Lines Matching +full:un +full:- +full:masked
1 // SPDX-License-Identifier: GPL-2.0-only
3 * Kernel-based Virtual Machine driver for Linux
16 * Ben-Ami Yassour <benami@il.ibm.com>
48 #include <linux/user-return-notifier.h>
62 #include <linux/entry-kvm.h>
108 ((struct kvm_vcpu *)(ctxt)->vcpu)
111 * - enable syscall per default because its emulated by KVM
112 * - enable LME and LMA per default on 64 bit KVM
143 *(((struct kvm_x86_ops *)0)->func));
146 #include <asm/kvm-x86-ops.h>
163 /* tsc tolerance in parts per million - default to 1/2 of the NTP threshold */
175 * Flags to manipulate forced emulation behavior (any non-zero value will
182 int __read_mostly pi_inject_timer = -1;
448 * List of MSRs that control the existence of MSR-based features, i.e. MSRs
461 (KVM_LAST_EMULATED_VMX_MSR - KVM_FIRST_EMULATED_VMX_MSR + 1)];
553 size - useroffset, NULL); in kvm_alloc_emulator_cache()
562 vcpu->arch.apf.gfns[i] = ~0; in kvm_async_pf_hash_reset()
578 if (msrs->registered) { in kvm_on_user_return()
579 msrs->registered = false; in kvm_on_user_return()
584 values = &msrs->values[slot]; in kvm_on_user_return()
585 if (values->host != values->curr) { in kvm_on_user_return()
586 wrmsrq(kvm_uret_msrs_list[slot], values->host); in kvm_on_user_return()
587 values->curr = values->host; in kvm_on_user_return()
612 return -1; in kvm_add_user_return_msr()
627 return -1; in kvm_find_user_return_msr()
639 msrs->values[i].host = value; in kvm_user_return_msr_cpu_online()
640 msrs->values[i].curr = value; in kvm_user_return_msr_cpu_online()
646 if (!msrs->registered) { in kvm_user_return_register_notifier()
647 msrs->urn.on_user_return = kvm_on_user_return; in kvm_user_return_register_notifier()
648 user_return_notifier_register(&msrs->urn); in kvm_user_return_register_notifier()
649 msrs->registered = true; in kvm_user_return_register_notifier()
658 value = (value & mask) | (msrs->values[slot].host & ~mask); in kvm_set_user_return_msr()
659 if (value == msrs->values[slot].curr) in kvm_set_user_return_msr()
665 msrs->values[slot].curr = value; in kvm_set_user_return_msr()
675 msrs->values[slot].curr = value; in kvm_user_return_msr_update_cache()
684 if (msrs->registered) in drop_user_return_notifiers()
685 kvm_on_user_return(&msrs->urn); in drop_user_return_notifiers()
739 * #DBs can be trap-like or fault-like, the caller must check other CPU in exception_type()
758 if (!ex->has_payload) in kvm_deliver_exception_payload()
761 switch (ex->vector) { in kvm_deliver_exception_payload()
764 * "Certain debug exceptions may clear bit 0-3. The in kvm_deliver_exception_payload()
768 vcpu->arch.dr6 &= ~DR_TRAP_BITS; in kvm_deliver_exception_payload()
777 * Active low bits should be cleared if 1-setting in payload. in kvm_deliver_exception_payload()
778 * Active high bits should be set if 1-setting in payload. in kvm_deliver_exception_payload()
785 vcpu->arch.dr6 |= DR6_ACTIVE_LOW; in kvm_deliver_exception_payload()
786 vcpu->arch.dr6 |= ex->payload; in kvm_deliver_exception_payload()
787 vcpu->arch.dr6 ^= ex->payload & DR6_ACTIVE_LOW; in kvm_deliver_exception_payload()
795 vcpu->arch.dr6 &= ~BIT(12); in kvm_deliver_exception_payload()
798 vcpu->arch.cr2 = ex->payload; in kvm_deliver_exception_payload()
802 ex->has_payload = false; in kvm_deliver_exception_payload()
803 ex->payload = 0; in kvm_deliver_exception_payload()
811 struct kvm_queued_exception *ex = &vcpu->arch.exception_vmexit; in kvm_queue_exception_vmexit()
813 ex->vector = vector; in kvm_queue_exception_vmexit()
814 ex->injected = false; in kvm_queue_exception_vmexit()
815 ex->pending = true; in kvm_queue_exception_vmexit()
816 ex->has_error_code = has_error_code; in kvm_queue_exception_vmexit()
817 ex->error_code = error_code; in kvm_queue_exception_vmexit()
818 ex->has_payload = has_payload; in kvm_queue_exception_vmexit()
819 ex->payload = payload; in kvm_queue_exception_vmexit()
832 * If the exception is destined for L2, morph it to a VM-Exit if L1 in kvm_multiple_exception()
836 kvm_x86_ops.nested_ops->is_exception_vmexit(vcpu, nr, error_code)) { in kvm_multiple_exception()
842 if (!vcpu->arch.exception.pending && !vcpu->arch.exception.injected) { in kvm_multiple_exception()
844 vcpu->arch.exception.pending = true; in kvm_multiple_exception()
845 vcpu->arch.exception.injected = false; in kvm_multiple_exception()
847 vcpu->arch.exception.has_error_code = has_error; in kvm_multiple_exception()
848 vcpu->arch.exception.vector = nr; in kvm_multiple_exception()
849 vcpu->arch.exception.error_code = error_code; in kvm_multiple_exception()
850 vcpu->arch.exception.has_payload = has_payload; in kvm_multiple_exception()
851 vcpu->arch.exception.payload = payload; in kvm_multiple_exception()
854 &vcpu->arch.exception); in kvm_multiple_exception()
859 prev_nr = vcpu->arch.exception.vector; in kvm_multiple_exception()
861 /* triple fault -> shutdown */ in kvm_multiple_exception()
873 vcpu->arch.exception.injected = false; in kvm_multiple_exception()
874 vcpu->arch.exception.pending = false; in kvm_multiple_exception()
879 that instruction re-execution will regenerate lost in kvm_multiple_exception()
910 * On VM-Entry, an exception can be pending if and only if event in kvm_requeue_exception()
920 * re-checking is incorrect if _L1_ injected the exception, in which in kvm_requeue_exception()
925 vcpu->arch.exception.injected = true; in kvm_requeue_exception()
926 vcpu->arch.exception.has_error_code = has_error_code; in kvm_requeue_exception()
927 vcpu->arch.exception.vector = nr; in kvm_requeue_exception()
928 vcpu->arch.exception.error_code = error_code; in kvm_requeue_exception()
929 vcpu->arch.exception.has_payload = false; in kvm_requeue_exception()
930 vcpu->arch.exception.payload = 0; in kvm_requeue_exception()
958 ++vcpu->stat.pf_guest; in kvm_inject_page_fault()
961 * Async #PF in L2 is always forwarded to L1 as a VM-Exit regardless of in kvm_inject_page_fault()
964 if (is_guest_mode(vcpu) && fault->async_page_fault) in kvm_inject_page_fault()
966 true, fault->error_code, in kvm_inject_page_fault()
967 true, fault->address); in kvm_inject_page_fault()
969 kvm_queue_exception_e_p(vcpu, PF_VECTOR, fault->error_code, in kvm_inject_page_fault()
970 fault->address); in kvm_inject_page_fault()
977 WARN_ON_ONCE(fault->vector != PF_VECTOR); in kvm_inject_emulated_page_fault()
979 fault_mmu = fault->nested_page_fault ? vcpu->arch.mmu : in kvm_inject_emulated_page_fault()
980 vcpu->arch.walk_mmu; in kvm_inject_emulated_page_fault()
986 if ((fault->error_code & PFERR_PRESENT_MASK) && in kvm_inject_emulated_page_fault()
987 !(fault->error_code & PFERR_RSVD_MASK)) in kvm_inject_emulated_page_fault()
988 kvm_mmu_invalidate_addr(vcpu, fault_mmu, fault->address, in kvm_inject_emulated_page_fault()
991 fault_mmu->inject_page_fault(vcpu, fault); in kvm_inject_emulated_page_fault()
997 atomic_inc(&vcpu->arch.nmi_queued); in kvm_inject_nmi()
1031 return vcpu->arch.reserved_gpa_bits | rsvd_bits(5, 8) | rsvd_bits(1, 2); in pdptr_rsvd_bits()
1039 struct kvm_mmu *mmu = vcpu->arch.walk_mmu; in load_pdptrs()
1044 u64 pdpte[ARRAY_SIZE(mmu->pdptrs)]; in load_pdptrs()
1072 if (!tdp_enabled && memcmp(mmu->pdptrs, pdpte, sizeof(mmu->pdptrs))) in load_pdptrs()
1073 kvm_mmu_free_roots(vcpu->kvm, mmu, KVM_MMU_ROOT_CURRENT); in load_pdptrs()
1075 memcpy(mmu->pdptrs, pdpte, sizeof(mmu->pdptrs)); in load_pdptrs()
1078 vcpu->arch.pdptrs_from_userspace = false; in load_pdptrs()
1103 * CR0.WP is incorporated into the MMU role, but only for non-nested, in kvm_post_set_cr0()
1150 if ((vcpu->arch.efer & EFER_LME) && !is_paging(vcpu) && in kvm_set_cr0()
1161 if (!(vcpu->arch.efer & EFER_LME) && (cr0 & X86_CR0_PG) && in kvm_set_cr0()
1186 if (vcpu->arch.guest_state_protected) in kvm_load_guest_xsave_state()
1191 if (vcpu->arch.xcr0 != kvm_host.xcr0) in kvm_load_guest_xsave_state()
1192 xsetbv(XCR_XFEATURE_ENABLED_MASK, vcpu->arch.xcr0); in kvm_load_guest_xsave_state()
1195 vcpu->arch.ia32_xss != kvm_host.xss) in kvm_load_guest_xsave_state()
1196 wrmsrq(MSR_IA32_XSS, vcpu->arch.ia32_xss); in kvm_load_guest_xsave_state()
1200 vcpu->arch.pkru != vcpu->arch.host_pkru && in kvm_load_guest_xsave_state()
1201 ((vcpu->arch.xcr0 & XFEATURE_MASK_PKRU) || in kvm_load_guest_xsave_state()
1203 wrpkru(vcpu->arch.pkru); in kvm_load_guest_xsave_state()
1209 if (vcpu->arch.guest_state_protected) in kvm_load_host_xsave_state()
1213 ((vcpu->arch.xcr0 & XFEATURE_MASK_PKRU) || in kvm_load_host_xsave_state()
1215 vcpu->arch.pkru = rdpkru(); in kvm_load_host_xsave_state()
1216 if (vcpu->arch.pkru != vcpu->arch.host_pkru) in kvm_load_host_xsave_state()
1217 wrpkru(vcpu->arch.host_pkru); in kvm_load_host_xsave_state()
1222 if (vcpu->arch.xcr0 != kvm_host.xcr0) in kvm_load_host_xsave_state()
1226 vcpu->arch.ia32_xss != kvm_host.xss) in kvm_load_host_xsave_state()
1236 return vcpu->arch.guest_supported_xcr0 & XFEATURE_MASK_USER_DYNAMIC; in kvm_guest_supported_xfd()
1243 u64 old_xcr0 = vcpu->arch.xcr0; in __kvm_set_xcr()
1259 valid_bits = vcpu->arch.guest_supported_xcr0 | XFEATURE_MASK_FP; in __kvm_set_xcr()
1278 vcpu->arch.xcr0 = xcr0; in __kvm_set_xcr()
1281 vcpu->arch.cpuid_dynamic_bits_dirty = true; in __kvm_set_xcr()
1310 * If CR4.PCIDE is changed 0 -> 1, there is no need to flush the TLB in kvm_post_set_cr4()
1324 * - CR4.PCIDE is changed from 1 to 0 in kvm_post_set_cr4()
1325 * - CR4.PGE is toggled in kvm_post_set_cr4()
1336 * - CR4.SMEP is changed from 0 to 1 in kvm_post_set_cr4()
1337 * - CR4.PAE is toggled in kvm_post_set_cr4()
1379 struct kvm_mmu *mmu = vcpu->arch.mmu; in kvm_invalidate_pcid()
1414 if (kvm_get_pcid(vcpu, mmu->prev_roots[i].pgd) == pcid) in kvm_invalidate_pcid()
1417 kvm_mmu_free_roots(vcpu->kvm, mmu, roots_to_free); in kvm_invalidate_pcid()
1450 vcpu->arch.cr3 = cr3; in kvm_set_cr3()
1459 * and it's impossible to use a non-zero PCID when PCID is disabled, in kvm_set_cr3()
1476 vcpu->arch.cr8 = cr8; in kvm_set_cr8()
1486 return vcpu->arch.cr8; in kvm_get_cr8()
1494 if (!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP)) { in kvm_update_dr0123()
1496 vcpu->arch.eff_db[i] = vcpu->arch.db[i]; in kvm_update_dr0123()
1504 if (vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP) in kvm_update_dr7()
1505 dr7 = vcpu->arch.guest_debug_dr7; in kvm_update_dr7()
1507 dr7 = vcpu->arch.dr7; in kvm_update_dr7()
1509 vcpu->arch.switch_db_regs &= ~KVM_DEBUGREG_BP_ENABLED; in kvm_update_dr7()
1511 vcpu->arch.switch_db_regs |= KVM_DEBUGREG_BP_ENABLED; in kvm_update_dr7()
1529 size_t size = ARRAY_SIZE(vcpu->arch.db); in kvm_set_dr()
1533 vcpu->arch.db[array_index_nospec(dr, size)] = val; in kvm_set_dr()
1534 if (!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP)) in kvm_set_dr()
1535 vcpu->arch.eff_db[dr] = val; in kvm_set_dr()
1541 vcpu->arch.dr6 = (val & DR6_VOLATILE) | kvm_dr6_fixed(vcpu); in kvm_set_dr()
1547 vcpu->arch.dr7 = (val & DR7_VOLATILE) | DR7_FIXED_1; in kvm_set_dr()
1558 size_t size = ARRAY_SIZE(vcpu->arch.db); in kvm_get_dr()
1562 return vcpu->arch.db[array_index_nospec(dr, size)]; in kvm_get_dr()
1565 return vcpu->arch.dr6; in kvm_get_dr()
1568 return vcpu->arch.dr7; in kvm_get_dr()
1592 * 10 - MISC_PACKAGE_CTRLS
1593 * 11 - ENERGY_FILTERING_CTL
1594 * 12 - DOITM
1595 * 18 - FB_CLEAR_CTRL
1596 * 21 - XAPIC_DISABLE_STATUS
1597 * 23 - OVERCLOCKING_STATUS
1730 u64 old_efer = vcpu->arch.efer; in set_efer()
1731 u64 efer = msr_info->data; in set_efer()
1737 if (!msr_info->host_initiated) { in set_efer()
1742 (vcpu->arch.efer & EFER_LME) != (efer & EFER_LME)) in set_efer()
1747 efer |= vcpu->arch.efer & EFER_LMA; in set_efer()
1775 struct kvm *kvm = vcpu->kvm; in kvm_msr_allowed()
1784 idx = srcu_read_lock(&kvm->srcu); in kvm_msr_allowed()
1786 msr_filter = srcu_dereference(kvm->arch.msr_filter, &kvm->srcu); in kvm_msr_allowed()
1792 allowed = msr_filter->default_allow; in kvm_msr_allowed()
1793 ranges = msr_filter->ranges; in kvm_msr_allowed()
1795 for (i = 0; i < msr_filter->count; i++) { in kvm_msr_allowed()
1802 allowed = test_bit(index - start, bitmap); in kvm_msr_allowed()
1808 srcu_read_unlock(&kvm->srcu, idx); in kvm_msr_allowed()
1817 * Returns 0 on success, non-0 otherwise.
1838 * non-canonical address is written on Intel but not on in __kvm_set_msr()
1839 * AMD (which ignores the top 32-bits, because it does in __kvm_set_msr()
1840 * not implement 64-bit SYSENTER). in __kvm_set_msr()
1842 * 64-bit code should hence be able to write a non-canonical in __kvm_set_msr()
1844 * vmentry does not fail on Intel after writing a non-canonical in __kvm_set_msr()
1846 * invokes 64-bit SYSENTER. in __kvm_set_msr()
1865 * clear the bits. This ensures cross-vendor migration will in __kvm_set_msr()
1898 * Returns 0 on success, non-0 otherwise.
1965 if (!vcpu->run->msr.error) { in complete_userspace_rdmsr()
1966 kvm_rax_write(vcpu, (u32)vcpu->run->msr.data); in complete_userspace_rdmsr()
1967 kvm_rdx_write(vcpu, vcpu->run->msr.data >> 32); in complete_userspace_rdmsr()
1973 return complete_emulated_insn_gp(vcpu, vcpu->run->msr.error); in complete_emulated_msr_access()
1984 return kvm_x86_call(complete_emulated_msr)(vcpu, vcpu->run->msr.error); in complete_fast_msr_access()
2013 if (!(vcpu->kvm->arch.user_space_msr_mask & msr_reason)) in kvm_msr_user_space()
2016 vcpu->run->exit_reason = exit_reason; in kvm_msr_user_space()
2017 vcpu->run->msr.error = 0; in kvm_msr_user_space()
2018 memset(vcpu->run->msr.pad, 0, sizeof(vcpu->run->msr.pad)); in kvm_msr_user_space()
2019 vcpu->run->msr.reason = msr_reason; in kvm_msr_user_space()
2020 vcpu->run->msr.index = index; in kvm_msr_user_space()
2021 vcpu->run->msr.data = data; in kvm_msr_user_space()
2022 vcpu->arch.complete_userspace_io = completion; in kvm_msr_user_space()
2038 kvm_rax_write(vcpu, data & -1u); in kvm_emulate_rdmsr()
2039 kvm_rdx_write(vcpu, (data >> 32) & -1u); in kvm_emulate_rdmsr()
2101 if (kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_MWAIT_NEVER_UD_FAULTS)) in kvm_emulate_monitor_mwait()
2104 if (kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_MISC_ENABLE_NO_MWAIT)) in kvm_emulate_monitor_mwait()
2107 enabled = vcpu->arch.ia32_misc_enable_msr & MSR_IA32_MISC_ENABLE_MWAIT; in kvm_emulate_monitor_mwait()
2132 return READ_ONCE(vcpu->mode) == EXITING_GUEST_MODE || in kvm_vcpu_exit_request()
2138 * i.e. the sending of IPI, sending IPI early in the VM-Exit flow reduces
2145 if (!lapic_in_kernel(vcpu) || !apic_x2apic_mode(vcpu->arch.apic)) in handle_fastpath_set_x2apic_icr_irqoff()
2152 return kvm_x2apic_icr_write(vcpu->arch.apic, data); in handle_fastpath_set_x2apic_icr_irqoff()
2226 return -EINVAL; in do_set_msr()
2258 write_seqcount_begin(&vdata->seq); in update_pvclock_gtod()
2261 vdata->clock.vclock_mode = tk->tkr_mono.clock->vdso_clock_mode; in update_pvclock_gtod()
2262 vdata->clock.cycle_last = tk->tkr_mono.cycle_last; in update_pvclock_gtod()
2263 vdata->clock.mask = tk->tkr_mono.mask; in update_pvclock_gtod()
2264 vdata->clock.mult = tk->tkr_mono.mult; in update_pvclock_gtod()
2265 vdata->clock.shift = tk->tkr_mono.shift; in update_pvclock_gtod()
2266 vdata->clock.base_cycles = tk->tkr_mono.xtime_nsec; in update_pvclock_gtod()
2267 vdata->clock.offset = tk->tkr_mono.base; in update_pvclock_gtod()
2269 vdata->raw_clock.vclock_mode = tk->tkr_raw.clock->vdso_clock_mode; in update_pvclock_gtod()
2270 vdata->raw_clock.cycle_last = tk->tkr_raw.cycle_last; in update_pvclock_gtod()
2271 vdata->raw_clock.mask = tk->tkr_raw.mask; in update_pvclock_gtod()
2272 vdata->raw_clock.mult = tk->tkr_raw.mult; in update_pvclock_gtod()
2273 vdata->raw_clock.shift = tk->tkr_raw.shift; in update_pvclock_gtod()
2274 vdata->raw_clock.base_cycles = tk->tkr_raw.xtime_nsec; in update_pvclock_gtod()
2275 vdata->raw_clock.offset = tk->tkr_raw.base; in update_pvclock_gtod()
2277 vdata->wall_time_sec = tk->xtime_sec; in update_pvclock_gtod()
2279 vdata->offs_boot = tk->offs_boot; in update_pvclock_gtod()
2281 write_seqcount_end(&vdata->seq); in update_pvclock_gtod()
2341 struct kvm_arch *ka = &vcpu->kvm->arch; in kvm_write_system_time()
2343 if (vcpu->vcpu_id == 0 && !host_initiated) { in kvm_write_system_time()
2344 if (ka->boot_vcpu_runs_old_kvmclock != old_msr) in kvm_write_system_time()
2347 ka->boot_vcpu_runs_old_kvmclock = old_msr; in kvm_write_system_time()
2350 vcpu->arch.time = system_time; in kvm_write_system_time()
2355 kvm_gpc_activate(&vcpu->arch.pv_time, system_time & ~1ULL, in kvm_write_system_time()
2358 kvm_gpc_deactivate(&vcpu->arch.pv_time); in kvm_write_system_time()
2381 shift--; in kvm_get_time_scale()
2426 vcpu->arch.tsc_catchup = 1; in set_tsc_khz()
2427 vcpu->arch.tsc_always_catchup = 1; in set_tsc_khz()
2431 return -1; in set_tsc_khz()
2435 /* TSC scaling required - calculate ratio */ in set_tsc_khz()
2440 pr_warn_ratelimited("Invalid TSC scaling ratio - virtual-tsc-khz=%u\n", in set_tsc_khz()
2442 return -1; in set_tsc_khz()
2458 return -1; in kvm_set_tsc_khz()
2463 &vcpu->arch.virtual_tsc_shift, in kvm_set_tsc_khz()
2464 &vcpu->arch.virtual_tsc_mult); in kvm_set_tsc_khz()
2465 vcpu->arch.virtual_tsc_khz = user_tsc_khz; in kvm_set_tsc_khz()
2473 thresh_lo = adjust_tsc_khz(tsc_khz, -tsc_tolerance_ppm); in kvm_set_tsc_khz()
2485 u64 tsc = pvclock_scale_delta(kernel_ns-vcpu->arch.this_tsc_nsec, in compute_guest_tsc()
2486 vcpu->arch.virtual_tsc_mult, in compute_guest_tsc()
2487 vcpu->arch.virtual_tsc_shift); in compute_guest_tsc()
2488 tsc += vcpu->arch.this_tsc_write; in compute_guest_tsc()
2502 struct kvm_arch *ka = &vcpu->kvm->arch; in kvm_track_tsc_matching()
2510 bool use_master_clock = (ka->nr_vcpus_matched_tsc + 1 == in kvm_track_tsc_matching()
2511 atomic_read(&vcpu->kvm->online_vcpus)) && in kvm_track_tsc_matching()
2512 gtod_is_based_on_tsc(gtod->clock.vclock_mode); in kvm_track_tsc_matching()
2520 if ((ka->use_master_clock && new_generation) || in kvm_track_tsc_matching()
2521 (ka->use_master_clock != use_master_clock)) in kvm_track_tsc_matching()
2524 trace_kvm_track_tsc(vcpu->vcpu_id, ka->nr_vcpus_matched_tsc, in kvm_track_tsc_matching()
2525 atomic_read(&vcpu->kvm->online_vcpus), in kvm_track_tsc_matching()
2526 ka->use_master_clock, gtod->clock.vclock_mode); in kvm_track_tsc_matching()
2533 * The most significant 64-N bits (mult) of ratio represent the
2536 * point number (mult + frac * 2^(-N)).
2559 tsc = kvm_scale_tsc(rdtsc(), vcpu->arch.l1_tsc_scaling_ratio); in kvm_compute_l1_tsc_offset()
2561 return target_tsc - tsc; in kvm_compute_l1_tsc_offset()
2566 return vcpu->arch.l1_tsc_offset + in kvm_read_l1_tsc()
2567 kvm_scale_tsc(host_tsc, vcpu->arch.l1_tsc_scaling_ratio); in kvm_read_l1_tsc()
2598 if (vcpu->arch.guest_tsc_protected) in kvm_vcpu_write_tsc_offset()
2601 trace_kvm_write_tsc_offset(vcpu->vcpu_id, in kvm_vcpu_write_tsc_offset()
2602 vcpu->arch.l1_tsc_offset, in kvm_vcpu_write_tsc_offset()
2605 vcpu->arch.l1_tsc_offset = l1_offset; in kvm_vcpu_write_tsc_offset()
2613 vcpu->arch.tsc_offset = kvm_calc_nested_tsc_offset( in kvm_vcpu_write_tsc_offset()
2618 vcpu->arch.tsc_offset = l1_offset; in kvm_vcpu_write_tsc_offset()
2625 vcpu->arch.l1_tsc_scaling_ratio = l1_multiplier; in kvm_vcpu_write_tsc_multiplier()
2629 vcpu->arch.tsc_scaling_ratio = kvm_calc_nested_tsc_multiplier( in kvm_vcpu_write_tsc_multiplier()
2633 vcpu->arch.tsc_scaling_ratio = l1_multiplier; in kvm_vcpu_write_tsc_multiplier()
2643 * TSC is marked unstable when we're running on Hyper-V, in kvm_check_tsc_unstable()
2660 struct kvm *kvm = vcpu->kvm; in __kvm_synchronize_tsc()
2662 lockdep_assert_held(&kvm->arch.tsc_write_lock); in __kvm_synchronize_tsc()
2664 if (vcpu->arch.guest_tsc_protected) in __kvm_synchronize_tsc()
2668 vcpu->kvm->arch.user_set_tsc = true; in __kvm_synchronize_tsc()
2674 kvm->arch.last_tsc_nsec = ns; in __kvm_synchronize_tsc()
2675 kvm->arch.last_tsc_write = tsc; in __kvm_synchronize_tsc()
2676 kvm->arch.last_tsc_khz = vcpu->arch.virtual_tsc_khz; in __kvm_synchronize_tsc()
2677 kvm->arch.last_tsc_offset = offset; in __kvm_synchronize_tsc()
2679 vcpu->arch.last_guest_tsc = tsc; in __kvm_synchronize_tsc()
2691 * These values are tracked in kvm->arch.cur_xxx variables. in __kvm_synchronize_tsc()
2693 kvm->arch.cur_tsc_generation++; in __kvm_synchronize_tsc()
2694 kvm->arch.cur_tsc_nsec = ns; in __kvm_synchronize_tsc()
2695 kvm->arch.cur_tsc_write = tsc; in __kvm_synchronize_tsc()
2696 kvm->arch.cur_tsc_offset = offset; in __kvm_synchronize_tsc()
2697 kvm->arch.nr_vcpus_matched_tsc = 0; in __kvm_synchronize_tsc()
2698 } else if (vcpu->arch.this_tsc_generation != kvm->arch.cur_tsc_generation) { in __kvm_synchronize_tsc()
2699 kvm->arch.nr_vcpus_matched_tsc++; in __kvm_synchronize_tsc()
2703 vcpu->arch.this_tsc_generation = kvm->arch.cur_tsc_generation; in __kvm_synchronize_tsc()
2704 vcpu->arch.this_tsc_nsec = kvm->arch.cur_tsc_nsec; in __kvm_synchronize_tsc()
2705 vcpu->arch.this_tsc_write = kvm->arch.cur_tsc_write; in __kvm_synchronize_tsc()
2713 struct kvm *kvm = vcpu->kvm; in kvm_synchronize_tsc()
2719 raw_spin_lock_irqsave(&kvm->arch.tsc_write_lock, flags); in kvm_synchronize_tsc()
2722 elapsed = ns - kvm->arch.last_tsc_nsec; in kvm_synchronize_tsc()
2724 if (vcpu->arch.virtual_tsc_khz) { in kvm_synchronize_tsc()
2731 } else if (kvm->arch.user_set_tsc) { in kvm_synchronize_tsc()
2732 u64 tsc_exp = kvm->arch.last_tsc_write + in kvm_synchronize_tsc()
2734 u64 tsc_hz = vcpu->arch.virtual_tsc_khz * 1000LL; in kvm_synchronize_tsc()
2736 * Here lies UAPI baggage: when a user-initiated TSC write has in kvm_synchronize_tsc()
2747 * come from the kernel's default vCPU creation. Make the 1-second in kvm_synchronize_tsc()
2763 vcpu->arch.virtual_tsc_khz == kvm->arch.last_tsc_khz) { in kvm_synchronize_tsc()
2765 offset = kvm->arch.cur_tsc_offset; in kvm_synchronize_tsc()
2775 raw_spin_unlock_irqrestore(&kvm->arch.tsc_write_lock, flags); in kvm_synchronize_tsc()
2781 u64 tsc_offset = vcpu->arch.l1_tsc_offset; in adjust_tsc_offset_guest()
2787 if (vcpu->arch.l1_tsc_scaling_ratio != kvm_caps.default_tsc_scaling_ratio) in adjust_tsc_offset_host()
2790 vcpu->arch.l1_tsc_scaling_ratio); in adjust_tsc_offset_host()
2822 switch (clock->vclock_mode) { in vgettsc()
2828 v = (tsc_pg_val - clock->cycle_last) & in vgettsc()
2829 clock->mask; in vgettsc()
2838 v = (*tsc_timestamp - clock->cycle_last) & in vgettsc()
2839 clock->mask; in vgettsc()
2848 return v * clock->mult; in vgettsc()
2853 * frequency of CLOCK_MONOTONIC_RAW (hence adding gtos->offs_boot).
2863 seq = read_seqcount_begin(>od->seq); in do_kvmclock_base()
2864 ns = gtod->raw_clock.base_cycles; in do_kvmclock_base()
2865 ns += vgettsc(>od->raw_clock, tsc_timestamp, &mode); in do_kvmclock_base()
2866 ns >>= gtod->raw_clock.shift; in do_kvmclock_base()
2867 ns += ktime_to_ns(ktime_add(gtod->raw_clock.offset, gtod->offs_boot)); in do_kvmclock_base()
2868 } while (unlikely(read_seqcount_retry(>od->seq, seq))); in do_kvmclock_base()
2886 seq = read_seqcount_begin(>od->seq); in do_monotonic()
2887 ns = gtod->clock.base_cycles; in do_monotonic()
2888 ns += vgettsc(>od->clock, tsc_timestamp, &mode); in do_monotonic()
2889 ns >>= gtod->clock.shift; in do_monotonic()
2890 ns += ktime_to_ns(gtod->clock.offset); in do_monotonic()
2891 } while (unlikely(read_seqcount_retry(>od->seq, seq))); in do_monotonic()
2905 seq = read_seqcount_begin(>od->seq); in do_realtime()
2906 ts->tv_sec = gtod->wall_time_sec; in do_realtime()
2907 ns = gtod->clock.base_cycles; in do_realtime()
2908 ns += vgettsc(>od->clock, tsc_timestamp, &mode); in do_realtime()
2909 ns >>= gtod->clock.shift; in do_realtime()
2910 } while (unlikely(read_seqcount_retry(>od->seq, seq))); in do_realtime()
2912 ts->tv_sec += __iter_div_u64_rem(ns, NSEC_PER_SEC, &ns); in do_realtime()
2913 ts->tv_nsec = ns; in do_realtime()
2981 * 4. ret0 = timespec0 + (rdtsc - tsc0) |
2982 * 5. | ret1 = timespec1 + (rdtsc - tsc1)
2983 * | ret1 = timespec0 + N + (rdtsc - (tsc0 + M))
2987 * - ret0 < ret1
2988 * - timespec0 + (rdtsc - tsc0) < timespec0 + N + (rdtsc - (tsc0 + M))
2990 * - 0 < N - M => M < N
3009 struct kvm_arch *ka = &kvm->arch; in pvclock_update_vm_gtod_copy()
3013 lockdep_assert_held(&kvm->arch.tsc_write_lock); in pvclock_update_vm_gtod_copy()
3014 vcpus_matched = (ka->nr_vcpus_matched_tsc + 1 == in pvclock_update_vm_gtod_copy()
3015 atomic_read(&kvm->online_vcpus)); in pvclock_update_vm_gtod_copy()
3022 &ka->master_kernel_ns, in pvclock_update_vm_gtod_copy()
3023 &ka->master_cycle_now); in pvclock_update_vm_gtod_copy()
3025 ka->use_master_clock = host_tsc_clocksource && vcpus_matched in pvclock_update_vm_gtod_copy()
3026 && !ka->backwards_tsc_observed in pvclock_update_vm_gtod_copy()
3027 && !ka->boot_vcpu_runs_old_kvmclock; in pvclock_update_vm_gtod_copy()
3029 if (ka->use_master_clock) in pvclock_update_vm_gtod_copy()
3033 trace_kvm_update_master_clock(ka->use_master_clock, vclock_mode, in pvclock_update_vm_gtod_copy()
3045 raw_spin_lock_irq(&kvm->arch.tsc_write_lock); in __kvm_start_pvclock_update()
3046 write_seqcount_begin(&kvm->arch.pvclock_sc); in __kvm_start_pvclock_update()
3059 struct kvm_arch *ka = &kvm->arch; in kvm_end_pvclock_update()
3063 write_seqcount_end(&ka->pvclock_sc); in kvm_end_pvclock_update()
3064 raw_spin_unlock_irq(&ka->tsc_write_lock); in kvm_end_pvclock_update()
3083 * per-CPU value (which may be zero if a CPU is going offline). Note, tsc_khz
3097 /* Called within read_seqcount_begin/retry for kvm->pvclock_sc. */
3100 struct kvm_arch *ka = &kvm->arch; in __get_kvmclock()
3106 data->flags = 0; in __get_kvmclock()
3107 if (ka->use_master_clock && in __get_kvmclock()
3112 if (kvm_get_walltime_and_clockread(&ts, &data->host_tsc)) { in __get_kvmclock()
3113 data->realtime = ts.tv_nsec + NSEC_PER_SEC * ts.tv_sec; in __get_kvmclock()
3114 data->flags |= KVM_CLOCK_REALTIME | KVM_CLOCK_HOST_TSC; in __get_kvmclock()
3117 data->host_tsc = rdtsc(); in __get_kvmclock()
3119 data->flags |= KVM_CLOCK_TSC_STABLE; in __get_kvmclock()
3120 hv_clock.tsc_timestamp = ka->master_cycle_now; in __get_kvmclock()
3121 hv_clock.system_time = ka->master_kernel_ns + ka->kvmclock_offset; in __get_kvmclock()
3125 data->clock = __pvclock_read_cycles(&hv_clock, data->host_tsc); in __get_kvmclock()
3127 data->clock = get_kvmclock_base_ns() + ka->kvmclock_offset; in __get_kvmclock()
3135 struct kvm_arch *ka = &kvm->arch; in get_kvmclock()
3139 seq = read_seqcount_begin(&ka->pvclock_sc); in get_kvmclock()
3141 } while (read_seqcount_retry(&ka->pvclock_sc, seq)); in get_kvmclock()
3163 read_lock_irqsave(&gpc->lock, flags); in kvm_setup_guest_pvclock()
3165 read_unlock_irqrestore(&gpc->lock, flags); in kvm_setup_guest_pvclock()
3170 read_lock_irqsave(&gpc->lock, flags); in kvm_setup_guest_pvclock()
3173 guest_hv_clock = (void *)(gpc->khva + offset); in kvm_setup_guest_pvclock()
3182 guest_hv_clock->version = hv_clock.version = (guest_hv_clock->version + 1) | 1; in kvm_setup_guest_pvclock()
3186 hv_clock.flags |= (guest_hv_clock->flags & PVCLOCK_GUEST_STOPPED); in kvm_setup_guest_pvclock()
3192 guest_hv_clock->version = ++hv_clock.version; in kvm_setup_guest_pvclock()
3195 read_unlock_irqrestore(&gpc->lock, flags); in kvm_setup_guest_pvclock()
3197 trace_kvm_pvclock_update(vcpu->vcpu_id, &hv_clock); in kvm_setup_guest_pvclock()
3205 struct kvm_vcpu_arch *vcpu = &v->arch; in kvm_guest_time_update()
3206 struct kvm_arch *ka = &v->kvm->arch; in kvm_guest_time_update()
3219 seq = read_seqcount_begin(&ka->pvclock_sc); in kvm_guest_time_update()
3220 use_master_clock = ka->use_master_clock; in kvm_guest_time_update()
3222 host_tsc = ka->master_cycle_now; in kvm_guest_time_update()
3223 kernel_ns = ka->master_kernel_ns; in kvm_guest_time_update()
3225 } while (read_seqcount_retry(&ka->pvclock_sc, seq)); in kvm_guest_time_update()
3252 if (vcpu->tsc_catchup) { in kvm_guest_time_update()
3255 adjust_tsc_offset_guest(v, tsc - tsc_timestamp); in kvm_guest_time_update()
3266 v->arch.l1_tsc_scaling_ratio); in kvm_guest_time_update()
3270 if (unlikely(vcpu->hw_tsc_khz != tgt_tsc_khz)) { in kvm_guest_time_update()
3272 &vcpu->pvclock_tsc_shift, in kvm_guest_time_update()
3273 &vcpu->pvclock_tsc_mul); in kvm_guest_time_update()
3274 vcpu->hw_tsc_khz = tgt_tsc_khz; in kvm_guest_time_update()
3277 hv_clock.tsc_shift = vcpu->pvclock_tsc_shift; in kvm_guest_time_update()
3278 hv_clock.tsc_to_system_mul = vcpu->pvclock_tsc_mul; in kvm_guest_time_update()
3280 hv_clock.system_time = kernel_ns + v->kvm->arch.kvmclock_offset; in kvm_guest_time_update()
3281 vcpu->last_guest_tsc = tsc_timestamp; in kvm_guest_time_update()
3288 if (vcpu->pv_time.active) { in kvm_guest_time_update()
3294 if (vcpu->pvclock_set_guest_stopped_request) { in kvm_guest_time_update()
3296 vcpu->pvclock_set_guest_stopped_request = false; in kvm_guest_time_update()
3298 kvm_setup_guest_pvclock(&hv_clock, v, &vcpu->pv_time, 0); in kvm_guest_time_update()
3303 kvm_hv_setup_tsc_page(v->kvm, &hv_clock); in kvm_guest_time_update()
3314 if (ka->xen.hvm_config.flags & KVM_XEN_HVM_CONFIG_PVCLOCK_TSC_UNSTABLE) in kvm_guest_time_update()
3317 if (vcpu->xen.vcpu_info_cache.active) in kvm_guest_time_update()
3318 kvm_setup_guest_pvclock(&hv_clock, v, &vcpu->xen.vcpu_info_cache, in kvm_guest_time_update()
3320 if (vcpu->xen.vcpu_time_info_cache.active) in kvm_guest_time_update()
3321 kvm_setup_guest_pvclock(&hv_clock, v, &vcpu->xen.vcpu_time_info_cache, 0); in kvm_guest_time_update()
3349 struct kvm_arch *ka = &kvm->arch; in kvm_get_wall_clock_epoch()
3355 seq = read_seqcount_begin(&ka->pvclock_sc); in kvm_get_wall_clock_epoch()
3358 if (!ka->use_master_clock) in kvm_get_wall_clock_epoch()
3380 hv_clock.tsc_timestamp = ka->master_cycle_now; in kvm_get_wall_clock_epoch()
3381 hv_clock.system_time = ka->master_kernel_ns + ka->kvmclock_offset; in kvm_get_wall_clock_epoch()
3383 } while (read_seqcount_retry(&ka->pvclock_sc, seq)); in kvm_get_wall_clock_epoch()
3389 * since 1970-01-01. in kvm_get_wall_clock_epoch()
3395 return ts.tv_nsec + NSEC_PER_SEC * ts.tv_sec - in kvm_get_wall_clock_epoch()
3399 return ktime_get_real_ns() - get_kvmclock_ns(kvm); in kvm_get_wall_clock_epoch()
3404 * vcpu->cpu migration, should not allow system_timestamp from
3410 * We need to rate-limit these requests though, as they can
3413 * by the delay we use to rate-limit the updates.
3435 struct kvm *kvm = v->kvm; in kvm_gen_kvmclock_update()
3438 schedule_delayed_work(&kvm->arch.kvmclock_update_work, in kvm_gen_kvmclock_update()
3451 schedule_delayed_work(&kvm->arch.kvmclock_update_work, 0); in kvmclock_sync_fn()
3452 schedule_delayed_work(&kvm->arch.kvmclock_sync_work, in kvmclock_sync_fn()
3473 return !!(vcpu->arch.msr_hwcr & BIT_ULL(18)); in can_set_mci_status()
3480 u64 mcg_cap = vcpu->arch.mcg_cap; in set_msr_mce()
3482 u32 msr = msr_info->index; in set_msr_mce()
3483 u64 data = msr_info->data; in set_msr_mce()
3488 vcpu->arch.mcg_status = data; in set_msr_mce()
3492 (data || !msr_info->host_initiated)) in set_msr_mce()
3496 vcpu->arch.mcg_ctl = data; in set_msr_mce()
3498 case MSR_IA32_MC0_CTL2 ... MSR_IA32_MCx_CTL2(KVM_MAX_MCE_BANKS) - 1: in set_msr_mce()
3499 last_msr = MSR_IA32_MCx_CTL2(bank_num) - 1; in set_msr_mce()
3503 if (!(mcg_cap & MCG_CMCI_P) && (data || !msr_info->host_initiated)) in set_msr_mce()
3508 offset = array_index_nospec(msr - MSR_IA32_MC0_CTL2, in set_msr_mce()
3509 last_msr + 1 - MSR_IA32_MC0_CTL2); in set_msr_mce()
3510 vcpu->arch.mci_ctl2_banks[offset] = data; in set_msr_mce()
3512 case MSR_IA32_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1: in set_msr_mce()
3513 last_msr = MSR_IA32_MCx_CTL(bank_num) - 1; in set_msr_mce()
3525 * single-bit ECC data errors. in set_msr_mce()
3533 * AMD-based CPUs allow non-zero values, but if and only if in set_msr_mce()
3536 if (!msr_info->host_initiated && is_mci_status_msr(msr) && in set_msr_mce()
3540 offset = array_index_nospec(msr - MSR_IA32_MC0_CTL, in set_msr_mce()
3541 last_msr + 1 - MSR_IA32_MC0_CTL); in set_msr_mce()
3542 vcpu->arch.mce_banks[offset] = data; in set_msr_mce()
3554 return (vcpu->arch.apf.msr_en_val & mask) == mask; in kvm_pv_async_pf_enabled()
3576 vcpu->arch.apf.msr_en_val = data; in kvm_pv_enable_async_pf()
3584 if (kvm_gfn_to_hva_cache_init(vcpu->kvm, &vcpu->arch.apf.data, gpa, in kvm_pv_enable_async_pf()
3588 vcpu->arch.apf.send_always = (data & KVM_ASYNC_PF_SEND_ALWAYS); in kvm_pv_enable_async_pf()
3589 vcpu->arch.apf.delivery_as_pf_vmexit = data & KVM_ASYNC_PF_DELIVERY_AS_PF_VMEXIT; in kvm_pv_enable_async_pf()
3598 /* Bits 8-63 are reserved */ in kvm_pv_enable_async_pf_int()
3605 vcpu->arch.apf.msr_int_val = data; in kvm_pv_enable_async_pf_int()
3607 vcpu->arch.apf.vec = data & KVM_ASYNC_PF_VEC_MASK; in kvm_pv_enable_async_pf_int()
3614 kvm_gpc_deactivate(&vcpu->arch.pv_time); in kvmclock_reset()
3615 vcpu->arch.time = 0; in kvmclock_reset()
3620 ++vcpu->stat.tlb_flush; in kvm_vcpu_flush_tlb_all()
3629 ++vcpu->stat.tlb_flush; in kvm_vcpu_flush_tlb_guest()
3645 * Flushing all "guest" TLB is always a superset of Hyper-V's fine in kvm_vcpu_flush_tlb_guest()
3654 ++vcpu->stat.tlb_flush; in kvm_vcpu_flush_tlb_current()
3662 * prior before nested VM-Enter/VM-Exit.
3676 struct gfn_to_hva_cache *ghc = &vcpu->arch.st.cache; in record_steal_time()
3679 gpa_t gpa = vcpu->arch.st.msr_val & KVM_STEAL_VALID_BITS; in record_steal_time()
3683 if (kvm_xen_msr_enabled(vcpu->kvm)) { in record_steal_time()
3688 if (!(vcpu->arch.st.msr_val & KVM_MSR_ENABLED)) in record_steal_time()
3691 if (WARN_ON_ONCE(current->mm != vcpu->kvm->mm)) in record_steal_time()
3694 slots = kvm_memslots(vcpu->kvm); in record_steal_time()
3696 if (unlikely(slots->generation != ghc->generation || in record_steal_time()
3697 gpa != ghc->gpa || in record_steal_time()
3698 kvm_is_error_hva(ghc->hva) || !ghc->memslot)) { in record_steal_time()
3700 BUILD_BUG_ON((sizeof(*st) - 1) & KVM_STEAL_VALID_BITS); in record_steal_time()
3702 if (kvm_gfn_to_hva_cache_init(vcpu->kvm, ghc, gpa, sizeof(*st)) || in record_steal_time()
3703 kvm_is_error_hva(ghc->hva) || !ghc->memslot) in record_steal_time()
3707 st = (struct kvm_steal_time __user *)ghc->hva; in record_steal_time()
3714 int err = -EFAULT; in record_steal_time()
3725 "+m" (st->preempted)); in record_steal_time()
3731 vcpu->arch.st.preempted = 0; in record_steal_time()
3733 trace_kvm_pv_tlb_flush(vcpu->vcpu_id, in record_steal_time()
3744 unsafe_put_user(0, &st->preempted, out); in record_steal_time()
3745 vcpu->arch.st.preempted = 0; in record_steal_time()
3748 unsafe_get_user(version, &st->version, out); in record_steal_time()
3753 unsafe_put_user(version, &st->version, out); in record_steal_time()
3757 unsafe_get_user(steal, &st->steal, out); in record_steal_time()
3758 steal += current->sched_info.run_delay - in record_steal_time()
3759 vcpu->arch.st.last_steal; in record_steal_time()
3760 vcpu->arch.st.last_steal = current->sched_info.run_delay; in record_steal_time()
3761 unsafe_put_user(steal, &st->steal, out); in record_steal_time()
3764 unsafe_put_user(version, &st->version, out); in record_steal_time()
3769 mark_page_dirty_in_slot(vcpu->kvm, ghc->memslot, gpa_to_gfn(ghc->gpa)); in record_steal_time()
3774 u32 msr = msr_info->index; in kvm_set_msr_common()
3775 u64 data = msr_info->data; in kvm_set_msr_common()
3778 * Do not allow host-initiated writes to trigger the Xen hypercall in kvm_set_msr_common()
3782 if (kvm_xen_is_hypercall_page_msr(vcpu->kvm, msr) && in kvm_set_msr_common()
3783 !msr_info->host_initiated) in kvm_set_msr_common()
3798 if (msr_info->host_initiated) in kvm_set_msr_common()
3799 vcpu->arch.microcode_version = data; in kvm_set_msr_common()
3802 if (!msr_info->host_initiated || in kvm_set_msr_common()
3805 vcpu->arch.arch_capabilities = data; in kvm_set_msr_common()
3808 if (!msr_info->host_initiated || in kvm_set_msr_common()
3820 if (vcpu->arch.perf_capabilities == data) in kvm_set_msr_common()
3823 vcpu->arch.perf_capabilities = data; in kvm_set_msr_common()
3829 if (!msr_info->host_initiated) { in kvm_set_msr_common()
3857 if (!msr_info->host_initiated && in kvm_set_msr_common()
3884 vcpu->arch.msr_hwcr = data; in kvm_set_msr_common()
3896 vcpu->arch.pat = data; in kvm_set_msr_common()
3902 return kvm_apic_set_base(vcpu, data, msr_info->host_initiated); in kvm_set_msr_common()
3910 if (!msr_info->host_initiated) { in kvm_set_msr_common()
3911 s64 adj = data - vcpu->arch.ia32_tsc_adjust_msr; in kvm_set_msr_common()
3918 vcpu->arch.ia32_tsc_adjust_msr = data; in kvm_set_msr_common()
3922 u64 old_val = vcpu->arch.ia32_misc_enable_msr; in kvm_set_msr_common()
3924 if (!msr_info->host_initiated) { in kvm_set_msr_common()
3934 if (!kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_MISC_ENABLE_NO_MWAIT) && in kvm_set_msr_common()
3938 vcpu->arch.ia32_misc_enable_msr = data; in kvm_set_msr_common()
3939 vcpu->arch.cpuid_dynamic_bits_dirty = true; in kvm_set_msr_common()
3941 vcpu->arch.ia32_misc_enable_msr = data; in kvm_set_msr_common()
3946 if (!IS_ENABLED(CONFIG_KVM_SMM) || !msr_info->host_initiated) in kvm_set_msr_common()
3948 vcpu->arch.smbase = data; in kvm_set_msr_common()
3951 vcpu->arch.msr_ia32_power_ctl = data; in kvm_set_msr_common()
3954 if (msr_info->host_initiated) { in kvm_set_msr_common()
3956 } else if (!vcpu->arch.guest_tsc_protected) { in kvm_set_msr_common()
3957 u64 adj = kvm_compute_l1_tsc_offset(vcpu, data) - vcpu->arch.l1_tsc_offset; in kvm_set_msr_common()
3959 vcpu->arch.ia32_tsc_adjust_msr += adj; in kvm_set_msr_common()
3963 if (!msr_info->host_initiated && in kvm_set_msr_common()
3973 vcpu->arch.ia32_xss = data; in kvm_set_msr_common()
3974 vcpu->arch.cpuid_dynamic_bits_dirty = true; in kvm_set_msr_common()
3977 if (!msr_info->host_initiated) in kvm_set_msr_common()
3979 vcpu->arch.smi_count = data; in kvm_set_msr_common()
3985 vcpu->kvm->arch.wall_clock = data; in kvm_set_msr_common()
3986 kvm_write_wall_clock(vcpu->kvm, data, 0); in kvm_set_msr_common()
3992 vcpu->kvm->arch.wall_clock = data; in kvm_set_msr_common()
3993 kvm_write_wall_clock(vcpu->kvm, data, 0); in kvm_set_msr_common()
3999 kvm_write_system_time(vcpu, data, false, msr_info->host_initiated); in kvm_set_msr_common()
4005 kvm_write_system_time(vcpu, data, true, msr_info->host_initiated); in kvm_set_msr_common()
4025 vcpu->arch.apf.pageready_pending = false; in kvm_set_msr_common()
4039 vcpu->arch.st.msr_val = data; in kvm_set_msr_common()
4060 if (data & (-1ULL << 1)) in kvm_set_msr_common()
4063 vcpu->arch.msr_kvm_poll_control = data; in kvm_set_msr_common()
4068 case MSR_IA32_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1: in kvm_set_msr_common()
4069 case MSR_IA32_MC0_CTL2 ... MSR_IA32_MCx_CTL2(KVM_MAX_MCE_BANKS) - 1: in kvm_set_msr_common()
4086 * all pre-dating SVM, but a recommended workaround from in kvm_set_msr_common()
4104 msr_info->host_initiated); in kvm_set_msr_common()
4107 /* Drop writes to this legacy MSR -- see rdmsr in kvm_set_msr_common()
4115 vcpu->arch.osvw.length = data; in kvm_set_msr_common()
4120 vcpu->arch.osvw.status = data; in kvm_set_msr_common()
4123 if (!msr_info->host_initiated) in kvm_set_msr_common()
4125 vcpu->arch.msr_platform_info = data; in kvm_set_msr_common()
4132 vcpu->arch.msr_misc_features_enables = data; in kvm_set_msr_common()
4136 if (!msr_info->host_initiated && in kvm_set_msr_common()
4143 fpu_update_guest_xfd(&vcpu->arch.guest_fpu, data); in kvm_set_msr_common()
4146 if (!msr_info->host_initiated && in kvm_set_msr_common()
4153 vcpu->arch.guest_fpu.xfd_err = data; in kvm_set_msr_common()
4169 u64 mcg_cap = vcpu->arch.mcg_cap; in get_msr_mce()
4179 data = vcpu->arch.mcg_cap; in get_msr_mce()
4184 data = vcpu->arch.mcg_ctl; in get_msr_mce()
4187 data = vcpu->arch.mcg_status; in get_msr_mce()
4189 case MSR_IA32_MC0_CTL2 ... MSR_IA32_MCx_CTL2(KVM_MAX_MCE_BANKS) - 1: in get_msr_mce()
4190 last_msr = MSR_IA32_MCx_CTL2(bank_num) - 1; in get_msr_mce()
4196 offset = array_index_nospec(msr - MSR_IA32_MC0_CTL2, in get_msr_mce()
4197 last_msr + 1 - MSR_IA32_MC0_CTL2); in get_msr_mce()
4198 data = vcpu->arch.mci_ctl2_banks[offset]; in get_msr_mce()
4200 case MSR_IA32_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1: in get_msr_mce()
4201 last_msr = MSR_IA32_MCx_CTL(bank_num) - 1; in get_msr_mce()
4205 offset = array_index_nospec(msr - MSR_IA32_MC0_CTL, in get_msr_mce()
4206 last_msr + 1 - MSR_IA32_MC0_CTL); in get_msr_mce()
4207 data = vcpu->arch.mce_banks[offset]; in get_msr_mce()
4218 switch (msr_info->index) { in kvm_get_msr_common()
4241 * so for existing CPU-specific MSRs. in kvm_get_msr_common()
4248 msr_info->data = 0; in kvm_get_msr_common()
4254 if (kvm_pmu_is_valid_msr(vcpu, msr_info->index)) in kvm_get_msr_common()
4256 msr_info->data = 0; in kvm_get_msr_common()
4259 msr_info->data = vcpu->arch.microcode_version; in kvm_get_msr_common()
4264 msr_info->data = vcpu->arch.arch_capabilities; in kvm_get_msr_common()
4269 msr_info->data = vcpu->arch.perf_capabilities; in kvm_get_msr_common()
4272 msr_info->data = vcpu->arch.msr_ia32_power_ctl; in kvm_get_msr_common()
4281 * return L1's TSC value to ensure backwards-compatible in kvm_get_msr_common()
4286 if (msr_info->host_initiated) { in kvm_get_msr_common()
4287 offset = vcpu->arch.l1_tsc_offset; in kvm_get_msr_common()
4288 ratio = vcpu->arch.l1_tsc_scaling_ratio; in kvm_get_msr_common()
4290 offset = vcpu->arch.tsc_offset; in kvm_get_msr_common()
4291 ratio = vcpu->arch.tsc_scaling_ratio; in kvm_get_msr_common()
4294 msr_info->data = kvm_scale_tsc(rdtsc(), ratio) + offset; in kvm_get_msr_common()
4298 msr_info->data = vcpu->arch.pat; in kvm_get_msr_common()
4303 return kvm_mtrr_get_msr(vcpu, msr_info->index, &msr_info->data); in kvm_get_msr_common()
4305 msr_info->data = 3; in kvm_get_msr_common()
4319 msr_info->data = 1 << 24; in kvm_get_msr_common()
4322 msr_info->data = vcpu->arch.apic_base; in kvm_get_msr_common()
4325 return kvm_x2apic_msr_read(vcpu, msr_info->index, &msr_info->data); in kvm_get_msr_common()
4327 msr_info->data = kvm_get_lapic_tscdeadline_msr(vcpu); in kvm_get_msr_common()
4330 msr_info->data = (u64)vcpu->arch.ia32_tsc_adjust_msr; in kvm_get_msr_common()
4333 msr_info->data = vcpu->arch.ia32_misc_enable_msr; in kvm_get_msr_common()
4336 if (!IS_ENABLED(CONFIG_KVM_SMM) || !msr_info->host_initiated) in kvm_get_msr_common()
4338 msr_info->data = vcpu->arch.smbase; in kvm_get_msr_common()
4341 msr_info->data = vcpu->arch.smi_count; in kvm_get_msr_common()
4345 msr_info->data = 1000ULL; in kvm_get_msr_common()
4347 msr_info->data |= (((uint64_t)4ULL) << 40); in kvm_get_msr_common()
4350 msr_info->data = vcpu->arch.efer; in kvm_get_msr_common()
4356 msr_info->data = vcpu->kvm->arch.wall_clock; in kvm_get_msr_common()
4362 msr_info->data = vcpu->kvm->arch.wall_clock; in kvm_get_msr_common()
4368 msr_info->data = vcpu->arch.time; in kvm_get_msr_common()
4374 msr_info->data = vcpu->arch.time; in kvm_get_msr_common()
4380 msr_info->data = vcpu->arch.apf.msr_en_val; in kvm_get_msr_common()
4386 msr_info->data = vcpu->arch.apf.msr_int_val; in kvm_get_msr_common()
4392 msr_info->data = 0; in kvm_get_msr_common()
4398 msr_info->data = vcpu->arch.st.msr_val; in kvm_get_msr_common()
4404 msr_info->data = vcpu->arch.pv_eoi.msr_val; in kvm_get_msr_common()
4410 msr_info->data = vcpu->arch.msr_kvm_poll_control; in kvm_get_msr_common()
4417 case MSR_IA32_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1: in kvm_get_msr_common()
4418 case MSR_IA32_MC0_CTL2 ... MSR_IA32_MCx_CTL2(KVM_MAX_MCE_BANKS) - 1: in kvm_get_msr_common()
4419 return get_msr_mce(vcpu, msr_info->index, &msr_info->data, in kvm_get_msr_common()
4420 msr_info->host_initiated); in kvm_get_msr_common()
4422 if (!msr_info->host_initiated && in kvm_get_msr_common()
4425 msr_info->data = vcpu->arch.ia32_xss; in kvm_get_msr_common()
4429 * Provide expected ramp-up count for K7. All other in kvm_get_msr_common()
4437 msr_info->data = 0x20000000; in kvm_get_msr_common()
4451 msr_info->index, &msr_info->data, in kvm_get_msr_common()
4452 msr_info->host_initiated); in kvm_get_msr_common()
4465 msr_info->data = 0xbe702111; in kvm_get_msr_common()
4470 msr_info->data = vcpu->arch.osvw.length; in kvm_get_msr_common()
4475 msr_info->data = vcpu->arch.osvw.status; in kvm_get_msr_common()
4478 if (!msr_info->host_initiated && in kvm_get_msr_common()
4479 !vcpu->kvm->arch.guest_can_read_msr_platform_info) in kvm_get_msr_common()
4481 msr_info->data = vcpu->arch.msr_platform_info; in kvm_get_msr_common()
4484 msr_info->data = vcpu->arch.msr_misc_features_enables; in kvm_get_msr_common()
4487 msr_info->data = vcpu->arch.msr_hwcr; in kvm_get_msr_common()
4491 if (!msr_info->host_initiated && in kvm_get_msr_common()
4495 msr_info->data = vcpu->arch.guest_fpu.fpstate->xfd; in kvm_get_msr_common()
4498 if (!msr_info->host_initiated && in kvm_get_msr_common()
4502 msr_info->data = vcpu->arch.guest_fpu.xfd_err; in kvm_get_msr_common()
4506 if (kvm_pmu_is_valid_msr(vcpu, msr_info->index)) in kvm_get_msr_common()
4527 for (i = 0; i < msrs->nmsrs; ++i) in __msr_io()
4549 r = -EFAULT; in msr_io()
4553 r = -E2BIG; in msr_io()
4558 entries = memdup_user(user_msrs->entries, size); in msr_io()
4566 if (writeback && copy_to_user(user_msrs->entries, entries, size)) in msr_io()
4567 r = -EFAULT; in msr_io()
4605 r = -EFAULT; in kvm_ioctl_get_supported_hv_cpuid()
4609 r = kvm_get_hv_cpuid(vcpu, &cpuid, cpuid_arg->entries); in kvm_ioctl_get_supported_hv_cpuid()
4613 r = -EFAULT; in kvm_ioctl_get_supported_hv_cpuid()
4628 return kvm && kvm->arch.has_protected_state ? 0 : KVM_SYNC_X86_VALID_FIELDS; in kvm_sync_valid_fields()
4770 r = kvm->max_vcpus; in kvm_vm_ioctl_check_extension()
4792 r = kvm_x86_ops.nested_ops->get_state ? in kvm_vm_ioctl_check_extension()
4793 kvm_x86_ops.nested_ops->get_state(NULL, NULL, 0) : 0; in kvm_vm_ioctl_check_extension()
4800 r = kvm_x86_ops.nested_ops->enable_evmcs != NULL; in kvm_vm_ioctl_check_extension()
4845 if (attr->group) { in __kvm_x86_dev_get_attr()
4847 return kvm_x86_call(dev_get_attr)(attr->group, attr->attr, val); in __kvm_x86_dev_get_attr()
4848 return -ENXIO; in __kvm_x86_dev_get_attr()
4851 switch (attr->attr) { in __kvm_x86_dev_get_attr()
4856 return -ENXIO; in __kvm_x86_dev_get_attr()
4862 u64 __user *uaddr = u64_to_user_ptr(attr->addr); in kvm_x86_dev_get_attr()
4871 return -EFAULT; in kvm_x86_dev_get_attr()
4895 r = -EFAULT; in kvm_arch_dev_ioctl()
4902 r = -E2BIG; in kvm_arch_dev_ioctl()
4905 r = -EFAULT; in kvm_arch_dev_ioctl()
4906 if (copy_to_user(user_msr_list->indices, &msrs_to_save, in kvm_arch_dev_ioctl()
4909 if (copy_to_user(user_msr_list->indices + num_msrs_to_save, in kvm_arch_dev_ioctl()
4921 r = -EFAULT; in kvm_arch_dev_ioctl()
4925 r = kvm_dev_ioctl_get_cpuid(&cpuid, cpuid_arg->entries, in kvm_arch_dev_ioctl()
4930 r = -EFAULT; in kvm_arch_dev_ioctl()
4937 r = -EFAULT; in kvm_arch_dev_ioctl()
4948 r = -EFAULT; in kvm_arch_dev_ioctl()
4955 r = -E2BIG; in kvm_arch_dev_ioctl()
4958 r = -EFAULT; in kvm_arch_dev_ioctl()
4959 if (copy_to_user(user_msr_list->indices, &msr_based_features, in kvm_arch_dev_ioctl()
4975 r = -EFAULT; in kvm_arch_dev_ioctl()
4983 r = -EFAULT; in kvm_arch_dev_ioctl()
4990 r = -EINVAL; in kvm_arch_dev_ioctl()
4999 return kvm_arch_has_noncoherent_dma(vcpu->kvm); in need_emulate_wbinvd()
5008 vcpu->arch.l1tf_flush_l1d = true; in kvm_arch_vcpu_load()
5010 if (vcpu->scheduled_out && pmu->version && pmu->event_count) { in kvm_arch_vcpu_load()
5011 pmu->need_cleanup = true; in kvm_arch_vcpu_load()
5018 cpumask_set_cpu(cpu, vcpu->arch.wbinvd_dirty_mask); in kvm_arch_vcpu_load()
5019 else if (vcpu->cpu != -1 && vcpu->cpu != cpu) in kvm_arch_vcpu_load()
5020 wbinvd_on_cpu(vcpu->cpu); in kvm_arch_vcpu_load()
5031 * is handled on the nested VM-Exit path. in kvm_arch_vcpu_load()
5039 vcpu->arch.host_pkru = read_pkru(); in kvm_arch_vcpu_load()
5042 if (unlikely(vcpu->arch.tsc_offset_adjustment)) { in kvm_arch_vcpu_load()
5043 adjust_tsc_offset_host(vcpu, vcpu->arch.tsc_offset_adjustment); in kvm_arch_vcpu_load()
5044 vcpu->arch.tsc_offset_adjustment = 0; in kvm_arch_vcpu_load()
5048 if (unlikely(vcpu->cpu != cpu) || kvm_check_tsc_unstable()) { in kvm_arch_vcpu_load()
5049 s64 tsc_delta = !vcpu->arch.last_host_tsc ? 0 : in kvm_arch_vcpu_load()
5050 rdtsc() - vcpu->arch.last_host_tsc; in kvm_arch_vcpu_load()
5056 vcpu->arch.last_guest_tsc); in kvm_arch_vcpu_load()
5058 if (!vcpu->arch.guest_tsc_protected) in kvm_arch_vcpu_load()
5059 vcpu->arch.tsc_catchup = 1; in kvm_arch_vcpu_load()
5067 * kvmclock on vcpu->cpu migration in kvm_arch_vcpu_load()
5069 if (!vcpu->kvm->arch.use_master_clock || vcpu->cpu == -1) in kvm_arch_vcpu_load()
5071 if (vcpu->cpu != cpu) in kvm_arch_vcpu_load()
5073 vcpu->cpu = cpu; in kvm_arch_vcpu_load()
5081 struct gfn_to_hva_cache *ghc = &vcpu->arch.st.cache; in kvm_steal_time_set_preempted()
5085 gpa_t gpa = vcpu->arch.st.msr_val & KVM_STEAL_VALID_BITS; in kvm_steal_time_set_preempted()
5088 * The vCPU can be marked preempted if and only if the VM-Exit was on in kvm_steal_time_set_preempted()
5092 * preempted if and only if the VM-Exit was due to a host interrupt. in kvm_steal_time_set_preempted()
5094 if (!vcpu->arch.at_instruction_boundary) { in kvm_steal_time_set_preempted()
5095 vcpu->stat.preemption_other++; in kvm_steal_time_set_preempted()
5099 vcpu->stat.preemption_reported++; in kvm_steal_time_set_preempted()
5100 if (!(vcpu->arch.st.msr_val & KVM_MSR_ENABLED)) in kvm_steal_time_set_preempted()
5103 if (vcpu->arch.st.preempted) in kvm_steal_time_set_preempted()
5107 if (unlikely(current->mm != vcpu->kvm->mm)) in kvm_steal_time_set_preempted()
5110 slots = kvm_memslots(vcpu->kvm); in kvm_steal_time_set_preempted()
5112 if (unlikely(slots->generation != ghc->generation || in kvm_steal_time_set_preempted()
5113 gpa != ghc->gpa || in kvm_steal_time_set_preempted()
5114 kvm_is_error_hva(ghc->hva) || !ghc->memslot)) in kvm_steal_time_set_preempted()
5117 st = (struct kvm_steal_time __user *)ghc->hva; in kvm_steal_time_set_preempted()
5118 BUILD_BUG_ON(sizeof(st->preempted) != sizeof(preempted)); in kvm_steal_time_set_preempted()
5120 if (!copy_to_user_nofault(&st->preempted, &preempted, sizeof(preempted))) in kvm_steal_time_set_preempted()
5121 vcpu->arch.st.preempted = KVM_VCPU_PREEMPTED; in kvm_steal_time_set_preempted()
5123 mark_page_dirty_in_slot(vcpu->kvm, ghc->memslot, gpa_to_gfn(ghc->gpa)); in kvm_steal_time_set_preempted()
5130 if (vcpu->preempted) { in kvm_arch_vcpu_put()
5132 * Assume protected guests are in-kernel. Inefficient yielding in kvm_arch_vcpu_put()
5136 vcpu->arch.preempted_in_kernel = vcpu->arch.guest_state_protected || in kvm_arch_vcpu_put()
5143 idx = srcu_read_lock(&vcpu->kvm->srcu); in kvm_arch_vcpu_put()
5144 if (kvm_xen_msr_enabled(vcpu->kvm)) in kvm_arch_vcpu_put()
5148 srcu_read_unlock(&vcpu->kvm->srcu, idx); in kvm_arch_vcpu_put()
5152 vcpu->arch.last_host_tsc = rdtsc(); in kvm_arch_vcpu_put()
5158 if (vcpu->arch.apic->guest_apic_protected) in kvm_vcpu_ioctl_get_lapic()
5159 return -EINVAL; in kvm_vcpu_ioctl_get_lapic()
5171 if (vcpu->arch.apic->guest_apic_protected) in kvm_vcpu_ioctl_set_lapic()
5172 return -EINVAL; in kvm_vcpu_ioctl_set_lapic()
5193 /* Acknowledging ExtINT does not happen if LINT0 is masked. */ in kvm_cpu_accept_dm_intr()
5205 * instruction boundary and with no events half-injected. in kvm_vcpu_ready_for_interrupt_injection()
5216 if (irq->irq >= KVM_NR_INTERRUPTS) in kvm_vcpu_ioctl_interrupt()
5217 return -EINVAL; in kvm_vcpu_ioctl_interrupt()
5219 if (!irqchip_in_kernel(vcpu->kvm)) { in kvm_vcpu_ioctl_interrupt()
5220 kvm_queue_interrupt(vcpu, irq->irq, false); in kvm_vcpu_ioctl_interrupt()
5226 * With in-kernel LAPIC, we only use this to inject EXTINT, so in kvm_vcpu_ioctl_interrupt()
5227 * fail for in-kernel 8259. in kvm_vcpu_ioctl_interrupt()
5229 if (pic_in_kernel(vcpu->kvm)) in kvm_vcpu_ioctl_interrupt()
5230 return -ENXIO; in kvm_vcpu_ioctl_interrupt()
5232 if (vcpu->arch.pending_external_vector != -1) in kvm_vcpu_ioctl_interrupt()
5233 return -EEXIST; in kvm_vcpu_ioctl_interrupt()
5235 vcpu->arch.pending_external_vector = irq->irq; in kvm_vcpu_ioctl_interrupt()
5250 if (tac->flags) in vcpu_ioctl_tpr_access_reporting()
5251 return -EINVAL; in vcpu_ioctl_tpr_access_reporting()
5252 vcpu->arch.tpr_access_reporting = !!tac->enabled; in vcpu_ioctl_tpr_access_reporting()
5262 r = -EINVAL; in kvm_vcpu_ioctl_x86_setup_mce()
5268 vcpu->arch.mcg_cap = mcg_cap; in kvm_vcpu_ioctl_x86_setup_mce()
5271 vcpu->arch.mcg_ctl = ~(u64)0; in kvm_vcpu_ioctl_x86_setup_mce()
5274 vcpu->arch.mce_banks[bank*4] = ~(u64)0; in kvm_vcpu_ioctl_x86_setup_mce()
5276 vcpu->arch.mci_ctl2_banks[bank] = 0; in kvm_vcpu_ioctl_x86_setup_mce()
5289 * - none of the bits for Machine Check Exceptions are set
5290 * - both the VAL (valid) and UC (uncorrectable) bits are set
5291 * MCI_STATUS_PCC - Processor Context Corrupted
5292 * MCI_STATUS_S - Signaled as a Machine Check Exception
5293 * MCI_STATUS_AR - Software recoverable Action Required
5297 return !mce->mcg_status && in is_ucna()
5298 !(mce->status & (MCI_STATUS_PCC | MCI_STATUS_S | MCI_STATUS_AR)) && in is_ucna()
5299 (mce->status & MCI_STATUS_VAL) && in is_ucna()
5300 (mce->status & MCI_STATUS_UC); in is_ucna()
5305 u64 mcg_cap = vcpu->arch.mcg_cap; in kvm_vcpu_x86_set_ucna()
5307 banks[1] = mce->status; in kvm_vcpu_x86_set_ucna()
5308 banks[2] = mce->addr; in kvm_vcpu_x86_set_ucna()
5309 banks[3] = mce->misc; in kvm_vcpu_x86_set_ucna()
5310 vcpu->arch.mcg_status = mce->mcg_status; in kvm_vcpu_x86_set_ucna()
5313 !(vcpu->arch.mci_ctl2_banks[mce->bank] & MCI_CTL2_CMCI_EN)) in kvm_vcpu_x86_set_ucna()
5317 kvm_apic_local_deliver(vcpu->arch.apic, APIC_LVTCMCI); in kvm_vcpu_x86_set_ucna()
5325 u64 mcg_cap = vcpu->arch.mcg_cap; in kvm_vcpu_ioctl_x86_set_mce()
5327 u64 *banks = vcpu->arch.mce_banks; in kvm_vcpu_ioctl_x86_set_mce()
5329 if (mce->bank >= bank_num || !(mce->status & MCI_STATUS_VAL)) in kvm_vcpu_ioctl_x86_set_mce()
5330 return -EINVAL; in kvm_vcpu_ioctl_x86_set_mce()
5332 banks += array_index_nospec(4 * mce->bank, 4 * bank_num); in kvm_vcpu_ioctl_x86_set_mce()
5341 if ((mce->status & MCI_STATUS_UC) && (mcg_cap & MCG_CTL_P) && in kvm_vcpu_ioctl_x86_set_mce()
5342 vcpu->arch.mcg_ctl != ~(u64)0) in kvm_vcpu_ioctl_x86_set_mce()
5348 if ((mce->status & MCI_STATUS_UC) && banks[0] != ~(u64)0) in kvm_vcpu_ioctl_x86_set_mce()
5350 if (mce->status & MCI_STATUS_UC) { in kvm_vcpu_ioctl_x86_set_mce()
5351 if ((vcpu->arch.mcg_status & MCG_STATUS_MCIP) || in kvm_vcpu_ioctl_x86_set_mce()
5357 mce->status |= MCI_STATUS_OVER; in kvm_vcpu_ioctl_x86_set_mce()
5358 banks[2] = mce->addr; in kvm_vcpu_ioctl_x86_set_mce()
5359 banks[3] = mce->misc; in kvm_vcpu_ioctl_x86_set_mce()
5360 vcpu->arch.mcg_status = mce->mcg_status; in kvm_vcpu_ioctl_x86_set_mce()
5361 banks[1] = mce->status; in kvm_vcpu_ioctl_x86_set_mce()
5366 mce->status |= MCI_STATUS_OVER; in kvm_vcpu_ioctl_x86_set_mce()
5367 banks[2] = mce->addr; in kvm_vcpu_ioctl_x86_set_mce()
5368 banks[3] = mce->misc; in kvm_vcpu_ioctl_x86_set_mce()
5369 banks[1] = mce->status; in kvm_vcpu_ioctl_x86_set_mce()
5390 * non-exiting _injected_ exception, and a pending exiting exception. in kvm_vcpu_ioctl_x86_get_vcpu_events()
5391 * In that case, ignore the VM-Exiting exception as it's an extension in kvm_vcpu_ioctl_x86_get_vcpu_events()
5394 if (vcpu->arch.exception_vmexit.pending && in kvm_vcpu_ioctl_x86_get_vcpu_events()
5395 !vcpu->arch.exception.pending && in kvm_vcpu_ioctl_x86_get_vcpu_events()
5396 !vcpu->arch.exception.injected) in kvm_vcpu_ioctl_x86_get_vcpu_events()
5397 ex = &vcpu->arch.exception_vmexit; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5399 ex = &vcpu->arch.exception; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5404 * intercepts #PF, ditto for DR6 and #DBs. If the per-VM capability, in kvm_vcpu_ioctl_x86_get_vcpu_events()
5409 if (!vcpu->kvm->arch.exception_payload_enabled && in kvm_vcpu_ioctl_x86_get_vcpu_events()
5410 ex->pending && ex->has_payload) in kvm_vcpu_ioctl_x86_get_vcpu_events()
5421 if (!kvm_exception_is_soft(ex->vector)) { in kvm_vcpu_ioctl_x86_get_vcpu_events()
5422 events->exception.injected = ex->injected; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5423 events->exception.pending = ex->pending; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5429 if (!vcpu->kvm->arch.exception_payload_enabled) in kvm_vcpu_ioctl_x86_get_vcpu_events()
5430 events->exception.injected |= ex->pending; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5432 events->exception.nr = ex->vector; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5433 events->exception.has_error_code = ex->has_error_code; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5434 events->exception.error_code = ex->error_code; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5435 events->exception_has_payload = ex->has_payload; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5436 events->exception_payload = ex->payload; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5438 events->interrupt.injected = in kvm_vcpu_ioctl_x86_get_vcpu_events()
5439 vcpu->arch.interrupt.injected && !vcpu->arch.interrupt.soft; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5440 events->interrupt.nr = vcpu->arch.interrupt.nr; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5441 events->interrupt.shadow = kvm_x86_call(get_interrupt_shadow)(vcpu); in kvm_vcpu_ioctl_x86_get_vcpu_events()
5443 events->nmi.injected = vcpu->arch.nmi_injected; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5444 events->nmi.pending = kvm_get_nr_pending_nmis(vcpu); in kvm_vcpu_ioctl_x86_get_vcpu_events()
5445 events->nmi.masked = kvm_x86_call(get_nmi_mask)(vcpu); in kvm_vcpu_ioctl_x86_get_vcpu_events()
5447 /* events->sipi_vector is never valid when reporting to user space */ in kvm_vcpu_ioctl_x86_get_vcpu_events()
5450 events->smi.smm = is_smm(vcpu); in kvm_vcpu_ioctl_x86_get_vcpu_events()
5451 events->smi.pending = vcpu->arch.smi_pending; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5452 events->smi.smm_inside_nmi = in kvm_vcpu_ioctl_x86_get_vcpu_events()
5453 !!(vcpu->arch.hflags & HF_SMM_INSIDE_NMI_MASK); in kvm_vcpu_ioctl_x86_get_vcpu_events()
5455 events->smi.latched_init = kvm_lapic_latched_init(vcpu); in kvm_vcpu_ioctl_x86_get_vcpu_events()
5457 events->flags = (KVM_VCPUEVENT_VALID_NMI_PENDING in kvm_vcpu_ioctl_x86_get_vcpu_events()
5460 if (vcpu->kvm->arch.exception_payload_enabled) in kvm_vcpu_ioctl_x86_get_vcpu_events()
5461 events->flags |= KVM_VCPUEVENT_VALID_PAYLOAD; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5462 if (vcpu->kvm->arch.triple_fault_event) { in kvm_vcpu_ioctl_x86_get_vcpu_events()
5463 events->triple_fault.pending = kvm_test_request(KVM_REQ_TRIPLE_FAULT, vcpu); in kvm_vcpu_ioctl_x86_get_vcpu_events()
5464 events->flags |= KVM_VCPUEVENT_VALID_TRIPLE_FAULT; in kvm_vcpu_ioctl_x86_get_vcpu_events()
5471 if (events->flags & ~(KVM_VCPUEVENT_VALID_NMI_PENDING in kvm_vcpu_ioctl_x86_set_vcpu_events()
5477 return -EINVAL; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5479 if (events->flags & KVM_VCPUEVENT_VALID_PAYLOAD) { in kvm_vcpu_ioctl_x86_set_vcpu_events()
5480 if (!vcpu->kvm->arch.exception_payload_enabled) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5481 return -EINVAL; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5482 if (events->exception.pending) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5483 events->exception.injected = 0; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5485 events->exception_has_payload = 0; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5487 events->exception.pending = 0; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5488 events->exception_has_payload = 0; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5491 if ((events->exception.injected || events->exception.pending) && in kvm_vcpu_ioctl_x86_set_vcpu_events()
5492 (events->exception.nr > 31 || events->exception.nr == NMI_VECTOR)) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5493 return -EINVAL; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5499 * morph the exception to a VM-Exit if appropriate. Do this only for in kvm_vcpu_ioctl_x86_set_vcpu_events()
5500 * pending exceptions, already-injected exceptions are not subject to in kvm_vcpu_ioctl_x86_set_vcpu_events()
5503 * pending exception, which in turn may cause a spurious VM-Exit. in kvm_vcpu_ioctl_x86_set_vcpu_events()
5505 vcpu->arch.exception_from_userspace = events->exception.pending; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5507 vcpu->arch.exception_vmexit.pending = false; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5509 vcpu->arch.exception.injected = events->exception.injected; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5510 vcpu->arch.exception.pending = events->exception.pending; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5511 vcpu->arch.exception.vector = events->exception.nr; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5512 vcpu->arch.exception.has_error_code = events->exception.has_error_code; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5513 vcpu->arch.exception.error_code = events->exception.error_code; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5514 vcpu->arch.exception.has_payload = events->exception_has_payload; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5515 vcpu->arch.exception.payload = events->exception_payload; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5517 vcpu->arch.interrupt.injected = events->interrupt.injected; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5518 vcpu->arch.interrupt.nr = events->interrupt.nr; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5519 vcpu->arch.interrupt.soft = events->interrupt.soft; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5520 if (events->flags & KVM_VCPUEVENT_VALID_SHADOW) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5522 events->interrupt.shadow); in kvm_vcpu_ioctl_x86_set_vcpu_events()
5524 vcpu->arch.nmi_injected = events->nmi.injected; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5525 if (events->flags & KVM_VCPUEVENT_VALID_NMI_PENDING) { in kvm_vcpu_ioctl_x86_set_vcpu_events()
5526 vcpu->arch.nmi_pending = 0; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5527 atomic_set(&vcpu->arch.nmi_queued, events->nmi.pending); in kvm_vcpu_ioctl_x86_set_vcpu_events()
5528 if (events->nmi.pending) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5531 kvm_x86_call(set_nmi_mask)(vcpu, events->nmi.masked); in kvm_vcpu_ioctl_x86_set_vcpu_events()
5533 if (events->flags & KVM_VCPUEVENT_VALID_SIPI_VECTOR && in kvm_vcpu_ioctl_x86_set_vcpu_events()
5535 vcpu->arch.apic->sipi_vector = events->sipi_vector; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5537 if (events->flags & KVM_VCPUEVENT_VALID_SMM) { in kvm_vcpu_ioctl_x86_set_vcpu_events()
5539 if (!!(vcpu->arch.hflags & HF_SMM_MASK) != events->smi.smm) { in kvm_vcpu_ioctl_x86_set_vcpu_events()
5541 kvm_smm_changed(vcpu, events->smi.smm); in kvm_vcpu_ioctl_x86_set_vcpu_events()
5544 vcpu->arch.smi_pending = events->smi.pending; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5546 if (events->smi.smm) { in kvm_vcpu_ioctl_x86_set_vcpu_events()
5547 if (events->smi.smm_inside_nmi) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5548 vcpu->arch.hflags |= HF_SMM_INSIDE_NMI_MASK; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5550 vcpu->arch.hflags &= ~HF_SMM_INSIDE_NMI_MASK; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5554 if (events->smi.smm || events->smi.pending || in kvm_vcpu_ioctl_x86_set_vcpu_events()
5555 events->smi.smm_inside_nmi) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5556 return -EINVAL; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5560 if (events->smi.latched_init) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5561 set_bit(KVM_APIC_INIT, &vcpu->arch.apic->pending_events); in kvm_vcpu_ioctl_x86_set_vcpu_events()
5563 clear_bit(KVM_APIC_INIT, &vcpu->arch.apic->pending_events); in kvm_vcpu_ioctl_x86_set_vcpu_events()
5567 if (events->flags & KVM_VCPUEVENT_VALID_TRIPLE_FAULT) { in kvm_vcpu_ioctl_x86_set_vcpu_events()
5568 if (!vcpu->kvm->arch.triple_fault_event) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5569 return -EINVAL; in kvm_vcpu_ioctl_x86_set_vcpu_events()
5570 if (events->triple_fault.pending) in kvm_vcpu_ioctl_x86_set_vcpu_events()
5586 if (vcpu->kvm->arch.has_protected_state && in kvm_vcpu_ioctl_x86_get_debugregs()
5587 vcpu->arch.guest_state_protected) in kvm_vcpu_ioctl_x86_get_debugregs()
5588 return -EINVAL; in kvm_vcpu_ioctl_x86_get_debugregs()
5592 BUILD_BUG_ON(ARRAY_SIZE(vcpu->arch.db) != ARRAY_SIZE(dbgregs->db)); in kvm_vcpu_ioctl_x86_get_debugregs()
5593 for (i = 0; i < ARRAY_SIZE(vcpu->arch.db); i++) in kvm_vcpu_ioctl_x86_get_debugregs()
5594 dbgregs->db[i] = vcpu->arch.db[i]; in kvm_vcpu_ioctl_x86_get_debugregs()
5596 dbgregs->dr6 = vcpu->arch.dr6; in kvm_vcpu_ioctl_x86_get_debugregs()
5597 dbgregs->dr7 = vcpu->arch.dr7; in kvm_vcpu_ioctl_x86_get_debugregs()
5606 if (vcpu->kvm->arch.has_protected_state && in kvm_vcpu_ioctl_x86_set_debugregs()
5607 vcpu->arch.guest_state_protected) in kvm_vcpu_ioctl_x86_set_debugregs()
5608 return -EINVAL; in kvm_vcpu_ioctl_x86_set_debugregs()
5610 if (dbgregs->flags) in kvm_vcpu_ioctl_x86_set_debugregs()
5611 return -EINVAL; in kvm_vcpu_ioctl_x86_set_debugregs()
5613 if (!kvm_dr6_valid(dbgregs->dr6)) in kvm_vcpu_ioctl_x86_set_debugregs()
5614 return -EINVAL; in kvm_vcpu_ioctl_x86_set_debugregs()
5615 if (!kvm_dr7_valid(dbgregs->dr7)) in kvm_vcpu_ioctl_x86_set_debugregs()
5616 return -EINVAL; in kvm_vcpu_ioctl_x86_set_debugregs()
5618 for (i = 0; i < ARRAY_SIZE(vcpu->arch.db); i++) in kvm_vcpu_ioctl_x86_set_debugregs()
5619 vcpu->arch.db[i] = dbgregs->db[i]; in kvm_vcpu_ioctl_x86_set_debugregs()
5622 vcpu->arch.dr6 = dbgregs->dr6; in kvm_vcpu_ioctl_x86_set_debugregs()
5623 vcpu->arch.dr7 = dbgregs->dr7; in kvm_vcpu_ioctl_x86_set_debugregs()
5645 u64 supported_xcr0 = vcpu->arch.guest_supported_xcr0 | in kvm_vcpu_ioctl_x86_get_xsave2()
5648 if (fpstate_is_confidential(&vcpu->arch.guest_fpu)) in kvm_vcpu_ioctl_x86_get_xsave2()
5649 return vcpu->kvm->arch.has_protected_state ? -EINVAL : 0; in kvm_vcpu_ioctl_x86_get_xsave2()
5651 fpu_copy_guest_fpstate_to_uabi(&vcpu->arch.guest_fpu, state, size, in kvm_vcpu_ioctl_x86_get_xsave2()
5652 supported_xcr0, vcpu->arch.pkru); in kvm_vcpu_ioctl_x86_get_xsave2()
5659 return kvm_vcpu_ioctl_x86_get_xsave2(vcpu, (void *)guest_xsave->region, in kvm_vcpu_ioctl_x86_get_xsave()
5660 sizeof(guest_xsave->region)); in kvm_vcpu_ioctl_x86_get_xsave()
5666 if (fpstate_is_confidential(&vcpu->arch.guest_fpu)) in kvm_vcpu_ioctl_x86_set_xsave()
5667 return vcpu->kvm->arch.has_protected_state ? -EINVAL : 0; in kvm_vcpu_ioctl_x86_set_xsave()
5669 return fpu_copy_uabi_to_guest_fpstate(&vcpu->arch.guest_fpu, in kvm_vcpu_ioctl_x86_set_xsave()
5670 guest_xsave->region, in kvm_vcpu_ioctl_x86_set_xsave()
5672 &vcpu->arch.pkru); in kvm_vcpu_ioctl_x86_set_xsave()
5678 if (vcpu->kvm->arch.has_protected_state && in kvm_vcpu_ioctl_x86_get_xcrs()
5679 vcpu->arch.guest_state_protected) in kvm_vcpu_ioctl_x86_get_xcrs()
5680 return -EINVAL; in kvm_vcpu_ioctl_x86_get_xcrs()
5683 guest_xcrs->nr_xcrs = 0; in kvm_vcpu_ioctl_x86_get_xcrs()
5687 guest_xcrs->nr_xcrs = 1; in kvm_vcpu_ioctl_x86_get_xcrs()
5688 guest_xcrs->flags = 0; in kvm_vcpu_ioctl_x86_get_xcrs()
5689 guest_xcrs->xcrs[0].xcr = XCR_XFEATURE_ENABLED_MASK; in kvm_vcpu_ioctl_x86_get_xcrs()
5690 guest_xcrs->xcrs[0].value = vcpu->arch.xcr0; in kvm_vcpu_ioctl_x86_get_xcrs()
5699 if (vcpu->kvm->arch.has_protected_state && in kvm_vcpu_ioctl_x86_set_xcrs()
5700 vcpu->arch.guest_state_protected) in kvm_vcpu_ioctl_x86_set_xcrs()
5701 return -EINVAL; in kvm_vcpu_ioctl_x86_set_xcrs()
5704 return -EINVAL; in kvm_vcpu_ioctl_x86_set_xcrs()
5706 if (guest_xcrs->nr_xcrs > KVM_MAX_XCRS || guest_xcrs->flags) in kvm_vcpu_ioctl_x86_set_xcrs()
5707 return -EINVAL; in kvm_vcpu_ioctl_x86_set_xcrs()
5709 for (i = 0; i < guest_xcrs->nr_xcrs; i++) in kvm_vcpu_ioctl_x86_set_xcrs()
5711 if (guest_xcrs->xcrs[i].xcr == XCR_XFEATURE_ENABLED_MASK) { in kvm_vcpu_ioctl_x86_set_xcrs()
5713 guest_xcrs->xcrs[i].value); in kvm_vcpu_ioctl_x86_set_xcrs()
5717 r = -EINVAL; in kvm_vcpu_ioctl_x86_set_xcrs()
5729 if (!vcpu->arch.pv_time.active) in kvm_set_guest_paused()
5730 return -EINVAL; in kvm_set_guest_paused()
5731 vcpu->arch.pvclock_set_guest_stopped_request = true; in kvm_set_guest_paused()
5741 switch (attr->attr) { in kvm_arch_tsc_has_attr()
5746 r = -ENXIO; in kvm_arch_tsc_has_attr()
5755 u64 __user *uaddr = u64_to_user_ptr(attr->addr); in kvm_arch_tsc_get_attr()
5758 switch (attr->attr) { in kvm_arch_tsc_get_attr()
5760 r = -EFAULT; in kvm_arch_tsc_get_attr()
5761 if (put_user(vcpu->arch.l1_tsc_offset, uaddr)) in kvm_arch_tsc_get_attr()
5766 r = -ENXIO; in kvm_arch_tsc_get_attr()
5775 u64 __user *uaddr = u64_to_user_ptr(attr->addr); in kvm_arch_tsc_set_attr()
5776 struct kvm *kvm = vcpu->kvm; in kvm_arch_tsc_set_attr()
5779 switch (attr->attr) { in kvm_arch_tsc_set_attr()
5785 r = -EFAULT; in kvm_arch_tsc_set_attr()
5789 raw_spin_lock_irqsave(&kvm->arch.tsc_write_lock, flags); in kvm_arch_tsc_set_attr()
5791 matched = (vcpu->arch.virtual_tsc_khz && in kvm_arch_tsc_set_attr()
5792 kvm->arch.last_tsc_khz == vcpu->arch.virtual_tsc_khz && in kvm_arch_tsc_set_attr()
5793 kvm->arch.last_tsc_offset == offset); in kvm_arch_tsc_set_attr()
5795 tsc = kvm_scale_tsc(rdtsc(), vcpu->arch.l1_tsc_scaling_ratio) + offset; in kvm_arch_tsc_set_attr()
5799 raw_spin_unlock_irqrestore(&kvm->arch.tsc_write_lock, flags); in kvm_arch_tsc_set_attr()
5805 r = -ENXIO; in kvm_arch_tsc_set_attr()
5819 return -EFAULT; in kvm_vcpu_ioctl_device_attr()
5822 return -ENXIO; in kvm_vcpu_ioctl_device_attr()
5842 if (cap->flags) in kvm_vcpu_ioctl_enable_cap()
5843 return -EINVAL; in kvm_vcpu_ioctl_enable_cap()
5845 switch (cap->cap) { in kvm_vcpu_ioctl_enable_cap()
5848 if (cap->args[0]) in kvm_vcpu_ioctl_enable_cap()
5849 return -EINVAL; in kvm_vcpu_ioctl_enable_cap()
5853 if (!irqchip_in_kernel(vcpu->kvm)) in kvm_vcpu_ioctl_enable_cap()
5854 return -EINVAL; in kvm_vcpu_ioctl_enable_cap()
5855 return kvm_hv_activate_synic(vcpu, cap->cap == in kvm_vcpu_ioctl_enable_cap()
5863 if (!kvm_x86_ops.nested_ops->enable_evmcs) in kvm_vcpu_ioctl_enable_cap()
5864 return -ENOTTY; in kvm_vcpu_ioctl_enable_cap()
5865 r = kvm_x86_ops.nested_ops->enable_evmcs(vcpu, &vmcs_version); in kvm_vcpu_ioctl_enable_cap()
5867 user_ptr = (void __user *)(uintptr_t)cap->args[0]; in kvm_vcpu_ioctl_enable_cap()
5870 r = -EFAULT; in kvm_vcpu_ioctl_enable_cap()
5876 return -ENOTTY; in kvm_vcpu_ioctl_enable_cap()
5881 return kvm_hv_set_enforce_cpuid(vcpu, cap->args[0]); in kvm_vcpu_ioctl_enable_cap()
5885 vcpu->arch.pv_cpuid.enforce = cap->args[0]; in kvm_vcpu_ioctl_enable_cap()
5888 return -EINVAL; in kvm_vcpu_ioctl_enable_cap()
5895 struct kvm_vcpu *vcpu = filp->private_data; in kvm_arch_vcpu_ioctl()
5911 r = -EINVAL; in kvm_arch_vcpu_ioctl()
5916 r = -ENOMEM; in kvm_arch_vcpu_ioctl()
5922 r = -EFAULT; in kvm_arch_vcpu_ioctl()
5929 r = -EINVAL; in kvm_arch_vcpu_ioctl()
5944 r = -EFAULT; in kvm_arch_vcpu_ioctl()
5962 r = -EFAULT; in kvm_arch_vcpu_ioctl()
5965 r = kvm_vcpu_ioctl_set_cpuid(vcpu, &cpuid, cpuid_arg->entries); in kvm_arch_vcpu_ioctl()
5972 r = -EFAULT; in kvm_arch_vcpu_ioctl()
5976 cpuid_arg->entries); in kvm_arch_vcpu_ioctl()
5983 r = -EFAULT; in kvm_arch_vcpu_ioctl()
5987 cpuid_arg->entries); in kvm_arch_vcpu_ioctl()
5990 r = -EFAULT; in kvm_arch_vcpu_ioctl()
5997 int idx = srcu_read_lock(&vcpu->kvm->srcu); in kvm_arch_vcpu_ioctl()
5999 srcu_read_unlock(&vcpu->kvm->srcu, idx); in kvm_arch_vcpu_ioctl()
6003 int idx = srcu_read_lock(&vcpu->kvm->srcu); in kvm_arch_vcpu_ioctl()
6005 srcu_read_unlock(&vcpu->kvm->srcu, idx); in kvm_arch_vcpu_ioctl()
6011 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6017 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6027 r = -EINVAL; in kvm_arch_vcpu_ioctl()
6030 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6033 idx = srcu_read_lock(&vcpu->kvm->srcu); in kvm_arch_vcpu_ioctl()
6035 srcu_read_unlock(&vcpu->kvm->srcu, idx); in kvm_arch_vcpu_ioctl()
6041 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6050 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6061 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6070 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6086 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6096 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6105 r = -EINVAL; in kvm_arch_vcpu_ioctl()
6106 if (vcpu->arch.guest_fpu.uabi_size > sizeof(struct kvm_xsave)) in kvm_arch_vcpu_ioctl()
6110 r = -ENOMEM; in kvm_arch_vcpu_ioctl()
6118 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6125 int size = vcpu->arch.guest_fpu.uabi_size; in kvm_arch_vcpu_ioctl()
6138 int size = vcpu->arch.guest_fpu.uabi_size; in kvm_arch_vcpu_ioctl()
6141 r = -ENOMEM; in kvm_arch_vcpu_ioctl()
6149 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6159 r = -ENOMEM; in kvm_arch_vcpu_ioctl()
6167 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6187 r = -EINVAL; in kvm_arch_vcpu_ioctl()
6189 if (vcpu->arch.guest_tsc_protected) in kvm_arch_vcpu_ioctl()
6207 r = vcpu->arch.virtual_tsc_khz; in kvm_arch_vcpu_ioctl()
6217 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6227 r = -EINVAL; in kvm_arch_vcpu_ioctl()
6228 if (!kvm_x86_ops.nested_ops->get_state) in kvm_arch_vcpu_ioctl()
6231 BUILD_BUG_ON(sizeof(user_data_size) != sizeof(user_kvm_nested_state->size)); in kvm_arch_vcpu_ioctl()
6232 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6233 if (get_user(user_data_size, &user_kvm_nested_state->size)) in kvm_arch_vcpu_ioctl()
6236 r = kvm_x86_ops.nested_ops->get_state(vcpu, user_kvm_nested_state, in kvm_arch_vcpu_ioctl()
6242 if (put_user(r, &user_kvm_nested_state->size)) in kvm_arch_vcpu_ioctl()
6243 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6245 r = -E2BIG; in kvm_arch_vcpu_ioctl()
6257 r = -EINVAL; in kvm_arch_vcpu_ioctl()
6258 if (!kvm_x86_ops.nested_ops->set_state) in kvm_arch_vcpu_ioctl()
6261 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6265 r = -EINVAL; in kvm_arch_vcpu_ioctl()
6280 idx = srcu_read_lock(&vcpu->kvm->srcu); in kvm_arch_vcpu_ioctl()
6281 r = kvm_x86_ops.nested_ops->set_state(vcpu, user_kvm_nested_state, &kvm_state); in kvm_arch_vcpu_ioctl()
6282 srcu_read_unlock(&vcpu->kvm->srcu, idx); in kvm_arch_vcpu_ioctl()
6294 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6299 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6305 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6313 r = -EINVAL; in kvm_arch_vcpu_ioctl()
6314 if (vcpu->kvm->arch.has_protected_state && in kvm_arch_vcpu_ioctl()
6315 vcpu->arch.guest_state_protected) in kvm_arch_vcpu_ioctl()
6319 r = -ENOMEM; in kvm_arch_vcpu_ioctl()
6323 r = -EFAULT; in kvm_arch_vcpu_ioctl()
6330 r = -EINVAL; in kvm_arch_vcpu_ioctl()
6331 if (vcpu->kvm->arch.has_protected_state && in kvm_arch_vcpu_ioctl()
6332 vcpu->arch.guest_state_protected) in kvm_arch_vcpu_ioctl()
6350 r = -ENOTTY; in kvm_arch_vcpu_ioctl()
6356 r = -EINVAL; in kvm_arch_vcpu_ioctl()
6374 if (addr > (unsigned int)(-3 * PAGE_SIZE)) in kvm_vm_ioctl_set_tss_addr()
6375 return -EINVAL; in kvm_vm_ioctl_set_tss_addr()
6390 return -EINVAL; in kvm_vm_ioctl_set_nr_mmu_pages()
6392 mutex_lock(&kvm->slots_lock); in kvm_vm_ioctl_set_nr_mmu_pages()
6395 kvm->arch.n_requested_mmu_pages = kvm_nr_mmu_pages; in kvm_vm_ioctl_set_nr_mmu_pages()
6397 mutex_unlock(&kvm->slots_lock); in kvm_vm_ioctl_set_nr_mmu_pages()
6407 * on all VM-Exits, thus we only need to kick running vCPUs to force a in kvm_arch_sync_dirty_log()
6408 * VM-Exit. in kvm_arch_sync_dirty_log()
6413 if (!kvm->arch.cpu_dirty_log_size) in kvm_arch_sync_dirty_log()
6425 if (cap->flags) in kvm_vm_ioctl_enable_cap()
6426 return -EINVAL; in kvm_vm_ioctl_enable_cap()
6428 switch (cap->cap) { in kvm_vm_ioctl_enable_cap()
6430 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6431 if (cap->args[0] & ~kvm_caps.supported_quirks) in kvm_vm_ioctl_enable_cap()
6435 kvm->arch.disabled_quirks |= cap->args[0] & kvm_caps.supported_quirks; in kvm_vm_ioctl_enable_cap()
6439 mutex_lock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6440 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6441 if (cap->args[0] > MAX_NR_RESERVED_IOAPIC_PINS) in kvm_vm_ioctl_enable_cap()
6443 r = -EEXIST; in kvm_vm_ioctl_enable_cap()
6446 if (kvm->created_vcpus) in kvm_vm_ioctl_enable_cap()
6450 kvm->arch.irqchip_mode = KVM_IRQCHIP_SPLIT; in kvm_vm_ioctl_enable_cap()
6451 kvm->arch.nr_reserved_ioapic_pins = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6455 mutex_unlock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6459 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6460 if (cap->args[0] & ~KVM_X2APIC_API_VALID_FLAGS) in kvm_vm_ioctl_enable_cap()
6463 if (cap->args[0] & KVM_X2APIC_API_USE_32BIT_IDS) in kvm_vm_ioctl_enable_cap()
6464 kvm->arch.x2apic_format = true; in kvm_vm_ioctl_enable_cap()
6465 if (cap->args[0] & KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK) in kvm_vm_ioctl_enable_cap()
6466 kvm->arch.x2apic_broadcast_quirk_disabled = true; in kvm_vm_ioctl_enable_cap()
6471 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6472 if (cap->args[0] & ~kvm_get_allowed_disable_exits()) in kvm_vm_ioctl_enable_cap()
6475 mutex_lock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6476 if (kvm->created_vcpus) in kvm_vm_ioctl_enable_cap()
6479 #define SMT_RSB_MSG "This processor is affected by the Cross-Thread Return Predictions vulnerabilit… in kvm_vm_ioctl_enable_cap()
6484 (cap->args[0] & ~(KVM_X86_DISABLE_EXITS_PAUSE | in kvm_vm_ioctl_enable_cap()
6488 kvm_disable_exits(kvm, cap->args[0]); in kvm_vm_ioctl_enable_cap()
6491 mutex_unlock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6494 kvm->arch.guest_can_read_msr_platform_info = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6498 kvm->arch.exception_payload_enabled = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6502 kvm->arch.triple_fault_event = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6506 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6507 if (cap->args[0] & ~KVM_MSR_EXIT_REASON_VALID_MASK) in kvm_vm_ioctl_enable_cap()
6509 kvm->arch.user_space_msr_mask = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6513 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6514 if (cap->args[0] & ~KVM_BUS_LOCK_DETECTION_VALID_MODE) in kvm_vm_ioctl_enable_cap()
6517 if ((cap->args[0] & KVM_BUS_LOCK_DETECTION_OFF) && in kvm_vm_ioctl_enable_cap()
6518 (cap->args[0] & KVM_BUS_LOCK_DETECTION_EXIT)) in kvm_vm_ioctl_enable_cap()
6522 cap->args[0] & KVM_BUS_LOCK_DETECTION_EXIT) in kvm_vm_ioctl_enable_cap()
6523 kvm->arch.bus_lock_detection_enabled = true; in kvm_vm_ioctl_enable_cap()
6530 r = sgx_set_attribute(&allowed_attributes, cap->args[0]); in kvm_vm_ioctl_enable_cap()
6537 kvm->arch.sgx_provisioning_allowed = true; in kvm_vm_ioctl_enable_cap()
6539 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6544 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6548 r = kvm_x86_call(vm_copy_enc_context_from)(kvm, cap->args[0]); in kvm_vm_ioctl_enable_cap()
6551 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6555 r = kvm_x86_call(vm_move_enc_context_from)(kvm, cap->args[0]); in kvm_vm_ioctl_enable_cap()
6558 if (cap->args[0] & ~KVM_EXIT_HYPERCALL_VALID_MASK) { in kvm_vm_ioctl_enable_cap()
6559 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6562 kvm->arch.hypercall_exit_enabled = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6566 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6567 if (cap->args[0] & ~1) in kvm_vm_ioctl_enable_cap()
6569 kvm->arch.exit_on_emulation_error = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6573 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6574 if (!enable_pmu || (cap->args[0] & ~KVM_CAP_PMU_VALID_MASK)) in kvm_vm_ioctl_enable_cap()
6577 mutex_lock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6578 if (!kvm->created_vcpus) { in kvm_vm_ioctl_enable_cap()
6579 kvm->arch.enable_pmu = !(cap->args[0] & KVM_PMU_CAP_DISABLE); in kvm_vm_ioctl_enable_cap()
6582 mutex_unlock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6585 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6586 if (cap->args[0] > KVM_MAX_VCPU_IDS) in kvm_vm_ioctl_enable_cap()
6589 mutex_lock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6590 if (kvm->arch.bsp_vcpu_id > cap->args[0]) { in kvm_vm_ioctl_enable_cap()
6592 } else if (kvm->arch.max_vcpu_ids == cap->args[0]) { in kvm_vm_ioctl_enable_cap()
6594 } else if (!kvm->arch.max_vcpu_ids) { in kvm_vm_ioctl_enable_cap()
6595 kvm->arch.max_vcpu_ids = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6598 mutex_unlock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6601 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6602 if ((u32)cap->args[0] & ~KVM_X86_NOTIFY_VMEXIT_VALID_BITS) in kvm_vm_ioctl_enable_cap()
6606 if (!((u32)cap->args[0] & KVM_X86_NOTIFY_VMEXIT_ENABLED)) in kvm_vm_ioctl_enable_cap()
6608 mutex_lock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6609 if (!kvm->created_vcpus) { in kvm_vm_ioctl_enable_cap()
6610 kvm->arch.notify_window = cap->args[0] >> 32; in kvm_vm_ioctl_enable_cap()
6611 kvm->arch.notify_vmexit_flags = (u32)cap->args[0]; in kvm_vm_ioctl_enable_cap()
6614 mutex_unlock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6617 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6631 r = -EPERM; in kvm_vm_ioctl_enable_cap()
6635 if (cap->args[0]) in kvm_vm_ioctl_enable_cap()
6638 mutex_lock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6639 if (!kvm->created_vcpus) { in kvm_vm_ioctl_enable_cap()
6640 kvm->arch.disable_nx_huge_pages = true; in kvm_vm_ioctl_enable_cap()
6643 mutex_unlock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6646 u64 bus_cycle_ns = cap->args[0]; in kvm_vm_ioctl_enable_cap()
6653 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6659 mutex_lock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6661 r = -ENXIO; in kvm_vm_ioctl_enable_cap()
6662 else if (kvm->created_vcpus) in kvm_vm_ioctl_enable_cap()
6663 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6665 kvm->arch.apic_bus_cycle_ns = bus_cycle_ns; in kvm_vm_ioctl_enable_cap()
6666 mutex_unlock(&kvm->lock); in kvm_vm_ioctl_enable_cap()
6670 r = -EINVAL; in kvm_vm_ioctl_enable_cap()
6684 msr_filter->default_allow = default_allow; in kvm_alloc_msr_filter()
6695 for (i = 0; i < msr_filter->count; i++) in kvm_free_msr_filter()
6696 kfree(msr_filter->ranges[i].bitmap); in kvm_free_msr_filter()
6707 if (!user_range->nmsrs) in kvm_add_msr_filter()
6710 if (user_range->flags & ~KVM_MSR_FILTER_RANGE_VALID_MASK) in kvm_add_msr_filter()
6711 return -EINVAL; in kvm_add_msr_filter()
6713 if (!user_range->flags) in kvm_add_msr_filter()
6714 return -EINVAL; in kvm_add_msr_filter()
6716 bitmap_size = BITS_TO_LONGS(user_range->nmsrs) * sizeof(long); in kvm_add_msr_filter()
6718 return -EINVAL; in kvm_add_msr_filter()
6720 bitmap = memdup_user((__user u8*)user_range->bitmap, bitmap_size); in kvm_add_msr_filter()
6724 msr_filter->ranges[msr_filter->count] = (struct msr_bitmap_range) { in kvm_add_msr_filter()
6725 .flags = user_range->flags, in kvm_add_msr_filter()
6726 .base = user_range->base, in kvm_add_msr_filter()
6727 .nmsrs = user_range->nmsrs, in kvm_add_msr_filter()
6731 msr_filter->count++; in kvm_add_msr_filter()
6744 if (filter->flags & ~KVM_MSR_FILTER_VALID_MASK) in kvm_vm_ioctl_set_msr_filter()
6745 return -EINVAL; in kvm_vm_ioctl_set_msr_filter()
6747 for (i = 0; i < ARRAY_SIZE(filter->ranges); i++) in kvm_vm_ioctl_set_msr_filter()
6748 empty &= !filter->ranges[i].nmsrs; in kvm_vm_ioctl_set_msr_filter()
6750 default_allow = !(filter->flags & KVM_MSR_FILTER_DEFAULT_DENY); in kvm_vm_ioctl_set_msr_filter()
6752 return -EINVAL; in kvm_vm_ioctl_set_msr_filter()
6756 return -ENOMEM; in kvm_vm_ioctl_set_msr_filter()
6758 for (i = 0; i < ARRAY_SIZE(filter->ranges); i++) { in kvm_vm_ioctl_set_msr_filter()
6759 r = kvm_add_msr_filter(new_filter, &filter->ranges[i]); in kvm_vm_ioctl_set_msr_filter()
6766 mutex_lock(&kvm->lock); in kvm_vm_ioctl_set_msr_filter()
6767 old_filter = rcu_replace_pointer(kvm->arch.msr_filter, new_filter, in kvm_vm_ioctl_set_msr_filter()
6768 mutex_is_locked(&kvm->lock)); in kvm_vm_ioctl_set_msr_filter()
6769 mutex_unlock(&kvm->lock); in kvm_vm_ioctl_set_msr_filter()
6770 synchronize_srcu(&kvm->srcu); in kvm_vm_ioctl_set_msr_filter()
6799 struct kvm *kvm = filp->private_data; in kvm_arch_vm_compat_ioctl()
6800 long r = -ENOTTY; in kvm_arch_vm_compat_ioctl()
6811 return -EFAULT; in kvm_arch_vm_compat_ioctl()
6819 .flags = cr->flags, in kvm_arch_vm_compat_ioctl()
6820 .nmsrs = cr->nmsrs, in kvm_arch_vm_compat_ioctl()
6821 .base = cr->base, in kvm_arch_vm_compat_ioctl()
6822 .bitmap = (__u8 *)(ulong)cr->bitmap, in kvm_arch_vm_compat_ioctl()
6869 return -EFAULT; in kvm_vm_ioctl_get_clock()
6876 struct kvm_arch *ka = &kvm->arch; in kvm_vm_ioctl_set_clock()
6881 return -EFAULT; in kvm_vm_ioctl_set_clock()
6888 return -EINVAL; in kvm_vm_ioctl_set_clock()
6908 data.clock += now_real_ns - data.realtime; in kvm_vm_ioctl_set_clock()
6911 if (ka->use_master_clock) in kvm_vm_ioctl_set_clock()
6912 now_raw_ns = ka->master_kernel_ns; in kvm_vm_ioctl_set_clock()
6915 ka->kvmclock_offset = data.clock - now_raw_ns; in kvm_vm_ioctl_set_clock()
6922 struct kvm *kvm = filp->private_data; in kvm_arch_vm_ioctl()
6924 int r = -ENOTTY; in kvm_arch_vm_ioctl()
6928 * This union makes it completely explicit to gcc-3.x in kvm_arch_vm_ioctl()
6946 mutex_lock(&kvm->lock); in kvm_arch_vm_ioctl()
6947 r = -EINVAL; in kvm_arch_vm_ioctl()
6948 if (kvm->created_vcpus) in kvm_arch_vm_ioctl()
6950 r = -EFAULT; in kvm_arch_vm_ioctl()
6955 mutex_unlock(&kvm->lock); in kvm_arch_vm_ioctl()
6963 mutex_lock(&kvm->lock); in kvm_arch_vm_ioctl()
6965 r = -EEXIST; in kvm_arch_vm_ioctl()
6969 r = -EINVAL; in kvm_arch_vm_ioctl()
6970 if (kvm->created_vcpus) in kvm_arch_vm_ioctl()
6989 /* Write kvm->irq_routing before enabling irqchip_in_kernel. */ in kvm_arch_vm_ioctl()
6991 kvm->arch.irqchip_mode = KVM_IRQCHIP_KERNEL; in kvm_arch_vm_ioctl()
6994 mutex_unlock(&kvm->lock); in kvm_arch_vm_ioctl()
7001 r = -EFAULT; in kvm_arch_vm_ioctl()
7006 mutex_lock(&kvm->lock); in kvm_arch_vm_ioctl()
7007 r = -EEXIST; in kvm_arch_vm_ioctl()
7008 if (kvm->arch.vpit) in kvm_arch_vm_ioctl()
7010 r = -ENOENT; in kvm_arch_vm_ioctl()
7013 r = -ENOMEM; in kvm_arch_vm_ioctl()
7014 kvm->arch.vpit = kvm_create_pit(kvm, u.pit_config.flags); in kvm_arch_vm_ioctl()
7015 if (kvm->arch.vpit) in kvm_arch_vm_ioctl()
7018 mutex_unlock(&kvm->lock); in kvm_arch_vm_ioctl()
7030 r = -ENXIO; in kvm_arch_vm_ioctl()
7036 r = -EFAULT; in kvm_arch_vm_ioctl()
7054 r = -ENXIO; in kvm_arch_vm_ioctl()
7063 r = -EFAULT; in kvm_arch_vm_ioctl()
7066 r = -ENXIO; in kvm_arch_vm_ioctl()
7067 if (!kvm->arch.vpit) in kvm_arch_vm_ioctl()
7072 r = -EFAULT; in kvm_arch_vm_ioctl()
7079 r = -EFAULT; in kvm_arch_vm_ioctl()
7082 mutex_lock(&kvm->lock); in kvm_arch_vm_ioctl()
7083 r = -ENXIO; in kvm_arch_vm_ioctl()
7084 if (!kvm->arch.vpit) in kvm_arch_vm_ioctl()
7088 mutex_unlock(&kvm->lock); in kvm_arch_vm_ioctl()
7092 r = -ENXIO; in kvm_arch_vm_ioctl()
7093 if (!kvm->arch.vpit) in kvm_arch_vm_ioctl()
7098 r = -EFAULT; in kvm_arch_vm_ioctl()
7105 r = -EFAULT; in kvm_arch_vm_ioctl()
7108 mutex_lock(&kvm->lock); in kvm_arch_vm_ioctl()
7109 r = -ENXIO; in kvm_arch_vm_ioctl()
7110 if (!kvm->arch.vpit) in kvm_arch_vm_ioctl()
7114 mutex_unlock(&kvm->lock); in kvm_arch_vm_ioctl()
7119 r = -EFAULT; in kvm_arch_vm_ioctl()
7122 r = -ENXIO; in kvm_arch_vm_ioctl()
7123 if (!kvm->arch.vpit) in kvm_arch_vm_ioctl()
7131 mutex_lock(&kvm->lock); in kvm_arch_vm_ioctl()
7132 if (kvm->created_vcpus) in kvm_arch_vm_ioctl()
7133 r = -EBUSY; in kvm_arch_vm_ioctl()
7135 (kvm->arch.max_vcpu_ids && arg > kvm->arch.max_vcpu_ids)) in kvm_arch_vm_ioctl()
7136 r = -EINVAL; in kvm_arch_vm_ioctl()
7138 kvm->arch.bsp_vcpu_id = arg; in kvm_arch_vm_ioctl()
7139 mutex_unlock(&kvm->lock); in kvm_arch_vm_ioctl()
7144 r = -EFAULT; in kvm_arch_vm_ioctl()
7153 r = -EFAULT; in kvm_arch_vm_ioctl()
7158 r = -EFAULT; in kvm_arch_vm_ioctl()
7164 r = -EFAULT; in kvm_arch_vm_ioctl()
7173 r = -EFAULT; in kvm_arch_vm_ioctl()
7189 r = -EINVAL; in kvm_arch_vm_ioctl()
7199 mutex_lock(&kvm->lock); in kvm_arch_vm_ioctl()
7200 if (!kvm->created_vcpus) { in kvm_arch_vm_ioctl()
7201 WRITE_ONCE(kvm->arch.default_tsc_khz, user_tsc_khz); in kvm_arch_vm_ioctl()
7204 mutex_unlock(&kvm->lock); in kvm_arch_vm_ioctl()
7208 r = READ_ONCE(kvm->arch.default_tsc_khz); in kvm_arch_vm_ioctl()
7212 r = -ENOTTY; in kvm_arch_vm_ioctl()
7221 r = -EFAULT; in kvm_arch_vm_ioctl()
7225 r = -ENOTTY; in kvm_arch_vm_ioctl()
7235 r = -EFAULT; in kvm_arch_vm_ioctl()
7239 r = -ENOTTY; in kvm_arch_vm_ioctl()
7250 r = -EFAULT; in kvm_arch_vm_ioctl()
7265 return -EFAULT; in kvm_arch_vm_ioctl()
7271 r = -ENOTTY; in kvm_arch_vm_ioctl()
7331 (msr_index - MSR_IA32_RTIT_ADDR0_A >= in kvm_probe_msr_to_save()
7336 MSR_ARCH_PERFMON_PERFCTR0 + KVM_MAX_NR_GP_COUNTERS - 1: in kvm_probe_msr_to_save()
7337 if (msr_index - MSR_ARCH_PERFMON_PERFCTR0 >= in kvm_probe_msr_to_save()
7342 MSR_ARCH_PERFMON_EVENTSEL0 + KVM_MAX_NR_GP_COUNTERS - 1: in kvm_probe_msr_to_save()
7343 if (msr_index - MSR_ARCH_PERFMON_EVENTSEL0 >= in kvm_probe_msr_to_save()
7348 MSR_ARCH_PERFMON_FIXED_CTR0 + KVM_MAX_NR_FIXED_COUNTERS - 1: in kvm_probe_msr_to_save()
7349 if (msr_index - MSR_ARCH_PERFMON_FIXED_CTR0 >= in kvm_probe_msr_to_save()
7418 !kvm_iodevice_write(vcpu, &vcpu->arch.apic->dev, addr, n, v)) in vcpu_mmio_write()
7423 len -= n; in vcpu_mmio_write()
7438 !kvm_iodevice_read(vcpu, &vcpu->arch.apic->dev, in vcpu_mmio_read()
7445 len -= n; in vcpu_mmio_read()
7467 struct kvm_mmu *mmu = vcpu->arch.mmu; in translate_nested_gpa()
7472 /* NPT walks are always user-walks */ in translate_nested_gpa()
7474 t_gpa = mmu->gva_to_gpa(vcpu, mmu, gpa, access, exception); in translate_nested_gpa()
7482 struct kvm_mmu *mmu = vcpu->arch.walk_mmu; in kvm_mmu_gva_to_gpa_read()
7485 return mmu->gva_to_gpa(vcpu, mmu, gva, access, exception); in kvm_mmu_gva_to_gpa_read()
7492 struct kvm_mmu *mmu = vcpu->arch.walk_mmu; in kvm_mmu_gva_to_gpa_write()
7496 return mmu->gva_to_gpa(vcpu, mmu, gva, access, exception); in kvm_mmu_gva_to_gpa_write()
7504 struct kvm_mmu *mmu = vcpu->arch.walk_mmu; in kvm_mmu_gva_to_gpa_system()
7506 return mmu->gva_to_gpa(vcpu, mmu, gva, 0, exception); in kvm_mmu_gva_to_gpa_system()
7513 struct kvm_mmu *mmu = vcpu->arch.walk_mmu; in kvm_read_guest_virt_helper()
7518 gpa_t gpa = mmu->gva_to_gpa(vcpu, mmu, addr, access, exception); in kvm_read_guest_virt_helper()
7519 unsigned offset = addr & (PAGE_SIZE-1); in kvm_read_guest_virt_helper()
7520 unsigned toread = min(bytes, (unsigned)PAGE_SIZE - offset); in kvm_read_guest_virt_helper()
7532 bytes -= toread; in kvm_read_guest_virt_helper()
7546 struct kvm_mmu *mmu = vcpu->arch.walk_mmu; in kvm_fetch_guest_virt()
7552 gpa_t gpa = mmu->gva_to_gpa(vcpu, mmu, addr, access|PFERR_FETCH_MASK, in kvm_fetch_guest_virt()
7557 offset = addr & (PAGE_SIZE-1); in kvm_fetch_guest_virt()
7559 bytes = (unsigned)PAGE_SIZE - offset; in kvm_fetch_guest_virt()
7605 struct kvm_mmu *mmu = vcpu->arch.walk_mmu; in kvm_write_guest_virt_helper()
7610 gpa_t gpa = mmu->gva_to_gpa(vcpu, mmu, addr, access, exception); in kvm_write_guest_virt_helper()
7611 unsigned offset = addr & (PAGE_SIZE-1); in kvm_write_guest_virt_helper()
7612 unsigned towrite = min(bytes, (unsigned)PAGE_SIZE - offset); in kvm_write_guest_virt_helper()
7623 bytes -= towrite; in kvm_write_guest_virt_helper()
7651 vcpu->arch.l1tf_flush_l1d = true; in kvm_write_guest_virt_system()
7711 struct kvm_mmu *mmu = vcpu->arch.walk_mmu; in vcpu_mmio_gva_to_gpa()
7721 !permission_fault(vcpu, vcpu->arch.walk_mmu, in vcpu_mmio_gva_to_gpa()
7722 vcpu->arch.mmio_access, 0, access))) { in vcpu_mmio_gva_to_gpa()
7723 *gpa = vcpu->arch.mmio_gfn << PAGE_SHIFT | in vcpu_mmio_gva_to_gpa()
7724 (gva & (PAGE_SIZE - 1)); in vcpu_mmio_gva_to_gpa()
7729 *gpa = mmu->gva_to_gpa(vcpu, mmu, gva, access, exception); in vcpu_mmio_gva_to_gpa()
7732 return -1; in vcpu_mmio_gva_to_gpa()
7763 if (vcpu->mmio_read_completed) { in read_prepare()
7765 vcpu->mmio_fragments[0].gpa, val); in read_prepare()
7766 vcpu->mmio_read_completed = 0; in read_prepare()
7801 struct kvm_mmio_fragment *frag = &vcpu->mmio_fragments[0]; in write_exit_mmio()
7803 memcpy(vcpu->run->mmio.data, frag->data, min(8u, frag->len)); in write_exit_mmio()
7829 bool write = ops->write; in emulator_read_write_onepage()
7831 struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt; in emulator_read_write_onepage()
7840 if (ctxt->gpa_available && emulator_can_use_gpa(ctxt) && in emulator_read_write_onepage()
7841 (addr & ~PAGE_MASK) == (ctxt->gpa_val & ~PAGE_MASK)) { in emulator_read_write_onepage()
7842 gpa = ctxt->gpa_val; in emulator_read_write_onepage()
7850 if (!ret && ops->read_write_emulate(vcpu, gpa, val, bytes)) in emulator_read_write_onepage()
7856 handled = ops->read_write_mmio(vcpu, gpa, bytes, val); in emulator_read_write_onepage()
7861 bytes -= handled; in emulator_read_write_onepage()
7864 WARN_ON(vcpu->mmio_nr_fragments >= KVM_MAX_MMIO_FRAGMENTS); in emulator_read_write_onepage()
7865 frag = &vcpu->mmio_fragments[vcpu->mmio_nr_fragments++]; in emulator_read_write_onepage()
7866 frag->gpa = gpa; in emulator_read_write_onepage()
7867 frag->data = val; in emulator_read_write_onepage()
7868 frag->len = bytes; in emulator_read_write_onepage()
7882 if (ops->read_write_prepare && in emulator_read_write()
7883 ops->read_write_prepare(vcpu, val, bytes)) in emulator_read_write()
7886 vcpu->mmio_nr_fragments = 0; in emulator_read_write()
7889 if (((addr + bytes - 1) ^ addr) & PAGE_MASK) { in emulator_read_write()
7892 now = -addr & ~PAGE_MASK; in emulator_read_write()
7899 if (ctxt->mode != X86EMUL_MODE_PROT64) in emulator_read_write()
7902 bytes -= now; in emulator_read_write()
7910 if (!vcpu->mmio_nr_fragments) in emulator_read_write()
7913 gpa = vcpu->mmio_fragments[0].gpa; in emulator_read_write()
7915 vcpu->mmio_needed = 1; in emulator_read_write()
7916 vcpu->mmio_cur_fragment = 0; in emulator_read_write()
7918 vcpu->run->mmio.len = min(8u, vcpu->mmio_fragments[0].len); in emulator_read_write()
7919 vcpu->run->mmio.is_write = vcpu->mmio_is_write = ops->write; in emulator_read_write()
7920 vcpu->run->exit_reason = KVM_EXIT_MMIO; in emulator_read_write()
7921 vcpu->run->mmio.phys_addr = gpa; in emulator_read_write()
7923 return ops->read_write_exit_mmio(vcpu, gpa, val, bytes); in emulator_read_write()
7963 if (bytes > 8 || (bytes & (bytes - 1))) in emulator_cmpxchg_emulated()
7977 page_line_mask = ~(cache_line_size() - 1); in emulator_cmpxchg_emulated()
7981 if (((gpa + bytes - 1) & page_line_mask) != (gpa & page_line_mask)) in emulator_cmpxchg_emulated()
8039 WARN_ON_ONCE(vcpu->arch.pio.count); in emulator_pio_in_out()
8055 memset(data, 0, size * (count - i)); in emulator_pio_in_out()
8064 vcpu->arch.pio.port = port; in emulator_pio_in_out()
8065 vcpu->arch.pio.in = in; in emulator_pio_in_out()
8066 vcpu->arch.pio.count = count; in emulator_pio_in_out()
8067 vcpu->arch.pio.size = size; in emulator_pio_in_out()
8070 memset(vcpu->arch.pio_data, 0, size * count); in emulator_pio_in_out()
8072 memcpy(vcpu->arch.pio_data, data, size * count); in emulator_pio_in_out()
8074 vcpu->run->exit_reason = KVM_EXIT_IO; in emulator_pio_in_out()
8075 vcpu->run->io.direction = in ? KVM_EXIT_IO_IN : KVM_EXIT_IO_OUT; in emulator_pio_in_out()
8076 vcpu->run->io.size = size; in emulator_pio_in_out()
8077 vcpu->run->io.data_offset = KVM_PIO_PAGE_OFFSET * PAGE_SIZE; in emulator_pio_in_out()
8078 vcpu->run->io.count = count; in emulator_pio_in_out()
8079 vcpu->run->io.port = port; in emulator_pio_in_out()
8095 int size = vcpu->arch.pio.size; in complete_emulator_pio_in()
8096 unsigned int count = vcpu->arch.pio.count; in complete_emulator_pio_in()
8097 memcpy(val, vcpu->arch.pio_data, size * count); in complete_emulator_pio_in()
8098 trace_kvm_pio(KVM_PIO_IN, vcpu->arch.pio.port, size, count, vcpu->arch.pio_data); in complete_emulator_pio_in()
8099 vcpu->arch.pio.count = 0; in complete_emulator_pio_in()
8107 if (vcpu->arch.pio.count) { in emulator_pio_in_emulated()
8155 cpumask_set_cpu(cpu, vcpu->arch.wbinvd_dirty_mask); in kvm_emulate_wbinvd_noskip()
8156 wbinvd_on_cpus_mask(vcpu->arch.wbinvd_dirty_mask); in kvm_emulate_wbinvd_noskip()
8158 cpumask_clear(vcpu->arch.wbinvd_dirty_mask); in kvm_emulate_wbinvd_noskip()
8192 return (curr_cr & ~((1ULL << 32) - 1)) | new_val; in mk_cr_64()
8205 value = vcpu->arch.cr2; in emulator_get_cr()
8234 vcpu->arch.cr2 = val; in emulator_set_cr()
8247 res = -1; in emulator_set_cr()
8308 desc->type = var.type; in emulator_get_segment()
8309 desc->s = var.s; in emulator_get_segment()
8310 desc->dpl = var.dpl; in emulator_get_segment()
8311 desc->p = var.present; in emulator_get_segment()
8312 desc->avl = var.avl; in emulator_get_segment()
8313 desc->l = var.l; in emulator_get_segment()
8314 desc->d = var.db; in emulator_get_segment()
8315 desc->g = var.g; in emulator_get_segment()
8333 if (desc->g) in emulator_set_segment()
8335 var.type = desc->type; in emulator_set_segment()
8336 var.dpl = desc->dpl; in emulator_set_segment()
8337 var.db = desc->d; in emulator_set_segment()
8338 var.s = desc->s; in emulator_set_segment()
8339 var.l = desc->l; in emulator_set_segment()
8340 var.g = desc->g; in emulator_set_segment()
8341 var.avl = desc->avl; in emulator_set_segment()
8342 var.present = desc->p; in emulator_set_segment()
8415 emul_to_vcpu(ctxt)->arch.halt_request = 1; in emulator_halt()
8423 &ctxt->exception); in emulator_intercept()
8463 static void emulator_set_nmi_mask(struct x86_emulate_ctxt *ctxt, bool masked) in emulator_set_nmi_mask() argument
8465 kvm_x86_call(set_nmi_mask)(emul_to_vcpu(ctxt), masked); in emulator_set_nmi_mask()
8498 struct kvm *kvm = emul_to_vcpu(ctxt)->kvm; in emulator_vm_bugged()
8500 if (!kvm->vm_bugged) in emulator_vm_bugged()
8590 struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt; in inject_emulated_exception()
8592 if (ctxt->exception.vector == PF_VECTOR) in inject_emulated_exception()
8593 kvm_inject_emulated_page_fault(vcpu, &ctxt->exception); in inject_emulated_exception()
8594 else if (ctxt->exception.error_code_valid) in inject_emulated_exception()
8595 kvm_queue_exception_e(vcpu, ctxt->exception.vector, in inject_emulated_exception()
8596 ctxt->exception.error_code); in inject_emulated_exception()
8598 kvm_queue_exception(vcpu, ctxt->exception.vector); in inject_emulated_exception()
8611 ctxt->vcpu = vcpu; in alloc_emulate_ctxt()
8612 ctxt->ops = &emulate_ops; in alloc_emulate_ctxt()
8613 vcpu->arch.emulate_ctxt = ctxt; in alloc_emulate_ctxt()
8620 struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt; in init_emulate_ctxt()
8625 ctxt->gpa_available = false; in init_emulate_ctxt()
8626 ctxt->eflags = kvm_get_rflags(vcpu); in init_emulate_ctxt()
8627 ctxt->tf = (ctxt->eflags & X86_EFLAGS_TF) != 0; in init_emulate_ctxt()
8629 ctxt->eip = kvm_rip_read(vcpu); in init_emulate_ctxt()
8630 ctxt->mode = (!is_protmode(vcpu)) ? X86EMUL_MODE_REAL : in init_emulate_ctxt()
8631 (ctxt->eflags & X86_EFLAGS_VM) ? X86EMUL_MODE_VM86 : in init_emulate_ctxt()
8635 ctxt->interruptibility = 0; in init_emulate_ctxt()
8636 ctxt->have_exception = false; in init_emulate_ctxt()
8637 ctxt->exception.vector = -1; in init_emulate_ctxt()
8638 ctxt->perm_ok = false; in init_emulate_ctxt()
8641 vcpu->arch.emulate_regs_need_sync_from_vcpu = false; in init_emulate_ctxt()
8646 struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt; in kvm_inject_realmode_interrupt()
8651 ctxt->op_bytes = 2; in kvm_inject_realmode_interrupt()
8652 ctxt->ad_bytes = 2; in kvm_inject_realmode_interrupt()
8653 ctxt->_eip = ctxt->eip + inc_eip; in kvm_inject_realmode_interrupt()
8659 ctxt->eip = ctxt->_eip; in kvm_inject_realmode_interrupt()
8660 kvm_rip_write(vcpu, ctxt->eip); in kvm_inject_realmode_interrupt()
8661 kvm_set_rflags(vcpu, ctxt->eflags); in kvm_inject_realmode_interrupt()
8669 struct kvm_run *run = vcpu->run; in prepare_emulation_failure_exit()
8682 run->exit_reason = KVM_EXIT_INTERNAL_ERROR; in prepare_emulation_failure_exit()
8683 run->emulation_failure.suberror = KVM_INTERNAL_ERROR_EMULATION; in prepare_emulation_failure_exit()
8695 run->emulation_failure.flags = 0; in prepare_emulation_failure_exit()
8698 BUILD_BUG_ON((sizeof(run->emulation_failure.insn_size) + in prepare_emulation_failure_exit()
8699 sizeof(run->emulation_failure.insn_bytes) != 16)); in prepare_emulation_failure_exit()
8701 run->emulation_failure.flags |= in prepare_emulation_failure_exit()
8703 run->emulation_failure.insn_size = insn_size; in prepare_emulation_failure_exit()
8704 memset(run->emulation_failure.insn_bytes, 0x90, in prepare_emulation_failure_exit()
8705 sizeof(run->emulation_failure.insn_bytes)); in prepare_emulation_failure_exit()
8706 memcpy(run->emulation_failure.insn_bytes, insn_bytes, insn_size); in prepare_emulation_failure_exit()
8709 memcpy(&run->internal.data[info_start], info, sizeof(info)); in prepare_emulation_failure_exit()
8710 memcpy(&run->internal.data[info_start + ARRAY_SIZE(info)], data, in prepare_emulation_failure_exit()
8713 run->emulation_failure.ndata = info_start + ARRAY_SIZE(info) + ndata; in prepare_emulation_failure_exit()
8718 struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt; in prepare_emulation_ctxt_failure_exit()
8720 prepare_emulation_failure_exit(vcpu, NULL, 0, ctxt->fetch.data, in prepare_emulation_ctxt_failure_exit()
8721 ctxt->fetch.end - ctxt->fetch.data); in prepare_emulation_ctxt_failure_exit()
8740 struct kvm_run *run = vcpu->run; in kvm_prepare_event_vectoring_exit()
8747 run->internal.data[ndata++] = info2; in kvm_prepare_event_vectoring_exit()
8748 run->internal.data[ndata++] = reason; in kvm_prepare_event_vectoring_exit()
8749 run->internal.data[ndata++] = info1; in kvm_prepare_event_vectoring_exit()
8750 run->internal.data[ndata++] = gpa; in kvm_prepare_event_vectoring_exit()
8751 run->internal.data[ndata++] = vcpu->arch.last_vmentry_cpu; in kvm_prepare_event_vectoring_exit()
8753 run->exit_reason = KVM_EXIT_INTERNAL_ERROR; in kvm_prepare_event_vectoring_exit()
8754 run->internal.suberror = KVM_INTERNAL_ERROR_DELIVERY_EV; in kvm_prepare_event_vectoring_exit()
8755 run->internal.ndata = ndata; in kvm_prepare_event_vectoring_exit()
8761 struct kvm *kvm = vcpu->kvm; in handle_emulation_failure()
8763 ++vcpu->stat.insn_emulation_fail; in handle_emulation_failure()
8771 if (kvm->arch.exit_on_emulation_error || in handle_emulation_failure()
8800 * a SPTE and write-protect the gfn to resolve the !PRESENT fault, and in kvm_unprotect_and_retry_on_failure()
8809 * table, unprotect the gfn (zap any relevant SPTEs) and re-enter the in kvm_unprotect_and_retry_on_failure()
8810 * guest to let the CPU re-execute the instruction in the hope that the in kvm_unprotect_and_retry_on_failure()
8845 struct kvm_run *kvm_run = vcpu->run; in kvm_vcpu_do_singlestep()
8847 if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP) { in kvm_vcpu_do_singlestep()
8848 kvm_run->debug.arch.dr6 = DR6_BS | DR6_ACTIVE_LOW; in kvm_vcpu_do_singlestep()
8849 kvm_run->debug.arch.pc = kvm_get_linear_rip(vcpu); in kvm_vcpu_do_singlestep()
8850 kvm_run->debug.arch.exception = DB_VECTOR; in kvm_vcpu_do_singlestep()
8851 kvm_run->exit_reason = KVM_EXIT_DEBUG; in kvm_vcpu_do_singlestep()
8921 if (unlikely(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP) && in kvm_vcpu_check_code_breakpoint()
8922 (vcpu->arch.guest_debug_dr7 & DR7_BP_EN_MASK)) { in kvm_vcpu_check_code_breakpoint()
8923 struct kvm_run *kvm_run = vcpu->run; in kvm_vcpu_check_code_breakpoint()
8926 vcpu->arch.guest_debug_dr7, in kvm_vcpu_check_code_breakpoint()
8927 vcpu->arch.eff_db); in kvm_vcpu_check_code_breakpoint()
8930 kvm_run->debug.arch.dr6 = dr6 | DR6_ACTIVE_LOW; in kvm_vcpu_check_code_breakpoint()
8931 kvm_run->debug.arch.pc = eip; in kvm_vcpu_check_code_breakpoint()
8932 kvm_run->debug.arch.exception = DB_VECTOR; in kvm_vcpu_check_code_breakpoint()
8933 kvm_run->exit_reason = KVM_EXIT_DEBUG; in kvm_vcpu_check_code_breakpoint()
8939 if (unlikely(vcpu->arch.dr7 & DR7_BP_EN_MASK) && in kvm_vcpu_check_code_breakpoint()
8943 vcpu->arch.dr7, in kvm_vcpu_check_code_breakpoint()
8944 vcpu->arch.db); in kvm_vcpu_check_code_breakpoint()
8958 switch (ctxt->opcode_len) { in is_vmware_backdoor_opcode()
8960 switch (ctxt->b) { in is_vmware_backdoor_opcode()
8977 switch (ctxt->b) { in is_vmware_backdoor_opcode()
8990 * (and wrong) when emulating on an intercepted fault-like exception[*], as
9000 struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt; in x86_decode_emulated_instruction()
9008 ++vcpu->stat.insn_emulation; in x86_decode_emulated_instruction()
9018 struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt; in x86_emulate_instruction()
9044 vcpu->arch.l1tf_flush_l1d = true; in x86_emulate_instruction()
9051 * are fault-like and are higher priority than any faults on in x86_emulate_instruction()
9069 if (ctxt->have_exception && in x86_emulate_instruction()
9072 * #UD should result in just EMULATION_FAILED, and trap-like in x86_emulate_instruction()
9075 WARN_ON_ONCE(ctxt->exception.vector == UD_VECTOR || in x86_emulate_instruction()
9076 exception_type(ctxt->exception.vector) == EXCPT_TRAP); in x86_emulate_instruction()
9094 * injecting single-step #DBs. in x86_emulate_instruction()
9097 if (ctxt->mode != X86EMUL_MODE_PROT64) in x86_emulate_instruction()
9098 ctxt->eip = (u32)ctxt->_eip; in x86_emulate_instruction()
9100 ctxt->eip = ctxt->_eip; in x86_emulate_instruction()
9107 kvm_rip_write(vcpu, ctxt->eip); in x86_emulate_instruction()
9108 if (ctxt->eflags & X86_EFLAGS_RF) in x86_emulate_instruction()
9109 kvm_set_rflags(vcpu, ctxt->eflags & ~X86_EFLAGS_RF); in x86_emulate_instruction()
9114 * If emulation was caused by a write-protection #PF on a non-page_table in x86_emulate_instruction()
9126 if (vcpu->arch.emulate_regs_need_sync_from_vcpu) { in x86_emulate_instruction()
9127 vcpu->arch.emulate_regs_need_sync_from_vcpu = false; in x86_emulate_instruction()
9134 ctxt->exception.address = cr2_or_gpa; in x86_emulate_instruction()
9137 if (vcpu->arch.mmu->root_role.direct) { in x86_emulate_instruction()
9138 ctxt->gpa_available = true; in x86_emulate_instruction()
9139 ctxt->gpa_val = cr2_or_gpa; in x86_emulate_instruction()
9143 ctxt->exception.address = 0; in x86_emulate_instruction()
9159 if (ctxt->have_exception) { in x86_emulate_instruction()
9160 WARN_ON_ONCE(vcpu->mmio_needed && !vcpu->mmio_is_write); in x86_emulate_instruction()
9161 vcpu->mmio_needed = false; in x86_emulate_instruction()
9164 } else if (vcpu->arch.pio.count) { in x86_emulate_instruction()
9165 if (!vcpu->arch.pio.in) { in x86_emulate_instruction()
9166 /* FIXME: return into emulator if single-stepping. */ in x86_emulate_instruction()
9167 vcpu->arch.pio.count = 0; in x86_emulate_instruction()
9170 vcpu->arch.complete_userspace_io = complete_emulated_pio; in x86_emulate_instruction()
9173 } else if (vcpu->mmio_needed) { in x86_emulate_instruction()
9174 ++vcpu->stat.mmio_exits; in x86_emulate_instruction()
9176 if (!vcpu->mmio_is_write) in x86_emulate_instruction()
9179 vcpu->arch.complete_userspace_io = complete_emulated_mmio; in x86_emulate_instruction()
9180 } else if (vcpu->arch.complete_userspace_io) { in x86_emulate_instruction()
9191 toggle_interruptibility(vcpu, ctxt->interruptibility); in x86_emulate_instruction()
9192 vcpu->arch.emulate_regs_need_sync_to_vcpu = false; in x86_emulate_instruction()
9195 * Note, EXCPT_DB is assumed to be fault-like as the emulator in x86_emulate_instruction()
9197 * of which are fault-like. in x86_emulate_instruction()
9199 if (!ctxt->have_exception || in x86_emulate_instruction()
9200 exception_type(ctxt->exception.vector) == EXCPT_TRAP) { in x86_emulate_instruction()
9202 if (ctxt->is_branch) in x86_emulate_instruction()
9204 kvm_rip_write(vcpu, ctxt->eip); in x86_emulate_instruction()
9205 if (r && (ctxt->tf || (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP))) in x86_emulate_instruction()
9208 __kvm_set_rflags(vcpu, ctxt->eflags); in x86_emulate_instruction()
9217 if (unlikely((ctxt->eflags & ~rflags) & X86_EFLAGS_IF)) in x86_emulate_instruction()
9220 vcpu->arch.emulate_regs_need_sync_to_vcpu = true; in x86_emulate_instruction()
9240 vcpu->arch.pio.count = 0; in complete_fast_pio_out_port_0x7e()
9246 vcpu->arch.pio.count = 0; in complete_fast_pio_out()
9248 if (unlikely(!kvm_is_linear_rip(vcpu, vcpu->arch.cui_linear_rip))) in complete_fast_pio_out()
9268 kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_OUT_7E_INC_RIP)) { in kvm_fast_pio_out()
9269 vcpu->arch.complete_userspace_io = in kvm_fast_pio_out()
9273 vcpu->arch.cui_linear_rip = kvm_get_linear_rip(vcpu); in kvm_fast_pio_out()
9274 vcpu->arch.complete_userspace_io = complete_fast_pio_out; in kvm_fast_pio_out()
9284 BUG_ON(vcpu->arch.pio.count != 1); in complete_fast_pio_in()
9286 if (unlikely(!kvm_is_linear_rip(vcpu, vcpu->arch.cui_linear_rip))) { in complete_fast_pio_in()
9287 vcpu->arch.pio.count = 0; in complete_fast_pio_in()
9292 val = (vcpu->arch.pio.size < 4) ? kvm_rax_read(vcpu) : 0; in complete_fast_pio_in()
9315 vcpu->arch.cui_linear_rip = kvm_get_linear_rip(vcpu); in kvm_fast_pio_in()
9316 vcpu->arch.complete_userspace_io = complete_fast_pio_in; in kvm_fast_pio_in()
9347 khz = freq->new; in tsc_khz_changed()
9368 /* TSC frequency always matches when on Hyper-V */ in kvm_hyperv_tsc_notifier()
9436 if (vcpu->cpu != cpu) in __kvmclock_cpufreq_notifier()
9439 if (vcpu->cpu != raw_smp_processor_id()) in __kvmclock_cpufreq_notifier()
9445 if (freq->old < freq->new && send_ipi) { in __kvmclock_cpufreq_notifier()
9468 if (val == CPUFREQ_PRECHANGE && freq->old > freq->new) in kvmclock_cpufreq_notifier()
9470 if (val == CPUFREQ_POSTCHANGE && freq->old < freq->new) in kvmclock_cpufreq_notifier()
9473 for_each_cpu(cpu, freq->policy->cpus) in kvmclock_cpufreq_notifier()
9501 if (policy->cpuinfo.max_freq) in kvm_timer_init()
9502 max_tsc_khz = policy->cpuinfo.max_freq; in kvm_timer_init()
9560 if (!gtod_is_based_on_tsc(gtod->clock.vclock_mode) && in pvclock_gtod_notify()
9573 memcpy(&kvm_x86_ops, ops->runtime_ops, sizeof(kvm_x86_ops)); in kvm_ops_update()
9583 #include <asm/kvm-x86-ops.h> in kvm_ops_update()
9586 kvm_pmu_ops_update(ops->pmu_ops); in kvm_ops_update()
9604 return -EIO; in kvm_x86_check_processor_compatibility()
9623 return -EEXIST; in kvm_x86_vendor_init()
9633 return -EOPNOTSUPP; in kvm_x86_vendor_init()
9638 return -EOPNOTSUPP; in kvm_x86_vendor_init()
9651 return -EIO; in kvm_x86_vendor_init()
9659 return -ENOMEM; in kvm_x86_vendor_init()
9665 r = -ENOMEM; in kvm_x86_vendor_init()
9689 kvm_init_pmu_capability(ops->pmu_ops); in kvm_x86_vendor_init()
9694 r = ops->hardware_setup(); in kvm_x86_vendor_init()
9716 if (pi_inject_timer == -1) in kvm_x86_vendor_init()
9725 kvm_register_perf_callbacks(ops->handle_intel_pt_intr); in kvm_x86_vendor_init()
9809 return -KVM_EOPNOTSUPP; in kvm_pv_clock_pairing()
9815 if (vcpu->arch.tsc_always_catchup) in kvm_pv_clock_pairing()
9816 return -KVM_EOPNOTSUPP; in kvm_pv_clock_pairing()
9819 return -KVM_EOPNOTSUPP; in kvm_pv_clock_pairing()
9828 if (kvm_write_guest(vcpu->kvm, paddr, &clock_pairing, in kvm_pv_clock_pairing()
9830 ret = -KVM_EFAULT; in kvm_pv_clock_pairing()
9839 * @apicid - apicid of vcpu to be kicked.
9859 return (READ_ONCE(kvm->arch.apicv_inhibit_reasons) == 0); in kvm_apicv_activated()
9865 ulong vm_reasons = READ_ONCE(vcpu->kvm->arch.apicv_inhibit_reasons); in kvm_vcpu_apicv_activated()
9893 set_or_clear_apicv_inhibit(&kvm->arch.apicv_inhibit_reasons, reason, true); in kvm_apicv_init()
9895 init_rwsem(&kvm->arch.apicv_update_lock); in kvm_apicv_init()
9903 vcpu->stat.directed_yield_attempted++; in kvm_sched_yield()
9909 map = rcu_dereference(vcpu->kvm->arch.apic_map); in kvm_sched_yield()
9911 if (likely(map) && dest_id <= map->max_apic_id) { in kvm_sched_yield()
9912 dest_id = array_index_nospec(dest_id, map->max_apic_id + 1); in kvm_sched_yield()
9913 if (map->phys_map[dest_id]) in kvm_sched_yield()
9914 target = map->phys_map[dest_id]->vcpu; in kvm_sched_yield()
9919 if (!target || !READ_ONCE(target->ready)) in kvm_sched_yield()
9929 vcpu->stat.directed_yield_successful++; in kvm_sched_yield()
9937 u64 ret = vcpu->run->hypercall.ret; in complete_hypercall_exit()
9956 ++vcpu->stat.hypercalls; in ____kvm_emulate_hypercall()
9969 ret = -KVM_EPERM; in ____kvm_emulate_hypercall()
9973 ret = -KVM_ENOSYS; in ____kvm_emulate_hypercall()
9983 kvm_pv_kick_cpu_op(vcpu->kvm, a1); in ____kvm_emulate_hypercall()
9996 ret = kvm_pv_send_ipi(vcpu->kvm, a0, a1, a2, a3, op_64_bit); in ____kvm_emulate_hypercall()
10008 ret = -KVM_ENOSYS; in ____kvm_emulate_hypercall()
10009 if (!user_exit_on_hypercall(vcpu->kvm, KVM_HC_MAP_GPA_RANGE)) in ____kvm_emulate_hypercall()
10014 ret = -KVM_EINVAL; in ____kvm_emulate_hypercall()
10018 vcpu->run->exit_reason = KVM_EXIT_HYPERCALL; in ____kvm_emulate_hypercall()
10019 vcpu->run->hypercall.nr = KVM_HC_MAP_GPA_RANGE; in ____kvm_emulate_hypercall()
10021 * In principle this should have been -KVM_ENOSYS, but userspace (QEMU <=9.2) in ____kvm_emulate_hypercall()
10022 * assumed that vcpu->run->hypercall.ret is never changed by KVM and thus that in ____kvm_emulate_hypercall()
10024 * vcpu->run->hypercall.ret, ensuring that it is zero to not break QEMU. in ____kvm_emulate_hypercall()
10026 vcpu->run->hypercall.ret = 0; in ____kvm_emulate_hypercall()
10027 vcpu->run->hypercall.args[0] = gpa; in ____kvm_emulate_hypercall()
10028 vcpu->run->hypercall.args[1] = npages; in ____kvm_emulate_hypercall()
10029 vcpu->run->hypercall.args[2] = attrs; in ____kvm_emulate_hypercall()
10030 vcpu->run->hypercall.flags = 0; in ____kvm_emulate_hypercall()
10032 vcpu->run->hypercall.flags |= KVM_EXIT_HYPERCALL_LONG_MODE; in ____kvm_emulate_hypercall()
10034 WARN_ON_ONCE(vcpu->run->hypercall.flags & KVM_EXIT_HYPERCALL_MBZ); in ____kvm_emulate_hypercall()
10035 vcpu->arch.complete_userspace_io = complete_hypercall; in ____kvm_emulate_hypercall()
10039 ret = -KVM_ENOSYS; in ____kvm_emulate_hypercall()
10044 vcpu->run->hypercall.ret = ret; in ____kvm_emulate_hypercall()
10051 if (kvm_xen_hypercall_enabled(vcpu->kvm)) in kvm_emulate_hypercall()
10072 if (!kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_FIX_HYPERCALL_INSN)) { in emulator_fix_hypercall()
10073 ctxt->exception.error_code_valid = false; in emulator_fix_hypercall()
10074 ctxt->exception.vector = UD_VECTOR; in emulator_fix_hypercall()
10075 ctxt->have_exception = true; in emulator_fix_hypercall()
10082 &ctxt->exception); in emulator_fix_hypercall()
10087 return vcpu->run->request_interrupt_window && in dm_request_for_irq_injection()
10088 likely(!pic_in_kernel(vcpu->kvm)); in dm_request_for_irq_injection()
10091 /* Called within kvm->srcu read side. */
10094 struct kvm_run *kvm_run = vcpu->run; in post_kvm_run_save()
10096 kvm_run->if_flag = kvm_x86_call(get_if_flag)(vcpu); in post_kvm_run_save()
10097 kvm_run->cr8 = kvm_get_cr8(vcpu); in post_kvm_run_save()
10098 kvm_run->apic_base = vcpu->arch.apic_base; in post_kvm_run_save()
10100 kvm_run->ready_for_interrupt_injection = in post_kvm_run_save()
10101 pic_in_kernel(vcpu->kvm) || in post_kvm_run_save()
10105 kvm_run->flags |= KVM_RUN_X86_SMM; in post_kvm_run_save()
10107 kvm_run->flags |= KVM_RUN_X86_GUEST_MODE; in post_kvm_run_save()
10120 if (vcpu->arch.apic->apicv_active) in update_cr8_intercept()
10123 if (!vcpu->arch.apic->vapic_addr) in update_cr8_intercept()
10126 max_irr = -1; in update_cr8_intercept()
10128 if (max_irr != -1) in update_cr8_intercept()
10140 kvm_x86_ops.nested_ops->triple_fault(vcpu); in kvm_check_nested_events()
10144 return kvm_x86_ops.nested_ops->check_events(vcpu); in kvm_check_nested_events()
10153 * is injected as intercepted #PF VM-Exits for AMD's Paged Real Mode do in kvm_inject_exception()
10156 vcpu->arch.exception.has_error_code &= is_protmode(vcpu); in kvm_inject_exception()
10158 trace_kvm_inj_exception(vcpu->arch.exception.vector, in kvm_inject_exception()
10159 vcpu->arch.exception.has_error_code, in kvm_inject_exception()
10160 vcpu->arch.exception.error_code, in kvm_inject_exception()
10161 vcpu->arch.exception.injected); in kvm_inject_exception()
10171 * injected as part of a previous VM-Enter, but weren't successfully delivered
10172 * and need to be re-injected.
10177 * also be able to re-inject NMIs and IRQs in the middle of an instruction.
10178 * I.e. for exceptions and re-injected events, NOT invoking this on instruction
10183 * instruction boundaries for asynchronous events. However, because VM-Exits
10189 * But, if a VM-Exit occurs during instruction execution, and KVM does NOT skip
10212 * Process nested events first, as nested VM-Exit supersedes event in kvm_check_and_inject_events()
10213 * re-injection. If there's an event queued for re-injection, it will in kvm_check_and_inject_events()
10214 * be saved into the appropriate vmc{b,s}12 fields on nested VM-Exit. in kvm_check_and_inject_events()
10222 * Re-inject exceptions and events *especially* if immediate entry+exit in kvm_check_and_inject_events()
10226 * Don't re-inject an NMI or interrupt if there is a pending exception. in kvm_check_and_inject_events()
10235 * as the exception "occurred" before the exit to userspace. Trap-like in kvm_check_and_inject_events()
10237 * And while fault-like exceptions, e.g. #GP and #PF, are the lowest in kvm_check_and_inject_events()
10240 * Thus a pending fault-like exception means the fault occurred on the in kvm_check_and_inject_events()
10244 if (vcpu->arch.exception.injected) in kvm_check_and_inject_events()
10248 else if (vcpu->arch.nmi_injected) in kvm_check_and_inject_events()
10250 else if (vcpu->arch.interrupt.injected) in kvm_check_and_inject_events()
10254 * Exceptions that morph to VM-Exits are handled above, and pending in kvm_check_and_inject_events()
10255 * exceptions on top of injected exceptions that do not VM-Exit should in kvm_check_and_inject_events()
10258 WARN_ON_ONCE(vcpu->arch.exception.injected && in kvm_check_and_inject_events()
10259 vcpu->arch.exception.pending); in kvm_check_and_inject_events()
10263 * nested VM-Enter or event re-injection so that a different pending in kvm_check_and_inject_events()
10266 * Otherwise, continue processing events even if VM-Exit occurred. The in kvm_check_and_inject_events()
10267 * VM-Exit will have cleared exceptions that were meant for L2, but in kvm_check_and_inject_events()
10274 * A pending exception VM-Exit should either result in nested VM-Exit in kvm_check_and_inject_events()
10275 * or force an immediate re-entry and exit to/from L2, and exception in kvm_check_and_inject_events()
10276 * VM-Exits cannot be injected (flag should _never_ be set). in kvm_check_and_inject_events()
10278 WARN_ON_ONCE(vcpu->arch.exception_vmexit.injected || in kvm_check_and_inject_events()
10279 vcpu->arch.exception_vmexit.pending); in kvm_check_and_inject_events()
10283 * to re-inject a previous event. See above comments on re-injecting in kvm_check_and_inject_events()
10288 if (vcpu->arch.exception.pending) { in kvm_check_and_inject_events()
10290 * Fault-class exceptions, except #DBs, set RF=1 in the RFLAGS in kvm_check_and_inject_events()
10291 * value pushed on the stack. Trap-like exception and all #DBs in kvm_check_and_inject_events()
10292 * leave RF as-is (KVM follows Intel's behavior in this regard; in kvm_check_and_inject_events()
10297 * fault-like. They do _not_ set RF, a la code breakpoints. in kvm_check_and_inject_events()
10299 if (exception_type(vcpu->arch.exception.vector) == EXCPT_FAULT) in kvm_check_and_inject_events()
10303 if (vcpu->arch.exception.vector == DB_VECTOR) { in kvm_check_and_inject_events()
10304 kvm_deliver_exception_payload(vcpu, &vcpu->arch.exception); in kvm_check_and_inject_events()
10305 if (vcpu->arch.dr7 & DR7_GD) { in kvm_check_and_inject_events()
10306 vcpu->arch.dr7 &= ~DR7_GD; in kvm_check_and_inject_events()
10313 vcpu->arch.exception.pending = false; in kvm_check_and_inject_events()
10314 vcpu->arch.exception.injected = true; in kvm_check_and_inject_events()
10320 if (vcpu->guest_debug & KVM_GUESTDBG_BLOCKIRQ) in kvm_check_and_inject_events()
10325 * due to architectural conditions (e.g. IF=0) a window-open exit in kvm_check_and_inject_events()
10326 * will re-request KVM_REQ_EVENT. Sometimes however an event is pending in kvm_check_and_inject_events()
10332 * The kvm_x86_ops hooks communicate this by returning -EBUSY. in kvm_check_and_inject_events()
10335 if (vcpu->arch.smi_pending) { in kvm_check_and_inject_events()
10337 -EBUSY; in kvm_check_and_inject_events()
10341 vcpu->arch.smi_pending = false; in kvm_check_and_inject_events()
10342 ++vcpu->arch.smi_count; in kvm_check_and_inject_events()
10350 if (vcpu->arch.nmi_pending) { in kvm_check_and_inject_events()
10352 -EBUSY; in kvm_check_and_inject_events()
10356 --vcpu->arch.nmi_pending; in kvm_check_and_inject_events()
10357 vcpu->arch.nmi_injected = true; in kvm_check_and_inject_events()
10362 if (vcpu->arch.nmi_pending) in kvm_check_and_inject_events()
10368 -EBUSY; in kvm_check_and_inject_events()
10374 if (!WARN_ON_ONCE(irq == -1)) { in kvm_check_and_inject_events()
10385 kvm_x86_ops.nested_ops->has_events && in kvm_check_and_inject_events()
10386 kvm_x86_ops.nested_ops->has_events(vcpu, true)) in kvm_check_and_inject_events()
10391 * is done emulating and should only propagate the to-be-injected event in kvm_check_and_inject_events()
10393 * infinite loop as KVM will bail from VM-Enter to inject the pending in kvm_check_and_inject_events()
10401 WARN_ON_ONCE(vcpu->arch.exception.pending || in kvm_check_and_inject_events()
10402 vcpu->arch.exception_vmexit.pending); in kvm_check_and_inject_events()
10406 if (r == -EBUSY) { in kvm_check_and_inject_events()
10427 if (kvm_x86_call(get_nmi_mask)(vcpu) || vcpu->arch.nmi_injected) in process_nmi()
10434 * tracked in vcpu->arch.nmi_pending. in process_nmi()
10437 limit--; in process_nmi()
10439 vcpu->arch.nmi_pending += atomic_xchg(&vcpu->arch.nmi_queued, 0); in process_nmi()
10440 vcpu->arch.nmi_pending = min(vcpu->arch.nmi_pending, limit); in process_nmi()
10442 if (vcpu->arch.nmi_pending && in process_nmi()
10444 vcpu->arch.nmi_pending--; in process_nmi()
10446 if (vcpu->arch.nmi_pending) in process_nmi()
10453 return vcpu->arch.nmi_pending + in kvm_get_nr_pending_nmis()
10470 struct kvm_lapic *apic = vcpu->arch.apic; in __kvm_vcpu_update_apicv()
10476 down_read(&vcpu->kvm->arch.apicv_update_lock); in __kvm_vcpu_update_apicv()
10483 if (apic->apicv_active == activate) in __kvm_vcpu_update_apicv()
10486 apic->apicv_active = activate; in __kvm_vcpu_update_apicv()
10496 if (!apic->apicv_active) in __kvm_vcpu_update_apicv()
10501 up_read(&vcpu->kvm->arch.apicv_update_lock); in __kvm_vcpu_update_apicv()
10521 if (apic_x2apic_mode(vcpu->arch.apic) && in kvm_vcpu_update_apicv()
10533 lockdep_assert_held_write(&kvm->arch.apicv_update_lock); in __kvm_set_or_clear_apicv_inhibit()
10538 old = new = kvm->arch.apicv_inhibit_reasons; in __kvm_set_or_clear_apicv_inhibit()
10556 kvm->arch.apicv_inhibit_reasons = new; in __kvm_set_or_clear_apicv_inhibit()
10559 int idx = srcu_read_lock(&kvm->srcu); in __kvm_set_or_clear_apicv_inhibit()
10562 srcu_read_unlock(&kvm->srcu, idx); in __kvm_set_or_clear_apicv_inhibit()
10565 kvm->arch.apicv_inhibit_reasons = new; in __kvm_set_or_clear_apicv_inhibit()
10575 down_write(&kvm->arch.apicv_update_lock); in kvm_set_or_clear_apicv_inhibit()
10577 up_write(&kvm->arch.apicv_update_lock); in kvm_set_or_clear_apicv_inhibit()
10586 bitmap_zero(vcpu->arch.ioapic_handled_vectors, 256); in vcpu_scan_ioapic()
10587 vcpu->arch.highest_stale_pending_ioapic_eoi = -1; in vcpu_scan_ioapic()
10591 if (irqchip_split(vcpu->kvm)) in vcpu_scan_ioapic()
10592 kvm_scan_ioapic_routes(vcpu, vcpu->arch.ioapic_handled_vectors); in vcpu_scan_ioapic()
10594 else if (ioapic_in_kernel(vcpu->kvm)) in vcpu_scan_ioapic()
10595 kvm_ioapic_scan_entry(vcpu, vcpu->arch.ioapic_handled_vectors); in vcpu_scan_ioapic()
10599 vcpu->arch.load_eoi_exitmap_pending = true; in vcpu_scan_ioapic()
10606 if (!kvm_apic_hw_enabled(vcpu->arch.apic)) in vcpu_load_eoi_exitmap()
10614 vcpu->arch.ioapic_handled_vectors, in vcpu_load_eoi_exitmap()
10615 to_hv_synic(vcpu)->vec_bitmap, 256); in vcpu_load_eoi_exitmap()
10621 vcpu, (u64 *)vcpu->arch.ioapic_handled_vectors); in vcpu_load_eoi_exitmap()
10638 * Called within kvm->srcu read side.
10656 r = -EIO; in vcpu_enter_guest()
10666 if (unlikely(!kvm_x86_ops.nested_ops->get_nested_state_pages(vcpu))) { in vcpu_enter_guest()
10676 kvm_update_masterclock(vcpu->kvm); in vcpu_enter_guest()
10700 * Fall back to a "full" guest flush if Hyper-V's precise in vcpu_enter_guest()
10701 * flushing fails. Note, Hyper-V's flushing is per-vCPU, but in vcpu_enter_guest()
10712 vcpu->run->exit_reason = KVM_EXIT_TPR_ACCESS; in vcpu_enter_guest()
10718 kvm_x86_ops.nested_ops->triple_fault(vcpu); in vcpu_enter_guest()
10721 vcpu->run->exit_reason = KVM_EXIT_SHUTDOWN; in vcpu_enter_guest()
10722 vcpu->mmio_needed = 0; in vcpu_enter_guest()
10729 vcpu->arch.apf.halted = true; in vcpu_enter_guest()
10746 BUG_ON(vcpu->arch.pending_ioapic_eoi > 255); in vcpu_enter_guest()
10747 if (test_bit(vcpu->arch.pending_ioapic_eoi, in vcpu_enter_guest()
10748 vcpu->arch.ioapic_handled_vectors)) { in vcpu_enter_guest()
10749 vcpu->run->exit_reason = KVM_EXIT_IOAPIC_EOI; in vcpu_enter_guest()
10750 vcpu->run->eoi.vector = in vcpu_enter_guest()
10751 vcpu->arch.pending_ioapic_eoi; in vcpu_enter_guest()
10764 vcpu->run->exit_reason = KVM_EXIT_SYSTEM_EVENT; in vcpu_enter_guest()
10765 vcpu->run->system_event.type = KVM_SYSTEM_EVENT_CRASH; in vcpu_enter_guest()
10766 vcpu->run->system_event.ndata = 0; in vcpu_enter_guest()
10771 vcpu->run->exit_reason = KVM_EXIT_SYSTEM_EVENT; in vcpu_enter_guest()
10772 vcpu->run->system_event.type = KVM_SYSTEM_EVENT_RESET; in vcpu_enter_guest()
10773 vcpu->run->system_event.ndata = 0; in vcpu_enter_guest()
10780 vcpu->run->exit_reason = KVM_EXIT_HYPERV; in vcpu_enter_guest()
10781 vcpu->run->hyperv = hv_vcpu->exit; in vcpu_enter_guest()
10788 * KVM_REQ_CLOCK_UPDATE, because Hyper-V SynIC timers in vcpu_enter_guest()
10789 * depend on the guest clock being up-to-date in vcpu_enter_guest()
10812 if (vcpu->arch.mp_state != KVM_MP_STATE_RUNNABLE) { in vcpu_enter_guest()
10821 ++vcpu->stat.req_event; in vcpu_enter_guest()
10827 if (vcpu->arch.mp_state == KVM_MP_STATE_INIT_RECEIVED) { in vcpu_enter_guest()
10862 /* Store vcpu->apicv_active before vcpu->mode. */ in vcpu_enter_guest()
10863 smp_store_release(&vcpu->mode, IN_GUEST_MODE); in vcpu_enter_guest()
10868 * 1) We should set ->mode before checking ->requests. Please see in vcpu_enter_guest()
10871 * 2) For APICv, we should set ->mode before checking PID.ON. This in vcpu_enter_guest()
10892 vcpu->mode = OUTSIDE_GUEST_MODE; in vcpu_enter_guest()
10911 if (vcpu->arch.guest_fpu.xfd_err) in vcpu_enter_guest()
10912 wrmsrq(MSR_IA32_XFD_ERR, vcpu->arch.guest_fpu.xfd_err); in vcpu_enter_guest()
10914 if (unlikely(vcpu->arch.switch_db_regs && in vcpu_enter_guest()
10915 !(vcpu->arch.switch_db_regs & KVM_DEBUGREG_AUTO_SWITCH))) { in vcpu_enter_guest()
10917 set_debugreg(vcpu->arch.eff_db[0], 0); in vcpu_enter_guest()
10918 set_debugreg(vcpu->arch.eff_db[1], 1); in vcpu_enter_guest()
10919 set_debugreg(vcpu->arch.eff_db[2], 2); in vcpu_enter_guest()
10920 set_debugreg(vcpu->arch.eff_db[3], 3); in vcpu_enter_guest()
10922 if (unlikely(vcpu->arch.switch_db_regs & KVM_DEBUGREG_WONT_EXIT)) in vcpu_enter_guest()
10931 * vendor code if any host-owned bits were changed, e.g. so that the in vcpu_enter_guest()
10935 if ((debug_ctl ^ vcpu->arch.host_debugctl) & kvm_x86_ops.HOST_OWNED_DEBUGCTL && in vcpu_enter_guest()
10936 !vcpu->arch.guest_state_protected) in vcpu_enter_guest()
10938 vcpu->arch.host_debugctl = debug_ctl; in vcpu_enter_guest()
10946 * per-VM state, and responding vCPUs must wait for the update in vcpu_enter_guest()
10966 /* Note, VM-Exits that go down the "slow" path are accounted below. */ in vcpu_enter_guest()
10967 ++vcpu->stat.exits; in vcpu_enter_guest()
10976 if (unlikely(vcpu->arch.switch_db_regs & KVM_DEBUGREG_WONT_EXIT)) { in vcpu_enter_guest()
10977 WARN_ON(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP); in vcpu_enter_guest()
10978 WARN_ON(vcpu->arch.switch_db_regs & KVM_DEBUGREG_AUTO_SWITCH); in vcpu_enter_guest()
10994 vcpu->arch.last_vmentry_cpu = vcpu->cpu; in vcpu_enter_guest()
10995 vcpu->arch.last_guest_tsc = kvm_read_l1_tsc(vcpu, rdtsc()); in vcpu_enter_guest()
10997 vcpu->mode = OUTSIDE_GUEST_MODE; in vcpu_enter_guest()
11002 * rely on the fact that guest_fpu::xfd is up-to-date (e.g. in vcpu_enter_guest()
11005 if (vcpu->arch.xfd_no_write_intercept) in vcpu_enter_guest()
11010 if (vcpu->arch.guest_fpu.xfd_err) in vcpu_enter_guest()
11015 * VM-Exit on SVM and any ticks that occur between VM-Exit and now. in vcpu_enter_guest()
11022 ++vcpu->stat.exits; in vcpu_enter_guest()
11050 !vcpu->arch.guest_state_protected)) { in vcpu_enter_guest()
11055 if (unlikely(vcpu->arch.tsc_always_catchup)) in vcpu_enter_guest()
11058 if (vcpu->arch.apic_attention) in vcpu_enter_guest()
11071 if (unlikely(vcpu->arch.apic_attention)) in vcpu_enter_guest()
11079 return (vcpu->arch.mp_state == KVM_MP_STATE_RUNNABLE && in kvm_vcpu_running()
11080 !vcpu->arch.apf.halted); in kvm_vcpu_running()
11085 if (!list_empty_careful(&vcpu->async_pf.done)) in kvm_vcpu_has_events()
11096 (vcpu->arch.nmi_pending && in kvm_vcpu_has_events()
11102 (vcpu->arch.smi_pending && in kvm_vcpu_has_events()
11120 kvm_x86_ops.nested_ops->has_events && in kvm_vcpu_has_events()
11121 kvm_x86_ops.nested_ops->has_events(vcpu, false)) in kvm_vcpu_has_events()
11133 return kvm_vcpu_running(vcpu) || vcpu->arch.pv.pv_unhalted || in kvm_arch_vcpu_runnable()
11137 /* Called within kvm->srcu read side. */
11144 * Switch to the software timer before halt-polling/blocking as in vcpu_block()
11147 * Switch before halt-polling so that KVM recognizes an expired in vcpu_block()
11155 if (vcpu->arch.mp_state == KVM_MP_STATE_HALTED) in vcpu_block()
11176 * state field (AMD does not have a similar field and a VM-Exit always in vcpu_block()
11182 WARN_ON_ONCE(r == -EBUSY); in vcpu_block()
11189 switch(vcpu->arch.mp_state) { in vcpu_block()
11195 vcpu->arch.apf.halted = false; in vcpu_block()
11206 /* Called within kvm->srcu read side. */
11211 vcpu->run->exit_reason = KVM_EXIT_UNKNOWN; in vcpu_run()
11220 vcpu->arch.at_instruction_boundary = false; in vcpu_run()
11240 vcpu->run->exit_reason = KVM_EXIT_IRQ_WINDOW_OPEN; in vcpu_run()
11241 ++vcpu->stat.request_irq_exits; in vcpu_run()
11261 * local APIC is in-kernel, the run loop will detect the non-runnable in __kvm_emulate_halt()
11266 ++vcpu->stat.halt_exits; in __kvm_emulate_halt()
11268 if (kvm_vcpu_has_events(vcpu) || vcpu->arch.pv.pv_unhalted) in __kvm_emulate_halt()
11273 vcpu->run->exit_reason = reason; in __kvm_emulate_halt()
11288 * TODO: we might be squashing a GUESTDBG_SINGLESTEP-triggered in kvm_emulate_halt()
11330 return vcpu->arch.preempted_in_kernel; in kvm_arch_vcpu_preempted_in_kernel()
11335 if (READ_ONCE(vcpu->arch.pv.pv_unhalted)) in kvm_arch_dy_runnable()
11355 BUG_ON(!vcpu->arch.pio.count); in complete_emulated_pio()
11380 struct kvm_run *run = vcpu->run; in complete_emulated_mmio()
11384 BUG_ON(!vcpu->mmio_needed); in complete_emulated_mmio()
11387 frag = &vcpu->mmio_fragments[vcpu->mmio_cur_fragment]; in complete_emulated_mmio()
11388 len = min(8u, frag->len); in complete_emulated_mmio()
11389 if (!vcpu->mmio_is_write) in complete_emulated_mmio()
11390 memcpy(frag->data, run->mmio.data, len); in complete_emulated_mmio()
11392 if (frag->len <= 8) { in complete_emulated_mmio()
11395 vcpu->mmio_cur_fragment++; in complete_emulated_mmio()
11398 frag->data += len; in complete_emulated_mmio()
11399 frag->gpa += len; in complete_emulated_mmio()
11400 frag->len -= len; in complete_emulated_mmio()
11403 if (vcpu->mmio_cur_fragment >= vcpu->mmio_nr_fragments) { in complete_emulated_mmio()
11404 vcpu->mmio_needed = 0; in complete_emulated_mmio()
11406 /* FIXME: return into emulator if single-stepping. */ in complete_emulated_mmio()
11407 if (vcpu->mmio_is_write) in complete_emulated_mmio()
11409 vcpu->mmio_read_completed = 1; in complete_emulated_mmio()
11413 run->exit_reason = KVM_EXIT_MMIO; in complete_emulated_mmio()
11414 run->mmio.phys_addr = frag->gpa; in complete_emulated_mmio()
11415 if (vcpu->mmio_is_write) in complete_emulated_mmio()
11416 memcpy(run->mmio.data, frag->data, min(8u, frag->len)); in complete_emulated_mmio()
11417 run->mmio.len = min(8u, frag->len); in complete_emulated_mmio()
11418 run->mmio.is_write = vcpu->mmio_is_write; in complete_emulated_mmio()
11419 vcpu->arch.complete_userspace_io = complete_emulated_mmio; in complete_emulated_mmio()
11426 /* Exclude PKRU, it's restored separately immediately after VM-Exit. */ in kvm_load_guest_fpu()
11427 fpu_swap_kvm_fpstate(&vcpu->arch.guest_fpu, true); in kvm_load_guest_fpu()
11434 fpu_swap_kvm_fpstate(&vcpu->arch.guest_fpu, false); in kvm_put_guest_fpu()
11435 ++vcpu->stat.fpu_reload; in kvm_put_guest_fpu()
11442 * SIPI_RECEIVED is obsolete; KVM leaves the vCPU in Wait-For-SIPI and in kvm_x86_vcpu_pre_run()
11447 if (WARN_ON_ONCE(vcpu->arch.mp_state == KVM_MP_STATE_SIPI_RECEIVED)) in kvm_x86_vcpu_pre_run()
11448 return -EINVAL; in kvm_x86_vcpu_pre_run()
11454 if (vcpu->arch.mp_state == KVM_MP_STATE_INIT_RECEIVED && in kvm_x86_vcpu_pre_run()
11456 return -EINVAL; in kvm_x86_vcpu_pre_run()
11463 struct kvm_queued_exception *ex = &vcpu->arch.exception; in kvm_arch_vcpu_ioctl_run()
11464 struct kvm_run *kvm_run = vcpu->run; in kvm_arch_vcpu_ioctl_run()
11468 r = kvm_mmu_post_init_vm(vcpu->kvm); in kvm_arch_vcpu_ioctl_run()
11474 kvm_run->flags = 0; in kvm_arch_vcpu_ioctl_run()
11478 if (unlikely(vcpu->arch.mp_state == KVM_MP_STATE_UNINITIALIZED)) { in kvm_arch_vcpu_ioctl_run()
11479 if (!vcpu->wants_to_run) { in kvm_arch_vcpu_ioctl_run()
11480 r = -EINTR; in kvm_arch_vcpu_ioctl_run()
11502 r = -EAGAIN; in kvm_arch_vcpu_ioctl_run()
11504 r = -EINTR; in kvm_arch_vcpu_ioctl_run()
11505 kvm_run->exit_reason = KVM_EXIT_INTR; in kvm_arch_vcpu_ioctl_run()
11506 ++vcpu->stat.signal_exits; in kvm_arch_vcpu_ioctl_run()
11511 sync_valid_fields = kvm_sync_valid_fields(vcpu->kvm); in kvm_arch_vcpu_ioctl_run()
11512 if ((kvm_run->kvm_valid_regs & ~sync_valid_fields) || in kvm_arch_vcpu_ioctl_run()
11513 (kvm_run->kvm_dirty_regs & ~sync_valid_fields)) { in kvm_arch_vcpu_ioctl_run()
11514 r = -EINVAL; in kvm_arch_vcpu_ioctl_run()
11518 if (kvm_run->kvm_dirty_regs) { in kvm_arch_vcpu_ioctl_run()
11524 /* re-sync apic's tpr */ in kvm_arch_vcpu_ioctl_run()
11526 if (kvm_set_cr8(vcpu, kvm_run->cr8) != 0) { in kvm_arch_vcpu_ioctl_run()
11527 r = -EINVAL; in kvm_arch_vcpu_ioctl_run()
11534 * a pending VM-Exit if L1 wants to intercept the exception. in kvm_arch_vcpu_ioctl_run()
11536 if (vcpu->arch.exception_from_userspace && is_guest_mode(vcpu) && in kvm_arch_vcpu_ioctl_run()
11537 kvm_x86_ops.nested_ops->is_exception_vmexit(vcpu, ex->vector, in kvm_arch_vcpu_ioctl_run()
11538 ex->error_code)) { in kvm_arch_vcpu_ioctl_run()
11539 kvm_queue_exception_vmexit(vcpu, ex->vector, in kvm_arch_vcpu_ioctl_run()
11540 ex->has_error_code, ex->error_code, in kvm_arch_vcpu_ioctl_run()
11541 ex->has_payload, ex->payload); in kvm_arch_vcpu_ioctl_run()
11542 ex->injected = false; in kvm_arch_vcpu_ioctl_run()
11543 ex->pending = false; in kvm_arch_vcpu_ioctl_run()
11545 vcpu->arch.exception_from_userspace = false; in kvm_arch_vcpu_ioctl_run()
11547 if (unlikely(vcpu->arch.complete_userspace_io)) { in kvm_arch_vcpu_ioctl_run()
11548 int (*cui)(struct kvm_vcpu *) = vcpu->arch.complete_userspace_io; in kvm_arch_vcpu_ioctl_run()
11549 vcpu->arch.complete_userspace_io = NULL; in kvm_arch_vcpu_ioctl_run()
11554 WARN_ON_ONCE(vcpu->arch.pio.count); in kvm_arch_vcpu_ioctl_run()
11555 WARN_ON_ONCE(vcpu->mmio_needed); in kvm_arch_vcpu_ioctl_run()
11558 if (!vcpu->wants_to_run) { in kvm_arch_vcpu_ioctl_run()
11559 r = -EINTR; in kvm_arch_vcpu_ioctl_run()
11571 if (kvm_run->kvm_valid_regs && likely(!vcpu->arch.guest_state_protected)) in kvm_arch_vcpu_ioctl_run()
11583 if (vcpu->arch.emulate_regs_need_sync_to_vcpu) { in __get_regs()
11591 emulator_writeback_register_cache(vcpu->arch.emulate_ctxt); in __get_regs()
11592 vcpu->arch.emulate_regs_need_sync_to_vcpu = false; in __get_regs()
11594 regs->rax = kvm_rax_read(vcpu); in __get_regs()
11595 regs->rbx = kvm_rbx_read(vcpu); in __get_regs()
11596 regs->rcx = kvm_rcx_read(vcpu); in __get_regs()
11597 regs->rdx = kvm_rdx_read(vcpu); in __get_regs()
11598 regs->rsi = kvm_rsi_read(vcpu); in __get_regs()
11599 regs->rdi = kvm_rdi_read(vcpu); in __get_regs()
11600 regs->rsp = kvm_rsp_read(vcpu); in __get_regs()
11601 regs->rbp = kvm_rbp_read(vcpu); in __get_regs()
11603 regs->r8 = kvm_r8_read(vcpu); in __get_regs()
11604 regs->r9 = kvm_r9_read(vcpu); in __get_regs()
11605 regs->r10 = kvm_r10_read(vcpu); in __get_regs()
11606 regs->r11 = kvm_r11_read(vcpu); in __get_regs()
11607 regs->r12 = kvm_r12_read(vcpu); in __get_regs()
11608 regs->r13 = kvm_r13_read(vcpu); in __get_regs()
11609 regs->r14 = kvm_r14_read(vcpu); in __get_regs()
11610 regs->r15 = kvm_r15_read(vcpu); in __get_regs()
11613 regs->rip = kvm_rip_read(vcpu); in __get_regs()
11614 regs->rflags = kvm_get_rflags(vcpu); in __get_regs()
11619 if (vcpu->kvm->arch.has_protected_state && in kvm_arch_vcpu_ioctl_get_regs()
11620 vcpu->arch.guest_state_protected) in kvm_arch_vcpu_ioctl_get_regs()
11621 return -EINVAL; in kvm_arch_vcpu_ioctl_get_regs()
11631 vcpu->arch.emulate_regs_need_sync_from_vcpu = true; in __set_regs()
11632 vcpu->arch.emulate_regs_need_sync_to_vcpu = false; in __set_regs()
11634 kvm_rax_write(vcpu, regs->rax); in __set_regs()
11635 kvm_rbx_write(vcpu, regs->rbx); in __set_regs()
11636 kvm_rcx_write(vcpu, regs->rcx); in __set_regs()
11637 kvm_rdx_write(vcpu, regs->rdx); in __set_regs()
11638 kvm_rsi_write(vcpu, regs->rsi); in __set_regs()
11639 kvm_rdi_write(vcpu, regs->rdi); in __set_regs()
11640 kvm_rsp_write(vcpu, regs->rsp); in __set_regs()
11641 kvm_rbp_write(vcpu, regs->rbp); in __set_regs()
11643 kvm_r8_write(vcpu, regs->r8); in __set_regs()
11644 kvm_r9_write(vcpu, regs->r9); in __set_regs()
11645 kvm_r10_write(vcpu, regs->r10); in __set_regs()
11646 kvm_r11_write(vcpu, regs->r11); in __set_regs()
11647 kvm_r12_write(vcpu, regs->r12); in __set_regs()
11648 kvm_r13_write(vcpu, regs->r13); in __set_regs()
11649 kvm_r14_write(vcpu, regs->r14); in __set_regs()
11650 kvm_r15_write(vcpu, regs->r15); in __set_regs()
11653 kvm_rip_write(vcpu, regs->rip); in __set_regs()
11654 kvm_set_rflags(vcpu, regs->rflags | X86_EFLAGS_FIXED); in __set_regs()
11656 vcpu->arch.exception.pending = false; in __set_regs()
11657 vcpu->arch.exception_vmexit.pending = false; in __set_regs()
11664 if (vcpu->kvm->arch.has_protected_state && in kvm_arch_vcpu_ioctl_set_regs()
11665 vcpu->arch.guest_state_protected) in kvm_arch_vcpu_ioctl_set_regs()
11666 return -EINVAL; in kvm_arch_vcpu_ioctl_set_regs()
11678 if (vcpu->arch.guest_state_protected) in __get_sregs_common()
11681 kvm_get_segment(vcpu, &sregs->cs, VCPU_SREG_CS); in __get_sregs_common()
11682 kvm_get_segment(vcpu, &sregs->ds, VCPU_SREG_DS); in __get_sregs_common()
11683 kvm_get_segment(vcpu, &sregs->es, VCPU_SREG_ES); in __get_sregs_common()
11684 kvm_get_segment(vcpu, &sregs->fs, VCPU_SREG_FS); in __get_sregs_common()
11685 kvm_get_segment(vcpu, &sregs->gs, VCPU_SREG_GS); in __get_sregs_common()
11686 kvm_get_segment(vcpu, &sregs->ss, VCPU_SREG_SS); in __get_sregs_common()
11688 kvm_get_segment(vcpu, &sregs->tr, VCPU_SREG_TR); in __get_sregs_common()
11689 kvm_get_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR); in __get_sregs_common()
11692 sregs->idt.limit = dt.size; in __get_sregs_common()
11693 sregs->idt.base = dt.address; in __get_sregs_common()
11695 sregs->gdt.limit = dt.size; in __get_sregs_common()
11696 sregs->gdt.base = dt.address; in __get_sregs_common()
11698 sregs->cr2 = vcpu->arch.cr2; in __get_sregs_common()
11699 sregs->cr3 = kvm_read_cr3(vcpu); in __get_sregs_common()
11702 sregs->cr0 = kvm_read_cr0(vcpu); in __get_sregs_common()
11703 sregs->cr4 = kvm_read_cr4(vcpu); in __get_sregs_common()
11704 sregs->cr8 = kvm_get_cr8(vcpu); in __get_sregs_common()
11705 sregs->efer = vcpu->arch.efer; in __get_sregs_common()
11706 sregs->apic_base = vcpu->arch.apic_base; in __get_sregs_common()
11713 if (vcpu->arch.guest_state_protected) in __get_sregs()
11716 if (vcpu->arch.interrupt.injected && !vcpu->arch.interrupt.soft) in __get_sregs()
11717 set_bit(vcpu->arch.interrupt.nr, in __get_sregs()
11718 (unsigned long *)sregs->interrupt_bitmap); in __get_sregs()
11727 if (vcpu->arch.guest_state_protected) in __get_sregs2()
11732 sregs2->pdptrs[i] = kvm_pdptr_read(vcpu, i); in __get_sregs2()
11733 sregs2->flags |= KVM_SREGS2_FLAGS_PDPTRS_VALID; in __get_sregs2()
11740 if (vcpu->kvm->arch.has_protected_state && in kvm_arch_vcpu_ioctl_get_sregs()
11741 vcpu->arch.guest_state_protected) in kvm_arch_vcpu_ioctl_get_sregs()
11742 return -EINVAL; in kvm_arch_vcpu_ioctl_get_sregs()
11766 if ((vcpu->arch.mp_state == KVM_MP_STATE_HALTED || in kvm_arch_vcpu_ioctl_get_mpstate()
11767 vcpu->arch.mp_state == KVM_MP_STATE_AP_RESET_HOLD) && in kvm_arch_vcpu_ioctl_get_mpstate()
11768 vcpu->arch.pv.pv_unhalted) in kvm_arch_vcpu_ioctl_get_mpstate()
11769 mp_state->mp_state = KVM_MP_STATE_RUNNABLE; in kvm_arch_vcpu_ioctl_get_mpstate()
11771 mp_state->mp_state = vcpu->arch.mp_state; in kvm_arch_vcpu_ioctl_get_mpstate()
11785 int ret = -EINVAL; in kvm_arch_vcpu_ioctl_set_mpstate()
11789 switch (mp_state->mp_state) { in kvm_arch_vcpu_ioctl_set_mpstate()
11808 * leaves the vCPU in INIT_RECIEVED (Wait-For-SIPI) and pends the SIPI. in kvm_arch_vcpu_ioctl_set_mpstate()
11811 if (mp_state->mp_state == KVM_MP_STATE_SIPI_RECEIVED) { in kvm_arch_vcpu_ioctl_set_mpstate()
11812 mp_state->mp_state = KVM_MP_STATE_INIT_RECEIVED; in kvm_arch_vcpu_ioctl_set_mpstate()
11813 set_bit(KVM_APIC_SIPI, &vcpu->arch.apic->pending_events); in kvm_arch_vcpu_ioctl_set_mpstate()
11816 kvm_set_mp_state(vcpu, mp_state->mp_state); in kvm_arch_vcpu_ioctl_set_mpstate()
11828 struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt; in kvm_task_switch()
11840 if (ret || vcpu->mmio_needed) { in kvm_task_switch()
11841 vcpu->mmio_needed = false; in kvm_task_switch()
11842 vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; in kvm_task_switch()
11843 vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_EMULATION; in kvm_task_switch()
11844 vcpu->run->internal.ndata = 0; in kvm_task_switch()
11848 kvm_rip_write(vcpu, ctxt->eip); in kvm_task_switch()
11849 kvm_set_rflags(vcpu, ctxt->eflags); in kvm_task_switch()
11856 if ((sregs->efer & EFER_LME) && (sregs->cr0 & X86_CR0_PG)) { in kvm_is_valid_sregs()
11859 * 64-bit mode (though maybe in a 32-bit code segment). in kvm_is_valid_sregs()
11862 if (!(sregs->cr4 & X86_CR4_PAE) || !(sregs->efer & EFER_LMA)) in kvm_is_valid_sregs()
11864 if (!kvm_vcpu_is_legal_cr3(vcpu, sregs->cr3)) in kvm_is_valid_sregs()
11868 * Not in 64-bit mode: EFER.LMA is clear and the code in kvm_is_valid_sregs()
11869 * segment cannot be 64-bit. in kvm_is_valid_sregs()
11871 if (sregs->efer & EFER_LMA || sregs->cs.l) in kvm_is_valid_sregs()
11875 return kvm_is_valid_cr4(vcpu, sregs->cr4) && in kvm_is_valid_sregs()
11876 kvm_is_valid_cr0(vcpu, sregs->cr0); in kvm_is_valid_sregs()
11886 return -EINVAL; in __set_sregs_common()
11888 if (kvm_apic_set_base(vcpu, sregs->apic_base, true)) in __set_sregs_common()
11889 return -EINVAL; in __set_sregs_common()
11891 if (vcpu->arch.guest_state_protected) in __set_sregs_common()
11894 dt.size = sregs->idt.limit; in __set_sregs_common()
11895 dt.address = sregs->idt.base; in __set_sregs_common()
11897 dt.size = sregs->gdt.limit; in __set_sregs_common()
11898 dt.address = sregs->gdt.base; in __set_sregs_common()
11901 vcpu->arch.cr2 = sregs->cr2; in __set_sregs_common()
11902 *mmu_reset_needed |= kvm_read_cr3(vcpu) != sregs->cr3; in __set_sregs_common()
11903 vcpu->arch.cr3 = sregs->cr3; in __set_sregs_common()
11905 kvm_x86_call(post_set_cr3)(vcpu, sregs->cr3); in __set_sregs_common()
11907 kvm_set_cr8(vcpu, sregs->cr8); in __set_sregs_common()
11909 *mmu_reset_needed |= vcpu->arch.efer != sregs->efer; in __set_sregs_common()
11910 kvm_x86_call(set_efer)(vcpu, sregs->efer); in __set_sregs_common()
11912 *mmu_reset_needed |= kvm_read_cr0(vcpu) != sregs->cr0; in __set_sregs_common()
11913 kvm_x86_call(set_cr0)(vcpu, sregs->cr0); in __set_sregs_common()
11915 *mmu_reset_needed |= kvm_read_cr4(vcpu) != sregs->cr4; in __set_sregs_common()
11916 kvm_x86_call(set_cr4)(vcpu, sregs->cr4); in __set_sregs_common()
11919 idx = srcu_read_lock(&vcpu->kvm->srcu); in __set_sregs_common()
11924 srcu_read_unlock(&vcpu->kvm->srcu, idx); in __set_sregs_common()
11927 kvm_set_segment(vcpu, &sregs->cs, VCPU_SREG_CS); in __set_sregs_common()
11928 kvm_set_segment(vcpu, &sregs->ds, VCPU_SREG_DS); in __set_sregs_common()
11929 kvm_set_segment(vcpu, &sregs->es, VCPU_SREG_ES); in __set_sregs_common()
11930 kvm_set_segment(vcpu, &sregs->fs, VCPU_SREG_FS); in __set_sregs_common()
11931 kvm_set_segment(vcpu, &sregs->gs, VCPU_SREG_GS); in __set_sregs_common()
11932 kvm_set_segment(vcpu, &sregs->ss, VCPU_SREG_SS); in __set_sregs_common()
11934 kvm_set_segment(vcpu, &sregs->tr, VCPU_SREG_TR); in __set_sregs_common()
11935 kvm_set_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR); in __set_sregs_common()
11941 sregs->cs.selector == 0xf000 && sregs->cs.base == 0xffff0000 && in __set_sregs_common()
11964 (const unsigned long *)sregs->interrupt_bitmap, max_bits); in __set_sregs()
11977 bool valid_pdptrs = sregs2->flags & KVM_SREGS2_FLAGS_PDPTRS_VALID; in __set_sregs2()
11978 bool pae = (sregs2->cr0 & X86_CR0_PG) && (sregs2->cr4 & X86_CR4_PAE) && in __set_sregs2()
11979 !(sregs2->efer & EFER_LMA); in __set_sregs2()
11982 if (sregs2->flags & ~KVM_SREGS2_FLAGS_PDPTRS_VALID) in __set_sregs2()
11983 return -EINVAL; in __set_sregs2()
11985 if (valid_pdptrs && (!pae || vcpu->arch.guest_state_protected)) in __set_sregs2()
11986 return -EINVAL; in __set_sregs2()
11995 kvm_pdptr_write(vcpu, i, sregs2->pdptrs[i]); in __set_sregs2()
11999 vcpu->arch.pdptrs_from_userspace = true; in __set_sregs2()
12013 if (vcpu->kvm->arch.has_protected_state && in kvm_arch_vcpu_ioctl_set_sregs()
12014 vcpu->arch.guest_state_protected) in kvm_arch_vcpu_ioctl_set_sregs()
12015 return -EINVAL; in kvm_arch_vcpu_ioctl_set_sregs()
12032 down_write(&kvm->arch.apicv_update_lock); in kvm_arch_vcpu_guestdbg_update_apicv_inhibit()
12035 if (vcpu->guest_debug & KVM_GUESTDBG_BLOCKIRQ) { in kvm_arch_vcpu_guestdbg_update_apicv_inhibit()
12041 up_write(&kvm->arch.apicv_update_lock); in kvm_arch_vcpu_guestdbg_update_apicv_inhibit()
12050 if (vcpu->arch.guest_state_protected) in kvm_arch_vcpu_ioctl_set_guest_debug()
12051 return -EINVAL; in kvm_arch_vcpu_ioctl_set_guest_debug()
12055 if (dbg->control & (KVM_GUESTDBG_INJECT_DB | KVM_GUESTDBG_INJECT_BP)) { in kvm_arch_vcpu_ioctl_set_guest_debug()
12056 r = -EBUSY; in kvm_arch_vcpu_ioctl_set_guest_debug()
12059 if (dbg->control & KVM_GUESTDBG_INJECT_DB) in kvm_arch_vcpu_ioctl_set_guest_debug()
12071 vcpu->guest_debug = dbg->control; in kvm_arch_vcpu_ioctl_set_guest_debug()
12072 if (!(vcpu->guest_debug & KVM_GUESTDBG_ENABLE)) in kvm_arch_vcpu_ioctl_set_guest_debug()
12073 vcpu->guest_debug = 0; in kvm_arch_vcpu_ioctl_set_guest_debug()
12075 if (vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP) { in kvm_arch_vcpu_ioctl_set_guest_debug()
12077 vcpu->arch.eff_db[i] = dbg->arch.debugreg[i]; in kvm_arch_vcpu_ioctl_set_guest_debug()
12078 vcpu->arch.guest_debug_dr7 = dbg->arch.debugreg[7]; in kvm_arch_vcpu_ioctl_set_guest_debug()
12081 vcpu->arch.eff_db[i] = vcpu->arch.db[i]; in kvm_arch_vcpu_ioctl_set_guest_debug()
12085 if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP) in kvm_arch_vcpu_ioctl_set_guest_debug()
12086 vcpu->arch.singlestep_rip = kvm_get_linear_rip(vcpu); in kvm_arch_vcpu_ioctl_set_guest_debug()
12096 kvm_arch_vcpu_guestdbg_update_apicv_inhibit(vcpu->kvm); in kvm_arch_vcpu_ioctl_set_guest_debug()
12111 unsigned long vaddr = tr->linear_address; in kvm_arch_vcpu_ioctl_translate()
12117 idx = srcu_read_lock(&vcpu->kvm->srcu); in kvm_arch_vcpu_ioctl_translate()
12119 srcu_read_unlock(&vcpu->kvm->srcu, idx); in kvm_arch_vcpu_ioctl_translate()
12120 tr->physical_address = gpa; in kvm_arch_vcpu_ioctl_translate()
12121 tr->valid = gpa != INVALID_GPA; in kvm_arch_vcpu_ioctl_translate()
12122 tr->writeable = 1; in kvm_arch_vcpu_ioctl_translate()
12123 tr->usermode = 0; in kvm_arch_vcpu_ioctl_translate()
12133 if (fpstate_is_confidential(&vcpu->arch.guest_fpu)) in kvm_arch_vcpu_ioctl_get_fpu()
12134 return vcpu->kvm->arch.has_protected_state ? -EINVAL : 0; in kvm_arch_vcpu_ioctl_get_fpu()
12138 fxsave = &vcpu->arch.guest_fpu.fpstate->regs.fxsave; in kvm_arch_vcpu_ioctl_get_fpu()
12139 memcpy(fpu->fpr, fxsave->st_space, 128); in kvm_arch_vcpu_ioctl_get_fpu()
12140 fpu->fcw = fxsave->cwd; in kvm_arch_vcpu_ioctl_get_fpu()
12141 fpu->fsw = fxsave->swd; in kvm_arch_vcpu_ioctl_get_fpu()
12142 fpu->ftwx = fxsave->twd; in kvm_arch_vcpu_ioctl_get_fpu()
12143 fpu->last_opcode = fxsave->fop; in kvm_arch_vcpu_ioctl_get_fpu()
12144 fpu->last_ip = fxsave->rip; in kvm_arch_vcpu_ioctl_get_fpu()
12145 fpu->last_dp = fxsave->rdp; in kvm_arch_vcpu_ioctl_get_fpu()
12146 memcpy(fpu->xmm, fxsave->xmm_space, sizeof(fxsave->xmm_space)); in kvm_arch_vcpu_ioctl_get_fpu()
12156 if (fpstate_is_confidential(&vcpu->arch.guest_fpu)) in kvm_arch_vcpu_ioctl_set_fpu()
12157 return vcpu->kvm->arch.has_protected_state ? -EINVAL : 0; in kvm_arch_vcpu_ioctl_set_fpu()
12161 fxsave = &vcpu->arch.guest_fpu.fpstate->regs.fxsave; in kvm_arch_vcpu_ioctl_set_fpu()
12163 memcpy(fxsave->st_space, fpu->fpr, 128); in kvm_arch_vcpu_ioctl_set_fpu()
12164 fxsave->cwd = fpu->fcw; in kvm_arch_vcpu_ioctl_set_fpu()
12165 fxsave->swd = fpu->fsw; in kvm_arch_vcpu_ioctl_set_fpu()
12166 fxsave->twd = fpu->ftwx; in kvm_arch_vcpu_ioctl_set_fpu()
12167 fxsave->fop = fpu->last_opcode; in kvm_arch_vcpu_ioctl_set_fpu()
12168 fxsave->rip = fpu->last_ip; in kvm_arch_vcpu_ioctl_set_fpu()
12169 fxsave->rdp = fpu->last_dp; in kvm_arch_vcpu_ioctl_set_fpu()
12170 memcpy(fxsave->xmm_space, fpu->xmm, sizeof(fxsave->xmm_space)); in kvm_arch_vcpu_ioctl_set_fpu()
12180 if (vcpu->run->kvm_valid_regs & KVM_SYNC_X86_REGS) in store_regs()
12181 __get_regs(vcpu, &vcpu->run->s.regs.regs); in store_regs()
12183 if (vcpu->run->kvm_valid_regs & KVM_SYNC_X86_SREGS) in store_regs()
12184 __get_sregs(vcpu, &vcpu->run->s.regs.sregs); in store_regs()
12186 if (vcpu->run->kvm_valid_regs & KVM_SYNC_X86_EVENTS) in store_regs()
12188 vcpu, &vcpu->run->s.regs.events); in store_regs()
12193 if (vcpu->run->kvm_dirty_regs & KVM_SYNC_X86_REGS) { in sync_regs()
12194 __set_regs(vcpu, &vcpu->run->s.regs.regs); in sync_regs()
12195 vcpu->run->kvm_dirty_regs &= ~KVM_SYNC_X86_REGS; in sync_regs()
12198 if (vcpu->run->kvm_dirty_regs & KVM_SYNC_X86_SREGS) { in sync_regs()
12199 struct kvm_sregs sregs = vcpu->run->s.regs.sregs; in sync_regs()
12202 return -EINVAL; in sync_regs()
12204 vcpu->run->kvm_dirty_regs &= ~KVM_SYNC_X86_SREGS; in sync_regs()
12207 if (vcpu->run->kvm_dirty_regs & KVM_SYNC_X86_EVENTS) { in sync_regs()
12208 struct kvm_vcpu_events events = vcpu->run->s.regs.events; in sync_regs()
12211 return -EINVAL; in sync_regs()
12213 vcpu->run->kvm_dirty_regs &= ~KVM_SYNC_X86_EVENTS; in sync_regs()
12221 if (kvm_check_tsc_unstable() && kvm->created_vcpus) in kvm_arch_vcpu_precreate()
12225 if (!kvm->arch.max_vcpu_ids) in kvm_arch_vcpu_precreate()
12226 kvm->arch.max_vcpu_ids = KVM_MAX_VCPU_IDS; in kvm_arch_vcpu_precreate()
12228 if (id >= kvm->arch.max_vcpu_ids) in kvm_arch_vcpu_precreate()
12229 return -EINVAL; in kvm_arch_vcpu_precreate()
12239 vcpu->arch.last_vmentry_cpu = -1; in kvm_arch_vcpu_create()
12240 vcpu->arch.regs_avail = ~0; in kvm_arch_vcpu_create()
12241 vcpu->arch.regs_dirty = ~0; in kvm_arch_vcpu_create()
12243 kvm_gpc_init(&vcpu->arch.pv_time, vcpu->kvm); in kvm_arch_vcpu_create()
12245 if (!irqchip_in_kernel(vcpu->kvm) || kvm_vcpu_is_reset_bsp(vcpu)) in kvm_arch_vcpu_create()
12258 r = -ENOMEM; in kvm_arch_vcpu_create()
12263 vcpu->arch.pio_data = page_address(page); in kvm_arch_vcpu_create()
12265 vcpu->arch.mce_banks = kcalloc(KVM_MAX_MCE_BANKS * 4, sizeof(u64), in kvm_arch_vcpu_create()
12267 vcpu->arch.mci_ctl2_banks = kcalloc(KVM_MAX_MCE_BANKS, sizeof(u64), in kvm_arch_vcpu_create()
12269 if (!vcpu->arch.mce_banks || !vcpu->arch.mci_ctl2_banks) in kvm_arch_vcpu_create()
12271 vcpu->arch.mcg_cap = KVM_MAX_MCE_BANKS; in kvm_arch_vcpu_create()
12273 if (!zalloc_cpumask_var(&vcpu->arch.wbinvd_dirty_mask, in kvm_arch_vcpu_create()
12280 if (!fpu_alloc_guest_fpstate(&vcpu->arch.guest_fpu)) { in kvm_arch_vcpu_create()
12287 if (kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_STUFF_FEATURE_MSRS)) { in kvm_arch_vcpu_create()
12288 vcpu->arch.arch_capabilities = kvm_get_arch_capabilities(); in kvm_arch_vcpu_create()
12289 vcpu->arch.msr_platform_info = MSR_PLATFORM_INFO_CPUID_FAULT; in kvm_arch_vcpu_create()
12290 vcpu->arch.perf_capabilities = kvm_caps.supported_perf_cap; in kvm_arch_vcpu_create()
12294 vcpu->arch.pending_external_vector = -1; in kvm_arch_vcpu_create()
12295 vcpu->arch.preempted_in_kernel = false; in kvm_arch_vcpu_create()
12298 vcpu->arch.hv_root_tdp = INVALID_PAGE; in kvm_arch_vcpu_create()
12308 kvm_set_tsc_khz(vcpu, vcpu->kvm->arch.default_tsc_khz); in kvm_arch_vcpu_create()
12315 fpu_free_guest_fpstate(&vcpu->arch.guest_fpu); in kvm_arch_vcpu_create()
12317 kmem_cache_free(x86_emulator_cache, vcpu->arch.emulate_ctxt); in kvm_arch_vcpu_create()
12319 free_cpumask_var(vcpu->arch.wbinvd_dirty_mask); in kvm_arch_vcpu_create()
12321 kfree(vcpu->arch.mce_banks); in kvm_arch_vcpu_create()
12322 kfree(vcpu->arch.mci_ctl2_banks); in kvm_arch_vcpu_create()
12323 free_page((unsigned long)vcpu->arch.pio_data); in kvm_arch_vcpu_create()
12333 struct kvm *kvm = vcpu->kvm; in kvm_arch_vcpu_postcreate()
12335 if (mutex_lock_killable(&vcpu->mutex)) in kvm_arch_vcpu_postcreate()
12342 vcpu->arch.msr_kvm_poll_control = 1; in kvm_arch_vcpu_postcreate()
12344 mutex_unlock(&vcpu->mutex); in kvm_arch_vcpu_postcreate()
12346 if (kvmclock_periodic_sync && vcpu->vcpu_idx == 0) in kvm_arch_vcpu_postcreate()
12347 schedule_delayed_work(&kvm->arch.kvmclock_sync_work, in kvm_arch_vcpu_postcreate()
12365 kmem_cache_free(x86_emulator_cache, vcpu->arch.emulate_ctxt); in kvm_arch_vcpu_destroy()
12366 free_cpumask_var(vcpu->arch.wbinvd_dirty_mask); in kvm_arch_vcpu_destroy()
12367 fpu_free_guest_fpstate(&vcpu->arch.guest_fpu); in kvm_arch_vcpu_destroy()
12372 kfree(vcpu->arch.mce_banks); in kvm_arch_vcpu_destroy()
12373 kfree(vcpu->arch.mci_ctl2_banks); in kvm_arch_vcpu_destroy()
12375 idx = srcu_read_lock(&vcpu->kvm->srcu); in kvm_arch_vcpu_destroy()
12377 srcu_read_unlock(&vcpu->kvm->srcu, idx); in kvm_arch_vcpu_destroy()
12378 free_page((unsigned long)vcpu->arch.pio_data); in kvm_arch_vcpu_destroy()
12379 kvfree(vcpu->arch.cpuid_entries); in kvm_arch_vcpu_destroy()
12389 * Several of the "set" flows, e.g. ->set_cr0(), read other registers in kvm_vcpu_reset()
12399 * SVM doesn't unconditionally VM-Exit on INIT and SHUTDOWN, thus it's in kvm_vcpu_reset()
12410 vcpu->arch.hflags = 0; in kvm_vcpu_reset()
12412 vcpu->arch.smi_pending = 0; in kvm_vcpu_reset()
12413 vcpu->arch.smi_count = 0; in kvm_vcpu_reset()
12414 atomic_set(&vcpu->arch.nmi_queued, 0); in kvm_vcpu_reset()
12415 vcpu->arch.nmi_pending = 0; in kvm_vcpu_reset()
12416 vcpu->arch.nmi_injected = false; in kvm_vcpu_reset()
12420 memset(vcpu->arch.db, 0, sizeof(vcpu->arch.db)); in kvm_vcpu_reset()
12422 vcpu->arch.dr6 = DR6_ACTIVE_LOW; in kvm_vcpu_reset()
12423 vcpu->arch.dr7 = DR7_FIXED_1; in kvm_vcpu_reset()
12426 vcpu->arch.cr2 = 0; in kvm_vcpu_reset()
12429 vcpu->arch.apf.msr_en_val = 0; in kvm_vcpu_reset()
12430 vcpu->arch.apf.msr_int_val = 0; in kvm_vcpu_reset()
12431 vcpu->arch.st.msr_val = 0; in kvm_vcpu_reset()
12437 vcpu->arch.apf.halted = false; in kvm_vcpu_reset()
12439 if (vcpu->arch.guest_fpu.fpstate && kvm_mpx_supported()) { in kvm_vcpu_reset()
12440 struct fpstate *fpstate = vcpu->arch.guest_fpu.fpstate; in kvm_vcpu_reset()
12457 vcpu->arch.smbase = 0x30000; in kvm_vcpu_reset()
12459 vcpu->arch.pat = MSR_IA32_CR_PAT_DEFAULT; in kvm_vcpu_reset()
12461 vcpu->arch.msr_misc_features_enables = 0; in kvm_vcpu_reset()
12462 vcpu->arch.ia32_misc_enable_msr = MSR_IA32_MISC_ENABLE_PEBS_UNAVAIL | in kvm_vcpu_reset()
12470 memset(vcpu->arch.regs, 0, sizeof(vcpu->arch.regs)); in kvm_vcpu_reset()
12481 kvm_rdx_write(vcpu, cpuid_0x1 ? cpuid_0x1->eax : 0x600); in kvm_vcpu_reset()
12488 vcpu->arch.cr3 = 0; in kvm_vcpu_reset()
12510 * which PCIDs have to be flushed. However, CR0.WP and the paging-related in kvm_vcpu_reset()
12580 if (!stable && vcpu->cpu == smp_processor_id()) in kvm_arch_enable_virtualization_cpu()
12582 if (stable && vcpu->arch.last_host_tsc > local_tsc) { in kvm_arch_enable_virtualization_cpu()
12584 if (vcpu->arch.last_host_tsc > max_tsc) in kvm_arch_enable_virtualization_cpu()
12585 max_tsc = vcpu->arch.last_host_tsc; in kvm_arch_enable_virtualization_cpu()
12615 * N.B. - this code below runs only on platforms with reliable TSC, in kvm_arch_enable_virtualization_cpu()
12629 u64 delta_cyc = max_tsc - local_tsc; in kvm_arch_enable_virtualization_cpu()
12631 kvm->arch.backwards_tsc_observed = true; in kvm_arch_enable_virtualization_cpu()
12633 vcpu->arch.tsc_offset_adjustment += delta_cyc; in kvm_arch_enable_virtualization_cpu()
12634 vcpu->arch.last_host_tsc = local_tsc; in kvm_arch_enable_virtualization_cpu()
12644 kvm->arch.last_tsc_nsec = 0; in kvm_arch_enable_virtualization_cpu()
12645 kvm->arch.last_tsc_write = 0; in kvm_arch_enable_virtualization_cpu()
12660 return vcpu->kvm->arch.bsp_vcpu_id == vcpu->vcpu_id; in kvm_vcpu_is_reset_bsp()
12666 return (vcpu->arch.apic_base & MSR_IA32_APICBASE_BSP) != 0; in kvm_vcpu_is_bsp()
12672 kfree(kvm->arch.hv_pa_pg); in kvm_arch_free_vm()
12684 return -EINVAL; in kvm_arch_init_vm()
12686 kvm->arch.vm_type = type; in kvm_arch_init_vm()
12687 kvm->arch.has_private_mem = in kvm_arch_init_vm()
12690 kvm->arch.pre_fault_allowed = in kvm_arch_init_vm()
12692 kvm->arch.disabled_quirks = kvm_caps.inapplicable_quirks & kvm_caps.supported_quirks; in kvm_arch_init_vm()
12706 atomic_set(&kvm->arch.noncoherent_dma_count, 0); in kvm_arch_init_vm()
12708 raw_spin_lock_init(&kvm->arch.tsc_write_lock); in kvm_arch_init_vm()
12709 mutex_init(&kvm->arch.apic_map_lock); in kvm_arch_init_vm()
12710 seqcount_raw_spinlock_init(&kvm->arch.pvclock_sc, &kvm->arch.tsc_write_lock); in kvm_arch_init_vm()
12711 kvm->arch.kvmclock_offset = -get_kvmclock_base_ns(); in kvm_arch_init_vm()
12713 raw_spin_lock_irqsave(&kvm->arch.tsc_write_lock, flags); in kvm_arch_init_vm()
12715 raw_spin_unlock_irqrestore(&kvm->arch.tsc_write_lock, flags); in kvm_arch_init_vm()
12717 kvm->arch.default_tsc_khz = max_tsc_khz ? : tsc_khz; in kvm_arch_init_vm()
12718 kvm->arch.apic_bus_cycle_ns = APIC_BUS_CYCLE_NS_DEFAULT; in kvm_arch_init_vm()
12719 kvm->arch.guest_can_read_msr_platform_info = true; in kvm_arch_init_vm()
12720 kvm->arch.enable_pmu = enable_pmu; in kvm_arch_init_vm()
12723 spin_lock_init(&kvm->arch.hv_root_tdp_lock); in kvm_arch_init_vm()
12724 kvm->arch.hv_root_tdp = INVALID_PAGE; in kvm_arch_init_vm()
12727 INIT_DELAYED_WORK(&kvm->arch.kvmclock_update_work, kvmclock_update_fn); in kvm_arch_init_vm()
12728 INIT_DELAYED_WORK(&kvm->arch.kvmclock_sync_work, kvmclock_sync_fn); in kvm_arch_init_vm()
12741 once_init(&kvm->arch.nx_once); in kvm_arch_init_vm()
12765 * -errno: on error
12770 * GPA->HVA translation will not change. However, the HVA is a user
12782 lockdep_assert_held(&kvm->slots_lock); in __x86_set_memory_region()
12785 return ERR_PTR_USR(-EINVAL); in __x86_set_memory_region()
12789 if (slot && slot->npages) in __x86_set_memory_region()
12790 return ERR_PTR_USR(-EEXIST); in __x86_set_memory_region()
12801 if (!slot || !slot->npages) in __x86_set_memory_region()
12804 old_npages = slot->npages; in __x86_set_memory_region()
12805 hva = slot->userspace_addr; in __x86_set_memory_region()
12833 * is unsafe, i.e. will lead to use-after-free. The PIT also needs to in kvm_arch_pre_destroy_vm()
12836 cancel_delayed_work_sync(&kvm->arch.kvmclock_sync_work); in kvm_arch_pre_destroy_vm()
12837 cancel_delayed_work_sync(&kvm->arch.kvmclock_update_work); in kvm_arch_pre_destroy_vm()
12849 if (current->mm == kvm->mm) { in kvm_arch_destroy_vm()
12855 mutex_lock(&kvm->slots_lock); in kvm_arch_destroy_vm()
12861 mutex_unlock(&kvm->slots_lock); in kvm_arch_destroy_vm()
12864 kvm_free_msr_filter(srcu_dereference_check(kvm->arch.msr_filter, &kvm->srcu, 1)); in kvm_arch_destroy_vm()
12869 kvfree(rcu_dereference_check(kvm->arch.apic_map, 1)); in kvm_arch_destroy_vm()
12870 kfree(srcu_dereference_check(kvm->arch.pmu_event_filter, &kvm->srcu, 1)); in kvm_arch_destroy_vm()
12883 vfree(slot->arch.rmap[i]); in memslot_rmap_free()
12884 slot->arch.rmap[i] = NULL; in memslot_rmap_free()
12895 vfree(slot->arch.lpage_info[i - 1]); in kvm_arch_free_memslot()
12896 slot->arch.lpage_info[i - 1] = NULL; in kvm_arch_free_memslot()
12904 const int sz = sizeof(*slot->arch.rmap[0]); in memslot_rmap_alloc()
12911 if (slot->arch.rmap[i]) in memslot_rmap_alloc()
12914 slot->arch.rmap[i] = __vcalloc(lpages, sz, GFP_KERNEL_ACCOUNT); in memslot_rmap_alloc()
12915 if (!slot->arch.rmap[i]) { in memslot_rmap_alloc()
12917 return -ENOMEM; in memslot_rmap_alloc()
12927 unsigned long npages = slot->npages; in kvm_alloc_memslot_metadata()
12935 memset(&slot->arch, 0, sizeof(slot->arch)); in kvm_alloc_memslot_metadata()
12955 slot->arch.lpage_info[i - 1] = linfo; in kvm_alloc_memslot_metadata()
12957 if (slot->base_gfn & (KVM_PAGES_PER_HPAGE(level) - 1)) in kvm_alloc_memslot_metadata()
12959 if ((slot->base_gfn + npages) & (KVM_PAGES_PER_HPAGE(level) - 1)) in kvm_alloc_memslot_metadata()
12960 linfo[lpages - 1].disallow_lpage = 1; in kvm_alloc_memslot_metadata()
12961 ugfn = slot->userspace_addr >> PAGE_SHIFT; in kvm_alloc_memslot_metadata()
12966 if ((slot->base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1)) { in kvm_alloc_memslot_metadata()
12987 vfree(slot->arch.lpage_info[i - 1]); in kvm_alloc_memslot_metadata()
12988 slot->arch.lpage_info[i - 1] = NULL; in kvm_alloc_memslot_metadata()
12990 return -ENOMEM; in kvm_alloc_memslot_metadata()
12999 * memslots->generation has been incremented. in kvm_arch_memslots_updated()
13004 /* Force re-initialization of steal_time cache */ in kvm_arch_memslots_updated()
13019 return -EINVAL; in kvm_arch_prepare_memory_region()
13022 if ((new->base_gfn + new->npages - 1) > kvm_mmu_max_gfn()) in kvm_arch_prepare_memory_region()
13023 return -EINVAL; in kvm_arch_prepare_memory_region()
13025 if (kvm_is_gfn_alias(kvm, new->base_gfn + new->npages - 1)) in kvm_arch_prepare_memory_region()
13026 return -EINVAL; in kvm_arch_prepare_memory_region()
13032 memcpy(&new->arch, &old->arch, sizeof(old->arch)); in kvm_arch_prepare_memory_region()
13034 return -EIO; in kvm_arch_prepare_memory_region()
13044 if (!kvm->arch.cpu_dirty_log_size) in kvm_mmu_update_cpu_dirty_logging()
13047 nr_slots = atomic_read(&kvm->nr_memslots_dirty_logging); in kvm_mmu_update_cpu_dirty_logging()
13057 u32 old_flags = old ? old->flags : 0; in kvm_mmu_slot_apply_flags()
13058 u32 new_flags = new ? new->flags : 0; in kvm_mmu_slot_apply_flags()
13078 * CREATE: No shadow pages exist, thus nothing to write-protect in kvm_mmu_slot_apply_flags()
13087 * READONLY and non-flags changes were filtered out above, and the only in kvm_mmu_slot_apply_flags()
13107 * Initially-all-set does not require write protecting any page, in kvm_mmu_slot_apply_flags()
13116 if (kvm->arch.cpu_dirty_log_size) { in kvm_mmu_slot_apply_flags()
13132 * write-protected before returning to userspace, i.e. before in kvm_mmu_slot_apply_flags()
13139 * Specifically, KVM also write-protects guest page tables to in kvm_mmu_slot_apply_flags()
13148 * To handle these scenarios, KVM uses a separate software-only in kvm_mmu_slot_apply_flags()
13149 * bit (MMU-writable) to track if a SPTE is !writable due to in kvm_mmu_slot_apply_flags()
13150 * a guest page table being write-protected (KVM clears the in kvm_mmu_slot_apply_flags()
13151 * MMU-writable flag when write-protecting for shadow paging). in kvm_mmu_slot_apply_flags()
13153 * The use of MMU-writable is also the primary motivation for in kvm_mmu_slot_apply_flags()
13156 * !MMU-writable SPTE, KVM must flush if it encounters any in kvm_mmu_slot_apply_flags()
13157 * MMU-writable SPTE regardless of whether the actual hardware in kvm_mmu_slot_apply_flags()
13160 * write access" helpers to ignore MMU-writable entirely. in kvm_mmu_slot_apply_flags()
13163 * access-tracked SPTEs is particularly relevant). in kvm_mmu_slot_apply_flags()
13177 if (!kvm->arch.n_requested_mmu_pages && in kvm_arch_commit_memory_region()
13181 nr_mmu_pages = kvm->nr_memslot_pages / KVM_MEMSLOT_PAGES_TO_MMU_PAGES_RATIO; in kvm_arch_commit_memory_region()
13197 if (vcpu->arch.guest_state_protected) in kvm_arch_vcpu_in_kernel()
13207 if (vcpu->arch.guest_state_protected) in kvm_arch_vcpu_get_ip()
13226 if (vcpu->arch.guest_state_protected) in kvm_get_linear_rip()
13247 if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP) in kvm_get_rflags()
13255 if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP && in __kvm_set_rflags()
13256 kvm_is_linear_rip(vcpu, vcpu->arch.singlestep_rip)) in __kvm_set_rflags()
13277 return (key + 1) & (ASYNC_PF_PER_VCPU - 1); in kvm_async_pf_next_probe()
13284 while (vcpu->arch.apf.gfns[key] != ~0) in kvm_add_async_pf_gfn()
13287 vcpu->arch.apf.gfns[key] = gfn; in kvm_add_async_pf_gfn()
13296 (vcpu->arch.apf.gfns[key] != gfn && in kvm_async_pf_gfn_slot()
13297 vcpu->arch.apf.gfns[key] != ~0); i++) in kvm_async_pf_gfn_slot()
13305 return vcpu->arch.apf.gfns[kvm_async_pf_gfn_slot(vcpu, gfn)] == gfn; in kvm_find_async_pf_gfn()
13314 if (WARN_ON_ONCE(vcpu->arch.apf.gfns[i] != gfn)) in kvm_del_async_pf_gfn()
13318 vcpu->arch.apf.gfns[i] = ~0; in kvm_del_async_pf_gfn()
13321 if (vcpu->arch.apf.gfns[j] == ~0) in kvm_del_async_pf_gfn()
13323 k = kvm_async_pf_hash_fn(vcpu->arch.apf.gfns[j]); in kvm_del_async_pf_gfn()
13330 vcpu->arch.apf.gfns[i] = vcpu->arch.apf.gfns[j]; in kvm_del_async_pf_gfn()
13339 return kvm_write_guest_cached(vcpu->kvm, &vcpu->arch.apf.data, &reason, in apf_put_user_notpresent()
13347 return kvm_write_guest_offset_cached(vcpu->kvm, &vcpu->arch.apf.data, in apf_put_user_ready()
13356 if (kvm_read_guest_offset_cached(vcpu->kvm, &vcpu->arch.apf.data, in apf_pageready_slot_free()
13369 if (!vcpu->arch.apf.send_always && in kvm_can_deliver_async_pf()
13370 (vcpu->arch.guest_state_protected || !kvm_x86_call(get_cpl)(vcpu))) in kvm_can_deliver_async_pf()
13378 return vcpu->arch.apf.delivery_as_pf_vmexit; in kvm_can_deliver_async_pf()
13396 if (kvm_hlt_in_guest(vcpu->kvm) && !kvm_can_deliver_async_pf(vcpu)) in kvm_can_do_async_pf()
13411 trace_kvm_async_pf_not_present(work->arch.token, work->cr2_or_gpa); in kvm_arch_async_page_not_present()
13412 kvm_add_async_pf_gfn(vcpu, work->arch.gfn); in kvm_arch_async_page_not_present()
13420 fault.address = work->arch.token; in kvm_arch_async_page_not_present()
13443 .vector = vcpu->arch.apf.vec in kvm_arch_async_page_present()
13446 if (work->wakeup_all) in kvm_arch_async_page_present()
13447 work->arch.token = ~0; /* broadcast wakeup */ in kvm_arch_async_page_present()
13449 kvm_del_async_pf_gfn(vcpu, work->arch.gfn); in kvm_arch_async_page_present()
13450 trace_kvm_async_pf_ready(work->arch.token, work->cr2_or_gpa); in kvm_arch_async_page_present()
13452 if ((work->wakeup_all || work->notpresent_injected) && in kvm_arch_async_page_present()
13454 !apf_put_user_ready(vcpu, work->arch.token)) { in kvm_arch_async_page_present()
13455 vcpu->arch.apf.pageready_pending = true; in kvm_arch_async_page_present()
13459 vcpu->arch.apf.halted = false; in kvm_arch_async_page_present()
13466 if (!vcpu->arch.apf.pageready_pending) in kvm_arch_async_page_present_queued()
13481 * Non-coherent DMA assignment and de-assignment may affect whether or in kvm_noncoherent_dma_assignment_start_or_stop()
13484 * (or last) non-coherent device is (un)registered to so that new SPTEs in kvm_noncoherent_dma_assignment_start_or_stop()
13495 if (atomic_inc_return(&kvm->arch.noncoherent_dma_count) == 1) in kvm_arch_register_noncoherent_dma()
13502 if (!atomic_dec_return(&kvm->arch.noncoherent_dma_count)) in kvm_arch_unregister_noncoherent_dma()
13509 return atomic_read(&kvm->arch.noncoherent_dma_count); in kvm_arch_has_noncoherent_dma()
13520 return (vcpu->arch.msr_kvm_poll_control & 1) == 0; in kvm_arch_no_poll()
13566 struct kvm_mmu *mmu = vcpu->arch.walk_mmu; in kvm_fixup_and_inject_pf_error()
13572 mmu->gva_to_gpa(vcpu, mmu, gva, access, &fault) != INVALID_GPA) { in kvm_fixup_and_inject_pf_error()
13574 * If vcpu->arch.walk_mmu->gva_to_gpa succeeded, the page in kvm_fixup_and_inject_pf_error()
13585 vcpu->arch.walk_mmu->inject_page_fault(vcpu, &fault); in kvm_fixup_and_inject_pf_error()
13598 if (KVM_BUG_ON(!e, vcpu->kvm)) in kvm_handle_memory_failure()
13599 return -EIO; in kvm_handle_memory_failure()
13609 * doesn't seem to be a real use-case behind such requests, just return in kvm_handle_memory_failure()
13665 * page tables, so a non-global flush just degenerates to a in kvm_handle_invpcid()
13684 struct kvm_run *run = vcpu->run; in complete_sev_es_emulated_mmio()
13688 BUG_ON(!vcpu->mmio_needed); in complete_sev_es_emulated_mmio()
13691 frag = &vcpu->mmio_fragments[vcpu->mmio_cur_fragment]; in complete_sev_es_emulated_mmio()
13692 len = min(8u, frag->len); in complete_sev_es_emulated_mmio()
13693 if (!vcpu->mmio_is_write) in complete_sev_es_emulated_mmio()
13694 memcpy(frag->data, run->mmio.data, len); in complete_sev_es_emulated_mmio()
13696 if (frag->len <= 8) { in complete_sev_es_emulated_mmio()
13699 vcpu->mmio_cur_fragment++; in complete_sev_es_emulated_mmio()
13702 frag->data += len; in complete_sev_es_emulated_mmio()
13703 frag->gpa += len; in complete_sev_es_emulated_mmio()
13704 frag->len -= len; in complete_sev_es_emulated_mmio()
13707 if (vcpu->mmio_cur_fragment >= vcpu->mmio_nr_fragments) { in complete_sev_es_emulated_mmio()
13708 vcpu->mmio_needed = 0; in complete_sev_es_emulated_mmio()
13716 run->mmio.phys_addr = frag->gpa; in complete_sev_es_emulated_mmio()
13717 run->mmio.len = min(8u, frag->len); in complete_sev_es_emulated_mmio()
13718 run->mmio.is_write = vcpu->mmio_is_write; in complete_sev_es_emulated_mmio()
13719 if (run->mmio.is_write) in complete_sev_es_emulated_mmio()
13720 memcpy(run->mmio.data, frag->data, min(8u, frag->len)); in complete_sev_es_emulated_mmio()
13721 run->exit_reason = KVM_EXIT_MMIO; in complete_sev_es_emulated_mmio()
13723 vcpu->arch.complete_userspace_io = complete_sev_es_emulated_mmio; in complete_sev_es_emulated_mmio()
13735 return -EINVAL; in kvm_sev_es_mmio_write()
13741 bytes -= handled; in kvm_sev_es_mmio_write()
13746 frag = vcpu->mmio_fragments; in kvm_sev_es_mmio_write()
13747 vcpu->mmio_nr_fragments = 1; in kvm_sev_es_mmio_write()
13748 frag->len = bytes; in kvm_sev_es_mmio_write()
13749 frag->gpa = gpa; in kvm_sev_es_mmio_write()
13750 frag->data = data; in kvm_sev_es_mmio_write()
13752 vcpu->mmio_needed = 1; in kvm_sev_es_mmio_write()
13753 vcpu->mmio_cur_fragment = 0; in kvm_sev_es_mmio_write()
13755 vcpu->run->mmio.phys_addr = gpa; in kvm_sev_es_mmio_write()
13756 vcpu->run->mmio.len = min(8u, frag->len); in kvm_sev_es_mmio_write()
13757 vcpu->run->mmio.is_write = 1; in kvm_sev_es_mmio_write()
13758 memcpy(vcpu->run->mmio.data, frag->data, min(8u, frag->len)); in kvm_sev_es_mmio_write()
13759 vcpu->run->exit_reason = KVM_EXIT_MMIO; in kvm_sev_es_mmio_write()
13761 vcpu->arch.complete_userspace_io = complete_sev_es_emulated_mmio; in kvm_sev_es_mmio_write()
13774 return -EINVAL; in kvm_sev_es_mmio_read()
13780 bytes -= handled; in kvm_sev_es_mmio_read()
13785 frag = vcpu->mmio_fragments; in kvm_sev_es_mmio_read()
13786 vcpu->mmio_nr_fragments = 1; in kvm_sev_es_mmio_read()
13787 frag->len = bytes; in kvm_sev_es_mmio_read()
13788 frag->gpa = gpa; in kvm_sev_es_mmio_read()
13789 frag->data = data; in kvm_sev_es_mmio_read()
13791 vcpu->mmio_needed = 1; in kvm_sev_es_mmio_read()
13792 vcpu->mmio_cur_fragment = 0; in kvm_sev_es_mmio_read()
13794 vcpu->run->mmio.phys_addr = gpa; in kvm_sev_es_mmio_read()
13795 vcpu->run->mmio.len = min(8u, frag->len); in kvm_sev_es_mmio_read()
13796 vcpu->run->mmio.is_write = 0; in kvm_sev_es_mmio_read()
13797 vcpu->run->exit_reason = KVM_EXIT_MMIO; in kvm_sev_es_mmio_read()
13799 vcpu->arch.complete_userspace_io = complete_sev_es_emulated_mmio; in kvm_sev_es_mmio_read()
13807 vcpu->arch.sev_pio_count -= count; in advance_sev_es_emulated_pio()
13808 vcpu->arch.sev_pio_data += count * size; in advance_sev_es_emulated_pio()
13816 int size = vcpu->arch.pio.size; in complete_sev_es_emulated_outs()
13817 int port = vcpu->arch.pio.port; in complete_sev_es_emulated_outs()
13819 vcpu->arch.pio.count = 0; in complete_sev_es_emulated_outs()
13820 if (vcpu->arch.sev_pio_count) in complete_sev_es_emulated_outs()
13830 min_t(unsigned int, PAGE_SIZE / size, vcpu->arch.sev_pio_count); in kvm_sev_es_outs()
13831 int ret = emulator_pio_out(vcpu, size, port, vcpu->arch.sev_pio_data, count); in kvm_sev_es_outs()
13839 if (!vcpu->arch.sev_pio_count) in kvm_sev_es_outs()
13843 vcpu->arch.complete_userspace_io = complete_sev_es_emulated_outs; in kvm_sev_es_outs()
13852 unsigned count = vcpu->arch.pio.count; in complete_sev_es_emulated_ins()
13853 int size = vcpu->arch.pio.size; in complete_sev_es_emulated_ins()
13854 int port = vcpu->arch.pio.port; in complete_sev_es_emulated_ins()
13856 complete_emulator_pio_in(vcpu, vcpu->arch.sev_pio_data); in complete_sev_es_emulated_ins()
13858 if (vcpu->arch.sev_pio_count) in complete_sev_es_emulated_ins()
13868 min_t(unsigned int, PAGE_SIZE / size, vcpu->arch.sev_pio_count); in kvm_sev_es_ins()
13869 if (!emulator_pio_in(vcpu, size, port, vcpu->arch.sev_pio_data, count)) in kvm_sev_es_ins()
13874 if (!vcpu->arch.sev_pio_count) in kvm_sev_es_ins()
13878 vcpu->arch.complete_userspace_io = complete_sev_es_emulated_ins; in kvm_sev_es_ins()
13886 vcpu->arch.sev_pio_data = data; in kvm_sev_es_string_io()
13887 vcpu->arch.sev_pio_count = count; in kvm_sev_es_string_io()