm_arch_hardware_disable(void *garbage) | |
5878 | { | |
5879 | kvm_x86_ops->hardware_disable(garbage); | |
3548bab5 | 5880 | drop_user_return_notifiers(garbage); |
e9b11c17 ZX |
5881 | } |
5882 | ||
5883 | int kvm_arch_hardware_setup(void) | |
5884 | { | |
5885 | return kvm_x86_ops->hardware_setup(); | |
5886 | } | |
5887 | ||
5888 | void kvm_arch_hardware_unsetup(void) | |
5889 | { | |
5890 | kvm_x86_ops->hardware_unsetup(); | |
5891 | } | |
5892 | ||
5893 | void kvm_arch_check_processor_compat(void *rtn) | |
5894 | { | |
5895 | kvm_x86_ops->check_processor_compatibility(rtn); | |
5896 | } | |
5897 | ||
5898 | int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu) | |
5899 | { | |
5900 | struct page *page; | |
5901 | struct kvm *kvm; | |
5902 | int r; | |
5903 | ||
5904 | BUG_ON(vcpu->kvm == NULL); | |
5905 | kvm = vcpu->kvm; | |
5906 | ||
9aabc88f | 5907 | vcpu->arch.emulate_ctxt.ops = &emulate_ops; |
14dfe855 | 5908 | vcpu->arch.walk_mmu = &vcpu->arch.mmu; |
ad312c7c | 5909 | vcpu->arch.mmu.root_hpa = INVALID_PAGE; |
c30a358d | 5910 | vcpu->arch.mmu.translate_gpa = translate_gpa; |
02f59dc9 | 5911 | vcpu->arch.nested_mmu.translate_gpa = translate_nested_gpa; |
c5af89b6 | 5912 | if (!irqchip_in_kernel(kvm) || kvm_vcpu_is_bsp(vcpu)) |
a4535290 | 5913 | vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE; |
e9b11c17 | 5914 | else |
a4535290 | 5915 | vcpu->arch.mp_state = KVM_MP_STATE_UNINITIALIZED; |
e9b11c17 ZX |
5916 | |
5917 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
5918 | if (!page) { | |
5919 | r = -ENOMEM; | |
5920 | goto fail; | |
5921 | } | |
ad312c7c | 5922 | vcpu->arch.pio_data = page_address(page); |
e9b11c17 | 5923 | |
c285545f ZA |
5924 | if (!kvm->arch.virtual_tsc_khz) |
5925 | kvm_arch_set_tsc_khz(kvm, max_tsc_khz); | |
5926 | ||
e9b11c17 ZX |
5927 | r = kvm_mmu_create(vcpu); |
5928 | if (r < 0) | |
5929 | goto fail_free_pio_data; | |
5930 | ||
5931 | if (irqchip_in_kernel(kvm)) { | |
5932 | r = kvm_create_lapic(vcpu); | |
5933 | if (r < 0) | |
5934 | goto fail_mmu_destroy; | |
5935 | } | |
5936 | ||
890ca9ae HY |
5937 | vcpu->arch.mce_banks = kzalloc(KVM_MAX_MCE_BANKS * sizeof(u64) * 4, |
5938 | GFP_KERNEL); | |
5939 | if (!vcpu->arch.mce_banks) { | |
5940 | r = -ENOMEM; | |
443c39bc | 5941 | goto fail_free_lapic; |
890ca9ae HY |
5942 | } |
5943 | vcpu->arch.mcg_cap = KVM_MAX_MCE_BANKS; | |
5944 | ||
f5f48ee1 SY |
5945 | if (!zalloc_cpumask_var(&vcpu->arch.wbinvd_dirty_mask, GFP_KERNEL)) |
5946 | goto fail_free_mce_banks; | |
5947 | ||
af585b92 GN |
5948 | kvm_async_pf_hash_reset(vcpu); |
5949 | ||
e9b11c17 | 5950 | return 0; |
f5f48ee1 SY |
5951 | fail_free_mce_banks: |
5952 | kfree(vcpu->arch.mce_banks); | |
443c39bc WY |
5953 | fail_free_lapic: |
5954 | kvm_free_lapic(vcpu); | |
e9b11c17 ZX |
5955 | fail_mmu_destroy: |
5956 | kvm_mmu_destroy(vcpu); | |
5957 | fail_free_pio_data: | |
ad312c7c | 5958 | free_page((unsigned long)vcpu->arch.pio_data); |
e9b11c17 ZX |
5959 | fail: |
5960 | return r; | |
5961 | } | |
5962 | ||
5963 | void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu) | |
5964 | { | |
f656ce01 MT |
5965 | int idx; |
5966 | ||
36cb93fd | 5967 | kfree(vcpu->arch.mce_banks); |
e9b11c17 | 5968 | kvm_free_lapic(vcpu); |
f656ce01 | 5969 | idx = srcu_read_lock(&vcpu->kvm->srcu); |
e9b11c17 | 5970 | kvm_mmu_destroy(vcpu); |
f656ce01 | 5971 | srcu_read_unlock(&vcpu->kvm->srcu, idx); |
ad312c7c | 5972 | free_page((unsigned long)vcpu->arch.pio_data); |
e9b11c17 | 5973 | } |
d19a9cd2 | 5974 | |
d89f5eff | 5975 | int kvm_arch_init_vm(struct kvm *kvm) |
d19a9cd2 | 5976 | { |
f05e70ac | 5977 | INIT_LIST_HEAD(&kvm->arch.active_mmu_pages); |
4d5c5d0f | 5978 | INIT_LIST_HEAD(&kvm->arch.assigned_dev_head); |
d19a9cd2 | 5979 | |
5550af4d SY |
5980 | /* Reserve bit 0 of irq_sources_bitmap for userspace irq source */ |
5981 | set_bit(KVM_USERSPACE_IRQ_SOURCE_ID, &kvm->arch.irq_sources_bitmap); | |
5982 | ||
99e3e30a | 5983 | spin_lock_init(&kvm->arch.tsc_write_lock); |
53f658b3 | 5984 | |
d89f5eff | 5985 | return 0; |
d19a9cd2 ZX |
5986 | } |
5987 | ||
5988 | static void kvm_unload_vcpu_mmu(struct kvm_vcpu *vcpu) | |
5989 | { | |
5990 | vcpu_load(vcpu); | |
5991 | kvm_mmu_unload(vcpu); | |
5992 | vcpu_put(vcpu); | |
5993 | } | |
5994 | ||
5995 | static void kvm_free_vcpus(struct kvm *kvm) | |
5996 | { | |
5997 | unsigned int i; | |
988a2cae | 5998 | struct kvm_vcpu *vcpu; |
d19a9cd2 ZX |
5999 | |
6000 | /* | |
6001 | * Unpin any mmu pages first. | |
6002 | */ | |
af585b92 GN |
6003 | kvm_for_each_vcpu(i, vcpu, kvm) { |
6004 | kvm_clear_async_pf_completion_queue(vcpu); | |
988a2cae | 6005 | kvm_unload_vcpu_mmu(vcpu); |
af585b92 | 6006 | } |
988a2cae GN |
6007 | kvm_for_each_vcpu(i, vcpu, kvm) |
6008 | kvm_arch_vcpu_free(vcpu); | |
6009 | ||
6010 | mutex_lock(&kvm->lock); | |
6011 | for (i = 0; i < atomic_read(&kvm->online_vcpus); i++) | |
6012 | kvm->vcpus[i] = NULL; | |
d19a9cd2 | 6013 | |
988a2cae GN |
6014 | atomic_set(&kvm->online_vcpus, 0); |
6015 | mutex_unlock(&kvm->lock); | |
d19a9cd2 ZX |
6016 | } |
6017 | ||
ad8ba2cd SY |
6018 | void kvm_arch_sync_events(struct kvm *kvm) |
6019 | { | |
ba4cef31 | 6020 | kvm_free_all_assigned_devices(kvm); |
aea924f6 | 6021 | kvm_free_pit(kvm); |
ad8ba2cd SY |
6022 | } |
6023 | ||
d19a9cd2 ZX |
6024 | void kvm_arch_destroy_vm(struct kvm *kvm) |
6025 | { | |
6eb55818 | 6026 | kvm_iommu_unmap_guest(kvm); |
d7deeeb0 ZX |
6027 | kfree(kvm->arch.vpic); |
6028 | kfree(kvm->arch.vioapic); | |
d19a9cd2 | 6029 | kvm_free_vcpus(kvm); |
3d45830c AK |
6030 | if (kvm->arch.apic_access_page) |
6031 | put_page(kvm->arch.apic_access_page); | |
b7ebfb05 SY |
6032 | if (kvm->arch.ept_identity_pagetable) |
6033 | put_page(kvm->arch.ept_identity_pagetable); | |
d19a9cd2 | 6034 | } |
0de10343 | 6035 | |
f7784b8e MT |
6036 | int kvm_arch_prepare_memory_region(struct kvm *kvm, |
6037 | struct kvm_memory_slot *memslot, | |
0de10343 | 6038 | struct kvm_memory_slot old, |
f7784b8e | 6039 | struct kvm_userspace_memory_region *mem, |
0de10343 ZX |
6040 | int user_alloc) |
6041 | { | |
f7784b8e | 6042 | int npages = memslot->npages; |
7ac77099 AK |
6043 | int map_flags = MAP_PRIVATE | MAP_ANONYMOUS; |
6044 | ||
6045 | /* Prevent internal slot pages from being moved by fork()/COW. */ | |
6046 | if (memslot->id >= KVM_MEMORY_SLOTS) | |
6047 | map_flags = MAP_SHARED | MAP_ANONYMOUS; | |
0de10343 ZX |
6048 | |
6049 | /*To keep backward compatibility with older userspace, | |
6050 | *x86 needs to hanlde !user_alloc case. | |
6051 | */ | |
6052 | if (!user_alloc) { | |
6053 | if (npages && !old.rmap) { | |
604b38ac AA |
6054 | unsigned long userspace_addr; |
6055 | ||
72dc67a6 | 6056 | down_write(¤t->mm->mmap_sem); |
604b38ac AA |
6057 | userspace_addr = do_mmap(NULL, 0, |
6058 | npages * PAGE_SIZE, | |
6059 | PROT_READ | PROT_WRITE, | |
7ac77099 | 6060 | map_flags, |
604b38ac | 6061 | 0); |
72dc67a6 | 6062 | up_write(¤t->mm->mmap_sem); |
0de10343 | 6063 | |
604b38ac AA |
6064 | if (IS_ERR((void *)userspace_addr)) |
6065 | return PTR_ERR((void *)userspace_addr); | |
6066 | ||
604b38ac | 6067 | memslot->userspace_addr = userspace_addr; |
0de10343 ZX |
6068 | } |
6069 | } | |
6070 | ||
f7784b8e MT |
6071 | |
6072 | return 0; | |
6073 | } | |
6074 | ||
6075 | void kvm_arch_commit_memory_region(struct kvm *kvm, | |
6076 | struct kvm_userspace_memory_region *mem, | |
6077 | struct kvm_memory_slot old, | |
6078 | int user_alloc) | |
6079 | { | |
6080 | ||
6081 | int npages = mem->memory_size >> PAGE_SHIFT; | |
6082 | ||
6083 | if (!user_alloc && !old.user_alloc && old.rmap && !npages) { | |
6084 | int ret; | |
6085 | ||
6086 | down_write(¤t->mm->mmap_sem); | |
6087 | ret = do_munmap(current->mm, old.userspace_addr, | |
6088 | old.npages * PAGE_SIZE); | |
6089 | up_write(¤t->mm->mmap_sem); | |
6090 | if (ret < 0) | |
6091 | printk(KERN_WARNING | |
6092 | "kvm_vm_ioctl_set_memory_region: " | |
6093 | "failed to munmap memory\n"); | |
6094 | } | |
6095 | ||
7c8a83b7 | 6096 | spin_lock(&kvm->mmu_lock); |
f05e70ac | 6097 | if (!kvm->arch.n_requested_mmu_pages) { |
0de10343 ZX |
6098 | unsigned int nr_mmu_pages = kvm_mmu_calculate_mmu_pages(kvm); |
6099 | kvm_mmu_change_mmu_pages(kvm, nr_mmu_pages); | |
6100 | } | |
6101 | ||
6102 | kvm_mmu_slot_remove_write_access(kvm, mem->slot); | |
7c8a83b7 | 6103 | spin_unlock(&kvm->mmu_lock); |
0de10343 | 6104 | } |
1d737c8a | 6105 | |
34d4cb8f MT |
6106 | void kvm_arch_flush_shadow(struct kvm *kvm) |
6107 | { | |
6108 | kvm_mmu_zap_all(kvm); | |
8986ecc0 | 6109 | kvm_reload_remote_mmus(kvm); |
34d4cb8f MT |
6110 | } |
6111 | ||
1d737c8a ZX |
6112 | int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu) |
6113 | { | |
af585b92 GN |
6114 | return (vcpu->arch.mp_state == KVM_MP_STATE_RUNNABLE && |
6115 | !vcpu->arch.apf.halted) | |
6116 | || !list_empty_careful(&vcpu->async_pf.done) | |
a1b37100 GN |
6117 | || vcpu->arch.mp_state == KVM_MP_STATE_SIPI_RECEIVED |
6118 | || vcpu->arch.nmi_pending || | |
6119 | (kvm_arch_interrupt_allowed(vcpu) && | |
6120 | kvm_cpu_has_interrupt(vcpu)); | |
1d737c8a | 6121 | } |
5736199a | 6122 | |
5736199a ZX |
6123 | void kvm_vcpu_kick(struct kvm_vcpu *vcpu) |
6124 | { | |
32f88400 MT |
6125 | int me; |
6126 | int cpu = vcpu->cpu; | |
5736199a ZX |
6127 | |
6128 | if (waitqueue_active(&vcpu->wq)) { | |
6129 | wake_up_interruptible(&vcpu->wq); | |
6130 | ++vcpu->stat.halt_wakeup; | |
6131 | } | |
32f88400 MT |
6132 | |
6133 | me = get_cpu(); | |
6134 | if (cpu != me && (unsigned)cpu < nr_cpu_ids && cpu_online(cpu)) | |
d94e1dc9 | 6135 | if (atomic_xchg(&vcpu->guest_mode, 0)) |
32f88400 | 6136 | smp_send_reschedule(cpu); |
e9571ed5 | 6137 | put_cpu(); |
5736199a | 6138 | } |
78646121 GN |
6139 | |
6140 | int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu) | |
6141 | { | |
6142 | return kvm_x86_ops->interrupt_allowed(vcpu); | |
6143 | } | |
229456fc | 6144 | |
f92653ee JK |
6145 | bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip) |
6146 | { | |
6147 | unsigned long current_rip = kvm_rip_read(vcpu) + | |
6148 | get_segment_base(vcpu, VCPU_SREG_CS); | |
6149 | ||
6150 | return current_rip == linear_rip; | |
6151 | } | |
6152 | EXPORT_SYMBOL_GPL(kvm_is_linear_rip); | |
6153 | ||
94fe45da JK |
6154 | unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu) |
6155 | { | |
6156 | unsigned long rflags; | |
6157 | ||
6158 | rflags = kvm_x86_ops->get_rflags(vcpu); | |
6159 | if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP) | |
c310bac5 | 6160 | rflags &= ~X86_EFLAGS_TF; |
94fe45da JK |
6161 | return rflags; |
6162 | } | |
6163 | EXPORT_SYMBOL_GPL(kvm_get_rflags); | |
6164 | ||
6165 | void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags) | |
6166 | { | |
6167 | if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP && | |
f92653ee | 6168 | kvm_is_linear_rip(vcpu, vcpu->arch.singlestep_rip)) |
c310bac5 | 6169 | rflags |= X86_EFLAGS_TF; |
94fe45da | 6170 | kvm_x86_ops->set_rflags(vcpu, rflags); |
3842d135 | 6171 | kvm_make_request(KVM_REQ_EVENT, vcpu); |
94fe45da JK |
6172 | } |
6173 | EXPORT_SYMBOL_GPL(kvm_set_rflags); | |
6174 | ||
56028d08 GN |
6175 | void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu, struct kvm_async_pf *work) |
6176 | { | |
6177 | int r; | |
6178 | ||
c4806acd XG |
6179 | if (!vcpu->arch.mmu.direct_map || !work->arch.direct_map || |
6180 | is_error_page(work->page)) | |
56028d08 GN |
6181 | return; |
6182 | ||
6183 | r = kvm_mmu_reload(vcpu); | |
6184 | if (unlikely(r)) | |
6185 | return; | |
6186 | ||
6187 | vcpu->arch.mmu.page_fault(vcpu, work->gva, 0, true); | |
6188 | } | |
6189 | ||
af585b92 GN |
6190 | static inline u32 kvm_async_pf_hash_fn(gfn_t gfn) |
6191 | { | |
6192 | return hash_32(gfn & 0xffffffff, order_base_2(ASYNC_PF_PER_VCPU)); | |
6193 | } | |
6194 | ||
6195 | static inline u32 kvm_async_pf_next_probe(u32 key) | |
6196 | { | |
6197 | return (key + 1) & (roundup_pow_of_two(ASYNC_PF_PER_VCPU) - 1); | |
6198 | } | |
6199 | ||
6200 | static void kvm_add_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn) | |
6201 | { | |
6202 | u32 key = kvm_async_pf_hash_fn(gfn); | |
6203 | ||
6204 | while (vcpu->arch.apf.gfns[key] != ~0) | |
6205 | key = kvm_async_pf_next_probe(key); | |
6206 | ||
6207 | vcpu->arch.apf.gfns[key] = gfn; | |
6208 | } | |
6209 | ||
6210 | static u32 kvm_async_pf_gfn_slot(struct kvm_vcpu *vcpu, gfn_t gfn) | |
6211 | { | |
6212 | int i; | |
6213 | u32 key = kvm_async_pf_hash_fn(gfn); | |
6214 | ||
6215 | for (i = 0; i < roundup_pow_of_two(ASYNC_PF_PER_VCPU) && | |
c7d28c24 XG |
6216 | (vcpu->arch.apf.gfns[key] != gfn && |
6217 | vcpu->arch.apf.gfns[key] != ~0); i++) | |
af585b92 GN |
6218 | key = kvm_async_pf_next_probe(key); |
6219 | ||
6220 | return key; | |
6221 | } | |
6222 | ||
6223 | bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn) | |
6224 | { | |
6225 | return vcpu->arch.apf.gfns[kvm_async_pf_gfn_slot(vcpu, gfn)] == gfn; | |
6226 | } | |
6227 | ||
6228 | static void kvm_del_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn) | |
6229 | { | |
6230 | u32 i, j, k; | |
6231 | ||
6232 | i = j = kvm_async_pf_gfn_slot(vcpu, gfn); | |
6233 | while (true) { | |
6234 | vcpu->arch.apf.gfns[i] = ~0; | |
6235 | do { | |
6236 | j = kvm_async_pf_next_probe(j); | |
6237 | if (vcpu->arch.apf.gfns[j] == ~0) | |
6238 | return; | |
6239 | k = kvm_async_pf_hash_fn(vcpu->arch.apf.gfns[j]); | |
6240 | /* | |
6241 | * k lies cyclically in ]i,j] | |
6242 | * | i.k.j | | |
6243 | * |....j i.k.| or |.k..j i...| | |
6244 | */ | |
6245 | } while ((i <= j) ? (i < k && k <= j) : (i < k || k <= j)); | |
6246 | vcpu->arch.apf.gfns[i] = vcpu->arch.apf.gfns[j]; | |
6247 | i = j; | |
6248 | } | |
6249 | } | |
6250 | ||
7c90705b GN |
6251 | static int apf_put_user(struct kvm_vcpu *vcpu, u32 val) |
6252 | { | |
6253 | ||
6254 | return kvm_write_guest_cached(vcpu->kvm, &vcpu->arch.apf.data, &val, | |
6255 | sizeof(val)); | |
6256 | } | |
6257 | ||
af585b92 GN |
6258 | void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu, |
6259 | struct kvm_async_pf *work) | |
6260 | { | |
6389ee94 AK |
6261 | struct x86_exception fault; |
6262 | ||
7c90705b | 6263 | trace_kvm_async_pf_not_present(work->arch.token, work->gva); |
af585b92 | 6264 | kvm_add_async_pf_gfn(vcpu, work->arch.gfn); |
7c90705b GN |
6265 | |
6266 | if (!(vcpu->arch.apf.msr_val & KVM_ASYNC_PF_ENABLED) || | |
fc5f06fa GN |
6267 | (vcpu->arch.apf.send_user_only && |
6268 | kvm_x86_ops->get_cpl(vcpu) == 0)) | |
7c90705b GN |
6269 | kvm_make_request(KVM_REQ_APF_HALT, vcpu); |
6270 | else if (!apf_put_user(vcpu, KVM_PV_REASON_PAGE_NOT_PRESENT)) { | |
6389ee94 AK |
6271 | fault.vector = PF_VECTOR; |
6272 | fault.error_code_valid = true; | |
6273 | fault.error_code = 0; | |
6274 | fault.nested_page_fault = false; | |
6275 | fault.address = work->arch.token; | |
6276 | kvm_inject_page_fault(vcpu, &fault); | |
7c90705b | 6277 | } |
af585b92 GN |
6278 | } |
6279 | ||
6280 | void kvm_arch_async_page_present(struct kvm_vcpu *vcpu, | |
6281 | struct kvm_async_pf *work) | |
6282 | { | |
6389ee94 AK |
6283 | struct x86_exception fault; |
6284 | ||
7c90705b GN |
6285 | trace_kvm_async_pf_ready(work->arch.token, work->gva); |
6286 | if (is_error_page(work->page)) | |
6287 | work->arch.token = ~0; /* broadcast wakeup */ | |
6288 | else | |
6289 | kvm_del_async_pf_gfn(vcpu, work->arch.gfn); | |
6290 | ||
6291 | if ((vcpu->arch.apf.msr_val & KVM_ASYNC_PF_ENABLED) && | |
6292 | !apf_put_user(vcpu, KVM_PV_REASON_PAGE_READY)) { | |
6389ee94 AK |
6293 | fault.vector = PF_VECTOR; |
6294 | fault.error_code_valid = true; | |
6295 | fault.error_code = 0; | |
6296 | fault.nested_page_fault = false; | |
6297 | fault.address = work->arch.token; | |
6298 | kvm_inject_page_fault(vcpu, &fault); | |
7c90705b | 6299 | } |
e6d53e3b | 6300 | vcpu->arch.apf.halted = false; |
7c90705b GN |
6301 | } |
6302 | ||
6303 | bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu) | |
6304 | { | |
6305 | if (!(vcpu->arch.apf.msr_val & KVM_ASYNC_PF_ENABLED)) | |
6306 | return true; | |
6307 | else | |
6308 | return !kvm_event_needs_reinjection(vcpu) && | |
6309 | kvm_x86_ops->interrupt_allowed(vcpu); | |
af585b92 GN |
6310 | } |
6311 | ||
229456fc MT |
6312 | EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_exit); |
6313 | EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_inj_virq); | |
6314 | EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_page_fault); | |
6315 | EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_msr); | |
6316 | EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_cr); | |
0ac406de | 6317 | EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_nested_vmrun); |
d8cabddf | 6318 | EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_nested_vmexit); |
17897f36 | 6319 | EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_nested_vmexit_inject); |
236649de | 6320 | EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_nested_intr_vmexit); |
ec1ff790 | 6321 | EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_invlpga); |
532a46b9 | 6322 | EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_skinit); |
2e554e8d | 6323 | EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_nested_intercepts); |