1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef ARCH_X86_KVM_X86_H
3 #define ARCH_X86_KVM_X86_H
5 #include <linux/kvm_host.h>
6 #include <asm/fpu/xstate.h>
8 #include <asm/pvclock.h>
9 #include "kvm_cache_regs.h"
10 #include "kvm_emulate.h"
14 /* control of guest tsc rate supported? */
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;
31 u64 supported_mce_cap;
34 u64 supported_perf_cap;
37 struct kvm_host_values {
39 * The host's raw MAXPHYADDR, i.e. the number of non-reserved physical
40 * address bits irrespective of features that repurpose legal bits,
48 u64 arch_capabilities;
51 void kvm_spurious_fault(void);
53 #define KVM_NESTED_VMENTER_CONSISTENCY_CHECK(consistency_check) \
55 bool failed = (consistency_check); \
57 trace_kvm_nested_vmenter_failed(#consistency_check, 0); \
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.
66 #define KVM_FIRST_EMULATED_VMX_MSR MSR_IA32_VMX_BASIC
67 #define KVM_LAST_EMULATED_VMX_MSR MSR_IA32_VMX_VMFUNC
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
77 static inline unsigned int __grow_ple_window(unsigned int val,
78 unsigned int base, unsigned int modifier, unsigned int max)
90 return min(ret, (u64)max);
93 static inline unsigned int __shrink_ple_window(unsigned int val,
94 unsigned int base, unsigned int modifier, unsigned int min)
104 return max(val, min);
107 #define MSR_IA32_CR_PAT_DEFAULT \
108 PAT_VALUE(WB, WT, UC_MINUS, UC, WB, WT, UC_MINUS, UC)
110 void kvm_service_local_tlb_flush_requests(struct kvm_vcpu *vcpu);
111 int kvm_check_nested_events(struct kvm_vcpu *vcpu);
113 /* Forcibly leave the nested mode in cases like a vCPU reset */
114 static inline void kvm_leave_nested(struct kvm_vcpu *vcpu)
116 kvm_x86_ops.nested_ops->leave_nested(vcpu);
119 static inline bool kvm_vcpu_has_run(struct kvm_vcpu *vcpu)
121 return vcpu->arch.last_vmentry_cpu != -1;
124 static inline bool kvm_is_exception_pending(struct kvm_vcpu *vcpu)
126 return vcpu->arch.exception.pending ||
127 vcpu->arch.exception_vmexit.pending ||
128 kvm_test_request(KVM_REQ_TRIPLE_FAULT, vcpu);
131 static inline void kvm_clear_exception_queue(struct kvm_vcpu *vcpu)
133 vcpu->arch.exception.pending = false;
134 vcpu->arch.exception.injected = false;
135 vcpu->arch.exception_vmexit.pending = false;
138 static inline void kvm_queue_interrupt(struct kvm_vcpu *vcpu, u8 vector,
141 vcpu->arch.interrupt.injected = true;
142 vcpu->arch.interrupt.soft = soft;
143 vcpu->arch.interrupt.nr = vector;
146 static inline void kvm_clear_interrupt_queue(struct kvm_vcpu *vcpu)
148 vcpu->arch.interrupt.injected = false;
151 static inline bool kvm_event_needs_reinjection(struct kvm_vcpu *vcpu)
153 return vcpu->arch.exception.injected || vcpu->arch.interrupt.injected ||
154 vcpu->arch.nmi_injected;
157 static inline bool kvm_exception_is_soft(unsigned int nr)
159 return (nr == BP_VECTOR) || (nr == OF_VECTOR);
162 static inline bool is_protmode(struct kvm_vcpu *vcpu)
164 return kvm_is_cr0_bit_set(vcpu, X86_CR0_PE);
167 static inline bool is_long_mode(struct kvm_vcpu *vcpu)
170 return !!(vcpu->arch.efer & EFER_LMA);
176 static inline bool is_64_bit_mode(struct kvm_vcpu *vcpu)
180 WARN_ON_ONCE(vcpu->arch.guest_state_protected);
182 if (!is_long_mode(vcpu))
184 kvm_x86_call(get_cs_db_l_bits)(vcpu, &cs_db, &cs_l);
188 static inline bool is_64_bit_hypercall(struct kvm_vcpu *vcpu)
191 * If running with protected guest state, the CS register is not
192 * accessible. The hypercall register values will have had to been
193 * provided in 64-bit mode, so assume the guest is in 64-bit.
195 return vcpu->arch.guest_state_protected || is_64_bit_mode(vcpu);
198 static inline bool x86_exception_has_error_code(unsigned int vector)
200 static u32 exception_has_error_code = BIT(DF_VECTOR) | BIT(TS_VECTOR) |
201 BIT(NP_VECTOR) | BIT(SS_VECTOR) | BIT(GP_VECTOR) |
202 BIT(PF_VECTOR) | BIT(AC_VECTOR);
204 return (1U << vector) & exception_has_error_code;
207 static inline bool mmu_is_nested(struct kvm_vcpu *vcpu)
209 return vcpu->arch.walk_mmu == &vcpu->arch.nested_mmu;
212 static inline bool is_pae(struct kvm_vcpu *vcpu)
214 return kvm_is_cr4_bit_set(vcpu, X86_CR4_PAE);
217 static inline bool is_pse(struct kvm_vcpu *vcpu)
219 return kvm_is_cr4_bit_set(vcpu, X86_CR4_PSE);
222 static inline bool is_paging(struct kvm_vcpu *vcpu)
224 return likely(kvm_is_cr0_bit_set(vcpu, X86_CR0_PG));
227 static inline bool is_pae_paging(struct kvm_vcpu *vcpu)
229 return !is_long_mode(vcpu) && is_pae(vcpu) && is_paging(vcpu);
232 static inline u8 vcpu_virt_addr_bits(struct kvm_vcpu *vcpu)
234 return kvm_is_cr4_bit_set(vcpu, X86_CR4_LA57) ? 57 : 48;
237 static inline u8 max_host_virt_addr_bits(void)
239 return kvm_cpu_cap_has(X86_FEATURE_LA57) ? 57 : 48;
243 * x86 MSRs which contain linear addresses, x86 hidden segment bases, and
244 * IDT/GDT bases have static canonicality checks, the size of which depends
245 * only on the CPU's support for 5-level paging, rather than on the state of
246 * CR4.LA57. This applies to both WRMSR and to other instructions that set
247 * their values, e.g. SGDT.
249 * KVM passes through most of these MSRS and also doesn't intercept the
250 * instructions that set the hidden segment bases.
252 * Because of this, to be consistent with hardware, even if the guest doesn't
253 * have LA57 enabled in its CPUID, perform canonicality checks based on *host*
254 * support for 5 level paging.
256 * Finally, instructions which are related to MMU invalidation of a given
257 * linear address, also have a similar static canonical check on address.
258 * This allows for example to invalidate 5-level addresses of a guest from a
259 * host which uses 4-level paging.
261 static inline bool is_noncanonical_address(u64 la, struct kvm_vcpu *vcpu,
264 if (flags & (X86EMUL_F_INVLPG | X86EMUL_F_MSR | X86EMUL_F_DT_LOAD))
265 return !__is_canonical_address(la, max_host_virt_addr_bits());
267 return !__is_canonical_address(la, vcpu_virt_addr_bits(vcpu));
270 static inline bool is_noncanonical_msr_address(u64 la, struct kvm_vcpu *vcpu)
272 return is_noncanonical_address(la, vcpu, X86EMUL_F_MSR);
275 static inline bool is_noncanonical_base_address(u64 la, struct kvm_vcpu *vcpu)
277 return is_noncanonical_address(la, vcpu, X86EMUL_F_DT_LOAD);
280 static inline bool is_noncanonical_invlpg_address(u64 la, struct kvm_vcpu *vcpu)
282 return is_noncanonical_address(la, vcpu, X86EMUL_F_INVLPG);
285 static inline void vcpu_cache_mmio_info(struct kvm_vcpu *vcpu,
286 gva_t gva, gfn_t gfn, unsigned access)
288 u64 gen = kvm_memslots(vcpu->kvm)->generation;
290 if (unlikely(gen & KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS))
294 * If this is a shadow nested page table, the "GVA" is
297 vcpu->arch.mmio_gva = mmu_is_nested(vcpu) ? 0 : gva & PAGE_MASK;
298 vcpu->arch.mmio_access = access;
299 vcpu->arch.mmio_gfn = gfn;
300 vcpu->arch.mmio_gen = gen;
303 static inline bool vcpu_match_mmio_gen(struct kvm_vcpu *vcpu)
305 return vcpu->arch.mmio_gen == kvm_memslots(vcpu->kvm)->generation;
309 * Clear the mmio cache info for the given gva. If gva is MMIO_GVA_ANY, we
310 * clear all mmio cache info.
312 #define MMIO_GVA_ANY (~(gva_t)0)
314 static inline void vcpu_clear_mmio_info(struct kvm_vcpu *vcpu, gva_t gva)
316 if (gva != MMIO_GVA_ANY && vcpu->arch.mmio_gva != (gva & PAGE_MASK))
319 vcpu->arch.mmio_gva = 0;
322 static inline bool vcpu_match_mmio_gva(struct kvm_vcpu *vcpu, unsigned long gva)
324 if (vcpu_match_mmio_gen(vcpu) && vcpu->arch.mmio_gva &&
325 vcpu->arch.mmio_gva == (gva & PAGE_MASK))
331 static inline bool vcpu_match_mmio_gpa(struct kvm_vcpu *vcpu, gpa_t gpa)
333 if (vcpu_match_mmio_gen(vcpu) && vcpu->arch.mmio_gfn &&
334 vcpu->arch.mmio_gfn == gpa >> PAGE_SHIFT)
340 static inline unsigned long kvm_register_read(struct kvm_vcpu *vcpu, int reg)
342 unsigned long val = kvm_register_read_raw(vcpu, reg);
344 return is_64_bit_mode(vcpu) ? val : (u32)val;
347 static inline void kvm_register_write(struct kvm_vcpu *vcpu,
348 int reg, unsigned long val)
350 if (!is_64_bit_mode(vcpu))
352 return kvm_register_write_raw(vcpu, reg, val);
355 static inline bool kvm_check_has_quirk(struct kvm *kvm, u64 quirk)
357 return !(kvm->arch.disabled_quirks & quirk);
360 void kvm_inject_realmode_interrupt(struct kvm_vcpu *vcpu, int irq, int inc_eip);
362 u64 get_kvmclock_ns(struct kvm *kvm);
363 uint64_t kvm_get_wall_clock_epoch(struct kvm *kvm);
364 bool kvm_get_monotonic_and_clockread(s64 *kernel_ns, u64 *tsc_timestamp);
366 int kvm_read_guest_virt(struct kvm_vcpu *vcpu,
367 gva_t addr, void *val, unsigned int bytes,
368 struct x86_exception *exception);
370 int kvm_write_guest_virt_system(struct kvm_vcpu *vcpu,
371 gva_t addr, void *val, unsigned int bytes,
372 struct x86_exception *exception);
374 int handle_ud(struct kvm_vcpu *vcpu);
376 void kvm_deliver_exception_payload(struct kvm_vcpu *vcpu,
377 struct kvm_queued_exception *ex);
379 int kvm_mtrr_set_msr(struct kvm_vcpu *vcpu, u32 msr, u64 data);
380 int kvm_mtrr_get_msr(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
381 bool kvm_vector_hashing_enabled(void);
382 void kvm_fixup_and_inject_pf_error(struct kvm_vcpu *vcpu, gva_t gva, u16 error_code);
383 int x86_decode_emulated_instruction(struct kvm_vcpu *vcpu, int emulation_type,
384 void *insn, int insn_len);
385 int x86_emulate_instruction(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
386 int emulation_type, void *insn, int insn_len);
387 fastpath_t handle_fastpath_set_msr_irqoff(struct kvm_vcpu *vcpu);
388 fastpath_t handle_fastpath_hlt(struct kvm_vcpu *vcpu);
390 extern struct kvm_caps kvm_caps;
391 extern struct kvm_host_values kvm_host;
393 extern bool enable_pmu;
396 * Get a filtered version of KVM's supported XCR0 that strips out dynamic
397 * features for which the current process doesn't (yet) have permission to use.
398 * This is intended to be used only when enumerating support to userspace,
399 * e.g. in KVM_GET_SUPPORTED_CPUID and KVM_CAP_XSAVE2, it does NOT need to be
400 * used to check/restrict guest behavior as KVM rejects KVM_SET_CPUID{2} if
401 * userspace attempts to enable unpermitted features.
403 static inline u64 kvm_get_filtered_xcr0(void)
405 u64 permitted_xcr0 = kvm_caps.supported_xcr0;
407 BUILD_BUG_ON(XFEATURE_MASK_USER_DYNAMIC != XFEATURE_MASK_XTILE_DATA);
409 if (permitted_xcr0 & XFEATURE_MASK_USER_DYNAMIC) {
410 permitted_xcr0 &= xstate_get_guest_group_perm();
413 * Treat XTILE_CFG as unsupported if the current process isn't
414 * allowed to use XTILE_DATA, as attempting to set XTILE_CFG in
415 * XCR0 without setting XTILE_DATA is architecturally illegal.
417 if (!(permitted_xcr0 & XFEATURE_MASK_XTILE_DATA))
418 permitted_xcr0 &= ~XFEATURE_MASK_XTILE_CFG;
420 return permitted_xcr0;
423 static inline bool kvm_mpx_supported(void)
425 return (kvm_caps.supported_xcr0 & (XFEATURE_MASK_BNDREGS | XFEATURE_MASK_BNDCSR))
426 == (XFEATURE_MASK_BNDREGS | XFEATURE_MASK_BNDCSR);
429 extern unsigned int min_timer_period_us;
431 extern bool enable_vmware_backdoor;
433 extern int pi_inject_timer;
435 extern bool report_ignored_msrs;
437 extern bool eager_page_split;
439 static inline void kvm_pr_unimpl_wrmsr(struct kvm_vcpu *vcpu, u32 msr, u64 data)
441 if (report_ignored_msrs)
442 vcpu_unimpl(vcpu, "Unhandled WRMSR(0x%x) = 0x%llx\n", msr, data);
445 static inline void kvm_pr_unimpl_rdmsr(struct kvm_vcpu *vcpu, u32 msr)
447 if (report_ignored_msrs)
448 vcpu_unimpl(vcpu, "Unhandled RDMSR(0x%x)\n", msr);
451 static inline u64 nsec_to_cycles(struct kvm_vcpu *vcpu, u64 nsec)
453 return pvclock_scale_delta(nsec, vcpu->arch.virtual_tsc_mult,
454 vcpu->arch.virtual_tsc_shift);
457 /* Same "calling convention" as do_div:
458 * - divide (n << 32) by base
462 #define do_shl32_div32(n, base) \
465 asm("divl %2" : "=a" (__quot), "=d" (__rem) \
466 : "rm" (base), "0" (0), "1" ((u32) n)); \
471 static inline bool kvm_mwait_in_guest(struct kvm *kvm)
473 return kvm->arch.mwait_in_guest;
476 static inline bool kvm_hlt_in_guest(struct kvm *kvm)
478 return kvm->arch.hlt_in_guest;
481 static inline bool kvm_pause_in_guest(struct kvm *kvm)
483 return kvm->arch.pause_in_guest;
486 static inline bool kvm_cstate_in_guest(struct kvm *kvm)
488 return kvm->arch.cstate_in_guest;
491 static inline bool kvm_notify_vmexit_enabled(struct kvm *kvm)
493 return kvm->arch.notify_vmexit_flags & KVM_X86_NOTIFY_VMEXIT_ENABLED;
496 static __always_inline void kvm_before_interrupt(struct kvm_vcpu *vcpu,
497 enum kvm_intr_type intr)
499 WRITE_ONCE(vcpu->arch.handling_intr_from_guest, (u8)intr);
502 static __always_inline void kvm_after_interrupt(struct kvm_vcpu *vcpu)
504 WRITE_ONCE(vcpu->arch.handling_intr_from_guest, 0);
507 static inline bool kvm_handling_nmi_from_guest(struct kvm_vcpu *vcpu)
509 return vcpu->arch.handling_intr_from_guest == KVM_HANDLING_NMI;
512 static inline bool kvm_pat_valid(u64 data)
514 if (data & 0xF8F8F8F8F8F8F8F8ull)
516 /* 0, 1, 4, 5, 6, 7 are valid values. */
517 return (data | ((data & 0x0202020202020202ull) << 1)) == data;
520 static inline bool kvm_dr7_valid(u64 data)
522 /* Bits [63:32] are reserved */
523 return !(data >> 32);
525 static inline bool kvm_dr6_valid(u64 data)
527 /* Bits [63:32] are reserved */
528 return !(data >> 32);
532 * Trigger machine check on the host. We assume all the MSRs are already set up
533 * by the CPU and that we still run on the same CPU as the MCE occurred on.
534 * We pass a fake environment to the machine check handler because we want
535 * the guest to be always treated like user space, no matter what context
536 * it used internally.
538 static inline void kvm_machine_check(void)
540 #if defined(CONFIG_X86_MCE)
541 struct pt_regs regs = {
542 .cs = 3, /* Fake ring 3 no matter what the guest ran on */
543 .flags = X86_EFLAGS_IF,
546 do_machine_check(®s);
550 void kvm_load_guest_xsave_state(struct kvm_vcpu *vcpu);
551 void kvm_load_host_xsave_state(struct kvm_vcpu *vcpu);
552 int kvm_spec_ctrl_test_value(u64 value);
553 bool __kvm_is_valid_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
554 int kvm_handle_memory_failure(struct kvm_vcpu *vcpu, int r,
555 struct x86_exception *e);
556 int kvm_handle_invpcid(struct kvm_vcpu *vcpu, unsigned long type, gva_t gva);
557 bool kvm_msr_allowed(struct kvm_vcpu *vcpu, u32 index, u32 type);
559 enum kvm_msr_access {
562 MSR_TYPE_RW = MSR_TYPE_R | MSR_TYPE_W,
566 * Internal error codes that are used to indicate that MSR emulation encountered
567 * an error that should result in #GP in the guest, unless userspace handles it.
568 * Note, '1', '0', and negative numbers are off limits, as they are used by KVM
569 * as part of KVM's lightly documented internal KVM_RUN return codes.
571 * UNSUPPORTED - The MSR isn't supported, either because it is completely
572 * unknown to KVM, or because the MSR should not exist according
575 * FILTERED - Access to the MSR is denied by a userspace MSR filter.
577 #define KVM_MSR_RET_UNSUPPORTED 2
578 #define KVM_MSR_RET_FILTERED 3
580 #define __cr4_reserved_bits(__cpu_has, __c) \
582 u64 __reserved_bits = CR4_RESERVED_BITS; \
584 if (!__cpu_has(__c, X86_FEATURE_XSAVE)) \
585 __reserved_bits |= X86_CR4_OSXSAVE; \
586 if (!__cpu_has(__c, X86_FEATURE_SMEP)) \
587 __reserved_bits |= X86_CR4_SMEP; \
588 if (!__cpu_has(__c, X86_FEATURE_SMAP)) \
589 __reserved_bits |= X86_CR4_SMAP; \
590 if (!__cpu_has(__c, X86_FEATURE_FSGSBASE)) \
591 __reserved_bits |= X86_CR4_FSGSBASE; \
592 if (!__cpu_has(__c, X86_FEATURE_PKU)) \
593 __reserved_bits |= X86_CR4_PKE; \
594 if (!__cpu_has(__c, X86_FEATURE_LA57)) \
595 __reserved_bits |= X86_CR4_LA57; \
596 if (!__cpu_has(__c, X86_FEATURE_UMIP)) \
597 __reserved_bits |= X86_CR4_UMIP; \
598 if (!__cpu_has(__c, X86_FEATURE_VMX)) \
599 __reserved_bits |= X86_CR4_VMXE; \
600 if (!__cpu_has(__c, X86_FEATURE_PCID)) \
601 __reserved_bits |= X86_CR4_PCIDE; \
602 if (!__cpu_has(__c, X86_FEATURE_LAM)) \
603 __reserved_bits |= X86_CR4_LAM_SUP; \
607 int kvm_sev_es_mmio_write(struct kvm_vcpu *vcpu, gpa_t src, unsigned int bytes,
609 int kvm_sev_es_mmio_read(struct kvm_vcpu *vcpu, gpa_t src, unsigned int bytes,
611 int kvm_sev_es_string_io(struct kvm_vcpu *vcpu, unsigned int size,
612 unsigned int port, void *data, unsigned int count,