1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef ARCH_X86_KVM_X86_H
3 #define ARCH_X86_KVM_X86_H
4
5 #include <linux/kvm_host.h>
6 #include <asm/fpu/xstate.h>
7 #include <asm/mce.h>
8 #include <asm/pvclock.h>
9 #include "kvm_cache_regs.h"
10 #include "kvm_emulate.h"
11 #include "cpuid.h"
12
13 struct kvm_caps {
14 /* control of guest tsc rate supported? */
15 bool has_tsc_control;
16 /* maximum supported tsc_khz for guests */
17 u32 max_guest_tsc_khz;
18 /* number of bits of the fractional part of the TSC scaling ratio */
19 u8 tsc_scaling_ratio_frac_bits;
20 /* maximum allowed value of TSC scaling ratio */
21 u64 max_tsc_scaling_ratio;
22 /* 1ull << kvm_caps.tsc_scaling_ratio_frac_bits */
23 u64 default_tsc_scaling_ratio;
24 /* bus lock detection supported? */
25 bool has_bus_lock_exit;
26 /* notify VM exit supported? */
27 bool has_notify_vmexit;
28 /* bit mask of VM types */
29 u32 supported_vm_types;
30
31 u64 supported_mce_cap;
32 u64 supported_xcr0;
33 u64 supported_xss;
34 u64 supported_perf_cap;
35 };
36
37 struct kvm_host_values {
38 /*
39 * The host's raw MAXPHYADDR, i.e. the number of non-reserved physical
40 * address bits irrespective of features that repurpose legal bits,
41 * e.g. MKTME.
42 */
43 u8 maxphyaddr;
44
45 u64 efer;
46 u64 xcr0;
47 u64 xss;
48 u64 arch_capabilities;
49 };
50
51 void kvm_spurious_fault(void);
52
53 #define KVM_NESTED_VMENTER_CONSISTENCY_CHECK(consistency_check) \
54 ({ \
55 bool failed = (consistency_check); \
56 if (failed) \
57 trace_kvm_nested_vmenter_failed(#consistency_check, 0); \
58 failed; \
59 })
60
61 /*
62 * The first...last VMX feature MSRs that are emulated by KVM. This may or may
63 * not cover all known VMX MSRs, as KVM doesn't emulate an MSR until there's an
64 * associated feature that KVM supports for nested virtualization.
65 */
66 #define KVM_FIRST_EMULATED_VMX_MSR MSR_IA32_VMX_BASIC
67 #define KVM_LAST_EMULATED_VMX_MSR MSR_IA32_VMX_VMFUNC
68
69 #define KVM_DEFAULT_PLE_GAP 128
70 #define KVM_VMX_DEFAULT_PLE_WINDOW 4096
71 #define KVM_DEFAULT_PLE_WINDOW_GROW 2
72 #define KVM_DEFAULT_PLE_WINDOW_SHRINK 0
73 #define KVM_VMX_DEFAULT_PLE_WINDOW_MAX UINT_MAX
74 #define KVM_SVM_DEFAULT_PLE_WINDOW_MAX USHRT_MAX
75 #define KVM_SVM_DEFAULT_PLE_WINDOW 3000
76
__grow_ple_window(unsigned int val,unsigned int base,unsigned int modifier,unsigned int max)77 static inline unsigned int __grow_ple_window(unsigned int val,
78 unsigned int base, unsigned int modifier, unsigned int max)
79 {
80 u64 ret = val;
81
82 if (modifier < 1)
83 return base;
84
85 if (modifier < base)
86 ret *= modifier;
87 else
88 ret += modifier;
89
90 return min(ret, (u64)max);
91 }
92
__shrink_ple_window(unsigned int val,unsigned int base,unsigned int modifier,unsigned int min)93 static inline unsigned int __shrink_ple_window(unsigned int val,
94 unsigned int base, unsigned int modifier, unsigned int min)
95 {
96 if (modifier < 1)
97 return base;
98
99 if (modifier < base)
100 val /= modifier;
101 else
102 val -= modifier;
103
104 return max(val, min);
105 }
106
107 #define MSR_IA32_CR_PAT_DEFAULT \
108 PAT_VALUE(WB, WT, UC_MINUS, UC, WB, WT, UC_MINUS, UC)
109
110 void kvm_service_local_tlb_flush_requests(struct kvm_vcpu *vcpu);
111 int kvm_check_nested_events(struct kvm_vcpu *vcpu);
112
113 /* Forcibly leave the nested mode in cases like a vCPU reset */
kvm_leave_nested(struct kvm_vcpu * vcpu)114 static inline void kvm_leave_nested(struct kvm_vcpu *vcpu)
115 {
116 kvm_x86_ops.nested_ops->leave_nested(vcpu);
117 }
118
kvm_vcpu_has_run(struct kvm_vcpu * vcpu)119 static inline bool kvm_vcpu_has_run(struct kvm_vcpu *vcpu)
120 {
121 return vcpu->arch.last_vmentry_cpu != -1;
122 }
123
kvm_set_mp_state(struct kvm_vcpu * vcpu,int mp_state)124 static inline void kvm_set_mp_state(struct kvm_vcpu *vcpu, int mp_state)
125 {
126 vcpu->arch.mp_state = mp_state;
127 if (mp_state == KVM_MP_STATE_RUNNABLE)
128 vcpu->arch.pv.pv_unhalted = false;
129 }
130
kvm_is_exception_pending(struct kvm_vcpu * vcpu)131 static inline bool kvm_is_exception_pending(struct kvm_vcpu *vcpu)
132 {
133 return vcpu->arch.exception.pending ||
134 vcpu->arch.exception_vmexit.pending ||
135 kvm_test_request(KVM_REQ_TRIPLE_FAULT, vcpu);
136 }
137
kvm_clear_exception_queue(struct kvm_vcpu * vcpu)138 static inline void kvm_clear_exception_queue(struct kvm_vcpu *vcpu)
139 {
140 vcpu->arch.exception.pending = false;
141 vcpu->arch.exception.injected = false;
142 vcpu->arch.exception_vmexit.pending = false;
143 }
144
kvm_queue_interrupt(struct kvm_vcpu * vcpu,u8 vector,bool soft)145 static inline void kvm_queue_interrupt(struct kvm_vcpu *vcpu, u8 vector,
146 bool soft)
147 {
148 vcpu->arch.interrupt.injected = true;
149 vcpu->arch.interrupt.soft = soft;
150 vcpu->arch.interrupt.nr = vector;
151 }
152
kvm_clear_interrupt_queue(struct kvm_vcpu * vcpu)153 static inline void kvm_clear_interrupt_queue(struct kvm_vcpu *vcpu)
154 {
155 vcpu->arch.interrupt.injected = false;
156 }
157
kvm_event_needs_reinjection(struct kvm_vcpu * vcpu)158 static inline bool kvm_event_needs_reinjection(struct kvm_vcpu *vcpu)
159 {
160 return vcpu->arch.exception.injected || vcpu->arch.interrupt.injected ||
161 vcpu->arch.nmi_injected;
162 }
163
kvm_exception_is_soft(unsigned int nr)164 static inline bool kvm_exception_is_soft(unsigned int nr)
165 {
166 return (nr == BP_VECTOR) || (nr == OF_VECTOR);
167 }
168
is_protmode(struct kvm_vcpu * vcpu)169 static inline bool is_protmode(struct kvm_vcpu *vcpu)
170 {
171 return kvm_is_cr0_bit_set(vcpu, X86_CR0_PE);
172 }
173
is_long_mode(struct kvm_vcpu * vcpu)174 static inline bool is_long_mode(struct kvm_vcpu *vcpu)
175 {
176 #ifdef CONFIG_X86_64
177 return !!(vcpu->arch.efer & EFER_LMA);
178 #else
179 return false;
180 #endif
181 }
182
is_64_bit_mode(struct kvm_vcpu * vcpu)183 static inline bool is_64_bit_mode(struct kvm_vcpu *vcpu)
184 {
185 int cs_db, cs_l;
186
187 WARN_ON_ONCE(vcpu->arch.guest_state_protected);
188
189 if (!is_long_mode(vcpu))
190 return false;
191 kvm_x86_call(get_cs_db_l_bits)(vcpu, &cs_db, &cs_l);
192 return cs_l;
193 }
194
is_64_bit_hypercall(struct kvm_vcpu * vcpu)195 static inline bool is_64_bit_hypercall(struct kvm_vcpu *vcpu)
196 {
197 /*
198 * If running with protected guest state, the CS register is not
199 * accessible. The hypercall register values will have had to been
200 * provided in 64-bit mode, so assume the guest is in 64-bit.
201 */
202 return vcpu->arch.guest_state_protected || is_64_bit_mode(vcpu);
203 }
204
x86_exception_has_error_code(unsigned int vector)205 static inline bool x86_exception_has_error_code(unsigned int vector)
206 {
207 static u32 exception_has_error_code = BIT(DF_VECTOR) | BIT(TS_VECTOR) |
208 BIT(NP_VECTOR) | BIT(SS_VECTOR) | BIT(GP_VECTOR) |
209 BIT(PF_VECTOR) | BIT(AC_VECTOR);
210
211 return (1U << vector) & exception_has_error_code;
212 }
213
mmu_is_nested(struct kvm_vcpu * vcpu)214 static inline bool mmu_is_nested(struct kvm_vcpu *vcpu)
215 {
216 return vcpu->arch.walk_mmu == &vcpu->arch.nested_mmu;
217 }
218
is_pae(struct kvm_vcpu * vcpu)219 static inline bool is_pae(struct kvm_vcpu *vcpu)
220 {
221 return kvm_is_cr4_bit_set(vcpu, X86_CR4_PAE);
222 }
223
is_pse(struct kvm_vcpu * vcpu)224 static inline bool is_pse(struct kvm_vcpu *vcpu)
225 {
226 return kvm_is_cr4_bit_set(vcpu, X86_CR4_PSE);
227 }
228
is_paging(struct kvm_vcpu * vcpu)229 static inline bool is_paging(struct kvm_vcpu *vcpu)
230 {
231 return likely(kvm_is_cr0_bit_set(vcpu, X86_CR0_PG));
232 }
233
is_pae_paging(struct kvm_vcpu * vcpu)234 static inline bool is_pae_paging(struct kvm_vcpu *vcpu)
235 {
236 return !is_long_mode(vcpu) && is_pae(vcpu) && is_paging(vcpu);
237 }
238
vcpu_virt_addr_bits(struct kvm_vcpu * vcpu)239 static inline u8 vcpu_virt_addr_bits(struct kvm_vcpu *vcpu)
240 {
241 return kvm_is_cr4_bit_set(vcpu, X86_CR4_LA57) ? 57 : 48;
242 }
243
max_host_virt_addr_bits(void)244 static inline u8 max_host_virt_addr_bits(void)
245 {
246 return kvm_cpu_cap_has(X86_FEATURE_LA57) ? 57 : 48;
247 }
248
249 /*
250 * x86 MSRs which contain linear addresses, x86 hidden segment bases, and
251 * IDT/GDT bases have static canonicality checks, the size of which depends
252 * only on the CPU's support for 5-level paging, rather than on the state of
253 * CR4.LA57. This applies to both WRMSR and to other instructions that set
254 * their values, e.g. SGDT.
255 *
256 * KVM passes through most of these MSRS and also doesn't intercept the
257 * instructions that set the hidden segment bases.
258 *
259 * Because of this, to be consistent with hardware, even if the guest doesn't
260 * have LA57 enabled in its CPUID, perform canonicality checks based on *host*
261 * support for 5 level paging.
262 *
263 * Finally, instructions which are related to MMU invalidation of a given
264 * linear address, also have a similar static canonical check on address.
265 * This allows for example to invalidate 5-level addresses of a guest from a
266 * host which uses 4-level paging.
267 */
is_noncanonical_address(u64 la,struct kvm_vcpu * vcpu,unsigned int flags)268 static inline bool is_noncanonical_address(u64 la, struct kvm_vcpu *vcpu,
269 unsigned int flags)
270 {
271 if (flags & (X86EMUL_F_INVLPG | X86EMUL_F_MSR | X86EMUL_F_DT_LOAD))
272 return !__is_canonical_address(la, max_host_virt_addr_bits());
273 else
274 return !__is_canonical_address(la, vcpu_virt_addr_bits(vcpu));
275 }
276
is_noncanonical_msr_address(u64 la,struct kvm_vcpu * vcpu)277 static inline bool is_noncanonical_msr_address(u64 la, struct kvm_vcpu *vcpu)
278 {
279 return is_noncanonical_address(la, vcpu, X86EMUL_F_MSR);
280 }
281
is_noncanonical_base_address(u64 la,struct kvm_vcpu * vcpu)282 static inline bool is_noncanonical_base_address(u64 la, struct kvm_vcpu *vcpu)
283 {
284 return is_noncanonical_address(la, vcpu, X86EMUL_F_DT_LOAD);
285 }
286
is_noncanonical_invlpg_address(u64 la,struct kvm_vcpu * vcpu)287 static inline bool is_noncanonical_invlpg_address(u64 la, struct kvm_vcpu *vcpu)
288 {
289 return is_noncanonical_address(la, vcpu, X86EMUL_F_INVLPG);
290 }
291
vcpu_cache_mmio_info(struct kvm_vcpu * vcpu,gva_t gva,gfn_t gfn,unsigned access)292 static inline void vcpu_cache_mmio_info(struct kvm_vcpu *vcpu,
293 gva_t gva, gfn_t gfn, unsigned access)
294 {
295 u64 gen = kvm_memslots(vcpu->kvm)->generation;
296
297 if (unlikely(gen & KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS))
298 return;
299
300 /*
301 * If this is a shadow nested page table, the "GVA" is
302 * actually a nGPA.
303 */
304 vcpu->arch.mmio_gva = mmu_is_nested(vcpu) ? 0 : gva & PAGE_MASK;
305 vcpu->arch.mmio_access = access;
306 vcpu->arch.mmio_gfn = gfn;
307 vcpu->arch.mmio_gen = gen;
308 }
309
vcpu_match_mmio_gen(struct kvm_vcpu * vcpu)310 static inline bool vcpu_match_mmio_gen(struct kvm_vcpu *vcpu)
311 {
312 return vcpu->arch.mmio_gen == kvm_memslots(vcpu->kvm)->generation;
313 }
314
315 /*
316 * Clear the mmio cache info for the given gva. If gva is MMIO_GVA_ANY, we
317 * clear all mmio cache info.
318 */
319 #define MMIO_GVA_ANY (~(gva_t)0)
320
vcpu_clear_mmio_info(struct kvm_vcpu * vcpu,gva_t gva)321 static inline void vcpu_clear_mmio_info(struct kvm_vcpu *vcpu, gva_t gva)
322 {
323 if (gva != MMIO_GVA_ANY && vcpu->arch.mmio_gva != (gva & PAGE_MASK))
324 return;
325
326 vcpu->arch.mmio_gva = 0;
327 }
328
vcpu_match_mmio_gva(struct kvm_vcpu * vcpu,unsigned long gva)329 static inline bool vcpu_match_mmio_gva(struct kvm_vcpu *vcpu, unsigned long gva)
330 {
331 if (vcpu_match_mmio_gen(vcpu) && vcpu->arch.mmio_gva &&
332 vcpu->arch.mmio_gva == (gva & PAGE_MASK))
333 return true;
334
335 return false;
336 }
337
vcpu_match_mmio_gpa(struct kvm_vcpu * vcpu,gpa_t gpa)338 static inline bool vcpu_match_mmio_gpa(struct kvm_vcpu *vcpu, gpa_t gpa)
339 {
340 if (vcpu_match_mmio_gen(vcpu) && vcpu->arch.mmio_gfn &&
341 vcpu->arch.mmio_gfn == gpa >> PAGE_SHIFT)
342 return true;
343
344 return false;
345 }
346
kvm_register_read(struct kvm_vcpu * vcpu,int reg)347 static inline unsigned long kvm_register_read(struct kvm_vcpu *vcpu, int reg)
348 {
349 unsigned long val = kvm_register_read_raw(vcpu, reg);
350
351 return is_64_bit_mode(vcpu) ? val : (u32)val;
352 }
353
kvm_register_write(struct kvm_vcpu * vcpu,int reg,unsigned long val)354 static inline void kvm_register_write(struct kvm_vcpu *vcpu,
355 int reg, unsigned long val)
356 {
357 if (!is_64_bit_mode(vcpu))
358 val = (u32)val;
359 return kvm_register_write_raw(vcpu, reg, val);
360 }
361
kvm_check_has_quirk(struct kvm * kvm,u64 quirk)362 static inline bool kvm_check_has_quirk(struct kvm *kvm, u64 quirk)
363 {
364 return !(kvm->arch.disabled_quirks & quirk);
365 }
366
367 void kvm_inject_realmode_interrupt(struct kvm_vcpu *vcpu, int irq, int inc_eip);
368
369 u64 get_kvmclock_ns(struct kvm *kvm);
370 uint64_t kvm_get_wall_clock_epoch(struct kvm *kvm);
371 bool kvm_get_monotonic_and_clockread(s64 *kernel_ns, u64 *tsc_timestamp);
372 int kvm_guest_time_update(struct kvm_vcpu *v);
373
374 int kvm_read_guest_virt(struct kvm_vcpu *vcpu,
375 gva_t addr, void *val, unsigned int bytes,
376 struct x86_exception *exception);
377
378 int kvm_write_guest_virt_system(struct kvm_vcpu *vcpu,
379 gva_t addr, void *val, unsigned int bytes,
380 struct x86_exception *exception);
381
382 int handle_ud(struct kvm_vcpu *vcpu);
383
384 void kvm_deliver_exception_payload(struct kvm_vcpu *vcpu,
385 struct kvm_queued_exception *ex);
386
387 int kvm_mtrr_set_msr(struct kvm_vcpu *vcpu, u32 msr, u64 data);
388 int kvm_mtrr_get_msr(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
389 bool kvm_vector_hashing_enabled(void);
390 void kvm_fixup_and_inject_pf_error(struct kvm_vcpu *vcpu, gva_t gva, u16 error_code);
391 int x86_decode_emulated_instruction(struct kvm_vcpu *vcpu, int emulation_type,
392 void *insn, int insn_len);
393 int x86_emulate_instruction(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
394 int emulation_type, void *insn, int insn_len);
395 fastpath_t handle_fastpath_set_msr_irqoff(struct kvm_vcpu *vcpu);
396 fastpath_t handle_fastpath_hlt(struct kvm_vcpu *vcpu);
397
398 extern struct kvm_caps kvm_caps;
399 extern struct kvm_host_values kvm_host;
400
401 extern bool enable_pmu;
402
403 /*
404 * Get a filtered version of KVM's supported XCR0 that strips out dynamic
405 * features for which the current process doesn't (yet) have permission to use.
406 * This is intended to be used only when enumerating support to userspace,
407 * e.g. in KVM_GET_SUPPORTED_CPUID and KVM_CAP_XSAVE2, it does NOT need to be
408 * used to check/restrict guest behavior as KVM rejects KVM_SET_CPUID{2} if
409 * userspace attempts to enable unpermitted features.
410 */
kvm_get_filtered_xcr0(void)411 static inline u64 kvm_get_filtered_xcr0(void)
412 {
413 u64 permitted_xcr0 = kvm_caps.supported_xcr0;
414
415 BUILD_BUG_ON(XFEATURE_MASK_USER_DYNAMIC != XFEATURE_MASK_XTILE_DATA);
416
417 if (permitted_xcr0 & XFEATURE_MASK_USER_DYNAMIC) {
418 permitted_xcr0 &= xstate_get_guest_group_perm();
419
420 /*
421 * Treat XTILE_CFG as unsupported if the current process isn't
422 * allowed to use XTILE_DATA, as attempting to set XTILE_CFG in
423 * XCR0 without setting XTILE_DATA is architecturally illegal.
424 */
425 if (!(permitted_xcr0 & XFEATURE_MASK_XTILE_DATA))
426 permitted_xcr0 &= ~XFEATURE_MASK_XTILE_CFG;
427 }
428 return permitted_xcr0;
429 }
430
kvm_mpx_supported(void)431 static inline bool kvm_mpx_supported(void)
432 {
433 return (kvm_caps.supported_xcr0 & (XFEATURE_MASK_BNDREGS | XFEATURE_MASK_BNDCSR))
434 == (XFEATURE_MASK_BNDREGS | XFEATURE_MASK_BNDCSR);
435 }
436
437 extern unsigned int min_timer_period_us;
438
439 extern bool enable_vmware_backdoor;
440
441 extern int pi_inject_timer;
442
443 extern bool report_ignored_msrs;
444
445 extern bool eager_page_split;
446
kvm_pr_unimpl_wrmsr(struct kvm_vcpu * vcpu,u32 msr,u64 data)447 static inline void kvm_pr_unimpl_wrmsr(struct kvm_vcpu *vcpu, u32 msr, u64 data)
448 {
449 if (report_ignored_msrs)
450 vcpu_unimpl(vcpu, "Unhandled WRMSR(0x%x) = 0x%llx\n", msr, data);
451 }
452
kvm_pr_unimpl_rdmsr(struct kvm_vcpu * vcpu,u32 msr)453 static inline void kvm_pr_unimpl_rdmsr(struct kvm_vcpu *vcpu, u32 msr)
454 {
455 if (report_ignored_msrs)
456 vcpu_unimpl(vcpu, "Unhandled RDMSR(0x%x)\n", msr);
457 }
458
nsec_to_cycles(struct kvm_vcpu * vcpu,u64 nsec)459 static inline u64 nsec_to_cycles(struct kvm_vcpu *vcpu, u64 nsec)
460 {
461 return pvclock_scale_delta(nsec, vcpu->arch.virtual_tsc_mult,
462 vcpu->arch.virtual_tsc_shift);
463 }
464
465 /* Same "calling convention" as do_div:
466 * - divide (n << 32) by base
467 * - put result in n
468 * - return remainder
469 */
470 #define do_shl32_div32(n, base) \
471 ({ \
472 u32 __quot, __rem; \
473 asm("divl %2" : "=a" (__quot), "=d" (__rem) \
474 : "rm" (base), "0" (0), "1" ((u32) n)); \
475 n = __quot; \
476 __rem; \
477 })
478
kvm_mwait_in_guest(struct kvm * kvm)479 static inline bool kvm_mwait_in_guest(struct kvm *kvm)
480 {
481 return kvm->arch.mwait_in_guest;
482 }
483
kvm_hlt_in_guest(struct kvm * kvm)484 static inline bool kvm_hlt_in_guest(struct kvm *kvm)
485 {
486 return kvm->arch.hlt_in_guest;
487 }
488
kvm_pause_in_guest(struct kvm * kvm)489 static inline bool kvm_pause_in_guest(struct kvm *kvm)
490 {
491 return kvm->arch.pause_in_guest;
492 }
493
kvm_cstate_in_guest(struct kvm * kvm)494 static inline bool kvm_cstate_in_guest(struct kvm *kvm)
495 {
496 return kvm->arch.cstate_in_guest;
497 }
498
kvm_notify_vmexit_enabled(struct kvm * kvm)499 static inline bool kvm_notify_vmexit_enabled(struct kvm *kvm)
500 {
501 return kvm->arch.notify_vmexit_flags & KVM_X86_NOTIFY_VMEXIT_ENABLED;
502 }
503
kvm_before_interrupt(struct kvm_vcpu * vcpu,enum kvm_intr_type intr)504 static __always_inline void kvm_before_interrupt(struct kvm_vcpu *vcpu,
505 enum kvm_intr_type intr)
506 {
507 WRITE_ONCE(vcpu->arch.handling_intr_from_guest, (u8)intr);
508 }
509
kvm_after_interrupt(struct kvm_vcpu * vcpu)510 static __always_inline void kvm_after_interrupt(struct kvm_vcpu *vcpu)
511 {
512 WRITE_ONCE(vcpu->arch.handling_intr_from_guest, 0);
513 }
514
kvm_handling_nmi_from_guest(struct kvm_vcpu * vcpu)515 static inline bool kvm_handling_nmi_from_guest(struct kvm_vcpu *vcpu)
516 {
517 return vcpu->arch.handling_intr_from_guest == KVM_HANDLING_NMI;
518 }
519
kvm_pat_valid(u64 data)520 static inline bool kvm_pat_valid(u64 data)
521 {
522 if (data & 0xF8F8F8F8F8F8F8F8ull)
523 return false;
524 /* 0, 1, 4, 5, 6, 7 are valid values. */
525 return (data | ((data & 0x0202020202020202ull) << 1)) == data;
526 }
527
kvm_dr7_valid(u64 data)528 static inline bool kvm_dr7_valid(u64 data)
529 {
530 /* Bits [63:32] are reserved */
531 return !(data >> 32);
532 }
kvm_dr6_valid(u64 data)533 static inline bool kvm_dr6_valid(u64 data)
534 {
535 /* Bits [63:32] are reserved */
536 return !(data >> 32);
537 }
538
539 /*
540 * Trigger machine check on the host. We assume all the MSRs are already set up
541 * by the CPU and that we still run on the same CPU as the MCE occurred on.
542 * We pass a fake environment to the machine check handler because we want
543 * the guest to be always treated like user space, no matter what context
544 * it used internally.
545 */
kvm_machine_check(void)546 static inline void kvm_machine_check(void)
547 {
548 #if defined(CONFIG_X86_MCE)
549 struct pt_regs regs = {
550 .cs = 3, /* Fake ring 3 no matter what the guest ran on */
551 .flags = X86_EFLAGS_IF,
552 };
553
554 do_machine_check(®s);
555 #endif
556 }
557
558 void kvm_load_guest_xsave_state(struct kvm_vcpu *vcpu);
559 void kvm_load_host_xsave_state(struct kvm_vcpu *vcpu);
560 int kvm_spec_ctrl_test_value(u64 value);
561 int kvm_handle_memory_failure(struct kvm_vcpu *vcpu, int r,
562 struct x86_exception *e);
563 int kvm_handle_invpcid(struct kvm_vcpu *vcpu, unsigned long type, gva_t gva);
564 bool kvm_msr_allowed(struct kvm_vcpu *vcpu, u32 index, u32 type);
565
566 enum kvm_msr_access {
567 MSR_TYPE_R = BIT(0),
568 MSR_TYPE_W = BIT(1),
569 MSR_TYPE_RW = MSR_TYPE_R | MSR_TYPE_W,
570 };
571
572 /*
573 * Internal error codes that are used to indicate that MSR emulation encountered
574 * an error that should result in #GP in the guest, unless userspace handles it.
575 * Note, '1', '0', and negative numbers are off limits, as they are used by KVM
576 * as part of KVM's lightly documented internal KVM_RUN return codes.
577 *
578 * UNSUPPORTED - The MSR isn't supported, either because it is completely
579 * unknown to KVM, or because the MSR should not exist according
580 * to the vCPU model.
581 *
582 * FILTERED - Access to the MSR is denied by a userspace MSR filter.
583 */
584 #define KVM_MSR_RET_UNSUPPORTED 2
585 #define KVM_MSR_RET_FILTERED 3
586
__kvm_is_valid_cr4(struct kvm_vcpu * vcpu,unsigned long cr4)587 static inline bool __kvm_is_valid_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
588 {
589 return !(cr4 & vcpu->arch.cr4_guest_rsvd_bits);
590 }
591
592 #define __cr4_reserved_bits(__cpu_has, __c) \
593 ({ \
594 u64 __reserved_bits = CR4_RESERVED_BITS; \
595 \
596 if (!__cpu_has(__c, X86_FEATURE_XSAVE)) \
597 __reserved_bits |= X86_CR4_OSXSAVE; \
598 if (!__cpu_has(__c, X86_FEATURE_SMEP)) \
599 __reserved_bits |= X86_CR4_SMEP; \
600 if (!__cpu_has(__c, X86_FEATURE_SMAP)) \
601 __reserved_bits |= X86_CR4_SMAP; \
602 if (!__cpu_has(__c, X86_FEATURE_FSGSBASE)) \
603 __reserved_bits |= X86_CR4_FSGSBASE; \
604 if (!__cpu_has(__c, X86_FEATURE_PKU)) \
605 __reserved_bits |= X86_CR4_PKE; \
606 if (!__cpu_has(__c, X86_FEATURE_LA57)) \
607 __reserved_bits |= X86_CR4_LA57; \
608 if (!__cpu_has(__c, X86_FEATURE_UMIP)) \
609 __reserved_bits |= X86_CR4_UMIP; \
610 if (!__cpu_has(__c, X86_FEATURE_VMX)) \
611 __reserved_bits |= X86_CR4_VMXE; \
612 if (!__cpu_has(__c, X86_FEATURE_PCID)) \
613 __reserved_bits |= X86_CR4_PCIDE; \
614 if (!__cpu_has(__c, X86_FEATURE_LAM)) \
615 __reserved_bits |= X86_CR4_LAM_SUP; \
616 __reserved_bits; \
617 })
618
619 int kvm_sev_es_mmio_write(struct kvm_vcpu *vcpu, gpa_t src, unsigned int bytes,
620 void *dst);
621 int kvm_sev_es_mmio_read(struct kvm_vcpu *vcpu, gpa_t src, unsigned int bytes,
622 void *dst);
623 int kvm_sev_es_string_io(struct kvm_vcpu *vcpu, unsigned int size,
624 unsigned int port, void *data, unsigned int count,
625 int in);
626
user_exit_on_hypercall(struct kvm * kvm,unsigned long hc_nr)627 static inline bool user_exit_on_hypercall(struct kvm *kvm, unsigned long hc_nr)
628 {
629 return kvm->arch.hypercall_exit_enabled & BIT(hc_nr);
630 }
631
632 int ____kvm_emulate_hypercall(struct kvm_vcpu *vcpu, unsigned long nr,
633 unsigned long a0, unsigned long a1,
634 unsigned long a2, unsigned long a3,
635 int op_64_bit, int cpl,
636 int (*complete_hypercall)(struct kvm_vcpu *));
637
638 #define __kvm_emulate_hypercall(_vcpu, nr, a0, a1, a2, a3, op_64_bit, cpl, complete_hypercall) \
639 ({ \
640 int __ret; \
641 \
642 __ret = ____kvm_emulate_hypercall(_vcpu, \
643 kvm_##nr##_read(_vcpu), kvm_##a0##_read(_vcpu), \
644 kvm_##a1##_read(_vcpu), kvm_##a2##_read(_vcpu), \
645 kvm_##a3##_read(_vcpu), op_64_bit, cpl, \
646 complete_hypercall); \
647 \
648 if (__ret > 0) \
649 __ret = complete_hypercall(_vcpu); \
650 __ret; \
651 })
652
653 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
654
655 #endif
656